You are on page 1of 512

DRINKING COFFEE, MATE

,
AND VERY HOT BEVERAGES
VOLUME 116

IARC MONOGRAPHS
ON THE EVALUATION
OF CARCINOGENIC RISKS
TO HUMANS
DRINKING COFFEE, MATE,
AND VERY HOT BEVERAGES
VOLUME 116

This publication represents the views and expert
opinions of an IARC Working Group on the
Evaluation of Carcinogenic Risks to Humans,
which met in Lyon, 24–31 May 2016

LYON, FRANCE - 2018

IARC MONOGRAPHS
ON THE EVALUATION
OF CARCINOGENIC RISKS
TO HUMANS
IARC MONOGRAPHS
In 1969, the International Agency for Research on Cancer (IARC) initiated a programme on the evaluation of the
carcinogenic risk of chemicals to humans involving the production of critically evaluated monographs on individual chemicals.
The programme was subsequently expanded to include evaluations of carcinogenic risks associated with exposures to complex
mixtures, lifestyle factors and biological and physical agents, as well as those in specific occupations. The objective of the
programme is to elaborate and publish in the form of monographs critical reviews of data on carcinogenicity for agents to
which humans are known to be exposed and on specific exposure situations; to evaluate these data in terms of human risk
with the help of international working groups of experts in carcinogenesis and related fields; and to indicate where additional
research efforts are needed. The lists of IARC evaluations are regularly updated and are available on the Internet at http://
monographs.iarc.fr/.
This programme has been supported since 1982 by Cooperative Agreement U01 CA33193 with the United States
National Cancer Institute, Department of Health and Human Services. Additional support has been provided since 1986 by
the European Commission Directorate-General for Employment, Social Affairs, and Inclusion, initially by the Unit of Health,
Safety and Hygiene at Work, and since 2014 by the European Union Programme for Employment and Social Innovation “EaSI”
(2014–2020) (for further information please consult: http://ec.europa.eu/social/easi). Support has also been provided since
1992 by the United States National Institute of Environmental Health Sciences, Department of Health and Human Services.
The contents of this volume are solely the responsibility of the Working Group and do not necessarily represent the official
views of the United States National Cancer Institute, the United States National Institute of Environmental Health Sciences, the
United States Department of Health and Human Services, or the European Commission.
Published by the International Agency for Research on Cancer,
150 cours Albert Thomas, 69372 Lyon Cedex 08, France
©International Agency for Research on Cancer, 2018
On-line publication, June 2018
Distributed by WHO Press, World Health Organization, 20 Avenue Appia, 1211 Geneva 27, Switzerland
(tel.: +41 22 791 3264; fax: +41 22 791 4857; email: bookorders@who.int).
Publications of the World Health Organization enjoy copyright protection in accordance with the provisions
of Protocol 2 of the Universal Copyright Convention. All rights reserved.
Corrigenda to the IARC Monographs are published online at http://monographs.iarc.fr/ENG/Publications/corrigenda.php
To report an error, please contact: editimo@iarc.fr

Co-funded by the European Union

The International Agency for Research on Cancer welcomes requests for permission to reproduce or translate its
publications, in part or in full. Requests for permission to reproduce or translate IARC publications – whether for sale or for
non-commercial distribution – should be addressed to the IARC Communications Group at: publications@iarc.fr.
The designations employed and the presentation of the material in this publication do not imply the expression of any
opinion whatsoever on the part of the Secretariat of the World Health Organization concerning the legal status of any country,
territory, city, or area or of its authorities, or concerning the delimitation of its frontiers or boundaries.
The mention of specific companies or of certain manufacturers’ products does not imply that they are endorsed or
recommended by the World Health Organization in preference to others of a similar nature that are not mentioned. Errors and
omissions excepted, the names of proprietary products are distinguished by initial capital letters.
The IARC Monographs Working Group alone is responsible for the views expressed in this publication.
IARC Library Cataloguing in Publication Data
Drinking Coffee, Mate, and Very Hot Beverages / IARC Working Group on the Evaluation of Carcinogenic Risks to
Humans (2016: Lyon, France)
(IARC monographs on the evaluation of carcinogenic risks to humans ; volume 116)
1. Carcinogens 2. Coffee – adverse effects 3. Ilex paraguariensis – adverse effects 4. Beverages – adverse effects
5. Hot Temperature – adverse effects 6. Risk Factors
I. International Agency for Research on Cancer II. Series
ISBN 978-92-832-0183-0 (NLM Classification: W1)
ISSN 1017-1606
CONTENTS

NOTE TO THE READER. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

LIST OF PARTICIPANTS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

PREAMBLE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
A. GENERAL PRINCIPLES AND PROCEDURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1. Background. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2. Objective and scope. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3. Selection of agents for review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4. Data for the Monographs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
5. Meeting participants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
6. Working procedures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
B. SCIENTIFIC REVIEW AND EVALUATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1. Exposure data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2. Studies of cancer in humans. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3. Studies of cancer in experimental animals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4. Mechanistic and other relevant data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5. Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6. Evaluation and rationale. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

GENERAL REMARKS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

DRINKING COFFEE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

1. Exposure Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.1 Identification of the agent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.2 Methods of production, uses, and preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.3 Exposure assessment and biological markers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
1.4 Chemical constituents. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

III
IARC MONOGRAPHS - 116

2. Cancer in Humans. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.1 Cancer of the bladder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.2 Cancer of the pancreas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
2.3 Cancer of the liver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
2.4 Cancer of the breast in women. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
2.5 Cancer of the endometrium. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
2.6 Cancer of the prostate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
2.7 Cancer of the lung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
2.8 Cancer of the larynx. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
2.9 Cancer of the ovary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
2.10 Childhood cancer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
2.11 Cancer of the oral cavity and pharynx . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
2.12 Cancer of the oesophagus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
2.13 Cancer of the stomach, small intestine, gall bladder, and biliary tract. . . . . . . . . . . . . . . . . . . . . . . 300
2.14 Cancer of the colorectum. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
2.15 Cancer of the kidney. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
2.16 Malignant melanoma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
2.17 Non-melanoma cancer of the skin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
2.18 Adult cancer of the brain. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
2.19 Adult haematopoietic cancers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
2.20 Other cancers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
2.21 All cancers combined . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311

3. Cancer in Experimental Animals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
3.1 Studies of carcinogenicity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
3.2 Co-carcinogenicity and initiation–promotion studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352

4. Mechanistic and Other Relevant Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
4.1 Toxicokinetic data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
4.2 Mechanisms of carcinogenesis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364
4.3 Genetic susceptibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392
4.4 Other effects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 398

5. Summary of Data Reported . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415
5.1 Exposure data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415
5.2 Human carcinogenicity data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415
5.3 Animal carcinogenicity data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 420
5.4 Mechanistic and other relevant data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 421

6. Evaluation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
6.1 Cancer in humans. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
6.2 Cancer in experimental animals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
6.3 Overall evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425

IV
Coffee and Very Hot Beverages

DRINKING MATE AND VERY HOT BEVERAGES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427

1. Exposure Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427
1.1 Identification of the agent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427
1.2 Production and use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429
1.3 Production and consumption data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434
1.4 Methods of measurement and exposure assessment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434

2. Cancer in Humans. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 437
2.1 Mate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 437
2.2 Very hot beverages other than mate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449

3. Cancer in Experimental Animals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 476
3.1 Mate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 476
3.2 Very hot water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478

4. Mechanistic and Other Relevant Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 479
4.1 Absorption, distribution, metabolism, and excretion of mate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 479
4.2 Mechanisms of carcinogenesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 480
4.3 Other adverse effects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485

5. Summary of Data Reported . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 486
5.1 Exposure data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 486
5.2 Human carcinogenicity data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 487
5.3 Animal carcinogenicity data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 488
5.4 Mechanistic and other relevant data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 489

6. Evaluation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 490
6.1 Cancer in humans. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 490
6.2 Cancer in experimental animals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 490
6.3 Overall evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 490
6.4 Rationale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 490
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 490

LIST OF ABBREVIATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 497

V
NOTE TO THE READER

The term ‘carcinogenic risk’ in the IARC Monographs series is taken to mean that an agent is
capable of causing cancer. The Monographs evaluate cancer hazards, despite the historical presence
of the word ‘risks’ in the title.
Inclusion of an agent in the Monographs does not imply that it is a carcinogen, only that the
published data have been examined. Equally, the fact that an agent has not yet been evaluated in a
Monograph does not mean that it is not carcinogenic. Similarly, identification of cancer sites with
sufficient evidence or limited evidence in humans should not be viewed as precluding the possibility
that an agent may cause cancer at other sites.
The evaluations of carcinogenic risk are made by international working groups of independent
scientists and are qualitative in nature. No recommendation is given for regulation or legislation.
Anyone who is aware of published data that may alter the evaluation of the carcinogenic risk
of an agent to humans is encouraged to make this information available to the Section of IARC
Monographs, International Agency for Research on Cancer, 150 cours Albert Thomas, 69372 Lyon
Cedex 08, France, in order that the agent may be considered for re-evaluation by a future Working
Group.
Although every effort is made to prepare the Monographs as accurately as possible, mistakes may
occur. Readers are requested to communicate any errors to the Section of IARC Monographs, so that
corrections can be reported in future volumes.

1
LIST OF PARTICIPANTS
Members 1

Christina Bamia Natasa Djordjevic
Dept of Hygiene, Epidemiology and Medical Dept of Pharmacology and Toxicology
Statistics, Medical School University of Kragujevac
National and Kapodistrian University of Kragujevac
Athens Serbia
Athens
Greece
Adriana Farah
Nutrition Institute, Centre of Health Sciences
John A. Baron Federal University of Rio de Janeiro
Dept of Medicine and Epidemiology Rio de Janeiro
University of North Carolina at Chapel Hill Brazil
Chapel Hill, NC
USA Elvira Gonzalez de Mejia
Dept of Food Science and Human Nutrition
University of Illinois
Urbana, IL
USA

1
Working Group Members and Invited Specialists serve in their individual capacities as scientists and not as
representatives of their government or any organization with which they are affiliated. Affiliations are provided for iden-
tification purposes only. Invited Specialists do not serve as Meeting Chair or Subgroup Chair, draft text that pertains to
the description or interpretation of cancer data, or participate in the evaluations.
Each participant was asked to disclose pertinent research, employment, and financial interests. Current financial
interests and research and employment interests during the past 4 years or anticipated in the future are identified here.
Minor pertinent interests are not listed and include stock valued at no more than US$ 1000 overall, grants that provide
no more than 5% of the research budget of the expert’s organization and that do not support the expert’s research or
position, and consulting or speaking on matters not before a court or government agency that does not exceed 2% of
total professional time or compensation. All grants that support the expert’s research or position and all consulting
or speaking on behalf of an interested party on matters before a court or government agency are listed as significant
pertinent interests.

3
IARC MONOGRAPHS – 116

Peter C. Hollman Dirk W. Lachenmeier (Subgroup Chair,
Division of Human Nutrition Exposure Data)
Wageningen University Chemical and Veterinary Investigation Agency
Wageningen Karlsruhe
The Netherlands Karlsruhe
Germany
Manami Inoue
Dept of Health and Human Security David L. McCormick (Subgroup Chair, Cancer
Graduate School of Medicine in Experimental Animals)
University of Tokyo Illinois Institute of Technology Research
Tokyo Institute
Japan Chicago, IL
USA
Farhad Islami (unable to attend)
Surveillance and Health Services Research Elizabeth Milne
American Cancer Society Cancer Epidemiology Group
Atlanta, GA Telethon Kids Institute
USA West Perth, WA
Australia
Charles William Jameson
CWJ Consulting, LLC Igor Pogribny
Cape Coral, FL Division of Biochemical Toxicology
USA National Center for Toxicological Research
Jefferson, AR
USA
Farin Kamangar
Dept of Public Health Analysis
School of Community Health and Policy Luis Felipe Ribeiro Pinto
Morgan State University Brazilian National Cancer Institute (INCA)
Baltimore, MD Rio de Janeiro
USA Brazil

Siegfried Knasmüller Ivan I. Rusyn (Subgroup Chair, Mechanistic
and Other Relevant Data)
Institute of Cancer Research
Medical University of Vienna College of Veterinary Medicine & Biomedical
Vienna Sciences
Austria Texas A&M University
College Station, TX
USA

4
Participants

Rashmi Sinha (Subgroup Chair, Cancer in Kathryn M. Wilson
Humans) Dept of Epidemiology
National Cancer Institute Harvard T.H. Chan School of Public Health
Division of Cancer Epidemiology & Genetics Boston, MA
Bethesda, MD USA
USA

Invited Specialists
Leslie Stayner (Overall Chair)
Division of Epidemiology and Biostatistics
School of Public Health None
University of Illinois at Chicago
Chicago, IL
USA Representatives

Mariana C. Stern
Norris Comprehensive Cancer Center Anne Morise
University of Southern California Risk Assessment Department
Los Angeles, CA French Agency for Food, Environmental and
USA Occupational Health & Safety (ANSES)
Maisons-Alfort
Alessandra Tavani France

Mario Negri Institute of Pharmacological
Research (IRCCS) Observers 2
Milan
Italy
Valentina Guercio 3
Piet van den Brandt
Mario Negri Institute of Pharmacological
Dept of Epidemiology Research (IRCCS)
Maastricht University Milan
Maastricht Italy
The Netherlands

2
Each Observer agreed to respect the Guidelines for Observers at IARC Monographs meetings. Observers did not serve
as Meeting Chair or Subgroup Chair, draft any part of a Monograph, or participate in the evaluations. They also agreed
not to contact participants before the meeting, not to lobby them at any time, not to send them written materials, and
not to offer them meals or other favours. IARC asked and reminded Working Group Members to report any contact or
attempt to influence that they may have encountered, either before or during the meeting.
3
Valentina Guercio attended as an Observer for the Mario Negri Institute of Pharmacological Research, Italy.

5
IARC MONOGRAPHS – 116

Thomas Hatzold 4 Michael Wilde 9
Munich Dept of Philosophy
Germany School of European Culture and Languages
University of Kent
Canterbury, Kent
Alan Leviton 5 England
Children’s Hospital
Boston, MA
USA IARC/WHO Secretariat

Behnoush Abedi-Ardekani
Frankie Phillips 6
Lamia Benbrahim-Tallaa (Rapporteur,
Exeter Mechanistic and Other Relevant Data)
England Véronique Bouvard (Rapporteur,
Exposure Data)
Valentin Thomas 7 Fatiha El Ghissassi (Rapporteur, Mechanistic
and Other Relevant Data)
University Paris Dauphine Yann Grosse (Rapporteur, Cancer in
Paris Experimental Animals)
France Neela Guha (Rapporteur, Cancer in Humans)
Kathryn Guyton (Rapporteur, Mechanistic
Christian Wallmann 8 and Other Relevant Data)
Inge Huybrechts
Dept of Philosophy Béatrice Lauby-Secretan
School of European Culture and Languages Dana Loomis (Responsible Officer)
University of Kent Heidi Mattock (Scientific Editor)
Canterbury, Kent Hwayoung Noh
England Silvia Pisanu
Joseph Rothwell
Augustin Scalbert
Kurt Straif (Head of Programme)
Maria Zhivagui

4
Thomas Hatzold was until 2015 employed by and still holds stock options from Mondelez International, a multina-
tional confectionery, food, and beverage conglomerate with headquarters in Deerfield, IL, USA. He is an Observer for
the Institute for Scientific Information on Coffee, an organization whose members include six of the major European
coffee companies.
5
Alan Leviton has provided consulting services for the National Coffee Association, a trade association, based in New
York, USA. He is an Observer for the National Coffee Association.
6
Frankie Phillips attended as a Freelance Dietitian and Nutrition Consultant.
7
Valentin Thomas attended as an Observer for the University of Paris Dauphine, France.
8
Christian Wallmann attended as an Observer for the University of Kent, England.
9
Michael Wilde attended as an Observer for the University of Kent, England.

6
Participants

Administrative Assistance

Natacha Blavoyer
Marieke Dusenberg
Sandrine Egraz
Michel Javin
Helene Lorenzen-Augros

Post-meeting Assistance

Elaine Rowan (Technical Editor)

Production Team

Elisabeth Elbers
Fiona Gould
Solène Quennehen

7
PREAMBLE
The Preamble to the IARC Monographs describes the objective and scope of the programme,
the scientific principles and procedures used in developing a Monograph, the types of
evidence considered and the scientific criteria that guide the evaluations. The Preamble
should be consulted when reading a Monograph or list of evaluations.

A. GENERAL PRINCIPLES AND of carcinogenic risk of chemicals to man, which
PROCEDURES became the initial title of the series.
In the succeeding years, the scope of the
programme broadened as Monographs were
1. Background developed for groups of related chemicals,
Soon after IARC was established in 1965, complex mixtures, occupational exposures, phys-
it received frequent requests for advice on ical and biological agents and lifestyle factors. In
the carcinogenic risk of chemicals, including 1988, the phrase ‘of chemicals’ was dropped from
requests for lists of known and suspected human the title, which assumed its present form, IARC
carcinogens. It was clear that it would not be Monographs on the Evaluation of Carcinogenic
a simple task to summarize adequately the Risks to Humans.
complexity of the information that was avail- Through the Monographs programme, IARC
able, and IARC began to consider means of seeks to identify the causes of human cancer. This
obtaining international expert opinion on this is the first step in cancer prevention, which is
topic. In 1970, the IARC Advisory Committee on needed as much today as when IARC was estab-
Environmental Carcinogenesis recommended ‘... lished. The global burden of cancer is high and
that a compendium on carcinogenic chemicals continues to increase: the annual number of new
be prepared by experts. The biological activity cases was estimated at 10.1 million in 2000 and
and evaluation of practical importance to public is expected to reach 15 million by 2020 (Stewart
health should be referenced and documented.’ & Kleihues, 2003). With current trends in demo-
The IARC Governing Council adopted a resolu- graphics and exposure, the cancer burden has
tion concerning the role of IARC in providing been shifting from high-resource countries to
government authorities with expert, inde- low- and medium-resource countries. As a result
pendent, scientific opinion on environmental of Monographs evaluations, national health agen-
carcinogenesis. As one means to that end, the cies have been able, on scientific grounds, to take
Governing Council recommended that IARC measures to reduce human exposure to carcino-
should prepare monographs on the evaluation gens in the workplace and in the environment.

9
IARC MONOGRAPHS – 116

The criteria established in 1971 to evaluate as causation of, and susceptibility to, malignant
carcinogenic risks to humans were adopted by the disease become more fully understood.
Working Groups whose deliberations resulted in A cancer ‘hazard’ is an agent that is capable
the first 16 volumes of the Monographs series. of causing cancer under some circumstances,
Those criteria were subsequently updated by while a cancer ‘risk’ is an estimate of the carcino-
further ad hoc Advisory Groups (IARC, 1977, genic effects expected from exposure to a cancer
1978, 1979, 1982, 1983, 1987, 1988, 1991; Vainio hazard. The Monographs are an exercise in evalu-
et al., 1992; IARC, 2005, 2006). ating cancer hazards, despite the historical pres-
The Preamble is primarily a statement of ence of the word ‘risks’ in the title. The distinction
scientific principles, rather than a specification between hazard and risk is important, and the
of working procedures. The procedures through Monographs identify cancer hazards even when
which a Working Group implements these prin- risks are very low at current exposure levels,
ciples are not specified in detail. They usually because new uses or unforeseen exposures could
involve operations that have been established engender risks that are significantly higher.
as being effective during previous Monograph In the Monographs, an agent is termed
meetings but remain, predominantly, the prerog- ‘carcinogenic’ if it is capable of increasing the
ative of each individual Working Group. incidence of malignant neoplasms, reducing
their latency, or increasing their severity or
multiplicity. The induction of benign neoplasms
2. Objective and scope may in some circumstances (see Part B, Section
The objective of the programme is to 3a) contribute to the judgement that the agent is
prepare, with the help of international Working carcinogenic. The terms ‘neoplasm’ and ‘tumour’
Groups of experts, and to publish in the form of are used interchangeably.
Monographs, critical reviews and evaluations of The Preamble continues the previous usage
evidence on the carcinogenicity of a wide range of the phrase ‘strength of evidence’ as a matter of
of human exposures. The Monographs represent historical continuity, although it should be under-
the first step in carcinogen risk assessment, which stood that Monographs evaluations consider
involves examination of all relevant information studies that support a finding of a cancer hazard
to assess the strength of the available evidence as well as studies that do not.
that an agent could alter the age-specific inci- Some epidemiological and experimental
dence of cancer in humans. The Monographs may studies indicate that different agents may act at
also indicate where additional research efforts different stages in the carcinogenic process, and
are needed, specifically when data immediately several different mechanisms may be involved.
relevant to an evaluation are not available. The aim of the Monographs has been, from their
In this Preamble, the term ‘agent’ refers to inception, to evaluate evidence of carcinogenicity
any entity or circumstance that is subject to at any stage in the carcinogenesis process,
evaluation in a Monograph. As the scope of the independently of the underlying mechanisms.
programme has broadened, categories of agents Information on mechanisms may, however, be
now include specific chemicals, groups of related used in making the overall evaluation (IARC,
chemicals, complex mixtures, occupational or 1991; Vainio et al., 1992; IARC, 2005, 2006; see
environmental exposures, cultural or behav- also Part B, Sections 4 and 6). As mechanisms
ioural practices, biological organisms and phys- of carcinogenesis are elucidated, IARC convenes
ical agents. This list of categories may expand international scientific conferences to determine
whether a broad-based consensus has emerged

10
Preamble

on how specific mechanistic data can be used 3. Selection of agents for review
in an evaluation of human carcinogenicity. The
results of such conferences are reported in IARC Agents are selected for review on the basis
Scientific Publications, which, as long as they still of two main criteria: (a) there is evidence of
reflect the current state of scientific knowledge, human exposure and (b) there is some evidence
may guide subsequent Working Groups. or suspicion of carcinogenicity. Mixed exposures
Although the Monographs have emphasized may occur in occupational and environmental
hazard identification, important issues may also settings and as a result of individual and cultural
involve dose–response assessment. In many habits (such as tobacco smoking and dietary
cases, the same epidemiological and experi- practices). Chemical analogues and compounds
mental studies used to evaluate a cancer hazard with biological or physical characteristics similar
can also be used to estimate a dose–response to those of suspected carcinogens may also be
relationship. A Monograph may undertake to considered, even in the absence of data on a
estimate dose–response relationships within possible carcinogenic effect in humans or exper-
the range of the available epidemiological data, imental animals.
or it may compare the dose–response informa- The scientific literature is surveyed for
tion from experimental and epidemiological published data relevant to an assessment of
studies. In some cases, a subsequent publication carcinogenicity. Ad hoc Advisory Groups
may be prepared by a separate Working Group convened by IARC in 1984, 1989, 1991, 1993, 1998
with expertise in quantitative dose–response and 2003 made recommendations as to which
assessment. agents should be evaluated in the Monographs
The Monographs are used by national and series. Recent recommendations are available
international authorities to make risk assess- on the Monographs programme web site 
(http://
ments, formulate decisions concerning preven- monographs.iarc.fr). IARC may schedule other
tive measures, provide effective cancer control agents for review as it becomes aware of new
programmes and decide among alternative scientific information or as national health agen-
options for public health decisions. The evalu- cies identify an urgent public health need related
ations of IARC Working Groups are scientific, to cancer.
qualitative judgements on the evidence for or As significant new data become available on
against carcinogenicity provided by the available an agent for which a Monograph exists, a re-eval-
data. These evaluations represent only one part of uation may be made at a subsequent meeting, and
the body of information on which public health a new Monograph published. In some cases it may
decisions may be based. Public health options be appropriate to review only the data published
vary from one situation to another and from since a prior evaluation. This can be useful for
country to country and relate to many factors, updating a database, reviewing new data to
including different socioeconomic and national resolve a previously open question or identifying
priorities. Therefore, no recommendation is given new tumour sites associated with a carcinogenic
with regard to regulation or legislation, which agent. Major changes in an evaluation (e.g. a new
are the responsibility of individual governments classification in Group 1 or a determination that a
or other international organizations. mechanism does not operate in humans, see Part
B, Section 6) are more appropriately addressed
by a full review.

11
IARC MONOGRAPHS – 116

4. Data for the Monographs 5. Meeting participants
Each Monograph reviews all pertinent epide- Five categories of participant can be present
miological studies and cancer bioassays in exper- at Monograph meetings.
imental animals. Those judged inadequate or
irrelevant to the evaluation may be cited but not (a) The Working Group
summarized. If a group of similar studies is not
reviewed, the reasons are indicated. The Working Group is responsible for the
Mechanistic and other relevant data are also critical reviews and evaluations that are devel-
reviewed. A Monograph does not necessarily oped during the meeting. The tasks of Working
cite all the mechanistic literature concerning Group Members are: (i) to ascertain that all
the agent being evaluated (see Part B, Section appropriate data have been collected; (ii) to
4). Only those data considered by the Working select the data relevant for the evaluation on the
Group to be relevant to making the evaluation basis of scientific merit; (iii) to prepare accurate
are included. summaries of the data to enable the reader to
With regard to epidemiological studies, follow the reasoning of the Working Group; (iv)
cancer bioassays, and mechanistic and other rele- to evaluate the results of epidemiological and
vant data, only reports that have been published experimental studies on cancer; (v) to evaluate
or accepted for publication in the openly available data relevant to the understanding of mecha-
scientific literature are reviewed. The same publi- nisms of carcinogenesis; and (vi) to make an
cation requirement applies to studies originating overall evaluation of the carcinogenicity of the
from IARC, including meta-analyses or pooled exposure to humans. Working Group Members
analyses commissioned by IARC in advance of generally have published significant research
a meeting (see Part B, Section 2c). Data from related to the carcinogenicity of the agents being
government agency reports that are publicly reviewed, and IARC uses literature searches to
available are also considered. Exceptionally, identify most experts. Working Group Members
doctoral theses and other material that are in are selected on the basis of (a) knowledge and
their final form and publicly available may be experience and (b) absence of real or apparent
reviewed. conflicts of interests. Consideration is also given
Exposure data and other information on an to demographic diversity and balance of scien-
agent under consideration are also reviewed. In tific findings and views.
the sections on chemical and physical proper-
ties, on analysis, on production and use and on (b) Invited Specialists
occurrence, published and unpublished sources Invited Specialists are experts who also have
of information may be considered. critical knowledge and experience but have
Inclusion of a study does not imply accept- a real or apparent conflict of interests. These
ance of the adequacy of the study design or of experts are invited when necessary to assist in
the analysis and interpretation of the results, and the Working Group by contributing their unique
limitations are clearly outlined in square brackets knowledge and experience during subgroup and
at the end of each study description (see Part B). plenary discussions. They may also contribute
The reasons for not giving further consideration text on non-influential issues in the section on
to an individual study also are indicated in the exposure, such as a general description of data
square brackets. on production and use (see Part B, Section 1).
Invited Specialists do not serve as meeting chair

12
Preamble

or subgroup chair, draft text that pertains to the to report financial interests, employment and
description or interpretation of cancer data, or consulting, and individual and institutional
participate in the evaluations. research support related to the subject of the
meeting. IARC assesses these interests to deter-
(c) Representatives of national and mine whether there is a conflict that warrants
international health agencies some limitation on participation. The declarations
are updated and reviewed again at the opening
Representatives of national and interna- of the meeting. Interests related to the subject of
tional health agencies often attend meetings the meeting are disclosed to the meeting partic-
because their agencies sponsor the programme ipants and in the published volume (Cogliano
or are interested in the subject of a meeting. et al., 2004).
Representatives do not serve as meeting chair or The names and principal affiliations of
subgroup chair, draft any part of a Monograph, participants are available on the Monographs
or participate in the evaluations. programme web site (http://monographs.iarc.fr)
approximately two months before each meeting.
(d) Observers with relevant scientific It is not acceptable for Observers or third parties
credentials to contact other participants before a meeting or
to lobby them at any time. Meeting participants
Observers with relevant scientific credentials
are asked to report all such contacts to IARC
may be admitted to a meeting by IARC in limited
(Cogliano et al., 2005).
numbers. Attention will be given to achieving a
All participants are listed, with their prin-
balance of Observers from constituencies with
cipal affiliations, at the beginning of each volume.
differing perspectives. They are invited to observe
Each participant who is a Member of a Working
the meeting and should not attempt to influence
Group serves as an individual scientist and not as
it. Observers do not serve as meeting chair or
a representative of any organization, government
subgroup chair, draft any part of a Monograph,
or industry.
or participate in the evaluations. At the meeting,
the meeting chair and subgroup chairs may grant
Observers an opportunity to speak, generally 6. Working procedures
after they have observed a discussion. Observers
agree to respect the Guidelines for Observers at A separate Working Group is responsible
IARC Monographs meetings (available at 
http:// for developing each volume of Monographs. A
monographs.iarc.fr). volume contains one or more Monographs, which
can cover either a single agent or several related
agents. Approximately one year in advance of
(e) The IARC Secretariat
the meeting of a Working Group, the agents to
The IARC Secretariat consists of scientists be reviewed are announced on the Monographs
who are designated by IARC and who have rele- programme web site (http://monographs.iarc.fr)
vant expertise. They serve as rapporteurs and and participants are selected by IARC staff in
participate in all discussions. When requested by consultation with other experts. Subsequently,
the meeting chair or subgroup chair, they may relevant biological and epidemiological data are
also draft text or prepare tables and analyses. collected by IARC from recognized sources of
Before an invitation is extended, each poten- information on carcinogenesis, including data
tial participant, including the IARC Secretariat, storage and retrieval systems such as PubMed.
completes the WHO Declaration of Interests Meeting participants who are asked to prepare

13
IARC MONOGRAPHS – 116

preliminary working papers for specific sections IARC Working Groups strive to achieve a
are expected to supplement the IARC literature consensus evaluation. Consensus reflects broad
searches with their own searches. agreement among Working Group Members, but
Industrial associations, labour unions not necessarily unanimity. The chair may elect
and other knowledgeable organizations may to poll Working Group Members to determine
be asked to provide input to the sections on the diversity of scientific opinion on issues where
production and use, although this involvement consensus is not readily apparent.
is not required as a general rule. Information on After the meeting, the master copy is verified
production and trade is obtained from govern- by consulting the original literature, edited and
mental, trade and market research publications prepared for publication. The aim is to publish
and, in some cases, by direct contact with indus- the volume within six months of the Working
tries. Separate production data on some agents Group meeting. A summary of the outcome is
may not be available for a variety of reasons (e.g. available on the Monographs programme web
not collected or made public in all producing site soon after the meeting.
countries, production is small). Information on
uses may be obtained from published sources
but is often complemented by direct contact with B. SCIENTIFIC REVIEW AND
manufacturers. Efforts are made to supplement EVALUATION
this information with data from other national
and international sources. The available studies are summarized by the
Six months before the meeting, the material Working Group, with particular regard to the
obtained is sent to meeting participants to prepare qualitative aspects discussed below. In general,
preliminary working papers. The working papers numerical findings are indicated as they appear
are compiled by IARC staff and sent, before in the original report; units are converted when
the meeting, to Working Group Members and necessary for easier comparison. The Working
Invited Specialists for review. Group may conduct additional analyses of the
The Working Group meets at IARC for seven published data and use them in their assessment
to eight days to discuss and finalize the texts and of the evidence; the results of such supplemen-
to formulate the evaluations. The objectives of the tary analyses are given in square brackets. When
meeting are peer review and consensus. During an important aspect of a study that directly
the first few days, four subgroups (covering expo- impinges on its interpretation should be brought
sure data, cancer in humans, cancer in experi- to the attention of the reader, a Working Group
mental animals, and mechanistic and other comment is given in square brackets.
relevant data) review the working papers, develop The scope of the IARC Monographs
a joint subgroup draft and write summaries. Care programme has expanded beyond chemicals to
is taken to ensure that each study summary is include complex mixtures, occupational expo-
written or reviewed by someone not associated sures, physical and biological agents, lifestyle
with the study being considered. During the last factors and other potentially carcinogenic expo-
few days, the Working Group meets in plenary sures. Over time, the structure of a Monograph
session to review the subgroup drafts and develop has evolved to include the following sections:
the evaluations. As a result, the entire volume is
the joint product of the Working Group, and Exposure data
there are no individually authored sections. Studies of cancer in humans

14
Preamble

Studies of cancer in experimental animals which the agent being evaluated is only one of
Mechanistic and other relevant data the ingredients.
Summary For biological agents, taxonomy, structure
and biology are described, and the degree of
Evaluation and rationale variability is indicated. Mode of replication,
In addition, a section of General Remarks at life cycle, target cells, persistence, latency, host
the front of the volume discusses the reasons the response and clinical disease other than cancer
agents were scheduled for evaluation and some are also presented.
key issues the Working Group encountered For physical agents that are forms of radiation,
during the meeting. energy and range of the radiation are included.
This part of the Preamble discusses the types For foreign bodies, fibres and respirable particles,
of evidence considered and summarized in each size range and relative dimensions are indicated.
section of a Monograph, followed by the scientific For agents such as mixtures, drugs or lifestyle
criteria that guide the evaluations. factors, a description of the agent, including its
composition, is given.
Whenever appropriate, other information,
1. Exposure data such as historical perspectives or the description
Each Monograph includes general infor- of an industry or habit, may be included.
mation on the agent: this information may
vary substantially between agents and must be (b) Analysis and detection
adapted accordingly. Also included is informa-
An overview of methods of analysis and
tion on production and use (when appropriate),
detection of the agent is presented, including
methods of analysis and detection, occurrence,
their sensitivity, specificity and reproducibility.
and sources and routes of human occupational
Methods widely used for regulatory purposes
and environmental exposures. Depending on the
are emphasized. Methods for monitoring human
agent, regulations and guidelines for use may be
exposure are also given. No critical evaluation
presented.
or recommendation of any method is meant or
implied.
(a) General information on the agent
For chemical agents, sections on chemical (c) Production and use
and physical data are included: the Chemical
The dates of first synthesis and of first
Abstracts Service Registry Number, the latest
commercial production of a chemical, mixture
primary name and the IUPAC systematic name
or other agent are provided when available; for
are recorded; other synonyms are given, but the
agents that do not occur naturally, this informa-
list is not necessarily comprehensive. Information
tion may allow a reasonable estimate to be made
on chemical and physical properties that are rele-
of the date before which no human exposure
vant to identification, occurrence and biological
to the agent could have occurred. The dates of
activity is included. A description of technical
first reported occurrence of an exposure are also
products of chemicals includes trade names,
provided when available. In addition, methods
relevant specifications and available informa-
of synthesis used in past and present commercial
tion on composition and impurities. Some of the
production and different methods of production,
trade names given may be those of mixtures in

15
IARC MONOGRAPHS – 116

which may give rise to different impurities, are with date and place. For biological agents, the
described. epidemiology of infection is described.
The countries where companies report produc-
tion of the agent, and the number of companies (e) Regulations and guidelines
in each country, are identified. Available data
on production, international trade and uses are Statements concerning regulations and
obtained for representative regions. It should not, guidelines (e.g. occupational exposure limits,
however, be inferred that those areas or nations maximal levels permitted in foods and water,
are necessarily the sole or major sources or users pesticide registrations) are included, but they
of the agent. Some identified uses may not be may not reflect the most recent situation, since
current or major applications, and the coverage such limits are continuously reviewed and modi-
is not necessarily comprehensive. In the case of fied. The absence of information on regulatory
drugs, mention of their therapeutic uses does not status for a country should not be taken to imply
necessarily represent current practice nor does it that that country does not have regulations with
imply judgement as to their therapeutic efficacy. regard to the exposure. For biological agents,
legislation and control, including vaccination
and therapy, are described.
(d) Occurrence and exposure
Information on the occurrence of an agent in
the environment is obtained from data derived
2. Studies of cancer in humans
from the monitoring and surveillance of levels This section includes all pertinent epidemio-
in occupational environments, air, water, soil, logical studies (see Part A, Section 4). Studies of
plants, foods and animal and human tissues. biomarkers are included when they are relevant
When available, data on the generation, persis- to an evaluation of carcinogenicity to humans.
tence and bioaccumulation of the agent are
also included. Such data may be available from (a) Types of study considered
national databases.
Data that indicate the extent of past and Several types of epidemiological study
present human exposure, the sources of expo- contribute to the assessment of carcinogenicity in
sure, the people most likely to be exposed and humans — cohort studies, case–control studies,
the factors that contribute to the exposure are correlation (or ecological) studies and interven-
reported. Information is presented on the range tion studies. Rarely, results from randomized
of human exposure, including occupational and trials may be available. Case reports and case
environmental exposures. This includes relevant series of cancer in humans may also be reviewed.
findings from both developed and developing Cohort and case–control studies relate indi-
countries. Some of these data are not distrib- vidual exposures under study to the occurrence of
uted widely and may be available from govern- cancer in individuals and provide an estimate of
ment reports and other sources. In the case of effect (such as relative risk) as the main measure
mixtures, industries, occupations or processes, of association. Intervention studies may provide
information is given about all agents known to strong evidence for making causal inferences,
be present. For processes, industries and occupa- as exemplified by cessation of smoking and the
tions, a historical description is also given, noting subsequent decrease in risk for lung cancer.
variations in chemical composition, physical In correlation studies, the units of inves-
properties and levels of occupational exposure tigation are usually whole populations (e.g. in

16
Preamble

particular geographical areas or at particular Bias is the effect of factors in study design or
times), and cancer frequency is related to a execution that lead erroneously to a stronger or
summary measure of the exposure of the popu- weaker association than in fact exists between an
lation to the agent under study. In correlation agent and disease. Confounding is a form of bias
studies, individual exposure is not documented, that occurs when the relationship with disease
which renders this kind of study more prone to is made to appear stronger or weaker than it
confounding. In some circumstances, however, truly is as a result of an association between the
correlation studies may be more informative apparent causal factor and another factor that is
than analytical study designs (see, for example, associated with either an increase or decrease in
the Monograph on arsenic in drinking-water; the incidence of the disease. The role of chance is
IARC, 2004). related to biological variability and the influence
In some instances, case reports and case series of sample size on the precision of estimates of
have provided important information about the effect.
carcinogenicity of an agent. These types of study In evaluating the extent to which these factors
generally arise from a suspicion, based on clinical have been minimized in an individual study,
experience, that the concurrence of two events — consideration is given to several aspects of design
that is, a particular exposure and occurrence of and analysis as described in the report of the
a cancer — has happened rather more frequently study. For example, when suspicion of carcino-
than would be expected by chance. Case reports genicity arises largely from a single small study,
and case series usually lack complete ascertain- careful consideration is given when interpreting
ment of cases in any population, definition or subsequent studies that included these data in
enumeration of the population at risk and esti- an enlarged population. Most of these consider-
mation of the expected number of cases in the ations apply equally to case–control, cohort and
absence of exposure. correlation studies. Lack of clarity of any of these
The uncertainties that surround the interpre- aspects in the reporting of a study can decrease
tation of case reports, case series and correlation its credibility and the weight given to it in the
studies make them inadequate, except in rare final evaluation of the exposure.
instances, to form the sole basis for inferring a First, the study population, disease (or
causal relationship. When taken together with diseases) and exposure should have been well
case–control and cohort studies, however, these defined by the authors. Cases of disease in the
types of study may add materially to the judge- study population should have been identified in
ment that a causal relationship exists. a way that was independent of the exposure of
Epidemiological studies of benign neoplasms, interest, and exposure should have been assessed
presumed preneoplastic lesions and other in a way that was not related to disease status.
end-points thought to be relevant to cancer are Second, the authors should have taken into
also reviewed. They may, in some instances, account — in the study design and analysis —
strengthen inferences drawn from studies of other variables that can influence the risk of
cancer itself. disease and may have been related to the expo-
sure of interest. Potential confounding by such
(b) Quality of studies considered variables should have been dealt with either in
the design of the study, such as by matching,
It is necessary to take into account the or in the analysis, by statistical adjustment. In
possible roles of bias, confounding and chance cohort studies, comparisons with local rates of
in the interpretation of epidemiological studies. disease may or may not be more appropriate than

17
IARC MONOGRAPHS – 116

those with national rates. Internal comparisons individual studies (pooled analysis) (Greenland,
of frequency of disease among individuals at 1998).
different levels of exposure are also desirable in The advantages of combined analyses are
cohort studies, since they minimize the potential increased precision due to increased sample
for confounding related to the difference in risk size and the opportunity to explore potential
factors between an external reference group and confounders, interactions and modifying effects
the study population. that may explain heterogeneity among studies
Third, the authors should have reported the in more detail. A disadvantage of combined
basic data on which the conclusions are founded, analyses is the possible lack of compatibility of
even if sophisticated statistical analyses were data from various studies due to differences in
employed. At the very least, they should have subject recruitment, procedures of data collec-
given the numbers of exposed and unexposed tion, methods of measurement and effects of
cases and controls in a case–control study and unmeasured co-variates that may differ among
the numbers of cases observed and expected in studies. Despite these limitations, well conducted
a cohort study. Further tabulations by time since combined analyses may provide a firmer basis
exposure began and other temporal factors are than individual studies for drawing conclusions
also important. In a cohort study, data on all about the potential carcinogenicity of agents.
cancer sites and all causes of death should have IARC may commission a meta-analysis or
been given, to reveal the possibility of reporting pooled analysis that is pertinent to a particular
bias. In a case–control study, the effects of inves- Monograph (see Part A, Section 4). Additionally,
tigated factors other than the exposure of interest as a means of gaining insight from the results of
should have been reported. multiple individual studies, ad hoc calculations
Finally, the statistical methods used to obtain that combine data from different studies may
estimates of relative risk, absolute rates of cancer, be conducted by the Working Group during the
confidence intervals and significance tests, and course of a Monograph meeting. The results of
to adjust for confounding should have been such original calculations, which would be speci-
clearly stated by the authors. These methods have fied in the text by presentation in square brackets,
been reviewed for case–control studies (Breslow might involve updates of previously conducted
& Day, 1980) and for cohort studies (Breslow & analyses that incorporate the results of more
Day, 1987). recent studies or de-novo analyses. Irrespective
of the source of data for the meta-analyses and
(c) Meta-analyses and pooled analyses pooled analyses, it is important that the same
criteria for data quality be applied as those that
Independent epidemiological studies of the would be applied to individual studies and to
same agent may lead to results that are difficult ensure also that sources of heterogeneity between
to interpret. Combined analyses of data from studies be taken into account.
multiple studies are a means of resolving this
ambiguity, and well conducted analyses can be
(d) Temporal effects
considered. There are two types of combined
analysis. The first involves combining summary Detailed analyses of both relative and abso-
statistics such as relative risks from individual lute risks in relation to temporal variables, such
studies (meta-analysis) and the second involves as age at first exposure, time since first expo-
a pooled analysis of the raw data from the sure, duration of exposure, cumulative expo-
sure, peak exposure (when appropriate) and

18
Preamble

time since cessation of exposure, are reviewed known phenotype of a genetic polymorphism
and summarized when available. Analyses of can explain the carcinogenic mechanism of the
temporal relationships may be useful in making agent being evaluated, data on this phenotype
causal inferences. In addition, such analyses may may be useful in making causal inferences.
suggest whether a carcinogen acts early or late in
the process of carcinogenesis, although, at best, (f) Criteria for causality
they allow only indirect inferences about mech-
anisms of carcinogenesis. After the quality of individual epidemiolog-
ical studies of cancer has been summarized and
assessed, a judgement is made concerning the
(e) Use of biomarkers in epidemiological
strength of evidence that the agent in question
studies is carcinogenic to humans. In making its judge-
Biomarkers indicate molecular, cellular or ment, the Working Group considers several
other biological changes and are increasingly criteria for causality (Hill, 1965). A strong asso-
used in epidemiological studies for various ciation 
(e.g. a large relative risk) is more likely
purposes (IARC, 1991; Vainio et al., 1992; Toniolo to indicate causality than a weak association,
et al., 1997; Vineis et al., 1999; Buffler et al., 2004). although it is recognized that estimates of effect
These may include evidence of exposure, of early of small magnitude do not imply lack of causality
effects, of cellular, tissue or organism responses, and may be important if the disease or exposure
of individual susceptibility or host responses, is common. Associations that are replicated in
and inference of a mechanism (see Part B, Section several studies of the same design or that use
4b). This is a rapidly evolving field that encom- different epidemiological approaches or under
passes developments in genomics, epigenomics different circumstances of exposure are more
and other emerging technologies. likely to represent a causal relationship than
Molecular epidemiological data that identify isolated observations from single studies. If there
associations between genetic polymorphisms are inconsistent results among investigations,
and interindividual differences in susceptibility possible reasons are sought (such as differences in
to the agent(s) being evaluated may contribute exposure), and results of studies that are judged
to the identification of carcinogenic hazards to to be of high quality are given more weight than
humans. If the polymorphism has been demon- those of studies that are judged to be methodo-
strated experimentally to modify the functional logically less sound.
activity of the gene product in a manner that is If the risk increases with the exposure, this is
consistent with increased susceptibility, these considered to be a strong indication of causality,
data may be useful in making causal inferences. although the absence of a graded response is not
Similarly, molecular epidemiological studies that necessarily evidence against a causal relation-
measure cell functions, enzymes or metabolites ship. The demonstration of a decline in risk after
that are thought to be the basis of susceptibility cessation of or reduction in exposure in indi-
may provide evidence that reinforces biological viduals or in whole populations also supports a
plausibility. It should be noted, however, that causal interpretation of the findings.
when data on genetic susceptibility originate from Several scenarios may increase confidence in
multiple comparisons that arise from subgroup a causal relationship. On the one hand, an agent
analyses, this can generate false-positive results may be specific in causing tumours at one site or
and inconsistencies across studies, and such of one morphological type. On the other, carcino-
data therefore require careful evaluation. If the genicity may be evident through the causation of

19
IARC MONOGRAPHS – 116

multiple tumour types. Temporality, precision 3. Studies of cancer in
of estimates of effect, biological plausibility and experimental animals
coherence of the overall database are considered.
Data on biomarkers may be employed in an All known human carcinogens that have been
assessment of the biological plausibility of epide- studied adequately for carcinogenicity in exper-
miological observations. imental animals have produced positive results
Although rarely available, results from rand- in one or more animal species (Wilbourn et al.,
omized trials that show different rates of cancer 1986; Tomatis et al., 1989). For several agents
among exposed and unexposed individuals (e.g. aflatoxins, diethylstilbestrol, solar radiation,
provide particularly strong evidence for causality. vinyl chloride), carcinogenicity in experimental
When several epidemiological studies show animals was established or highly suspected
little or no indication of an association between before epidemiological studies confirmed their
an exposure and cancer, a judgement may be carcinogenicity in humans (Vainio et al., 1995).
made that, in the aggregate, they show evidence Although this association cannot establish that
of lack of carcinogenicity. Such a judgement all agents that cause cancer in experimental
requires first that the studies meet, to a suffi- animals also cause cancer in humans, it is biolog-
cient degree, the standards of design and anal- ically plausible that agents for which there is suffi-
ysis described above. Specifically, the possibility cient evidence of carcinogenicity in experimental
that bias, confounding or misclassification of animals (see Part B, Section 6b) also present a
exposure or outcome could explain the observed carcinogenic hazard to humans. Accordingly, in
results should be considered and excluded with the absence of additional scientific information,
reasonable certainty. In addition, all studies that these agents are considered to pose a carcino-
are judged to be methodologically sound should genic hazard to humans. Examples of additional
(a) be consistent with an estimate of effect of scientific information are data that demonstrate
unity for any observed level of exposure, (b) when that a given agent causes cancer in animals
considered together, provide a pooled estimate of through a species-specific mechanism that does
relative risk that is at or near to unity, and (c) not operate in humans or data that demonstrate
have a narrow confidence interval, due to suffi- that the mechanism in experimental animals
cient population size. Moreover, no individual also operates in humans (see Part B, Section 6).
study nor the pooled results of all the studies Consideration is given to all available long-
should show any consistent tendency that the term studies of cancer in experimental animals
relative risk of cancer increases with increasing with the agent under review (see Part A, Section
level of exposure. It is important to note that 4). In all experimental settings, the nature and
evidence of lack of carcinogenicity obtained extent of impurities or contaminants present in
from several epidemiological studies can apply the agent being evaluated are given when avail-
only to the type(s) of cancer studied, to the dose able. Animal species, strain (including genetic
levels reported, and to the intervals between first background where applicable), sex, numbers per
exposure and disease onset observed in these group, age at start of treatment, route of expo-
studies. Experience with human cancer indicates sure, dose levels, duration of exposure, survival
that the period from first exposure to the devel- and information on tumours (incidence, latency,
opment of clinical cancer is sometimes longer severity or multiplicity of neoplasms or prene-
than 20 years; latent periods substantially shorter oplastic lesions) are reported. Those studies in
than 30 years cannot provide evidence for lack of experimental animals that are judged to be irrel-
carcinogenicity. evant to the evaluation or judged to be inadequate

20
Preamble

(e.g. too short a duration, too few animals, poor (a) Qualitative aspects
survival; see below) may be omitted. Guidelines
for conducting long-term carcinogenicity exper- An assessment of carcinogenicity involves
iments have been published (e.g. OECD, 2002). several considerations of qualitative importance,
Other studies considered may include: exper- including (i) the experimental conditions under
iments in which the agent was administered in which the test was performed, including route,
the presence of factors that modify carcinogenic schedule and duration of exposure, species,
effects (e.g. initiation–promotion studies, co-car- strain (including genetic background where
cinogenicity studies and studies in genetically applicable), sex, age and duration of follow-up; (ii)
modified animals); studies in which the end-point the consistency of the results, for example, across
was not cancer but a defined precancerous lesion; species and target organ(s); (iii) the spectrum of
experiments on the carcinogenicity of known neoplastic response, from preneoplastic lesions
metabolites and derivatives; and studies of and benign tumours to malignant neoplasms;
cancer in non-laboratory animals (e.g. livestock and (iv) the possible role of modifying factors.
and companion animals) exposed to the agent. Considerations of importance in the inter-
For studies of mixtures, consideration is pretation and evaluation of a particular study
given to the possibility that changes in the include: (i) how clearly the agent was defined
physicochemical properties of the individual and, in the case of mixtures, how adequately
substances may occur during collection, storage, the sample characterization was reported; (ii)
extraction, concentration and delivery. Another whether the dose was monitored adequately,
consideration is that chemical and toxicological particularly in inhalation experiments; (iii)
interactions of components in a mixture may whether the doses, duration of treatment and
alter dose–response relationships. The relevance route of exposure were appropriate; (iv) whether
to human exposure of the test mixture adminis- the survival of treated animals was similar to
tered in the animal experiment is also assessed. that of controls; (v) whether there were adequate
This may involve consideration of the following numbers of animals per group; (vi) whether
aspects of the mixture tested: (i) physical and both male and female animals were used; (vii)
chemical characteristics, (ii) identified constitu- whether animals were allocated randomly to
ents that may indicate the presence of a class of groups; (viii) whether the duration of observa-
substances and (iii) the results of genetic toxicity tion was adequate; and (ix) whether the data were
and related tests. reported and analysed adequately.
The relevance of results obtained with an When benign tumours (a) occur together
agent that is analogous (e.g. similar in structure with and originate from the same cell type as
or of a similar virus genus) to that being evalu- malignant tumours in an organ or tissue in a
ated is also considered. Such results may provide particular study and (b) appear to represent a
biological and mechanistic information that is stage in the progression to malignancy, they are
relevant to the understanding of the process of usually combined in the assessment of tumour
carcinogenesis in humans and may strengthen incidence (Huff et al., 1989). The occurrence of
the biological plausibility that the agent being lesions presumed to be preneoplastic may in
evaluated is carcinogenic to humans (see Part B, certain instances aid in assessing the biological
Section 2f). plausibility of any neoplastic response observed.
If an agent induces only benign neoplasms that
appear to be end-points that do not readily
undergo transition to malignancy, the agent

21
IARC MONOGRAPHS – 116

should nevertheless be suspected of being Gart et al., 1986; Portier & Bailer, 1989; Bieler &
carcinogenic and requires further investigation. Williams, 1993). The choice of the most appro-
priate statistical method requires consideration
(b) Quantitative aspects of whether or not there are differences in survival
among the treatment groups; for example,
The probability that tumours will occur reduced survival because of non-tumour-re-
may depend on the species, sex, strain, genetic lated mortality can preclude the occurrence of
background and age of the animal, and on the tumours later in life. When detailed information
dose, route, timing and duration of the exposure. on survival is not available, comparisons of the
Evidence of an increased incidence of neoplasms proportions of tumour-bearing animals among
with increasing levels of exposure strengthens the effective number of animals (alive at the time
the inference of a causal association between the the first tumour was discovered) can be useful
exposure and the development of neoplasms. when significant differences in survival occur
The form of the dose–response relationship before tumours appear. The lethality of the
can vary widely, depending on the particular agent tumour also requires consideration: for rapidly
under study and the target organ. Mechanisms fatal tumours, the time of death provides an indi-
such as induction of DNA damage or inhibition cation of the time of tumour onset and can be
of repair, altered cell division and cell death rates assessed using life-table methods; non-fatal or
and changes in intercellular communication incidental tumours that do not affect survival can
are important determinants of dose–response be assessed using methods such as the Mantel-
relationships for some carcinogens. Since many Haenzel test for changes in tumour prevalence.
chemicals require metabolic activation before Because tumour lethality is often difficult to
being converted to their reactive intermediates, determine, methods such as the Poly-K test that
both metabolic and toxicokinetic aspects are do not require such information can also be used.
important in determining the dose–response When results are available on the number and
pattern. Saturation of steps such as absorption, size of tumours seen in experimental animals
activation, inactivation and elimination may (e.g. papillomas on mouse skin, liver tumours
produce nonlinearity in the dose–response rela- observed through nuclear magnetic resonance
tionship (Hoel et al., 1983; Gart et al., 1986), tomography), other more complicated statistical
as could saturation of processes such as DNA procedures may be needed (Sherman et al., 1994;
repair. The dose–response relationship can also Dunson et al., 2003).
be affected by differences in survival among the Formal statistical methods have been devel-
treatment groups. oped to incorporate historical control data into the
analysis of data from a given experiment. These
(c) Statistical analyses methods assign an appropriate weight to histor-
Factors considered include the adequacy of ical and concurrent controls on the basis of the
the information given for each treatment group: extent of between-study and within-study vari-
(i) number of animals studied and number exam- ability: less weight is given to historical controls
ined histologically, (ii) number of animals with a when they show a high degree of variability, and
given tumour type and (iii) length of survival. greater weight when they show little variability. It
The statistical methods used should be clearly is generally not appropriate to discount a tumour
stated and should be the generally accepted tech- response that is significantly increased compared
niques refined for this purpose (Peto et al., 1980; with concurrent controls by arguing that it falls
within the range of historical controls, particularly

22
Preamble

when historical controls show high between- one subsection. For example, a mutation in a
study variability and are, thus, of little relevance gene that codes for an enzyme that metabolizes
to the current experiment. In analysing results the agent under study could be discussed in the
for uncommon tumours, however, the anal- subsections on toxicokinetics, mechanisms and
ysis may be improved by considering historical individual susceptibility if it also exists as an
control data, particularly when between-study inherited polymorphism.
variability is low. Historical controls should be
selected to resemble the concurrent controls as (a) Toxicokinetic data
closely as possible with respect to species, gender
and strain, as well as other factors such as basal Toxicokinetics refers to the absorption,
diet and general laboratory environment, which distribution, metabolism and elimination of
may affect tumour-response rates in control agents in humans, experimental animals and,
animals (Haseman et al., 1984; Fung et al., 1996; where relevant, cellular systems. Examples of
Greim et al., 2003). kinetic factors that may affect dose–response
Although meta-analyses and combined anal- relationships include uptake, deposition, bioper-
yses are conducted less frequently for animal sistence and half-life in tissues, protein binding,
experiments than for epidemiological studies metabolic activation and detoxification. Studies
due to differences in animal strains, they can be that indicate the metabolic fate of the agent
useful aids in interpreting animal data when the in humans and in experimental animals are
experimental protocols are sufficiently similar. summarized briefly, and comparisons of data
from humans and animals are made when
possible. Comparative information on the rela-
4. Mechanistic and other relevant tionship between exposure and the dose that
data reaches the target site may be important for the
extrapolation of hazards between species and in
Mechanistic and other relevant data may clarifying the role of in-vitro findings.
provide evidence of carcinogenicity and also
help in assessing the relevance and importance (b) Data on mechanisms of carcinogenesis
of findings of cancer in animals and in humans.
The nature of the mechanistic and other rele- To provide focus, the Working Group
vant data depends on the biological activity of attempts to identify the possible mechanisms by
the agent being considered. The Working Group which the agent may increase the risk of cancer.
considers representative studies to give a concise For each possible mechanism, a representative
description of the relevant data and issues that selection of key data from humans and experi-
they consider to be important; thus, not every mental systems is summarized. Attention is given
available study is cited. Relevant topics may to gaps in the data and to data that suggests that
include toxicokinetics, mechanisms of carcino- more than one mechanism may be operating.
genesis, susceptible individuals, populations and The relevance of the mechanism to humans is
life-stages, other relevant data and other adverse discussed, in particular, when mechanistic data
effects. When data on biomarkers are informa- are derived from experimental model systems.
tive about the mechanisms of carcinogenesis, Changes in the affected organs, tissues or cells
they are included in this section. can be divided into three non-exclusive levels as
These topics are not mutually exclusive; thus, described below.
the same studies may be discussed in more than

23
IARC MONOGRAPHS – 116

(i) Changes in physiology described for every possible level and mechanism
Physiological changes refer to exposure-re- discussed above.
lated modifications to the physiology and/or Genotoxicity data are discussed here to illus-
response of cells, tissues and organs. Examples trate the key issues involved in the evaluation of
of potentially adverse physiological changes mechanistic data.
include mitogenesis, compensatory cell division, Tests for genetic and related effects are
escape from apoptosis and/or senescence, pres- described in view of the relevance of gene muta-
ence of inflammation, hyperplasia, metaplasia tion and chromosomal aberration/aneuploidy
and/or preneoplasia, angiogenesis, alterations in to carcinogenesis (Vainio et al., 1992; McGregor
cellular adhesion, changes in steroidal hormones et al., 1999). The adequacy of the reporting of
and changes in immune surveillance. sample characterization is considered and, when
necessary, commented upon; with regard to
(ii) Functional changes at the cellular level complex mixtures, such comments are similar
to those described for animal carcinogenicity
Functional changes refer to exposure-re-
tests. The available data are interpreted critically
lated alterations in the signalling pathways used
according to the end-points detected, which
by cells to manage critical processes that are
may include DNA damage, gene mutation, sister
related to increased risk for cancer. Examples
chromatid exchange, micronucleus formation,
of functional changes include modified activ-
chromosomal aberrations and aneuploidy. The
ities of enzymes involved in the metabolism
concentrations employed are given, and mention
of xenobiotics, alterations in the expression
is made of whether the use of an exogenous
of key genes that regulate DNA repair, altera-
metabolic system in vitro affected the test result.
tions in cyclin-dependent kinases that govern
These data are listed in tabular form by phyloge-
cell cycle progression, changes in the patterns
netic classification.
of post-translational modifications of proteins,
Positive results in tests using prokaryotes,
changes in regulatory factors that alter apoptotic
lower eukaryotes, insects, plants and cultured
rates, changes in the secretion of factors related
mammalian cells suggest that genetic and related
to the stimulation of DNA replication and tran-
effects could occur in mammals. Results from
scription and changes in gap–junction-mediated
such tests may also give information on the types
intercellular communication.
of genetic effect produced and on the involve-
(iii) Changes at the molecular level ment of metabolic activation. Some end-points
described are clearly genetic in nature (e.g. gene
Molecular changes refer to exposure-related
mutations), while others are associated with
changes in key cellular structures at the molec-
genetic effects (e.g. unscheduled DNA synthesis).
ular level, including, in particular, genotoxicity.
In-vitro tests for tumour promotion, cell transfor-
Examples of molecular changes include forma-
mation and gap–junction intercellular commu-
tion of DNA adducts and DNA strand breaks,
nication may be sensitive to changes that are not
mutations in genes, chromosomal aberrations,
necessarily the result of genetic alterations but
aneuploidy and changes in DNA methylation
that may have specific relevance to the process of
patterns. Greater emphasis is given to irreversible
carcinogenesis. Critical appraisals of these tests
effects.
have been published (Montesano et al., 1986;
The use of mechanistic data in the identifi-
McGregor et al., 1999).
cation of a carcinogenic hazard is specific to the
Genetic or other activity manifest in humans
mechanism being addressed and is not readily
and experimental mammals is regarded to be of

24
Preamble

greater relevance than that in other organisms. surgical implants of various kinds, and poorly
The demonstration that an agent can induce soluble fibres, dusts and particles of various
gene and chromosomal mutations in mammals sizes, the pathogenic effects of which are a result
in vivo indicates that it may have carcinogenic of their physical presence in tissues or body
activity. Negative results in tests for mutagenicity cavities. Other relevant data for such materials
in selected tissues from animals treated in vivo may include characterization of cellular, tissue
provide less weight, partly because they do not and physiological reactions to these materials
exclude the possibility of an effect in tissues other and descriptions of pathological conditions
than those examined. Moreover, negative results other than neoplasia with which they may be
in short-term tests with genetic end-points associated.
cannot be considered to provide evidence that
rules out the carcinogenicity of agents that act (c) Other data relevant to mechanisms
through other mechanisms (e.g. receptor-medi-
ated effects, cellular toxicity with regenerative A description is provided of any structure–
cell division, peroxisome proliferation) (Vainio activity relationships that may be relevant to an
et al., 1992). Factors that may give misleading evaluation of the carcinogenicity of an agent, the
results in short-term tests have been discussed toxicological implications of the physical and
in detail elsewhere (Montesano et al., 1986; chemical properties, and any other data relevant
McGregor et al., 1999). to the evaluation that are not included elsewhere.
When there is evidence that an agent acts by High-output data, such as those derived
a specific mechanism that does not involve geno- from gene expression microarrays, and high-
toxicity (e.g. hormonal dysregulation, immune throughput data, such as those that result from
suppression, and formation of calculi and other testing hundreds of agents for a single end-point,
deposits that cause chronic irritation), that pose a unique problem for the use of mecha-
evidence is presented and reviewed critically in nistic data in the evaluation of a carcinogenic
the context of rigorous criteria for the operation hazard. In the case of high-output data, there is
of that mechanism in carcinogenesis (e.g. Capen the possibility to overinterpret changes in indi-
et al., 1999). vidual end-points (e.g. changes in expression in
For biological agents such as viruses, one gene) without considering the consistency of
bacteria and parasites, other data relevant to that finding in the broader context of the other
carcinogenicity may include descriptions of the end-points (e.g. other genes with linked transcrip-
pathology of infection, integration and expres- tional control). High-output data can be used in
sion of viruses, and genetic alterations seen in assessing mechanisms, but all end-points meas-
human tumours. Other observations that might ured in a single experiment need to be considered
comprise cellular and tissue responses to infec- in the proper context. For high-throughput data,
tion, immune response and the presence of where the number of observations far exceeds
tumour markers are also considered. the number of end-points measured, their utility
For physical agents that are forms of radia- for identifying common mechanisms across
tion, other data relevant to carcinogenicity may multiple agents is enhanced. These data can be
include descriptions of damaging effects at the used to identify mechanisms that not only seem
physiological, cellular and molecular level, as plausible, but also have a consistent pattern of
for chemical agents, and descriptions of how carcinogenic response across entire classes of
these effects occur. ‘Physical agents’ may also be related compounds.
considered to comprise foreign bodies, such as

25
IARC MONOGRAPHS – 116

(d) Susceptibility data 5. Summary
Individuals, populations and life-stages may This section is a summary of data presented
have greater or lesser susceptibility to an agent, in the preceding sections. Summaries can be
based on toxicokinetics, mechanisms of carcino- found on the Monographs programme web site
genesis and other factors. Examples of host and (http://monographs.iarc.fr).
genetic factors that affect individual susceptibility
include sex, genetic polymorphisms of genes (a) Exposure data
involved in the metabolism of the agent under
evaluation, differences in metabolic capacity due Data are summarized, as appropriate, on
to life-stage or the presence of disease, differ- the basis of elements such as production, use,
ences in DNA repair capacity, competition for occurrence and exposure levels in the work-
or alteration of metabolic capacity by medica- place and environment and measurements in
tions or other chemical exposures, pre-existing human tissues and body fluids. Quantitative
hormonal imbalance that is exacerbated by a data and time trends are given to compare
chemical exposure, a suppressed immune system, exposures in different occupations and environ-
periods of higher-than-usual tissue growth or mental settings. Exposure to biological agents is
regeneration and genetic polymorphisms that described in terms of transmission, prevalence
lead to differences in behaviour (e.g. addiction). and persistence of infection.
Such data can substantially increase the strength
of the evidence from epidemiological data and (b) Cancer in humans
enhance the linkage of in-vivo and in-vitro labo-
Results of epidemiological studies pertinent
ratory studies to humans.
to an assessment of human carcinogenicity are
summarized. When relevant, case reports and
(e) Data on other adverse effects correlation studies are also summarized. The
Data on acute, subchronic and chronic target organ(s) or tissue(s) in which an increase in
adverse effects relevant to the cancer evaluation cancer was observed is identified. Dose–response
are summarized. Adverse effects that confirm and other quantitative data may be summarized
distribution and biological effects at the sites of when available.
tumour development, or alterations in physi-
ology that could lead to tumour development, are (c) Cancer in experimental animals
emphasized. Effects on reproduction, embryonic
Data relevant to an evaluation of carcino-
and fetal survival and development are summa-
genicity in animals are summarized. For each
rized briefly. The adequacy of epidemiological
animal species, study design and route of admin-
studies of reproductive outcome and genetic
istration, it is stated whether an increased inci-
and related effects in humans is judged by the
dence, reduced latency, or increased severity
same criteria as those applied to epidemiological
or multiplicity of neoplasms or preneoplastic
studies of cancer, but fewer details are given.
lesions were observed, and the tumour sites are
indicated. If the agent produced tumours after
prenatal exposure or in single-dose experiments,
this is also mentioned. Negative findings, inverse
relationships, dose–response and other quantita-
tive data are also summarized.

26
Preamble

(d) Mechanistic and other relevant data (a) Carcinogenicity in humans
Data relevant to the toxicokinetics (absorp- The evidence relevant to carcinogenicity
tion, distribution, metabolism, elimination) and from studies in humans is classified into one of
the possible mechanism(s) of carcinogenesis (e.g. the following categories:
genetic toxicity, epigenetic effects) are summa- Sufficient evidence of carcinogenicity:
rized. In addition, information on susceptible The Working Group considers that a causal
individuals, populations and life-stages is relationship has been established between expo-
summarized. This section also reports on other sure to the agent and human cancer. That is, a
toxic effects, including reproductive and devel- positive relationship has been observed between
opmental effects, as well as additional relevant the exposure and cancer in studies in which
data that are considered to be important. chance, bias and confounding could be ruled
out with reasonable confidence. A statement that
there is sufficient evidence is followed by a sepa-
6. Evaluation and rationale rate sentence that identifies the target organ(s) or
Evaluations of the strength of the evidence for tissue(s) where an increased risk of cancer was
carcinogenicity arising from human and exper- observed in humans. Identification of a specific
imental animal data are made, using standard target organ or tissue does not preclude the
terms. The strength of the mechanistic evidence possibility that the agent may cause cancer at
is also characterized. other sites.
It is recognized that the criteria for these Limited evidence of carcinogenicity:
evaluations, described below, cannot encompass A positive association has been observed
all of the factors that may be relevant to an eval- between exposure to the agent and cancer for
uation of carcinogenicity. In considering all of which a causal interpretation is considered by
the relevant scientific data, the Working Group the Working Group to be credible, but chance,
may assign the agent to a higher or lower cate- bias or confounding could not be ruled out with
gory than a strict interpretation of these criteria reasonable confidence.
would indicate. Inadequate evidence of carcinogenicity:
These categories refer only to the strength of The available studies are of insufficient
the evidence that an exposure is carcinogenic quality, consistency or statistical power to permit
and not to the extent of its carcinogenic activity a conclusion regarding the presence or absence
(potency). A classification may change as new of a causal association between exposure and
information becomes available. cancer, or no data on cancer in humans are
An evaluation of the degree of evidence is available.
limited to the materials tested, as defined phys- Evidence suggesting lack of carcinogenicity:
ically, chemically or biologically. When the There are several adequate studies covering
agents evaluated are considered by the Working the full range of levels of exposure that humans
Group to be sufficiently closely related, they may are known to encounter, which are mutually
be grouped together for the purpose of a single consistent in not showing a positive association
evaluation of the degree of evidence. between exposure to the agent and any studied
cancer at any observed level of exposure. The
results from these studies alone or combined
should have narrow confidence intervals with an
upper limit close to the null value (e.g. a relative

27
IARC MONOGRAPHS – 116

risk of 1.0). Bias and confounding should be ruled or more species of animals or (b) two or more
out with reasonable confidence, and the studies independent studies in one species carried out
should have an adequate length of follow-up. A at different times or in different laboratories or
conclusion of evidence suggesting lack of carcino- under different protocols. An increased incidence
genicity is inevitably limited to the cancer sites, of tumours in both sexes of a single species in a
conditions and levels of exposure, and length of well conducted study, ideally conducted under
observation covered by the available studies. In Good Laboratory Practices, can also provide
addition, the possibility of a very small risk at the sufficient evidence.
levels of exposure studied can never be excluded. A single study in one species and sex might
In some instances, the above categories may be considered to provide sufficient evidence of
be used to classify the degree of evidence related carcinogenicity when malignant neoplasms occur
to carcinogenicity in specific organs or tissues. to an unusual degree with regard to incidence,
When the available epidemiological studies site, type of tumour or age at onset, or when there
pertain to a mixture, process, occupation or are strong findings of tumours at multiple sites.
industry, the Working Group seeks to identify Limited evidence of carcinogenicity:
the specific agent considered most likely to be The data suggest a carcinogenic effect but
responsible for any excess risk. The evaluation are limited for making a definitive evaluation
is focused as narrowly as the available data on because, e.g. (a) the evidence of carcinogenicity
exposure and other aspects permit. is restricted to a single experiment; (b) there are
unresolved questions regarding the adequacy
(b) Carcinogenicity in experimental of the design, conduct or interpretation of the
animals studies; (c) the agent increases the incidence
only of benign neoplasms or lesions of uncer-
Carcinogenicity in experimental animals tain neoplastic potential; or (d) the evidence
can be evaluated using conventional bioassays, of carcinogenicity is restricted to studies that
bioassays that employ genetically modified demonstrate only promoting activity in a narrow
animals, and other in-vivo bioassays that focus range of tissues or organs.
on one or more of the critical stages of carcino- Inadequate evidence of carcinogenicity:
genesis. In the absence of data from conventional The studies cannot be interpreted as showing
long-term bioassays or from assays with neoplasia either the presence or absence of a carcinogenic
as the end-point, consistently positive results in effect because of major qualitative or quantitative
several models that address several stages in the limitations, or no data on cancer in experimental
multistage process of carcinogenesis should be animals are available.
considered in evaluating the degree of evidence Evidence suggesting lack of carcinogenicity:
of carcinogenicity in experimental animals. Adequate studies involving at least two
The evidence relevant to carcinogenicity in species are available which show that, within the
experimental animals is classified into one of the limits of the tests used, the agent is not carcino-
following categories: genic. A conclusion of evidence suggesting lack
Sufficient evidence of carcinogenicity: of carcinogenicity is inevitably limited to the
The Working Group considers that a causal species, tumour sites, age at exposure, and condi-
relationship has been established between the tions and levels of exposure studied.
agent and an increased incidence of malignant
neoplasms or of an appropriate combination
of benign and malignant neoplasms in (a) two

28
Preamble

(c) Mechanistic and other relevant data experimental animals and whether a unique
mechanism might operate in a susceptible group.
Mechanistic and other evidence judged to be The possible contribution of alternative mecha-
relevant to an evaluation of carcinogenicity and nisms must be considered before concluding
of sufficient importance to affect the overall eval- that tumours observed in experimental animals
uation is highlighted. This may include data on are not relevant to humans. An uneven level of
preneoplastic lesions, tumour pathology, genetic experimental support for different mechanisms
and related effects, structure–activity relation- may reflect that disproportionate resources
ships, metabolism and toxicokinetics, physico- have been focused on investigating a favoured
chemical parameters and analogous biological mechanism.
agents. For complex exposures, including occupa-
The strength of the evidence that any carcino- tional and industrial exposures, the chemical
genic effect observed is due to a particular mech- composition and the potential contribution of
anism is evaluated, using terms such as ‘weak’, carcinogens known to be present are considered
‘moderate’ or ‘strong’. The Working Group then by the Working Group in its overall evaluation
assesses whether that particular mechanism is of human carcinogenicity. The Working Group
likely to be operative in humans. The strongest also determines the extent to which the mate-
indications that a particular mechanism oper- rials tested in experimental systems are related
ates in humans derive from data on humans to those to which humans are exposed.
or biological specimens obtained from exposed
humans. The data may be considered to be espe-
cially relevant if they show that the agent in
(d) Overall evaluation
question has caused changes in exposed humans Finally, the body of evidence is considered
that are on the causal pathway to carcinogenesis. as a whole, to reach an overall evaluation of the
Such data may, however, never become available, carcinogenicity of the agent to humans.
because it is at least conceivable that certain An evaluation may be made for a group of
compounds may be kept from human use solely agents that have been evaluated by the Working
on the basis of evidence of their toxicity and/or Group. In addition, when supporting data indi-
carcinogenicity in experimental systems. cate that other related agents, for which there is no
The conclusion that a mechanism operates direct evidence of their capacity to induce cancer
in experimental animals is strengthened by in humans or in animals, may also be carcino-
findings of consistent results in different experi- genic, a statement describing the rationale for
mental systems, by the demonstration of biolog- this conclusion is added to the evaluation narra-
ical plausibility and by coherence of the overall tive; an additional evaluation may be made for
database. Strong support can be obtained from this broader group of agents if the strength of the
studies that challenge the hypothesized mecha- evidence warrants it.
nism experimentally, by demonstrating that the The agent is described according to the
suppression of key mechanistic processes leads wording of one of the following categories, and
to the suppression of tumour development. The the designated group is given. The categorization
Working Group considers whether multiple of an agent is a matter of scientific judgement that
mechanisms might contribute to tumour devel- reflects the strength of the evidence derived from
opment, whether different mechanisms might studies in humans and in experimental animals
operate in different dose ranges, whether sepa- and from mechanistic and other relevant data.
rate mechanisms might operate in humans and

29
IARC MONOGRAPHS – 116

Group 1: The agent is carcinogenic to be classified in this category solely on the basis of
humans. limited evidence of carcinogenicity in humans. An
This category is used when there is suffi- agent may be assigned to this category if it clearly
cient evidence of carcinogenicity in humans. belongs, based on mechanistic considerations, to
Exceptionally, an agent may be placed in this a class of agents for which one or more members
category when evidence of carcinogenicity in have been classified in Group 1 or Group 2A.
humans is less than sufficient but there is suffi- Group 2B: The agent is possibly
cient evidence of carcinogenicity in experimental carcinogenic to humans.
animals and strong evidence in exposed humans
that the agent acts through a relevant mechanism This category is used for agents for which
of carcinogenicity. there is limited evidence of carcinogenicity in
humans and less than sufficient evidence of
Group 2. carcinogenicity in experimental animals. It may
This category includes agents for which, at also be used when there is inadequate evidence
one extreme, the degree of evidence of carcino- of carcinogenicity in humans but there is suffi-
genicity in humans is almost sufficient, as well as cient evidence of carcinogenicity in experimental
those for which, at the other extreme, there are animals. In some instances, an agent for which
no human data but for which there is evidence there is inadequate evidence of carcinogenicity
of carcinogenicity in experimental animals. in humans and less than sufficient evidence of
Agents are assigned to either Group 2A (probably carcinogenicity in experimental animals together
carcinogenic to humans) or Group 2B (possibly with supporting evidence from mechanistic and
carcinogenic to humans) on the basis of epidemi- other relevant data may be placed in this group.
ological and experimental evidence of carcino- An agent may be classified in this category solely
genicity and mechanistic and other relevant data. on the basis of strong evidence from mechanistic
The terms probably carcinogenic and possibly and other relevant data.
carcinogenic have no quantitative significance
and are used simply as descriptors of different Group 3: The agent is not classifiable as
levels of evidence of human carcinogenicity, with to its carcinogenicity to humans.
probably carcinogenic signifying a higher level of This category is used most commonly for
evidence than possibly carcinogenic. agents for which the evidence of carcinogenicity
Group 2A: The agent is probably is inadequate in humans and inadequate or
carcinogenic to humans. limited in experimental animals.
Exceptionally, agents for which the evidence
This category is used when there is limited of carcinogenicity is inadequate in humans but
evidence of carcinogenicity in humans and suffi- sufficient in experimental animals may be placed
cient evidence of carcinogenicity in experimental in this category when there is strong evidence
animals. In some cases, an agent may be clas- that the mechanism of carcinogenicity in exper-
sified in this category when there is inadequate imental animals does not operate in humans.
evidence of carcinogenicity in humans and suffi- Agents that do not fall into any other group
cient evidence of carcinogenicity in experimental are also placed in this category.
animals and strong evidence that the carcino- An evaluation in Group 3 is not a determi-
genesis is mediated by a mechanism that also nation of non-carcinogenicity or overall safety.
operates in humans. Exceptionally, an agent may It often means that further research is needed,

30
Preamble

especially when exposures are widespread or References
the cancer data are consistent with differing
interpretations. Bieler GS, Williams RL (1993). Ratio estimates, the
delta method, and quantal response tests for
Group 4: The agent is probably not increased carcinogenicity. Biometrics, 49:793–801.
carcinogenic to humans. doi:10.2307/2532200 PMID:8241374
Breslow NE, Day NE (1980). Statistical methods in cancer
This category is used for agents for which research. Volume I - The analysis of case-control
there is evidence suggesting lack of carcinogenicity studies. IARC Sci Publ, 32:5–338. PMID:7216345
Breslow NE, Day NE (1987). Statistical methods in cancer
in humans and in experimental animals. In research. Volume II–The design and analysis of cohort
some instances, agents for which there is inad- studies. IARC Sci Publ, 82:1–406. PMID:3329634
equate evidence of carcinogenicity in humans Buffler P, Rice J, Baan R et al. (2004). Workshop on mech-
but evidence suggesting lack of carcinogenicity in anisms of carcinogenesis: contributions of molecular
epidemiology. Lyon, 14–17 November 2001. Workshop
experimental animals, consistently and strongly report. IARC Sci Publ, 157:1–27. PMID:15055286
supported by a broad range of mechanistic and Capen CC, Dybing E, Rice JM, Wilbourn JD (1999).
other relevant data, may be classified in this Species differences in thyroid, kidney and urinary
bladder carcinogenesis. Proceedings of a consensus
group. conference. Lyon, France, 3–7 November 1997. IARC
Sci Publ, 147:1–225. PMID:10627184
(e) Rationale Cogliano V, Baan R, Straif K et al. (2005). Transparency
in IARC Monographs. Lancet Oncol, 6:747. doi:10.1016/
The reasoning that the Working Group used S1470-2045(05)70380-6
to reach its evaluation is presented and discussed. Cogliano VJ, Baan RA, Straif K et al. (2004). The science
and practice of carcinogen identification and eval-
This section integrates the major findings from uation. Environ Health Perspect, 112:1269–1274.
studies of cancer in humans, studies of cancer doi:10.1289/ehp.6950 PMID:15345338
in experimental animals, and mechanistic and Dunson DB, Chen Z, Harry J (2003). A Bayesian approach
for joint modeling of cluster size and subunit-specific
other relevant data. It includes concise state- outcomes. Biometrics, 59:521–530. doi:10.1111/1541-
ments of the principal line(s) of argument that 0420.00062 PMID:14601753
emerged, the conclusions of the Working Group Fung KY, Krewski D, Smythe RT (1996). A comparison
on the strength of the evidence for each group of tests for trend with historical controls in carcinogen
bioassay. Can J Stat, 24:431–454. doi:10.2307/3315326
of studies, citations to indicate which studies Gart JJ, Krewski D, Lee PN et al. (1986). Statistical
were pivotal to these conclusions, and an expla- methods in cancer research. Volume III–The design
nation of the reasoning of the Working Group and analysis of long-term animal experiments. IARC
in weighing data and making evaluations. When Sci Publ, 79:1–219. PMID:3301661
Greenland S (1998). Meta-analysis. In: Rothman
there are significant differences of scientific KJ, Greenland S, editors. Modern epidemiology.
interpretation among Working Group Members, Philadelphia: Lippincott Williams & Wilkins, pp.
a brief summary of the alternative interpreta- 643–673.
Greim H, Gelbke H-P, Reuter U et al. (2003).
tions is provided, together with their scientific Evaluation of historical control data in carcino-
rationale and an indication of the relative degree genicity studies. Hum Exp Toxicol, 22:541–549.
of support for each alternative. doi:10.1191/0960327103ht394oa PMID:14655720
Haseman JK, Huff J, Boorman GA (1984). Use
of historical control data in carcinogenicity
studies in rodents. Toxicol Pathol, 12:126–135.
doi:10.1177/019262338401200203 PMID:11478313
Hill AB (1965). The environment and disease:
Association or causation? Proc R Soc Med, 58:295–300.
PMID:14283879

31
IARC MONOGRAPHS – 116

Hoel DG, Kaplan NL, Anderson MW (1983). Implication of OECD (2002). Guidance notes for analysis and evaluation
nonlinear kinetics on risk estimation in carcinogenesis. of chronic toxicity and carcinogenicity studies (Series
Science, 219:1032–1037. doi:10.1126/science.6823565 on Testing and Assessment No. 35), Paris: OECD.
PMID:6823565 Peto R, Pike MC, Day NE et al. (1980). Guidelines for
Huff JE, Eustis SL, Haseman JK (1989). Occurrence and simple, sensitive significance tests for carcinogenic
relevance of chemically induced benign neoplasms in effects in long-term animal experiments. IARC
long-term carcinogenicity studies. Cancer Metastasis Monogr Eval Carcinog Risk Chem Hum Suppl, 2:Suppl:
Rev, 8:1–22. doi:10.1007/BF00047055 PMID:2667783 311–426. PMID:6935185
IARC (1977). IARC Monographs Programme on the Portier CJ, Bailer AJ (1989). Testing for increased carcino-
Evaluation of the Carcinogenic Risk of Chemicals to genicity using a survival-adjusted quantal response test.
Humans. Preamble (IARC Intern Tech Rep No. 77/002). Fundam Appl Toxicol, 12:731–737. doi:10.1016/0272-
IARC (1978). Chemicals with sufficient evidence of 0590(89)90004-3 PMID:2744275
carcinogenicity in experimental animals – IARC Sherman CD, Portier CJ, Kopp-Schneider A (1994).
Monographs Volumes 1–17 (IARC Intern Tech Rep Multistage models of carcinogenesis: an approx-
No. 78/003). imation for the size and number distribution
IARC (1979). Criteria to select chemicals for IARC of late-stage clones. Risk Anal, 14:1039–1048.
Monographs (IARC Intern Tech Rep No. 79/003). doi:10.1111/j.1539-6924.1994.tb00074.x PMID:7846311
IARC (1982). Chemicals, industrial processes and Stewart BW, Kleihues P, editors (2003). World cancer
industries associated with cancer in humans (IARC report, Lyon: IARC.
Monographs, volumes 1 to 29). IARC Monogr Eval Tomatis L, Aitio A, Wilbourn J, Shuker L (1989). Human
Carcinog Risk Chem Hum Suppl, 4:1–292. carcinogens so far identified. Jpn J Cancer Res,
IARC (1983). Approaches to classifying chemical carcin- 80:795–807. doi:10.1111/j.1349-7006.1989.tb01717.x
ogens according to mechanism of action (IARC Intern PMID:2513295
Tech Rep No. 83/001). Toniolo P, Boffetta P, Shuker DEG et al. (1997). Proceedings
IARC (1987). Overall evaluations of carcinogenicity: an of the workshop on application of biomarkers to cancer
updating of IARC Monographs volumes 1 to 42. IARC epidemiology. Lyon, France, 20–23 February 1996.
Monogr Eval Carcinog Risks Hum Suppl, 7:1–440. IARC Sci Publ, 142:1–318. PMID:9410826
PMID:3482203 Vainio H, Magee P, McGregor D, McMichael A (1992).
IARC (1988). Report of an IARC Working Group to Review Mechanisms of carcinogenesis in risk identification.
the Approaches and Processes Used to Evaluate the IARC Working Group Meeting. Lyon, 11–18 June 1991.
Carcinogenicity of Mixtures and Groups of Chemicals IARC Sci Publ, 116:1–608. PMID:1428077
(IARC Intern Tech Rep No. 88/002). Vainio H, Wilbourn JD, Sasco AJ et al. (1995).
IARC (1991). A consensus report of an IARC Monographs [Identification of human carcinogenic risks in IARC
Working Group on the Use of Mechanisms of monographs] Bull Cancer, 82:339–348. PMID:7626841
Carcinogenesis in Risk Identification (IARC Intern Vineis P, Malats N, Lang M et al., editors (1999). Metabolic
Tech Rep No. 91/002). polymorphisms and susceptibility to cancer. IARC Sci
IARC (2004). Some drinking-water disinfectants and Publ, 148:1–510. PMID:10493243
contaminants, including arsenic. IARC Monogr Eval Wilbourn J, Haroun L, Heseltine E et al. (1986).
Carcinog Risks Hum, 84:1–477. PMID:15645577 Response of experimental animals to human carcin-
IARC (2005). Report of the Advisory Group to Recommend ogens: an analysis based upon the IARC Monographs
Updates to the Preamble to the IARC Monographs programme. Carcinogenesis, 7:1853–1863. doi:10.1093/
(IARC Intern Rep No. 05/001). carcin/7.11.1853 PMID:3769134
IARC (2006). Report of the Advisory Group to Review the
Amended Preamble to the IARC Monographs (IARC
Intern Rep No. 06/001).
McGregor DB, Rice JM, Venitt S (1999). The use of
short-and medium-term tests for carcinogens and
data on genetic effects in carcinogenic hazard eval-
uation. Consensus report. IARC Sci Publ, 146:1–18.
PMID:10353381
Montesano R, Bartsch H, Vainio H et al., editors(1986).
Long-term and short-term assays for carcinogene-
sis—a critical appraisal. IARC Sci Publ, 83:1–564.
PMID:3623675

32
GENERAL REMARKS
This one-hundred-and-sixteenth volume of the IARC Monographs presents evaluations of
the carcinogenic hazard to humans of drinking coffee, mate, and very hot beverages. A
summary of the findings of this volume appears in The Lancet Oncology (Loomis et al., 2016).

The carcinogenicity of coffee was previously for cancer of the bladder that is often strongly
evaluated in Volume 51 of the IARC Monographs associated with coffee drinking. In considering
(IARC, 1991). After reviewing the data available the data now available for more than 20 other
at that time, the Working Group had classi- cancer sites in humans, the Working Group
fied coffee as possibly carcinogenic to humans found evidence suggesting lack of carcinogenicity
(Group 2B) based on limited evidence in humans for cancers of the female breast, uterine endome-
– derived from some 20 epidemiological case– trium, prostate, pancreas, and liver, and inade-
control studies – that coffee causes cancer of quate evidence in humans for cancers at all other
the urinary bladder, and inadequate evidence in sites. The Working Group’s review of other rele-
experimental animals. The same Working Group vant data found strong evidence in humans that
also concluded that there was evidence suggesting coffee has antioxidant effects. As a result of this
lack of carcinogenicity for cancers of the female re-evaluation, the Working Group concluded
breast and the colorectum. that drinking coffee is not classifiable as to its
In the current evaluation, based on a much carcinogenicity to humans (Group 3).
larger volume of data comprising more than An earlier evaluation of the carcinogenicity
1000 observational and experimental studies, the of mate was also reported in Volume 51 (IARC,
Working Group concluded there is inadequate 1991). The evidence available at that time was
evidence in humans and experimental animals obtained entirely from epidemiological case–
for the carcinogenicity of coffee drinking. With control studies. In that review, the Working
the expanded literature, the Working Group Group drew a distinction between mate itself
focused their review on higher-quality epide- and drinking hot mate, concluding that mate
miological studies of cancer of the bladder and (without further specification) was not classifiable
coffee drinking; these did not show a consistent as to its carcinogenicity to humans (Group 3), but
association or a dose–response relationship. The that drinking hot mate was probably carcinogenic
Working Group judged that the positive associa- to humans (Group 2A). Taking into account the
tions between coffee drinking and cancer of the previous evaluation, in addition to new data in
bladder observed in some studies were probably humans and experimental animals, an Advisory
due to inadequate control for the confounding Group that met in 2014 gave high priority to eval-
effects of tobacco smoking, a major risk factor uating the carcinogenicity of drinking hot mate

33
IARC MONOGRAPHS – 116

and other hot beverages (Straif et al., 2014). In
light of the evidence available at the present time,
the current Working Group chose to evaluate the
carcinogenicity of very hot beverages, including,
but not limited to, mate. Epidemiological studies
of cancer risk and drinking temperature for a
variety of hot beverages, as well as co-carcino-
genicity experiments in which hot liquids were
administered to animals, were accordingly
taken into consideration. The Working Group
concluded that drinking very hot beverages
(>  65  °C) is probably carcinogenic to humans
(Group 2A) based on epidemiological studies
showing limited evidence of a causal association
with cancer of the oesophagus in humans and
limited evidence in experimental animals. The
Working Group noted that a causal relation-
ship between consuming very hot beverages and
cancer of the oesophagus is biologically plausible
through mechanisms linking thermal injury to
cancer. Drinking mate that is not very hot was
classified in Group 3 (not classifiable as to its
carcinogenicity to humans).

References
IARC (1991). Coffee, tea, mate, methylxanthines and
methylglyoxal. IARC Monogr Eval Carcinog Risks Hum,
51:1–513. Available from: http://monographs.iarc.fr/
ENG/Monographs/vol51/index.php PMID:1674554
Loomis D, Guyton KZ, Grosse Y, Lauby-Secretan B,
El Ghissassi F, Bouvard V, et al.; International Agency
for Research on Cancer Monograph Working Group
(2016). Carcinogenicity of drinking coffee, mate,
and very hot beverages. Lancet Oncol, 17(7):877–8.
doi:10.1016/S1470-2045(16)30239-X PMID:27318851
Straif K, Loomis D, Guyton K, Grosse Y, Lauby-Secretan
B, El Ghissassi F, et al. (2014). Future priorities for
the IARC Monographs. Lancet Oncol, 15(7):683–4.
doi:10.1016/S1470-2045(14)70168-8

34
DRINKING COFFEE
1. EXPOSURE DATA

Coffee seeds (known as beans) are contained colonies in North America (Wellman, 1961;
in fruits from trees and shrubs grown naturally IARC, 1991).
in the shade of eastern African forests encom- During the 17th century the cultivation
passing Ethiopia and the islands of Madagascar of coffee spread to the Malabar coast of India
and Mauritius, among other countries. Coffee and to Ceylon. From the beginning of the 18th
has attracted the attention of explorers and century, seedlings of Coffea arabica L. cultivated
botanists from all over the world from the 16th in European glasshouses, as first described by
century, when the first coffee trees were reported Linnaeus in 1737 (Debry, 1989), were introduced
in the literature (Charrier & Berthaud, 1987); in to the Dutch West Indies and to the French,
particular, many new species were discovered in Portuguese, and Spanish colonies of America
the second half of the 19th century. The interest and Asia (Wellman, 1961; IARC, 1991; Davis
may have been partly due to its stimulating effects et al., 2007; Farah & Ferreira dos Santos, 2015).
in animals and humans, compounded with its
enchanting aroma after roasting. Today, it is
known that different coffee species originated
1.1 Identification of the agent
in different parts of Africa and that there are 1.1.1 Botanical data
still species being discovered, some in Ethiopia
(Farah & Ferreira dos Santos, 2015). (a) Nomenclature
The year 575 AD is often cited as the date of
Botanical name: Coffea arabica L., Coffea
the arrival of coffee on the Arabian peninsula
canephora Pierre ex A. Froehner, and various
from Ethiopia. Commercial and political links
other species in the genus Coffea
were at that time strengthening across the Red
Sea (Wellman, 1961; IARC, 1991). Coffee cher- Family: Rubiaceae
ries (bun or bon) were then probably only dried Subfamily: Ixoroideae
and chewed as a stimulant against fatigue. It was Tribe: Coffeeae
only by the middle of the 15th century that coffee Genus: Coffea
as a beverage (kahwah in Arabic), an infusion of
Subgenus: Coffea
roasted and ground coffee beans cultivated in
Yemen, near the harbour of Mocha, came into Common names: coffee, Arabica coffee,
general use throughout the Ottoman empire. By Robusta coffee
the end of the 16th century it had crossed the GRIN (2016)
Mediterranean Sea, and in less than a century it The coffee tree belongs to the botanical
had spread throughout Europe and to the British family Rubiaceae, with the genus Coffea being

37
IARC MONOGRAPHS – 116

Fig. 1.1 The coffee plant, Coffea arabica L.

The figure shows the leaves, fruits, and berries.
From Spohn (2015) with permission from Dr Roland Spohn, www.spohns.de

the most economically important member of Lashermes et al., 1999; Anthony et al., 2002;
this family (Murthy & Naidu, 2012). Because of Farah & Ferreira dos Santos, 2015).
the great variation in the types of coffee plants
and seeds, botanists have failed to agree on a (b) Description of the coffee plant
precise single system to classify them or even to Coffee is an evergreen perennial plant. The
designate some plants as true members of the shapes of the coffee tree and roots vary depending
Coffea genus. Although it is said that hundreds of on the species and, in some cases, variety (Murthy
species have been described, the National Center & Naidu, 2012; FAO, 2014). In general, the coffee
of Biotechnology Information (NCBI) in the tree consists of an upright main shoot (trunk)
USA and Davis et al. (2006; 2007) have described with primary, secondary, and tertiary lateral
over 90 species within the Coffea genus, from branches. Naturally grown trees may be 4–6 m
which 25 have been more extensively studied tall for C. arabica and 8–12 m for C. canephora
(Davis et al., 2006; 2007; NCBI, 2014; Farah & (Illy & Viani, 2005).
Ferreira dos Santos, 2015). From these 25 species, Each leaf pair is opposite to the next leaf pair.
only two have major commercial importance: Leaves appear shiny, wavy, and dark green in
Coffea arabica and Coffea canephora. It has been colour with conspicuous veins. In the axil of each
suggested that C. arabica, a tetraploid species leaf are four to six serial buds, which can develop
(2n = 4x = 44), originated from natural hybridi- into an inflorescence or secondary branches
zation between C. canephora and C. eugenioides, (Farah & Ferreira dos Santos, 2015; Fig. 1.1). The
or ecotypes related to these two diploid mature fruit, or “cherry” (Fig.  1.2), comprises:
(2n = 2x = 22) species (Charrier & Berthaud 1987; (1) skin (epicarp or exocarp), which is a red, dark

38
Drinking coffee

Fig. 1.2 Diagram of the coffee cherry fruit

Reproduced from Farah & Ferreira dos Santos (2015). The coffee plant and beans: introduction. In: Preedy V, editor, Coffee in health and disease
prevention, 1st ed. San Diego (CA) USA: Academic Press; 5–10.

pink or yellow monocellular layer covered with Examples of countries that have this climate are
a waxy substance protecting the fruit; (2) pulp Colombia, Costa Rica, Ethiopia, and Kenya (Illy
(mesocarp); (3) parchment or parch (endocarp); & Viani, 2005).
(4) silverskin, which is the seed coat composed C. canephora trees also grow at low elevations
mainly of polysaccharides (especially cellulose, (from sea level to 900 m) in warmer climates in
hemicelluloses, and mannans); (5) two elliptical the equatorial regions at latitudes below 10°, and
seeds (beans) containing the endosperm and demonstrate higher resistance to diseases, but
embryo (CAC, 2009; Murthy & Naidu, 2012; yield a beverage of inferior quality and lower
Sánchez & Anzola, 2012; FAO, 2014; Farah & market value compared to Arabica species. The
Ferreira dos Santos, 2015). species Coffea liberica, commanding less than
Arabica coffee grows optimally at altitudes 1% of the market, grows in warm climates and
of 550–1100 m in subtropical regions of latitudes at low elevations; it is however susceptible to
16–24° with well-defined rainy and dry seasons. diseases and yields a beverage of poor quality
The Brazilian regions of Minas Gerais and São (Illy & Illy, 1989; Hendre et al., 2008; FAO, 2014;
Paulo, and Jamaica, Mexico, and Zimbabwe are Farah & Ferreira dos Santos, 2015; ICO, 2016).
examples of areas with these climate conditions
(Illy & Viani, 2005). In the equatorial regions
at latitudes below 10° coffee grows well at the
higher altitudes of 1100–1900 m. Frequent rain-
fall causes almost continuous flowering, which
results in two coffee harvesting seasons per year.

39
IARC MONOGRAPHS – 116

1.2 Methods of production, uses, been valued in the market for the preparation of
and preparation blends, as the silverskin confers more “body” or
thickness to the brew (Borrelli et al., 2004).
1.2.1 Green coffee production The wet process is more sophisticated and
tends to generate a higher-quality beverage;
(a) Harvesting
generally only ripe cherries are processed this
Harvesting begins when about 80% of the way. They can be selectively hand-picked or sepa-
fruits are ripe. Coffee cherries (ripe fruits) tend rated mechanically or in flotation tanks. Sorting
to yield better-quality beverages, whereas imma- is followed by mechanical (de)pulping, soaking,
ture and overripe fruits yield defective low-quality and fermenting in tanks, where the remaining
beans (Toci & Farah, 2008). Harvesting may be pulp and silverskin are removed; acidity increases
undertaken manually or mechanically. Manual during this process and the pH may reduce to 4.5.
picking tends to yield better-quality beans The naked beans are washed and then dried in
(Farah, 2009; Filho et al., 2015). yards or in ventilated tables, possibly combined
with hot air drying. After drying, the remaining
(b) Post-harvest processing
part of the hull is often mechanically removed.
Fig.  1.3 summarizes the steps involved in Wet processing is frequently used in places where
dry, semi-dry (or semi-wet/semi-washed), and coffee is harvested by manual picking such as
wet methods used for coffee primary processing. Asia, Central America, and Colombia; due to
After being harvested, the fruits are either the higher market value, however, various farms
sorted manually or washed and separated in in larger coffee-producing countries such as
flotation tanks, followed by processing for the Brazil have also adopted this processing method.
separation of the seeds from the rest of the fruit. Wet-processed beans are called pulped coffees
In the original dry processing method, (Flament, 2002, Bee et al., 2005; Knopp et al.,
harvested seeds are parched by sun exposure 2006; CAC 2009; Farah, 2009; Farah & Ferreira
outdoors and/or by air dryers until the moisture dos Santos, 2015).
content is about 10–12% (Trugo, 2003; Farah, At all steps of coffee processing, gram-nega-
2009). Unless air dryers are available, protec- tive and gram-positive bacteria, yeasts, and fila-
tion from rain during the harvesting period is mentous fungi are present at high levels. There
required to avoid the growth of microorganisms has been a concern about the potential prod-
and to produce good-quality coffee (CAC, 2009; uction of ochratoxin A and other mycotoxins by
Farah, 2009). Once the fruits are dried, they are microorganisms during the fermentation in wet
cleaned and the dried pericarp (endocarp, meso- processing, when the natural growth of micro-
carp, and epicarp) is removed mechanically organisms can occur (see Section 1.4.2).
(Fig. 1.2), leaving the mucilaginous material that After the beans are treated by the dry or wet
envelops the seeds (silverskin) still adhering to method, they are either stored or mechanically,
their surface (Geromel et al., 2006; CAC, 2009; manually, and/or electronically sized and sorted
Farah, 2009). The product of dry processed fruits for quality control and commercialization. This
is called “natural” green coffee (CAC, 2009; process may be followed by an additional sorting
Farah, 2009). The dry method is commonly used with UV excitation.
in Brazil and Africa (Farah & Ferreira dos Santos, Sorting yields coffee beans with extrinsic
2015). Seeds produced by the dry method keep defects, such as stones, twigs, or other foreign
the silverskin adhered to their surface and have matter, and intrinsic defects, such as immature,

40
Drinking coffee

Fig. 1.3 Coffee post-harvest processing: flow of dry, wet, and semi-dry/semi-washed methods

Reproduced from Farah & Ferreira dos Santos (2015). The coffee plant and beans: Introduction. In: Preedy V; ed., Coffee in health and disease
prevention, 1st ed. San Diego (CA), USA: Academic Press; 5–10.

41
IARC MONOGRAPHS – 116

sour, or insect-damaged beans (Toci & Farah, employed at temperatures and pressures of
2008; Farah, 2009; Farah & Ferreira dos Santos, 40–80 °C and 200–300 bar (i.e. above its critical
2015). Avoiding undesirable contamination point of 31.06  °C and 73.8  bar) for 5–30 hours
and the growth of microorganisms (especially (IARC, 1991). Dichloromethane or ethyl acetate
mould) during harvesting, drying, and storage are used in Central and South America,
of the seeds is also critical. Defective beans are depending on the country.
sold at a low price, roasted, blended with better-
quality beans, and sold in the market as popular 1.2.3 Roasting, grinding, and packing
blends. The major concern here is the presence of
moulded and oxidized defects, and therefore the It is only during roasting that coffee acquires
its characteristic aroma, a consequence of a
potential presence of contaminants.
dramatic change in chemical composition of
After marketing, green coffee beans are ready
the beans. The two main systems used for heat
to undergo roasting. Optionally, they may be
transfer in coffee roasters are conduction and
decaffeinated, steam-treated, or stored before
convection. In roasters that use conduction, heat
roasting. If stored, this is another critical stage
is transferred by direct contact with a hot surface
where the growth of microorganisms is common.
and/or fire. In convection roasters heat is trans-
ferred by hot air circulation in the roasting
1.2.2 Decaffeination chamber, distributing heat evenly. Most modern
Decaffeination is traditionally performed roasters work this way, and some use both convec-
before roasting. For decaffeinated coffee, most tion and conduction. Convection roasters roast
countries require that the content of caffeine be faster than conduction roasters, and this will
reduced to less than 0.1% on a dry weight basis; have an influence on the chemical reactions that
however, decaffeinated coffees with up to 0.3% of occur during the process (IARC, 1991; Holman,
caffeine can be found on the market. In addition 2009; Soares et al., 2009; Fernandes, 2017). In
to extracting caffeine, other substances are also addition to the different types of roasters, there
extracted during the process including aroma are several variables that can be applied to the
precursors and bioactive compounds (Farah process such as time and temperature control.
et al., 2006; Toci et al., 2006). As a result, decaf- During the roasting process the beans
feinated products can taste very different from increase in volume, develop a brittle structure,
regular coffee. and acquire a light to dark brown colour, while
Different solvents and adsorbent substances their composition changes dramatically as a
can be used for decaffeination, among which consequence of pyrolysis, caramelization, and
dichloromethane (see IARC, 1986, 1987, 2017), Maillard reactions. The roasting intensity will
ethyl acetate (see IARC, 1979, 1987), edible vary for different cultures and types of coffee.
fats and oils, supercritical carbon dioxide, and Roasting chamber temperatures generally vary
acid-activated carbon (used in extract decaffein- over 190–270  °C, although the use of tempera-
ation) are well known (IARC, 1991). The selec- tures of 210–230 °C is more common in industry.
tivity of solvents and adsorbent materials varies, After cooling, which can be accelerated by
along with the sensory result. Processes that quenching with vapourized water or a cool air
do not employ an organic solvent are known as stream, residual carbon dioxide trapped in the
“water decaffeination”. Currently, the industry bean is slowly released over a period of days
in Europe and in the USA uses mostly water or up to 48  hours after grinding (IARC, 1991;
and supercritical carbon dioxide; the latter is Farah, 2012).

42
Drinking coffee

Coffee can be sold pre-ground and pack- (a) Decoction
aged after short degassing (2–4 hours) or under (i) Boiled coffee
an initial slight vacuum in flexible bags (Viani,
1986). Unlike green beans, roasted coffee spoils To prepare a boiled brew (most commonly
relatively quickly if unprotected from oxygen and consumed in northern Scandinavian countries),
moisture; at ambient temperature, whole beans boiling water is poured onto coarsely ground
stale 4–6 weeks and ground coffee 2 weeks after roasted Arabica coffee and the decoction is boiled
roasting (Toci et al., 2013). for up to 10 minutes. The decoction may also be
allowed to sit without boiling. The brew is made
at about 5% (weight/volume) [50 g coffee grounds
1.2.4 Brewing techniques in 1 L of water]. The ground coffee settles at the
Brewing can be defined as the preparation bottom and the brew is consumed. Cup volume
of a beverage by “mixing, steeping, soaking, or is about 150  mL (IARC, 1991; van Dusseldorp
boiling a solid in water” (Thesaurus, 2016). Coffee et al., 1991; [expert knowledge of the Working
brewing is the process by which coffee soluble Group]).
solids are extracted by water. Brewing techniques (ii) “Turkish” coffee
encompass a wide range of procedures used in
different parts of the world, which are based on For coffee consumed in Greece, Turkey, parts
the types of coffee and roasting degrees tradition- of the Balkans, and parts of the Middle East,
ally used. Local cultural practices and the cup very finely ground coffee is brewed with sugar
size used, associated with the variety of prepa- in a copper pot (ibrik) at about 8% w/v by gentle
ration methods, result in large differences in the boiling. The ingredients are heated until a large
chemical composition of the brew and a range of bubble or foam is formed in the centre of the
individual consumption patterns. Globalization ibrick. The heat is interrupted and the process
has reduced cultural distances however, and a can be repeated up to three times. Cup volume is
diversity of techniques and methods is available generally 60 mL (IARC, 1991).
both for home preparation and in coffeehouses. (iii) Kopi tubruk
An overview of the variety of coffee preparation Another variation of boiled coffee is kopi
methods is provided in Hatzold (2012). tubruk, also called mud coffee. This is a common
In addition to the type of coffee and roasting brewing method brought to Indonesia by Middle
degree, the particle size and the ratio of ground Eastern traders. The concentration is similar to
coffee to water vary considerably with tech- “Turkish” coffee, but a medium to coarse grind is
niques. Generally speaking, water temperature used instead of a fine grind. Coffee and sugar are
may vary from about 7–10 °C in cold brewing to placed in a cup or mug, boiling water is added,
100 °C in boiling methods (see the monograph and the coffee is “cooked” until the grounds
on Very Hot Beverages in the present volume for settle at the bottom. A variation is to heat water,
more information). After the extraction of coffee coffee grounds, and sugar together and let them
components, some techniques use filter paper to boil until the grounds settle. Cup/mug volume
separate grounds from brew while others use a varies over 150–190 mL [expert knowledge of the
strainer, plunger, or no device at all. If no filter Working Group].
paper is used, the coffee will contain the diter-
penes cafestol and kahweol (Urgert et al., 1995).
The most common brewing techniques are
described below.

43
IARC MONOGRAPHS – 116

(b) Infusion roasting degree becomes lighter. Light to medium
In this technique roasted coarse coffee roasts predominate in the USA, medium to dark
grounds are infused, usually followed by the use roasts prevail in South America, and dark roasts
of a device to separate the grounds from the brew. are favoured in France and Italy. Cup size varies
over 40–150  mL. Automatic coffee makers are
(i) Plunger pot available worldwide and have also been widely
In this system, also called a “French press” adopted by the food-services industry (IARC,
or piston system, hot water is a poured over 1991; [expert knowledge of the Working Group]).
coarsely ground coffee with a concentration of
(ii) Vapourization under pressure (moka pot)
about 4–10% (w/v), and a metal strainer is pushed
down the coffee pot to separate the grounds from In this technique water is heated to just
the brew after infusion. This system is used in above boiling point and forced by slight excess
Australia, Europe, and North America. A vari- pressure through coarse medium-/dark-roasted
ation of this system uses a paper filter instead coffee. Continuous recirculation over the coffee
of a metal plunger. Cup/mug size varies over grounds occurs until the desired brew strength is
150–190 mL (IARC, 1991) [expert knowledge of reached. Ground coffee concentrations normally
the Working Group]. range over 8–12%. This method is traditional in
Italian and Spanish households, and is becoming
(ii) Cold brewing popular all over the world [expert knowledge of
In this system, very coarsely ground coffee the Working Group].
is placed in a receptacle with a lid. Cold to
(iii) Vapourization under pressure (espresso)
room-temperature water is poured over the
ground coffee (about 12.5% w/v). The mixture This method, which originated in Italy and
is well stirred and the jar is covered and left is now popular wordwide, allows rapid extrac-
for 12–24  hours. When brewed, the mixture is tion. In espresso machines, water at 92–95 °C is
strained to remove solid residue. The resulting driven through a medium-/dark-roasted ground
brew can be served cold or mixed with boiling coffee packed bed by a pressure pump (8–12 bar)
water to serve hot. The main sensory character- to extract soluble material over a period of
istics of this brew are low acidity, gentleness, and 15–35 seconds. About 5–8 g of roasted coffee is
sweetness [expert knowledge of the Working used for each 25–60 mL cup. The extraction yield
Group]. of coffee soluble solids from the roasted coffee
is 18–26% with a soluble solids concentration in
(c) Percolation the cup of 20–60 g/L brew; 70–85% caffeine is
(i) Filtration recovered (Illy & Viani, 2005; Petracco, 2005;
Corrochano et al., 2015). Automated methods
Filtered coffee is one of the most common for coffee preparation have gained popularity
brewing methods around the world; it is made during the last decades, including fully auto-
by percolating pre-boiling water (95–98  °C) matic espresso-type machines that use coffee
through medium-ground roasted coffee in a pods or capsules (Gloess et al., 2013).
filter (usually paper but may be metal, nylon,
or ceramic) set in a funnel. The brew drips into
1.2.5 Instant coffee production
a warmed pot within about 2–5  minutes. The
strength of the brew will vary with cultural Instant coffee is a dried water extract of
habits, which includes roasting degree and coffee roasted and ground coffee which readily dissolves
to water ratio of 7–14% w/v, which increases as the in both cold and hot water, eliminating the need

44
Drinking coffee

for brewing equipment. The unit operations be similar to those applied to instant coffee prod-
performed during the manufacture of instant uction. Decaffeinated hydroalcoholic extracts
coffee are the same as for roasted coffee, followed and alternative technologies are also available.
by extraction, concentration, and drying of the Coffee silverskin is another byproduct of coffee
extract (to a maximum of 5% moisture by spray- production; its use for human consumption can
drying or freeze-drying), followed by agglom- be an alternative to its environmental disposal.
eration, aromatization, and packaging of the The high fibre and antioxidant compound
powder. Packaging is performed under vacuum content means that use of coffee silverskin (as
or in an inert atmosphere in jars or flexible bags, well as its extract) as a supplement for different
and the packaged product can be stable for more purported purposes such as weight control or for
than two years if unopened (Viani, 1986; IARC, antioxidant intake has been proposed (Borrelli
1991). et al., 2004; Narita & Inouye, 2012).

1.2.6 Other beverages containing coffee 1.3 Exposure assessment and
Coffee may also be sold as a “convenience” biological markers
product in a ready-to-drink form in cans or
aseptic carton packaging, typically premixed This section reviews the methods used to
with milk or other ingredients (Waizenegger assess coffee consumption and exposure to coffee
et al., 2011). components. Food consumption data, which are
typically obtained by questionnaire, provide
estimates of external exposures to coffee and
1.2.7 Other uses of coffee and coffee
related substances, while biological markers may
byproducts
be used to assess internal exposures.
Numerous other products containing coffee
are available on the market. Coffee and all forms 1.3.1 Questionnaires
of coffee extracts or instant coffee may be used
as an ingredient in various foods, such as the Dietary assessment methodologies are
flavouring of chocolate or in various bakery under continuous development in an attempt to
products. Infusions may be prepared with coffee improve the validity of dietary exposure data,
byproducts from post-harvesting processing (dry while also profiting from rapid evolution in
cherry pulp) or from coffee leaves. Coffee can innovative technologies such as mobile applica-
also be consumed as chocolate-coated roasted tions, scan- and sensor-based technologies, and
coffee beans. many other upcoming technologies (Illner et al.,
Non-traditional uses of coffee or coffee 2012). A comprehensive review of dietary assess-
extracts include the use in food supplements. ment methodologies and technologies used in
The US Dietary Supplements Label Database, epidemiological studies is beyond the scope of
for example, lists more than 100 products this report, but can be found in Thompson &
that contain the word “coffee” in the product Subar (2013) and Slimani et al. (2015).
name (NLM, 2016). Specifically, green coffee (a) Concepts, design, and applications
extract and roasted and green blends have been
marketed for purported effects such as weight The majority of epidemiological studies inves-
loss and intake of antioxidants. The methods of tigating associations between coffee consump-
encapsulated green coffee extract production can tion and cancer risk have used a food frequency
questionnaire (FFQ) to assess individual usual

45
IARC MONOGRAPHS – 116

coffee intake and/or particular coffee compo- different types of FFQ can be distinguished
nents (e.g. caffeine). The food frequency approach depending on the portion size information
asks respondents to report their usual frequency required in the questionnaire. If no portion size
of consumption of specific food items from a list information is included then the questionnaire
covering a specific period of time. FFQs are typi- is called a qualitative or non-quantitative FFQ,
cally used in epidemiological studies to assess while a questionnaire including detailed portion
“usual” dietary intake in individuals for several size information (e.g. in grams or number of
reasons. First, FFQs are the only feasible approach units) is called a quantitative FFQ. Some ques-
in case–control studies where usual diet (often in tionnaires include several portion size categories
the distant past) must be ascertained retrospec- (e.g. ≤ 1 cup; 2–3 cups; ≥ 4 cups) which the subject
tively. Second, in large prospective cohort studies can choose from, and is called a semi-quantitative
FFQs are often the instrument of choice because FFQ. Researchers have used methods to improve
of their time- and cost-efficient characteristics, assessment of portion size (e.g. picture booklets
including low investigator burden and cost. The to estimate cup size) in the studies included in
FFQ can be distributed by mail or online to a this report (see Tables 1.1 and 1.2).
large number of participants, can be self-ad- Some FFQs also include extra questions
ministered, may be optically scanned, computer regarding food preparation methods (e.g.
assisted, or web-based, and is often pre-coded brewing method for coffee), and identification
to facilitate data handling. Third, the FFQ has of the type (e.g. caffeinated versus decaffeinated)
the advantage that it does not affect the respond- and brand of certain types of foods.
ent’s eating behaviour and that usual individual FFQs are used widely in case–control and
intake is being requested (over a long timeframe), cohort studies to assess associations between
avoiding the need for repeated measurements. dietary intake and disease risk but, importantly,
Nevertheless, the completion of an FFQ may they are generally used for ranking subjects
be a challenge for respondents as usual consump- according to food or nutrient intake rather than
tions, and particularly portion sizes, are difficult for estimating absolute levels of intake (Beaton,
to estimate precisely. The ability to quantify total 1994; Kushi, 1994).
dietary intake depends on the number of food The FFQs used in the epidemiological studies
items listed in the FFQ and on the level of detail included in this report were developed with a
collected within the questionnaire, whether or broader aim than only coffee assessment; details
not portion sizes are included, what timeframe regarding the type of coffee and/or preparation
of intake or reference period is used, and, for method were therefore often lacking. This limit-
caffeine intake specifically, the differentiation ation should be considered when comparing
between caffeinated and decaffeinated coffee in results from different regions/countries as quan-
the food list (Block et al., 1986; Rimm et al., 1992). tities of coffee powder/beans used to brew a
FFQs generally include 50–150 (mostly coffee may differ between countries and cultures,
generic) food items, with the number of frequency although this information is lacking in most
categories varying according to the study objec- studies.
tives and designs. The appropriateness of the food
list is crucial as the full variability of an individ- (b) Validation and calibration
ual’s diet, which includes many foods and mixed Assessments of the relative validity of a FFQ
dishes, cannot be captured by a finite food list. provide information about how well the instru-
Whether portion size information is also ment is measuring what it is intended to measure.
required depends on the study aims. Three This is evaluated by comparing dietary intake

46
Table 1.1 Summary of methods used in cohort studies investigating the relationship between coffee consumption and
cancer risk

Study, country, Information-collection method Period of Respondents Distinction Exposure metrics
reference information between
collection caffeinated/
decaffeinated
Melbourne Collaborative General: FFQ At baseline 24 500 women and 17 000 No 1–3 cups/mo,
Cohort Study (MCCS), Coffee-specific questions: frequency of men aged 40–69 yr 1 cup/wk,
Australia coffee consumption, consumption of 2–4 cups/wk,
(Ireland et al., 1994) milk with coffee 5–6 cups/wk,
1 cup/day,
2–3 cups/day,
4–5 cups/day,
> 6 cups/day
The Singapore Chinese General: 165-item FFQ, in conjunction At baseline 63 257 Chinese men and No Never or hardly ever,
Health Study, Singapore with the Singapore Food Composition women aged 45–74 yr 1–3 times/mo,
(Ainslie-Waldman et al., Database for the ascertainment of 96 once/wk,
2014) items including caffeine 2–3 times/wk,
Coffee-specific questions: frequency of 4–6 times/wk,
caffeinated coffee consumption once/day,
Questionnaire not available 2–3 times/day,
4–5 times/day,
≥ 6 times/day
Life Span Study, Japan General: 22-item FFQ At baseline 40 349 Japanese men and No 1 cup/wk,
(Sauvaget et al., 2002) Coffee-specific questions: frequency of women 2–4 cups/wk,
coffee consumption Almost daily,
Questionnaire not available Do not eat/drink

Drinking coffee
47
48

IARC MONOGRAPHS – 116
Table 1.1 (continued)

Study, country, Information-collection method Period of Respondents Distinction Exposure metrics
reference information between
collection caffeinated/
decaffeinated
Japan Public Health General: self-administered At baseline 140 420 male and female No Studies using baseline
Center-based questionnaire including questions on Japanese subjects aged questionnaire:
Prospective (JPHC) beverage consumption 40–69 yr at baseline Almost never
study, Japan Coffee-specific questions: circle the 1–2 days/wk,
(Makiuchi et al., 2016) frequency of your average consumption 3–4 days/wk,
of coffee; how many teaspoons of sugar 1–2 cups/day,
do you use per cup of coffee? 3–4 cups/day,
> 5 cups/day
Study using 5-year follow-up
questionnaire:
0 cups/wk,
1–2 cups/wk,
3–4 cups/wk,
5–6 cups/wk,
1 cup/day
2–3 cups/day,
4–6 cups/day,
7–9 cups/day,
10 cups/day
Miyagi Cohort, Japan General: self-administered At baseline 25 279 men and 26 642 No Never,
(Naganuma et al., 2008) questionnaire with 36 food items and 4 women 40–64 yr at < 1 cup/day
beverages including coffee baseline 1–2 cups/day,
Coffee-specific questions: frequency of 3–4 cups/day,
coffee consumption 5 cups/day
Questionnaire not available
Japan Collaborative General: self-administered At baseline 110 792 Japanese men and No Seldom or never,
Cohort (JACC) study, questionnaire including questions on women aged 40–79 yr at 1–2 cups/mo,
Japan beverage consumption baseline 1–4 cups/wk,
(Yamada et al., 2014) Coffee-specific questions: frequency of 1 cup/day,
coffee consumption; addition of sugar 2–3 cups/day,
and milk in coffee > 4 cups/day
Questionnaire not available
Takayama city cohort, General: 169- item FFQ At baseline 13 392 men and 15 695 No Never,
Japan Coffee-specific questions: frequency of women aged ≥ 35 yr > 1 cup mo to 4–6 cups/wk,
(Oba et al., 2008) coffee consumption > 1 cup/day
Questionnaire not available
Table 1.1 (continued)

Study, country, Information-collection method Period of Respondents Distinction Exposure metrics
reference information between
collection caffeinated/
decaffeinated
National Breast General: 86-item FFQ At baseline 89 835 women, aged 40–59 No None,
Screening Study (NBSS), Coffee-specific questions: frequency of yr 0–1 cup/day,
Canada coffee consumption 2–3 cups/day,
(Silvera et al., 2007) Questionnaire not available ≥ 4 cups/day
Health Professional General: 130-item FFQ At baseline in 47 911 male health Yes None,
Follow-up study, United Coffee-specific questions: how 1986 and every professionals aged 40–75 < 1 cup/day,
States frequently did you drink a cup of 4 yr thereafter yr at baseline 1–3 cups/day,
(Wilson et al., 2011) caffeinated coffee? 4–5 cups/day,
How frequently did you drink a cup of 6 cups/day
decaffeinated coffee?
Iowa Womens’ Health General: 127-item FFQ At baseline 98 826 women in Iowa Yes < 1/mo,
Study, Iowa Coffee-specific questions: number of aged 55–69 yr at baseline 1–3 cups/mo,
(Lueth et al., 2008) cups per day for normal or decaffeinated 1 cup/wk,
coffee 2–4 cups/wk,
Questionnaire not available 5–6 cups/wk,
1 cup/day,
2–3 cups/day,
4–5 cups/day,
6 cups/day
NIH-AARP Diet and General: 124-item FFQ At baseline 3.5 million men and Yes 10 frequency categories
Health Study, USA Coffee-specific questions: how women from American ranging from never to
(Dubrow et al., 2012) many cups of coffee caffeinated or Association of Retired > 6 times/day
decaffeinated did you drink? Persons
When you drank coffee, mark whether
you usually drank caffeine-free or
caffeine-containing types (didn’t drink
this beverage, more than half the time I
drank caffeine-free, more than half the
time I drank caffeine containing)
Black Women’s Health General: 68-item modified version of the At baseline and 59 000 African-American Yes Nine frequency categories
study, United States National Cancer Institute Block FFQ, after 6 yr women ranging from never or

Drinking coffee
(Boggs et al., 2010) 85-item version in 2001 1 time/mo to 6 times/day
Coffee-specific questions: how often did
you drink coffee with caffeine?
How often did you drink decaffeinated
coffee?
Milk or cream in coffee
49
50

IARC MONOGRAPHS – 116
Table 1.1 (continued)

Study, country, Information-collection method Period of Respondents Distinction Exposure metrics
reference information between
collection caffeinated/
decaffeinated
Nurses’ Health Study General: NHS1: 61-item semi- At baseline and 121 700 female nurses Yes 0–1 cups/day,
and Nurses’ Health quantitative FFQ at baseline, after 130- every 3 yr 30–55 yr old at baseline 2 cups/day,
Study 2, USA item FFQ NHS2: 131-FFQ (NHS1), 3 cups/day,
(Holick et al., 2010) Coffee-specific questions: how often did 116 686 female nurses 4 cups/day,
you use coffee not decaffeinated (cups) in 25–42 yr old at baseline 5 cups/day
the precedent year? (NHS2)
Prostate, Lung General: 77-item FFQ, NIH Health At baseline 154 901 men and women, Yes None,
Colorectal and Diet History Questionnaire (DHQ) in aged 55–74 yr at baseline < 1 cup/day,
Ovarian(PLCO) cohort, addition for coffee intake 1 cup/day,
USA Coffee-specific questions: how 2–3 cups/day,
(Dominianni et al., 2013) many cups of coffee caffeinated or > 4 cups/day
decaffeinated did you drink?
How often was the coffee you drank
decaffeinated?
How often did you add sugar or honey to
your coffee?
Each time sugar or honey was added
to your coffee how much was usually
added?
How often did you add artificial
sweetener to your coffee?
What kind of artificial sweetener do you
usually use?
How often was non-dairy creamer added
to your coffee?
Women’s Health General: questionnaires on demographic At baseline and 93 676 women aged 50–79 Yes 0 or 1 cup/day,
Initiative, USA characteristics, medical history, family after 3 yr yr at baseline 1 cup/day,
(Giri et al., 2011) history, reproductive history, lifestyle/ 2–3 cups/day,
behavioural factors, and quality of life 4 cups/day
Coffee-specific questions: do you usually 4–5 cups/day
drink coffee each day? > 6 cups/day
Number of cups of coffee
7th Day Adventists, USA General: Lifestyle questionnaire, 51-item At baseline 34 192 members of 7th No < 1 cup
(Phillips & Snowdon, FFQ Day Adventist church in 1–2 cups/day
1983; Butler et al., 2008) Coffee-specific questions: frequency of California, > 25 yr old; 3–4 cups/ day
coffee consumption non-Hispanic whites 5+ cups/day
Table 1.1 (continued)

Study, country, Information-collection method Period of Respondents Distinction Exposure metrics
reference information between
collection caffeinated/
decaffeinated
Lutheran Brotherhood General: FFQ At baseline 17 818 male, white policy No None
Insurance Study, USA Coffee-specific questions: frequency of holders, aged ≥ 35 yr, of < 1 cup/day
(Murray et al., 1981) coffee consumption the Lutheran Brotherhood 1–2 cups/day
Questionnaire not available Insurance Society < 3 cups/day
3–4 cups/day,
5–6 cups/day,
≥ 7 cups/day
Leisure World Cohort, General: FFQ At baseline 13 978 residents of Leisure Yes None
USA Coffee-specific questions: frequency of World, California; men < 1 cup/day
(Paganini-Hill et al., coffee consumption and women. The mean of 1 cup/day,
2007) Questionnaire not available age at entry was 75.0 yr for 2–3 cups/day,
men and 73.8 yr for women ≥ 4 cups/day
Kaiser Permanente General: lifestyle questionnaire At baseline 182 357 Kaiser Foundation No ≤ 6 cups/day,
Medical Care Program Coffee-specific questions: frequency of Health Plan members > 6 cups/day
Study, USA coffee consumption
(Efird et al., 2004) Questionnaire not available
Cancer Prevention Study General: 66-item FFQ At baseline 968 432; men and women Yes < 1 cup/day,
II, USA Coffee-specific questions: frequency of (average age 57 yr) 1–2 cups/day,
(Hildebrand et al., 2013) coffee consumption (currently and in the 3–4 cups/day,
previous year) > 4 cups/day
Lutheran Brotherhood General: FFQ At baseline 17 633 male white policy No < 3 cups/day
Insurance Study, USA Coffee-specific questions: frequency of holders, aged ≥ 35 yr, of 3–4 cups/day,
(Murray et al., 1981) coffee consumption the Lutheran Brotherhood 5–6 cups/day,
Questionnaire not available Insurance Society ≥ 7 cups/day
The Glostrup Population General: standardized questionnaire for At baseline 5207 Danish women No 0–2 cups/day,
Studies, Denmark coffee consumption 3–6 cups/day,
(Sjøl et al., 1991) Coffee-specific questions: number of > 7 cups/day
cups of coffee
Questionnaire not available
The ATBC study, General: FFQ in conjunction with At baseline 29 133 Finnish male No Participants indicated the

Drinking coffee
Finland a validated comprehensive nutrient smokers average number of cups of
(Lai et al., 2013) database coffee consumed per day or
Coffee-specific questions: how many per week in the previous year
cups did you drink per week or per day?
Sugar, whipping cream, coffee cream,
light cream, milk
51
52

IARC MONOGRAPHS – 116
Table 1.1 (continued)

Study, country, Information-collection method Period of Respondents Distinction Exposure metrics
reference information between
collection caffeinated/
decaffeinated
Hu et al. (2008), Finland General: lifestyle questionnaire At baseline 62 015 Finish participants No 0–1cups/day,
Coffee-specific questions: frequency of (seven for seven surveys, aged 2–3 cups/day,
coffee consumption independent 25–74 yr 4–5 cups/day,
Questionnaire not available cross-sectional 6–7 cups/day,
population ≥ 8 cups/day
surveys
were carried
out in six
geographical
areas of
Finland in
1972, 1977,
1982, 1987,
1992, 1997 and
2002)
Bidel et al. (2013), General: lifestyle questionnaire At baseline 60 041 Finnish men and No 1–2 cups/day,
Finland Coffee-specific questions: frequency of women aged 26–74 yr 3–4 cups/day,
coffee consumption 5–6 cups/day,
Questionnaire not available 7–9 cups/day,
≥ 10 cups/day
Norwegian National General: semi-quantitative At baseline 25 708 men and 25 049 No < 2 cups/day,
Health Screening Service questionnaire with 54 questions women aged 16–56 yr 3–4 cups/day,
for CVD, Norway obtained information on general meal at baseline attending 5–6 cups/day,
(Veierød et al., 1997) pattern, amounts, frequencies and types a Norwegian health > 7 cups/day
of specified foods and beverages screening in 1977–1983
Coffee-specific questions: number of
cups of coffee
Questionnaire not available
Norwegian Women and General: questionnaire on health, At baseline 97 926 women resident in No ≤ 1cup/day, 2–3 cups/day,
Cancer (NOWAC) study, lifestyle, and reproductive factors Norway, aged 30–70 yr at 4–7 cups/day,
Norway Coffee-specific questions: how many baseline ≥ 8 cups/day
(Gavrilyuk et al., 2014) cups of each kind of coffee/tea do you Almost never
usually drink? (filtered, boiled, instant)
Do you use sugar, milk, or cream in
coffee?
Table 1.1 (continued)

Study, country, Information-collection method Period of Respondents Distinction Exposure metrics
reference information between
collection caffeinated/
decaffeinated
Swedish Women’s General: self-administered FFQ At baseline 48 249 women residing in No No
Lifestyle and Health Coffee-specific questions: how many the Uppsala Health Care
cohort study, Sweden cups per day or per week during the Region in Sweden between
(Weiderpass et al., 2014) preceding year 1991 and 1992
Swedish Mammography General: 67-item FFQ At baseline and 60 634 women born No 1 cup/day,
Cohort, Sweden Coffee-specific questions: how often do after 7 yr between 1914–1948 living 2–3 cups/day,
(Friberg et al., 2009) you drink coffee? in Uppsala County of 4 cups/day
Central Sweden, women
born during 1917–1948
living in Västmanland
county
Västerbotten General: 84-item VIP FFQ (1992–1996), At baseline 64 603 residents of the No Never,
Intervention Project 64 items VIP FFQ (1997–2007) county of Västerbotten A few times/yr,
(VIP), Sweden Coffee-specific questions: two questions turning 40, 50, and 60 yr 1–3 times/mo,
(Norberg et al., 2010) on coffee, one for filtered and one for of age 1 time/wk,
boiled coffee 2–3 times/wk,
Questionnaire not available 4–6 times/wk,
1 occasion/day,
2–3 occasions/day,
4 occasions/day
Swedish Twin Registry General: lifestyle questionnaire At baseline 1884 men No 0–2 cups/day,
Study, Sweden Coffee-specific questions: frequency of and women 3–6 cups/day,
(Isaksson et al., 2002) coffee consumption recruited in 1961, ≥ 7 cups/day
Questionnaire not available aged 36–75 yr
Discacciati et al. (2013), General: 96-item FFQ At baseline 48 645 men aged 45–79 yr No None,
Sweden Coffee-specific questions: frequency of residing in Västmanland < 1 cup/day,
coffee consumption and Örebro counties in 1–3 cups/day,
Questionnaire not available central Sweden 4–5 cups/day,
≥ 6 cups/day
Netherlands Cohort General: FFQ including question about At baseline 120 852 men and women No 0 – < 1 cups/day

Drinking coffee
Study, coffee intake (yes/no, how many cups aged 55–69 yr at baseline 1 – < 3 cups/day
Netherlands per day) 3 to < 5 cups/day
(Steevens et al., 2007) Coffee-specific questions: number of > 5 cups/day
cups of coffee of coffee
Questionnaire not available
53
54

IARC MONOGRAPHS – 116
Table 1.1 (continued)

Study, country, Information-collection method Period of Respondents Distinction Exposure metrics
reference information between
collection caffeinated/
decaffeinated
Million Women Study, General: questionnaire on health, At baseline and 1.3 million middle-aged No < 1 cup/day, 1–2 cups/day,
UK lifestyle, and reproductive factors approximately women 3–4 cups/day,
(Yang et al., 2015) Coffee-specific questions: after 3 yr ≥ 5 cups/day
How many teaspoons of sugar do you
add to coffee?
Do you add milk to your coffee?
Supplementation en General: 24 hours dietary recall At baseline and 7876 women aged 35–60 yr No No
Vitamines et Mineraux Coffee-specific questions: indication of every 2 mo and 5141 men aged 45–60
Antioxydants Study the portion size yr
(SUVIMAX), France
(Mennen et al., 2007;
Hercberg et al., 2004)
Fagherazzi et al. (2011), General: 208-item FFQ At baseline and 67 703 women Yes ≤ 1 cup/day,
France Coffee-specific questions: frequency of after 3 yr 1–3 cups/day,
coffee consumption > 3 cups/day
Questionnaire not available
EPIC, 10 European General: Country-specific At baseline 521 448 men and women Yes, except Different exposure metrics
countries questionnaires: self-administered semi- for Denmark depending on the country
(Bhoo-Pathy et al., 2015) quantitative FFQ (± 260 food items), and France
dietary history questionnaires (> 600
food items), semi-quantitative food-
frequency questionnaires combined with
a food record
Coffee-specific questions: number of
cups of coffee
Questionnaires not available
Multiethnic Cohort General: lifestyle questionnaire At baseline 215 000 men and women Yes Never or hardly never,
Study of Diet and Cancer including diet history primarily of African- Once a month,
(MEC) Coffee-specific questions: frequency of American, Japanese, 2–3 times/mo,
(Kolonel et al., 2000) coffee consumption Latino, native Hawaiian Once a week,
Do you add any of the following to and Caucasian origin 2–3 times/wk,
coffee: cream or half and half; milk; 4–6 times/wk
non-diary cream; sugar or honey; sugar Once a day,
substitute 2–3 times/day,
> 4 times/day
mo, month(s); wk, week(s); yr, year(s)
Drinking coffee

Table 1.2 Summary description of methods used in cohort studies investigating the relationship
between coffee consumption and cancer risk

Study, country, reference Portion size Specific Method validated for coffee
components consumption
measured
Melbourne Collaborative Cohort Study No No No
(MCCS), Australia
(Ireland et al., 1994)
Singapore Chinese Health Study, Yes, three possible portion Daily caffeine No
Singapore sizes intake
(Ainslie-Waldman et al., 2014)
Life Span Study, Japan No No Yes, correlation with 24 h diary:
(Sauvaget et al., 2002) 0.51
Japan Public Health Center-based No No Yes, Spearman correlation
Prospective (JPHC) study, Japana coefficient with diet record data:
(Makiuchi et al., 2016) Baseline Q: 0.42 for men and
0.38 for women
5 yr follow-up Q: 0.75 in men
and 0.80 in women
Miyagi Cohort, Japan No No Yes, Spearman rank correlation
(Naganuma et al., 2008) with 3-day diet records: 0.70
Japan Collaborative Cohort (JACC) No No Yes, Spearman rank correlation
study, Japan 12-day dietary record: 0.81
(Yamada et al., 2014)
Takayama City Cohort, Japan No No No
(Oba et al., 2008)
National Breast Screening Study (NBSS), Yes No No
Canada
(Silvera et al., 2007)
Health Professional Follow-up study, No Daily caffeine Yes, correlation with two week-
USA intake long diet records: 0.93
(Wilson et al., 2011)
Iowa Womens’ Health Study, Iowa No Daily caffeine Yes, correlation between caffeine
(Lueth et al., 2008) intake intake estimates from dietary
recalls and FFQ: 0.95
NIH-AARP Diet and Health Study, USA No Daily caffeine No
(Dubrow et al., 2012) intake
Black Women’s Health study, USA Yes (small, medium, or Daily caffeine Yes
(Boggs et al., 2010) large) intake
Nurses’ Health Study and Nurses’ No Daily caffeine Yes, Pearson correlation with 1
Health Study 2, USA intake wk diet record: 0.93
(Holick et al., 2010)
Prostate, Lung Colorectal and Ovarian No Daily caffeine No
(PLCO) cohort, USA intake
(Dominianni et al., 2013)
Women’s Health Initiative, USA No No No
(Giri et al., 2011)
7th Day Adventists, USA No No No
(Phillips and Snowdon, 1983)
Leisure World Cohort, USA No No No
(Paganini-Hill et al., 2007)

55
IARC MONOGRAPHS – 116

Table 1.2 (continued)

Study, country, reference Portion size Specific Method validated for coffee
components consumption
measured
Kaiser Permanente Medical Care No No No
Program Study, USA
(Efird et al., 2004)
Cancer Prevention Study II, USA No No No
(Hildebrand et al., 2013)
Lutheran Brotherhood Insurance Study, No No No
USA
(Murray et al., 1981)
Glostrup Population Studies, Denmark No No No
(Sjøl et al., 1991)
ATBC study, Finlandb Yes, using a colour picture No Yes, correlation with diet
(Lai et al., 2013) booklet (four possible records: 0.72–0.79
portion sizes)
Hu et al. (2008), Bidel et al. (2013) No No Yes, Spearman correlation with
Finland food records: 0.89 in men, 0.85
in women
Norwegian National Health Screening No No No
Service for CVD, Norway
(Veierød et al., 1997)
Norwegian Women and Cancer No No Yes, Spearman correlation with
(NOWAC) study, Norway 24 h recall: 0.82
(Gavrilyuk et al., 2014)
Swedish Women’s Lifestyle and Health Yes (small, medium, or Daily caffeine Yes, Spearman correlation with
cohort study, Sweden large) intake weighted record: 0.60
(Weiderpass et al., 2014)
Swedish Mammography Cohort, No No Yes, Spearman correlation with
Sweden weighted record: 0.60
(Friberg et al., 2009)
Västerbotten Intervention Project (VIP) No No Yes, correlation with 24 h recall:
b, Sweden 0.72–0.84
(Norberg et al., 2010)
Swedish Twin Registry Study, Sweden No No No
(Isaksson et al., 2002)
Discacciati et al. (2013), Sweden No No Yes, Spearman correlation with
24 h recall: 0.71
Netherlands Cohort Study, No No Yes, validated against a 9-day
Netherlands diet record
(Steevens et al., 2007)
Million Women Study, UK No No No
(Yang et al., 2015)
Supplementation en Vitamines Yes No Yes
et Mineraux Antioxydants Study
(SUVIMAX), France
(Mennen et al., 2007)
Fagherazzi et al. (2011), France Yes Daily caffeine No
intake

56
Drinking coffee

Table 1.2 (continued)

Study, country, reference Portion size Specific Method validated for coffee
components consumption
measured
EPIC, 10 European countries Yes, except for Denmark, No No
(Bhoo-Pathy et al., 2015) Italy, Norway, and Umeå
(Sweden)
Multiethnic Cohort Study of Diet and No Daily caffeine Yes, Spearman correlation with
Cancer (MEC), Hawai intake 24 h recall: 0.72
(Kolonel et al., 2000)
All studies in the table used retrospective dietary assessment methods to assess coffee exposure
Temperature at which the coffee was consumed was not assessed in any of these studies
a In the Japan Public Health Center-based Prospective (JPHC) study, canned coffee was also assessed as specific coffee type

b In the ATBC study, Norwegian Women and Cancer (NOWAC) study and the Västerbotten Intervention Project (VIP), the preparation

method was specified as filtered or boiled
h, hour(s); wk, week(s); yr, year(s)

assessed using an FFQ to intake assessed in the In the European Prospective Investigation
same individuals using a reference method that into Cancer and Nutrition (EPIC), dietary intakes
is deemed to be superior, but that may be prohib- over the previous year were assessed at enrolment
itive to use in large epidemiological studies due through validated study centre specific question-
to participant burden or cost. naires which also enquired about coffee intake.
To illustrate, in the National Institutes of This method was reported to yield very good
Health–American Association of Retired Persons reliability of coffee consumption compared with
(NIH-AARP) cohort study, the FFQ used was repeated 24-hour recalls (r = 0.70) (Aleksandrova
validated against two non-consecutive 24-hour et al., 2015).
dietary recalls (Thompson et al., 2008). In a vali- Calibration studies are used to calibrate a
dation set of participants, Spearman correlations FFQ to a reference method using a regression
between 24-hour dietary recalls and the food model (e.g. an interviewer-led diet history or
frequency questionnaire were 0.80 for coffee, multiple 24-hour recalls). For example, the EPIC
0.64 for caffeinated coffee, and 0.48 for decaf- study used a computerized 24-hour diet recall
feinated coffee (Thompson et al., 2008; Sinha method to calibrate dietary measurements across
et al., 2012). Further, data obtained through a countries and to correct for systematic over- or
semi-quantitative FFQ in the Health Professionals underestimation of dietary intakes (Slimani
Follow-Up Study (HPFS) and the FFQs used in et al., 2002).
the different waves of the Nurses’ Health Study Each epidemiological study included in
have been tested for reproducibility in a subgroup Section 2 of this monograph has been examined
of participants who completed two FFQs 1 year to determine whether the FFQ used to assess
apart and two 1-week diet records 6  months coffee exposure has been validated.
apart during the intervening year. Pearson
correlations between the average coffee intake, (c) Cohort studies
assessed by two 1-week diet records completed A major strength of cohort studies in nutri-
6  months apart, and the baseline FFQ was tional epidemiology is the ability to demonstrate
0.93 in the HPFS and 0.78 in the Nurses’ Health a temporal relationship between dietary expo-
Study (Holick et al., 2010). sure and cancer risk, as all dietary assessments

57
IARC MONOGRAPHS – 116

are completed before diagnosis. This mitigates extensively validated against 24-hour dietary
concerns related to recall bias and reverse causa- recalls, showing good validity of coffee estimates
tion. However, a limitation of many cohort (Thompson et al., 2008). Limitations were the
studies is that exposures are often measured only lack of an assessment of preparation methods,
once, usually during enrolment, whereas cancer which likely affect the concentration of different
cases develop over a long period of time. In the compounds in coffee, and the fact that decaffein-
case of coffee consumption, however, there is a ated coffee drinkers were also defined on the basis
high correlation between successive measure- of drinking either beverage more than half of the
ments taken over time. time, which could have led to misclassification.]
In the Nurses’ Health Study (Willett et al., In the EPIC cohort study (Riboli et al., 2002;
1985; 1988), data were obtained at baseline in Aleksandrova et al., 2015), dietary intake over
1980 through a validated, self-administered, the 12 months before enrolment was measured
61-item, semi-quantitative FFQ which was later by country-specific validated dietary question-
expanded and applied every 4  years thereafter naires (88–266 food items, depending on the
(Michels et al., 2005). Essentially the same vali- country), self-administered in most countries. A
dated, self-administered, semi-quantitative FFQ second dietary measurement was taken from an
questionnaire with 131 items was used in the 8% random sample of the cohort (36 900 partic-
HPFS (Rimm et al., 1992). For each item, partic- ipants) using a computerized 24-hour diet recall
ipants were asked to report their average use of method to calibrate dietary measurements across
each food and beverage over the preceding year. countries and to correct for systematic over- or
Consumption of caffeinated and decaffeinated underestimation of dietary intakes. [A major
coffee was measured in cups per day. Most anal- strength of this study was the large variability
yses from the Nurses’ Health Study and HPFS of in dietary intake across populations and the use
coffee intake have used the cumulative average of a computerized 24-hour diet recall method
intake of coffee over time, incorporating informa- to calibrate dietary measurements across coun-
tion from the repeated questionnaires (Michels tries. A limitation was the use of different dietary
et al., 2005). [A strength of these studies was that questionnaires in each participating country,
the FFQs used were extensively validated and requiring post-harmonization and calibration of
tested for reproducibility, demonstrating good the dietary data.]
validity for coffee intake estimations. A limita- In the Alpha-Tocopherol, Beta-Carotene
tion was the lack of an assessment of preparation Cancer Prevention (ATBC) study (Lai et al., 2014),
methods, which likely affects the concentration researchers used a self-administered, modified
of different compounds in coffee.] dietary history method to capture usual dietary
Sinha et al. (2012) used data from the intake 12 months before recruitment. The
NIH-AARP Diet and Health Study in which dietary history method included 276 food items
the National Cancer Institute’s Diet History and a picture booklet of photographs illustrating
Questionnaire, a 124-item FFQ with informa- different portion sizes (Pietinen et al., 1988). To
tion on the frequency of intake and portion sizes assess coffee consumption, participants were
over the past year, was used. Coffee intake was also asked to indicate their typical cup size.
assessed from 0 to 6 cups/day and participants [A strength of this study was that the assess-
were dichotomized according to whether they ment of coffee intake was shown to be valid after
reported drinking caffeinated or decaffeinated comparison with food records. Another impor-
coffee more than half the time. [A strength tant advantage of this study was that information
of this study was that the FFQs used were on coffee preparation methods was collected,

58
Drinking coffee

allowing exploration of whether the associations in cups per day or week were also different
between coffee intake and disease risk differed by in two versions of the questionnaire. Post-
brewing method. A limitation was that no infor- harmonization of the dietary data was therefore
mation was collected on whether intake of coffee needed to create a common version of frequen-
was caffeinated or not, and coffee intake was only cies. Based on a 24-hour recall investigation in
assessed a single time.] the NOWAC cohort, a standard cup size of 2.1 dL
In the Iowa Women’s Health Study cohort, (7.1 oz) was assumed. [An advantage of this study
usual dietary intake during the previous year was was the broad range of exposures in the cohort
assessed by a 127-item, validated, semi-quanti- and information on several coffee brewing
tative, self-administered FFQ, virtually identical methods. Limitations were the lack of informa-
to that used in the 1984 survey of the Nurses’ tion about decaffeinated coffee and preparation
Health Study (Lueth et al., 2008). For decaffein- methods, which can also influence the level and
ated coffee and regular coffee, the specified daily properties of some coffee compounds.]
portion size was one cup (8 fluid ounces). The In the multiethnic, prospective, popu-
validity of the FFQ was evaluated by comparing lation-based Northern Manhattan Study
nutrient values from the FFQ to those from the (NOMAS), participants were administered a
average of five 24-hour dietary recall surveys for modified Block National Cancer Institute FFQ
44 study participants. Reliability was assessed by at baseline. Questions assessed the average
repeating the FFQ after 3–6 months. [This study consumption of decaffeinated and regular
had several strengths, including the assessment (caffeinated) coffee in units of medium cups
of reliability and accuracy of the FFQ in the according to nine different frequency categories
Iowa cohort and the possibility of allowing for (Gardener et al., 2013). [Despite the use of a vali-
multivariable adjustment due to the extensive dated and reliable Block FFQ, there were some
FFQ. A limitation was the lack of an assessment limitations to the coffee exposure data. The anal-
of preparation methods, which likely affects the ysis focuses on frequency of consumption and is
concentration of different compounds in coffee.] not standardized for cup size. Although the FFQ
In the Singapore Chinese Health Study, was designed to measure average consumption
cohort members completed an in-person inter- over the previous year, dietary information was
view that included a validated 165-item FFQ collected at one time (baseline) and information
that assessed coffee intake at nine predefined on duration of coffee consumption as well as
levels (Johnson et al., 2011). Decaffeinated coffee changes during follow-up was lacking. Further,
consumption was not assessed. [A strength of the questionnaire did not determine whether
this study was that the FFQ was shown to be individuals were drinking boiled unfiltered or
valid, while the fact that only a baseline assess- filtered coffee.]
ment of self-reported coffee intake was available In the Japan Collaborative Cohort Study
should be considered a limitation because of the (Yamada et al., 2014), information about coffee
potential for non-differential misclassification.] consumption and other lifestyle factors was
In the Norwegian Women and Cancer obtained using a self-administered question-
(NOWAC) study the questions assessing coffee naire. The question regarding coffee consump-
consumption varied according to the year of tion was previously assessed by a validation
enrolment; women were asked about either their study, which reported a strong agreement with
total coffee consumption or their consumption 12-day weighted dietary records. [A limitation
of filtered, boiled, and instant coffee (Gavrilyuk of this study was that only baseline data were
et al., 2014). The categories of coffee consumption collected; no details of coffee consumption,

59
IARC MONOGRAPHS – 116

such as the use of caffeinated or decaffeinated self-administered questionnaire (Green et al.,
coffee and the method of coffee preparation (e.g. 2014). The questions were adapted from the
filtered or boiled), were collected.] Arizona Tea Questionnaire, which was shown to
have high test–retest reliability and high relative
(d) Case–control studies validity relative to 4-day food records. Data were
A description of the case–control studies collected on the frequency of consumption of hot
included in this report is provided in Section 2. caffeinated coffee, hot decaffeinated coffee, and
Case–control studies investigating the associ- iced coffee. [A limitation of this study was that
ation between coffee intake and cancer risk are asking participants to recall their dietary intake
limited as they assess dietary intake after cancer 10 years before may have affected the quality of
has been diagnosed, which can lead to recall recall, leading to increased likelihood of expo-
bias if intakes of the distant or recent past are sure misclassification which could bias the risk
assessed. In addition, for some cancer types, estimates towards the null.]
notably cancers of the digestive tract, patients may
change their dietary intake with the potential to (e) Covariates of coffee consumption
bias assessment of the usual coffee intake in the The estimation of cancer risk associated with
past. As a result, investigators usually ask cases coffee intake may be influenced by other dietary
included in their studies to recall dietary intake and/or lifestyle factors that are correlated with
in a period before the diagnosis of cancer in an coffee consumption. However, few studies inves-
attempt to capture usual diet before diagnosis. tigated associations between coffee consump-
For these reasons, the approach used to assess tion and other lifestyle factors, which can vary
dietary intake in the distant past is often concep- depending on the population under study.
tualized differently from the typical FFQs used Freedman et al. (2012) investigated associa-
in cohort studies by including extra questions tions of coffee consumption with other dietary
to help the respondent remember details of past and lifestyle factors in the NIH-AARP Diet and
consumption. For example, in the Yale Study of Health Study in the USA. Coffee drinkers were
Skin Health in Young People, Ferrucci et al. (2014) more likely than non-drinkers of coffee to smoke
participants were first asked whether they drunk cigarettes and also consume more alcoholic
at least one cup of caffeinated coffee per week for drinks and more red meat; coffee drinkers also
at least 6 months. Those who responded affirma- tended to have lower levels of education, vigorous
tively were then asked the age at which they began physical activity, and intake of fruits and vegeta-
drinking caffeinated coffee at this frequency, as bles (Freedman et al., 2012).
well as whether they were currently drinking it In a Japanese study, Yamada et al. (2014)
at least weekly or, if not, the age at which they reported that subjects who consumed high quan-
had stopped. Thereafter, participants reported tities of coffee were also more likely to be smokers,
the average number of cups of coffee they drank alcohol drinkers, and to regularly eat beef or
per day and the number of years of consumption. pork, but were younger and better educated.
[Even though the possibility of recall bias was In the Singapore Chinese Health Study,
still a limitation in this case–control study, this Ainslie-Waldman et al. (2014) investigated
cognitive approach is considered an important differences in lifestyle and sociodemographic
strength in avoiding such bias.] factors by the amount of coffee consumption.
In the Western Australian Bowel Health Among both men and women, a higher level of
Study (WABOHS), data on coffee consump- coffee consumption was associated with a higher
tion 10  years previously were collected by prevalence of current smoking and alcohol

60
Drinking coffee

consumption, as well as lower proportions with to consider in the relationship between coffee
a higher education or a history of diabetes. Those consumption and cancer risk. Another limit-
who drank more coffee also consumed more total ation is the reporting of the exposure assess-
energy but less fruit and vegetables. However, ment, which is sometimes insufficiently detailed
men who drank more than 4 cups of coffee per to allow a critical and correct evaluation of the
day had lower levels of BMI compared to those results reported (Lachat et al., 2016). These limi-
consuming < 1 cup per day. tations in exposure assessment should be consid-
The effect of coffee consumption on the risk ered when interpreting the results reported in
of cancer may therefore be confounded by intake epidemiological studies.
of other foods and lifestyle factors, in particular
smoking status, but also drinking, BMI, meat (g) Biological markers
consumption, and age. Adequate adjustment for To date, very few studies have used
these potential confounders is therefore essen- biomarkers to estimate coffee intake in cancer
tial, but only possible if researchers have included epidemiological studies. However, the use of
robust measures of exposure to these potential mass spectrometry techniques and metabolomic
confounders and have used analytical methods approaches has recently allowed the identification
that adequately adjust for these variables. Each of several promising coffee biomarkers. A metab-
study reviewed in this monograph, considered by olomic analysis of baseline serum samples from
cancer site in Section  2, has been examined in participants in the Prostate, Lung, Colorectal,
terms of ability to adequately adjust for potential and Ovarian (PLCO) trial in the USA identi-
confounders. fied trigonelline, quinate, 1-methylxanthine
and paraxanthine, along with N-2-furoylglycine
(f) Limitations and catechol sulfate, as potential biomarkers of
There was substantial heterogeneity across coffee intake (Guertin et al., 2014). In a nested
the studies reviewed by the Working Group due case–control study of colorectal cancer risk in
to a variety of factors, such as methods of dietary the same cohort, plasma trigonelline and quinate
assessment and/or measurement, variable defi- concentrations were best correlated with coffee
nitions (e.g. food groups, serving sizes), levels intake as assessed by FFQ (Guertin et al., 2015).
of detail (e.g. caffeinated versus decaffeinated), Negative associations of these metabolites and
analytical categorizations (e.g. servings per week, several others with diagnosis of cancer of the
grams per day), exposure contrasts (analytical colorectum were observed.
cut-points and comparisons of intake levels), and Biomarkers of coffee consumption have
degree of adjustment for potential confounding also been investigated through comparisons of
factors. coffee consumers and non-consumers in studies
An important limitation in almost all of free-living subjects. Dihydrocaffeic acid and
epidemiological studies investigating the rela- its 3-glucuronide measured in 24-hour pooled
tionship between coffee consumption and cancer urine were found to discriminate between high
risk is the lack of details in the description of and low levels of coffee consumption with high
the coffee consumed. Descriptors such as the sensitivity and specificity in a dietary interven-
preparation method (e.g. filtered or not), the tion study in the UK, suggesting potential as
coffee concentration, and drinking temperature markers of habitual coffee consumption (Lloyd
(see the monograph on Very Hot Beverages in et al., 2013). In the EPIC cohort, concentra-
the present volume) are almost always missing, tions of 16 conjugated metabolites of phenolic
although some of these factors could be important acids, mostly glucuronide or sulfate esters, were

61
IARC MONOGRAPHS – 116

measured in 24-hour urine samples, and their metabolites of chlorogenic acids were found to
levels were found to be correlated with both discriminate coffee consumers from controls.
acute and regular coffee intake (Edmands et al., Dihydroferulic acid 4-O-sulfate and dihydro-
2015). Dihydroferulic acid sulfate, feruloylquinic caffeic acid 3-O-sulfate attained the highest
acid glucuronide, ferulic acid sulfate, and guai- plasma concentrations after coffee intake, while
acol glucuronide were the metabolites whose the latter compound and feruloylglycine were
measured intensities best predicted the highest reported as the most effective urinary biomarkers
or lowest quintile of coffee intake. Coffee intake of intake (Stalmach et al., 2009). Most of these
markers including non-phenolic metabolites were coffee metabolites are eliminated within one day
searched for in morning spot urine of 39 French (Stalmach et al., 2009; Lang et al., 2013). [The rapid
coffee consumers from the Supplémentation elimination of these metabolites should not be a
en Vitamines et Minéraux Anti-oxydants limitation to using them as coffee biomarkers,
(SUVIMAX) cohort (Rothwell et al., 2014). The due to the regular intake of coffee by most coffee
intensities of several coffee-derived metabolites consumers.]
accurately classified consumers into high- and [The Working Group noted that the speci-
low-intake groups. The most effective of these ficity of biomarkers for coffee drinking has not
were the diterpene atractyligenin glucuronide, always been assessed as it relies on the known
the cyclic amino acid cyclo(isoleucyl-prolyl), and presence or absence of such compounds in other
trignolline. foods or beverages, and this information is not
Several small, short-term intervention always available. In addition, some biomarkers
studies provided detailed information of coffee might be specific to particular coffee brews,
compounds found in blood or urine after coffee roasts, or varieties. The combinations of several
consumption. For example, urinary concentra- biomarkers may provide clues regarding the type
tions of trigonelline and the product of the coffee of coffee consumed (e.g. caffeinated vs decaf-
roasting process N-methylpyridinium best feinated coffee, varying degrees of roasting).
distinguished subjects given coffee from controls The validity of biomarkers identified in clinical
(Lang et al., 2011). Both compounds remained trials should be treated with caution and cannot
elevated in the urine of coffee consumers for at necessarily be generalized to diverse free-living
least two days after coffee consumption. Another populations.]
coffee roasting product, N-2-furoylglycine, was
identified as a promising biomarker of coffee 1.3.2 Production and consumption volumes
intake in a metabolomic study based on spot
urine profiles of five volunteers administered Coffee production in 2015 was estimated to
one cup of espresso coffee (Heinzmann et al., be about 9 million tonnes (ICO, 2016). In recent
2015). Trigonelline, caffeine, dimethylxanthine, years, about two thirds of the world production
methylxanthine, and ferulic acid in urine were has typically been Arabica coffee and one third
also found to discriminate subjects admin- Robusta. Brazilian production (mainly Arabica)
istered a standardized dose of coffee from accounted for 30%, followed by Viet Nam
controls (Lang et al., 2013). [Dimethylxanthine (19%, mainly Robusta), Colombia (9%, mainly
and methylxanthine are metabolites of caffeine, Arabica), Indonesia (8%), and Ethiopia (4%),
so intake of other beverages containing caffeine among others (ICO, 2016; USDA, 2016).
(e.g. soft drinks, energy drinks, or tea) limit World coffee consumption in 2014 and 2015
the specificity of caffeine and its metabolites was estimated to be about 9 million tonnes (ICO,
as biomarkers of coffee intake.] Several other 2016; USDA, 2016). Together, the European

62
Drinking coffee

Fig. 1.4 Total annual coffee consumption in 2014, in selected countries or regions with high
consumption

3

2.53
2.5
Million metric tonnes

2

1.5 1.43
1.22

1

0.45
0.5
0.26 0.24 0.22 0.16 0.14 0.13 0.12 0.1
0

Venezuela
USA

Japan

India
European Union

Brazil

Mexico
Indonesia

Philippines
Ethiopia

Viet Nam
Russian Federation

From International Coffee Organization (ICO, 2016)

Union countries are responsible for 28% of the difference between gross imports and re-exports
world coffee consumption. Because of their large (ICO, 2016). An estimate of per capita coffee
territorial extension and populations associated consumption can be obtained by dividing disap-
with high consumptions, the USA and Brazil pearance (typically reported in kilograms per
are the individual countries with the highest year) by population. This indicator is primarily
consumption, accounting for about 16% and 13% useful for ranking countries with respect to rela-
of total world consumption, respectively. Japan tive levels of consumption. Disappearance data
(5.6%) and the Russian Federation (2.2%) follow for trade in green coffee beans indicate that the
(ICO, 2016; Fig. 1.4). Nordic countries Finland, Norway, Denmark,
The consumption of decaffeinated coffee and Iceland and Austria are leading consumers of
as a percentage of total consumption in 2009 coffee by this measure (Table 1.3). Other impor-
was estimated as being highest in Spain (17%), tant consumers are Switzerland, Montenegro,
USA (16%), UK (13%), Netherlands (12%), and Sweden, Lebanon and Germany (ICO, 2016).
Belgium/Luxembourg (10%). In all other coun- Historical consumption data from 2011 to 2014
tries the percentage was below 10% (ITC, 2012). (ICO, 2016) show a modest increase in worldwide
Coffee consumption at the population level consumption, with stagnation in countries with a
can be measured in several ways. Data on the high per capita consumption, mainly in Europe.
amounts of coffee in international trade are typi- A few countries in Central America, Mexico, and
cally expressed in terms of disappearance, defined South America also showed this trend. Increasing
as net imports into a country, estimated by the consumption is observed in countries that are

63
IARC MONOGRAPHS – 116

Table 1.3 Annual coffee consumption per person from disappearance data for green coffee beans
(10 highest consuming countries, 2013)

Country Arithmetic mean consumption
(kg/person per year)
Finland 12.07
Norway 9.01
Denmark 8.75
Austria 8.74
Iceland 8.43
Switzerland 8.29
Montenegro 7.61
Sweden 7.33
Lebanon 6.97
Germany 6.92
With permission from the International Coffee Organization (unpublished work)

currently lower per capita consumers, situated Brazil (ABIC, 2016). Consumption is organized
mainly in Asia, Oceania and Africa. Examples by range of consumption and average amount
are Egypt, India, Indonesia, Philippines, Saudi consumed daily. Based on these assessments,
Arabia, Thailand, Turkey, and Viet Nam. the highest individual coffee consumption is in
The preceding estimates of per capita con- Denmark followed by the USA, Netherlands, and
sumption of traded coffee beans may not reflect Germany. There is also a very large individual
the amounts of coffee beverage consumed by indi- variation (23–1914 g/day). [The Working Group
viduals, however. Individual-level consumption noted that the data from the different countries,
of coffee beverages is assessed by dietary surveys including within Europe, were collected from
(see Section 1.3.1) or specialized surveys of coffee different sources with differences in years and
drinking. An important consideration in survey types of questionnaires, among others.]
data is that the proportion of coffee drinkers in
a given population is less than 100%, with that
proportion varying between and within coun-
1.4 Chemical constituents
tries. Consequently, average consumption among 1.4.1 Major constituents
all respondents tends to be less than the average
among coffee drinkers only. The latter quantity An overview of compounds present in green,
corresponds most closely to dose and is compa- roasted, brewed, instant, and decaffeinated coffees
rable to the exposure indicators typically used was provided in the previous IARC Monograph
in epidemiological studies. Table  1.4 presents on coffee (IARC, 1991). More recent updates have
survey data on the average coffee consumption of been published (Farah, 2012; Oestreich-Janzen,
adult coffee drinkers in countries with available 2013).
data. A limitation of such survey data is that they The compounds in coffee are typically classi-
are not available for all countries and the details fied into non-volatile and volatile fractions. The
available within countries may not be compa- non-volatile compounds include water, carbohy-
rable. Data in Table  1.4 were obtained from a drates and fibre, proteins and free amino acids,
FFQ for the USA (NCA, 2016), 24-hour recall for lipids, minerals (40% potassium), organic acids,
Europe (EFSA, 2011), and purchasing data for chlorogenic acids, trigonelline, and caffeine.

64
Drinking coffee

Table 1.4 Mean daily coffee beverage consumption among coffee drinkers in selected countries

Country, year Arithmetic meand Ranged Proportion of consumers within population (%)
(mL/day) (mL/day)
Denmark, NRa 846 86–1914 83
USA, 2016b 740e NR 57 consuming on previous day, 76 in previous year
Netherlands, 2007–2010a 573 100–1253 75
Germany, NRa 539 100–1199 82
Sweden, 2010–2011a 431 75–875 78
Latvia, 2004a 309 90–650 83
Ireland, NRa 259 23–783 54
Brazil, 2016c 222 NR
Italy, 2005–2006a 108 20–240 89
Romania, 2005–2006a 93 11–253 68
UK, NRa 147 9–552 34
a European Food Safety Association (EFSA, 2011)
b National Coffee Association (NCA, 2016)
c Brazilian Coffee Industry Association (ABIC, 2016)

d Converted to L/day with a density of 1.0

e Reported mean of 2.96 cups/day converted to g/day assuming 250 g/cup

NR, not reported

The volatile fraction may contain more than Table  1.6 presents the typical composition of
950 different compounds from chemical classes coffee brew from medium-roasted coffee beans.
such as furans and pyrans, pyrazines, pyrroles, Caffeine is thought to be one of the prin-
ketones, phenols, hydrocarbons, alcohols, alde- cipal components with pharmacological effects
hydes, pyridines, and other compounds (IARC, in coffee, and its mild stimulating effect may be
1991; Farah, 2012). An overview of the composi- the reason for the popularity of this beverage
tion of roasted coffee seeds is provided in Table 1.5. (Lachenmeier et al., 2012). According to a review
Minor compounds such as 16-O-methylcafestol by the European Food Safety Authority (EFSA),
and kahweol (Monakhova et al., 2015) or a standard cup of coffee in Europe has a caffeine
minor chlorogenic acid compounds (Farah & concentration of about 38–69  mg/100  mL; an
Donangelo, 2006) vary between Arabica and average of 45  mg/100  mL was used for intake
Robusta species. assessment (EFSA, 2015).
As a natural product, the chemical composi- For further details on caffeine, see IARC
tion of coffee may vary to a wide degree depending (1991).
on the species and degree of maturation, as well
as soil composition, climate, agricultural prac- 1.4.2 Other constituents and contaminants
tices, and storage conditions (Farah, 2012). For
example, the caffeine content in coffee bever- Some compounds found in coffee have known
ages (including decaffeinated beverages) may toxic properties and, in some cases, carcinogenic
range over 0.3–380  mg/100  mL (Farah, 2012; effects. Coffee constituents and contaminants
Lachenmeier et al., 2013); the niacin content may that have been evaluated by IARC and classified
range over about 10–40  mg/100  mL (Adrian & as possibly carcinogenic to humans (Group 2B)
Frangne, 1991); and cafestol may be present in or higher are shown in Table 1.7. These include
quantities of 2–5 mg/100 mL (Zhang et al., 2012). heating-induced compounds benzo[a]pyrene
(Group 1), acrylamide (Group 2A), acetaldehyde

65
IARC MONOGRAPHS – 116

Table 1.5 Chemical composition of roasted Coffea arabica and Coffea canephora seeds

Compounds Concentrationa (g/100 g)
Coffea arabica Coffea canephora
Carbohydrates/fibre
Sucrose 4.2–tr 1.6–tr
Reducing sugars 0.3 0.3
Polysaccharides (arabinogalactan, mannan, and glucan) 31–33 37
Lignin 3.0 3.0
Pectins 2.0 2.0
Nitrogenous compounds
Protein 7.5–10 7.5–10
Free amino acids ND ND
Caffeine 1.1–1.3 2.4–2.5
Trigonelline 1.2–0.2 0.7–0.3
Nicotinic acid 0.016–0.026 0.014–0.025
Lipids
Coffee oil (triglycerides with unsaponifiables) 17.0 11.0
Diterpene esters 0.9 0.2
Minerals 4.5 4.7
Acids and esters
Chlorogenic acids 1.9–2.5 3.3–3.8
Aliphatic acids 1.6 1.6
Quinic acid 0.8 1.0
Melanoidins 25 25
Content varies according to cultivar, agricultural practices, climate, soil composition, methods of analysis, and roasting degree
a

ND, not detected; tr, trace
Reproduced from Farah (2012). Coffee Constituents. Coffee: Wiley-Blackwell; 2012. p. 21–58

and furan (both Group 2B), the mycotoxins afla- levels in coffee; reductions of the concentration
toxin (Group 1) and ochratoxin A (Group 2B), of ochratoxin A by more than 90% have been
and pesticides DDT, captafol (both Group 2A), observed, depending on the degree of roasting
and dichlorvos (Group 2B). (Soliman, 2002; Romani et al., 2003; Ferraz et al.,
The prevalence of mycotoxins, notably afla- 2010).
toxins and ochratoxin A, in green coffee beans Few studies have assessed the levels of ochra-
is high (Soliman, 2002; Paterson et al., 2014). In toxin A in coffee beverages. Only trace levels of
tropical and subtropical regions where coffee ochratoxin A (< 1 µg/L) were detected in ready-to-
is grown, ochratoxin A and aflatoxin B1 are drink coffee (Noba et al., 2009), and ochratoxin A
produced by Aspergillus species. Contamination has been detected in instant coffee (IARC, 1991;
will vary not only with country but with indi- Vecchio et al., 2012). Although aflatoxin has
vidual farms, depending on manufacturing been detected in green and roasted coffee beans
practices. It has been speculated that with (Soliman, 2002), data on aflatoxin occurrence in
climate change aflatoxin may gain importance brewed coffee are unavailable (Vieira et al., 2015).
as a hazard to coffee production (Paterson et al., Another group of coffee contaminants are
2014). biogenic amines, which may be produced by
The high temperatures of coffee roasting microorganisms in defective beans or during
significantly decrease aflatoxin and ochratoxin A storage. Putrescine, spermidine, and spermine

66
Drinking coffee

Table 1.6 Typical chemical composition per 100 mL of coffee brew from medium roasted coffee

Constituenta Concentration (mg/100 mL)
Caffeine 50–380
Chlorogenic acids 35–500
Trigonelline 40–50
Soluble fibre 200–800
Protein 100
Lipids 0.8
Minerals 250–700
Niacin 10
Melanoidins 500–1500
Brew composition varies according to blend, roasting degree, grind, and method of preparation
a

From Farah (2012)

Table 1.7 Summary of IARC-evaluated compounds that may be present in coffee

Agent IARC Monographs evaluation of carcinogenicity IARC Monographs
Volume (year of
In animals In humans IARC group publication in print)
Acetaldehyde Sufficient Inadequate 2B 71 (1999)
Acrylamide Sufficient Inadequate 2A 60 (1994)
Aflatoxins Sufficient Sufficient 1 100F (2012)
Benzo[a]pyrene Sufficient No data 1 100F (2012)
Caffeic acid Sufficient No data 2B 56 (1993)
Captafol Sufficient No data 2A 53 (1991)
DDT and associated compounds Sufficient Limited 2A 113 (2018)
Dichlorvos Sufficient Inadequate 2B 53 (1991)
Dichloromethane Sufficient Limited 2A 110 (2017)
Furan Sufficient Inadequate 2B 63 (1995)
Methyl isobutyl ketone Sufficient No data 2B 101 (2013)
Ochratoxin A Sufficient Inadequate 2B 56 (1993)

are considered to be major amines in coffee. (Houessou et al., 2007). It has been reported that
Tyramine (one of the most toxic amines), cadav- lower roasting temperatures and shorter roasting
erine, and others are considered to be minor times tend to reduce the formation of PAHs and
amines in coffee (Farah, 2012). acrylamide (Guenther et al., 2007; Houessou
While roasting largely destroys contaminants et al., 2007; Soares et al., 2009).
such as mycotoxins and thermolabile pesticide Among the heat-induced contaminants
residues that may exist in the green coffee, several in coffee, only furan has been systematically
heat-induced contaminants may be formed studied; it occurs in every coffee brew at about
during roasting. These include acetaldehyde 40–100  µg/L depending on the preparation
(Uebelacker & Lachenmeier, 2011), furan (Wenzl method (Waizenegger et al., 2012). The average
et al., 2007; Petisca et al., 2013; Lachenmeier, furan intake per cup of coffee was estimated as
2015), acrylamide (Lantz et al., 2006; Guenther 0.12 µg/kg body weight (Lachenmeier, 2015). In
et al., 2007), and polycyclic aromatic hydro- an experimental study, Houessou et al. (2007)
carbons (PAHs) including benzo[a]pyrene reported total PAH concentrations of <  1  µg/L

67
IARC MONOGRAPHS – 116

in brewed coffee. [The Working Group noted the Bhoo-Pathy N, Peeters PH, Uiterwaal CS, Bueno-de-
absence of systematic survey data on PAHs in Mesquita HB, Bulgiba AM, Bech BH, et al. (2015).
Coffee and tea consumption and risk of pre- and post-
coffee beans or coffee brews.] menopausal breast cancer in the European Prospective
Some cohort studies have reported coffee Investigation into Cancer and Nutrition (EPIC) cohort
as one of the largest contributor to acrylamide study. Breast Cancer Res, 17(15):15–0521. doi:10.1186/
s13058-015-0521-3 PMID:25637171
intake (e.g. Larsson et al., 2009; Wilson et al., Bidel S, Hu G, Jousilahti P, Pukkala E, Hakulinen T,
2012). [Other data indicated that fried potatoes Tuomilehto J (2013). Coffee consumption and risk of
and bread products are the major contributors gastric and pancreatic cancer–a prospective cohort
to the dietary exposures of acrylamide for most study. Int J Cancer, 132(7):1651–9. doi:10.1002/ijc.27773
PMID:22886387
countries (JECFA, 2011). The Working Group Block G, Hartman AM, Dresser CM, Carroll MD,
noted that the populations in some cohort studies Gannon J, Gardner L (1986). A data-based approach to
may therefore have been biased towards groups diet questionnaire design and testing. Am J Epidemiol,
with a high consumption of coffee relative to 124(3):453–69. PMID:3740045
Boggs DA, Palmer JR, Stampfer MJ, Spiegelman D, Adams-
fried foods.] Campbell LL, Rosenberg L (2010). Tea and coffee intake
in relation to risk of breast cancer in the Black Women’s
Health Study. Cancer Causes Control, 21(11):1941–8.
doi:10.1007/s10552-010-9622-6 PMID:20680436
References Borrelli RC, Esposito F, Napolitano A, Ritieni A, Fogliano
V (2004). Characterization of a new potential func-
ABIC (2016). Indicadores da indústria de café no tional ingredient: coffee silverskin. J Agric Food Chem,
Brasil – 2015. Available from: http://abic.com.br/ 52(5):1338–43. doi:10.1021/jf034974x PMID:14995143
estatisticas/indicadores-da-industria/indicadores-da- Butler TL, Fraser GE, Beeson WL, Knutsen SF, Herring
industria-2015/ [Portuguese] RP, Chan J, et al. (2008). Cohort profile: The Adventist
Adrian J, Frangne R (1991). Synthesis and availability of Health Study-2 (AHS-2). Int J Epidemiol, 37(2):260–5.
niacin in roasted coffee. Adv Exp Med Biol, 289:49–59. doi:10.1093/ije/dym165 PMID:17726038
doi:10.1007/978-1-4899-2626-5_4 PMID:1897406 CAC (2009). Codex Alimentarius Commission, Joint
Ainslie-Waldman CE, Koh WP, Jin A, Yeoh KG, Zhu FAO/WHO Food Standards Programme, Report of the
F, Wang R, et al. (2014). Coffee intake and gastric 3rd session of the Codex Committee on Contaminants
cancer risk: the Singapore Chinese health study. in Foods, Proposed Draft Code of Practice for
Cancer Epidemiol Biomarkers Prev, 23(4):638–47. the Prevention and Reduction of Ochratoxin A
doi:10.1158/1055-9965.EPI-13-0886 PMID:24608187 Contamination in Coffee, Appendix VI, page 73.
Aleksandrova K, Bamia C, Drogan D, Lagiou P, Agenda Item 8. ALINORM 09/32/41.
Trichopoulou A, Jenab M, et al. (2015). The association Charrier A, Berthaud J (1987). Botanical classifica-
of coffee intake with liver cancer risk is mediated by tion of coffee. In: Clifford MN, Willson KC, editors.
biomarkers of inflammation and hepatocellular injury: Coffee botany, biochemistry and production of beans
data from the European Prospective Investigation into and beverage. Connecticut, USA: the AVI publishing
Cancer and Nutrition. Am J Clin Nutr, 102(6):1498–508. company, Inc., 1985. pp. 13–47
doi:10.3945/ajcn.115.116095 PMID:26561631 Corrochano BR, Melrose JR, Bentley AC, Fryer PJ,
Anthony F, Combes C, Astorga C, Bertrand B, Graziosi Bakalis S (2015). A new methodology to estimate the
G, Lashermes P (2002). The origin of cultivated Coffea steady-state permeability of roast and ground coffee
arabica L. varieties revealed by AFLP and SSR markers. in packed beds. J Food Eng, 150:106–16. doi:10.1016/j.
Theor Appl Genet, 104(5):894–900. doi:10.1007/s00122- jfoodeng.2014.11.006
001-0798-8 PMID:12582651 Davis AP, Chester M, Maurin O, Fay MF (2007). Searching
Beaton GH (1994). Approaches to analysis of dietary data: for the relatives of Coffea (Rubiaceae, Ixoroideae): the
relationship between planned analyses and choice of circumscription and phylogeny of Coffeeae based
methodology. Am J Clin Nutr, 59(1) Suppl: 253S–61S. on plastid sequence data and morphology. Am J Bot,
PMID:8279436 94(3):313–29. doi:10.3732/ajb.94.3.313 PMID:21636404
Bee S, Brando CHJ, Brumen G, Carvalhaes N, Kolling- Davis AP, Govaerts R, Bridson DM, Stoffelen P (2006).
Speer I, Speer K, et al. (2005). The raw bean. In: Illy A, An annotated taxonomic conspectus of the genus
Viani R editors. Espresso coffee, the science of quality. Coffea (Rubiaceae). Bot J Linn Soc, 152(4):465–512.
Elsevier Academic Press; pp. 87–178. doi:10.1016/ doi:10.1111/j.1095-8339.2006.00584.x
B978-012370371-2/50004-4

68
Drinking coffee

Debry G (1989). Le Café [Coffee], Nancy, France: Centre Farah A (2012). Chapter 2: Coffee constituents. In:
de Nutrition Humaine. [French] Chu Y-F editor. Coffee: emerging health effects and
Discacciati A, Orsini N, Andersson SO, Andrén O, disease prevention. Oxford, UK: Wiley-Blackwell;
Johansson JE, Mantzoros CS, et al. (2013). Coffee doi:10.1002/9781119949893.ch2
consumption and risk of localized, advanced and fatal Farah A, de Paulis T, Moreira DP, Trugo LC, Martin PR
prostate cancer: a population-based prospective study. (2006). Chlorogenic acids and lactones in regular and
Ann Oncol, 24(7):1912–8. doi:10.1093/annonc/mdt105 water-decaffeinated arabica coffees. J Agric Food Chem,
PMID:23508823 54(2):374–81. doi:10.1021/jf0518305 PMID:16417293
Dominianni C, Huang WY, Berndt S, Hayes RB, Ahn J Farah A, Donangelo CM (2006). Phenolic compounds in
(2013). Prospective study of the relationship between coffee. Braz J Plant Physiol, 18(1):23–36. doi:10.1590/
coffee and tea with colorectal cancer risk: the PLCO S1677-04202006000100003
Cancer Screening Trial. Br J Cancer, 109(5):1352–9. Farah A, Ferreira dos Santos T (2015). The coffee plant
doi:10.1038/bjc.2013.434 PMID:23907431 and beans: an introduction. In: Preedy VR, editor.
Dubrow R, Darefsky AS, Freedman ND, Hollenbeck AR, Coffee in health and disease prevention. San Diego
Sinha R (2012). Coffee, tea, soda, and caffeine intake in (CA), USA: Academic Press; pp. 5–10. doi:10.1016/
relation to risk of adult glioma in the NIH-AARP Diet B978-0-12-409517-5.00001-2
and Health Study. Cancer Causes Control, 23(5):757– Fernandes F (2017). Chapter 8, Roasting. In: Farah A,
68. doi:10.1007/s10552-012-9945-6 PMID:22457000 editor. Coffee: Chemistry, quality and health implica-
Edmands WM, Ferrari P, Rothwell JA, Rinaldi S, Slimani tions. London, UK: Royal Society of Chemistry.
N, Barupal DK, et al. (2015). Polyphenol metabo- Ferraz MB, Farah A, Iamanaka BT, Perrone D, Copetti
lome in human urine and its association with intake MV, Marques VX, et al. (2010). Kinetics of ochratoxin
of polyphenol-rich foods across European countries. A destruction during coffee roasting. Food Contr,
Am J Clin Nutr, 102(4):905–13. doi:10.3945/ 21(6):872–7. doi:10.1016/j.foodcont.2009.12.001
ajcn.114.101881 PMID:26269369 Ferrucci LM, Cartmel B, Molinaro AM, Leffell DJ, Bale
Efird JT, Friedman GD, Sidney S, Klatsky A, Habel AE, Mayne ST (2014). Tea, coffee, and caffeine and
LA, Udaltsova NV, et al. (2004). The risk for malig- early-onset basal cell carcinoma in a case-control
nant primary adult-onset glioma in a large, multi- study. Eur J Cancer Prev, 23(4):296–302. doi:10.1097/
ethnic, managed-care cohort: cigarette smoking CEJ.0000000000000037 PMID:24841641
and other lifestyle behaviors. J Neurooncol, 68(1): Filho GRS, Vieira HD, Rodrigues WP, Partelli FL (2015).
57–69. doi:10.1023/B:NEON.0000024746.87666.ed Comparison between manual and semi-mechanical
PMID:15174522 harvest of coffee fruit in mountainous areas. Afr J Agric
EFSA (2011). The EFSA comprehensive European Res, 10(28):2724–30. doi:10.5897/AJAR2014.9356
Food Consumption Database. Parma, Italy: Flament I (2002). Coffee flavor chemistry. West Sussex,
European Food Safety Authority. Available from: England: John Wiley & Sons, Ltd.
http://www.efsa.europa.eu/en/food-consumption/ Freedman ND, Park Y, Abnet CC, Hollenbeck AR, Sinha
comprehensive-database. R (2012). Association of coffee drinking with total and
EFSA (2015). Scientific Opinion on the safety of caffeine. cause-specific mortality. N Engl J Med, 366(20):1891–
Panel on Dietetic Products, Nutrition and Allergies 904. doi:10.1056/NEJMoa1112010 PMID:22591295
EFSA J, 13(5):4120. Friberg E, Orsini N, Mantzoros CS, Wolk A (2009). Coffee
Fagherazzi G, Touillaud MS, Boutron-Ruault MC, Clavel- drinking and risk of endometrial cancer–a popula-
Chapelon F, Romieu I (2011). No association between tion-based cohort study. Int J Cancer, 125(10):2413–7.
coffee, tea or caffeine consumption and breast cancer doi:10.1002/ijc.24543 PMID:19585497
risk in a prospective cohort study. Public Health Gardener H, Rundek T, Wright CB, Elkind MS, Sacco RL
Nutr, 14(7):1315–20. doi:10.1017/S1368980011000371 (2013). Coffee and tea consumption are inversely asso-
PMID:21466740 ciated with mortality in a multiethnic urban popula-
FAO (2014). Arabica coffee manual for Lao PDR 2005, tion. J Nutr, 143(8):1299–308. doi:10.3945/jn.112.173807
January 2014. FAO Corporate Document Repository. PMID:23784068
Rome, Italy: Food and Agriculture Organization of the Gavrilyuk O, Braaten T, Skeie G, Weiderpass E, Dumeaux
United Nations. Available from: http://www.fao.org/ V, Lund E (2014). High coffee consumption and different
docrep/008/ae939e/ae939e00.htm. brewing methods in relation to postmenopausal endo-
Farah A (2009). Coffee as a speciality and functional metrial cancer risk in the Norwegian women and
beverage. In: Paquin P, editor. Functional and speciality cancer study: a population-based prospective study.
beverage technology. 1st ed. Cambridge, UK: Woodhead BMC Womens Health, 14(1):48–14. doi:10.1186/1472-
Publishing in FoodScience, Technology and Nutrition. 6874-14-48 PMID:24666820
doi:10.1533/9781845695569.3.370

69
IARC MONOGRAPHS – 116

Geromel C, Ferreira LP, Guerreiro SM, Cavalari AA, Pot studies. BMC Plant Biol, 8(1):51. doi:10.1186/1471-2229-
D, Pereira LF, et al. (2006). Biochemical and genomic 8-51 PMID:18447947
analysis of sucrose metabolism during coffee (Coffea Hercberg S, Galan P, Preziosi P, Bertrais S, Mennen L,
arabica) fruit development. J Exp Bot, 57(12):3243–58. Malvy D, et al. (2004). The SU.VI.MAX Study: a random-
doi:10.1093/jxb/erl084 PMID:16926239 ized, placebo-controlled trial of the health effects
Giri A, Sturgeon SR, Luisi N, Bertone-Johnson E, of antioxidant vitamins and minerals. Arch Intern
Balasubramanian R, Reeves KW (2011). Caffeinated Med, 164(21):2335–42. doi:10.1001/archinte.164.21.2335
coffee, decaffeinated coffee and endometrial cancer PMID:15557412
risk: a prospective cohort study among US postmeno- Hildebrand JS, Patel AV, McCullough ML, Gaudet MM,
pausal women. Nutrients, 3(11):937–50. doi:10.3390/ Chen AY, Hayes RB, et al. (2013). Coffee, tea, and
nu3110937 PMID:22254087 fatal oral/pharyngeal cancer in a large prospective US
Gloess AN, Schönbächler B, Klopprogge B, Lucio D, cohort. Am J Epidemiol, 177(1):50–8. doi:10.1093/aje/
Chatelain K, Bongartz A, et al. (2013). Comparison kws222 PMID:23230042
of nine common coffee extraction methods: instru- Holick CN, Smith SG, Giovannucci E, Michaud DS (2010).
mental and sensory analysis. Eur Food Res Technol, Coffee, tea, caffeine intake, and risk of adult glioma in
236(4):607–27. doi:10.1007/s00217-013-1917-x three prospective cohort studies. Cancer Epidemiol
Green CJ, de Dauwe P, Boyle T, Tabatabaei SM, Fritschi L, Biomarkers Prev, 19(1):39–47. doi:10.1158/1055-9965.
Heyworth JS (2014). Tea, coffee, and milk consumption EPI-09-0732 PMID:20056621
and colorectal cancer risk. J Epidemiol, 24(2):146–53. Holman JP (2009). Heat transfer. Chapters 5 and 6. New
doi:10.2188/jea.JE20130063 PMID:24531002 York, USA: McGraw-Hill Book Company.
GRIN (2016). Germplasm Resources Information Houessou JK, Maloug S, Leveque AS, Delteil C, Heyd B,
Network. Beltsville (MD), USA: United States Camel V (2007). Effect of roasting conditions on the
Department of Agriculture. Available from: https:// polycyclic aromatic hydrocarbon content in ground
www.ars-grin.gov/, accessed 22 January 2016. Arabica coffee and coffee brew. J Agric Food Chem,
Guenther H, Anklam E, Wenzl T, Stadler RH (2007). 55(23):9719–26. doi:10.1021/jf071745s PMID:17941690
Acrylamide in coffee: review of progress in analysis, Hu G, Tuomilehto J, Pukkala E, Hakulinen T, Antikainen
formation and level reduction. Food Addit Contam, R, Vartiainen E, et al. (2008). Joint effects of coffee
24(Suppl 1):60–70. doi:10.1080/02652030701243119 consumption and serum gamma-glutamyltransferase
PMID:17687700 on the risk of liver cancer. Hepatology, 48(1):129–36.
Guertin KA, Loftfield E, Boca SM, Sampson JN, Moore doi:10.1002/hep.22320 PMID:18537182
SC, Xiao Q, et al. (2015). Serum biomarkers of habitual IARC (1979). Some halogenated hydrocarbons. IARC
coffee consumption may provide insight into the Monogr Eval Carcinog Risk Chem Hum, 20:1–
mechanism underlying the association between 609. Available from: http://publications.iarc.fr/38
coffee consumption and colorectal cancer. Am J Clin PMID:296120
Nutr, 101(5):1000–11. doi:10.3945/ajcn.114.096099 IARC (1986). Some halogenated hydrocarbons and pesti-
PMID:25762808 cide exposures. IARC Monogr Eval Carcinog Risk Chem
Guertin KA, Moore SC, Sampson JN, Huang WY, Xiao Q, Hum, 41:1–407. Available from: http://publications.
Stolzenberg-Solomon RZ, et al. (2014). Metabolomics iarc.fr/59 PMID:3473020
in nutritional epidemiology: identifying metabolites IARC (1987). Overall evaluations of carcinogenicity:
associated with diet and quantifying their potential an updating of IARC Monographs volumes 1 to
to uncover diet-disease relations in populations. Am J 42. IARC Monogr Eval Carcinog Risks Hum Suppl,
Clin Nutr, 100(1):208–17. doi:10.3945/ajcn.113.078758 7:1–440. Available from: http://publications.iarc.fr/139
PMID:24740205 PMID:3482203
Hatzold T (2012). Introduction. In: Chu Y-F, editor. IARC (1991). Coffee, tea, mate, methylxanthines and
Coffee: emerging health effects and disease prevention. methylglyoxal. IARC Monogr Eval Carcinog Risks
Oxford, UK: Wiley-Blackwell. Hum, 51:1–513. Available from: http://publications.
Heinzmann SS, Holmes E, Kochhar S, Nicholson JK, iarc.fr/69 PMID:1674554
Schmitt-Kopplin P (2015). 2-Furoylglycine as a candi- IARC (2017). Some chemicals used as solvents and in
date biomarker of coffee consumption. J Agric Food polymer manufacture. IARC Monogr Eval Carcinog
Chem, 63(38):8615–21. doi:10.1021/acs.jafc.5b03040 Risks Hum, 110:1–276. Available from: http://
PMID:26357997 publications.iarc.fr/547.
Hendre PS, Phanindranath R, Annapurna V, Lalremruata ICO (2016) Historical Data on the Global Coffee Trade.
A, Aggarwal RK (2008). Development of new genomic London, UK: International Coffee Organization.
microsatellite markers from robusta coffee (Coffea Available from: http://www.ico.org/new_historical.
canephora Pierre ex A. Froehner) showing broad asp?section=Statistics, accessed February 2016.
cross-species transferability and utility in genetic

70
Drinking coffee

Illner AK, Freisling H, Boeing H, Huybrechts I, Crispim SP, Lachenmeier DW (2015). Furan in coffee products:
Slimani N (2012). Review and evaluation of innovative a probabilistic exposure estimation. In: Preedy VR,
technologies for measuring diet in nutritional epide- editor. Coffee in health and disease prevention. San
miology. Int J Epidemiol, 41(4):1187–203. doi:10.1093/ Diego, USA: Academic Press; pp. 887–93. doi:10.1016/
ije/dys105 PMID:22933652 B978-0-12-409517-5.00098-X
Illy A, Viani R (2005). Espresso coffee: the science of Lachenmeier DW, Steffen C, el-Atma O, Maixner S,
quality. 2nd ed. Elsevier Academic Press. Löbell-Behrends S, Kohl-Himmelseher M (2012).
Illy F, Illy R (1989). Dal Caffe al Espresso. Milano, Italy: What is a food and what is a medicinal product in the
A. Mondadoni [Italian]. European Union? Use of the benchmark dose (BMD)
Ireland P, Jolley D, Giles G, O’Dea K, Powles J, Rutishauser methodology to define a threshold for “pharmacolog-
I, et al. (1994). Development of the Melbourne FFQ: ical action”. Regul Toxicol Pharmacol, 64(2):286–95.
a food frequency questionnaire for use in an Australian doi:10.1016/j.yrtph.2012.08.017 PMID:22960033
prospective study involving an ethnically diverse Lachenmeier DW, Wegert K, Kuballa T, Schneider R,
cohort. Asia Pac J Clin Nutr, 3(1):19–31. PMID:24351203 Ruge W, Reusch H, et al. (2013). Caffeine intake from
Isaksson B, Jonsson F, Pedersen NL, Larsson J, Feychting beverages in German children, adolescents, and adults.
M, Permert J (2002). Lifestyle factors and pancre- J Caffeine Res, 3(1):47–53. doi:10.1089/jcr.2013.0008
atic cancer risk: a cohort study from the Swedish Lai GY, Weinstein SJ, Albanes D, Taylor PR, McGlynn KA,
Twin Registry. Int J Cancer, 98(3):480–2. doi:10.1002/ Virtamo J, et al. (2013). The association of coffee intake
ijc.10256 PMID:11920604 with liver cancer incidence and chronic liver disease
ITC (2012). The coffee guide. Trade practices of relevance mortality in male smokers. Br J Cancer, 109(5):1344–51.
to exporters in coffee-producing countries. Demand doi:10.1038/bjc.2013.405 PMID:23880821
– decaffeinated coffee. International Trade Centre. Lai GY, Weinstein SJ, Albanes D, Taylor PR, Virtamo J,
Available from: http://www.thecoffeeguide.org/ McGlynn KA, et al. (2014). Association of serum α-to-
coffee-guide/the-markets-for-coffee/. copherol, β-carotene, and retinol with liver cancer
JECFA (2011). Safety evaluation of certain contaminants incidence and chronic liver disease mortality. Br J
in food. WHO Food Additives Series: 63. Geneva, Cancer, 111(11):2163–71. doi:10.1038/bjc.2014.365
Switzerland: Joint FAO/WHO Expert Committee PMID:25314058
on Food Additives, World Health Organization. Lang R, Dieminger N, Beusch A, Lee YM, Dunkel A, Suess
Available from: http://www.inchem.org/documents/ B et al. (2013). Bioappearance and pharmacokinetics
jecfa/jecmono/v63je01.pdf. of bioactives upon coffee consumption. Anal Bioanal
Johnson S, Koh WP, Wang R, Govindarajan S, Yu MC, Chem, 405(26):8487–503. doi:10.1007/s00216-013-
Yuan JM (2011). Coffee consumption and reduced 7288-0 PMID:23982107
risk of hepatocellular carcinoma: findings from the Lang R, Wahl A, Stark T, Hofmann T (2011). Urinary
Singapore Chinese Health Study. Cancer Causes N-methylpyridinium and trigonelline as candidate
Control, 22(3):503–10. doi:10.1007/s10552-010-9725-0 dietary biomarkers of coffee consumption. Mol Nutr
PMID:21258859 Food Res, 55(11):1613–23. doi:10.1002/mnfr.201000656
Knopp S, Bytof G, Selmar D (2006). Influence of processing PMID:21618426
on the content of sugars in green Arabica coffee beans. Lantz I, Ternité R, Wilkens J, Hoenicke K, Guenther H,
Eur Food Res Technol, 223(2):195–201. doi:10.1007/ van der Stegen GH (2006). Studies on acrylamide levels
s00217-005-0172-1 in roasting, storage and brewing of coffee. Mol Nutr
Kolonel LN, Henderson BE, Hankin JH, Nomura AM, Food Res, 50(11):1039–46. doi:10.1002/mnfr.200600069
Wilkens LR, Pike MC, et al. (2000). A multiethnic PMID:17054100
cohort in Hawaii and Los Angeles: baseline charac- Larsson SC, Akesson A, Wolk A (2009). Long-term
teristics. Am J Epidemiol, 151(4):346–57. doi:10.1093/ dietary acrylamide intake and risk of epithelial ovarian
oxfordjournals.aje.a010213 PMID:10695593 cancer in a prospective cohort of Swedish women.
Kushi LH (1994). Gaps in epidemiologic research methods: Cancer Epidemiol Biomarkers Prev, 18(3):994–7.
design considerations for studies that use food- doi:10.1158/1055-9965.EPI-08-0868 PMID:19223560
frequency questionnaires. Am J Clin Nutr, 59(1) Suppl: Lashermes P, Combes M-C, Robert J, Trouslot P, D’Hont
180S–4S. PMID:8279420 A, Anthony F, et al. (1999). Molecular characterisa-
Lachat C, Hawwash D, Ocké MC, Berg C, Forsum E, tion and origin of the Coffea arabica L. genome. Mol
Hörnell A, et al. (2016). Strengthening the Reporting Gen Genet, 261(2):259–66. doi:10.1007/s004380050965
of Observational Studies in Epidemiology-Nutritional PMID:10102360
Epidemiology (STROBE-nut): An Extension of the
STROBE Statement. PLoS Med, 13(6):e1002036.
doi:10.1371/journal.pmed.1002036 PMID:27270749

71
IARC MONOGRAPHS – 116

Lloyd AJ, Beckmann M, Haldar S, Seal C, Brandt K, Taxonomy/Browser/wwwtax.cgi?id=13442, accessed
Draper J (2013). Data-driven strategy for the discovery January 2016.
of potential urinary biomarkers of habitual dietary NLM (2016). Coffee. Dietary supplements labels database.
exposure. Am J Clin Nutr, 97(2):377–89. doi:10.3945/ Bethesda (MD), USA: United States National Library of
ajcn.112.048033 PMID:23269817 Medicine. Available from: https://www.dsld.nlm.nih.
Lueth NA, Anderson KE, Harnack LJ, Fulkerson JA, gov/dsld/index.jsp, accessed 28 March 2016.
Robien K (2008). Coffee and caffeine intake and the Noba S, Uyama A, Mochizuki N (2009). Determination
risk of ovarian cancer: the Iowa Women’s Health Study. of ochratoxin a in ready-to-drink coffee by immuno-
Cancer Causes Control, 19(10):1365–72. doi:10.1007/ affinity cleanup and liquid chromatography-tandem
s10552-008-9208-8 PMID:18704717 mass spectrometry. J Agric Food Chem, 57(14):6036–40.
Makiuchi T, Sobue T, Kitamura T, Ishihara J, Sawada N, doi:10.1021/jf900546p PMID:19537783
Iwasaki M, et al. (2016). Association between green Norberg M, Wall S, Boman K, Weinehall L (2010). The
tea/coffee consumption and biliary tract cancer: Västerbotten Intervention Programme: background,
A population-based cohort study in Japan. Cancer Sci, design and implications. Glob Health Action, 3(1):10.
107(1):76–83. doi:10.1111/cas.12843 PMID:26530716 doi:10.3402/gha.v3i0.4643 PMID:20339479
Mennen L, Hirvonen T, Arnault N, Bertrais S, Galan Oba S, Nagata C, Nakamura K, Takatsuka N, Shimizu
P, Hercberg S (2007). Consumption of black, green H (2008). Self-reported diabetes mellitus and risk of
and herbal tea and iron status in French adults. Eur mortality from all causes, cardiovascular disease, and
J Clin Nutr, 61(10):1174–9. doi:10.1038/sj.ejcn.1602634 cancer in Takayama: a population-based prospective
PMID:17299492 cohort study in Japan. J Epidemiol, 18(5):197–203.
Michels KB, Willett WC, Fuchs CS, Giovannucci E (2005). doi:10.2188/jea.JE2008004 PMID:18753735
Coffee, tea, and caffeine consumption and incidence of Oestreich-Janzen S (2013). Chemistry of coffee.
colon and rectal cancer. J Natl Cancer Inst, 97(4):282– Reference module in chemistry, molecular sciences
92. doi:10.1093/jnci/dji039 PMID:15713963 and chemical engineering. Elsevier Inc. doi:10.1016/
Monakhova YB, Ruge W, Kuballa T, Ilse M, Winkelmann B978-0-12-409547-2.02786-4
O, Diehl B, et al. (2015). Rapid approach to identify Paganini-Hill A, Kawas CH, Corrada MM (2007).
the presence of Arabica and Robusta species in coffee Non-alcoholic beverage and caffeine consumption
using 1H NMR spectroscopy. Food Chem, 182:178–84. and mortality: the Leisure World Cohort Study. Prev
doi:10.1016/j.foodchem.2015.02.132 PMID:25842325 Med, 44(4):305–10. doi:10.1016/j.ypmed.2006.12.011
Murray SS, Bjelke E, Gibson RW, Schuman LM (1981). PMID:17275898
Coffee consumption and mortality from ischemic heart Paterson RRM, Lima N, Taniwaki MH (2014). Coffee,
disease and other causes: results from the Lutheran mycotoxins and climate change (2014). Food Res Int,
Brotherhood study, 1966–1978. Am J Epidemiol, 61:1–15. doi:10.1016/j.foodres.2014.03.037
113(6):661–7. doi:10.1093/oxfordjournals.aje.a113145 Petisca C, Perez-Palacios T, Farah A, Pinho O, Ferreira
PMID:7234854 IMPL (2013). Furans and other volatile compounds in
Murthy PS, Naidu MM (2012). Sustainable manage- ground roasted and espresso coffee using headspace
ment of coffee industry by-products and value addi- solid-phase microextraction: Effect of roasting speed.
tion — a review. Resour Conserv Recycling, 66:45–8. Food Bioprod Process, 91(3):233–41. doi:10.1016/j.
doi:10.1016/j.resconrec.2012.06.005 fbp.2012.10.003
Naganuma T, Kuriyama S, Kakizaki M, Sone T, Nakaya Petracco M (2005). The cup. In: Illy A, Viani R
N, Ohmori-Matsuda K, et al. (2008). Coffee consump- editors. Espresso coffee: the science of quality.
tion and the risk of oral, pharyngeal, and esophageal Elsevier Academic Press; pp. 259–89. doi:10.1016/
cancers in Japan: the Miyagi Cohort Study. Am J B978-012370371-2/50008-1
Epidemiol, 168(12):1425–32. doi:10.1093/aje/kwn282 Phillips RL, Snowdon DA (1983). Association of meat
PMID:18974083 and coffee use with cancers of the large bowel, breast,
Narita Y, Inouye K (2012). High antioxidant activity of and prostate among Seventh-Day Adventists: prelim-
coffee silverskin extracts obtained by the treatment of inary results. Cancer Res, 43(5) Suppl: 2403s–8s.
coffee silverskin with subcritical water. Food Chem, PMID:6831464
135(3):943–9. doi:10.1016/j.foodchem.2012.05.078 Pietinen P, Hartman AM, Haapa E, Räsänen L, Haapakoski
PMID:22953809 J, Palmgren J, et al. (1988). Reproducibility and validity
NCA (2016). National coffee drinking trends 2016. New of dietary assessment instruments. I. A self-adminis-
York, USA: National Coffee Association of USA. tered food use questionnaire with a portion size picture
Available from: http://www.ncausa.org/. booklet. Am J Epidemiol, 128(3):655–66. PMID:2458036
NCBI (2014). Coffea. Taxonomy Browser. Bethesda (MD), Riboli E, Hunt KJ, Slimani N, Ferrari P, Norat T, Fahey
USA: National Center for Biotechnology Information. M, et al. (2002). European Prospective Investigation
Available from: http://www.ncbi.nlm.nih.gov/ into Cancer and Nutrition (EPIC): study populations

72
Drinking coffee

and data collection. Public Health Nutr, 5(6B):1113–24. on Coffee Science, ASIC 2008, Campinas, Brazil;
doi:10.1079/PHN2002394 PMID:12639222 pp. 239–241.
Rimm EB, Giovannucci EL, Stampfer MJ, Colditz GA, Soliman KM (2002). Incidence, level, and behavior of
Litin LB, Willett WC (1992). Reproducibility and aflatoxins during coffee bean roasting and decaffein-
validity of an expanded self-administered semiquan- ation. J Agric Food Chem, 50(25):7477–81. doi:10.1021/
titative food frequency questionnaire among male jf011338v PMID:12452679
health professionals. Am J Epidemiol, 135(10):1114–26, Spohn R (2015). Water colour of Coffea arabica, Engen,
discussion 1127–36. PMID:1632423 Germany: Dr. Roland Spohn, Botanik-Fotoarchiv,
Romani S, Pinnavaia GG, Dalla Rosa M (2003). Influence Illustration, Malerei.
of roasting levels on ochratoxin a content in coffee. Stalmach A, Mullen W, Barron D, Uchida K, Yokota T,
J Agric Food Chem, 51(17):5168–71. doi:10.1021/ Cavin C, et al. (2009). Metabolite profiling of hydrox-
jf030116p PMID:12903986 ycinnamate derivatives in plasma and urine after
Rothwell JA, Fillâtre Y, Martin JF, Lyan B, Pujos- the ingestion of coffee by humans: identification
Guillot E, Fezeu L, et al. (2014). New biomarkers of of biomarkers of coffee consumption. Drug Metab
coffee consumption identified by the non-targeted Dispos, 37(8):1749–58. doi:10.1124/dmd.109.028019
metabolomic profiling of cohort study subjects. PLoS PMID:19460943
ONE, 9(4):e93474. doi:10.1371/journal.pone.0093474 Steevens J, Schouten LJ, Verhage BAJ, Goldbohm RA, van
PMID:24713823 den Brandt PA (2007). Tea and coffee drinking and
Sánchez DA, Anzola C (2012). Chemical characterization ovarian cancer risk: results from the Netherlands Cohort
of the coffee silverskin (Coffea arabica) in varieties Study and a meta-analysis. Br J Cancer, 97(9):1291–4.
Colombia and Caturra. Rev Colomb Quim, 41(2):211–26. doi:10.1038/sj.bjc.6604008 PMID:17923877
Sauvaget C, Allen N, Hayashi M, Spencer E, Nagano J Thesaurus (2016). Brewing. Online dictionary. Available
(2002). Validation of a food frequency questionnaire in from: http://www.dictionary.com/browse/brewing,
the Hiroshima/Nagasaki Life Span Study. J Epidemiol, accessed 11 February 2016.
12(5):394–401. doi:10.2188/jea.12.394 PMID:12395883 Thompson FE, Kipnis V, Midthune D, Freedman LS,
Silvera SA, Jain M, Howe GR, Miller AB, Rohan TE (2007). Carroll RJ, Subar AF, et al. (2008). Performance of a
Intake of coffee and tea and risk of ovarian cancer: food-frequency questionnaire in the US NIH-AARP
a prospective cohort study. Nutr Cancer, 58(1):22–7. (National Institutes of Health-American Association of
doi:10.1080/01635580701307945 PMID:17571963 Retired Persons) Diet and Health Study. Public Health
Sinha R, Cross AJ, Daniel CR, Graubard BI, Wu JW, Nutr, 11(2):183–95. doi:10.1017/S1368980007000419
Hollenbeck AR, et al. (2012). Caffeinated and decaffein- PMID:17610761
ated coffee and tea intakes and risk of colorectal cancer Thompson FE, Subar AF (2013). Dietary assessment meth-
in a large prospective study. Am J Clin Nutr, 96(2):374– odology. In: Coulston AM, Boushey CJ, Ferruzzi MG,
81. doi:10.3945/ajcn.111.031328 PMID:22695871 editors. Nutrition in the prevention and treatment of
Sjøl A, Grunnet K, Schroll M (1991). Secular trends in disease. 3rd ed. Elsevier Academic Press. doi:10.1016/
serum cholesterol, high density lipoproteins and B978-0-12-391884-0.00001-9
triglycerides 1964–1987. Int J Epidemiol, 20(1):105–13. Toci A, Farah A, Trugo LC (2006). Effect of decaffeination
doi:10.1093/ije/20.1.105 PMID:2066206 using dichloromethane on the chemical composition
Slimani N, Freisling H, Illner AK, Huybrechts I (2015). of arabica and robusta raw and roasted coffees. Quim
Methods to determine dietary intake. Chapter 4. In: Nova, 29(5):965–71. doi:10.1590/S0100-40422006000
Lovegrove JA, Hodson L, Sharma S, Lanham-New 500015
SA, editors. Nutrition research methodologies. Wiley- Toci AT, Farah A (2008). Volatile compounds as poten-
Blackwell; pp. 48–70. Available from: http://eu.wiley. tial defective coffee beans’ markers. Food Chem,
com/WileyCDA/WileyTitle/productCd-EHEP003370. 108(3):1133–41. doi:10.1016/j.foodchem.2007.11.064
html. PMID:26065781
Slimani N, Kaaks R, Ferrari P, Casagrande C, Clavel- Toci AT, Neto VJ, Torres AG, Farah A (2013). Changes
Chapelon F, Lotze G, et al. (2002). European Prospective in triacylglycerols and free fatty acids composition
Investigation into Cancer and Nutrition (EPIC) calibra- during storage of roasted coffee. Lebensm Wiss Technol,
tion study: rationale, design and population character- 50(2):581–90. doi:10.1016/j.lwt.2012.08.007
istics. Public Health Nutr, 5(6B):1125–45. doi:10.1079/ Trugo L (2003). Coffee analysis. In: Caballero B, Trugo
PHN2002395 PMID:12639223 LC, Finglas P, editors. Encyclopedia of food science
Soares C, Farah A, Fernandes F, Casal S, Fernandes J (2009). and nutrition. Volume 3. 2nd ed. Oxford, UK: Academic
Influence of the roasting conditions on the formation Press. doi:10.1016/B0-12-227055-X/00271-6
of acrylamide in Brazilian coffee: preliminary results. Uebelacker M, Lachenmeier DW (2011). Quantitative
In: Proceedings of the 22nd International Conference determination of acetaldehyde in foods using auto-
mated digestion with simulated gastric fluid followed

73
IARC MONOGRAPHS – 116

by headspace gas chromatography. J Autom Methods Willett WC, Sampson L, Browne ML, Stampfer MJ,
Manag Chem, 2011:907317. doi:10.1155/2011/907317 Rosner B, Hennekens CH, et al. (1988). The use of a self-
PMID:21747735 administered questionnaire to assess diet four years in
Urgert R, van der Weg G, Kosmeijer-Schuil TG, van de the past. Am J Epidemiol, 127(1):188–99. doi:10.1093/
Bovenkamp P, Hovenier R, Katan MB (1995). Levels oxfordjournals.aje.a114780 PMID:3337073
of the cholesterol-elevating diterpenes cafestol and Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C,
kahweol in various coffee brews. J Agric Food Chem, Witschi J, et al. (1985). Reproducibility and validity of
43(8):2167–72. doi:10.1021/jf00056a039 a semiquantitative food frequency questionnaire. Am J
USDA (2016). Coffee: world market and trade. June 2016. Epidemiol, 122(1):51–65. PMID:4014201
United States Department of Agriculture, Foreign Wilson KM, Giovannucci E, Stampfer MJ, Mucci LA
Agricultural Service. Available from: http://apps.fas. (2012). Dietary acrylamide and risk of prostate cancer.
usda.gov/psdonline/circulars/coffee.pdf. Int J Cancer, 131(2):479–87. doi:10.1002/ijc.26383
van Dusseldorp M, Smits P, Lenders JW, Thien T, PMID:21866549
Katan MB (1991). Boiled coffee and blood pressure. Wilson KM, Kasperzyk JL, Rider JR, Kenfield S, van
A 14-week controlled trial. Hypertension, 18(5):607–13. Dam RM, Stampfer MJ, et al. (2011). Coffee consump-
doi:10.1161/01.HYP.18.5.607 PMID:1937663 tion and prostate cancer risk and progression in
Vecchio A, Mineo V, Planeta D (2012). Ochratoxin A in the Health Professionals Follow-up Study. J Natl
instant coffee in Italy (2012). Food Contr, 28(2):220–3. Cancer Inst, 103(11):876–84. doi:10.1093/jnci/djr151
doi:10.1016/j.foodcont.2012.04.029 PMID:21586702
Veierød MB, Thelle DS, Laake P (1997). Diet and risk Yamada H, Kawado M, Aoyama N, Hashimoto S, Suzuki
of cutaneous malignant melanoma: a prospec- K, Wakai K, et al. ; JACC Study Group(2014). Coffee
tive study of 50,757 Norwegian men and women. consumption and risk of colorectal cancer: the Japan
Int J Cancer, 71(4):600–4. doi:10.1002/(SICI)1097- Collaborative Cohort Study. J Epidemiol, 24(5):370–8.
0215(19970516)71:4<600::AID-IJC15>3.0.CO;2-F doi:10.2188/jea.JE20130168 PMID:24857957
PMID:9178814 Yang TO, Crowe F, Cairns BJ, Reeves GK, Beral V (2015).
Viani R (1986). Coffee. Ullmann’s encyclopedia of indus- Tea and coffee and risk of endometrial cancer: cohort
trial chemistry. Volume A7. Weinheim, Germany: study and meta-analysis. Am J Clin Nutr, 101(3):570–8.
VCH Verlagsgesellschaft; pp. 315–39. doi:10.3945/ajcn.113.081836 PMID:25733642
Vieira T, Cunha S, Casal S (2015). Chapter 25: Mycotoxins Zhang C, Linforth R, Fisk ID (2012). Cafestol extraction
in coffee. In: Preed V, editor. Coffee in health and yield from different coffee brew mechanisms. Food Res
disease prevention. Academic Press, Elsevier; pp. Int, 49(1):27–31. doi:10.1016/j.foodres.2012.06.032
225–33. doi:10.1016/B978-0-12-409517-5.00025-5
Waizenegger J, Castriglia S, Winkler G, Schneider R,
Ruge W, Kersting M, et al. (2011). Caffeine exposure
in children and adolescents consuming ready-to-drink
coffee products. J Caffeine Res, 1(4):200–5. doi:10.1089/
jcr.2011.0024
Waizenegger J, Winkler G, Kuballa T, Ruge W, Kersting
M, Alexy U, et al. (2012). Analysis and risk assessment
of furan in coffee products targeted to adolescents.
Food Addit Contam Part A Chem Anal Control Expo
Risk Assess, 29(1):19–28. doi:10.1080/19440049.2011.61
7012 PMID:22035212
Weiderpass E, Sandin S, Lof M, Oh JK, Inoue M, Shimazu
T, et al. (2014). Endometrial cancer in relation to coffee,
tea, and caffeine consumption: a prospective cohort
study among middle-aged women in Sweden. Nutr
Cancer, 66(7):1132–43. doi:10.1080/01635581.2014.948
214 PMID:25181598
Wellman FL (1961). Coffee: botany, cultivation and utili-
zation. London, UK: Leonard Hill.
Wenzl T, Lachenmeier DW, Gökmen V (2007). Analysis
of heat-induced contaminants (acrylamide, chloro-
propanols and furan) in carbohydrate-rich food. Anal
Bioanal Chem, 389(1):119–37. doi:10.1007/s00216-007-
1459-9 PMID:17673989

74
2. CANCER IN HUMANS

2.1 Cancer of the bladder but did not conduct sensitivity analyses, as well
as one study that did neither, are then discussed.
2.1.1 Cohort studies Overall, studies with a large sample size are
See Table 2.1, Fig 2.1, Fig. 2.2, and Fig. 2.3 considered more informative as measures of
This section summarizes the results of the association will tend to be more precise; we
Working Group’s review of prospective cohort therefore discuss larger studies first, followed by
studies that reported on the association between smaller studies.
drinking coffee and the risk of cancer of the In the following paragraphs the cohort
bladder. One study that reported on bladder studies that were considered the most informa-
cancer mortality as an end-point (Snowdon & tive by the Working Group are described. These
Phillips, 1984) was excluded, as the role of coffee studies were given more weight in the evaluation.
in cancer etiology cannot be distinguished In the Netherlands Cohort Study (Zeegers
from its role in cancer progression or response et al., 2001), 569 incident cases of cancer of the
to treatment. Also excluded are three studies, urinary bladder were identified. Among men,
two of the same cohort, that did not report esti- the relative risk for the highest level of intake
mates of association (Schulte et al., 1985, 1986; (≥  7 cups/day) compared with the lowest (0 to
Whittemore et al., 1985). < 2 cups/day) was 1.33 (95% CI, 0.94–1.90), with
When reviewing the available studies, the an estimate per 1 cup/day of coffee of 1.04 (95%
Working Group considered two important CI, 1.00–1.09). The test for trend was not statis-
criteria in evaluating how informative each was. tically significant (P  for trend,  0.06). Among
One was appropriate adjustment for tobacco women, the relative risk for the highest level
smoking, given that this is an important bladder (≥  5 cups/day) was 0.36 (95% CI,  0.18–0.72),
cancer risk factor and is often reported to be with an estimate per 1 cup/day of 0.83 (95%
correlated with coffee drinking. The other was CI,  0.72–0.96). A statistically significant test
consideration of sensitivity analyses excluding for trend (P  for trend, <  0.01) was reported.
patients diagnosed too close to the start of the Sensitivity analyses excluding cases diagnosed
cohort; patients with bladder cancer might be in the first 1–2 years of follow-up did not change
likely to change their coffee drinking habits, results. [The limitations of this study were the
which might lead to bias in the analyses. Studies lack of consideration of coffee drinking history
that conducted such sensitivity analyses and and lack of stratification by smoking status.]
adjusted for tobacco smoking were therefore
considered to be the most informative, and are
discussed first. Studies that adjusted for smoking

75
76

IARC MONOGRAPHS – 116
Table 2.1 Cohort studies on cancer of the bladder and drinking coffee

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category cases/ (95% CI) controlled
enrolment/ assessment method or level deaths
follow-up
period
Zeegers et al. 3500, Netherlands Cohort Urinary Coffee consumption among men (cups/day) Age, numbers of Strengths: prospective, large
(2001) Study, men and women bladder: 0 to < 2 23 0.89 (0.51–1.54) cigarettes/day, years number of cases, detailed
Netherlands, (aged 55–69 yr), case– ~96% TCC 2 to < 3 32 0.72 (0.45–1.13) of cigarette smoking questionnaire including 19
1986 cohort approach beverages, both men and
3 to < 4 61 1.27 (0.87–1.87)
(enrolment), Exposure assessment women included, complete
1992 (follow-up) method: 4 to < 5 119 1.00 follow-up data
FFQ (non-validated 5 to < 6 72 0.98 (0.68–1.4) Limitations: no drinking
coffee questions, 6 to < 7 91 1.25 (0.89–1.76) history; no follow-up
self-administered, ≥ 7 93 1.33 (0.94–1.90) information
frequency and amount), Per 1 cup/day NR 1.04 (1.00–1.09)
caffeinated coffee only
Coffee consumption among women (cups/day)
(low consumption of
decaffeinated) 0 to < 2 11 1.23 (0.56–2.73)
2 to < 3 13 0.84 (0.4–1.76)
3 to < 4 20 1.00
4 to < 5 17 0.44 (0.22–0.86)
≥ 5 17 0.36 (0.18–0.72)
Per 1 cup/day NR 0.83 (0.72–0.96)
Ros et al. (2011) 233 236 (67 914 men and Urinary Coffee consumption (mL/day) Age, sex, centre, Strengths: prospective
10 European 165 322 women), EPIC, bladder: T1: < 429 133 1.00 smoking status, large cohort, extensive set
countries, subjects aged 25–70 yr UCC (men), 250 duration of smoking, of potential confounders,
1992–2000 Exposure assessment (women) lifetime intensity possible to distinguish
(enrolment), method: validated FFQ, T2: 429–874 179 1.11 (0.88–1.41) of smoking, energy between low- and high-risk
follow-up varied frequency and amount (men), intake from fat and urothelial bladder cancers
by country considered 250–469 non-fat sources Limitations: no history
(women) of coffee drinking, no
information about type of
T3: ≥ 875 201 1.11 (0.85–1.43)
coffee studied, results not
(men), ≥ 500
stratified by sex or smoking,
(women)
no follow-up information
Continuous 380 1.00 (0.98–1.03) on exposure
for every 100
mL increase
(observed)
Trend test P value, 0.5
Table 2.1 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category cases/ (95% CI) controlled
enrolment/ assessment method or level deaths
follow-up
period
Michaud et al. 47 909; HPFS, male Urinary Decaffeinated coffee consumption Geographic region, Strengths: prospective,
(1999) health professionals aged bladder: < 1 cup/mo 106 1.00 age, pack-years of follow-up information every
USA, 1986 40–75 yr in all 50 states, 90% TCC 1 cup/mo–6 65 0.94 (0.69–1.29) smoking, current 2 yr
(enrolment), predominantly white cups/wk smoking status, Limitations: restricted to
1996 (last Exposure assessment energy intake, mostly white professional
1–3 cups/day 72 1.20 (0.87–1.65)
follow-up) method: intake of fruits and men in USA (no women
validated FFQ by mail, ≥ 4 cups/day 9 0.83 (0.41–1.66) vegetables, intake of included), no history of
regular and decaffeinated Trend test P value, 0.47 all other beverages intake
coffee, frequency/serving Coffee consumption (water, milk, juice,
size assessed Per 240 mL of 252 0.93 (0.85–1.02) soda, lemonade, tea,
daily intake alcohol)
< 1 cup/mo 75 1.00
1 cup/mo–6 56 0.97 (0.68–1.37)
cups/wk
1–3 cups/day 98 1.00 (0.73–1.37)
≥ 4 cups/day 23 0.79 (0.48–1.30)
Trend test P value, 0.56
Nagano et al. 38 540 atomic bomb Urinary Coffee consumption frequency (times/wk) Age, sex, radiation Strengths: prospective
(2000) survivors, Life Span Study bladder 0 25 1.00 dose, smoking status Limitations: modest
Japan, (men and women) 1–4 32 0.73 (0.43–1.25) and cigarettes/day, numbers, not representative
1979–1981 Exposure assessment education level, of all Japanese population,
≥5 32 0.90 (0.52–1.56)
(enrolment), method: BMI, calendar time no information on serving
1980–1993 frequency only by Trend test P value, 0.78 sizes, consumption history,
(follow-up) self-administered or types of coffee
questionnaire

Drinking coffee
77
78

IARC MONOGRAPHS – 116
Table 2.1 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category cases/ (95% CI) controlled
enrolment/ assessment method or level deaths
follow-up
period
Jacobsen et al. 16 555; two cohorts Urinary Coffee consumption (cups/day): men only Age, residence, Strengths: prospective,
(1986) of Norwegian men bladder ≤ 2 20 1.00 smoking status, sensitivity analyses
Norway, 1964 (population sample and > 7 10 0.98 (NR) cigarettes/day considering time between
(enrolment), brothers of migrants to the diagnosis and baseline
Trend test P value, 0.88
1967 USA); spouses and siblings Limitations: no assessment
(questionnaire), of individuals enrolled in of duration of coffee
follow-up until a case–control study of drinking, unclear reference
1978 gastrointestinal cancer period for coffee intake,
were included coffee type only coffee
Exposure assessment (decaffeinated/instant not
method: commonly consumed)
validated self-administered
questionnaire with follow-
up
Stensvold & 43 973 men and women Urinary Coffee consumption (cups/day): men Age, cigarettes Strengths: population-
Jacobsen (1994) aged 35–54 yr participating bladder: ≤4 13 1.00 per day, county of based, included participants
Norway, in cardiovascular screening ICD-7, 181 5–6 8 0.70 residence in different parts of Norway
1977–1982 programme Limitations: no assessment
≥7 19 1.50
(enrolment), Exposure assessment of duration of coffee intake
follow-up until method: Per 2 cup/day NR 1.13 (0.87–1.49) or type of coffee/preparation
1990 validated self-administered increase method, modest sample size
FFQ Coffee consumption (cups/day): women
≤ 4 3 1.00
5–6 5 2.10
≥ 7 5 2.40
Per 2 cup/day NR 1.22 (0.73–2.05)
increase
Table 2.1 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category cases/ (95% CI) controlled
enrolment/ assessment method or level deaths
follow-up
period
Sugiyama et al. 73 346 (38 646 Miyagi, Urinary Coffee consumption (cups/day) Sex, age, BMI, Strengths: prospective, large
(2017) 34 700 Ohsaki) men and bladder: Never 63 1.00 history of cohorts, use of population-
Japan, 1990– women aged 40–79 yr, ICD-O-3 Occasionally 130 1.22 (0.90–1.66) hypertension, based registries
2007 (Miyagi), cohorts were pooled for C67–67.9 diabetes mellitus, Limitations: no history of
1–2 65 0.88 (0.61–1.26)
1994–2008 analyses myocardial drinking coffee assessed, no
(Ohsaki) Exposure assessment ≥ 3 16 0.56 (0.32–0.99) infarction, stroke, follow-up information (only
method: Trend test P value, 0.04 job status, years of baseline), no information on
validated self-administered Urinary Coffee consumption (cups/day) stratified by education, smoking brewing or type of coffee,
FFQ bladder: smoking: never smokers status and cigarettes/ no occupational exposures
ICD-O-3 Never 19 1.00 day, alcohol assessed, very few cases
C67–67.9 Occasionally 35 1.46 (0.82–2.58) consumption, green
tea consumption,
1–2 13 0.97 (0.47–2.01)
time spent walking
≥ 3 2 0.62 (0.14–2.72)
Coffee consumption (cups/day) stratified by
smoking: former or current smokers
Never 38 1.00
Occasionally 83 1.22 (0.83–1.81)
1–2 48 0.95 (0.61–1.47)
≥ 3 13 0.61 (0.32–1.17)

Drinking coffee
79
80

IARC MONOGRAPHS – 116
Table 2.1 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category cases/ (95% CI) controlled
enrolment/ assessment method or level deaths
follow-up
period
Kurahashi et al. 133 084 (65 660 men, Urinary Coffee consumption among men Age, area of Decaffeinated coffee is rare
(2009) 67 424 women), JPHC, bladder Almost never 50 1.00 recruitment, in Japan; no other cancers
Japan 104 440 residents of 11 1–4 times/wk 52 1.26 (0.84–1.88) smoking status/ reported in this paper
1990, 1993 public health centre pack-years, alcohol Strengths: prospective,
1–2 cups/day 43 1.53 (0.98–2.37)
(enrolment), areas across Japan of age drinking, green tea catchment area includes
2005 (follow-up) 40–69 yr included ≥ 3 cups/day 19 1.37 (0.75–2.51) most of the country,
Exposure assessment Trend test P value, 0.09 stratification by sex and
method: Coffee consumption among women smoking
validated, self- Almost never 19 1.00 Limitations: no assessment
administered 1–4 times/wk 15 1.03 (0.51–2.07) of drinking history, modest
questionnaire assessing numbers (especially for
≥ 1 cup/day 8 0.55 (0.23–1.33)
frequency/amount (no stratified analyses)
decaffeinated coffee Trend test P value, 0.23
considered) Coffee frequency among men stratified by  
smoking status
Among never smokers
Almost none 6 1.00
1–4 times/wk 9 1.89 (0.67–5.32)
≥ 1 cup/day 11 2.48 (0.88–7.05)
Among former smokers
Almost none 13 1.00
1–4 times/wk 13 1.25 (0.58–2.71)
≥ 1 cup/day 16 2.09 (0.96–4.54)
Among current smokers
Almost none 29 1.00
1–4 times/wk 30 1.11 (0.65–1.9)
≥ 1 cup/day 33 1.13 (0.65–1.97)
Table 2.1 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category cases/ (95% CI) controlled
enrolment/ assessment method or level deaths
follow-up
period
Chyou et al. 7355 Japanese men born Urinary Coffee consumption (cups/wk) Age, pack-years Strengths: prospective, good
(1993) during 1900–1919 (no bladder ≤ 1 5 1.00 smoking assessment of cancers
Hawaii (Oahu), other criteria mentioned) 2–4 5 3.52 (1.02–12.2) Limitations: modest sample
1965–1968 Exposure assessment size, only assessed past 24
≥ 5 86 2.07 (0.84–5.12)
(enrolment), method: hours of intake not long-
15 years of 24-hour recall Trend test P value, 0.174 term history of drinking,
follow-up questionnaire (no few criteria listed for study
decaffeinated coffee eligibility, only men
considered)
Mills et al. 34 198 non-Hispanic white Urinary Coffee intake frequency (cups/day) Age, sex, smoking Strengths: prospective, men
(1991) members of Seventh- bladder: Never 26 1.00 and women included
USA day Adventist church in 92% TCC < 1 7 0.98 (0.41–2.31) Limitations: no assessment
(California), California, > 25 yr old of duration of coffee
1 2 0.44 (0.11–1.83)
1974 Exposure assessment drinking, population
(recruitment), method: ≥ 2 12 1.99 (0.91–4.34) studied does not
1976 (survey), self-administered Trend test P value, 0.13 traditionally drink coffee
1982 (end of 51-item FFQ and Coffee frequency among never smokers so intake of coffee might be
follow-up) lifestyle questionnaire (cups/day) a proxy for other changes
(no decaffeinated coffee Never NR 1.00 from traditional Adventist
considered) < 1 NR 1.11 (0.31–3.95) lifestyle
≥ 1 NR 2.03 (0.70–5.87)
Coffee frequency among past/current smokers
(cups/day)
Never NR 1.00
< 1 NR 0.93 (0.29–2.96)
≥ 1 NR 1.14 (0.46–2.80)
BMI, body mass index; CI, confidence interval; EPIC, European Prospective Investigation into Cancer and Nutrition; FFQ, food frequency questionnaire; HPFS, Health Professionals
Follow-up Study; ICD-7, International Classification of Disease - Revision 7; ICD-O-3, International Classification of Disease – Oncology Revision 3; JPHC, Japan Public Health Center-
based Prospective; mo, month(s); NR, not reported; TCC, transitional cell carcinoma; UCC, urothelial cell carcinoma; wk, week(s); yr, year(s)

Drinking coffee
81
Fig. 2.1 Relative risk estimate for coffee and bladder cancer cohorts: both sexes
82

IARC MONOGRAPHS – 116
*

* CIs were forced to display on the plot
Compiled by the Working Group
Fig. 2.2 Relative risk estimate for coffee and bladder cohorts: men only

*

Drinking coffee
* CIs were forced to display on the plot
Compiled by the Working Group
83
Fig. 2.3 Relative risk estimate for coffee and bladder cohorts: women only
84

IARC MONOGRAPHS – 116
*

* CIs were forced to display on the plot
Compiled by the Working Group
Drinking coffee

In the European Prospective Investigation of dose–response or trend (P  for trend,  0.78).
into Cancer and Nutrition (EPIC) study, 513 inci- Sensitivity analyses excluding cases diagnosed
dent cases were identified during 1992–2000 (Ros during the first 2 years after a postal survey (a total
et al., 2011). The relative risk for every 100  mL of 96 cases) yielded the same results. [A weakness
of coffee increase was 1.0 (95% CI,  0.98–1.03). of this study was the limited assessment of coffee
The relative risk for the highest level of coffee consumption with no quantity/serving, history
intake (≥  875  mL/day for men and ≥  500  mL/ of intake, or follow-up data provided.]
day for women) compared with the lowest level In a study that included 94 bladder cancer
(<  429  mL/day for men and <  250  mL/day for cases diagnosed within two Norwegian cohorts
women) was 1.11 (95% CI,  0.85–1.43, P for of men (Jacobsen et al., 1986), the relative risk
trend,  0.5). Sensitivity analyses excluding cases for the highest level of intake (>  7  cups/day)
diagnosed within 2 years of recruitment did not compared with the lowest (≤ 2 cups/day) was 0.98.
change results. Stratification of cases by high No confidence intervals were provided. Similar
(≥  T1, CIS, WHO grade 3) or low (Ta grade 1, estimates were obtained for women, although
Ta grade 2) risk of progression also yielded no adjustment for smoking was possible among
comparable results. [Limitations noted were: them. Excluding cases diagnosed in the first
stratified results by smoking were conducted 4 years of the cohorts yielded comparable results.
and mentioned but estimates not shown; a lack [Weaknesses of this study were the lack of assess-
of consideration of coffee-drinking history; and ment of coffee-drinking history, no follow-up
no follow-up data on coffee drinking.] data regarding coffee, and no stratification of
In the Health Professionals Follow-up Study results by smoking. Even though decaffeinated
(HPFS) (Michaud et al., 1999) 252 incident cases coffee or instant coffee were not assessed, it was
of bladder cancer were identified during 1986– indicated that these were rarely consumed at the
1996. The relative risk for the highest level of time of the study.]
caffeinated coffee intake (≥ 4 cups/day) compared In the Norwegian National Health Screening
with the lowest (< 1 cup/month) was 0.79 (95% Service for cardiovascular disease (Stensvold &
CI, 0.48–1.30), with no evidence of dose–response Jacobsen, 1994) a total of 53 incident cases of
and trend (P for trend, 0.56). Similarly, for decaf- cancer of the bladder (40 men and 13 women)
feinated coffee the relative risk for the highest level were identified. Among men the relative risk
of coffee (≥ 4 cups/day) compared with the lowest per 2  cups/day increase in coffee drinking was
(<  1  cup/month) was 0.83 (95% CI,  0.41–1.66), [1.13 (95% CI,  0.87–1.49)]; among women the
with no evidence of dose–response and trend corresponding relative risk was [1.22 (95%
(P for trend, 0.47). Sensitivity analyses excluding CI,  0.73–2.05)] [the paper reports coefficients
cases diagnosed during the first 3  years of the for these estimates, which were exponentiated
study did not change findings. [A weakness here]. Analyses using tertiles of coffee intake
was the lack of consideration of coffee-drinking are presented without confidence intervals.
history.] Sensitivity analyses for the first 2 years of diag-
In the Life Span Study of atomic bomb noses in cohort were performed. [A main weak-
survivors in Japan (Nagano et al., 2000), 114 ness was the modest sample size, particularly for
incident cases of bladder cancer were identified women, and lack of consideration of duration of
between 1979 and 1983 (83 men and 31 women). coffee intake.]
The relative risk for the highest level of intake In the following, cohort studies that reported
(> 5 times/week) compared with never drinkers results for coffee intake but were given less weight
was 0.90 (95% CI,  0.52–1.56), with no evidence by the Working Group are described.

85
IARC MONOGRAPHS – 116

A study that combined data from the Miyagi There was no evidence of a dose–response rela-
Cohort Study and the Ohsaki Cohort Study in tionship. A previous study reported on a subset
Japan, including 272 bladder cancer cases, was of these men (Nomura et al., 1986). [A limitation
reported (Sugiyama et al., 2017). The relative risk of this study was the fact that coffee intake was
for the highest consumption level (≥ 3 cups/day) assessed via 24 hour recalls, which may not be
compared with never drinkers was 0.56 (95% representative of long-term coffee drinking. The
CI,  0.32–0.99; P for trend,  0.04). When strati- numbers of cases in lower-intake categories were
fying individuals by smoking status, the relative very small.]
risks for the same comparisons were 0.62 (95% A total of 52 bladder cancer cases were iden-
CI,  0.14–2.72) for never smokers and 0.61 (95% tified within the Seventh-day Adventist Church
CI,  0.32–1.17) for former or current smokers, Cohort study conducted in California (Mills
with a test of interaction P  =  0.99. Interaction et al., 1991). The relative risk for the highest
analyses were also performed for sex, age, body level of intake (≥  2  cups/day) compared with
mass index (BMI), diabetes, and alcohol; no never drinkers was 1.99 (95% CI, 0.91–4.34; P for
evidence of effect modification was obtained for trend, 0.13) with little evidence of a dose–response
any of these variables. [The number of cases was relation. Analyses stratifying by smoking status
small for stratified analyses, especially among showed that the relative risk for the highest cate-
never smokers.] gory (≥ 1 cup/day) compared with never drinkers
In the Japan Public Health Center-based was 2.03 (95% CI,  0.70–5.87) among never
Prospective (JPHC) study 206 (164 men and 42 smokers and 1.14 (95% CI,  0.46–2.80) among
women) bladder cancer cases were identified past or current smokers. [Key limitations were
(Kurahashi et al., 2009). Among men, the hazard overall small numbers (especially among never
ratio for the highest category of coffee intake smokers with only 25 cases), an unclear defini-
(≥ 3 cups/day) compared with those who consumed tion of smoking variables in regression, and a
almost no coffee was 1.37 (95% CI, 0.75–2.51; P concern for potential underreporting of tobacco
for trend, 0.09). [No evidence of a dose–response consumption.]
relationship was observed.] Among women the In the Iowa Women’s Health Study 112 inci-
hazard ratio for the highest category of intake dent bladder cancer cases were identified between
(≥  1  cup/day) compared with almost none was 1986 and 1998 among postmenopausal women
0.55 (95% CI, 0.23–1.33; P for trend, 0.23). Among (Tripathi et al., 2002). The relative risk for the
never smoking men, the hazard ratio for the highest highest frequency of coffee intake (≥ 4 times/day)
category (≥ 1 cup/day) compared with almost no compared with the lowest (never or <  1  time/
coffee drinking was 2.48 (95% CI, 0.88–7.05), 2.09 month) was 1.59 (95% CI,  0.95–2.68). [Since it
(95% CI, 0.96–4.54) among former smokers, and was not clear whether smoking was included as
1.13 (95% CI, 0.65–1.97) among current smokers. a confounder, the Working Group decided not to
A test of interaction was not statistically significant. include this study for final evaluation.]
[The main weaknesses were the modest sample
size among never smokers, and the lack of coffee- 2.1.2 Case–control studies
drinking history and follow-up exposure data.]
In a prospective study conducted in Hawaii, 96 See Table 2.2
men with bladder cancer were identified (Chyou The Working Group identified 64 case–
et al., 1993). The relative risk for high (≥ 5 cups/ control studies that reported on associations
week) compared with low (≤ 2 cups/week) intake between coffee intake and risk of cancer of the
was 2.07 (95% CI, 0.84–5.12; P for trend, 0.174). bladder. In reviewing the literature, the Working

86
Table 2.2 Case–control studies on cancer of the bladder and drinking coffee

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Cole (1971) Cases: 470 population- Urinary bladder: Coffee intake among men (cups/day) Age, cigarette Also presented
USA based, pathology logs TCC and SCC < 1 29 1.00 smoking (cigarettes analyses stratified by
(Massachusetts), of hospitals in the area 1 86 1.34 (NR) smoked/day), age and sex, although
1966–1968 were used occupation numbers were very
2–3 146 1.18 (NR)
Controls: 500 small; among men
population-based using ≥ 4 84 1.31 (NR) the association was
residents lists, matched ≥ 1 vs < 1 316 1.24 (0.80–1.93) stronger for older men,
to cases by sex and year Coffee intake among women (cups/day) and among women
of birth < 1 9 1.00 it was stronger for
Exposure assessment 1 19 1.60 (NR) women aged 60–74 yr
method: Strengths: population-
2–3 50 3.76 (NR)
interviewer- based, adequate sample
administered ≥ 4 22 2.19 (NR) size, consideration of
questionnaire, ≥ 1 vs < 1 100 2.58 (1.30–5.10) occupational exposures
frequency and amount Coffee intake among non-smokers without high- Limitations: no
of coffee, unclear risk occupations (cups/day) information on
validation < 1 10 1.00 drinking history
1 31 2.18 (NR) or types of coffee
consumed, no
2–3 37 1.84 (NR)
confidence intervals
≥ 4 12 2.60 (NR) shown for RR in dose–
response analyses,
small numbers for
some stratified analyses
Fraumeni et al. Cases: 493; NR see Urinary bladder Daily consumption coffee (cups/day) Age, cigarette Strengths: both white
(1971) Dunham et al. (1968) Any amount vs NR 1.50 smoking and black subjects
USA (New Controls: 527; NR see none Limitations: no
Orleans), Dunham et al. (1968) Daily consumption of coffee among white men confidence intervals
1958–1964 Exposure assessment (cups/day) shown for most
method: estimates
0 (reference) 5 1.00

Drinking coffee
Questionnaire; see
Dunham et al. (1968) 1–2 85 1.40 (NR)
3–4 76 1.96 (NR)
≥ 5 99 1.66 (NR)
87
88

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Fraumeni et al. Daily consumption of coffee among black men
(1971) (cups/day)
(cont.) 0 (reference) 6 1.00
1–2 23 2.13 (NR)
3–4 27 2.90 (NR)
≥ 5 13 2.10 (NR)
Daily consumption of coffee (cups/day)
Any daily 323 1.95 (NR)
amount vs none
(all men)
Any daily 260 1.78 (NR)
amount vs none
(white men)
Any daily 63 2.10 (NR)
amount vs none
(black men)
Daily consumption of coffee among white women
(cups/day)
0 (reference) 14 1.00
1–2 45 0.70 (NR)
3–4 29 0.47 (NR)
≥ 5 24 0.32 (NR)
Trend test P value, 0.04
Daily consumption of coffee among black women
(cups/day)
0 (reference) 2 1.00
1–2 27 10.00 (NR)
3–4 10 4.58 (NR)
≥ 5 8 2.30 (NR)
Trend test P value, 0.04
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Fraumeni et al. Daily consumption of coffee among all women  
(1971) (cups/day)
(cont.) Any daily 147 1.19 (NR)
amount vs none
(all women)
Daily amount 98 0.51 (NR)
vs none (white
women)
Daily amount 45 5.65 (NR)
vs none (black
women)
Trend test P value, 0.04
Daily consumption coffee (cups/day)  
Never smokers NR 1.00
(blacks)
Ever smokers NR 3.56 (NR)
(blacks)
Never smokers NR 1.00
(whites)
Ever smokers NR 0.67 (NR)
(whites)

Drinking coffee
89
90

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Simon et al. Cases: 135 hospital- Urinary bladder Coffee consumption among non- and light smokers None Strengths: Assessed
(1975) based (cups/day) coffee drinking
USA Controls: 390 hospital- 0 to < 1 9 1.0 strength and history
(Massachusetts, based, identified via ≥ 1 76 1.7 (0.8–3.5) Limitations: hospital-
Rhode Island), discharge lists of same based, controls not
Coffee consumption among moderate to heavy
1965–1971 hospitals as cases, excluded based on
smokers (cups/day)
free of urinary tract non-urinary tract
problems (no selection 0 to < 1 1 1.0 disease that may also
made related to other ≥ 1 45 3.7 (0.6–23.6) affect coffee drinking
diseases) (GI diseases), small
Exposure assessment numbers in stratified
method: analyses, no estimates
mailed questionnaire, provided adjusting for
validation unclear smoking, women only
(both regular and
decaffeinated coffee
considered)
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Mettlin & Cases: 569 hospital- Urinary bladder: Coffee consumption among men (cups/day) Cigarettes smoked/ Same patient
Graham (1979) based ICD-188 < 1 24 1.00 day population as
USA (Buffalo, Controls: 1025 hospital- 1 56 1.38 (NR) described by Bross &
New York), based, admitted to Tidings (1973)
2 73 1.16 (NR)
1957–1965 same hospital as cases Strengths: adequate
with non-neoplastic 3 76 2.11 (NR) sample size with large
complaints, no > 3 124 1.64 (NR) number of controls
matching performed Urinary bladder Coffee consumption among women (cups/day) Limitations: hospital-
Exposure assessment < 1 15 1.00 based, no drinking
method: 1 25 0.83 (NR) history, controls may
questionnaire, include patients with
2 34 1.03 (NR)
validation unclear, disorders that affect
administered in person, 3 13 1.25 (NR) coffee drinking,
frequency/amount of > 3 24 0.81 (NR) number of women
coffee Coffee consumption among men and women Sex, cigarettes for smoking stratified
(cups/day) smoked/day analyses was small (not
< 1 39 1.00 presented here)
1 81 1.15 (NR)
2 107 1.11 (NR)
3 89 1.82 (NR)
> 3 148 1.30 (NR)

Drinking coffee
91
92

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Mettlin & Coffee consumption among light-smoking Cigarettes smoked/
Graham (1979) (< half a pack/day) men (cups/day) day
(cont.) < 1 16 1.00
1 28 1.28 (0.58–2.82)
2 34 0.98 (0.46–2.09)
3 30 2.18 (0.97–4.93)
> 3 26 1.40 (0.62–3.15)
Coffee consumption among light-smoking
(< half a pack/day) women (cups/day)
< 1 14 1.00
1 24 0.80 (0.33–1.93)
2 30 0.93 (0.40–2.19)
3 12 1.17 (0.40–3.47)
> 3 15 0.66 (0.25–1.74)
Wynder & Cases: 732 hospital- Urinary bladder Coffee (cups/day) Smoking Strengths: adequate
Goldsmith based, from 17 hospitals None/ NR 1.0 numbers, includes
(1977) in New York (majority), occasionally cases from various
USA (various Houston, Los Angeles, 1–3 NR 1.4 (0.8–2.3) regions of the USA
states), Miami, Birmingham, Limitations: controls
4–6 NR 1.9 (1.0–3.6)
1969–1974 New Orleans, Virginia may include patients
Controls: 732 hospital- ≥ 7 NR 2.0 (0.8–4.9) with diseases that
based, patients without affect coffee intake, few
history of tobacco- details of statistical
related conditions, analyses, no history
matched to cases by sex, of coffee drinking
race, hospital status, age considered
at diagnosis
Exposure assessment
method:
questionnaire, in-
person interview,
frequency and amount
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Howe et al. Cases: 632 population- Urinary bladder Lifetime average total coffee for men Cigarettes smoked, Strengths: coffee
(1980) based, identified (cups/day) smoking status, drinking history and
Canada through cancer Never drinker NR 1.0 lifetime pipe coffee types, included
(Nova Scotia, registries 1–2 NR [1.6 (1.0–2.6)] use, inhales pipe men and women,
Newfoundland, Controls: 632 smoke heavily, comprehensive
3–4 NR [1.3 (0.7–2.3)]
British population-based, occupation, use of consideration of
Columbia), neighbourhood > 4 NR [1.5 (0.8–2.8)] non-public water confounders
1974–1976 controls, matched to supply, bladder Limitations: small
cases by age and sex infection, diabetes, numbers for stratified
Exposure assessment education, aspirin, analyses
method: artificial sweetener
interviewer- Lifetime average total coffee for women Cigarettes smoked,
administered (cups/day) smoking status,
questionnaire Never drinker NR 1.0 lifetime pipe
1–2 NR [0.7 (0.3–1.5)] use, inhales pipe
smoke heavily,
3–4 NR [1.7 (0.6–4.8)]
occupation, use of
> 4 NR [1.3 (0.4–4.1)] non-public water
supply, bladder
infection, kidney
infection, diabetes
Lifetime average instant coffee for men Cigarettes smoked,
(cups/day) smoking status,
Never drinker NR 1.0 lifetime pipe
1–2 NR [1.5 (1.0–2.3)] use, inhales pipe
smoke heavily,
3–4 NR [1.7 (0.9–3.3)]
occupation, use
> 4 NR [1.5 (0.7–3.1)] of non-public
water supply,
bladder infection,

Drinking coffee
education, aspirin,
artificial sweetener,
regular coffee
93
94

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Howe et al. Lifetime average instant coffee for women Cigarettes smoked,
(1980) (cups/day) smoking status,
(cont.) Never drinker NR 1.0 lifetime pipe
1–2 NR [1.1 (0.5–2.5)] use, inhales pipe
smoke heavily,
3–4 NR [1.2 (0.3–5.1)]
occupation, use of
> 4 NR [1.2 (0.2–5.5)] non-public water
supply, bladder
infection, kidney
infection, diabetes,
regular coffee
Lifetime average regular coffee for men Cigarettes smoked,
(cups/day) smoking status,
Never drinker NR 1.0 lifetime pipe
1–2 NR [2.0 (1.1–3.4)] use, inhales pipe
smoke heavily,
3–4 NR [1.5 (0.8–2.7)]
occupation, use
> 4 NR [1.8 (1.0–3.5)] of non-public
water supply,
bladder infection,
education, aspirin,
artificial sweetener,
instant coffee
Lifetime average regular coffee for women Cigarettes smoked,
(cups/day) smoking status,
Never drinker NR 1.0 lifetime pipe
1–2 NR [0.4 (0.2–8.0)] use, inhales pipe
smoke heavily,
3–4 NR [0.7 (0.2–14)]
occupation, use of
> 4 NR [0.7 (0.2–16)] non-public water
supply, bladder
infection, kidney
infection, diabetes,
instant coffee
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
  Lifetime average instant coffee for non-smoking NR
women (cups/day)
≤ 2 NR 1.0
> 2 NR 1.4 (0.4–4.4)
Morrison et al. Cases: 1666 population- Urinary bladder Coffee (cups/day): all studies combined Age, sex, study Strengths: large sample
(1982) based, identified < 1 514 1.0 area, cigarette size, comprehensive
USA through hospitals > 1 903 1.0 (0.8–1.2) smoking exposure assessment,
(Boston), UK Controls: 2229 consideration of
Coffee (cups/day): Boston study, men only
(Manchester), population-based, occupational exposure
Japan (Nagoya), randomly identified, < 1 23 1.0 and other confounders
1976–1978 matched to cases by age 1 98 0.8 (NR) Limitations: not all
and sex 2 95 0.7 (NR) analyses are shown, no
Exposure assessment 3 82 0.9 (NR) confidence intervals
method: 4 41 0.8 (NR) provided for most of
questionnaire the estimates
5 19 0.8 (NR)
(no information
about validation) ≥ 6 65 1.5 (NR)
administered in person Coffee (cups/day): Boston study, women only
< 1 20 1.0
1 59 0.8 (NR)
2 38 0.6 (NR)
3 19 1.7 (NR)
4 12 0.9 (NR)
5 10 0.7 (NR)
≥ 6 7 1 (NR)
Coffee (cups/day): Manchester study, men only
< 1 224 1.0
1 85 1.1 (NR)

Drinking coffee
2 40 0.9 (NR)
3–4 27 0.9 (NR)
≥ 5 12 0.8 (NR)
95
96

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Morrison et al. Coffee (cups/day): Manchester study, women only
(1982) < 1 79 1.0
(cont.) 1 46 1.4 (NR)
2 8 0.4 (NR)
3–4 14 1.2 (NR)
≥5 5 1 (NR)
Coffee (cups/day): Nagoya study, men only
< 1 116 1.0
1 43 1.0
2 38 1.2
3–4 20 1.3
≥5 7 1.9
Coffee (cups/day): Nagoya study, women only
< 1 52 1.0
1 11 0.7
> 2 2 0.7
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Hartge et al. Cases: 2982 population- Urinary bladder Coffee drinking history Sex, age, race, Strengths: large
(1983) based, identified Never drinker 98 1.0 geographic sample size, thorough
USA (10 through SEER cancer Ever drinker 2809 1.4 (1.1–1.8) area, tobacco confounding
geographical registries history (based on assessment, years
Men: never 58 1.0
regions), Controls: 5782 cigarettes/day and of coffee drinking
drinker
1977–1978 population-based, smoking status) assessed
identified through RDD Men: ever 2139 1.6 (1.2–2.2) Limitations: modest
or Medicare records, drinker numbers in some
frequency matched Women: never 40 1.0 stratified analyses,
to cases on age, sex, drinker small number in
and geographical Women: ever 670 1.2 (0.8–1.7) reference group
distribution drinker
Exposure assessment Coffee consumption (cups/wk) among men
method: ≤ 7 397 1.0
interviewer-
7.1–14 389 0.9 (0.8–1.1)
administered
questionnaire, different 41.1–21 381 1.0 (0.8–1.2)
types of coffee and 21.1–35 493 1.1 (0.9–1.3)
frequency of drinking 35.1–49 195 1.0 (0.8–1.3)
assessed 49.1–63 109 1.2 (0.9–1.6)
63.1–155 148 1.5 (1.1–1.9)
Coffee consumption (cups/wk) among women
≤7 164 1.0
7.1–14 161 0.9 (0.7–1.2)
41.1–21 110 0.8 (0.6–1.1)
21.1–35 133 0.9 (0.7–1.2)
35.1–49 49 0.7 (0.5–1.1)
49.1–63 21 0.9 (0.5–1.7)
63.1–155 26 0.8 (0.4–1.4)

Drinking coffee
97
98

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Hartge et al. Coffee drinking status by smoking status among
(1983) men
(cont.) Non-smokers NR –
Never drinkers 159 1.0
Ever drinkers NR 1.5 (0.9–2.5)
Past smokers NR –
Never drinkers 62 1.0
Ever drinkers NR 1.4 (0.8–2.6)
Smokers NR –
Never drinkers 56 1.0
Ever drinkers NR 2.1 (1.2–3.9)
Coffee drinking high/low by smoking status among
men
Non-smokers NR –
≤ 49 cups/wk NR 1.0
Ever drinkers 21 4.2 (1.7–10.0)
(> 49 cups/wk)
Past smokers NR –
≤ 49 cups/wk NR 1.0
Ever drinkers 208 1.3 (1–1.8)
(> 49 cups/wk)
Smokers NR –
≤ 49 cups/wk NR 1.0
Ever drinkers 302 1.2 (1–1.6)
(> 49 cups/wk)
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Hartge et al. Coffee drinking status by smoking status among Sex, age, race,
(1983) women geographical area,
(cont.) Non-smokers NR – amount of tobacco
Never drinkers 121 1.0
Ever drinkers NR 0.9 (0.6–1.5)
Past smokers NR –
Never drinkers 13 1.0
Ever drinkers NR 3.0 (0.8–12.0)
Smokers NR –
Never drinkers 27 1.0
Ever drinkers NR 1.3 (0.6–2.9)
Coffee drinking high/low by smoking status among
women
Non-smokers NR –
≤ 49 cups/wk NR 1.0
Ever drinkers 24 0.4 (0.2–1.5)
(> 49 cups/wk)
Past smokers NR –
≤ 49 cups/wk NR 1.0
Ever drinkers 25 1.7 (0.7–4.2)
(> 49 cups/wk)
Smokers NR –
≤ 49 cups/wk NR 1.0
Ever drinkers 67 1 (0.6–1.7)
(> 49 cups/wk)

Drinking coffee
99
100

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Sturgeon et al. Cases: 1860; see Hartge Urinary bladder: Coffee consumption (cups/wk) by tumour grade Age, sex, cigarette Same study as Hartge
(1994) et al. (1983) TCC Grade I, 326 1.0 use (status and et al. (1983)
USA (10 Controls: 3934; see consumption cigarettes/day), Strengths: large
geographical Hartge et al. (1983) < 50 history of urinary sample size, thorough
regions), Exposure assessment Grade I, 49 1.3 (0.9–1.8) infections, history confounding
1977–1978 method: consumption of bladder stones, assessment, years
questionnaire; see ≥ 50 artificial sweetener, of coffee drinking
Hartge et al. (1983) family history assessed, very
Grade II, 578 1.0
of urinary tract comprehensive and
consumption
cancer, high-risk thorough analyses
< 50
occupation, race, Limitations: modest
Grade II, 87 1.3 (1.0–1.7) education numbers for stage
consumption and grade combined
≥ 50 analyses
Grade III/IV, 562 1.0
consumption
< 50
Grade III/IV, 61 1.4
consumption
≥ 50
RR for coffee (cups/wk) by tumour stage
Non-invasive, 983 1.0
consumption
< 50
Non-invasive, 147 1.4 (1.1–1.7)
consumption
≥ 50
Invasive overall, 522 1.0
consumption
< 50
Invasive overall, 68 1.2 (0.9–1.6)
consumption
≥ 50
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Sturgeon et al. RR for coffee (cups/wk) by tumour grade and stage
(1994) Non-invasive 668 1.0
(cont.) low grade,
consumption
< 50
Non-invasive 109 1.4 (1.1–1.8)
low grade,
consumption
≥ 50
Non-invasive 156 1.0
high grade,
consumption
< 50
Non-invasive 15 1.3 (0.7–2.2)
high grade,
consumption
≥ 50
Invasive 197 1.0
low grade,
consumption
< 50
Invasive 23 1.0 (0.6–1.5)
low grade,
consumption
≥ 50
Invasive 293 1.0
high grade,
consumption
< 50

Drinking coffee
Invasive 43 1.4 (1.0–2.0)
high grade,
consumption
≥ 50
101
102

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Kantor et al. Cases: 2915; see Hartge Urinary bladder: Coffee consumption (cups/wk) Sex, age, cigarette Strengths:
(1988) et al. (1983) SCC 0–7 9 1.0 smoking consideration of
USA (10 Controls: 5782; see 8–21 12 0.9 (0.3–2.2) tumour subtypes, large
geographical Hartge et al. (1983) sample size for TCC
22–49 13 1.4 (0.5–3.5)
regions), Exposure assessment Limitations: very small
1977–1978 method: 50–63 3 2.1 (0.4–10.8) numbers for SCC and
questionnaire; see ≥ 64 2 1.1 (0.1–6.6) adenocarcinoma, small
Hartge et al. (1983) Urinary bladder: Coffee consumption (cups/wk) number in reference
adenocarcinomas 0–7 5 1.0 group (never drinkers)
8–21 13 2.1 (0.7–6.9)
22–49 11 2.8 (0.8–9.5)
50–63 1 2.7 (0.1–48.7)
≥ 64 2 5.2 (0.5–58.1)
Trend test P value, 0.049
Urinary bladder: RR for coffee (cups/wk)
TCC 0–7 625 1.0
8–21 932 1.0 (0.9–1.1)
22–49 761 1.1 (0.9–1.2)
50–63 110 1.4 (1.0–1.8)
≥ 64 153 1.5 (1.1–1.9)
Trend test P value, < 0.01
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Rebelakos et al. Cases: 300 hospital- Urinary bladder: Coffee consumption (cups/day): men and women Age, sex, smoking Strengths: proper
(1985) based 93% TCC 0 25 1.0 status adjustment
Greece (Athens), Controls: 300 hospital- 1 62 1.2 (0.8–2.2) Limitations: moderate
1980–1982 based, different size, but too small
2 150 1.7 (1.0–2.8)
hospitals from cases for stratified analyses
(majority traumatic 3 36 2.7 (0.9–8.2) by sex (few women),
fractures or conditions) ≥ 4 24 0.7 (0.2–2.7) no consideration of
Exposure assessment > 2 vs < 2 210 1.7 (1.2–2.3) drinking history,
method: (including 0) no mention of other
interviewer- Coffee consumption (cups/day): men confounders other
administered 0 15 1.0 than age and smoking,
questionnaire (no different types of coffee
1 41 1.1 (0.5–2.3)
information about not specified
validation), amount and 2 133 1.5 (0.8–2.7)
duration recorded 3 32 4.0 (1.2–13.4)
≥ 4 22 0.5 (0.1–2.5)
> 2 vs < 2 187 1.7 (1.2–2.4)
(including 0)
  Coffee consumption (cups/day): women
0 10 1.0
1 21 2.0 (0.9–5.0)
2 15 2.1 (0.9–5.0)
3 0 0.0 (0.0–0.0)
≥ 4 2 2.0 (0.2–23.6)
> 2 vs < 2 17 1.6 (0.8–3.2)
(including 0)

Drinking coffee
103
104

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Claude et al. Cases: 431 hospital- Urinary bladder Consumption of ground coffee (cups/day): men Smoking Strengths: adequate
(1986) based 0 NR 1.00 numbers
Germany Controls: 431 hospital- 1–2 NR 1.42 (0.70–2.80) Limitations: hospital-
(Lower Saxony), based, identified in based, possible bias due
3–4 NR 1.39 (0.70–2.60)
1977–1982 urology ward and for to selection of hospital-
older individuals from > 4 NR 2.29 (0.40–11.60) based controls with
elderly homes in town, Drinker vs NR 1.57 [(0.60–3.80)] urological diseases,
matched to cases 1:1 by non-drinker duration of intake not
age and sex Consumption of ground coffee (cups/day): women considered
Exposure assessment 0 NR 1.00
method: 1–2 NR 1.26 (0.80–2.00)
questionnaire
3–4 NR 1.89 (0.50–6.60)
administered in person,
frequency of intake > 4 NR 2.18 (0.50–10.00)
recorded, different Drinker vs NR 0.99 [(1.00–1.00)]
types (ground, regular, non-drinker
decaffeinated) of coffee
considered
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Kunze et al. Cases: 675 hospital- Urinary bladder: Coffee consumption (cups/day): women Smoking status, Extension of a study
(1992) based lower urinary 0 NR 1.0 pack-years reported by Claude
Germany Controls: 675 hospital- tract cancers, 1–2 47 1.4 (0.7–3.0) et al. (1986), so includes
(Lower Saxony), based, identified in majority bladder patients reported in
3–4 60 2.4 (1.0–5.4)
1977–1985 urology ward, matched but also others this previous study
by age and sex (64% of ≥ 5 24 2.7 (0.9–7.8) Strengths: adequate
men had hyperplasia Coffee consumption (cups/day): men numbers
of the prostate, 73% 0 NR 1.0 Limitations: hospital-
women had lower 1–2 168 1.3 (0.8–2.0) based, possible bias due
urinary infections) 3–4 205 1.5 (0.95–2.3) to selection of hospital-
Exposure assessment based controls with
≥ 5 102 2.0 (1.2–3.3)
method: urological diseases,
questionnaire no history of coffee
administered in person, drinking considered
frequency of intake
recorded, different
types (ground, regular,
decaffeinated) of coffee
considered

Drinking coffee
105
106

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Jensen et al. Cases: 371 population- Urinary bladder: Coffee consumption Age, smoking Strengths: adequate
(1986) based majority TCC Men per L/day NR [1.1 (0.9–1.4)] status, lifetime sample size,
Denmark Controls: 771 Women per NR [1.1 (0.7–1.9)] cigarette exposure comprehensive
(Copenhagen), population-based, L/day (pack-years) questionnaire
1979–1981 matched to cases by sex, Limitations: no
Coffee consumption (mL/day): men
age, and residential area information on
Exposure assessment 0 15 1.0 validation of
method: 1–499 69 0.9 (0.5–1.9) questionnaire, modest
questionnaire (0–2 cups) numbers for reference
administered in person, 500–999 90 0.8 (0.4–1.6) category for stratified
frequency of intake (2–4 cups) analyses by sex
recorded, different 1000–1499 56 0.9 (0.4–1.8)
types (ground, regular, (4–6 cups)
decaffeinated, instant) ≥ 1500 50 1 (0.5–2.1)
of coffee considered, (> 6 cups)
drinking history and
Trend test P value, 0.83
amount
Coffee consumption (mL/day): women
0 4 1.0
1–499 20 1.9 (0.6–6.7)
(0–2 cups)
500–999 33 1.2 (0.4–3.5)
(2–4 cups)
1000–1499 15 1.6 (0.4–6.0)
(4–6 cups)
≥1500 (> 6 cups) 13 2.7 (0.7–10.9)
Trend test P value, 0.37
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Kabat et al. Cases: 152; see Wynder Urinary bladder Brewed coffee consumption (cups/day): men None Strengths: focus
(1986) & Goldsmith (1977) None/occasional 40 1.00 on non-smokers
USA (various Controls: 492; see 1–2 18 0.91 (0.48–1.71) (important given the
states), 1976– Wynder & Goldsmith strong confounding
3–4 15 1.38 (0.69–2.79)
1983 (1977) effect of smoking),
Exposure assessment 5–6 3 1.38 (0.34–5.59) large catchment area
method: ≥ 7 0 0.46 (0.03–8.47) across the USA
questionnaire; see Brewed coffee consumption (cups/day): women Limitations: hospital-
Wynder & Goldsmith None/occasional 40 1.00 based controls (which
(1977) 1–2 24 1.51 (0.84–2.72) may introduce bias if
they had diseases that
3–4 8 0.81 (0.35–1.88)
affect coffee intake),
5–6 2 0.66 (0.14–3.10) small numbers for
≥7 2 2.43 (0.41–14.34) some of the coffee
Decaffeinated coffee consumption (cups/day): men drinking categories
None/occasional 60 1.00
1–2 cups/day 14 1.07 (0.54–2.11)
3–4 cups/day 2 0.40 (0.09–1.71)
≥ 5 cups/day 0 0.27 (0.02–4.10)
Decaffeinated coffee consumption (cups/day):
women
None/occasional 62 1.00
1–2 9 0.50 (0.24–1.07)
3–4 5 0.73 (0.26–2.01)
≥5 0 0.58 (0.03–11.82)

Drinking coffee
107
108

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Piper et al. Cases: 165 population- Urinary bladder Coffee consumption (cup-years) among women Race, level of Strengths: population-
(1986) based, identified Non-drinker NR 1.0 education, smoking based, adequate control
USA (New through cancer registry 1–50 NR 0.9 (0.5–2.3) (pack-years), for confounders
York), Controls: 165 phenacetin drugs Limitations: narrow
51–100 NR 1.9 (0.8–4.6)
1975–1980 population-based, use, bladder focus on young women,
identified through RDD, ≥ 101 NR 2.1 (0.7–6.3) infection, thyroid no history of coffee
paired to cases by strata uptake procedure drinking studied
defined by age and
residence
Exposure assessment
method:
telephone questionnaire,
no information about
validation, regular
coffee only
Iscovich et al. Cases: 117 hospital- Urinary bladder Coffee consumption (cups/day) Age, average Strengths: case
(1987) based, 60% of registered 0 35 1.00 cigarettes smoked recruitment
Argentina (La cases for catchment area 1 24 1.08 (NR) comparable to a
Plata), Controls: 117 hospital- population-based study
2 16 4.45 (NR)
1983–1985 based (16% digestive Limitations: modest
system problems, 17% ≥3 24 12 (NR) numbers, use of
heart disease, 12% Trend test P value, < 0.05 hospital-based controls
hypertension), 2:1 ratio: that included disorders
one recruited from same that may affect coffee
hospitals, another from intake (thus leading to
the neighbourhood of potential biases that
the case may inflate ORs), no
Exposure assessment confidence intervals
method: presented, no history
in-person of coffee drinking
questionnaire, coffee considered
frequency and amount
considered
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Ciccone & Cases: 512 hospital- Urinary bladder Current consumption (cups/day): men Age, smoking Strengths: stratification
Vineis (1988) based Non-drinker 88 1.0 status, lifelong use by smoking
Italy (Torino), Controls: 594 hospital- 1 93 0.8 (0.5–1.3) of cigarettes, high- Limitations: hospital-
1978–1983 based, patients with risk occupations based (therefore
2 122 1.0 (0.7–1.5)
urological or surgical concern about bias
conditions (~20% 3 122 1.2 (0.8–1.8) introduced), very small
from ‘other surgical ≥4 87 0.8 (0.5–1.2) numbers for stratified
departments’; no other Consumption (cups/day) 10 yr before diagnosis: analyses
information provided), men
no information on Non-drinker 39 1.0
matching to cases 1 65 1.2 (0.7–2.1)
Exposure assessment
2 97 1.5 (0.9–2.5)
method:
in-person questionnaire 3 104 1.1 (0.7–1.8)
(unclear validation), ≥4 139 1.1 (0.6–1.8)
coffee history and Current consumption (cups/day): women Age, smoking
frequency of intake Non-drinkers 8 1.0 status, lifelong use
1 17 1.4 (0.5–3.8) of cigarettes
2 12 1.0 (0.4–3.0)
3 8 0.7 (0.2–2.2)
≥4 7 0.8 (0.2–2.6)
Consumption (cups/day) 10 yr before diagnosis:
women
0–1 16 1.0
2 13 0.9 (0.4–2.3)
3 8 0.5 (0.2–1.5)
≥4 15 1.4 (0.6–3.5)

Drinking coffee
109
110

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Ciccone & Current consumption (cups/day): non-smoking Age, high-risk
Vineis (1988) men occupations
(cont.) Non-drinker 3 1.0
1 6 1.1 (0.2–5.4)
2 5 1.9 (0.4–9.3)
3 5 4.4 (0.8–25.1)
Consumption (cups/day) 10 yr before diagnosis: Age, smoking,
non-smoking men high-risk
Non-drinker 2 1.0 occupations
1 4 1.6 (0.2–10.4)
2 5 2.7 (0.4–17)
≥ 3 5 4.9 (0.8–31.6)
Current consumption (cups/day): non-smoking Age
women
Non-drinker 7 1.0
1 11 1.1 (0.4–3.3)
2 7 0.9 (0.3–3.2)
≥ 3 6 0.5 (0.1–1.5)
Urinary bladder: Consumption (cups/day) 10 yr before diagnosis:
no information non-smoking women
provided on 0–1 12 1.0
histological types 2 6 0.9 (0.3–2.6)
3 5 0.7 (0.2–2.2)
≥ 4 8 1.5 (0.6–3.5)
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Risch et al. Cases: 835 population- Urinary bladder Ever coffee drinking of total coffee: men Lifetime smoking Strengths: large sample
(1988) based, identified Ever drinker NR 0.86 (0.59–1.25) history (pack- size, comprehensive
Canada through hospital Ever drinker, NR 1.69 (0.30–9.59) years), history of questionnaire,
(South Central registries or regional non-smokers diabetes consideration of non-
Ontario), tumour registry smokers, different
Ever drinker, NR 0.64 (0.38–1.06)
1979–1982 Controls: 781 types of coffee and
non-user
population-based lifetime use
of artificial
identified from Limitations: sample
sweetener
population listings, size not shown for
matched by sex, birth Ever coffee drinking of total coffee: women different strata in
year, area of residence Ever drinker NR 1.87 (1.03–3.40) analyses, no tests for
Exposure assessment Ever drinker, NR 2.05 (0.69–6.15) trend shown
method: non-smokers
questionnaire, in- Ever drinker, NR 2.55 (1.05–6.22)
person interview (no non-user
information about of artificial
validation), different sweetener
types (ground, instant, Average consumption of total coffee (cups/day):
instant decaffeinated, men
espresso) of coffee
None NR 1.00
considered, frequency
and lifetime use > 1–3 NR 1.04 (0.76–1.41)
considered > 3–6 NR 1.15 (0.82–1.62)
> 6 NR 0.91 (0.58–1.44)
Total lifetime NR 0.95 (0.85–1.06)
intake
Average consumption of total coffee (cups/day):
women
None NR 1.00
> 1–3 NR 0.96 (0.57–1.61)

Drinking coffee
> 3–6 NR 1.85 (0.98–3.50)
> 6 NR 1.11 (0.46–2.71)
Total lifetime NR 1.16 (0.88–1.53)
intake
111
112

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Risch et al. Ever drinker of ground coffee: men
(1988) Ever NR 1.02 (0.78–1.33)
(cont.) Total lifetime NR 0.95 (0.85–1.08)
intake
Ever drinker of ground coffee: women
Ever NR 1.15 (0.75–1.76)
Total lifetime NR 1.11 (0.83–1.48)
intake
Ever drinker of instant coffee: men
Ever NR 0.93 (0.74–1.18)
Total lifetime NR 0.94 (0.83–1.07)
intake
Ever drinker of instant coffee: women
Ever NR 0.97 (0.65–1.47)
Total lifetime NR 0.95 (0.73–1.25)
intake
Ever drinker of instant decaffeinated coffee: men
Ever NR 1.12 (0.83–1.51)
Total lifetime NR 0.91 (0.76–1.10)
intake
Ever drinker of instant decaffeinated coffee:
women
Ever NR 1.5 (0.90–2.52)
Total lifetime NR 1.2 (0.87–1.67)
intake
Ever drinker of espresso coffee: men
Ever NR 1.64 (0.96–2.79)
Total lifetime NR 1.29 (0.96–1.74)
intake
Ever drinker of espresso coffee: women
Ever NR 1.50 (0.59–3.78)
Total lifetime NR 1.75 (0.91–3.39)
intake
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Slattery et al. Cases: 332 population- Urinary bladder Caffeinated coffee (cups/wk) Smoking status Possible overlap with
(1988a) based, cases identified Never drinkers NR 1.00 (never, ex, current) Slattery et al. (1988b)
USA (Utah), through population- 1–15 NR 1.32 (0.88–2.00) Strengths: population-
1977–1983 based cancer registry based, cases identified
16–30 NR 0.80 (0.50–1.26)
Controls: 686 via registry
population-based, > 30 NR 1.28 (0.76–2.17) Limitations: very
identified through Caffeinated coffee (cups/wk): never smokers None unique population with
RDD or social security Never drinkers NR 1.00 majority of Mormons
administration roster 1–15 NR 1.42 (0.69–2.90) (distinctive coffee
(Medicare), frequency 16–30 NR 1.36 (0.55–3.35) drinking and smoking
matched by age and sex habits)
> 30 NR 1.50 (0.39–2.17)
Exposure assessment
method: Caffeinated coffee (cups/wk): smokers
questionnaire, in- Never drinkers NR 1.00
person survey, lifetime 1–15 NR 1.36 (0.88–2.10)
coffee (only caffeinated) 16–30 NR 0.88 (0.56–1.39)
> 30 NR 1.54 (0.98–2.44)

Drinking coffee
113
114

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Slattery et al. Cases: 419 population- Urinary bladder Consumption of caffeinated coffee, number of 8 oz Age, sex, diabetes, Strengths: population-
(1988b) based, identified servings (~1 cup)/wk bladder infections, based, cases identified
USA (Utah), through population- 0 164 1.00 cigarette smoking via registry
1977–1983 based cancer registry 1–20 99 1.23 (0.88–1.72) (smoking status, Limitations: very
Controls: 889 pack-years) unique population with
21–40 93 1.05 (0.73–1.51)
population-based, majority of Mormons
identified through > 40 58 1.60 (1.00–2.56) (distinctive coffee
RDD or social security Consumption of caffeinated coffee, number of 8 oz drinking and smoking
administration roster servings (~1 cup)/wk habits)
(Medicare), 2:1 ratio 0 354 1.00
of controls to cases, ≥ 1 62 1.04 (0.73–1.48)
frequency matched by
age, sex
Exposure assessment
method:
questionnaire, in-
person survey, lifetime
coffee (only caffeinated)
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Clavel & Cordier Cases: 781 hospital- Urinary bladder Average daily coffee consumption (cups/day): men Age, hospital, Strengths: large sample
(1991) based and women residence, smoking size, several types of
France Controls: 781 hospital- 0 12 1.00 status coffee studied
1984–1987 based controls, 1–4 488 1.24 (0.56–2.74) Limitations: hospital-
identified in same based, 21% of controls
5–7 61 1.46 (0.6–3.51)
hospitals as cases (non- had gastrointestinal
cancer, no symptoms > 7 27 2.94 (1.06–8.15) disease and 30% men
of bladder cancer), Average daily coffee consumption (cups/day): non- Age, hospital, with heart disease
matched by sex, age, smoking women residence problems (which may
place of residence 0 3 1.00 affect coffee intake),
Exposure assessment 1 7 1.00 sample sizes too small
method: 2 16 0.99 (0.34–2.93) for stratified analyses
in-person questionnaire, (too many cells with
3 13 1.51 (0.48–4.74)
different types (regular, number of subjects
instant, caffeinated, > 3 15 2.29 (0.59–8.86) < 10), no combined
decaffeinated) of coffee Average daily coffee consumption (cups/day): non- analyses shown
considered, history of smoking men
consumption (average 0 1 1.00
daily consumption since 1 3 0.97 (0.08–11.43)
age 18) 2 9 2.93 (0.31–30.35)
≥3 29 5.10 (0.59–43.86)

Drinking coffee
115
116

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
D’Avanzo et al. Cases: 555 hospital- Urinary bladder Regular coffee duration of drinking (yr), both sites Age, sex, education Same design as La
(1992) based from Milan and combined level, smoking Vecchia et al. (1989a)
Italy (Milan), Pordenone Non-drinkers 71 1.0 (status, cigarettes/ so probably some
1985–1990 Controls: 855 hospital- < 30 219 1.2 (0.9–1.7) day), alcohol, overlap of cases
based, recruited in occupation Strengths: validated
≥ 30 267 1.4 (0.9–2.2)
same hospitals as cases questionnaire,
(no urological and Trend test P value, < 0.05 consideration of
non-cancer patients, Regular coffee drinking status, both sites combined duration of drinking
no specific diseases Non-drinkers 71 1.0 Limitations: possible
excluded were listed) Drinkers 484 1.3 (1.0–1.8) bias introduced by
Exposure assessment Trend test P value, > 0.05 use of hospital-based
method: controls, many of
Regular coffee drinking frequency (cups/day), both
validated in-person whom may have had
sites combined
questionnaire, regular disease that affect
and decaffeinated coffee Non-drinkers 71 1.0 coffee intake
1 126 1.2 (0.8–1.7)
2 167 1.4 (0.9–2.0)
3 109 1.5 (1.0–2.2)
≥4 82 1.4 (0.9–2.2)
Trend test P value, > 0.05
Decaffeinated coffee drinking status, both sites
combined
Non-drinkers 519 1.0
Drinkers 39 1.5 (0.9–2.4)
Trend test P value, > 0.05
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Nomura et al. Cases: 261 population- Urinary bladder Coffee consumption, all types (cup-years): men Pack-years of Strengths: validated
(1991) based, identified at 7 Non-drinker 7 1.0 smoking and thorough coffee
USA (Hawaii), hospitals Drinker 188 0.8 (0.3–2.0) drinking assessment,
1977–1986 Controls: 522 including years of
1–49 34 0.6 (0.2–1.6)
population-based, consumption
identified from state 50–109 74 0.9 (0.4–2.3) Limitations: no
survey, matched to cases ≥ 110 80 1.0 (0.4–2.7) adjustment for race,
by sex, ethnic group, Trend test P value, 0.12 analyses of different
age, residence Coffee consumption, regular ground (cup-years): coffee types not
Exposure assessment men adjusted for each other
method: Non-drinker 10 1.0
validated in-person
Drinker 185 0.9 (0.4–2.0)
questionnaire,
frequency and quantity 1–39 46 0.9 (0.4–2.0)
of coffee (regular, 40–89 58 0.9 (0.4–2.1)
decaffeinated, brewed, ≥ 90 81 1.0 (0.4–2.3)
instant, and all Trend test P value, 0.72
combinations of these) Coffee consumption, instant (cup-years): men
considered
Non-drinker 106 1.0
Drinker 89 1.0 (0.7–1.4)
1–14 37 0.8 (0.5–1.3)
≥ 15 52 1.2 (0.8–1.9)
Trend test P value, 0.26
Coffee consumption, instant decaffeinated
(cup-years): men
Non-drinker 144 1.0
Drinker 51 1.3 (0.8–2.0)
1–4 26 1.5 (0.8–2.6)

Drinking coffee
≥ 5 25 1.1 (0.6–1.9)
Trend test P value, 0.82
117
118

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Nomura et al. Coffee consumption, all types (cup-years): women
(1991) Non-drinker 6 1.0
(cont.) Drinker 60 0.8 (0.3–2.6)
1–49 24 0.9 (0.3–2.9)
50–109 24 0.9 (0.2–2.9)
≥ 110 12 0.5 (0.5–2.1)
Trend test P value, 0.26
Coffee consumption, regular ground (cup-years):
women
Non-drinker 9 1.0
Drinker 57 0.7 (0.2–1.9)
1–39 20 0.7 (0.2–2.1)
40–89 29 0.8 (0.3–3.6)
≥ 90 8 0.3 (0.1–1.0)
Trend test P value, 0.02
Coffee consumption, instant (cup-years): women
Non-drinker 32 1.0
Drinker 34 1.8 (0.9–3.3)
1–14 22 1.8 (0.9–3.7)
≥ 15 12 1.6 (0.6–4.0)
Trend test P value, 0.46
Urinary bladder: Coffee consumption, instant decaffeinated
98% bladder (cup-years): women
cancer, 83% of Non-drinker 53 1.0
which were SCC Drinker 13 0.6 (0.3–1.2)
or TCC
1–4 7 0.5 (0.2–1.4)
≥5 6 0.6 (0.2–1.6)
Trend test P value, 0.3
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Escolar Pujolar Cases: 497 hospital- Urinary bladder Coffee consumption status and frequency Smoking Strengths: adequate
et al. (1993) based but with good (cups/wk): men (cigarettes/day), sample size,
Spain (Cadiz, population coverage, Non-drinker 34 1.00 occupation, comprehensive
Barcelona, 51% identified using (reference) consumption assessment of coffee
Madrid, registries Ex-drinker 42 1.22 (0.69–2.15) of artificial drinking (taking into
Guipuzkoa, Controls: 1113, sweeteners, account frequency,
Current drinker 362 0.96 (0.62–1.49)
Bizcaya), ~50% hospital-based age, province of amount, and duration),
1983–1986 (excluding urological, Drinker 404 0.98 (0.64–1.52) residence stratification by
diabetes, heart or 2–7 138 0.99 (0.63–1.57) smoking
circulatory, cancer of 8–14 130 0.95 (0.59–1.51) Limitations: use of
respiratory or upper ≥ 15 135 1.02 (64.0–1.63) hospital-based controls
gastrointestinal tract), Coffee consumption status and frequency Smoking status, may introduce bias,
matched for sex, age, (cups/wk): women consumption very small numbers for
province of residence; of artificial stratified analyses by
Non-drinker 5 1.00
other ~50% were sweeteners, smoking
(reference)
population-based age, province of
controls identified from Ex-drinker 6 0.87 (0.20–3.77)
residence
electoral rolls Current drinker 48 0.98 (0.31–3.14)
Exposure assessment 2–7 17 1.02 (0.29–3.58)
method: 8–14 24 1.14 (0.34–3.85)
in-person questionnaire, ≥ 15 13 0.71 (0.20–2.56)
unclear validation,
Coffee lifelong consumption in cups (thousands):
coffee (regular, instant,
women
decaffeinated) history/
frequency considered 0 3 1.00
1–10 6 1.47 (0.29–7.58)
10–20 13 1.80 (0.41–7.90)
20–30 9 2.03 (0.43–9.70)
30–40 10 1.47 (0.31–6.89)
≥ 40 12 1.39 (0.31–6.25)

Drinking coffee
119
120

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Escolar Pujolar Coffee consumption status and frequency Smoking
et al. (1993) (cups/wk): non-smoking men (cigarettes/day),
(cont.) Non-drinker 3 1.00 occupation,
Ex-drinker 1 0.61 (0.06–6.26) consumption
of artificial
Current drinker 24 2.78 (0.78–9.87)
sweeteners,
Drinker 25 2.41 (0.68–8.46) age, province of
2–7 10 2.22 (0.57–8.66) residence
8–14 10 3.11 (0.79–12.27)
≥ 15 5 1.87 (0.41–8.47)
Coffee lifelong consumption in cups (thousands):
men
0 cups 28 1.00
1–10 70 1.09 (0.63–1.87)
10–20 86 0.91 (0.54–1.54)
20–30 69 1.11 (0.65–1.90)
30–40 52 0.99 (0.56–1.74)
≥ 40 128 1.14 (0.69–1.90)
Coffee lifelong consumption in cups (thousands):
non-smoking men
0 cups 3 1.00
1–10 5 1.74 (0.38–7.95)
10–20 6 2.42 (0.55–10.66)
20–30 5 2.67 (0.57–12.45)
30–40 4 3.67 (0.70–19.25)
≥ 40 5 2.08 (0.44–9.86)
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Vena et al. Cases: 351 hospital- Urinary bladder: Coffee consumption (cups/day) Age, education, Strengths: adequate
(1993) based, recruited at most TCC 0–1 60 1.0 cigarette smoking sample size, use of
USA (west New hospitals in the area 2 62 1.3 (0.8–2.0) (pack-years), other population-based
York), 1979– (Buffalo, Niagara Falls, liquids, sodium, controls, hospital-
3–4 114 1.6 (1.1–2.3)
1985 Rochester) carotene, calories based cases with
Controls: 855 ≥5 115 2.1 (1.3–3.2) ample catchment
population-based Trend test P value, < 0.001 area (comparable to
neighbourhood controls Coffee consumption (cups/day) for non-smokers population-based
in same counties as aged > 65 yr cases)
cases 0–1 NR 1.0 Limitations: no history
Exposure assessment 2 NR 2.3 of coffee consumption
method: recorded, patients
3–4 NR 3.3
in-person questionnaire, too ill to participate
validated for some of ≥5 NR 6.4 or deceased were
the factors via telephone Trend test P value, 0.02 not included, many
recalls, coffee (regular, controls declined to
decaffeinated, instant, participate because the
perk) frequency only survey was too long 

Drinking coffee
121
122

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Vena et al. Coffee consumption (cups/day) by coffee type: men Age, education
(1993) 0–1, any type 60 1.0
(cont.) 2–4 25 1.8 (1.0–3.2)
decaffeinated,
instant
≥5 2 0.4 (0.9–1.8)
decaffeinated,
instant
2–4 8 1.0 (0.5–2.4)
decaffeinated,
perk
≥5 7 2.8 (1.0–7.8)
decaffeinated,
perk
2–4 regular, 29 1.5 (0.9–2.5)
instant
≥ 5 regular, 19 1.6 (0.9–3.0)
instant
2–4 regular, 114 1.5 (1.0–2.1)
perk
≥ 5 regular, perk 87 2.5 (1.7–3.8)
Coffee consumption (cups/day) among those aged Age, education,
< 65 yr cigarette smoking
0–1 NR 1.0 pack-years, other
2 NR 1.3 (0.7–2.7) liquids, sodium,
carotene, calories
3–4 NR 1.4 (0.7–2.6)
≥5 NR 1.9 (1.0–3.7)
Trend test P value, 0.03
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Vena et al. Coffee consumption (cups/day) among those aged
(1993) > 65 yr
(cont.) 0–1 NR 1.0
2 NR 1.3 (0.7–2.2)
3–4 NR 1.7 (1.0–2.8)
≥5 NR 2.2 (1.2–4.1)
Trend test P value, < 0.01
Coffee consumption (cups/day) among non-
smokers aged < 65 yr
0–1 NR 1.0
2 NR 0.6
3–4 NR 1.0
≥5 NR 1.6
Trend test P value, 0.08
Momas et al. Cases: 219 population- Urinary bladder Lifelong coffee drinking (cups) Lifelong tobacco Strengths: population-
(1994) based, identified via < 365 8 1.0 smoking (cigarettes based study,
France (Herault cancer registry 365–25 000 36 1.6 (0.6–3.8) equivalent), spice consideration of coffee
district), Controls: 792 consumption, duration
25 001–60 000 59 1.6 (0.6–3.8)
1987–1989 population-based age, occupation, Limitations: very small
selected via electoral > 60 000 58 4.1 (1.7–10.0) residence, vegetable numbers for reference
rolls consumption, category used, only
Exposure assessment lifelong alcohol considered lifelong
method: drinking, coffee intake (not
in-person or mailed birthplace, frequency)
questionnaire, duration saccharin
and changes in coffee
intake

Drinking coffee
123
124

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Bruemmer et al. Cases: 262 population- Urinary bladder: Coffee consumption (cups/day): women Age, county, Pack-years was
(1997) based cases identified invasive or non- None 11 1.0 smoking status not found to be a
USA via cancer registry invasive (in situ ≤ 3 21 0.5 (0.2–1.2) (never, former, confounder, so it was
(Washington), (SEER) or papillary) current) not added
> 3–6 20 0.5 (0.5–1.3)
1987–1990 Controls: 405 Strengths: population-
population-based, > 6 8 0.6 (0.2–1.9) based, consideration of
identified via RDD Trend test P value, 0.46 decaffeinated
Exposure assessment Coffee consumption (cups/day): men Limitations: modest
method: None 24 1.0 numbers (especially for
telephone interview ≤ 3 50 1.1 (0.5–2.1) women), no consider
questionnaire, of duration of intake or
> 3–6 77 1.7 (0.9–3.4)
coffee (regular, amounts, participants
decaffeinated) > 6 51 1.2 (0.6–2.3) < 65 yr
frequency and amount Trend test P value, 0.38
of intake considered Decaffeinated coffee consumption: women
only ≤ 1 cup/mo 39 1.0
> 1 cup/mo – 1 12 1.6 (0.7–3.6)
cup/wk
> 7 cups/wk 9 2.1 (0.8–5.3)
Trend test P value, 0.08
Decaffeinated coffee consumption: men
≤ 1 cup/mo 148 1.0
> 1 cup/mo – 1 31 1.4 (0.8–2.6)
cup/wk
> 7 cups/wk 23 0.9 (0.5–1.8)
Trend test P value, 0.85
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Donato et al. Cases: 172 hospital- Urinary bladder Coffee consumption status and frequency: women Age, residence, Strengths: good
(1997) based Non-drinker 2 1.0 education, date of representation of
Italy (Brescia), Controls: 578 hospital- Ex-drinker 0 – interview, smoking underlying case
1991–1992 based identified from (lifetime cigarettes population
Current drinker 35 5.2 (1.0–30.4)
three hospitals (prostate smoked), alcohol Limitations: hospital-
adenoma, urolithiasis, 1–2 cups/day 27 4.3 (0.8–23.9) based controls (not
obstructive uropathy), 3–4 cups/day 8 4.9 (0.7–33.0) clear if the diseases
male controls were age- Coffee consumption status and frequency: men included may affect
matched (not possible Non-drinker 7 1.0 coffee intake), very
for women) Ex-drinker 6 2.7 (0.7–10.3) tiny numbers for
Exposure assessment stratified analyses by
Current drinker 122 2.6 (1.1–6.1)
method: sex (women)
in-person validated 1–2 cups/day 66 2.3 (0.9–5.6)
questionnaire, coffee 3–4 cups/day 44 2.8 (1.1–7.4)
quantity and frequency ≥ 5 cups/day 11 4.5 (1.2–16.8)
considered Trend test P value, > 0.1
Pohlabeln et al. Cases: 300 hospital- Urinary bladder Coffee amount: men and women Smoking status and With more thorough
(1999) based Heavy NR 1.52 (0.39–5.93) pack-years adjustment for
Germany Controls: 300 hospital- consumption smoking, weaker
(Hesse), based (identified from Coffee frequency (cups/day): men ORs were observed
1989–1992 same hospitals as cases), for coffee intake.
≤1 53 1.00
matched to cases on sex, Restricting analyses to
age, area of residence 2–4 128 1.51 (0.95–2.39) urinary bladder yielded
Exposure assessment ≥5 58 1.59 (0.87–2.91) similar results.
method: Coffee frequency (cups/day): women Strengths: thorough
questionnaire, in- ≤1 11 1.00 adjustment by
person interview, coffee 2–4 40 1.28 (0.50–3.31) smoking, stratification
frequency and amount by smoking
≥5 10 1.25 (0.29–5.30)
considered Limitations: hospital-
Coffee frequency (cups/day): non-smokers NR based (therefore

Drinking coffee
≤1 8 1.00 concerns about
2–4 15 2.29 (0.82–6.36) potential bias), very
≥5 1 0.69 (0.07–6.86) small number of
women
125
126

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Geoffroy-Perez Cases: 765 hospital- Urinary bladder Frequency of coffee intake (mL/wk): men Age, centre, place Strengths: large sample
& Cordier based ≤ 1050 83 1.00 of residence, size, duration of
(2001) Controls: 765 hospital- 1051–2050 116 1.45 (0.97–2.16) smoking status, drinking was taking
France, based (free of cancer, pack-years into account
2051–2400 133 1.54 (1.04–2.28)
1984–1987 respiratory diseases, Limitations: concern
and bladder cancer 2401–2800 127 1.62 (1.08–2.40) about controls
symptoms), matched > 2800 134 1.42 (0.94–2.14) with disease that
to cases based on Trend test P value, 0.14 may affect coffee
hospital, sex, age, area Frequency of coffee intake (mL/wk): women Age, centre, place intake (GI diseases,
of residence ≤ 1150 20 1.00 of residence, cardiovascular)
Exposure assessment smoking status
1151–2100 38 1.40 (0.63–3.12)
method:
questionnaire, in person 2101–2600 28 1.25 (0.53–2.98)
interview, drinking > 2600 19 0.74 (0.28–1.96)
history, frequency and Trend test P value, 0.63
amounts Frequency of coffee intake (mL/wk): non-smoking Age, centre, place
women of residence
≤ 1100 13 1.00
1101–2100 25 1.67 (0.66–4.21)
2101–2550 9 1.11 (0.35–3.51)
> 2550 19 1.28 (0.45–3.63)
Trend test P value, 0.69
Frequency of coffee intake (mL/wk): non-smoking
men
≤ 1050 7 1.00
1051–2050 8 1.41 (0.43–4.65)
2051–2600 28 3.78 (1.36–10.47)
> 2600 11 2.49 (0.73–8.49)
Trend test P value, 0.02
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Woolcott et al. Cases: 927 population- Urinary bladder: Coffee frequency (cups/day) for all individuals Age, sex, education Strengths: population-
(2002) based, identified via ICD-9 188 < 1 150 1.00 level, smoking based, large sample size
Canada, registry 1–2 320 1.03 (0.81–1.32) (ever, current, Limitations: controls
1992–1994 Controls: 2494 hospital- cumulative, were matched to other
3–4 278 0.88 (0.68–1.13)
based, identified intensity), energy cancer cases, few cases
through RDD, ≥ 5 165 1.06 (0.79–1.42) intake, calcium, were non-smokers
frequency matched Trend test P value, 0.76 fibre, beer
to cases on age and Coffee frequency (cups/day): never smokers
sex distribution of the < 1 NR 1.00
combined case series 1–2 NR 1.46 (0.91–2.35)
(bladder, colon, rectum)
3–4 NR 1.25 (0.73–2.13)
Exposure assessment
method: ≥ 5 NR 1.84 (0.80–4.22)
mailed questionnaire, Trend test P value, 0.23
coffee (brewed, iced) Coffee frequency (cups/day): ever smokers
considered < 1 NR 1.00
1–2 NR 0.90 (0.67–1.20)
3–4 NR 0.77 (0.58–1.03)
≥ 5 NR 0.92 (0.66–1.27)
Trend test P value, 0.39

Drinking coffee
127
128

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Radosavljević Cases: 130 hospital- Urinary bladder: Coffee intake Smoking soda, Limitations: hospital-
et al. (2003) based 93% TCC   NR 1.46 (1.05–2.01) spirit, mineral based, concern about
Serbia, 1997– Controls: 130 hospital- water, skim milk, controls; units of coffee
1999 based (no urological yogurt, frequency intake not clear
malignancies or of daily urination
diseases that change Coffee intake Smoking
diet), same hospital   NR 1.55 (1.24–1.94)
as cases, matched 1:1
by sex, age, place of
residence
Exposure assessment
method:
FFQ, unclear validation
and administration,
patterns of consumption
and changes in diet
in the past 10 yr
considered
Ugnat et al. Cases: 549 population- Urinary bladder Coffee consumption Age, province, Strengths: population-
(2004) based controls identified < 1 cup/mo 34 1.00 education, pack- based, adequate sample
Canada (British as part of a larger ≥ 1 cup/mo – 89 1.13 (0.69–1.83) years of smoking, Limitations: no
Columbia, population-based study ≤ 1 cup/day tea consideration of
Alberta, (NECSS) duration of intake of
2–3 cups/day 214 1.56 (0.99–2.46)
Saskatchewan, Controls: 1099 coffee, not clear if test
Manitoba), population-based ≥ 4 cups/day 210 1.77 (1.11–2.82) of trend corresponds to
1994–1997 matched to cases by Trend test P value, 0.0001 adjusted or unadjusted
distribution of age, Coffee frequency (cups/day): non-smokers NR model
identified randomly < 4 NR 1.00
from health insurance > 4 NR 6.17 (1.73–21.96)
plan lists or RDD
Exposure assessment
method:
mailed questionnaire,
unclear validation
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Wakai et al. Cases: 124 hospital- Urinary bladder: Coffee consumption (cups/day) Age, sex, year of Less than 3% of cases
(2004) based cases identified 90% TCC Almost never 26 1.00 first visit, pack- drank high levels of
Japan (Nagoya), from database of Occasionally 23 0.93 (0.52–1.66) years cigarette coffee
1994–2000 outpatients smoking Limitations: hospital-
1 28 0.82 (0.47–1.44)
Controls: 620 hospital- based, (therefore
based, randomly 2 26 1.07 (0.59–1.94) potential for bias
selected from ≥3 21 1.14 (0.58–2.23) among controls
outpatients in database Trend test P value, 0.68 depending on cause
without cancer, matched of outpatient visit), no
by age, sex, year of visit lifetime consumption
Exposure assessment of coffee considered,
method: few confounders
self-administered considered
questionnaire but
checked by interviewer,
frequency of coffee
intake
De Stefani et al. Cases: 255 hospital- Urinary bladder: Coffee with milk (cups/wk) Age, sex, residence, Some overlap in
(2007) based TCC Never drinkers 135 1.0 urban/rural status, patients between this
Uruguay, Controls: 501 hospital- 1–6 70 1.5 (1–2.2) family history of study and that by Balbi
1996–2000 based (excluding bladder cancer, et al. (2001)
≥7 24 1.9 (1–3.7)
diseases related to BMI, occupation, Limitations: data
tobacco, alcohol or Trend test P value, 0.01 smoking status, regarding drinking
recent changes in Pure coffee consumption (cups/wk) years since quitting history mentioned but
diet), identified at Never drinkers 135 1.0 smoking, number not provided
same hospital as cases, 1–6 22 1.6 (0.8–3.1) of cigarettes
frequency matched by ≥7 15 2.0 (0.9–4.4) smoked per day,
age, sex, and residence mate, soft drinks,
Trend test P value, 0.03
Exposure assessment milk, tea
method: Total coffee consumption (cups/wk)

Drinking coffee
in-person questionnaire, Never drinkers 135 1.0
coffee drinking history 1–6 84 1.5 (1.1–2.2)
considered ≥7 36 2.1 (1.2–3.6)
Trend test P value, < 0.01
129
130

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Covolo et al. Cases: 197 hospital- Urinary bladder Coffee consumption (cups/day) Age, education, Genotype data also
(2008) based Non-drinkers 26 1.00 PAHs and collected: GSTM1,
Italy (Brescia), Controls: 211 hospital- 1–3 125 0.76 (0.41–1.41) AA exposure, GSTT1, GSTP1, NAT1,
1997–2000 based, identified at cumulative lifetime NAT2, SULTIA1,
> 3 77 1.25 (0.59–2.67)
same hospital as cases smoking (pack- XRCC1–3, XPD.
(patients with urological Coffee consumption (cups/day): heavy smokers years) Combined estimates of
non-neoplastic Non-drinkers 12 1.00 genotypes and coffee
diseases), frequency 1–3 86 1.45 (0.56–3.70) were presented, but no
matched to cases on age, > 3 27 1.46 (0.49–4.36) tests of interaction.
period of recruitment, Coffee consumption (cups/day): non-smokers Strengths: Lifetime
and hospital history of coffee use
Non-drinkers 5 1.00
Exposure assessment Limitations: Hospital-
method: 1–3 10 0.42 (0.01–1.77) based controls
in-person > 3 2 0.35 (0.04–2.99) (therefore concern
questionnaire, coffee Coffee consumption (cups/day): light smokers about possible bias
(with milk, cappuccino, Non-drinkers 9 1.00 introduced by changes
decaffeinated) lifetime 1–3 29 0.47 (0.16–1.35) in coffee consumption),
consumption very small numbers in
> 3 17 3.04 (0.77–11.97)
stratified analyses by
smoking, very modest
sample size for GxE
interaction analyses
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Jiang et al. Cases: 1586 population- Urinary bladder Coffee consumption (cups/day) Level of education, Strengths: population-
(2008) based, identified via 0 129 1.00 use of NSAIDs, based, large sample size
USA (Los cancer registry (SEER) < 1 49 1.15 (0.71–1.85) intake of Limitations: no
Angeles), Controls: 1586 carotenoids, years long-term history of
1–2 501 1.04 (0.78–1.38)
1987–1999 population-based, as hairdresser/ consumption of coffee,
identified via 3–4 467 1.21 (0.89–1.64) barber, cigarette only recent (2 yr before
neighbourhoods of 5–6 226 1.19 (0.95–1.68) smoking status, diagnosis)
cases, matched to cases ≥ 7 210 1.38 (0.95–2.00) duration of
by age, sex, and race Trend test P value, 0.052 smoking, intensity
Exposure assessment of smoking, age,
method: sex, race
in-person questionnaire,
both regular and
decaffeinated
coffee considered
Villanueva et al. Cases: 1219 hospital- Urinary bladder Coffee consumption (cups/day) Age, sex, Strengths: large
(2009) based Never 120 1.00 area, intensity sample size, large
Spain Controls: 1271 hospital- Ever 1016 1.25 (0.95–1.64) of smoking representation of
(Barcelona, based, identified from (cigarettes/day) hospitals in this area,
1 336 1.24 (0.92–1.66)
Valles/Bages, same hospitals as cases coffee drinking history
Alicante, (disease unrelated to 2 303 1.11 (0.82–1.51) Limitations: hospital-
Tenerife, bladder cancer risk 3 223 1.57 (1.13–2.19) based controls could
Asturias), factors), individually ≥4 154 1.27 (0.88–1.81) induce bias if they
1998–2001 matched to cases by sex, Trend test P value, 0.082 altered coffee drinking
age and residence due to disease (does not
Coffee consumption (cups/day): current smokers
Exposure assessment seem likely in this case)
method: Never 46 1.00
questionnaire, Ever 468 1.20 (0.72–2.01)
computer-assisted 1 130 1.14 (0.65–2.00)
interview, coffee 2 143 1.20 (0.68–2.09)

Drinking coffee
assessment included 3 105 1.39 (0.77–2.53)
age started and stopped
≥4 90 1.13 (0.61–2.09)
drinking, and average
intake per day during Trend test P value, 0.559
adult life
131
132

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Villanueva et al. Coffee consumption (cups/day): former smokers
(2009) Never 34 1.00
(cont.) Ever 423 1.85 (1.16–2.95)
1 152 1.92 (1.16–3.17)
2 128 1.62 (0.97–2.70)
3 94 2.36 (1.36–4.11)
≥4 49 1.57 (0.86–2.90)
Trend test P value, 0.176
Coffee consumption (cups/day): never smokers
Never 40 1.00
Ever 125 0.85 (0.53–1.35)
1 54 0.91 (0.53–1.56)
2 32 0.61 (0.34–1.10)
3 24 1.06 (0.53–2.13)
≥4 15 1.23 (0.55–2.76)
Trend test P value, 0.961
Wang et al. Cases: 1007 hospital- Urinary bladder: Frequency of all coffee intake (servings/day) Age, sex, ethnicity, Assessed
(2013a) based TCC Never 155 1.00 energy intake, polymorphisms in
USA (Houston, Controls: 1299 clinic- 0.1–1.9 271 1.13 (0.87–1.47) smoking status UGT enzymes
Texas), 1999– based, identified at Strengths: large sample
≥2 581 1.14 (0.90–1.46)
ongoing clinics in the area for size
annual health check-ups Trend test P value, 0.336 Limitations: no lifetime
Exposure assessment Frequency of regular coffee intake (servings/day) history of coffee
method: Never 288 1.00 assessed
in-person questionnaire, 0.1–1.9 235 0.91 (0.72–1.15)
coffee (regular, ≥2 484 0.92 (0.74–1.13)
decaffeinated)
Trend test P value, 0.426
frequency and amount
Frequency of decaffeinated coffee intake (servings/
day)
Never 717 1.00
0.1–1.9 94 1.75 (1.28–2.41)
≥2 196 1.37 (1.09–1.73)
Trend test P value, 0.001
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Turati et al. Cases: 690 hospital- Urinary bladder Average lifetime coffee drinking (cups/day) Age, sex, study Strengths: thorough
(2015) based 0 to < 1 57 1.00 centre, year of exposure assessment
Italy (Aviano, Controls: 655 hospital- 1 to < 2 142 1.30 (0.83–2.03) interview, smoking Limitations: use
Pordenone, based (with acute, (status and cigs/ of hospital-based
2 to < 3 166 0.90 (0.58–1.38)
Milan, Naples, non-neoplastic diseases day among current controls, although
Catania), unrelated to smoking 3 to < 4 149 1.16 (0.74–1.82) smokers) it is noted that most
2003–2014 and alcohol or long- ≥4 176 1.73 (1.08–2.77) diseases among
term diet changes) 1 cup/day NR 1.06 (0.99–1.14) controls seem
identified from same increase unrelated to coffee
network of hospitals as Trend test P value, 0.049 intake
cases, matched by study Coffee drinking status
centre, sex, and age
Never 30 1.00
Exposure assessment
method: Ex 42 1.21 (0.61–2.40)
in-person questionnaire, Current 618 1.25 (0.74–2.10)
coffee (regular, Lifetime coffee drinking (1 cup/day increase) by Age, sex, study
cappuccino, age centre, year
decaffeinated) frequency Age < 65 yr NR 1.09 (0.99–1.21) of interview,
of consumption, age at Age > 65 yr NR 1.05 (0.95–1.16) tobacco smoking,
starting and quitting, education, alcohol,
Lifetime coffee drinking (1 cup/day increase) by sex
changes in drinking BMI, family history
during life, and average Men NR 1.05 (0.98–1.14) of bladder cancer,
lifetime coffee drinking Women NR 1.14 (0.90–1.45) history of cystitis
estimated Lifetime coffee drinking (1 cup/day increase) by
smoking status
Never smokers NR 1.18 (0.96–1.46)
Ex-smokers NR 1.07 (0.97–1.19)
Current NR 1.03 (0.92–1.15)
smokers

Drinking coffee
133
134

IARC MONOGRAPHS – 116
Table 2.2 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Turati et al. Duration of coffee drinking (yr) Age, sex, study
(2015) ≤ 35 146 1.00 centre, year
(cont.) 36–44 172 1.13 (0.79–1.63) of interview,
smoking (status
45–51 174 1.17 (0.79–1.72)
and cigarettes/day
≥ 52 185 1.20 (0.80–1.79) among current
10-yr increase NR 1.03 (0.95–1.13) smokers)
Coffee drinking frequency (cups/day)
0 to < 1 99 1.00
1 to < 2 128 1.13 (0.77–1.68)
2 to < 3 161 0.86 (0.60–1.24)
3 to < 4 146 1.15 (0.78–1.69)
≥4 156 1.28 (0.85–1.94)
1 cup/day NR 1.03 (0.96–1.10)
increase
Trend test P value, 0.305
Age at starting drinking (yr)
< 20 267 1.00
≥ 20 380 1 (0.78–1.28)
AA, aromatic amines; BMI, body mass index; CI, confidence interval; FFQ, food frequency questionnaire; GI, gastrointestinal; ICD, International Classification of Disease; mo,
month(s); NECSS, Canadian National Enhanced Cancer Surveillance System; NR, not reported; NSAID, nonsteroidal anti-inflammatory drugs; OR, odds ratio; PAH, polycyclic
aromatic hydrocarbons; RDD, random-digit dialling; RR, relative risk; SCC, squamous cell carcinoma; SEER, Surveillance, Epidemiology and End Results; TCC, transitional cell
carcinoma; UGT, UDP-glucuronosyltransferase; vs, versus; wk, week(s); yr, year(s)
Drinking coffee

Group considered the following criteria when Based on the criteria described above, of the
determining which studies would be informative 64 studies identified: seven studies were excluded
for evaluation of the association between risk of due to having a case sample size of < 100 (Sullivan,
bladder cancer and coffee consumption. 1982; Mommsen et al., 1983a; González et al.,
1. Sample size, which impacts statistical power. 1985; Restrepo et al., 1989; Bento & Barros, 1997;
As there were a large number of studies Lu et al., 1999; Kobeissi et al., 2013); one study was
published on this topic, the Working Group excluded because no potential confounders were
focussed its review on studies had a minimum considered (Demirel et al., 2008); four studies
of 100 cases. were excluded because risk estimates were not
reported (Morgan & Jain, 1974; Mommsen et al.,
2. Case and control selection: hospital-based 1983b; Wynder et al., 1985; Akdaş et al., 1990);
versus population-based control selection. one study was excluded because smoking was
Depending on the inclusion criteria for not adjusted for (Bravo et al., 1986); one study
hospital controls, these individuals may was excluded because no units were provided for
have diseases that could potentially lead the estimates of association (Boada et al., 2015);
to modification in coffee intake, making and five studies were excluded because they
them less representative of the underlying included cases and controls already included in
population to which the cases should be other studies (Bross & Tidings, 1973; Mettlin &
compared, and therefore result in selection Graham, 1979; Marrett et al., 1983; Ohno et al.,
biases. In particular, studies that included 1985; La Vecchia et al., 1989a).
hospital-based controls with gastrointes- The Working Group organized studies for
tinal diseases and cardiovascular disorders discussion into four main groups defined in
were considered potentially problematic. The Sections  2.1.2 (a)–(d). Given that studies with
Working Group considered whether studies larger sample sizes are likely to be more inform-
had specifically listed which diseases were ative, larger studies are described first followed
included among hospital-based controls, or by studies with smaller sample sizes.
provided some indication that diseases that
may affect coffee intake had been excluded. (a) Population-based studies
The Working Group gave more weight to
The population-based case–control studies
population-based studies.
that reported results for coffee intake and were
3. Adjustment for potential confounding considered informative by the Working Group
factors, in particular tobacco smoking. are described in the following. These studies
Given that smoking is a strong risk factor for were given more weight in the evaluation than
bladder cancer and tends to be highly corre- those described in Section 2.1.1 (b)–(d).
lated with coffee intake in many populations, Hartge et al. (1983) conducted a study in the
the Working Group considered only studies USA (2982 cases, 5782 controls) that reported
that evaluated smoking variables as possible a positive association between ever drinking
confounders. Although adjustment for other coffee and risk of bladder cancer among men
confounders was also favourably considered (OR, 1.6; 95% CI, 1.2–2.2), women (OR, 1.2; 95%
and noted (e.g. occupational exposure), none CI,  0.8–1.7) and for both combined (OR,  1.4;
of the other risk factors were deemed as 95% CI,  1.1–1.8). When various levels of coffee
important as tobacco smoking. consumption were considered, the only statisti-
cally significant association was for men drinking
over 63 cups of coffee per week (OR,  1.5; 95%

135
IARC MONOGRAPHS – 116

CI,  1.1–1.9) [equivalent to roughly 9  cups/day]. controls), and Nagoya, Japan (289 cases, 586
No dose–response relationship was evident for controls) for a total of 1666 cases and 2229
either men or women. Similarly, there was no controls. On pooling the three studies, there was
association with duration of coffee drinking. no association for drinking ≥  1 cup/day versus
When stratifying men by smoking status, no less (OR, 1.0; 95% CI, 0.8–1.2). Results stratifying
differences in the magnitude of estimates were by study area did not show consistent evidence of
observed when comparing ever drinkers to never a dose–response relationship [confidence inter-
drinkers. However, when comparing drinkers vals for estimates were not reported for any of the
of large quantities to drinkers of smaller quan- study-specific results].
tities (≤  49 cups/week), a stronger significant Jiang et al. (2008) conducted a study in Los
positive association was observed among never Angeles County, California, USA (1586 cases,
smokers [numbers of subjects were smaller and 1586 controls). They reported a positive associ-
confidence intervals very wide], whereas positive ation for heavy coffee drinking (≥  7  cups/day)
associations of smaller magnitude were observed versus non-drinkers with an odds ratio of 1.38
among past or current smokers. Results were less (95% CI, 0.95–2.00). There was weak evidence of
pronounced among women, and none of the esti- a dose–response relationship (P for trend, 0.052).
mates was statistically significant. A subsequent [The limitations included a lack of consider-
study by Kantor et al. (1988) reported estimates ation of coffee-drinking history; only coffee
by subtyping cases by three histological types; consumption from 2 years before diagnosis was
significant trends for positive associations with considered.]
risk of adenocarcinomas or transitional cell A population-based study was performed in
carcinomas (TCC) for men and women combined Ontario, Canada (Woolcott et al., 2002) involving
were reported, although only the estimate for the 927 cases and 2494 controls. No associations
highest intake (>  64  cups/week) versus lowest were noted when considering all individuals
(0–7  cups/week) was statistically significant for combined; positive associations were however
TCC (OR, 1.5; 95% CI, 1.1–1.9; P for trend, < 0.01 observed among never smokers, although the
[numbers for adenocarcinomas were extremely estimates were not statistically significant and
low (32 cases)]. There was no evidence of trend or there was no consistent dose–response trend. No
statistically significant point estimates for squa- evidence of positive associations was observed
mous cell carcinomas [numbers were too small among ever smokers. [The limitations of this
to interpret]. Another extension of the study by study include the fact that controls were recruited
Sturgeon et al. (1994) considered subtypes of cases for multiple cancers and matching for bladder
defined by tumour stage and grade. Positive asso- cancer might not be optimal. Further, only 15%
ciations of similar magnitude were observed for of cases were non-smokers (n = 139), which limits
high versus low intake of coffee among non-inva- power for smoking-stratified analyses.]
sive and invasive bladder cancer, as well as when Risch et al. (1988) (835 cases, 781 controls)
stratifying cases by grade (I, II, or III/IV). Even reported that ever drinkers of coffee had an odds
though some of the estimates were statistically ratio of 0.86 (95% CI,  0.59–1.25) in men and
significant in some strata and not others, all esti- 1.87 (95% CI,  1.03–3.4) in women. Restricting
mates were of comparable magnitude. analyses to non-smokers yielded positive associ-
Morrison et al. (1982) reported a study that ations for both men and women, but neither was
combined data from three population-based statistically significant. Analyses that considered
case–control studies in Boston, USA (587 cases, several categories of frequency of coffee intake
528 controls), Manchester, UK (541 cases, 725 showed little evidence for a dose–response

136
Drinking coffee

relationship or association for men or women. with evidence of a dose–response relation-
Similarly, estimates were close to null when shop. [It is unclear from the publication if the
considering ground, decaffeinated, or instant test for trend corresponds to the unadjusted or
coffee. For total lifetime intake or ever intake of adjusted estimates.] It is mentioned in the text
espresso coffee, positive associations were noted that a positive association was found among
for both men and women; neither reached statis- non-smokers (≥  4 cups/day vs <  1 cup/month:
tical significance, however, with a lifetime intake OR,  6.17; 95% CI,  1.73–21.96). [The number of
odds ratio of 1.29 (95% CI, 0.96–1.74) for men and cases in these analyses was not reported. Further,
1.75 (95% CI, 0.91–3.39) for women. [No tests for the low response rates of cases and controls raise
trend were presented. Smoking adjustment only some concern about possible bias introduced by
included pack-years of smoking, raising concerns responders. Only pack-years for smoking adjust-
about residual confounding.] ment were considered, raising concern about
Howe et al. (1980) reported on a study based residual confounding.]
in Nova Scotia, Newfoundland, and British Cole (1971) (470 cases, 500 controls)
Columbia in Canada, involving 632 cases and conducted a population-based case–control
632 controls. A non-statistically significant posi- study in Massachusetts, USA. Positive associa-
tive association between the highest level of life- tions between coffee intake and risk of bladder
time average consumption (> 4 cups/day) of total cancer were reported (> 4 cups/day vs < 1 cup/day:
coffee and risk of bladder cancer when compared OR, 1.31 for men and 2.19 for women) [no confi-
with never drinkers was reported (OR, 1.5; 95% dence intervals or a test for trend were presented],
CI, 0.8–2.8 for men and OR, 1.3; 95% CI, 0.4–4.1 with weak evidence for a dose–response trend.
for women). No tests of trend were presented, and The odds ratio for drinking > 1 cup/day versus
there was no evidence of a dose–response relation- < 1 cup/day was 1.24 (95% CI, 0.8–1.93) among
ship. Separate risk estimates are also presented men and 2.58 (95% CI, 1.30–5.10) among women.
for instant coffee and regular coffee, for men and When restricting analyses to non-smokers
women individually. A positive association was without high-risk occupational exposure and
reported for regular coffee for men (>  4 cups/ comparing the highest intake (> 4 cups/day) to
day vs never drinkers: OR, 1.8; 95% CI, 1.0–3.5), the lowest (<  1  cup/day), an odds ratio of 2.6
but there was weak evidence of a dose–response for men and women combined was reported
relationship. Analyses restricted to non-smokers [no confidence intervals were provided, and the
were conducted only among women and an odds reference category comprised only 10 cases].
ratio of 1.4 (95% CI, 0.4–4.4) was reported for a Jensen et al. (1986) (371 cases, 771 controls)
lifetime average of > 2 cups/day compared with conducted a population-based case–control
≤  2 cups/days. [Numbers were very small for study in Copenhagen, Denmark and reported
some of the cells in stratified analyses. All odds no association between coffee intake and risk
ratios and confidence intervals were estimated by of bladder cancer; per L/day of coffee intake,
the Working Group.] odds ratios were 1.1 (95% CI,  0.9–1.4) for men
Ugnat et al. (2004) (549 cases, 1099 controls) and 1.1 (95% CI,  0.7–1.9) for women. Analyses
conducted a population-based case–control considering quintiles showed estimates close to
study in Canada (British Columbia, Alberta, 1 for men, with no evidence of dose–response
Saskatchewan, and Manitoba provinces) and or trend (P for trend, 0.83). In contrast, positive
reported a positive association with high intake associations were reported among women for all
of coffee (≥  4 cups/day vs <  1 cup/month: categories in comparison to never drinkers with
OR, 1.77; 95% CI, 1.11–2.82; P for trend, < 0.001), an odds ratio of 2.7 (95% CI,  0.7–10.9) for the

137
IARC MONOGRAPHS – 116

highest category (>  1500 mL/day or >  6 cups), the highest category of regular coffee intake
but there was no evidence of a dose–response (>  6  cups/day) with non-drinkers was 1.2 (95%
relationship and the trend was not statistically CI,  0.6–2.3) for men and 0.6 (95% CI,  0.2–1.9)
significant (P for trend, 0.37). [It was noted that among women. There was no evidence of a dose–
the reference category for this analysis among response relationship and no statistically signif-
women had only 4 cases and the highest cate- icant trends. When considering decaffeinated
gory had only 13 cases.] No differences in age at coffee, the comparable odds ratios were 0.9 (95%
which coffee drinking started or in duration of CI, 0.5–1.8) for men and 2.1 (95% CI, 0.8–5.3) for
coffee drinking were observed between cases and women [there were only 9 cases in the highest
controls, and changes over time of the quantity intake category]. There was no evidence of a
of coffee consumed were similar for both cases trend among men; there was however a sugges-
and controls; however, no estimates were shown. tion of a trend among women with an estimate
Slattery et al. (1988a) reported the results of 1.6 (95% CI, 0.7–3.6; P for trend, 0.08) for the
of a population-based case–control study middle category. [The Working Group noted that
conducted in Utah, USA (332 cases, 686 this study only included men and women of age
controls). A non-statistically significant posi- up to 65 years and the models only adjusted for
tive association with caffeinated coffee (>  30 smoking status, raising concerns over residual
cups/week vs 1–15  cups/week OR,  1.28; 95% confounding.]
CI, 0.76–2.17) was reported, without evidence Nomura et al. (1991) reported on a study
of a dose–response relationship. Different conducted in Hawaii, USA (261 cases, 522
models adjusting for different smoking vari- controls). For ‘cup-years’ of coffee consumed
ables yielded comparable results, with the among men, estimates of association for all
exception of a model that adjusted for ‘years types of coffee combined or for regular ground
stopped smoking’ that yielded null results (> 30 coffee were around 1.0 with no evidence of a
cups/week vs 1–15 cups/week OR,  1.07; 95% dose–response relationship or trend. For both
CI, 0.62–1.85). Another paper published on the regular and decaffeinated instant coffee, some
same study (Slattery et al., 1988b) with slightly estimates were > 1 but there was no evidence of
larger numbers also reported a non-statistically a dose–response relationship. Among women,
significant association with no consistent dose– for all types of coffee combined and regular
response relationship (>  40 servings/week vs ground coffee there were inverse associations
never drinkers OR,  1.6; 95% CI,  1.00–2.56). In for the highest intake categories (regular ground
this study there was no evidence of an association coffee OR, 0.3; 95% CI, 0.1–1.0 for > 90 cup-years
between consumption of decaffeinated coffee and compared with non-drinkers), but the trend was
risk of bladder cancer. [It was noted in the study only statistically significant for regular ground
that a substantial proportion of the Utah popul- coffee (P  =  0.02). [The number of cases in the
ation belongs to the Mormon church, which highest intake category was 8 and there were 9
forbids the consumption of coffee and tea as well non-drinkers.] For regular instant coffee and
as alcohol and tobacco; there is therefore the decaffeinated instant coffee some of the estimates
potential for underreporting of both coffee and were either > 1.0 or < 1.0; none were statistically
smoking, which might lead to bias and residual significant however, and there was no evidence
confounding.] of a dose–response relationship or trend. [There
Bruemmer et al. (1997) reported on a popu- was no evidence that different coffee types were
lation-based study in Washington, USA (262 mutually adjusted, and there was no adjustment
cases, 405 controls). The odds ratio comparing for race even though this was a multiethnic study.

138
Drinking coffee

Adjustment for smoking only included pack- of a dose–response trend. When considering life-
years, raising concerns about potential residual long consumption in number of cups, the odds
confounding.] ratio for 40 000 cups versus none was 1.14 (95%
Momas et al. (1994) reported on a study CI, 0.69–1.9) for men and 1.39 (95% CI, 0.31–6.25)
conducted in the Herault district, France (219 for women. Analyses restricted to non-smoking
cases, 792 controls). They reported an odds men or women showed positive associations,
ratio for lifelong coffee drinking of 4.1 (95% although neither were significant [the numbers
CI, 1.7–10.0) for > 60 000 cups compared with of cases for many of the strata among men were
< 365 cups. Whereas estimates for lower strata <  10, and all of the strata among women were
were smaller, there was no clear dose–response < 10]. [The Working Group noted that very small
relationship. [No estimates of trend were numbers were employed in the stratified analyses
reported. It was also noted that the reference by smoking. Smoking adjustment may not have
category had only 8 cases and that adjustment been adequate, as only cigarettes/day for men
for smoking only included lifelong smoking and smoking status for women were considered.]
(cigarettes equivalent), raising concerns about Vena et al. (1993) reported results from a study
residual confounding.] carried out in western New York, USA (351 hospi-
Piper et al. (1986) reported results from a popu- tal-based cases, 855 population-based controls).
lation-based case–control study of bladder cancer When comparing the highest intake category
in women (aged 20–49 years) conducted in New (≥ 5 cups/day) to the lowest (0–1 cup/day) they
York State (165 cases, 165 controls). The odds ratio reported an odds ratio of 2.1 (95% CI, 1.3–3.2), and
for drinking more than 101 cup-years compared there was evidence of a dose–response relation-
with non-drinkers was 2.1 (95% CI, 0.7–6.3). [No ship with a significant trend (P for trend, <0.001).
test for trend estimate or counts for each exposure When restricting analyses to non-smokers there
level were presented. Adjustment for smoking was also evidence of a positive association, and
only included pack-years, raising concerns about among those > 65 years old there was evidence
residual confounding.] of a dose–response relationship and a signifi-
cant trend (P  for trend,  0.02). Positive associa-
(b) Hospital-based case-control studies that tions were also noted for decaffeinated instant,
used population-based controls decaffeinated perk, regular instant, and regular
Hospital-based case–control studies that used perk, although these analyses were only adjusted
population-based controls and reported results for age and education. [Among the weaknesses
for coffee intake are discussed in the following. of this study were the low response rates which,
The Working Group considered these studies to combined with the fact that deceased subjects
be slightly less informative than those described or those too ill to participate were not included,
in Section 2.1.2 (a) above, and they were corre- raises concerns about possible bias. Many of the
spondingly given less weight in the evaluation. controls declined to participate, which could also
Escolar Pujolar et al. (1993) reported findings introduce a bias. Many of the strata evaluated
from a study conducted in Spain (497 cases, 1113 had very small numbers. Subject numbers for
controls). They reported no evidence of associ- analyses stratifying by smoking status were not
ation between frequency of coffee consumption shown. Adjustment for smoking only considered
and risk of bladder cancer among men, with all pack-years which might not be adequate, raising
estimates close to 1.0. The highest versus lowest concerns about residual confounding.]
intake level odds ratio among women was 0.71
(95% CI,  0.20–2.56), but there was no evidence

139
IARC MONOGRAPHS – 116

(c) Hospital-based case-control studies that underlying population. Adjustment for smoking
excluded diseases that may affect coffee only included smoking status, raising concerns
intake about residual confounding.]
Hospital-based case–control studies that Turati et al. (2015) reported on a hospital-based
used hospital-based controls and reported study conducted in Italy (690 cases, 655 controls).
results for coffee intake are described in the When considering the average lifetime intake, the
following. The Working Group considered these odds ratio for the highest versus the lowest cate-
studies less informative than those described gory was 1.73 (95% CI, 1.08–2.77) and 1.06 for a
in Sections 2.2.1 (a) and (b) above, and so were 1 cup/day increase (95% CI, 0.99–1.14). There was
given less weight in the evaluation. no consistent evidence of a dose–response trend,
Villanueva et al. (2009) reported on a hospi- and the trend test P value was 0.049. Estimates
tal-based study conducted in Spain (1219 cases, for current drinking did not show statistically
1271 controls). The odds ratio for the highest significant associations or evidence of a dose–
level of consumption (≥  4  cups/day) compared response relationship. However, when analyses
with never drinkers was 1.27 (95% CI, 0.88–1.81; were restricted to non-smokers there was an
P for trend,  0.082) and there was no consistent odds ratio of 1.18 (95% CI, 0.96–1.46), whereas
dose–response relationship. They also reported estimates were around 1.0 among ex-smokers
estimates stratified by smoking status; the odds or current smokers. Comparable analyses
ratios for the highest intake versus never drinkers performed with lifetime coffee drinking showed
were >  1.0 among never, former, and current similar odds ratios (close to 1.0) across the three
smokers, but there was no consistent dose– categories of smoking. There was no significant
response relationship for any of the groups and association observed between years of drinking
none of the trend tests were significant. [Smoking or age at which coffee drinking began.
adjustment only included smoking intensity, so Rebelakos et al. (1985) conducted a study in
residual confounding cannot be ruled out.] Greece (300 cases, 300 controls) and reported that
Wang et al. (2013a) reported on a hospi- drinking > 2 cups/day compared with < 2 cups/day
tal-based case–control study conducted in had an odds ratio of 1.7 (95% CI, 1.2–2.3). Results
Houston, Texas, USA (1007 cases, 1299 controls). stratifying by sex showed estimates of similar
When comparing the highest intake level of all magnitude, although they were only significant
types of coffee combined (>  2  servings/day) among men. Analyses comparing cups/day to
with never drinkers, the odds ratio was 1.14 never drinkers showed no evidence of a dose–
(95% CI,  0.9–1.46; P for trend,  0.336). There response relationship. [The Working Group noted
was no evidence for a dose–response relation- that sample size among women was very small
ship. When considering decaffeinated coffee (these analyses were therefore underpowered)
only, the comparable odds ratio was 1.37 (95% and that adjustment for smoking only considered
CI, 1.09–1.73; P for trend, 0.001); however, there smoking status, raising concerns about residual
was no evidence of a dose–response relationship, confounding.]
with the middle category estimate being larger De Stefani et al. (2007) conducted a hospi-
than the highest category. Estimates for regular tal-based study in Uruguay (255 cases, 501
coffee only were no near 1.0. [Controls were indi- controls) and reported an odds ratio for the
viduals attending clinics for annual check-ups; highest intake (≥  7  cups/week) and interme-
there is therefore concern that their coffee- diate intake of coffee (1–6 cups/week) compared
drinking habits are not representative of the with never drinkers of 2.1 (95% CI, 1.2–3.6) and
1.5 (95% CI,  1.1–2.2), respectively, with a P for

140
Drinking coffee

trend of <0.01. Similar estimates were observed were observed for both men and women without
when considering pure coffee or coffee with consistent evidence of a dose–response relation-
milk. [Diseases among controls were listed; it is ship. [For analyses of non-smokers, the reference
unclear whether some of them could affect coffee category had 7 cases for men and 13 cases for
intake, raising concerns about possible bias in women. Control subjects had conditions that
estimates.] could affect coffee drinking habits (approxi-
mately 20% had gastrointestinal diseases and
(d) Hospital-based case-control studies that close to 30% had cardiovascular diseases), leading
used controls with diseases that may affect to concerns about possible selection bias.]
coffee intake, or where no information was Kunze et al. (1992) reported on a hospi-
provided tal-based study carried out in Germany (675
Hospital-based case–control studies that cases, 675 controls) which found an odds ratio
used hospital-based controls and included for the highest category of intake (>  5  cups/
diseases that may have affected coffee intake, or day) compared with never drinkers of 2.0 (95%
studies for which it is not clear if other diseases CI,  1.2–3.3) for men and 2.7 (95% CI,  0.9–7.8)
were considered (raising concerns about biased for women. There was also some evidence of a
estimates), are described in the following. The positive dose–response relationship, but no test
Working Group considered these studies to for trend was provided. A previous report was
be less informative than those described in published by Claude et al. (1986), reporting on
Sections 2.1.2 (a)–(c) above, and gave them little a subset of these patients. [A main limitation of
weight in the evaluation. this study was the use of controls with urolog-
Clavel & Cordier (1991) conducted a hospi- ical diseases, such as hyperplasia of the prostate
tal-based study in France (781 cases, 781 controls), in men and urinary infections in women, which
reporting positive associations for all individ- may affect their liquid intake and possibly intro-
uals combined and for non-smoking men and duce a bias in the estimates.]
women separately. [All analyses were conducted Wynder & Goldsmith (1977) reported find-
using never drinkers as the reference, and subject ings from a hospital-based study conducted in
numbers for this category are < 10 for both men the USA (732 cases, 732 controls). Compared
and women non-smokers (1 and 3, respect- with individuals with no or occasional intake,
ively); all estimates are therefore very unstable. the odds ratio for those who consumed ≥ 7 cups/
Adjustment for smoking was performed using day was 2.0 (95% CI,  0.8–4.9). [No definition
smoking status only, which may lead to residual of the smoking variable used for controlling
confounding. More than 50% of controls had a confounding was provided. Controls with
disease that may affect coffee intake, leading to diseases that may affect coffee intake were not
biased estimates.] excluded, raising concerns about bias.] An
Geoffroy-Perez & Cordier (2001) reported on expanded study (Kabat et al., 1986) included
a hospital-based study conducted in France (765 some of these cases as well as additional cases
cases, 765 controls). When comparing the highest recruited later (152 cases, 492 controls). No asso-
intake category with the lowest, they reported ciation between consumption of brewed coffee or
an odds ratio of 1.42 (95% CI, 0.94–2.14) among decaffeinated coffee and risk of bladder cancer
men and 0.74 (95% CI, 0.28–1.96) among women. was observed for either sex, with all estimates
There was no evidence for a dose–response trend being very close to unity and based on very small
for either men or women. When restricting numbers. [The Working Group noted the very
analyses to non-smokers, positive associations small numbers for stratified analyses, the same

141
IARC MONOGRAPHS – 116

concerns as for the parent study by Wynder & Fraumeni et al. (1971) reported on a study
Goldsmith (1977).] conducted in the USA (493 cases, 527 controls),
Mettlin & Graham (1979) reported results a reanalysis of a previous study conducted by
from a hospital-based study performed in the Dunham et al. (1968). A positive association was
USA (569 cases, 1025 controls) which showed found for black men and women (statistically
that consumption of ≥  3  cups/day compared significant in women only), without evidence of a
with <  1  cup/day was associated with an odds dose–response relationship. Positive associations
ratio of 1.30 for men and women combined [no were seen for white and black men, but neither
confidence intervals were provided]. The corre- was statistically significant. Overall, there was no
sponding results for men and women separately consistent dose–response relationship [no confi-
were 1.64 and 0.81. Among men classified as rela- dence intervals were presented].
tively light smokers (< half a pack/day) there was Pohlabeln et al. (1999) conducted a hospi-
still a positive association, whereas for women tal-based study in Germany (300 cases, 300
classified as relatively light smokers there was a controls). When comparing the highest
slight inverse association. Neither estimate was intake level of coffee (≥  5 cups/day) with the
statistically significant, and there was no evidence lowest (≤  1 cup/day), the odds ratios were 1.59
of a dose–response relationship [no definition (95% CI,  0.87–2.91) for men and 1.25 (95%
of diseases among controls]. A previous report CI, 0.29–5.30) for women. There was no evidence
by Bross & Tidings (1973) reported on the same of a dose–response relationship, as estimates for
patients in this study. the middle category (2–4  cups/day) were either
D’Avanzo et al. (1992) reported results from a higher than or similar to the highest category. No
hospital-based study performed in Italy (555 cases, test for trend was provided. They also reported
855 controls). The odds ratio for the highest intake analyses among non-smokers, but numbers
level of regular coffee (≥ 4  cups/day) compared were too small to be meaningful. [Among male
with non-drinkers was 1.4 (95% CI, 0.9–2.2; P for controls, 41% had prostatic adenoma and 30%
trend, > 0.05), with no evidence of a dose–response had kidney stones. Among women, 13% had
relationship. Coffee drinking for ≥  30 years urinary infections and 62% had kidney stones.
compared with no coffee drinking yielded an odds The Working Group considered that it is feasible
ratio of 1.4 (95% CI, 0.9–2.2), whereas drinking that patients with prostate adenoma may have
coffee for < 30 years had an odds ratio of 1.2 (95% changed coffee-drinking habits due to increased
CI, 0.9–1.7; P for trend, <  0.05). [The strengths urination, raising concerns about possible bias.]
of this study include consideration of drinking Covolo et al. (2008) reported on a hospi-
history.] A non-statistically significant positive tal-based study carried out in Italy (197 cases,
association was also reported for decaffeinated 211 controls). Comparing the highest level of
ever drinking versus never drinking. [No specific coffee intake (>  3  cups/day) with non-coffee
diseases excluded from controls were listed, drinkers resulted in an odds ratio of  1.25 (95%
raising concerns about possible bias.] CI,  0.59–2.67). There was no evidence of a
Ciccone & Vineis (1988) reported on a hospi- dose–response relationship and no test for trend
tal-based study in Italy (512 cases, 594 controls); presented. Results were also stratified by smoking,
none of the estimates were statistically signif- but numbers of non-smokers were too small to
icant. Analyses stratifying by smoking were be meaningful. Interactions were presented for
presented, but subject numbers were very small. the examined polymorphisms in metabolism
[No information was provided about the condi- enzymes, but no details of test of interaction were
tions of the controls.] presented. [The Working Group was concerned

142
Drinking coffee

about bias introduced by patient controls, as well There was no evidence of a dose–response trend
as the small numbers in many categories.] and the trend test was not statistically significant.
Donato et al. (1997) reported on another Iscovich et al. (1987) conducted a hospi-
hospital-based study in Italy (172 cases, 578 tal-based study in Argentina with 117 cases
controls). Among men, the odds ratio for and 234 controls (117 hospital and 117 neigh-
comparing the lowest (1–2  cups/day), interme- bourhood). The odds ratios for consumption of
diate (3–4 cups/day), and highest intake level 1 cup/day, 2 cups/day or ≥ 3 cups/day compared
(≥ 5 cups/day) with non-drinkers were 2.3 (95% with non-drinkers were 1.08, 4.45, and 12,
CI, 0.9–5.6), 2.8 (95% CI, 1.1–7.4), and 4.5 (95% respectively. [No confidence intervals or test for
CI, 1.2–16.8), respectively, without a statistically trend were provided. Hospital controls included
significant trend (P  for trend,  >0.1). Among patients with digestive system problems (16%),
women, the estimates for the lowest (1–2 cups/day) heart disease (17%), and hypertension diseases
and highest (3–4  cups/day) coffee intake levels (12%), all of which could affect coffee drinking
compared with non-coffee drinkers were 4.3 (95% and lead to bias.]
CI, 0.8–23.9) and 4.9 (95% CI, 0.7–33.0), respect-
ively. [Numbers for some of the categories were 2.1.3 Meta-analyses and pooled analyses
very small, in particular non-drinkers. Controls
included several benign urological diseases Sala et al. (2000) conducted a pooled
(prostate adenoma, urolithiasis, obstructive analyses of coffee intake and bladder cancer
uropathy), and it is not clear if these disorders among non-smokers that included ten case–
affect coffee intake. Prostate adenoma could control studies carried out in Europe, including
affect coffee intake, raising concerns about bias Rebelakos et al. (1985), Jensen et al. (1986),
in the results.] Ciccone & Vineis (1988), Clavel & Cordier
Simon et al. (1975) conducted a hospital-based (1991), Kunze et al. (1992), Escolar Pujolar et al.
study in the USA (135 cases, 390 controls) and (1993), Donato et al. (1997), and Pohlabeln et al.
reported non-statistically significant positive (1999), discussed in Section  2.1.2 above. These
associations among non-smokers/light smokers ten studies involved a total of 564 cases and 2929
and also among moderate–heavy smokers. controls. The pooled odds ratio from comparing
[Subject numbers for this analysis were very low.] the highest intake level (≥  10  cups/day) with
Radosavljević et al. (2003) conducted a hospi- never drinkers was 1.8 (95% CI,  1.0–3.3), with
tal-based study in Serbia (130 cases, 130 controls) no evidence of a dose–response relationship or
and reported an odds ratio for coffee intake of 1.46 a significant trend. When stratifying studies by
(95% CI, 1.05–2.01). [The units associated with the types of controls among studies that used hospi-
reported odds ratios are not clear from the paper. tal-based controls, the odds ratio was 3.2 (95%
The smoking variable used was not defined, so CI,  1.4–7.3) with a P for trend of  0.05. Among
there is concern over residual confounding. It is studies that used population-based controls, the
not clear if the diseases among controls may have odds ratio was 0.7 (95% CI, 0.2–2.0) with a P for
influenced coffee intake, leading to bias.] trend of 0.3 [the number of cases in the highest
Wakai et al. (2004) reported results from a category among population-based controls was
study conducted in Japan (124 cases, 620 controls). 4]. Similar estimates were observed when further
The odds ratio for comparing the highest level stratifying by sex although, among women, the
of coffee intake (≥  3 cups/day) with the lowest odds ratio for population-based controls was
(almost never) was 1.14 (95% CI,  0.58–2.23). > 1.0. Analyses taking into account duration of
consumption in years (six studies) showed an

143
IARC MONOGRAPHS – 116

odds ratio for the longest duration compared 1986; Jacobsen et al., 1986; Mack et al., 1986;
with never drinkers of 0.9 (95% CI, 0.6–1.2). Norell et al., 1986; Wynder et al., 1986; Raymond
Wu et al. (2015) conducted a meta-analysis et al., 1987; Pfeffer et al., 1989; Mizuno et al.,
that included 25 case–control (15 419 cases and 1992; Kalapothaki et al., 1993; Gullo et al., 1995;
23  585 controls) and five prospective studies Kokic et al., 1996; Mori et al., 1999). We also
(753 cases and 236  343 controls). The overall excluded studies that did not provide sufficient
pooled odds ratio for all studies was 1.33 (95% information regarding risk estimates associated
CI,  1.19–1.48), and heterogeneity was present with coffee intake (Kinlen & McPherson, 1984;
(P = 0.008; I2 = 38.4%). For case–control studies, the Baghurst et al., 1991, Chan et al., 2009).
combined odds ratio was 1.37 (95% CI, 1.22–1.53) If the 14 cohort studies included in a pooled
and also showed heterogeneity (P  =  0.017; I2 = analysis by Genkinger et al. (2012) are counted
37.1%). For cohort studies the corresponding individually, then evidence from 20 individual
odds ratio was 1.10 (95% CI, 0.78–1.54) with less cohort studies is available. In addition, 22 case–
heterogeneity (P  =  0.112; I2  =  44%). Subgroup control studies were available that controlled for
analyses were performed for various character- smoking, 14 of which were population-based
istics, such as type of control (hospital, popul- and 8 hospital-based. For the reviewed studies,
ation, or both). The meta-analysis odds ratio for detailed information is presented in Table  2.3
studies that used hospital-based controls (20 for cohort studies and Table 2.4 for case–control
studies) was 1.44 (95% CI, 1.21–1.72); for studies studies.
that used population-based controls (12 studies),
the meta-analysis odds ratio was 0.98 (95% 2.2.1 Cohort studies
CI,  0.63–1.52). While studies based in Europe
or America had comparable meta-analysis odds See Table 2.3
ratios of approximately 1.3, studies from Asia A nested case–control analysis of a cohort
had an odds ratio of 1.0 (95% CI, 0.7–1.4). [Of the study investigated pancreatic cancer mortality in
11 cohort studies with data available, Wu et al. a follow-up of 50 000 male former college students
(2015) only included 4; there are also 6 other (Whittemore et al., 1983). There were 84 deaths
studies that were published during the period from pancreatic cancer. Data on coffee and tea
considered in this meta-analysis that were not consumption and other variables were collected
included.] during a physical examination at the college. No
statistically significant association with coffee
consumption was noted; after adjustment for
2.2 Cancer of the pancreas smoking, age, college, and class year the relative
risk was 1.1 (95% CI, 0.7–1.9) when comparing
The Working Group reviewed all of the
those drinking ≥ 2 cups/day with those drinking
pertinent cohort studies (including nested
< 2 cups/day.
case–control or case–cohort studies), case–
In a Hawaiian cohort study of the association
control studies, and pooled and meta-analyses
between cancer incidence and coffee consump-
that assessed the association between coffee
tion, 7355 Japanese men were followed for a
consumption and cancer of the pancreas.
minimum of 14 years from the time of collection
Studies were excluded if statistical analyses
of a 24-hour dietary recall during 1965–1968
were not adjusted for smoking, since it is an
(Nomura et al., 1986). This is an update of an
important potential confounder (Jick & Dinan,
earlier study by the same group (Nomura et al.,
1981; Kessler, 1981; Goldstein, 1982; Heuch et al.,
1981). Incidence rates were adjusted for age or
1983; Snowdon & Phillips, 1984; Hsieh et al.,

144
Table 2.3 Prospective cohort studies on cancer of the pancreas and drinking coffee

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category or cases/deaths (95% CI)
enrolment/follow- assessment method level
up period
Whittemore et al. 50 000 (84 cases): college Pancreas Current coffee drinking (cups/day) Age, smoking, college, Strengths: nested case–
(1983) alumni, male students < 2 60 1.0 class year control with 4 controls
USA, 1962–1966 who entered Harvard ≥ 2 24 1.1 (0.7–1.9) per case, matched on
(enrolment), University during birth year
mortality until 1916–1950 or University Limitations: fatal cancer
1978 of Pennsylvania during only, small number of
1931–1940 cases, limited exposure
Exposure assessment information
method: questionnaire
Nomura et al. 7355 (21 cases): Japanese Pancreas Current coffee drinking (cups/day) Age, smoking Strengths: prospective
(1986) men born during 0 2 1.00 design
USA (Hawaii), 1900–1919 on Hawaiian 1–2 7 1.16 Limitations: very small
1965–1968 Island of Oahu, aged number of cases, intake
3–4 7 2.08
(enrolment), 45–68 yr at baseline based on 24-hour recall,
incidence until Exposure assessment ≥ 5 5 1.63 limited confounder
July 1983 method: 24-hour diet Trend test P value, 0.41 information
recall
Hiatt et al. (1988) 122 894 (49 cases): Pancreas Current coffee drinking (cups/day) Age, sex, ethnicity, Strengths: prospective
USA, 1978–1984 members (men and 0 NR 1.0 smoking, alcohol design
(enrolment), women) of the Kaiser < 1 NR 0.8 (0.3–2.6) consumption, diabetes, Limitations: short follow-
incidence 6 yr Permanente Medical blood glucose up, small number of cases
1–3 NR 0.9 (0.4–2.1)
Care Program in
Northern California ≥ 4 NR 0.7 (0.2–1.9)
who had a multiphasic
health check-up during
1978–1984, mean age at
baseline 41 yr
Exposure assessment
method: questionnaire

Drinking coffee
145
146

IARC MONOGRAPHS – 116
Table 2.3 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category or cases/deaths (95% CI)
enrolment/follow- assessment method level
up period
Mills et al. (1988) 34 198 (40 cases), Pancreas Coffee drinking status Age, smoking, sex, Strengths: prospective
USA, 1976 non-Hispanic white Not current NR 1.00 consumption of meat design
(enrolment), Californian Seventh-day Current NR 2.21 (0.61–7.99) and eggs Limitations: fatal cancer
mortality 1976– Adventists, men and only, low number of cases
1982 (6 yr) women, aged ≥ 25 yr due to short follow-
Exposure assessment up, only dichotomous
method: questionnaire exposure to coffee (few
heavy coffee drinkers),
generalizing findings
to general population
limited
Friedman & van 175 000 (450 cases, Pancreas Current coffee drinking (cups/day) Age, sex, smoking, Part of substantial
den Eeden (1993) 2687 controls in nested ≤6 NR 1.00 race, examination site, multiple-comparison
USA, case–control analysis), > 6 NR 0.95 (0.73–1.22) date of first check-up analysis
incidence members (men and Strengths: large cohort
Trend test P value, 0.672
1964–1988 women) of the Kaiser study, with relatively
Permanente Medical large number of cases
Care Program in Limitations: very limited
Northern California exposure information
Exposure assessment (single coffee intake
method: questionnaire, question of “Do you
focusing on large usually drink over 6 cups
volumes of consumption of coffee per day?”)
Zheng et al. (1993) 17 633 (57 cases), white Pancreas Current coffee drinking (cups/day) Age, smoking, alcohol Strengths: prospective
USA, men aged ≥ 35 yr, policy < 3 21 1.0 consumption design
1966 (enrolment), holders of the Lutheran 3–4 18 0.6 (0.3–1.2) Limitations: fatal cancer
mortality 1966– Brotherhood Life only, small number of
5–6 12 0.7 (0.4–1.6)
1986 (20 yr) Insurance Society (LBS) cases
Exposure assessment ≥ 7 5 0.9 (0.3–2.4)
method: FFQ
Table 2.3 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category or cases/deaths (95% CI)
enrolment/follow- assessment method level
up period
Shibata et al. 13 979 (65 cases), men Pancreas Current coffee drinking (cups/day) Age, sex, smoking Strengths: prospective
(1994) and women, mean < 1 7 1.00 design
USA, 1981–1985 age at entry (standard 1 16 1.82 (0.75–4.43) Limitations: upper–
(enrolment), deviation) 75.0 yr (men) middle socioeconomic
2–3 35 1.67 (0.74–3.77)
incidence 1981– and 73.8 yr (women) class considered only,
1990 (9 yr) Exposure assessment ≥ 4 5 0.88 (0.28–2.80) small number of cases,
method: FFQ limited confounder
information
Stensvold & 42 973 (41 cases) Pancreas Current coffee drinking (cups/day): women Age, smoking, Strengths: prospective
Jacobsen (1994) men and women ≤ 4 6 1.0 residence design, sex-specific
Norway, 1977– aged 35–54 yr, living ≥ 5 9 1.2 analyses
1982 (enrolment), in three counties Limitations: small
Current coffee drinking (cups/day): men
incidence until in different parts of number of pancreas
1990 (average Norway, participating ≤ 4 9 1.0 cancer cases overall,
10.1 yr) in a cardiovascular 5–6 9 1.0 and sex-specific, very
screening programme ≥ 7 8 0.6 few subjects drinking
organized by the 0–1 cups/day, limited
National Health confounder information,
Screening Service multiple comparisons (15
Exposure assessment cancer sites analysed)
method: FFQ

Drinking coffee
147
148

IARC MONOGRAPHS – 116
Table 2.3 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category or cases/deaths (95% CI)
enrolment/follow- assessment method level
up period
Harnack et al. 33 976 (66 cases) women Pancreas Coffee (cups/wk) Age, smoking Comparison of results in
(1997) living in Iowa aged ≤ 7 11 1.00 never smokers with total
USA, 1986 55–69 yr (Iowa Women’s 8–17.5 20 1.91 (0.92–40.00) cohort suggests residual
(enrolment), Health Study), 99% of confounding by smoking.
> 17.5 35 2.15 (1.08–4.30)
incidence 1986– cohort was white Updated version of this
1994 (9 yr) Exposure assessment Trend test P value, 0.03 study (with inverse
method: FFQ Coffee consumption (cups/wk): never smokers Age association) is reported
≤ 7 10 1.00 in pooled analysis of
8–17.5 11 1.36 (0.58–3.20) Genkinger et al. (2012)
> 17.5 17 1.74 (0.80–3.80) Strengths: population-
based cohort, validated
Trend test P value, 0.17
FFQ (from NHS),
prospective design
precludes recall bias,
separate results for never
smokers
Limitations: low
number of cases, limited
confounder information
Michaud et al. 88 799 in NHS (158 Pancreas Current coffee drinking (cups/day): women Age, sex, smoking, Strengths: validated
(2001) female cases), 47 794 in 0 39 1.00 BMI, diabetes, FFQ (from NHS), large
USA, 1980 (NHS HPFS (130 male cases) < 1 10 0.72 (0.36–1.44) cholecystectomy, cohorts with detailed
enrolment), 1986 Exposure assessment energy intake, period information, able to
1 14 0.71 (0.38–1.30)
(HPFS enrolment), method: FFQ control for multiple
1980–1996 (NHS 2–3 52 0.88 (0.58–1.34) confounders
incidence), > 3 43 0.88 (0.56–1.38) Limitations: limited to
1986–1998 (HPFS, Trend test P value, 0.92 health professionals
incidence) Current coffee drinking (cups/day): men
0 47 1.00
< 1 36 1.04 (0.67–1.61)
1 10 0.48 (0.24–0.95)
2–3 31 0.89 (0.56–1.40)
> 3 6 0.37 (0.16–0.88)
Trend test P value, 0.04
Table 2.3 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category or cases/deaths (95% CI)
enrolment/follow- assessment method level
up period
Michaud et al. Current coffee drinking (cups/day)
(2001) 0 86 1.00
(cont.) < 1 46 0.94 (0.65–1.36)
1 24 0.60 (0.38–0.94)
2–3 83 0.88 (0.65–1.21)
> 3 49 0.62 (0.27–1.43)
Trend test P value, 0.35
Isaksson et al. 21 884 (131 cases), Pancreas Current coffee drinking (cups/day) Age, sex, smoking Strengths: 90% of the
(2002) Swedish Twin Registry 0–2 29 1.00 pancreas tumours were
Sweden cohort: male and female 3–6 95 0.91 (0.60–1.38) histologically confirmed
1961 (enrolment), same-sexed twin pairs Limitations: no incidence
≥7 7 0.39 (0.17–0.89)
1969–1997 born during 1886–1925 data in period 1961–1969,
(incidence, 16 yr and both living in limited dietary and
median) Sweden in 1961 confounder information
Exposure assessment
method: questionnaire
Lin et al. (2002) 110 792, JACC (46 465 Pancreas Current coffee drinking: men Age, smoking pack- According to authors,
Japan, 1988–1990 men and 64 327 0 35 1.00 years the association between
(enrolment), women), inhabitants 1–2 cups/mo 12 0.74 (0.37–1.49) coffee consumption and
mortality until of 45 areas throughout pancreatic cancer risk
1–4 cups/wk 19 0.58 (0.32–1.08)
1997 (8.1 yr Japan aged 40–79 yr at was similar for non-
average) baseline 1 cup/day 8 0.59 (0.26–1.33) smokers and current
Exposure assessment 2–3 cups/day 11 0.75 (0.36–1.59) smokers (data not shown)
method: questionnaire ≥ 4 cups/day 5 3.19 (1.22–8.35) Strengths: large cohort
Trend test P value, 0.79 study with relatively large
Current coffee drinking: women number of cases
Limitations: fatal
0 27 1.00
cancer only, no data on
1–2 cups/mo 12 1.27 (0.64–2.54) histological confirmation,
1–4 cups/wk 11 0.74 (0.36–1.50) small proportion

Drinking coffee
1 cup/day 9 0.94 (0.44–2.01) drinking larger amounts
2–3 cups/day 2 0.31 (0.07–1.33) of coffee with very few
≥ 4 cups/day 1 1.8 (0.24–13.66) drinking > 4 cups/day,
limited confounder
Trend test P value, 0.21
information
149
150

IARC MONOGRAPHS – 116
Table 2.3 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category or cases/deaths (95% CI)
enrolment/follow- assessment method level
up period
Stolzenberg- 27 111 (163 cases), Pancreas Coffee consumption (g/day) Age, smoking years Strengths: detailed and
Solomon et al. participants ATBC, ≤ 321.4 NR 1.00 validated FFQ
(2002) smoking men aged 450 NR 1.48 (0.89–2.46) Limitations: male
Finland, 1985– 50–69 yr residing in smokers only, few people
624.9 NR 1.12 (0.61–2.03)
1988 (enrolment), southwestern with low intake of coffee
incidence until Finland, randomized to 878.6 NR 1.72 (1.01–2.86)
1997 (10.2 yr receive supplements or > 878.6 NR 0.95 (0.54–1.68)
median) placebo Trend test P value, 0.62
Exposure assessment
method: FFQ
Khan et al. (2004) 3158 (25 fatal cases), Pancreas Coffee drinking: men Age, smoking Limitations: no data on
Japan, mortality subjects aged ≥ 40 yr Non/occasional NR 1.0 histological confirmation,
1984–2002 (mean using the resident ≥ several NR 0.6 (0.2–2.2) very small number of
13.8 yr for men, registries of Hokkaido, times/wk cases, fatal cases only,
14.8 yr for women) Japan (1524 men and limited control for
Coffee drinking: women Age, health status,
1634 women) confounders 
Non/occasional NR 1.0 health education,
Exposure assessment
≥ several NR 0.2 (0–1.8) health screening,
method: questionnaire
times/wk smoking
Table 2.3 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category or cases/deaths (95% CI)
enrolment/follow- assessment method level
up period
Luo et al. (2007) 102 137 (233 cases), Pancreas Current coffee drinking: men Age, sex, smoking, Strengths: large number
Japan, incidence JPHC Study, conducted Rarely 54 1.0 BMI, physical activity, of incident cases
1990–2003 (mean in 11 public health 1–2 cups/wk 30 1.0 (0.6–1.5) alcohol, diabetes, Limitations: no data on
11 yr) centre-based areas cholelithiasis, study histological confirmation,
3–4 cups/wk 15 0.8 (0.5–1.5)
throughout Japan area, green tea relatively few people with
among residents aged 1–2 cups/day 25 0.7 (0.4–1.1) high coffee intake
40–69 yr (48 783 men ≥ 3 cups/day 11 0.6 (0.3–1.1)
and 53 354 women) Trend test P value, 0.04
Exposure assessment Current coffee drinking: women
method: FFQ Rarely 38 1.0
1–2 cups/wk 16 0.9 (0.5–1.7)
3–4 cups/wk 14 1.7 (0.9–3.1)    
1–2 cups/day 24 1.3 (0.8–2.3)
≥ 3 cups/day 6 1.3 (0.5–3.3)
Trend test P value, 0.2
   
Current coffee drinking: men and women combined
Rarely 92 1.0
1–2 cups/wk 46 1.0 (0.7–1.4)
3–4 cups/wk 29 1.1 (0.7–1.7)    
1–2 cups/day 49 0.9 (0.6–1.3)
≥ 3 cups/day 17 0.8 (0.4–1.3)
Trend test P value, 0.4    

Drinking coffee
151
152

IARC MONOGRAPHS – 116
Table 2.3 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category or cases/deaths (95% CI)
enrolment/follow- assessment method level
up period
Nilsson et al. 64 603 (74 cases), Pancreas All coffee (cups/day) Age, sex, smoking, Strengths: distinction
(2010) prospective cohort study < 1 5 1.00 BMI, education, between filtered and
Sweden, from the VIP, subjects 1–3 41 1.18 (0.47–3.02) physical activity boiled coffee
incidence aged 40–60 yr at start Limitations: small
≥ 4 28 1.50 (0.57–3.92)
1992–2007 Exposure assessment number of cases, no
(median 6 yr) method: FFQ Coffee intake, filtered method (cups/day) data on histological
< 1 23 1.00 confirmation
1–3 38 0.85 (0.50–1.44)
≥ 4 13 0.88 (0.44–1.76)
Coffee intake, boiled method (cups/day)
< 1 42 1.00
1–3 24 1.68 (1.01–2.81)
≥ 4 8 2.51 (1.15–5.50)
Nakamura et al. 30 826 (14 241 men Pancreas Current coffee drinking: men Age, smoking, BMI, Limitations: small
(2011) and 16 585 women; 52 Never 14 1.00 diabetes number of cases, only
Japan, mortality fatal cases) residents > 1 cup/mo to 11 0.67 (0.29–1.55) fatal cases, no histological
1992–1997 (5 yr) of Takayama, Gifu 4–6 cups/wk confirmation, low coffee
Prefecture, Japan, aged intake levels
≥ 1 cup/day 8 0.44 (0.15–1.29)
≥ 35 yr
Exposure assessment Trend test P value, 0.08
method: FFQ Current coffee drinking: women
Never 9 1.00
> 1 cup/mo to 5 0.62 (0.2–2)
   
4–6 cups/wk
≥ 1 cup/day 4 0.68 (0.17–2.78)
Trend test P value, 0.71    
Table 2.3 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category or cases/deaths (95% CI)
enrolment/follow- assessment method level
up period
Genkinger et al. 853 894 (317 828 men, Pancreas Coffee consumption (g/day): men and women Age, smoking, alcohol When the case
(2012) 536 066 women) and 0 149 1.00 consumption, diabetes, definition was limited
USA, Canada, 2185 cases (1047 men, 0.01 to < 150 135 1.16 (0.84–1.6) BMI, energy intake, to adenocarcinomas
Netherlands, 1138 women); pooling year of enrolment (n = 1554), no statistically
150 to < 400 316 1.01 (0.82–1.25)
Sweden, Australia, of 14 prospective cohort significant association
incidence studies (including 400 to < 900 738 1.08 (0.89–1.31) was observed with intake
1980–2005 (varies ATBC, BCDDP, CNBSS, ≥ 900 257 1.10 (0.81–1.48) of coffee.
by cohort) CPS-II, CTS, COSM, Continuous 1595 1.01 (0.97–1.04) Strengths: large size with
HPFS, IWHS, MCCS, for 237 g/day high number of cases,
NLCS, NYSC, NHS, increase enabling analyses of
PLCO, SMC) Trend test P value, 0.71 broad exposure range and
Exposure assessment Coffee consumption (g/day): men possibility to evaluate
method: FFQ effect modification
0 54 1.00
Limitations: none
0.01 to < 150 79 1.53 (1.03–2.26)
150 to < 400 163 1.02 (0.73–1.43)
400 to < 900 411 1.15 (0.84–1.58)
≥ 900 130 0.95 (0.67–1.36)
Continuous 837 0.98 (0.95–1.01)
for 237 g/day    
increase
Trend test P value, 0.06
   
Coffee consumption (g/day): women
0 95 1.00
0.01 to < 150 56 0.87 (0.53–1.43)
150 to < 400 153 1.00 (0.76–1.32)
400 to < 900 327 1.04 (0.8–1.34)    
≥ 900 127 1.18 (0.71–1.98)
Continuous 758 1.04 (0.97–1.11)
for 237 g/day

Drinking coffee
increase
Trend test P value, 0.5    
153
154

IARC MONOGRAPHS – 116
Table 2.3 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category or cases/deaths (95% CI)
enrolment/follow- assessment method level
up period
Bidel et al. (2013) 60 041 (29 159 men Pancreas Current coffee drinking (cups/day): men Age, smoking, study Coffee cup size was small
Finland, incidence and 30 882 women; 0 9 1.00 year, education, (100 mL)
1972–2006 (mean 235 cases) from six 1–2 14 0.72 (0.30–1.71) alcohol consumption, Strengths: prospective
18 yr) geographic areas of physical activity, design with long
3–4 32 0.76 (0.35–1.67)
Finland, random diabetes, tea, BMI follow-up (precluding
sampling of the 5–6 38 0.64 (0.29–1.41) recall bias), sex-specific
population aged 25–74 7–9 20 0.72 (0.31–1.68) analyses possible, wide
yr, stratified by area, sex, ≥ 10 16 0.80 (0.30–1.95) range of coffee intake
and 10-year age group Trend test P value, 0.91 analysed
Exposure assessment Current coffee drinking (cups/day): women
method: mailed,
0 3 1.00
self-administered
questionnaire 1–2 11 1.30 (0.36–4.77)
3–4 33 1.29 (0.39–4.31)
5–6 40 1.21 (0.36–4.07)
7–9 16 1.52 (0.42–5.43)
≥ 10 3 0.71 (0.14–3.63)
Trend test P value, 0.88
Bhoo-Pathy et al. 477 312 (865 cases), Pancreas Total coffee: country-specific quartiles Age, sex, centre, and Median total coffee
(2013) EPIC cohort Non-drinker 52 1.09 (0.8–1.5) age at diagnosis, intake ranged from
10 European Exposure assessment Q1 (ref) 237 1.00 height, weight, 92 mL/day in Italy to
countries method: FFQ smoking status, 900 mL/day in Denmark.
Q2 214 1.11 (0.92–1.34)
(Denmark, France, history of diabetes, Decaffeinated coffee also
Germany, Greece, Q3 196 0.99 (0.81–1.21) education, physical showed no association.
Italy, Netherlands, Q4 166 1.07 (0.86–1.33) activity, energy intake, Strengths: large study
Norway, Spain, Continuous 865 1.00 (0.97–1.02) red meat, processed size and number of cases,
Sweden, UK), for 100 mL/day meat, alcohol, tea, with large variation
1992–2000 increase soft drink, fruit, and in coffee intake, coffee
(enrolment), Trend test P value, 0.925 vegetable intake intake calibrated with
follow-up varied 24-hour recall
by country (mean Limitations: method
11.6 years) and source of follow-up
not described for most
countries
Table 2.3 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category or cases/deaths (95% CI)
enrolment/follow- assessment method level
up period
Guertin et al. 457 366 (1541 cases Pancreas Current coffee drinking (cups/day): men Age, smoking, The association did not
(2016) with exocrine pancreas 0 71 1.00 diabetes, race/ differ by tobacco smoking
USA, enrolment cancer); NIH-AARP < 1 153 1.14 (0.86–1.52) ethnicity, BMI, highest or self-reported history of
(NA), follow-up Diet and Health Study, level of education, diabetes.
1 146 1.02 (0.76–1.35)
incidence until participants aged 50–71 alcohol consumption, Strengths: large study size
2006 yr residing in one of six 2–3 427 1.05 (0.81–1.36) health status, use and number of cases
US states or two 4–5 142 1.06 (0.79–1.43) of nutritional
metropolitan areas ≥ 6 54 1.21 (0.84–1.75) supplements, current
Exposure assessment Trend test P value, 0.55 marital status, physical
method: FFQ Current coffee drinking (cups/day): women activity, history of
cardiovascular disease,
0 58 1.00
family history of
< 1 81 0.91 (0.65–1.28) cancer, energy intake,
1 112 1.12 (0.82–1.55) nutrient density-
2–3 218 1.01 (0.75–1.35) adjusted intakes of
4–5 53 0.89 (0.60–1.3) fruits, vegetables,
≥ 6 26 1.38 (0.85–2.22) folate, protein,
saturated fat, total fat
Trend test P value, 0.53
ATBC, Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study; BCDDP, Breast Cancer Detection Demonstration Project; BMI, body mass index; CNBSS, Canadian National Breast
Screening Study; CI, confidence interval; COSM, Cohort of Swedish Men; CPS-II, Cancer Prevention Study; CTS, California Teacher’s Study; EPIC, European Prospective Investigation
into Cancer and Nutrition; FFQ, food frequency questionnaire; HPFS, Health Professionals Follow-up Study; IWHS, Iowa Women’s Health Study; JACC, Japan Collaborative Cohort
Study for Evaluation of Cancer Risk; JPHC, Japan Public Health Center-based Prospective; MCCS, Melbourne Collaborative Cohort Study; mo, month(s); NA, not available; NHS,
Nurses’ Health Study; NIH-AARP, National Institutes of Health–Association of American Retired Persons; NLCS, Netherlands Cohort Study; NR, not reported; NYSC, New York State
Cohort; PLCO, Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial; SMC, Swedish Mammography Cohort; VIP, Västerbotten Intervention Project; wk, week(s); yr, year(s)

Drinking coffee
155
IARC MONOGRAPHS – 116

both age and smoking, using the entire cohort drinking > 6 cups/day of coffee was not associated
as the standard population. No significant asso- with increased pancreatic cancer risk (RR, 0.95;
ciation was reported between coffee drinking 95% CI, 0.73–1.22).
and risk of pancreatic cancer after adjusting for Via the Lutheran Brotherhood Life Insurance
smoking (P for trend, 0.41). [The Working Group Society (LBS) cohort, Zheng et al. (1993) studied
noted the very low number of cases; in addi- risk factors for pancreatic cancer mortality in a
tion, dietary information was based on a single cohort study of 17 633 white men in the USA who
24-hour recall.] responded to a mailed questionnaire in 1966 and
A cohort study in northern California inves- were followed up until 1986 for mortality. After
tigated a 6-year follow-up of pancreatic cancer 20 years of follow-up, 57 fatal pancreatic cancer
incidence among 122 894 men and women who cases were identified. Coffee consumption at
had completed a questionnaire collecting data baseline (current coffee drinking) was measured
on coffee, tea, smoking, and alcohol use during using a food frequency questionnaire (FFQ).
1978–1984 (Hiatt et al., 1988). There were 49 cases Coffee was not related to pancreatic cancer
of pancreatic cancer. A multivariate analysis mortality; the relative risk for those drinking
identified no increased pancreatic cancer risk ≥ 7 cups/day was 0.9 (95% CI, 0.3–2.4) compared
associated with increasing coffee consumption. with those drinking < 3 cups/day.
A cohort study (Mills et al., 1988) of 34 198 Shibata et al. (1994) examined risk factors
non-Hispanic, white Californian Seventh-day for pancreatic cancer in a cohort study of 13 979
Adventists followed participants for 6 years after men and women resident within a retirement
their completion of a questionnaire determining community in USA. After 9 years of follow-up,
exposure to several risk factors, including 65 incident cases of pancreatic cancer were
coffee consumption, in 1976. Forty deaths from identified. Coffee consumption at baseline was
pancreatic cancer were reported. Multivariate measured using a FFQ. Coffee was not related
analyses using the Cox proportional hazards to pancreatic cancer risk; the relative risk for
model resulted in a relative risk for current coffee those drinking ≥  4  cups/day compared with
consumption versus no coffee consumption, those drinking <  1 cup/day was 0.88 (95% CI,
adjusted for age, sex, and smoking, of 2.21 (95% 0.28–2.80).
CI, 0.61–7.99). [The Working Group noted that As part of a larger study on coffee drinking
the distribution of coffee drinking in this popul- and cancer incidence, Stensvold & Jacobsen
ation is unusual because there are few drinkers (1994) studied a cohort of 21 735 men and 21 238
of larger quantities of coffee; only 17–18% of the women aged 35–54 years. The study population
population drank ≥ 2 cups/day.] participated in a cardiovascular screening in
Friedman & van den Eeden (1993) conducted three counties in Norway during 1977–1982.
a nested case–control study within the Kaiser– After an average follow-up period of 10.1 years,
Permanente cohort study, consisting of people 41 incident cases were identified. Data on coffee
who had received multiphasic health check-ups habits at baseline were based on information
in the San Francisco Bay Area. Measurement from a self-administered FFQ. No statistically
of coffee intake was limited to one yes-or-no significant association was found between coffee
question in a questionnaire focusing on heavy drinking and incidence of cancer of the pancreas.
consumption: “Do you usually drink over 6 cups In men, the relative risk for those drinking
of coffee per day?” As part of an exploratory ≥ 7 cups/day compared with ≤ 4 cups/day was 0.6
analysis of 779 characteristics, coffee intake was (no confidence interval given) [coffee consump-
also analysed. After multivariate adjustment, tion is high in Norway]. In women, the relative

156
Drinking coffee

risk for those drinking ≥  5  cups/day compared Isaksson et al. (2002) studied the association
with those drinking ≤ 4 cups/day was 1.2. [The between coffee consumption and pancreatic
Working Group noted that the reference group cancer incidence in a cohort study of twins estab-
could include individuals who consumed signif- lished in 1958 and followed up by the Swedish
icant amounts of coffee.] Twin Registry. At 1961 (baseline), self-adminis-
Harnack et al. (1997) examined the relation- tered questionnaires regarding lifestyle factors
ship between coffee consumption and pancreatic were mailed. The analysis included 12 204 women
cancer incidence in the Iowa Women’s Health and 9680 men who responded to these question-
Study cohort. Data were available from 33  976 naires. For those who consumed ≥  7  cups/day
women aged 55–69 years in 1986 who responded compared with those who reported ≤ 2 cups/day,
to a mailed questionnaire and who were followed the relative risk of pancreatic cancer was 0.39
until 1994 (9 years) for cancer incidence. Coffee (95% CI, 0.17–0.89). [The Working Group noted
intake at baseline was estimated using a vali- that no incidence follow-up data were available
dated FFQ. The relative risk for those drinking for the period 1961–1969.]
≥ 17.5 cups/week compared with those drinking Lin et al. (2002) evaluated the association
≤  7  cups/week was 2.15 (95% CI, 1.08–4.30; P between coffee consumption and pancreatic
for trend, 0.03). Among never smokers, the rela- cancer mortality in a large-scale prospective
tive risk for the same consumption levels was not cohort study, the Japan Collaborative Cohort
statistically significant at 1.74 (95% CI, 0.80–3.80; Study for Evaluation of Cancer Risk (JACC
P for trend,  0.17). [The Working Group noted study). At baseline, a self-administered question-
that an updated version of this study with a naire was used to estimate coffee consumption.
longer follow-up, but with an inverse association, During the follow-up period (mean 8.1  years),
is reported in the pooled analysis of Genkinger 225 pancreatic cancer deaths were identified.
et al. (2012).] Overall, coffee intake was not associated with
Michaud et al. (2001) used data on coffee fatal pancreatic cancer. While the relative risks
intake from semiquantitative FFQs administered were inverse for those drinking up to 3 cups/day
at baseline in the Nurses’ Health Study (NHS) of coffee compared with non-consumers of coffee
and the Health Professionals Follow-Up Study (0 cups/day), the corresponding relative risk was
(HPFS), and in subsequent follow-up question- positive and statistically significant (RR,  3.19;
naires. In both the NHS and HPFS, repeated 95% CI, 1.22–8.35) for men who consumed ≥ 4
measurements for coffee intake were accounted cups/day of coffee. A similar, but less-pronounced
for in the analysis. The HPFS included 44  794 pattern of risks was observed among women.
men, while there were data available on 88 799 [The Working Group noted that, there was only
women from the NHS. Results revealed a signif- limited control for confounders.]
icant inverse association in men (RR for those Stolzenberg-Solomon et al. (2002) examined
drinking >  3 cups/day compared with those the association between coffee and exocrine
drinking 0 cups/day was 0.37; 95% CI, 0.16–0.88; pancreatic cancer in the Alpha-Tocopherol,
P for trend, 0.04), and no association in women Beta-Carotene (ATBC) Cancer Prevention Study
(RR, 0.88; 95% CI, 0.56–1.38; P for trend, 0.92). cohort in Finland among 27  111 male smokers
No associations between decaffeinated coffee or aged 50–69 years. Coffee intake was estimated
caffeine intake and pancreatic cancer, overall or with a self-administered FFQ given at baseline
by sex, were evident. [Data from the NHS and (1985–1988). Compared with those drinking
HPFS were included in the pooled analysis of ≤  321.4  mL/day of coffee, the relative risk for
Genkinger et al. (2012).] those drinking > 878.6 g/day was 0.95 (95% CI,

157
IARC MONOGRAPHS – 116

0.54–1.68; P for trend, 0.62). [The Working Group was modelled as a continuous variable, there was
noted that coffee consumption was not very low significant heterogeneity between filtered and
in the reference group. Data from this study were boiled coffee (P for trend, 0.013) with an elevated
included in the pooled analysis of Genkinger risk for boiled coffee.
et al. (2012).] Nakamura et al. (2011) evaluated the asso-
Khan et al. (2004) studied the association ciation between coffee consumption and risk of
between coffee drinking and pancreatic cancer death from pancreatic cancer in a prospective
mortality in a cohort study (1984–2002) in cohort study in Takayama, Japan. Coffee intake
Hokkaido, Japan, among 1524 men and 1634 was estimated with a self-administered FFQ
women aged 40 years and over at the beginning distributed at baseline. There was no significant
of the study period. Baseline coffee consump- association between intake of coffee and the risk
tion was assessed with a questionnaire. During of pancreatic cancer death; when comparing
follow-up until 2002, 25 fatal cases were detected. subjects drinking ≥  1 cup/day versus never
There was no significant association between drinkers of coffee, the relative risk was 0.44 (95%
coffee drinking and the incidence of pancreatic CI, 0.15–1.29; P for trend, 0.08) among men and
cancer in men or women. [The Working Group 0.68 (95% CI, 0.17–2.78; P for trend, 0.71) among
noted the extremely low number of cases in women. [The Working Group noted the very
sex-specific analyses.] small numbers of cases in sex-specific analyses.]
Luo et al. (2007) examined the association Genkinger et al. (2012) performed a pooled
between coffee drinking and the risk of pancre- analysis of primary data from 14 cohort studies
atic cancer in a large population-based cohort as part of the Prospective Studies of Diet and
study in Japan (JPHC study). A total of 233 inci- Cancer Pooling Project, a large international
dent cases of pancreatic cancer were identified. consortium. These studies included: the ATBC;
Baseline coffee consumption was assessed with a Breast Cancer Detection Demonstration Project
FFQ. Coffee drinking was not significantly asso- Follow-up Study (BCDDP); Canadian National
ciated with the risk of pancreatic cancer in men Breast Screening Study (CNBSS); Cancer
and women combined (P for trend, 0.4). Among Prevention Study II Nutrition Cohort (CPS
men, but not among women, there was a signif- II); California Teachers Study (CTS); Cohort
icant trend towards lower risk with increasing of Swedish Men (COSM); Health Professionals
coffee intake; the relative risk for ≥  3  cups/day Follow-up Study (HPFS); Iowa Women’s Health
versus rarely drinking coffee was 0.6 (95% CI, Study (IWHS); Melbourne Collaborative Cohort
0.3–1.1; P for trend, 0.04). Study (MCCS); the Netherlands Cohort Study
Nilsson et al. (2010) investigated total, filtered, (NLCS); New York State Cohort (NYSC); Nurses’
and boiled coffee consumption in relation to the Health Study (NHS); Prostate, Lung, Colorectal,
risk of incident cancer in a prospective cohort and Ovarian (PLCO) Cancer Screening Trial;
study from the ongoing, population-based and Swedish Mammography Cohort (SMC).
Västerbotten Intervention Project (VIP) estab- Baseline coffee consumption was measured
lished in 1985 in Sweden. Consumption of with FFQs as applied in each of the cohorts.
filtered and boiled coffee was assessed using a Estimated coffee intake levels were converted
FFQ. Total and filtered coffee were not associated into grams/day to avoid heterogeneity due to
with risk of pancreatic cancer, but boiled coffee different cup sizes between countries. Coffee
was positively associated with a relative risk of consumption was not associated with pancreatic
2.51 for ≥ 4 cups/day versus < 1 cups/day (95% CI, cancer risk overall, and there was no indication
1.15–5.50; P for trend, 0.006). When coffee intake of a dose–response association in categorical or

158
Drinking coffee

continuous analyses. When comparing intake P for trend, 0.88) for women, and 0.82 (95% CI,
of ≥ 900 g/day with 0 g/day, the pooled relative 0.38–1.76; P for trend, 0.95) for men and women
risk was 1.10 (95% CI, 0.81–1.48) with a P value combined.
of 0.08 in a test for between-study heterogeneity. Bhoo-Pathy et al. (2013) analysed the relation-
There was no indication of a differential associ- ship between coffee intake and pancreatic cancer
ation by sex (P value,  0.69 in test for between- in the EPIC cohort conducted in 10 European
study heterogeneity due to sex). The pooled countries: Denmark, France, Germany, Greece,
relative risks among women were 1.18 (95% CI, Italy, the Netherlands, Norway, Spain, Sweden,
0.71–1.98; P value in test for between-studies and the UK. The cohort included 477 312 partici-
heterogeneity,  0.01) and among men  0.95 (95% pants without cancer who completed a FFQ during
CI, 0.67–1.36; P value in test for between-studies 1992–2000 and were followed up for cancer inci-
heterogeneity,  0.83). Although not statistically dence. Estimated coffee intake from the FFQ was
significant, a suggestion of heterogeneity due to calibrated with a 24-hour recall. Median total
differences in the percentage of current smokers coffee intake ranged from 92  mL/day in Italy
in the female cohorts was present (P value for to 900  mL/day in Denmark. Consumption of
between-studies heterogeneity,  0.12). Expressed total coffee, caffeinated, and decaffeinated coffee
per increment of 237 mL/day, the pooled relative intake were not associated with risk of pancre-
risk was 1.01 (95% CI, 0.97–1.04) for women and atic cancer. For total coffee, the hazard ratio of
men combined with a P value for between-studies pancreatic cancer risk for the highest versus the
heterogeneity of 0.05. The large size of the pooled lowest quartile of consumption was 1.07 (95% CI,
analysis also permitted evaluation of the effect of 0.86–1.33; P for trend, 0.925). Hazard rations for
modification by other variables; however, there caffeinated and decaffeinated coffee were similar.
was no evidence of interaction by evaluated Continuous analyses for increments of 100 mL/
lifestyle or cohort characteristics. Among never day did not show any increase or decrease in
smokers (525 cases), the relative risk was 1.04 risk of pancreatic cancer for all coffee types. No
(95% CI, 0.95–1.15) per 237  mL/day. [The large material changes in risk estimates were observed
size of this pooled analysis of individual data when beverages were grouped using EPIC cohort-
with a high number of cases enabled analyses wide categories instead of country-specific
of broad exposure ranges and the possibility of intake. Associations between coffee intake and
evaluating effect modification.] pancreatic cancer were generally similar across
Bidel et al. (2013) examined the association subgroups as defined by sex, age group, smoking
between coffee and pancreatic cancer in a cohort status, and BMI categories.
study in six areas in Finland among 29 159 men Guertin et al. (2016) used data from the
and 30 882 women aged 25–74 years at baseline. National Institutes of Health–American
Coffee intake was estimated with a self-adminis- Association of Retired Persons (NIH-AARP)
tered questionnaire. Incident cancer cases were Diet and Health Study. At baseline, participants
identified through the country-wide Finnish were aged 50–71 years and resided in one of six US
Cancer Registry. Coffee consumption was not states or two metropolitan areas. For this analysis,
associated with an increased risk of pancreatic 457 366 participants (275 328 men and 182 038
cancer in men, women, or both sexes combined. women) with non-missing data on coffee intake
The hazard ratio of pancreatic cancer incidence and smoking were included. Cancer cases were
for ≥  10 cups/day of coffee compared with identified by linkage of the NIH-AARP cohort to
non-drinkers was 0.80 (95% CI, 0.30–1.95; P for 11 state cancer registries and the National Death
trend, 0.91) for men, and 0.71 (95% CI, 0.14–3.63; Index. Intakes of coffee and predominant type

159
IARC MONOGRAPHS – 116

of coffee consumed were assessed with a FFQ. interviewed by telephone. Participation was about
Although models adjusted only for age and sex 50% of eligible individuals in both control series.
suggested a statistically significant higher risk No significant associations were found between
of pancreatic cancer with higher coffee intake, pancreatic cancer and coffee drinking when
the association was substantially attenuated after using hospital- or population-based controls.
extensive adjustment for smoking. Adjustment The relative risks for those drinking ≥ 3 cups/day
for additional covariates did not appreciably alter versus 0 cups/day, while controlling for smoking
risk estimates. In the fully adjusted model, the status, were 1.68 (95% CI, 0.71–3.95) when using
hazard ratio of pancreatic cancer risk for men population controls and 1.52 (95% CI, 0.68–3.43)
drinking ≥ 6 cups/day of coffee versus 0 cups/day with hospital controls.
was 1.21 (95% CI, 0.84–1.75; P for trend,  0.55); A small study by Gorham et al. (1988) of 30
for women, the corresponding hazard ratio was cases and 47 controls was based only on death
1.38 (95% CI, 0.85–2.22; P for trend,  0.53). The certificates in Imperial County, California, USA,
association did not vary with tobacco smoking during 1978–1984. Controls were matched for
or self-reported history of diabetes. age, sex, race, and year of death; cancer patients
were excluded. The estimated relative risk for
2.2.2 Case–control studies pancreatic cancer mortality associated with
consumption of ≥  3  cups/day compared with
See Table 2.4 <  3 cups/day of coffee was 2.7, which dropped
(a) Population-based case–control studies to 1.9 and was non-significant after adjustment
for smoking. [The Working Group noted that
Severson et al. (1982) based their study on only 30 of 51 deaths from pancreatic cancer were
22 cases aged 40–79  years from a registry that included; hospital records were not examined.]
was part of the Surveillance, Epidemiology A case–control study in the USA involved 212
and End Results (SEER) Program in Seattle, cases identified from death certificates and 220
Washington, USA during 1977–1980, and on a population-based controls contacted by RDD
random population sample of controls (n = 485). and matched to cases by age within 5 years (Olsen
Next of kin were interviewed for most of the et al., 1989). Family members (usually widow
cases (20), whereas personal interviews were or spouse) were interviewed on the case’s use
obtained for controls. The odds ratio for current of cigarettes, alcohol, coffee, and other dietary
versus not current coffee drinking was 1.0 (95% factors 2 years before the death of the patient or
CI, 0.2–4.5). [This study was published as a letter, before interview for controls. Coffee intake was
which contained few details.] not associated with pancreatic cancer mortality
In the study of Gold et al. (1985), 201 cases (OR for ≥  7 cups/day versus <  1  cup/day, 0.60;
(94 men, 107 women) with pancreatic cancer 95% CI, 0.27–1.27.
from 16 hospitals in Baltimore, Maryland, USA Farrow & Davis (1990) conducted a case–
were included in a matched analysis. Of the control study with 148 cases and 188 controls
201, 25% had a personal interview. Two control among married men in Washington State, USA.
groups were used: a matched hospital series (for Cases residing in three counties of Washington
age, race, sex, hospital, date of admission) from State, aged 20–74 years at diagnosis, were identi-
which patients with other cancers were excluded; fied from the SEER Program. Population-based
and a population-based group that was chosen controls, matched to cases by age, were contacted
by random-digit dialling (RDD), matched by by RDD. Information about each man’s consump-
age, race, sex, and telephone exchange, and tion of coffee and other exposures was collected

160
161

Table 2.4 Case–control studies on cancer of the pancreas and drinking coffee

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
MacMahon Cases: 367 admitted to one Pancreas Coffee consumed (cups/day) Age, sex, Strengths: comparable
et al. (1981b) of 11 hospitals 0 20 1.0 smoking catchment area of cases and
USA, 1974–1979 Controls: 644 hospital- 1–2 153 1.8 (1.0–3.0) controls
based, other patients Limitations: many controls had
≥ 3 194 2.7 (1.6–4.7)
treated in same hospitals as gastrointestinal problems and
cases (excluding diseases Trend test P value, 0.001 may therefore have reduced
of biliary tract, pancreas, their coffee intake, response
CVD, diabetes, respiratory rates moderate, interviewers not
or bladder cancer, peptic blinded for case/control status
ulcer)
Exposure assessment
method: interview
Severson et al. Cases: 22 from SEER Pancreas Coffee drinking status Age, sex, Strengths: population-based
(1982) registry in Seattle, aged Not current NR 1.0 smoking study
USA, 1977–1980 40–79 yr at diagnosis Current NR 1.0 (0.2–4.5) Limitations: very small number
Controls: 485 population- of cases, cases information from
based, randomly selected two living patients and 20 from
from population in which next-of-kin because of death,
cases arose, aged 40–79 yr limited exposure information
Exposure assessment
method: interview
Wynder et al. Cases: 275 aged 20–80 yr, Pancreas Coffee consumed (cups/day): men Age, smoking Strengths: relatively large
(1983) admitted to 17 hospitals in 6 0 26 1.00 series with detailed control for
USA, 1977–1981 major cities 1 15 0.80 (0.40–1.48) smoking
Controls: 7994 hospital- Limitations: hospital-based
2 34 1.10 (0.68–1.95)
based controls, matched on controls, reduced response rates
age, race, sex, room status 3–5 50 1.00 (0.59–1.59) in cases and controls
from same hospital as cases ≥6 28 1.00 (0.59–1.79)
(diseases, some cancers, not Coffee consumed (cups/day): women Age, smoking
associated with tobacco) 0 25 1.0

Drinking coffee
Exposure assessment 1 19 0.90 (0.48–1.64)
method: interview
2 25 0.90 (0.51–1.59)
3–5 36 0.90 (0.53–1.50)
≥ 6 17 1.00 (0.52–1.83)
161
162

IARC MONOGRAPHS – 116
Table 2.4 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Kinlen & Cases: 216 aged > 40 yr, Pancreas Coffee drinking status: men Age, tea, Strengths: adjustment for tea
McPherson derived from an earlier Never 69 1.00 smoking Limitations: hospital-based,
(1984) study by Stocks (1957) Weekly 22 0.87 (0.48–1.54) little information about
UK, 1952–1954 conducted in 1952–1954 in cases, no information about
Daily 18 0.93 (0.49–1.76)
greater Liverpool area and response rates, limited exposure
north Wales Coffee drinking status: women Age, smoking, information
Controls: 432 hospital- Never 55 1.00 tea
based, cancer controls from Weekly 29 1.28 (0.71–2.28)
Stocks study (excluding Daily 23 0.86 (0.86–1.58)
smoking-related and GI
tract cancer, and ovarian
cancer), matching on sex,
age, residence area
Exposure assessment
method: interview
Gold et al. Cases: 201 from 16 major Pancreas Coffee consumed (cups/day): population Age, sex, Strengths: relatively large case
(1985) hospitals in Baltimore area controls smoking series with two types of control
USA, 1978– Controls: 201 population- 0 18 1.00 groups
1980 based, matched by age, race, 1–2 91 1.37 (0.59–3.18) Limitations: large difference in
sex and telephone exchange, proportion of proxy interviews
≥3 88 1.68 (0.71–3.95)
plus 201 hospital-based between cases (75%) and
(other cancers excluded) Coffee consumed (cups/day): hospital controls controls (0%), different response
controls matched for age, 0 18 1.00 rates between cases and controls
race, sex, hospital, date of 1–2 91 1.43 (0.65–3.14)
admission ≥3 88 1.52 (0.68–3.43)
Exposure assessment
method: interview (often
with next of kin)
Table 2.4 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Falk et al. Cases: 363 incident cases Pancreas Coffee consumed (cups/day): women Age, smoking, Strengths: questionnaire instead
(1988) from hospitals in Louisiana 0 32 1.00 alcohol of interview
USA, 1979– Controls: 1234 admitted 1–2 58 0.67 consumption, Limitations: hospital-based,
1983 to same hospital as cases, intake of fruit, interview for 50% of cases and
3–4 35 0.69
matched on sex, age, race income 13% of controls through next of
Exclusions: chronic 5–7 15 0.96 kin (potential for recall bias)
conditions (cancers, ≥ 8 20 0.92
diabetes, CVD, digestive Coffee consumed (cups/day): men
diseases, respiratory 0 34 1.00
diseases) suspected to be 1–2 64 0.66
related to lifestyle or diet
3–4 34 0.53
Exposure assessment
method: questionnaire 5–7 23 0.67
≥ 8 48 1.39
Gorham et al. Cases: 30 fatal pancreatic Pancreas Coffee consumed (cups/day) Age, smoking Strengths: comparison of fatal
(1988) cancer cases identified < 3 7 1.0 cases with dead controls should
USA, 1978– from death certificates in ≥ 3 16 1.9 lead to less information bias,
1984 Imperial County, California interviewers blinded with
Controls: 47 controls respect to cause of death
identified from death Limitations: only 30 of 51 deaths
certificates (excluding from pancreatic cancer were
deaths from cancer), included, hospital records were
matching on age, sex, race not examined, information
and year of death from next of kin, median length
Exposure assessment of time between death and date
method: questionnaire of interview was 6 yr in cases
and controls

Drinking coffee
163
164

IARC MONOGRAPHS – 116
Table 2.4 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Clavel et al. Cases: 161 cases (98 men) Pancreas Coffee consumed (cups/day): women Age, ethnicity, Unusually high risks were seen
(1989) with diagnosed cancer 0 4 1.00 education, in women and in persons who
France, of exocrine pancreas 1 24 3.94 (0.85–18.22) alcohol had never drank alcohol.
1982–1985 from public hospitals in consumption, Strengths: study of interaction
2–3 29 6.71 (1.47–30.65)
Paris (102 of 161 cases smoking with alcohol
histologically verified); ≥ 4 6 9.56 (1.29–70.71) Limitations: hospital–based,
mean age at diagnosis was Trend test P value, 0.006 interviewers not blinded,
62 yr in men and 64 yr in Coffee consumed (cups/day): men proportion of subjects born
women 0 6 1.00 outside France was higher
Controls: 268 hospital- 1 35 1.07 (0.30–3.88) among cases than controls (but
based controls, matched was adjusted for in analyses),
2–3 44 1.45 (0.41–5.04)
on age, sex, hospital, possible interview bias in study
interviewer; 129 controls ≥ 4 15 2.08 (0.49–8.86) period due to widely publicized
had other cancers Trend test P value, 0.14 study by MacMahon et al.
(excluding biliary, liver, (1981b)
stomach, oesophagus,
respiratory and bladder
cancers) and 139 had non-
neoplastic disorders
Exposure assessment
method: interview
Cuzick & Cases: 216 cases (30% Pancreas Coffee consumed currently (cups/day) Age, smoking, Strengths: analyses of coffee
Babiker (1989) histologically verified) from 0 97 1.00 sex consumption 10 yr previously
UK, 1983–1986 Leeds, London, Oxford 1–2 77 0.87 Limitations: mostly hospital-
Controls: 279, mix of based
3–4 19 0.63
hospital-based (212) and
population-based (67) ≥ 5 23 1.37
controls from same three Trend test P value, 0.23
areas Coffee consumed 10 yr previously (cups/day)
Exposure assessment 0 117 1.00
method: questionnaire 1–2 69 0.93
3–4 18 0.85
≥ 5 12 0.77
Trend test P value, 0.43
Table 2.4 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Olsen et al. Cases: 212 aged 40–84 Pancreas Coffee consumed (cups/day) Age, smoking, Strengths: dead cases are
(1989) yr identified from death < 1 29 1.00 education, compared with dead controls,
USA, 1980– certificates in Minneapolis– 1–3 60 0.50 (0.26–1.00) diabetes, meat comparable information more
1983 St Paul area intake, intake of likely
4–6 74 0.72 (0.37–1.45)
Controls: 220 population- vegetables Limitations: information
based white men contacted ≥ 7 49 0.60 (0.27–1.27) obtained from next of kin
by RDD, matched to cases
by age within 5 yr
Exposure assessment
method: FFQ
Farrow & Davis Cases: 148 men from SEER, Pancreas Coffee consumed (cups/day) Age, smoking, Strengths: surrogate interviews
(1990) Washington State, aged 0 18 1.0 race, education, for all cases and controls,
USA, 1982– 20–74 yr 1–2 27 0.7 (0.3–1.7) energy-adjusted comparable information more
1986 Controls: 188 population- intake of protein likely
3–5 55 1.0 (0.4–2.2)
based controls contacted and calcium Limitations: information
by RDD, matched to cases ≥ 6 62 1.1 (0.5–2.4) obtained from next of kin,
by age interviews were held 2.0–4.5 yr
Exposure assessment after the diagnosis
method: interview

Drinking coffee
165
166

IARC MONOGRAPHS – 116
Table 2.4 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Jain et al. (1991) Cases: 249 diagnosed in 20 Pancreas Lifetime coffee consumption (cup-years) Age, sex, Further analysis by type
Canada, hospitals in Toronto 0 25 1.00 smoking, of coffee (regular, instant,
1983–1986 Controls: 505 population- ≤ 39 69 0.94 (0.47–1.89) residence, proxy/ caffeinated, decaffeinated) also
based, matched by sex and direct interview, showed no evidence of an effect.
40–110 76 0.90 (0.45–1.79)
age from population lists energy intake, Strengths: relatively large study
Exposure assessment ≥ 110 76 0.90 (0.44–1.81) fibre with dietary history interview;
method: diet history Continuous 229 0.96 (0.77–1.19) lifetime history estimates
interview for 100 cup- of coffee, tea and alcohol
years consumption
Limitations: low response rates,
interview 3 mo after diagnosis
with high case fatality rate,
different proportions of cases
and controls interviewed by
proxy (possibly leading to bias),
194 of 249 cases interviewed by
proxy (62% with spouse, 31%
with daughters and sons, and
7% with others), 194 of 505
controls interviewed by proxy
(72% with spouse, 19% with
daughters and sons, and 9%
with others)
Ghadirian et al. Cases: 179 aged 35–79 yr, Pancreas Cumulative lifetime coffee consumption Age, sex, Further analysis by type
(1991) diagnosed in 19 hospitals Quintile 1 NR 1.00 smoking, of coffee (regular, instant,
Canada, located in greater Montreal Q2 vs Q1 NR 0.44 education, caffeinated, decaffeinated) also
1984–1988 Controls: 239 population- respondent type showed no evidence of an effect.
Q3 vs Q1 NR 0.82
based matched for age, Strengths: lifetime coffee
sex, and place of residence Q4 vs Q1 NR 0.51 drinking and coffee drinking
selected randomly from Q5 vs Q1 NR 0.55 (0.19–1.62) patterns (e.g. with meals) were
RDD Trend test P value, 0.53 studied
Exposure assessment Limitations: large difference
method: questionnaire, in proportion of interviews by
interviews proxy between cases (75%) and
controls (17%)
Table 2.4 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Bueno de Cases: 176 aged 35–79 Pancreas Cumulative lifetime coffee consumption (L) Age, sex, The suggestion of an inverse
Mesquita et al. yr in central part of the < 6 193 26 1.00 smoking, dose–response relationship
(1992) Netherlands < 9 012 23 0.72 (0.36–1.43) respondent with the lifetime consumption
Netherlands, Controls: 487 population- type, energy of coffee was not present in the
< 11 840 17 0.37 (0.18–0.79)
1984–1987 based controls aged intake, intake of analysis of direct responders
35–79 yr from municipal ≥ 11 840 24 0.58 (0.28–1.20) vegetables, tea only. Further analysis by type
population registries in Trend test P value, 0.06 of coffee (regular, instant,
the same area, frequency caffeinated, decaffeinated)
matched to the age-and-sex showed no evidence of an
distribution of the cases association.
Exposure assessment Strengths: lifetime coffee
method: interviewer- drinking
administered questionnaire Limitations: possible selection
on lifetime frequency bias due to relatively large
difference in response rate
between cases and controls and
different proportion of proxy
interviews between cases (42%)
and controls (29%)
Lyon et al. Cases: 149 with pancreatic Pancreas Cumulative lifetime coffee consumption (cups) Age, sex, Strengths: for all cases and
(1992) adenocarcinoma or 0–2000 38 1.00 smoking, controls, surrogate interviews
USA, 1984– carcinoma, from Utah 2001–50 000 44 1.34 (0.78–2.29) religion were held with next of kin
1987 Cancer Registry, aged 40–79 (comparable information more
≥ 50 000 40 2.38 (1.16–4.85)
years likely)
Controls: 363 population- Trend test P value, < 0.001 Limitations: non-response rate
based controls, frequency among controls was higher
matched to the distribution than among cases, surrogate
of cases by age, sex, and information obtained from next
county of residence at the of kin (information less reliable)
time of diagnosis

Drinking coffee
Exposure assessment
method: questionnaire,
telephone interview with
proxies
167
168

IARC MONOGRAPHS – 116
Table 2.4 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Zatonski et al. Cases: 110 identified Pancreas Cumulative lifetime coffee consumption (L) Age, sex, Strengths: substantial
(1993) through hospitals and the 0 58 1.00 smoking, proportion of never drinkers of
Poland, Cancer Registry located < 417 17 0.61 (0.30–1.23) education coffee
1985–1988 in the Opole Voivodeship Limitations: large difference
< 1916 18 0.63 (0.30–1.30)
Oncological Clinic in proportion of proxy
Controls: 195 population- ≥ 1916 16 0.48 (0.22–1.02) interviews between cases (71%)
based controls from same Trend test P value, 0.042 and controls (0%) leading to
area, frequency matched on information bias, few subjects
age, sex, place of residence drinking large amounts of
Exposure assessment coffee
method: interviewer-
administered questionnaire
on lifetime frequency of
the consumption of specific
beverages per age period
Partanen et al. Cases: 662 identified at the Pancreas Coffee consumed 20 yr previously (cups/day) Age, sex, Consumption of coffee is high
(1995) Finnish Cancer Registry None/ 24 1.00 smoking in Finland, with few people who
Finland, Controls: 1770 from Finnish occasional never or occasionally drink
1984–1987 Cancer Registry (1014 1–3 104 0.83 (0.50–1.38) coffee. ORs were lower (but NS)
stomach, 441 colon, 315 when rectum cancers were used
4–6 273 0.96 (0.59–1.56)
rectum cancer) as controls only, as opposed to
Exposure assessment > 6 91 0.71 (0.41–1.20) colon cancer controls only (OR
method: questionnaire, close to 1).
mail questionnaire, coffee Strengths: size, surrogate
use 20 yr before diagnosis interviews were held with next
considered, obtained from of kin for all cases and controls
next of kin (comparable information more
likely)
Limitations: use of cancer
controls possibly related to
coffee consumption, surrogate
information obtained from
next of kin (information less
reliable), response rates in cases
or controls were not provided
Table 2.4 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Nishi et al. Cases: 141 pancreas cancer Pancreas Coffee consumed (cups/day): men Age, smoking Reports a U-shape curve, with
(1996) diagnosed at Sapporo 0 NR 1.00 extra meta-analyses.
Japan, 1987– Medical University and its Occasionally NR 0.18 (0.07–0.43) Strengths: population-based
1992 affiliated hospitals Limitations: cases were
1–2 NR 0.53 (0.27–1.07)
Controls: 282 population- interviewed but controls
based controls from ≥ 3 NR 0.93 (0.44–1.96) received a questionnaire
Hokkaido, matched for sex, Coffee consumed (cups/day): women (possibly leading to information
age and place of residence 0 NR 1.00 bias), limited control for
Exposure assessment Occasionally NR 0.53 (0.20–1.38) confounders
method: cases interviewed 1–2 NR 0.70 (0.31–1.58)
and controls received a
≥3 NR 1.37 (0.46–4.14)
questionnaire
Silverman et al. Cases: 436 among Pancreas Coffee consumed (cups/day): men Age, race, study Strengths: size, high proportion
(1998) 30–79-year-old residents ≤ 1 53 1.0 area, smoking, of direct interviews
USA, 1986– of areas covered by cancer 2 57 1.1 (0.7–1.7) alcohol
1989 registries in Atlanta, consumption,
3 31 1.0 (0.6–1.7)
Detroit, and 10 New Jersey diabetes, BMI,
counties 4–5 23 0.8 (0.4–1.4) energy intake,
Controls: 2003, random ≥ 6 28 0.9 (0.5–1.7) cholecystectomy,
sample from general Non-drinker 26 1.0 income
population, frequency (reference)
matched on age, race, sex, Ever 192 0.9 (0.5–1.4)
and study area Coffee consumed (cups/day): women
Exposure assessment
≤ 1 65 1.0
method: interview
(sometimes with next of 2 52 1.0 (0.7–1.6)
kin) with FFQ 3 26 0.7 (0.4–1.1)
4–5 32 1.0 (0.6–1.7)
≥ 6 15 1.0 (0.5–2.2)
Non-drinker 23 1.0

Drinking coffee
(reference)
Ever 190 1.4 (0.9–2.4)
169
170

IARC MONOGRAPHS – 116
Table 2.4 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Villeneuve et al. Cases: 583 aged 30–76 yr Pancreas Coffee consumed: men Age, province Proxy interviews for 24% of
(2000) from eight provincial cancer < 3 cups/mo 34 1.00 of residence, cases but 0% of controls.
Canada, registries confirmed 1–6 cups/wk 33 1.23 (0.71–2.13) smoking, alcohol Strengths: large study
1994–1997 Controls: 4813 population- consumption, Limitations: large difference in
1 cup/day 33 0.70 (0.40–1.22)
based, frequency matched energy intake, proportion of proxy interviews
on age and sex 2–3 cups/ 124 1.11 (0.72–1.71) fat intake between cases and controls,
Exposure assessment day leading to information bias
method: FFQ ≥ 4 cups/day 91 1.23 (0.78–1.97)
Coffee consumed (cups/day): women Age, province
< 3 cups/mo 43 1.00 of residence,
1–6 cups/wk 29 0.90 (0.52–1.57) smoking, alcohol
consumption,
1 cup/day 40 1.00 (0.61–1.65)
energy intake,
2–3 cups/ 85 0.81 (0.53–1.33) fat intake,
day number of live
≥ 4 cups/day 55 1.02 (0.63–1.66) births
Turati et al. Cases: 688, pooling of data Pancreas Coffee consumed (cups/day)  Age, sex, Includes results from La
(2011a) from two hospital-based 0 78 1.00 smoking, year Vecchia et al. (1987) by pooling
Italy, 1983–2008 case–control studies in ≤ 1 171 1.41 (1.02–1.94) of enrolment, two case–control studies.
Milan (362 cases, 1983– education, No heterogeneity by age, sex,
≤ 2 199 1.29 (0.94–1.77)
1992) and Pordenone (326 BMI, alcohol smoking, other covariates. No
cases, 1992–2008) ≤ 3 133 1.23 (0.88–1.72) consumption, association with decaffeinated
Controls: 2204, hospital- > 3 107 1.46 (1.02–2.10) diabetes coffee.
based controls (admitted to Continuous 610 1.05 (0.98–1.11) Strengths: large pooled study
the same hospitals as cases for 1 cup/ with investigation of effect
for acute conditions other day modifiers
than neoplasia or diseases Trend test P value, 0.232 Limitations: hospital-based
of the digestive tract), controls
frequency matched with
cases by age and sex
Exposure assessment
method: questionnaire
Table 2.4 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Azeem et al. Cases: 309 (180 men, Pancreas All types of coffee consumed Age, sex, Limitations: interviewers were
(2013) 129 women) from three 0–1 cup/wk 53 1.00 smoking, BMI, not blinded, no indication of
Czech Republic, hospitals in three regions > 1 cup/wk – 202 1.02 (0.60–1.75) education, the cancer diagnosis method,
2006–2009 Controls: 220 (123 men, 97 2 cups/day physical response rates unknown
women) population-based, activity, alcohol
≥ 3 cups/day 38 0.78 (0.36–1.66)
matched on age, sex, health consumption,
status and region tea
Exposure assessment
method: questionnaire,
interview, measurements of
anthropometric data
BMI, body mass index; CI, confidence interval; CVD, cardiovascular disease; FFQ, food frequency questionnaire; GI, gastrointestinal; mo, month(s); NR, not reported; NS, not
significant; OR, odds ratio; RDD, random-digit dialling; SEER, Surveillance, Epidemiology and End Results; vs, versus; wk, week(s); yr, year(s)

Drinking coffee
171
IARC MONOGRAPHS – 116

in a telephone interview with his wife. Coffee evident in analyses by type of coffee consumed.
was not significantly associated with pancre- The authors noted that proxy respondents
atic cancer risk; the odds ratio for ≥ 6 cups/day reported higher amounts of total coffee intake
versus 0 cups/day was 1.1 (95% CI, 0.5–2.4). [The compared with direct respondents for all subjects
Working Group noted that the deliberate use of combined. [The Working Group noted a large
surrogate interviewees enhanced comparability difference in proportion of interviews by proxy
of information of cases and controls; neverthe- between cases (75%) and controls (17%), leading
less, both could have suffered from misclassifica- to possible information bias.]
tion. This problem may have been aggravated as a Bueno de Mesquita et al. (1992) conducted
result of the long period (2–4.5 years) between the a case–control study on pancreatic cancer and
times of diagnosis and interviews with spouses, coffee consumption in the Netherlands as part of
who were required to recall exposure details of IARC-SEARCH. Pancreatic cancer cases (alive or
more than 3 years before diagnosis.] dead) were 35–79 years of age, newly diagnosed
Jain et al. (1991) described results obtained between 1984 and 1987, and living in the central
in a population-based case–control study carried part of the Netherlands at the time of diagnosis
out in Toronto, Canada, as part of the IARC- of cancer of the exocrine pancreas. Population-
SEARCH programme. A quantitative diet history based controls were obtained from municipal
was used to estimate the lifetime consumption population registries in the area and matched to
of different types of coffee for 249 cases and 505 the age–sex distribution of the cases. A quanti-
controls. A total of 194 cases were interviewed tative diet history was used to estimate the life-
by proxy. A proxy control was obtained for each time consumption of total coffee and of different
case interviewed by proxy. Odds ratio estimates types of coffee for 176 cases and 487 controls. The
for quartiles of coffee consumption or per 100 results for lifetime drinking of coffee indicated
cup-years increment showed no evidence of an an inverse dose–response association between
association between coffee intake and pancreatic coffee intake and risk of pancreatic cancer, with
cancer risk. The odds ratio for ≥  110  cup-years the test for trend approaching statistical signif-
versus 0 cup-years was 0.90 (95% CI, 0.44–1.81). icance (P for trend, 0.06). The odds ratio for
The odds ratio for an increment of 100 cup-years ≥ 11 840 L coffee per life versus < 6193 L coffee
was 0.96 (95% CI, 0.77–1.19). Further analysis by per life was 0.58 (95% CI, 0.28–1.20). The sugges-
type of coffee (regular, instant, caffeinated, and tion of an inverse dose–response relationship
decaffeinated) also showed no evidence of an with the lifetime consumption of coffee was not
association. present in the analysis of direct responders only.
Ghadirian et al. (1991) described results from [The Working Group noted that possible selec-
another Canadian case–control study that was tion bias may have occurred due to relatively
part of IARC-SEARCH. A total of 179 cases, large differences in the response rate between
aged 35–79 years, were diagnosed in 19 hospitals cases and controls. The different proportions
located in Greater Montreal. Population-based of proxy interviews between cases and controls
controls (239) matched for age, sex, and place of (42% versus 29%) could also contribute to infor-
residence were selected by the RDD method or mation bias.]
randomly from the telephone directory. There Lyon et al. (1992) conducted a population-based
was an inverse association (P for trend,  0.53) case–control study of 149 cases of cancer of the
between cumulative lifetime coffee consumption exocrine pancreas (excluding insulinomas) and
in quintiles and pancreatic cancer risk (Q5 vs Q1 363 controls in Utah, USA. All information was
OR, 0.55; 95% CI, 0.19–1.62). Similar results were obtained from proxy respondents for cases and

172
Drinking coffee

controls. Pancreatic cancer risk increased with Nishi et al. (1996) conducted a case–control
the amount of coffee drunk with an odds ratio study in Hokkaido, Japan, employing 141 cases
of 2.38 (95% CI, 1.16–4.85) for those having at with cancer of the pancreas and 282 controls
least 50 000 lifetime cups (P for trend, < 0.001) (2 for each case) matched for sex, age, and place of
compared with those having 0–2000 lifetime residence. This is an update of an earlier study by
cups. Positive associations were also observed Goto et al. (1990). To estimate coffee intake, cases
for users of regular and decaffeinated coffee, but were interviewed by a trained interviewer while a
were stronger in magnitude for users of decaf- ‘self-rating questionnaire’ was distributed to the
feinated coffee than users of regular coffee. [The controls. Consumption of coffee was not signifi-
Working Group noted many limitations of this cantly associated with risk of pancreatic cancer;
study. The non-response rate among controls the odds ratio for ≥ 3 versus 0 cups/day was 0.93
(23%) was higher than among cases (12%), which (95% CI, 0.44–1.96) among men and 1.37 (95% CI,
might have led to selection bias. Since all infor- 0.46–4.14) among women. [The Working Group
mation was obtained from proxy respondents, noted that cases were interviewed but controls
it is possible that there was a difference in the received a questionnaire, possibly leading to
type of proxy respondents available for the cases information bias. There was also limited control
compared with the controls. Approximately 5% for confounders.]
more spouses were available as proxies for the Silverman et al. (1998) conducted a popu-
controls than for the cases, whereas about 7% lation-based case–control study of pancreatic
more children or children’s spouses were avail- cancer diagnosed in Atlanta, Detroit, and in 10
able as proxies for the cases than for the controls, New Jersey counties, USA, from August 1986
possibly resulting in information bias.] to April 1989. Reliable dietary histories were
Zatonski et al. (1992) conducted a case– obtained for 436 patients and 2003 general-pop-
control study on the association between pancre- ulation control subjects aged 30–79  years. Men
atic cancer and coffee consumption in Poland who were regular coffee drinkers experienced
as part of IARC-SEARCH. Of the 110 cases, 32 no overall increased risk, whereas women who
were interviewed directly and a proxy interview were regular drinkers had a non-significant 40%
was available for 78. All 195 controls were inter- increased risk of pancreatic cancer as compared
viewed directly following the very low acceptance with non-drinkers of coffee. Among coffee
rate among proxy controls found in a pilot study. drinkers, neither a gradient in risk with increasing
Lifetime coffee drinking was estimated for total amount of coffee consumed or increased risk
coffee and different types of coffee. Compared with any amount of consumption was observed
with never drinkers of coffee, the odds ratio of for either men or women.
risk of pancreatic cancer for ≥ 1916  L of coffee Villeneuve et al. (2000) conducted a popu-
per life was 0.48 (95% CI, 0.22–1.02). A signifi- lation-based case–control study of pancreatic
cant trend test (P for trend, 0.042) was observed, cancer diagnosed in eight Canadian provinces as
which remained when the analyses were limited part of the Canadian National Enhanced Cancer
to directly interviewed subjects only and when Surveillance System (NECSS) project. Cases
consumption of tea was additionally adjusted for. (n = 583) aged 30–76 years were identified from
[The Working Group noted a large difference in eight provincial cancer registries. Population-
the proportion of proxy interviews between cases based controls (4813), frequency-matched for
and controls, which may have led to information age and sex, were selected from health insur-
bias.] ance plans using stratified random sampling or
RDD, depending on province. Coffee intake was

173
IARC MONOGRAPHS – 116

estimated using a FFQ. Among cases, 24% were but these estimations were not adjusted for
proxy interviews with next of kin; among controls smoking. [The Working Group noted that many
the corresponding percentage was 0. Coffee controls had gastrointestinal problems, meaning
intake was not significantly associated with that subjects may have reduced their coffee
pancreatic cancer risk in either men or women. intake to relieve symptoms. For this reason,
The odds ratio for ≥ 4 cups/day versus < 3 cups/ the Working Group judged that the observed
month in men was 1.23 (95% CI, 0.78–1.97); in positive associations might have been spurious
women the respective association was 1.02 (95% effects due to selection bias.]
CI, 0.63–1.66). [The Working Group noted a large A study (part of a larger study of tobacco-re-
difference in proportion of proxy interviews lated cancers in six US cities) of 275 histologi-
between cases and controls, which may have led cally verified cases (153 men, 122 women) aged
to information bias.] 20–80 years, interviewed during 1977–1981, and
Azeem et al. (2013) conducted a popula- of 7994 hospital controls reported null associ-
tion-based case–control study (529 subjects, 303 ations between risk of pancreatic cancer and
men and 226 women, period of study 2006–2009) coffee intake (Wynder et al., 1983). Controls were
of lifestyle factors and risk of pancreatic cancer patients with diseases not related to tobacco.
in the Czech Republic. Newly diagnosed cases of Personal interviews were carried out within
pancreatic cancer (n = 309) were recruited from 6 months of diagnosis. The study found no asso-
three hospitals. [The Working Group noted that ciation between coffee consumption and pancre-
no information on how the diagnosis of pancreatic atic cancer. [The Working Group noted that the
cancer was established was provided.] Controls low response rate among cases and controls may
(n  =  220) were a population-based sample of have resulted in selection bias.]
individuals from the same regions as cases. Kinlen & McPherson (1984) re-evaluated data
After adjustment for other factors, no trend was from the case–control study of Stocks (partly
observed with respect to the amount of coffee reported by Stocks, 1957) collected from hospi-
consumption for ≥  3 cups/day compared with tals in north-western England and north Wales
0 to ≤ 1 cup/week (OR, 0.78; 95% CI, 0.36–1.66). during 1952–1954, including 216 cases (109 men,
107 women) aged > 40 years. These were compared
(b) Hospital-based case–control studies with 432 controls who were patients with other
MacMahon et al. (1981a, b; the latter study was cancers in the original study matched for age, sex,
reported in a letter) reported on a case–control and area of residence; patients with cancers of
study of 367 (216 men, 151 women) subjects the lung, bladder, mouth, pharynx, oesophagus,
with cancer of the pancreas (excluding islet cell gastrointestinal tract, and ovary were excluded.
tumours) under 80 years of age identified in 11 No association between pancreatic cancer risk
hospitals in Boston and Rhode Island, USA, and and coffee consumption was found either before
644 controls who had been at hospital for other or after adjustment for smoking.
diseases at the same time as the cases. Each case A case–control study by Falk et al. (1988),
and control pair was interviewed personally by based on 363 incident cases (203 men, 160
the same physician. Compared with non-drinkers women) and 1234 hospital controls, was carried
of coffee, the relative risks for those drinking out in Louisiana, USA. Control subjects were
1–2 cups/day and ≥ 3 cups/day were 1.8 (95% CI, matched for hospital, age, sex, and race. Patients
1.0–3.0) and 2.7 (95% CI, 1.6–4.7), respectively with cancer, diabetes, circulatory disorders, and
(P for trend, 0.001). Elevated relative risks were digestive or respiratory diseases were excluded
also reported among men and women separately, from the pool of potential controls. Direct

174
Drinking coffee

interviews were carried out with 50% of cases and consumption approximately 10 years before the
50% were with next of kin. For controls, direct interview was examined.
interviews were with 13%. No association was Partanen et al. (1995) conducted a case–
found between coffee drinking (any amount) and control study using pancreatic cancer deaths as
risk of pancreatic cancer for men or women after cases and patients with cancers other than that
adjusting for age, residence, smoking, alcohol, of the pancreas as controls during 1984–1987 in
fruit consumption, diabetes, and income. [The Finland, a country with very high coffee consump-
Working Group noted the high proportion of tion. Cases and controls were identified from the
proxy interviews, especially among controls.] Finnish Cancer Registry: 662 endocrine pancreas
Clavel et al. (1989) conducted a hospital-based cancer cases and 1770 controls (1014 stomach,
interview study in Paris, France, with 161 cases of 441 colon, and 315 rectum cancer). Using a
cancer of the pancreas (98 men, 63 women) during mail questionnaire, data on coffee consumption
1982–1985. There were 268 hospital controls, 129 20 years before diagnosis were obtained from next
of which had other cancers (excluding biliary, of kin. There was no association between coffee
liver, stomach, oesophagus, respiratory, and consumption and pancreatic cancer mortality;
bladder cancers) and 139 of which had non-neo- the odds ratio for those drinking >  6 cups/day
plastic disease. All were matched to cases for age, compared with never/occasional coffee drinkers
sex, hospital, and interviewer. None of the cases was 0.71 (95% CI, 0.41–1.20). Odds ratios were
and about 5% of controls refused to participate. lower (but non-significant) when rectum cancers
After adjustment for education, alcohol, and were used as controls only, as opposed to colon
smoking, a non-significant trend for pancreatic cancer controls only (ORs close to 1).
cancer was observed among men with a rela- Turati et al. (2011a) performed a pooled
tive risk of 2.08 for ≥ 4 cups/day compared with analysis of two earlier case–control studies from
0  cups/day (95% CI, 0.49–8.86). In women, the northern Italy, conducted between 1983 and
respective trend was statistically significant and 2008, including a total of 688 cases of cancer
the corresponding relative risk was 9.56 (95% of the pancreas and 2204 hospital controls with
CI, 1.29–70.71). [The Working Group noted that acute, non-neoplastic diseases. The first study,
unusually high relative risks were seen in women conducted during 1983–1992 in Milan, included
and in persons who had never drunk alcohol, 362 incident cases of pancreatic cancer (229
possibly due to interview bias from publicity men, 133 women) and 1552 controls and is an
about the topic.] update of an earlier study by La Vecchia et al.
A study of 216 cases of cancer of the pancreas (1987) and Soler et al. (1998). The second study,
(123 men, 93 women) and 279 controls was carried conducted between 1992 and 2008 in Milan and
out in the UK during 1983–1986 (Cuzick & Pordenone, northern Italy, included 326 incident
Babiker, 1989) based on personal interviews. The cases (174 men, 152 women) and 652 controls,
controls included 212 hospital controls without frequency-matched with cases by age and sex
cancers or other chronic medical conditions, (Rossi et al., 2010). In both studies, controls were
and the remaining 67 were population-based admitted to the same network of hospitals as
controls. The study reported essentially null cases for a wide spectrum of acute conditions
associations between pancreatic cancer risk and other than neoplasia or diseases of the diges-
coffee consumption, although a slightly elevated tive tract. Less than 5% of cases and controls
risk was seen in cases whose current consump- refused to participate in the interview. Cases
tion was ≥ 5 cups/day (RR, 1.4) as compared with and controls were interviewed using a structured
0 cups/day. This trend disappeared when coffee questionnaire regarding frequency of coffee

175
IARC MONOGRAPHS – 116

consumption. Compared with non-drinkers of between-study heterogeneity (P  =  0.002). This
coffee, the odds ratio for coffee drinkers was heterogeneity was not explained by study design,
1.34 (95% CI, 1.01–1.77). The odds ratio for sex, or geographic location. The summary relative
those drinking >  3  cups/day was 1.46 (95% CI, risk was 1.00 (95% CI, 0.83–1.19) for men and 1.15
1.02–2.10) compared with coffee non-drinkers. (95% CI, 0.94–1.41) for women when combining
However, there was no trend in risk of pancre- all smoking-adjusted studies (P heterogeneity
atic cancer with respect to dose (cups/day) (P for between sexes, 0.312). Per increment of 1 cup/day
trend, 0.232). The odds ratio for an increment of of coffee based on the smoking-adjusted studies,
1 cup/day was 1.05 (95% CI, 0.98–1.11). There was the summary relative risk was 1.04 (95% CI,
no heterogeneity in the apparent associations 1.00–1.09) for case–control studies and 1.00 (95%
in strata defined by age, sex, and other covari- CI, 0.95–1.05) for cohort studies. The authors
ates, including tobacco smoking. No association estimated a weak positive association between
emerged for drinkers of decaffeinated coffee coffee consumption and pancreatic cancer risk
compared with non-drinkers of decaffeinated when combining case–control studies that were
coffee (OR, 0.87; 95% CI, 0.60–1.26). not adjusted for tobacco, which can be attributed
to residual confounding by smoking.
2.2.3 Meta-analyses
Meta-analyses of cohort studies on the asso- 2.3 Cancer of the liver
ciation between coffee consumption and cancer
A total of 14 cohort and 11 case–control
of the pancreas were conducted by Dong et al.
studies that examined the association between
(2011), Yu et al. (2011), and Ran et al. (2016); these
coffee consumption and the risk of cancer of the
meta-analyses included studies that did not adjust
liver were available for review by the Working
for smoking, however, and also excluded several
Group.
studies. Because of the shortcomings of these
Regarding the cohort studies, seven were
meta-analyses, the Working Group focused on
conducted in Japan, three in the US, three in
the more rigorous meta-analysis by Turati et al.
Europe, and one in Singapore. Among these 14
(2012).
cohort studies, 11 focused on incidence (Inoue
Turati et al. (2012) conducted a meta-analysis
et al., 2005, 2009; Shimazu et al., 2005; Hu et al.,
on the association between coffee consumption
2008; Ohishi et al., 2008; Johnson et al., 2011;
and pancreatic cancer risk, using data from case–
Lai et al., 2013; Aleksandrova et al., 2015; Bamia
control and cohort studies that were published
et al., 2015; Petrick et al., 2015; Setiawan et al.,
until March 2011. They identified 37 case–control
2015) and 3 focused on mortality (Kurozawa
and 17 cohort studies (10  594 cases) as eligible
et al., 2004, 2005; Wakai et al., 2007). Inoue
for meta-analysis. Random-effects models were
et al. (2005, 2009) reported findings from the
used. When only smoking-adjusted studies were
same prospective cohort study, but the latter
considered, 22 case–control studies and 15 cohort
study (Inoue et al., 2009) reported the results
studies were suitable for meta-analysis. Among
from a subcohort with information on hepa-
the smoking-adjusted studies, Turati et al. estim-
titis C virus (HCV) and hepatitis B virus (HBV)
ated pooled relative risks of pancreatic cancer for
status. Kurozawa et al. (2004, 2005) and Wakai
high versus low coffee consumption of 1.10 (95%
et al. (2007) also reported results derived from
CI, 0.92–1.31) for case–control studies, 1.04 (95%
the same study population; the latter (Wakai
CI, 0.80–1.36) for cohort studies, and 1.08 (95%
et al., 2007) used a nested case–control analysis.
CI, 0.94–1.25) for all studies, with significant
Likewise, Bamia et al. (2015) and Aleksandrova

176
Drinking coffee

et al. (2015) reported results derived from the questionnaire at baseline. After adjusting for
same population; the latter used a nested case– potential confounders, those who consumed
control study analysis. Johnson et al. (2011) and coffee on a daily basis had a lower risk of HCC
Lai et al. (2013) reported results for both cohort than non-drinkers (HR, 0.49; 95% CI, 0.36–0.66).
and nested case–control analysis. Petrick et al. The risk decreased with the amount of coffee
(2015) reported results from a pooled analysis consumed; compared with non-drinkers, the
of the cohort studies. One pooled analysis of hazard ratio for drinking 1–2 cups/day was
US cohorts analysed the risk by histological 0.52 (95% CI, 0.38–0.73), for 3–4 cups/day 0.48
subtypes, hepatocellular carcinoma, and intra- (95% CI, 0.28–0.83), and for ≥  5 cups/day 0.24
hepatic cholangiocarcinoma (Petrick et al., 2015). (95% CI, 0.08–0.77). The P value for trend was
Case–control studies were conducted in < 0.001. The inverse association persisted when
various countries: three studies in Italy, one in the participants were stratified by age, smoking,
Greece, one in Italy and Greece, two in Japan, alcohol intake, green vegetable intake, green
and one each in Serbia, the Republic of Korea, tea intake, and history of chronic liver disease.
Hong Kong Special Administrative Region, Similar associations were observed when the
and India. All studies except one (Tanaka et al., analysis was restricted to HCV+ or HBV+ cases.
2007) were hospital-based. Tanaka et al. (2007) [The strengths of this study were its prospective
included both population-based and hospi- design and large scale. Limitations included
tal-based control groups. the facts that consumption was self-reported,
The Working Group also reviewed seven changes in coffee consumption were not consid-
meta-analyses of coffee drinking and cancer of ered, and the HCV/HBV status of controls was
the liver. not available.]
A cohort study (Kurozawa et al., 2004) Kurozawa et al. (2005) examined the associa-
reporting coffee consumption and risk of hepato- tion between coffee drinking and HCC mortality
cellular carcinoma (HCC) mortality by sex and in the JACC Study. In total, 110  688 men and
age group has been excluded from this review; women aged 40–79 years were grouped by coffee
the results were derived from univariate analysis intake categories. Information on habitual coffee
with no adjustment for other risk factors, and the consumption was obtained by self-reported
results controlling for confounding factors were questionnaire at baseline. On adjusting for
reported in another paper by Kurozawa et al. potential confounders, including history of
(2005). One case–control study (Kanazir et al., diabetes, liver diseases, and alcohol consump-
2010) was also excluded from this review because tion, the hazard ratio of HCC mortality for
it did not adjust for any covariates. drinkers of ≥  1 cups/day of coffee compared
with non-coffee drinkers was 0.50 (95% CI,
2.3.1 Cohort studies 0.31–0.79); the hazard ratio for drinkers of
< 1 cup/day was 0.83 (95% CI, 0.54–1.25). [The
See Table 2.5 strengths of this study were its large scale and
Inoue et al. (2005) investigated the associ- prospective design. Limitations included the
ation between coffee consumption and inci- absence of HCV and HBV markers.]
dence of HCC among 90  452 Japanese (43  109 Shimazu et al. (2005) examined the associ-
men and 47  343 women) aged 40–69  years at ation between coffee consumption and the risk
baseline in the JPHC-based prospective study, of cancer of the liver in a pooled analysis of
which began during 1990–1994. Information on data available from two cohort studies based in
coffee drinking was obtained by self-reported Miyagi, Japan. A self-administered questionnaire

177
178

IARC MONOGRAPHS – 116
Table 2.5 Cohort studies on cancer of the liver and drinking coffee

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Inoue et al. 90 452 (43 109 men and Liver/HCC Coffee consumption: men and women Sex, age, study area, Strengths: prospective,
(2005) 47 343 women), JPHC Almost never 161 1.00 smoking, alcohol large scale
Japan, Study subjects aged 1–2 days/wk 65 0.75 (0.56–1.01) drinking, green Limitations: self-
1990–1994 40–69 yr, 11 public vegetable intake, report, change not
3–4 days/wk 36 0.79 (0.55–1.14)
to 2001 health centre-based green tea drinking considered,
areas, residential register Almost 72 0.49 (0.36–0.66) HCV, HBV status of
Exposure assessment everyday controls unknown
method: questionnaire 1–2 cups/day 54 0.52 (0.38–0.73)
3–4 cups/day 15 0.48 (0.28–0.83)
≥ 5 cups/day 3 0.24 (0.08–0.77)
Trend test P value, < 0.001
Coffee consumption: men
Almost never 116 1.00
1–2 days/wk 43 0.74 (0.52–1.05)
3–4 days/wk 27 0.76 (0.50–1.16)
Almost 59 0.49 (0.35–0.69)
everyday
1–2 cups/day 45 0.55 (0.38–0.80)
3–4 cups/day 11 0.41 (0.21–0.77)
≥ 5 cups/day 3 0.27 (0.09–0.87)
Trend test P value, < 0.001
Coffee consumption: women
Almost never 45 1.00
1–2 days/wk 17 0.77 (0.43–1.37)
3–4 days/wk 9 0.89 (0.43–1.84)
Almost 13 0.48 (0.25–0.92)
everyday
1–2 cups/day 9 0.43 (0.20–0.90)
3–4 cups/day 4 0.89 (0.31–2.59)
≥ 5 cups/day 0 –
Trend test P value, 0.042
Table 2.5 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Inoue et al. Coffee consumption: HCC with HCV+, men and
(2005) women combined
(cont.) Almost never 86 1.00
1–2 days/wk 26 0.59 (0.38–0.91)
3–4 days/wk 15 0.66 (0.38–1.16)
Almost 37 0.57 (0.37–0.86)
everyday
1–2 cups/day 29 0.64 (0.41–0.99)
3–4 cups/day 6 0.42 (0.18–0.99)
≥ 5 cups/day 2 0.34 (0.08–1.41)
Trend test P value, 0.005
Coffee consumption: HCC with HBV+, men and
women combined (60 cases)
Almost never 24 1.00
1–2 days/wk 9 0.66 (0.31–1.43)
3–4 days/wk 9 1.14 (0.52–2.47)
Almost 18 0.60 (0.31–1.18)
everyday
1–2 cups/day 12 0.56 (0.26–1.21)
3–4 cups/day 5 0.81 (0.30–2.22)
≥ 5 cups/day 1 0.39 (0.05–2.98)
Trend test P value, 0.231
Coffee consumption: no history of CLD
Almost never NR 1.00
1–2 days/wk NR 0.85 (0.59–1.24)
3–4 days/wk NR 1.15 (0.76–1.74)
Almost NR 0.45 (0.3–0.67)

Drinking coffee
everyday
1–2 cups/day NR 0.46 (0.29–0.72)
3–4 cups/day NR 0.52 (0.26–1.05)
≥ 5 cups/day NR 0.15 (0.02–1.05)
Trend test P value, < 0.001
179
180

IARC MONOGRAPHS – 116
Table 2.5 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Inoue et al. Coffee consumption: history of CLD
(2005) Almost never NR 1.00
(cont.) 1–2 days/wk NR 0.79 (0.48–1.30)
3–4 days/wk NR 0.44 (0.18–1.11)
Almost NR 0.91 (0.58–1.41)
everyday
1–2 cups/day NR 0.99 (0.61–1.61)
3–4 cups/day NR 0.71 (0.31–1.67)
≥ 5 cups/day NR 0.76 (0.18–3.16)
Trend test P value, 0.432
Kurozawa 110 688 (46 399 men, Liver/HCC Coffee consumption (cups/day) Age, sex, education, Strengths: large-scale,
et al. (2005) 64 289 women), JACC All subjects     history of diabetes prospective design
Japan, Study, subjects aged Non-drinkers 103 1.00 and liver disease, Limitations: absence
1988–1990, 40–79 yr smoking and alcohol of HCV and HBV
< 1 57 0.83 (0.54–1.25)
follow-up Exposure assessment habits markers
until 1999 method: questionnaire ≥ 1 98 0.50 (0.31–0.79)
Trend test P value, 0.007
Coffee consumption (cups/day): men Age, education,
Men     history of diabetes
Non-drinkers 66 1.00 and liver disease,
smoking and alcohol
< 1 41 0.91 (0.57–1.45)
habits
≥ 1 71 0.49 (0.28–0.85)
Trend test P value, 0.007
Coffee consumption (cups/day): women
Non-drinkers 37 1.00
< 1 16 0.64 (0.27–1.51)
≥ 1 27 0.51 (0.20–1.31)
Trend test P value, 0.141
Table 2.5 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Kurozawa et Coffee consumption (cups/day): with history of liver Age, sex, education,
al. (2005) diseases history of diabetes,
(cont.) Non-drinkers 62 1.00 smoking and alcohol
< 1 35 0.94 (0.53–1.66) habits
≥ 1 54 0.44 (0.22–0.88)
Trend test P value, 0.028
Coffee consumption (cups/day): without history of
liver diseases
Non-drinkers 41 1.00
< 1 22 0.79 (0.44–1.41)
≥ 1 44 0.61 (0.32–1.16)
Trend test P value, 0.113
Shimazu Cohort 1: 22 404 (10 588 Liver/HCC Coffee consumption (cups/day): cohort 1 Age, sex, history Strengths: prospective,
et al. (2005) men and 11 816 women), Never 29 1.00 of liver disease, large scale
Japan aged ≥ 40 yr Occasionally 25 0.56 (0.33–0.97) alcohol consumption, Limitations: no
(Miyagi): (1) Cohort 2: 38 703 (18 869 smoking status information on
≥ 1 16 0.53 (0.28–1.00)
1984–1992 men, 19 834 women), HBV and HCV
and (2) aged 40–64 yr Trend test P value, 0.038 infection status,
1990–1997 Exposure assessment Coffee consumption (cups/day): cohort 2 DCO cases possibility
method: Never 12 1.00 of misclassifying
questionnaire Occasionally 21 1.05 (0.52–2.16) secondary metastasis
≥ 1 14 0.68 (0.31–1.51) to liver, former
drinkers not
Trend test P value, 0.3
distinguishable from
Coffee consumption (cups/day): pooled non-drinkers
Never 41 1.00
Occasionally 46 0.71 (0.46–1.09)
≥ 1 30 0.58 (0.36–0.96)

Drinking coffee
Trend test P value, 0.024
181
182

IARC MONOGRAPHS – 116
Table 2.5 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Wakai et al. Cases: 96 of HCC Liver/HCC Coffee consumption (cups/day): total Area, smoking and Strengths: nested case–
(2007) mortality, identified Total     drinking habits, control design (as part
Japan, from death certificates Non-drinkers 44 1.00 history of diabetes of JACC)
1988–1990 Controls: 420 HCV+ and mellitus and liver Limitations: mortality
< 1 34 0.77 (0.45–1.32)
3024 HCV– controls, diseases not incidence, coffee
matched for age, ≥ 1 18 0.49 (0.25–0.96) intake at baseline only
sex, HCV-antibody Trend test P value, 0.038
seropositivity Coffee consumption (cups/day): HCV-Ab-positive
Exposure assessment Non-drinkers 28 1.00
method: questionnaire; < 1 23 0.91 (0.41–2.04)
rank correlation r2 = 0.79
≥ 1 9 0.31 (0.11–0.85)
Trend test P value, 0.031
Coffee consumption (cups/day): HCV-Ab-negative
Non-drinkers 16 1.00
< 1 11 0.65 (0.29–1.46)
≥ 1 9 0.75 (0.29–1.92)
Trend test P value, 0.45
Hu et al. 60 323; seven Liver/HCC Daily coffee consumption (cups/day) Adjusted for age, Strengths: large-scale
(2008) independent cross- Total 128 – sex, study year, population-based,
Finland, sectional surveys in six 0–1 20 1.00 alcohol consumption, prospective,
1972–2006 geographic areas education, smoking, long follow-up (19.3 yr)
2–3 30 0.66 (0.37–1.16)
Exposure assessment diabetes, and CLD Limitations: self-
method: questionnaire 4–5 33 0.44 (0.25–0.77) report only at baseline,
6–7 28 0.38 (0.21–0.69) impossible to assess
≥ 8 17 0.32 (0.16–0.62) caffeine intake, no
Trend test P value, 0.003 data on HBV or HCV,
residual confounding
Table 2.5 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Hu et al. Daily coffee consumption (cups/day): men (82 cases)
(2008) 0–1 16 1.00
(cont.) 2–3 21 0.68 (0.35–1.31)
4–5 17 0.35 (0.18–0.71)
6–7 15 0.31 (0.15–0.63)
≥ 8 13 0.28 (0.13–0.61)
Trend test P value, 0.001
Daily coffee consumption (cups/day): women (46
cases)
0–1 4 1.00
2–3 9 0.62 (0.19–2.04)
4–5 16 0.60 (0.20–1.82)
6–7 13 0.58 (0.19–1.82)
≥ 8 4 0.41 (0.10–1.70)
Trend test P value, 0.82
Ohishi et al. Cases: 224 HCC Liver/HCC Coffee intake frequency Hepatitis virus Strengths: prospective,
(2008) identified from Never 187 1.00 infection, alcohol nested case–control,
Japan, Hiroshima and Tissue Daily 37 0.40 (0.16–1.02) consumption, HCV and HBV
1969–2002 Registry and Nagasaki smoking, BMI, infection considered
Trend test P value, 0.055
Cancer Registry diabetes mellitus, Limitations: severity of
Controls: 644 matched radiation dose of the liver fibrosis could not
from the cohort by liver be considered
sex, age, city, time of
serum storage, method
for serum storage and
radiation exposure
Exposure assessment
method: questionnaire

Drinking coffee
183
184

IARC MONOGRAPHS – 116
Table 2.5 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Inoue et al. 18 815; JPHC Cohort II Liver/HCC Coffee consumption (cups/day): total (110 cases) Sex, age, area, Strengths: prospective
(2009) Exposure assessment Almost never 67 1.00 smoking, alcohol analysis with blood
Japan, method: questionnaire < 1 35 0.67 (0.42–1.07) drinking, green tea samples
1993–2006 intake, BMI, history Limitations: relatively
1–2 18 0.49 (0.27–0.91)
of diabetes, serum small number of cases
≥ 3 6 0.54 (0.21–1.39) ALT, HCV and HBV
Trend test P value, 0.025 infection status
Coffee consumption (cups/day): HCV+ and/or HBV+
(92 cases)
Almost never 43 1.00
< 1 28 0.55 (0.33–0.93)
1–2 15 0.47 (0.24–0.93)
≥ 3 6 0.61 (0.23–1.62)
Trend test P value, 0.036
Coffee intake (cups/day): HCV+ (80 cases)
Almost never 38 1.00
< 1 cup/day 24 0.56 (0.32–0.99)
1–2 cups/day 12 0.40 (0.18–0.88)
≥ 3 cups/day 6 0.78 (0.28–2.15)
Trend test P value, 0.065
Johnson 63 257 Chinese aged Liver/HCC Coffee consumption (cups/day) Age, sex, dialect Strengths: prospective
et al. (2011) 45–74 yr Non-drinkers 69 1.00 group, years of with blood samples (in
Singapore, Exposure assessment 0 to < 1 38 0.94 (0.63–1.40) recruitment, part)
1993–1998 method: 165-item FFQ BMI, education, Limitations: lack of
1 to < 2 149 1.17 (0.87–1.56)
to 2006 consumption of HBV and HCV status
2 to < 3 92 0.78 (0.56–1.07) alcohol beverages, for all participants,
≥ 3 14 0.56 (0.31–1.00) cigarette smoking, participants not
Trend test P value, 0.05 black tea and green examined for liver
tea intake, and damage at baseline;
history of diabetes relatively small
number of cases
Table 2.5 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Johnson Cases: 92 HCC by Liver/HCC Coffee consumption (cups/day) Age, sex, dialect Case–control analysis
et al. (2011) national cancer registry Non-drinkers 17 1.00 group, years of of a subset of the
Singapore, Controls: 276 0 to < 1 11 0.77 (0.26–2.29) recruitment, cohort
1993–1998 individually matched BMI, education, Strengths: nested case–
1 to < 2 34 0.84 (0.38–1.85)
to 2006 by sex, dialect group, consumption of control, prospective
age at enrolment, date 2 to < 3 28 1.32 (0.56–3.14) alcohol beverages, HBV and HCV
of baseline interview ≥ 3 2 0.23 (0.05–1.21) cigarette smoking, information available
and date of biospecimen Trend test P value, 0.71 black tea and green Limitations:
collection (± 6 mo) tea intake, history of participants were not
Exposure assessment diabetes, and HBV/ examined for liver
method: 165-item FFQ HCV infection status damage at baseline,
relatively small
number of cases
Lai et al. 27 037; ATBC study Liver/HCC Coffee consumption (cups/day) Intervention arm, Strengths: prospective
(2013) male smokers aged Never drinker 9 1.00 age, BMI, education, study, long follow-up
Finland, 50–69 yr > 0 to < 1 36 1.35 (0.65–2.82) marital status, history Limitations: HCV/
1985–1988, Exposure assessment of diabetes, smoking, HBV status available
1 to < 2 60 0.73 (0.48–1.12)
follow-up to method: questionnaire alcohol consumption, for subset only
December 2 to < 3 47 0.52 (0.33–0.82) serum cholesterol
2009 3 to < 4 22 0.45 (0.26–0.78)
≥ 4 20 0.53 (0.30–0.95)
Unit change NR 0.82 (0.73–0.93)
(per cups/day)
Trend test P value, 0.0007

Drinking coffee
185
186

IARC MONOGRAPHS – 116
Table 2.5 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Lai et al. Coffee consumption, filtered method (cups/day) Type of coffee, ATBC
(2013) > 0 to < 1 16 1.00 intervention arm,
(cont.) 1 to < 2 34 0.80 (0.44–1.47) age, BMI, education,
marital status, history
2 to < 3 26 0.54 (0.29–1.03)
of diabetes, smoking,
3 to < 4 9 0.34 (0.15–0.78) alcohol consumption,
≥ 4 12 0.61 (0.28–1.34) serum cholesterol
Unit increase NR 0.82 (0.69–0.98)
(per cups/day)
Trend test P value, 0.03
Coffee consumption, boiled method (cups/day)
> 0 to < 1 7 1.00
1 to < 2 10 0.60 (0.23–1.57)
2 to < 3 5 0.25 (0.08–0.80)
3 to < 4 7 0.60 (0.21–1.75)
≥ 4 4 0.40 (0.12–1.40)
Unit increase NR 0.85 (0.65–1.11)
(per cups/day)
Trend test P value, 0.19
Table 2.5 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Bamia et al. 486 799; EPIC Study Liver/HCC Coffee intake, quintiles (mL/day) Sex, diabetes, Stratified for age
(2015) Exposure assessment Q1 (M: 0–83.3; 47 1.00 education, BMI, at recruitment and
Europe method: F: 0–60) smoking, physical centre.
(Denmark, questionnaire Q2 (M: 49 0.85 (0.56–1.29) activity, alcohol Strengths: cohort
France, 83.3–200.4; F: intake, energy intake, design,
Germany, 60–191.9) tea intake multicentre coverage
Greece, to examine variable
Q3 (M: 38 0.63 (0.39–1.02)
Italy, the range of intake
200.5–476.9; F:
Netherlands, across European
191.9–375)
Norway, countries, validated
Spain, Q4 (M: 36 0.49 (0.29–0.82) questionnaire,
Sweden, UK) 477.2–830.4; F: relatively long follow-
1992–2000 375–580.2) up
to 2004– Q5 (M 31 0.28 (0.16–0.50) Limitations: modest
2008 831.3–4500; F: number of HCC cases,
580.3–6250) lack of data on brewing
Trend test P value, < 0.001 methods
Petrick et al. 1 212 893; Liver Cancer Liver/HCC Coffee consumption (cups/day) Sex, age, race, cohort, Strengths: large sample
(2015) Pooling Project (LCPP), Non-drinker 85 1.00 BMI, smoking status, size allowed stratifying
USA, USA-based NCI cohort Ever 650 1.00 (0.79–1.27) cigarette smoking by caffeine content
1992–1995, consortium comprising intensity, alcohol, of coffee and sex,
> 0 to < 1 138 1.24 (0.94–1.64)
2007–2010 NIH-AARP, AHS, P-value for trend of histological subtype of
or variable USRTS, PLCO, 1 to < 2 149 1.16 (0.88–1.52) continuous variables liver cancer (HCC and
WHS, CPS-II, IWHS, 2–3 255 0.89 (0.68–1.15) ICC)
BWHS, WHI > 3 97 0.73 (0.53–0.99) Limitations: number of
Exposure assessment Continuous NR 0.90 (0.85–0.94) ICC limited
method: questionnaire Trend test P value, < 0.0001

Drinking coffee
187
188

IARC MONOGRAPHS – 116
Table 2.5 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Petrick et al. Coffee consumption (cups/day): men (530) Age, race, cohort,
(2015) Non-drinker 40 1.00 BMI, smoking status,
(cont.) Ever 490 1.21 (0.87–1.69) cigarette smoking
intensity, alcohol,
> 0 to < 1 113 1.57 (1.09–2.25)
P-value for trend of
1 to < 2 103 1.35 (0.93–1.95) continuous variable
2–3 195 1.06 (0.75–1.51)
> 3 79 0.93 (0.63–1.37)
Continuous NR 0.90 (0.86–0.96)
(cups/day)
Trend test P value, 0.0004
Coffee consumption (cups/day): women (205)
Non-drinker 45 1.00
Ever 160 0.78 (0.56–1.10)
> 0 to < 1 25 0.79 (0.47–1.33)
1 to < 2 46 1.01 (0.66–1.53)
2–3 60 0.71 (0.48–1.06)
> 3 18 0.46 (0.26–0.81)
Continuous NR 0.87 (0.79–0.96)
(cups/day)
Trend test P value, 0.004
Caffeinated coffee (cups/day) Sex, age, race, cohort,
Non-drinker 85 1.00 BMI, smoking status,
Ever 379 1.00 (0.77–1.28) cigarette smoking
intensity, alcohol,
> 0 to < 1 58 1.22 (0.87–1.73)
P value for trend of
1 to < 2 85 1.19 (0.87–1.62) continuous variables
2–3 174 0.95 (0.72–1.26)
> 3 62 0.71 (0.50–1.01)
Trend test P value, 0.002
Table 2.5 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Petrick et al. Decaffeinated coffee (cups/day)
(2015) Non-drinker 85 1.00
(cont.) Ever 204 1.16 (0.88–1.53)
0 63 1.00
> 0 to < 1 58 1.33 (0.92–1.91)
1 to < 2 51 1.38 (0.95–2.02)
2–3 64 0.97 (0.67–1.40)
> 3 21 0.92 (0.55–1.54)
Trend test P value, 0.1
Liver and bile Coffee consumption (cups/day)
ducts: ICC Non-drinker 33 1.00
Ever 199 0.93 (0.63–1.37)
> 0 to < 1 36 1.15 (0.70–1.89)
1 to < 2 33 0.79 (0.48–1.30)
2–3 85 0.93 (0.61–1.42)
> 3 40 1.00 (0.61–1.63)
Continuous, NR 1.00 (0.92–1.08)
cups/day
Trend test P value, 0.9
Caffeinated coffee (cups/day)
Non-drinker 33 1.00
Ever 119 0.91 (0.60–1.37)
0 33 1.00
> 0 to < 1 17 1.32 (0.71–2.43)
1 to < 2 15 0.59 (0.32–1.10)
2–3 57 0.91 (0.58–1.43)

Drinking coffee
> 3 30 1.08 (0.63–1.83)
Trend test P value, > 0.99
189
190

IARC MONOGRAPHS – 116
Table 2.5 (continued)

Reference, Population size, Organ site Exposure Exposed Risk estimate Covariates Comments
location, description, exposure category or cases/ (95% CI) controlled
enrolment/ assessment method level deaths
follow-up
period
Petrick et al. Decaffeinated coffee (cups/day)
(2015) Non-drinker 33 1.00
(cont.) Ever 56 0.95 (0.59–1.53)
0 18 1.00
> 0 to < 1 15 1.17 (0.58–2.35)
1 to < 2 10 0.94 (0.43–2.07)
2–3 20 1.11 (0.56–2.17)
> 3 6 1.03 (0.39–2.70)
Trend test P value, 0.6
Setiawan 162 022; multiethnic Liver/HCC Regular coffee (cups/day) Age, sex, ethnicity, Strengths: prospective,
et al. (2015) cohort (MEC) study, Never 119 1.00 education, BMI, long follow-up time,
US, 1993– Hawaii and California < 1 111 1.14 (0.88–1.48) alcohol intake, multiethnic and
1996, 18 yr Exposure assessment smoking status, large sample size,
1 137 0.87 (0.67–1.11)
follow-up method: questionnaire diabetes confounder adjustment
2–3 67 0.62 (0.46–0.84) Limitations: coffee
≥ 4 17 0.59 (0.35–0.99) assessment by single
Trend test P value, 0.002 self-report, lack of
Decaffeinated coffee (cups/day) information on liver
Never 287 1.00 disease other than
HCC, no information
< 1 128 0.87 (0.70–1.08)
on HBV/HCV status
≥ 2 21 0.86 (0.55–1.34)
Trend test P value, 0.2
Ab, antibody; AHS, Agricultural Health Study; ALT, alanine transaminase; ATBC, Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study; BMI, body mass index; BWHS, Black
Women’s Health Study; CI, confidence interval; CLD, chronic liver disease; CPS-II, Cancer Prevention Study-II; DCO, death certificate only; EPIC, European Prospective Investigation
into Cancer and Nutrition; F, female; FFQ, food frequency questionnaire; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; ICC, intrahepatic
cholangiocarcinoma; IWHS, Iowa Women’s Health Study; JACC Japan Collaborative Cohort Study; JPHC, Japan Public Health Center-based Prospective; LCPP, Liver Cancer Pooling
Project; M, male; MEC, multiethnic cohort; mo, month(s); NIH-AARP, National Institutes of Health–American Association of Retired Persons; NR, not reported; PLCO, Prostate, Lung,
Colorectal and Ovarian Cancer Screening Trial; USRTS, United States Radiologic Technologists Study; WHI, Women’s Health Initiative; WHS, Women’s Health Study; wk, week(s); yr,
year(s)
Drinking coffee

regarding the frequency of coffee consump- Hu et al. (2008) examined the single and
tion and other health habits was distributed to joint associations of coffee consumption and
22 404 women and men in Cohort 1 and 38 703 serum gamma-glutamyltransferase (GGT) with
subjects in Cohort 2. After adjustment for age, the risk of primary cancer of the liver. The study
sex, history of liver disease and diabetes, alcohol cohort included 60  323 Finnish subjects who
consumption, and smoking status, the pooled were aged 25–74 years and free from any cancer
hazard ratios (95% CI) of drinking coffee occa- at baseline. Information on coffee consumption
sionally and ≥ 1 cups/day compared with never was collected using mailed self-administered
were 0.71 (0.46–1.09) and 0.58 (0.36–0.96) (P for questionnaires. After adjustment for risk factors
trend, 0.024). [The strengths of this study were its including alcohol consumption, diabetes, and
prospective design and large scale. Limitations chronic liver disease at baseline and during
included: the lack of information regarding HBV follow-up, and BMI, hazard ratios (95% CI) of
and HCV infection status; death certificate only liver cancer in participants who drank 2–3, 4–5,
(DCO) cases meant it was possible to misclassify 6–7, and ≥  8 cups/day of coffee compared with
secondary metastasis as cancer of the liver; and none were 0.66 (0.37–1.16), 0.44 (0.25–0.77),
former drinkers were not distinguished from 0.38 (0.21–0.69), and 0.32 (0.16–0.62) (P for
non-drinkers.] trend, 0.003). Further adjustment for serum GGT
Wakai et al. (2007) examined HCC mortality in subgroup analysis did not substantially affect
in relation to coffee consumption and anti-HCV the results. This inverse association between
antibody (Ab) seropositivity. This study was coffee consumption and liver cancer risk persisted
carried out in Japan as a nested case–control in analyses stratified by several risk factors. [The
study as part of the JACC Study previously main strengths of this study were its large-scale,
reported by Kurozawa et al. (2005). The analyses population-based, prospective design and long
involved 96 HCC mortality cases with serum follow-up (19.3  years). Limitations included
samples. Among 39 242 subjects donating blood consideration of coffee consumption at baseline
samples at baseline, controls were matched for only, a lack of data on HBV or HCV, and residual
age, sex, and HCV-Ab seropositivity. Habitual confounding.]
coffee consumption was assessed by self-reported Ohishi et al. (2008) conducted a nested
questionnaire at baseline. Coffee drinking was case–control study using sera stored before HCC
significantly associated with a decreased risk of diagnosis in the longitudinal cohort of Japanese
death from HCC. After adjustment, including atomic bomb survivors, considering the joint
for history of diabetes and liver disease, odds effect (synergism) of HBV and HCV infections.
ratios (95% CI) for daily coffee drinkers versus The study included 224 incident HCC cases
non-drinkers were 0.49 (0.25–0.96), 0.31 and 644 controls who were matched to cases
(0.11–0.85), and 0.75 (0.29–1.92) for total subjects, on sex, age (± 2 years), city, and time (± 2 years)
HCV-Ab-positive subjects and HCV-Ab-negative and method of serum storage, and were coun-
subjects, respectively. The increased risk observed ter-matched on radiation dose. Information
among HCV-Ab-positive individuals with signif- on daily coffee drinking was obtained from a
icant trend (P for trend, 0.031) was not observed survey in 1978. After adjustment for HBV and
among HCV-Ab-negative individuals. [The main HCV infections, alcohol consumption, smoking
strength of this study was its nested case–control habits, BMI, and diabetes mellitus, the odds ratio
design. Limitations included the consideration of HCC for daily coffee drinking compared with
of mortality and not incidence, and coffee intake never drinking coffee was 0.4 (95% CI, 0.16–1.02;
was only recorded at baseline.] P for trend, 0.055). [The strengths of this study

191
IARC MONOGRAPHS – 116

were its prospective, nested case–control design 1.17 (0.87–1.56), 0.78 (0.56–1.07), and 0.56
and the fact that HCV and HBV infection status (0.31–1.00), respectively (P for trend,  0.05). All
was considered. The main limitation was that results were adjusted for age at recruitment, sex,
severity of liver fibrosis could not be considered.] dialect group, year of recruitment, BMI, level of
Inoue et al. (2009) examined whether coffee education, consumption of alcoholic beverages,
consumption was associated with a reduced risk cigarette smoking, frequency of black and green
of liver cancer by hepatitis virus infection status tea intake, and history of diabetes.
in the JPHC Study Cohort II. This study was a This study also provided results from the
subcohort analysis of Inoue et al. (2005), with subset of the cohort who provided blood samples
HCV and HBV infections determined by analyses at baseline. A total of 92 cases of HCC of the liver
of blood samples. Hazard ratios of liver cancer for and their controls matched for age, date of inter-
different levels of coffee consumption compared view, and date of blood sample collection were
with almost-never drinkers were estimated after analysed. On adjustment for HBV and HCV
adjusting for risk factors including smoking infection status, in addition to the factors previ-
status, ethanol intake, BMI, history of diabetes, ously indicated, the odds ratios of HCC and high
and HCV and HBV infection status. Increased consumption of coffee in the subset were similar
coffee consumption was associated with a to those based on the entire cohort, although not
reduced risk of liver cancer in all subjects; multi- all odds ratios were statistically significant. Odds
variate-adjusted hazard ratios (95% CI) for < 1, ratios (95% CI) of the risk of HCC for individuals
1–2, and ≥ 3 cups/day were 0.67 (0.42–1.07), 0.49 who consumed coffee at a frequency of 0 to < 1,
(0.27–0.91), and 0.54 (0.21–1.39), respectively. A 1 to <  2, 2 to <  3, and ≥  3 cups/day compared
similar trend in the hazard ratios was observed with non-drinkers were 0.77 (0.26–2.29), 0.84
in those with HCV and/or HBV infection. [The (0.38–1.85), 1.32 (0.56–3.14), and 0.23 (0.05–1.21),
Working Group considered the strengths of this respectively (P for trend, 0.71). [The strength of
study to be its prospective analysis with blood this study was its prospective nature and use of
samples as well as consideration of HCV and blood samples for part of the cohort. Its limita-
HBV infection status. Its main limitation was the tions included a lack of HBV and HCV status
relatively small number of cases.] for all cohort participants, participants in the
Johnson et al. (2011) examined the asso- cohort were not measured for the amount of liver
ciation between coffee consumption and the damage present at baseline, and the relatively
risk of developing HCC of the liver within the small number of cases.]
Singapore Chinese Health Study, a prospective Lai et al. (2013) evaluated the association
cohort of 63 257 Chinese men and women aged between coffee intake and incident cancer of
45–74 years (a relatively high-risk population the liver and chronic liver disease mortality in
for developing HCC). Data on coffee consump- 27 037 Finnish male smokers, aged 50–69 years,
tion were collected through in-person inter- in the ATBC Study. Coffee consumption was
views at baseline during 1993–1998. A total of recorded at baseline by FFQ and subjects were
362 cohort participants had developed HCC by followed up for 24 years for incident liver cancer.
2006. High levels of coffee consumption were Adjusted hazard ratios (95% CI) for the associ-
associated with reduced risk of HCC. Compared ation between coffee intake and incident liver
with non-drinkers, individuals who consumed cancer, compared with never drinkers, were 1.35
coffee at a frequency of 0 to <  1, 1 to <  2, 2 to (0.65–2.82), 0.73 (0.48–1.12), 0.52 (0.33–0.82),
< 3, and ≥ 3 cups/day had a reduced risk of HCC 0.45 (0.26–0.78), and 0.53 (0.30–0.95) for
with hazard ratios (95% CI) of 0.94 (0.63–1.40), drinking coffee at a frequency of 0 to < 1, 1 to

192
Drinking coffee

< 2, 2 to < 3, 3 to < 4, and ≥ 4 cups/day, respect- Its limitations were the modest number of HCC
ively (P for trend,  0.0007). Inverse associations cases and a lack of data on brewing methods.]
persisted in those without diabetes, among HBV- Aleksandrova et al. (2015) also used the EPIC
and HCV-negative subjects, and in analyses population to evaluate the potential mediating
stratified by age, BMI, alcohol consumption, and roles of inflammatory, metabolic, liver injury,
smoking dose. The study observed similar associ- and iron metabolism biomarkers on the asso-
ations for those drinking boiled or filtered coffee. ciation between coffee intake and risk of HCC
This study also provided results among those using a nested case–control study design. The
with information on HBV and HCV using 155 association between cancer of the liver and coffee
cases of cancer of the liver and 770 controls. The consumption was similar to that reported by
association was not appreciably different when Bamia et al. (2015), who also provided evidence
adjusted for HBV and HCV infection status. that this association was mediated by biomarkers
[The strengths of this study were its prospec- of inflammation and hepatocellular injury.
tive nature and long follow-up. However, it was Petrick et al. (2015) investigated whether
not reported whether the coffee consumed was caffeine is responsible for the inverse association
caffeinated or decaffeinated.] between coffee and cancer of the liver. Through
Bamia et al. (2015) investigated the associ- the Liver Cancer Pooling Project, a consortium of
ation between coffee consumption and risk of US-based cohort studies, data from 1 212 893 indi-
HCC in the EPIC study. Information on coffee viduals (860 cases of HCC and 260 cases of intra-
intake was obtained through centre-specific hepatic cholangiocarcinoma (ICC)) in 9 cohorts
questionnaires on cups per day, week, or month. were pooled. Hazard ratios and confidence inter-
Hazard ratios for HCC incidence in relation to vals were estimated adjusting for sex, age, race,
categories of coffee intake in mL/day were estim- cohort, BMI, smoking status, cigarette smoking
ated, adjusting for risk factors including self- intensity, and alcohol intake. Higher coffee
reported diabetes, ethanol intake, BMI, energy consumption was associated with a lower risk of
intake, and tea intake. Compared with the lowest HCC; the hazard ratio for consumption of >  3
quintile (Q1), coffee consumers in the higher cups/day of coffee compared with a non-drinker
quintiles had lower hazard ratios (95% CI) of 0.85 was 0.73 (95% CI, 0.53–0.99; P for trend, < 0.0001).
(0.56–1.29), 0.63 (0.39–1.02), 0.49 (0.29–0.82), When considering men and women separately, a
and 0.28 (0.16–0.50) for quintiles Q2, Q3, Q4, and reduced risk for consumption of > 3 cups/day of
Q5, respectively (P for trend, < 0.001). There was coffee compared with a non-drinker was notable
no compelling evidence of heterogeneity of these among women (HR, 0.46; 95% CI, 0.26–0.81;
associations across strata of important HCC risk P for trend,  0.004) compared with men (HR,
factors, including HBV or HCV infection status, 0.93; 95% CI, 0.63–1.37; P for trend,  0.0004).
in a nested case–control analysis. The inverse, The associations were stronger for caffeinated
monotonic associations of coffee intake with coffee; the hazard ratio for consumption of > 3
risk of HCC were apparent for caffeinated (P for cups/day of coffee compared with a non-drinker
trend, 0.009) but not decaffeinated coffee (P for was 0.71 (95% CI, 0.50–1.01; P for trend, 0.002)
trend, 0.45), but this information was only avail- for caffeinated coffee compared with 0.92 (95%
able for about one third of the study subjects. [The CI, 0.55–1.54; P for trend, 0.1) for decaffeinated
strengths of this study included its cohort design, coffee. There was no association between coffee
multicentre coverage to examine a variable range consumption and ICC. [The Working Group
of intake across European countries, a validated noted that the large sample size allowed stratifi-
questionnaire, and a relatively long follow-up. cation by caffeine content of coffee and sex. An

193
IARC MONOGRAPHS – 116

additional strength of the study was considera- controls. After adjustment for sex, age, heavy
tion of the histological subtype of liver cancer alcohol use, smoking status, and HBV and
(HCC and ICC). The number of cases of ICC HCV markers (except for community controls),
was however limited and no data on HBV/HCV coffee use during the previous 1–2  years was
status were provided.] associated with a decreased HCC risk using any
Setiawan et al. (2015) evaluated the associa- of the control groups. For coffee use 10  years
tion between coffee intake and HCC of the liver before, comparison between HCC cases and
in 162 022 African-American, Native Hawaiian, either community controls or chronic liver
Japanese-American, Latino, and white subjects disease (CLD) patients revealed a decreased risk.
in the US Multiethnic Cohort (MEC) of Hawaii Against community controls, adjusted odds
and California assembled in 1993–1996. During ratios (95% CI) for occasional use, 1–2  cups/
an 18-year follow-up period, there were 451 inci- day, and ≥  3 cups/day compared with no use
dent cases of HCC. Compared with non-coffee were 0.33 (0.22–0.48), 0.27 (0.15–0.48), and 0.22
drinkers, those who drank 2–3  cups/day had a (0.11–0.43), respectively (P for trend, <  0.001).
38% reduction in risk for HCC (HR, 0.62; 95% Against CLD controls, the equivalent odds ratios
CI, 0.46–0.84); those who drank ≥ 4 cups per day (95% CI) were 0.86 (0.55–1.34), 0.62 (0.32–1.21),
had a 41% reduction in HCC risk (HR, 0.59; 95% and 0.53 (0.25–1.12), respectively. No significant
CI, 0.35–0.99) (P  <  0.002). The inverse associa- trend was observed using hospital patients as
tions were similar regardless of the participants’ controls. [The strengths of this study include the
ethnicity, sex, BMI, smoking status, alcohol multiple centres and multiple types of controls
intake, or diabetes status. [The strengths of this (community, hospital, and CLD). Limitations
study included its prospective design, the long include the possible decrease of coffee use among
follow-up time, its multiethnicity, and the large HCC cases due to their advanced liver disease,
sample size. Limitations included coffee assess- and the fact that caffeine and unfiltered coffee
ment by a single self-report, a lack of information intake could not be evaluated due to uncommon
on liver disease other than HCC, and no infor- use.]
mation on HBV and HCV infection status.]
(b) Hospital-based case–control studies
2.3.2 Case–control studies La Vecchia et al. (1989b) investigated the
association between coffee drinking and the risk
See Table 2.6 of digestive tract neoplasms including cancer
(a) Population-based case–control studies of the liver in a hospital-based case–control
study; 151 cases of liver cancer and 1944 control
Tanaka et al. (2007) conducted a case–control subjects admitted for acute, non-digestive tract
study recruiting 209 incident cases of HCC and disorders in general hospitals from the Greater
three different control sets (1308 community Milan area, Italy, during 1983–1988 were
controls, 275 hospital controls, and 381 patients included. Information on coffee consumption
with chronic liver disease without HCC), all of was collected by interview using a standard ques-
whom were aged 40–79  years and residents of tionnaire. There was no significant or consistent
Saga Prefecture, Japan. A questionnaire survey association between coffee intake and liver
obtained information on coffee use during the cancer. The multivariate odds ratio for consump-
previous 1–2  years and 10 years before, and tion of 2  cups/day and ≥  3 cups/day compared
plasma HBV surface antigen (HBsAg) and with 0–1  cups/day were 0.79 and 0.78, respect-
HCV-Ab were tested for all but the community ively [confidence intervals were not reported].

194
Table 2.6 Case–control studies on cancer of the liver and drinking coffee

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category cases/ (95% CI)
enrolment/ assessment method or level deaths
follow-up
period
La Vecchia Cases: 151 (115 men, 36 Liver/ Coffee consumption (cups/day) Age, sex, social class, Strengths: multicentre
et al. (1989b) women) histologically HCC 0–1 71 1.00 education, marital network, well-defined
Italy, 1983– confirmed cases 2 39 0.79 (NR) status, smoking, alcohol catchment area
1988 Controls: 1944 (1334 consumption Limitations: hospital-based,
≥ 3 41 0.78 (NR)
men, 610 women) patients no virus infection status
admitted for acute, non- Trend test P value, 0.09 adjustment
digestive tract disorders
Exposure assessment
method: questionnaire
Kuper et al. Cases: 333 (283 men, 50 Liver/ Coffee consumption (cups/wk) Age and sex Strengths: virus infection
(2000a) women) HCC cases HCC All 333 – status considered
Greece, Controls: 360 (298 men, subjects Limitations: hospital-based
1995–1998 62 women) hospitalized Non- 36 1.0
for eye, ear, nose, throat, drinkers
or orthopaedic conditions
< 20 230 0.9 (0.5–1.5)
(matched for sex and
5-year age band) ≥ 20 67 0.7 (0.4–1.2)
Exposure assessment Coffee consumption for subjects with virus Age, sex, year of schooling,
method: questionnaire information (330) (cups/wk) HBsAg, and anti-HCV
Non- NR 1.0
drinkers
< 20 NR 1.1 (0.5–2.6)
≥ 20 NR 0.9 (0.4–2.5)
Trend test P value, 0.75
Coffee consumption for subjects without Age and sex
both HBsHg and anti-HCV (82) (cups/wk)
Non- NR 1.0
drinkers
< 20 NR 1.9 (0.6–5.9)

Drinking coffee
≥ 20 NR 1.7 (0.5–5.9)
Trend test P value, 0.66
195
196

IARC MONOGRAPHS – 116
Table 2.6 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category cases/ (95% CI)
enrolment/ assessment method or level deaths
follow-up
period
Gallus et al. Cases: 834 (661 men, 173 Liver/ Coffee consumption (cups/day): Greece and Age, sex, education, Analysis of data from La
(2002) women) HCC Italy combined tobacco smoking, alcohol Vecchia et al. (1989b) and
Italy and Controls: 1912 (1439 Non- 129 1.0 drinking, BMI, history of Gallus et al. (2002)
Greece, men, 473 women), Italian drinkers diabetes and hepatitis Strengths: participation
1984–1997 patients with acute non- Drinkers 705 1.0 (0.7–1.3) almost complete (< 5% refuse
(Italy), 1995– neoplastic conditions interview), confounding
1 231 1.2 (0.9–1.6)
1998 (Greece) (matched for area and factors considered
hospital) and Greek 2 292 1.0 (0.7–1.3) Limitations: hospital-based,
patients hospitalized for ≥ 3 178 0.7 (0.5–1.0) change of exposure after
eye, ear, nose, throat or Trend test P value, 0.015 hospital admission
orthopaedic conditions Duration (yr): Greece and Italy combined
(matched for sex and Non- 705 1.0
5-year age band) drinkers
Exposure assessment
< 30 161 1 (0.7–1.4)
method: questionnaire
30–39 243 1 (0.7–1.4)
≥ 40 294 1 (0.7–1.3)
Trend test P value, 0.864
Table 2.6 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category cases/ (95% CI)
enrolment/ assessment method or level deaths
follow-up
period
Gelatti et al. Cases: 250 (204 men, 46 Liver/ Coffee consumption (cups/day) Adjusted for HBV, HCV, Strengths: virus infection
(2005) women), first diagnosis HCC 0 44 1.0 alcohol intake, sex, age adjusted and stratified
Italy, 1994– of HCC admitted to two 1–2 119 0.8 (0.4–1.3) Limitations: hospital-based
2003 major hospitals
3–4 69 0.4 (0.2–0.8)
Controls: 500 (408 men,
92 women) admitted for ≥5 18 0.3 (0.1–0.7)
other than liver disease, Coffee consumption (cups/day) by HBV
matched with age, sex, infection
date of hospital admission HBV–, 1–2 129 1.0
Exposure assessment HBV–, > 2 61 0.5 (0.3–0.8)
method: questionnaire, HBV+, 1–2 35 16.4 (7.1–38.2)
interview
HBV+, > 2 25 7.3 (3.3–16.1)
Coffee consumption (cups/day) by HCV
infection
HCV–, 1–2 92 1.0
HCV–, > 2 53 0.6 (0.4–0.9)
HCV+, 1–2 70 38.2 (18.2–80.1)
HCV+, > 2 34 9.0 (4.5–17.8)
Ohfuji et al. Cases: 73 primary Liver/ Frequency of consumption (cups/day) before Duration from first Strengths: both cases and
(2006) HCC diagnosis by HCC identification of liver disease identification of liver controls were HCV infection
Japan, 2001– histopathologic Non- 25 1.00 disease, BMI at first positive
2002 examination or imaging drinker identification of liver Limitations: hospital-based,
study from the hospital < 1 19 0.61 (0.18–2.03) disease, disease severity selection bias (all subjects
record at first hospital visit, were HCV+), timing
≥ 1 29 0.38 (0.13–1.12)
Controls: 253, ratio of family history of liver of HCV infection was
1:1–5 matching for age Trend test P value, 0.171 disease, interferon therapy, known for 65% of subjects,
(± 2 yr), sex, the date of Frequency of consumption (cups/day) after smoking, alcohol drinking, imperfect memory of
first hospital visit identification of liver disease other caffeine-containing distant past history of coffee
Exposure assessment Non- 27 1.00 beverage consumption

Drinking coffee
method: questionnaire drinker
< 1 25 0.57 (0.20–1.67)
≥ 1 21 0.19 (0.05–0.71)
Trend test P value, 0.032
197
198

IARC MONOGRAPHS – 116
Table 2.6 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category cases/ (95% CI)
enrolment/ assessment method or level deaths
follow-up
period
Montella et al. Cases: 185 (149 men, 36 Liver/ Coffee consumption (cups/wk) Age, sex, centre, education, Strengths: virus infection
(2007) women) incident HCC HCC Abstainers 27 2.28 (0.99–5.24) smoking habits, maximal status considered, minimal
Italy, 1999– who had not yet received < 14 67 1.00 lifetime alcohol intake, information bias due to same
2002 any cancer treatment at HCV/HBV status interviewer under similar
14–20 50 0.54 (0.27–1.07)
study entry setting between cases and
Controls: 412 (281 men, 21–27 27 0.57 (0.25–1.32) controls
131 women) from same ≥ 28 14 0.43 (0.16–1.13) Limitations: hospital-
hospitals for acute, Trend test P value, 0.02 based, recall and selection
non-neoplastic diseases Decaffeinated coffee consumption (never/ bias, change of coffee
unrelated to diet ever) consumption not considered
Exposure assessment Never 174 1.00
method: FFQ
Ever 11 0.72 (0.21–2.50)
administered by trained
interviewer Coffee consumption (cups/wk) for
HCV–/HBV– (38 cases)
Abstainers 9 2.09 (0.72–6.07)
< 14 13 1.00
14–20 7 0.63 (0.22–1.82)
≥ 21 9 0.38 (0.13–1.09)
Trend test P value, < 0.01
Coffee consumption (cups/wk) for
HCV+/HBV+ (147 cases)
Abstainers 18 2.64 (0.59–11.93)
< 14 54 1.00
14–20 43 0.58 (0.21–1.52)
≥ 21 32 0.84 (0.23–3.01)
Trend test P value, 0.15
Table 2.6 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category cases/ (95% CI)
enrolment/ assessment method or level deaths
follow-up
period
Tanaka et al. Cases: 209 from two large Liver/ Coffee consumption (cups/day) during Sex, age, heavy alcohol Strengths: multicentre study,
(2007) hospitals HCC previous 1–2 yr: community controls drinking, smoking status multiple types of controls
Japan, 2001– Controls: 1308 None 135 1.00 (community, hospital, CLD)
2004 community control, 275 Occasional 53 0.31 (0.21–0.46) Limitations: possible
hospital control, 381 CLD decrease of coffee use among
1–2 15 0.11 (0.06–0.21)
control HCC cases due to their
Exposure assessment ≥ 3 6 0.10 (0.04–0.24) advanced liver disease
method: questionnaire, Trend test P value, < 0.001
interview Coffee consumption (cups/day) during Sex, age, heavy alcohol
previous 1–2 yr: hospital controls drinking, smoking status,
None 135 1.00 HBsAg, anti-HCV
Occasional 53 0.42 (0.19–0.95)
1–2 15 0.23 (0.08–0.68)
≥ 3 6 1.08 (0.22–5.35)
Trend test P value, 0.03
Coffee consumption (cups/day) during
previous 1–2 yr: CLD controls
None 135 1.00
Occasional 53 0.86 (0.55–1.35)
1–2 15 0.42 (0.21–0.84)
≥ 3 6 0.29 (0.11–0.75)
Trend test P value, 0.001

Drinking coffee
199
200

IARC MONOGRAPHS – 116
Table 2.6 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category cases/ (95% CI)
enrolment/ assessment method or level deaths
follow-up
period
Tanaka et al. Coffee consumption (cups/day) during Sex, age, heavy alcohol
(2007) previous 10 yr: community controls drinking, smoking status
(cont.) None 127 1.00
Occasional 53 0.33 (0.22–0.48)
1–2 17 0.27 (0.15–0.48)
≥ 3 12 0.22 (0.11–0.43)
Trend test P value, < 0.001
Coffee consumption (cups/day) during Sex, age, heavy alcohol
previous 10 yr: hospital controls drinking, smoking status,
None 135 1.00 HBsAg, anti-HCV
Occasional 53 0.99 (0.42–2.32)
1–2 15 0.95 (0.31–2.89)
≥ 3 6 2.59 (0.58–11.56)
Trend test P value, 0.47
Coffee consumption (cups/day) during
previous 10 yr: CLD controls
None 135 1.00
Occasional 53 0.86 (0.55–1.34)
1–2 15 0.62 (0.32–1.21)
≥ 3 6 0.53 (0.25–1.12)
Trend test P value, 0.05
Leung et al. Cases: 109 HCC by review Liver/ Coffee consumption (times/wk) Age, sex, cigarette Strengths: HBV carriers
(2011) of medical record HCC No 81 1.00 smoking, alcohol use, tea Limitations: hospital-based
China, Hong Controls: 125 HBV Yes 28 0.54 (0.30–0.97) consumption, and physical
Kong SAR, carriers at the same activity
< 1 86 1.00
2007–2008 hospital
Exposure assessment 1–3 11 0.58 (0.24–1.36)
method: questionnaire, ≥ 4 12 0.41 (0.19–0.89)
face-to-face interview Trend test P value, 0.02
Table 2.6 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, exposure site category cases/ (95% CI)
enrolment/ assessment method or level deaths
follow-up
period
Jang et al. Cases: 258 HCC Liver/ Lifetime amount (cups): HCE (480 cases) Age, sex, BMI, past medical Strengths: results from
(2013) Controls: 480 health- HCC ≤ 20 000 54 1.00 history of DM, lifetime endemic area, multiple
Republic check examinee (HCE), > 20 000 204 0.56 (0.33–0.95) smoking amount, lifetime control (HCE and CLD),
of Korea, 626 CLD alcohol consumption virus infection status
2007–2008 Exposure assessment Lifetime amount (cups): CLD (258 cases) Age, sex, BMI, past medical considered
method: questionnaire ≤ 20 000 54 1.00 history of DM, lifetime Limitations: hospital-based
> 20 000 204 0.55 (0.36–0.85) smoking amount, lifetime
alcohol drinking amount,
chronic liver disease (none,
HCV, HBV, both HCV and
HBV)
Lifetime amount (cups): patients without Age, sex, BMI, past medical
HBV (83 cases) history of DM, lifetime
≤ 20 000 NR 1.00 smoking amount, lifetime
> 20 000 NR 0.47 (0.23–0.94) alcohol consumption
Lifetime amount (cups): patients with HBV Age, sex, BMI, past medical
(170 cases) history of DM, lifetime
≤ 20 000 NR 1.00 smoking amount, lifetime
> 20 000 NR 0.64 (0.36–1.14) alcohol drinking amount,
HBV status
Patil et al. Cases: 141 HCC patients, Liver/ Coffee consumption (cups/day) Age, alcohol consumption, Strengths: viral infection
(2014) consecutive recruitment HCC Never 105 1.00 ALT level, ferritin level, positive only, ferritin level
India Controls: 240 patients Ever 36 2.00 (1.05–3.83) family income, sex, tobacco considered
(Mumbai), with CLD of viral consumption Limitations: hospital-based
≤ 2 20 1.00
2009–2011 etiology, consecutive
recruitment > 2 3 0.37 (0.10–1.34)
Exposure assessment
method: questionnaire
ALT, alanine aminotransferase; BMI, body mass index; CI, confidence interval; CLD, chronic liver disease; DM, diabetes mellitis; FFQ, food frequency questionnaire; HBsAg, hepatitis

Drinking coffee
B virus surface antigen; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCE, health-check examinee; HCV, hepatitis C virus; NR, not reported; SAR, Special Administrative
Region; wk, week(s); yr, year(s)
201
IARC MONOGRAPHS – 116

The inverse exposure–response trend was not Gelatti et al. (2005) conducted a hospital-based
significant (P for trend,  0.09). [The multicentre case–control study in an area of northern Italy.
network and well-defined catchment area were A total of 250 cases of HCC of the liver and 500
the strengths of this study, while limitations controls, hospitalized for any reason other than
included the hospital-based design and lack of neoplasms and liver and alcohol-related diseases,
adjustment for virus infection status.] were recruited during 1994–2003. Lifetime
Kuper et al. (2000a) conducted a hospi- history of coffee consumption was assessed using
tal-based case–control study in Greece. Blood a standardized questionnaire. Coffee consump-
samples and questionnaire data were obtained tion in the decade before the interview was asso-
from 333 incident cases of HCC of the liver ciated with a reduced risk of HCC with a clear
during 1995–1998, as well as from 360 controls inverse dose–response relationship. With respect
matched for sex and age (±  5  years) who were to non-drinking subjects, the odds ratio (95% CI)
hospitalized for eye, ear, nose, throat, or ortho- was 0.3 (0.1–0.7) for ≥ 5 cups/day. [The strengths
paedic conditions in Athens. Information on of this study included adjustment and stratifica-
coffee consumption was collected by interview. tion for virus infection. The hospital-based study
Hepatitis B surface antigen (HBsAg) and anti- design was a limitation.]
bodies to HCV (anti-HCV) were tested for the Ohfuji et al. (2006) conducted a hospital-based
study participants. Coffee intake was not associ- case–control study in Japan to assess the associa-
ated with HCC risk after controlling for age, sex, tion between coffee and HCC of the liver, in which
year of schooling, and HBV and HCV infection both 73 cases and 253 controls were patients
status. Compared with non-drinkers, odds ratios with chronic type C liver disease. A self-admin-
(95% CI) for consumption of <  20 cups/week istered questionnaire was used to assess coffee
and ≥  20 cups/week were 1.1 (0.5–2.6) and 0.9 consumption. The effect of coffee intake was
(0.4–2.5), respectively. [Consideration of virus estimated separately for before and after first
infection status was a strength of this study, identification of liver disease. Coffee drinking
while its main limitation was its hospital-based on a daily basis (≥ 1 cup/day) revealed lowered
design.] odds ratios as compared with non-drinkers both
Gallus et al. (2002) analysed the association before first identification of liver disease (OR,
between coffee consumption and HCC of the 0.38; 95% CI, 0.13–1.12; P for trend, 0.171) as well
liver in the two preceding case–control studies as after disease identification (OR, 0.19; 95% CI,
conducted in Italy and Greece (La Vecchia 0.05–0.71; P for trend, 0.032). Odds ratios were
et al., 1989b; Kuper et al. 2000a). Compared with adjusted for time from the first identification of
non-drinkers, the multivariate odds ratio (95% liver disease, BMI, smoking, alcohol drinking,
CI) adjusting for age, sex, education, tobacco consumption of other caffeine-containing bever-
smoking, alcohol drinking, BMI, and history ages, and clinical characteristics. The inverse
of diabetes and hepatitis was 0.7 (0.5–1.0) for association persisted after excluding subjects who
drinkers of ≥  3 cups/day (P for trend,  0.015). reported a reduction in the frequency of coffee
Duration (years) of coffee consumption was not intake after first identification of liver disease.
associated with risk of HCC. [The strengths of [The strength of this study was that both cases
this study were an almost-complete participation and controls were HCV positive. Limitations
rate (< 5% refused interviews) and consideration included: hospital-based design, generalizability
of confounding factors. It was limited by its (all subjects were HCV-positive), missing infor-
hospital-based design.] mation on the timing of HCV infection (known

202
Drinking coffee

for only 65% of subjects), and imperfect recall of CI, 0.30–0.97) compared with non-drinkers. The
distant past coffee consumption.] study also observed a significant dose–response
Montella et al. (2007) conducted a hospi- association (P for trend,  0.02), with a reduced
tal-based case–control study in Italy that included risk for moderate drinkers (≥  4  times/week) of
185 incident, histologically confirmed cases 59% (OR, 0.41; 95% CI, 0.19–0.89) compared
of HCC aged 43–84  years that were identified with those with no coffee habit (< 1 time/week).
during 1999–2002. Controls were 412 subjects [The main strength of this study was the use
admitted to the same hospital networks as the of HBV carriers to control for confounding by
cases for acute, non-neoplastic diseases unre- infection status. The hospital-based design was
lated to diet. Coffee consumption was assessed a limitation.]
using a validated FFQ. Compared with people Jang et al. (2013) performed a hospital-based
who drank < 14 cups/week of coffee, the adjusted case–control study in the Republic of Korea
risk of HCC decreased for increasing levels of to determine the association between lifetime
consumption with odds ratios (95% CI) of 0.54 coffee consumption and the risk of HCC devel-
(0.27–1.07) for 14–20 cups/week, 0.57 (0.25–1.32) opment in a HBV-prevalent region. A total of
for 21–27 cups/week, and 0.43 (0.16–1.13) for 1364 subjects – 258 HCC patients, 480 health-
≥ 28 cups/week (P for trend, 0.02). An increased check examinees (control group 1, HCE), and
risk was observed among abstainers of coffee 626 patients with chronic liver disease other
relative to people who drank <  14  cups/week than HCC (control group 2, CLD) – were inter-
of coffee (OR, 2.28; 95% CI, 0.99–5.24). Inverse viewed on smoking, alcohol consumption, and
associations were observed across strata of coffee drinking using a standardized question-
HCV and HBV infections and alcohol drinking. naire. HBV e-antigen (HBeAg) status and serum
A non-significant inverse association was HBV DNA levels were measured in patients
observed with consumption of decaffeinated infected with HBV. After adjustment for risk
coffee (OR, 0.72; 95% CI, 0.21–2.50). [The factors, including the presence of hepatitis virus
strengths of this study were the consideration (except for HCE) and lifetime alcohol drinking/
of hepatitis infection status, and minimal infor- smoking, a high lifetime consumption of coffee
mation bias due to the same interviewer being (>  20  000  cups) compared with a low lifetime
used under a similar setting between cases coffee consumption (≤ 20 000 cups) was associ-
and controls. The hospital-based design was a ated with a reduced risk of HCC using both HCE
limitation.] and CLD control groups, yielding odds ratios
Leung et al. (2011) examined whether coffee (95% CI) of 0.56 (0.33–0.95) and 0.55 (0.36–0.85),
has a protective effect in chronic HBV carriers, respectively. The high coffee consumption was
a group at high risk of developing liver cancer, not associated with a significantly increased risk
in a hospital-based case–control study in Hong of HCC; among patients with HBV, the odds
Kong Special Administrative Region, China. ratio was 0.64 (95% CI, 0.36–1.14) after adjust-
A total of 234 HBV chronic carriers (109 HCC ment for HBeAg status, serum HBV DNA level,
cases and 125 controls) were recruited from a and antiviral therapy. [The strengths of this
core hospital during 2007–2008. Data collection study included the fact that results were obtained
included review of medical records and face-to- from a hepatitis endemic area with consideration
face interview. On adjusting for age, sex, ciga- of infection status, the use of multiple controls
rette smoking, alcohol use, tea consumption, (HCE and CLD), . Limitations included its hospi-
and physical activity, coffee drinking signifi- tal-based design and the potential for selection
cantly reduced the risk of HCC (OR, 0.54; 95% bias with CLD controls.]

203
IARC MONOGRAPHS – 116

Patil et al. (2014) analysed the association 0.42–0.75, I2, 74.1%; P for trend, < 0.001) and the
between coffee consumption and HCC of the cohort studies 0.64 (RR, 0.64; 95% CI, 0.52–0.78;
liver in an Indian population that was HCV I2,  69.1%; P  for trend,  0.002). Compared with
and/or HBV positive. The study enrolled 141 no coffee consumption, the summary relative
patients with HCC and 240 patients with HBV risk was 0.72 (95% CI, 0.61–0.84; I2, 58.4%; P for
or HCV infection-related CLD. After adjusting trend, 0.003) for low consumption and 0.44 (95%
for alcohol consumption, ALT level, ferritin CI, 0.39–0.50; I2,  0.0%; P  for trend,  0.495) for
level, and other covariates, ever compared with high consumption. The relative risk was 0.80
never consumption of coffee was associated (95% CI, 0.77–0.84) for an increment of 1  cup/
with an increased risk of HCC (OR, 2.00; 95% day of coffee. The inverse association between
CI, 1.05–3.83) in patients with hepatitis-related coffee and HCC risk was consistent regardless of
CLD. [The strengths of the study included the subject sex, alcohol consumption, or history of
use of HBV- and/or HCV-positive subjects and hepatitis or liver disease. Several cohort studies
the consideration of ferritin level. Limitations reported after 2013 were not included in this
included the hospital-based design, the fact that meta-analysis.
controls were patients with CLD, and the cate- Bravi et al. (2017) recently conducted an
gories of coffee consumption being only never or updated meta-analysis of prospective studies,
ever.] including results from the recent cohort studies
which were not included in the previous
2.3.3 Meta-analyses meta-analysis by Bravi et al. (2013), by performing
a PubMed/MEDLINE and Embase search of arti-
Seven meta-analyses of the association cles published up to June 2015 on cohort studies.
between cancer of the liver and coffee drinking Twelve cohort studies (2154 cases in total) were
have been published (Bravi et al., 2007a, 2009, included in this meta-analysis. Compared with
2013, 2017; Larsson & Wolk, 2007; Yu et al., 2011; no consumption, the summary relative risks for
Sang et al., 2013). The most recent and compre- HCC by random-effect model were 0.66 (95% CI,
hensive meta-analyses are summarized here. 0.55–0.78) for regular, 0.78 (95% CI, 0.66–0.91) for
Bravi et al. (2013) conducted a meta-ana- low, and 0.50 (95% CI, 0.43–0.58) for high coffee
lysis of epidemiological studies that examined consumption, with a significant heterogeneity
the association between liver cancer and coffee (P < 0.001 for I2-statistic). The summary relative
consumption. A PubMed/MEDLINE search risk for an increment of 1 cup/day was 0.85 (95%
from 1966 to September 2012 was performed CI, 0.81–0.90). This meta-analysis supported the
to identify case–control or cohort studies inverse association between coffee consumption
that examined the association between coffee and the risk of HCC.
consumption and cancer or HCC of the liver.
The summary relative risks for any, low, and high
consumption of coffee versus no consumption 2.4 Cancer of the breast in women
were obtained from the results for eight cohort
A total of 23 cohort and 22 case–control
and eight case–control studies. The summary
studies that investigated the association between
relative risk for any coffee consumption versus
coffee intake and of cancer of the breast in
no consumption was 0.60 (95% CI, 0.50–0.71;
women were available for review by the Working
I2, 73.9%; P < 0.001) from 16 studies that included
Group. All but one of the cohort studies inves-
a total of 3153 HCC cases. The findings were
tigated incident breast cancer; the remaining
similar for the case–control (RR, 0.56; 95% CI,
study considered breast cancer mortality. Four

204
Drinking coffee

of the case–control studies investigated breast cancer cases in women (38/2891) and a lack of
cancer in women with known status regarding adjustment for reproductive factors or smoking.
BRCA1/BRCA2 mutations. Four meta-analyses
of the above-indicated studies, published from 2.4.1 Cohort studies
2009 to 2013, are also included in this review.
Thirteen (twelve case–control and one cohort) See Table 2.7
studies were excluded for the following reasons. (a) Incident cancer of the breast
The studies by Lawson et al. (1981), Lubin
et al. (1981), and Franceschi et al. (1995) were Vatten et al. (1990) studied the association
excluded because coffee and tea (and decaffein- between coffee consumption and breast cancer
ated coffee in Franceschi et al., 1995) were exam- incidence using a cohort of 14  593 Norwegian
ined as one combined exposure; the association women (aged 35–51 years) who participated
between coffee and risk of breast cancer could not in a health screening examination for cardio-
be separated from those of the other beverages. vascular disease (National Health Screening
The study by Mansel et al. (1982) was excluded Service) between 1974 and 1977. Age-adjusted
as the study design and analysis were unclear. incidence rate ratios (IRR) in relation to breast
The studies by Lê (1985), Rohan & McMichael cancer risk indicated an overall inverse, non-sta-
(1988), Smith et al. (1994), Zhang et al. (2007), tistically significant association between daily
and Ayari et al. (2013) were excluded as no intake of coffee and risk of breast cancer. There
measure of relative risk for coffee intake in rela- was an indication or effect modification of the
tion to risk of breast cancer was reported. association by BMI (P-interaction  =  0.02). The
The study by Pozner et al. (1986), which risk of breast cancer for coffee consumption of
examined caffeine and coffee intakes in women ≥ 5 cups/day compared with ≤ 2 cups/day yielded
with breast cancer to determine whether they an incidence rate ratio of 0.5 (95% CI, 0.3–0.9;
influence cell differentiation in tumours, was P for trend, 0.02) for BMI < 24. For BMI ≥ 24,
excluded since, as described in the previous IARC an equivalent comparison yielded an incidence
Monographs evaluation (Volume 51; IARC, 1991), rate ratio of 2.1 (95% CI, 0.8–5.2; P for trend,
this study is difficult to group with other studies 0.09). [The limitations of this study included the
of etiology. small number of cases and lack of information/
The study by Männistö et al. (1999), which adjustment for risk factors (apart from age) for
used the association between coffee consumption breast cancer incidence (i.e. reproductive history,
and breast cancer risk as an illustration paradigm hormones, smoking).]
when investigating a methodological issue, was Høyer & Engholm (1992) studied the associa-
excluded because of the influence of recall bias in tion between serum lipids and breast cancer risk,
previous knowledge of health status. reporting also for coffee intake, in a cohort of 5207
The study by Shirlina et al. (2015), which Danish female participants (aged 30–80  years)
investigated nutritional risk factors in association recruited in the Glostrup Population Studies
with breast cancer in the Russian Federation, was between 1964 and 1986. Participants were repre-
excluded as the full text (in Russian) could not be sentative of urban and suburban Danes with
obtained. respect to social class, housing, education, occu-
A cohort study by Jacobsen et al. (1986), pational conditions, and job categories (partic-
investigating the association between coffee and ipation rate 78.5%). During the 4–26 years of
cancer incidence using two Norwegian cohorts, follow-up, 51 incident cases of breast cancer
was excluded due to the small number of breast were identified by linkage to the Danish Cancer

205
206

IARC MONOGRAPHS – 116
Table 2.7 Cohort studies on cancer of the breast and drinking coffee

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, site category or cases/ (95% CI)
enrolment/ exposure assessment level deaths
follow-up method
period
Snowdon & 23 912 (176 BC Breast Coffee consumption (cups/day) Age, sex, meat Breast cancer mortality
Phillips (1984) deaths) among < 1 131 1.0 consumption, smoking Strengths: dietary questionnaire
USA, 1960– white Seventh-day 1 19 1.1 (0.7–1.8) was used by the ACS study;
1980 Adventists (aged record linkage for identification
≥2 26 0.9 (0.6–1.3)
≥ 30 yr in 1960) of cases
Exposure assessment Trend test P value, 0.62 Limitations: particular
method: self- characteristics of studied
administered population may have resulted
questionnaire in reporting bias, coffee
consumption rare, number
of events small (as cancer
mortality and not incidence is
the endpoint), no adjustment for
important risk factors (therefore
residual confounding)
Vatten et al. 14 593 (152 BC Breast Coffee consumption (cups/day) Age Strengths: comprehensive
(1990) cases) among ≤ 2 27 1.0 definition of cases, validation of
Norway, Norwegian women 3–4 62 0.9 (0.6–1.4) questionnaire for coffee intake
1974–1977 (aged 35–51 yr) Limitations: small number of
5–6 42 0.8 (0.5–1.3)
(enrolment), 12 who participated cases, possibility of information
yr follow-up in National Health ≥7 21 0.8 (0.5–1.4) bias, no information/adjustment
Screening Service Trend test P value, 0.37 for important risk factors (e.g.
Exposure assessment reproductive or smoking),
method: FFQ assessment of coffee at baseline
only
Høyer & 5207 (51 BC cases) Breast Coffee consumption (cups/day) Possibly for social Minimum analysis and focus
Engholm among Danish ≤ 2 NR 1.0 class, age at menarche, on coffee intake since main
(1992) women participants 3–6 NR 1.4 (0.6–3.4) menopause status, number exposure was serum lipids.
Denmark, (aged 30–80 yr) of full-term pregnancies, Strengths: random sample of
≥ 7 NR 1.7 (0.7–4.3)
1964–1986 Exposure height, weight, BMI, the general population, linkage
(enrolment), assessment method: Trend test P value, > 0.20 alcohol, smoking (not to cancer registry (regarded as
1964−1986 standardized clear) virtually complete)
(follow-up, questionnaires in all Limitations: most probably RR
4–26 yr) cohorts at baseline are crude
Table 2.7 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, site category or cases/ (95% CI)
enrolment/ exposure assessment level deaths
follow-up method
period
Folsom et al. 34 388 (580 BC Breast Coffee consumption among postmenopausal Age, waist/hip ratio, Caffeine was the main exposure
(1993) cases) among women women number of live births, age of interest.
USA, 1986 aged 55–69 yr in Never or 183 1.00 at first live birth, age at Strengths: use of a large cohort,
(enrolment), 1986 participating in < 1 time/mo menarche, family history the comprehensive identification
1990 (follow- the IWHS 1 time/mo – 78 0.87 (0.66–1.14) of BC, family history of cases, and validated Harvard
up) Exposure assessment 4 times/wk (including family waist/ semi-quantitative FFQ
method: FFQ, hip ratio and number of questionnaire for assessment of
5–7 times/wk 77 0.96 (0.73–1.27)
regular coffee and live births) exposures
caffeine intakes over 2–3 times/ 136 0.98 (0.78–1.23) Limitations: short follow-up
the previous year day period and therefore small
assessed ≥ 4 times/ 106 1.02 (0.79–1.30) number of cases, caffeine and not
day coffee was the main exposure of
Trend test P value, 0.6 interest (and therefore examined
in more detail)
Stensvold & 21 238 women Breast Coffee consumption (cups/day) Age, cigarettes per day, Strengths: comprehensive
Jacobsen (1994) resident in three ≤ 2 22 1.0 county of residence definition of cases, validation of
Norway, Norwegian counties 3–4 69 1.1 (NR) questionnaire for coffee intake
1977–1982 aged 35–54 yr Limitations: small number of
5–6 77 1.4 (NR)
Exposure assessment cases, possibility of information
method: validated ≥ 7 43 1.2 (NR) bias, no information/adjustment
FFQ for coffee Per category 211 1.07 (0.94–1.22) for important risk factors for
consumption increment BC incidence (i.e. reproductive),
assessment of coffee only at
baseline, no CI reported

Drinking coffee
207
208

IARC MONOGRAPHS – 116
Table 2.7 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, site category or cases/ (95% CI)
enrolment/ exposure assessment level deaths
follow-up method
period
Key et al. 34 759 (427 BC cases) Breast Coffee consumption (times/wk) Attained age, calendar Strengths: comprehensive
(1999) women in Hiroshima ≤ 1 151 1.00 period, city of residence, identification of cases and
Japan, and Nagasaki, 2–4 71 1.03 (0.78–1.37) age at the time of the adequate statistical analyses
1969–1970 and participants of bombing, radiation dose Limitations: major exposure
≥ 5 122 1.19 (0.93–1.52)
1979–1980 the Radiation studied was soya foods so coffee
(enrolment), Effects Research Unknown 83 1.11 (0.84–1.46) intake was not examined in
follow-up until Foundation’s Life Trend test P value, 0.258 detail, special characteristics
1993 Span Study of the studied populations,
Exposure assessment use of a non-validated
method: non- dietary questionnaire, lack
validated dietary of information regarding
questionnaire potentially important
confounders
Michels et al. 59 036 (1271 BC Breast Coffee consumption Age, family history of BC, Strengths: population with high
(2002) cases) among women ≤ 1 cup/wk 76 1.00 height, BMI, education, coffee intakes, high response
Sweden, aged 40–76 yr 2–4 cups/wk 33 0.81 (0.54–1.22) parity, age at first birth, rates, comprehensive endpoint
1987–1990 participating in the alcohol consumption, total ascertainment, FFQ validated for
1 cup/day 185 0.99 (0.75–1.28)
(enrolment), large population- caloric intake coffee intake
follow-up for based SMC cohort 2–3 cups/day 763 0.94 (0.79–1.12) Limitations: assessment of coffee
9.5 yr Exposure assessment ≥ 4 cups/day 214 0.94 (0.75–1.28) only at baseline
method: self- Trend test P value, 0.91
administered
semi-quantitative
FFQ, assessing diet
over the 6 mo before
recruitment
Table 2.7 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, site category or cases/ (95% CI)
enrolment/ exposure assessment level deaths
follow-up method
period
Suzuki et al. 14 409 (103 BC cases) Breast Coffee consumption Age, type of health Green tea was the main exposure.
(2004) in Cohort 1 and 20 Never NR 1.00 insurance, age at Strengths: based on two cohort
Japan 595 (119 BC cases) in Occasionally NR 0.78 (0.53–1.13) menarche, menopausal studies in Japan
Cohort 1: 1984 Cohort 2, comprising status, age at first birth, Limitations: small number
(enrolment), women aged > 40 yr ≥ 1 cup/day NR 0.81 (0.55–1.18) parity, mother’s history of cases, coffee not the main
9 yr follow- participating in two Trend test P value, 0.44 of BC, smoking, alcohol exposure so not examined in
up (111 267 population-based drinking, BMI detail
person-years) prospective cohort
Cohort 2: 1990 studies in Japan
(enrolment), Exposure assessment
7 yr follow- method: self-
up (151 882 administered
person-years) validated
questionnaires
covering recent or
usual consumption
Hirvonen et al. 4396 (95 BC cases) Breast Tertiles of coffee intake (mL/day) Age, smoking, menopausal Strengths: close monitoring and
(2006) apparently healthy 0–111 30 1.00 status, oral contraception efficient detection of BC cases
France, 1994 women aged 35–60 112–252 32 1.07 (0.64–1.79) use, family history of BC, due to frequent examination of
(enrolment), yr at recruitment, number of children participants (every year)
≥ 253 33 1.10 (0.66–1.84)
6.6 yr median participating Limitations: some reproductive
follow-up in a controlled, Trend test P value, 0.71 factors as well as HRT and
primary-prevention randomized treatment
trial of vitamins not adjusted for limited
and minerals (SU. generalizability
VI.MAX)
Exposure
assessment method:
questionnaire;
computerized

Drinking coffee
24-hour dietary
record every 2 mo
209
210

IARC MONOGRAPHS – 116
Table 2.7 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, site category or cases/ (95% CI)
enrolment/ exposure assessment level deaths
follow-up method
period
Ganmaa et al. 85 987 (5272 BC Breast All coffee consumption: cumulatively averaged Age, smoking status, Strengths: validated (for coffee)
(2008) cases) women aged and updated BMI, physical activity, FFQ, substantial number of
USA (11 30–55 yr, recruited < 1 cup/mo 837 1.00 height, history of benign cases, ability to examine BC
states), 1976 in the NHS 1 cup/mo – 745 1.01 (0.92–1.12) breast disease, family by ER/PR status, detailed
(enrolment), Exposure assessment 4.9 cups/wk history of BC, weight assessment and repeated
follow-up method: FFQ, change since age 18, age measures of coffee intakes,
5 cups/wk 1335 0.92 (0.84–1.01)
during coffee (caffeinated at menarche, parity, age comprehensive statistical
−1.9 cups/
1980–2002 or decaffeinated) at first birth, alcohol analysis, ability to extensively
day
assessed in 1980, intake, total energy adjust for potential confounders
1984, 1986, 1990, 2–3.9 cups/ 1718 0.93 (0.85–1.02) intake, age at menopause, Limitations: selected cohort of
1994, 1998, day postmenopausal hormone nurses
through a validated ≥ 4 cups/day 637 0.92 (0.82–1.03) use
(for coffee) Trend test P value, 0.14
FFQ, assessing
consumption over
the previous year
Ishitani et al. 38 432 (1188 BC Breast Coffee (caffeinated and decaffeinated) Age and randomized Strengths: validated FFQ, the
(2008) cases) among (cups/day) treatment, as well as, for: substantial number of cases, the
USA, 1992 female US health Almost 274 1.00 alcohol consumption, ability to examine BC by ER/PR
(enrolment), professionals, aged ≥ never BMI, family history of BC, status, comprehensive statistical
average follow- 45 yr when recruited < 1 145 0.97 (0.79–1.18) history of hysterectomy, analysis, ability to extensively
up of 10 yr to the WHS bilateral oophorectomy, adjust for potential confounders,
1 166 0.98 (0.81–1.19)
Exposure smoking status, history long follow-up
assessment method: 2–3 405 1.05 (0.89–1.22) of benign breast disease, Limitations: selected cohort
questionnaire; coffee ≥ 4 191 1.08 (0.89–1.3) age at menarche, parity, of health professionals (not
consumption over Trend test P value, 0.27 age at first birth, physical expected to bias the results),
the year before activity, total energy the lack of repeated measures of
recruitment, the intake, multivitamin coffee intake
validated FFQ from use, age at menopause,
the Nurses’ Health menopausal status, and
Study was used postmenopausal hormone
use
Table 2.7 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, site category or cases/ (95% CI)
enrolment/ exposure assessment level deaths
follow-up method
period
Larsson et al. 61 433 (2952 BC Breast Coffee consumption (cups/day) Age, education, BMI, Strengths: as for Michels et al.
(2009) cases) women aged < 1 251 1.00 height, parity, age at first (2002), repeated measures for
Sweden, 40–76 years from 1 486 1.05 (0.90–1.23) birth, age at menarche, coffee intake, follow-up resulted
1987–1990 the SMC, study age at menopause, use in a substantial number of BC
2–3 1723 0.97 (0.84–1.11)
(enrolment), design and BC of oral contraceptives, cases, information on ER/PR
mean follow- cases ascertainment ≥4 492 1.02 (0.87–1.2) use of postmenopausal status available for majority of
up until described by Michels Trend test P value, 0.74 hormones, family history cases
2009 (17.4 yr; et al. (2002) of BC, intakes of alcohol, Limitations: possibility of
1 071 164 Exposure assessment tea, total energy information bias
person-years) method: as in
Michels et al. (2002),
plus 1997 self-
administered FFQ
to assess long-term
effect of diet on BC
risk
Wilson et al. 90 628 (1179 Breast Coffee consumption: quintiles of servings/day Age, calendar year, BMI, Premenopausal BC was the end-
(2009) BC cases) 1st quintile 270 1.00 height, oral contraceptive point of interest.
USA, 1991 premenopausal 2nd quintile 155 1.11 (0.91–1.36) use, parity and age at first Acrylamide intake was the main
(enrolment), women aged birth, age at menarche, exposure studied.
3rd quintile 230 0.97 (0.81–1.16)
14 yr (945 764 26–46 yr family history of BC, Strengths: similar to those
person-years) Exposure assessment 4th quintile 266 1.01 (0.85–1.21) history of benign breast reported for Ganmaa et al. (2008)
of follow-up method: FFQ, 5th quintile 258 0.92 (0.77–1.11) disease, smoking, physical Limitations: similar to
similar assessment Trend test P value, 0.28 activity, animal fat, those reported for Ganmaa
as for Ganmaa et al. glycaemic load, alcohol et al. (2008), lack of detailed
(2008) intake, total energy intake examination of coffee in relation
to BC risk since acrylamide was
the exposure studied

Drinking coffee
211
212

IARC MONOGRAPHS – 116
Table 2.7 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, site category or cases/ (95% CI)
enrolment/ exposure assessment level deaths
follow-up method
period
Boggs et al. 52 062 (1268 BC Breast Coffee consumption Energy intake, age at Strengths: population-based
(2010) cases) African- Never or < 1 592 1.00 menarche, BMI at age sample, extended follow-up,
USA (all American women cup/mo 18, family history of BC, repeated measures of coffee
regions), 1995 aged 21–69 yr at < 1 cups/day 357 0.98 (0.85–1.12) education, geographic intake, advanced statistical
(enrolment), enrolment in the region, parity, age at first analysis with time-varying
1 cups/day 148 0.91 (0.76–1.09)
follow-up until BWHS birth, oral contraceptive covariates for exposures and
2007 (12 yr) Exposure assessment 2–3 cups/day 122 0.94 (0.77–1.15) use, menopausal status, potential confounders, control
method: validated ≥ 4 cups/day 49 1.03 (0.77–1.39) age at menopause, for a large number of BC risk
FFQ, self- Trend test P value, 0.9 menopausal hormone use, factors
administered at vigorous activity, smoking Limitations: results not
baseline in 1995 and status, intake of alcohol, generalizable to populations
in 2001 tea, decaffeinated coffee other than African-American
women
Nilsson et al. 32 178 (587 cases) Breast Boiled coffee (occasions/day) Sex, age, BMI, smoking, Method of coffee preparation was
(2010) women recruited in < 1 433 1.00 education, recreational the main interest of the study.
Sweden the VIP 1–3 141 1.02 (0.84–1.23) physical activity Discrepancies in tables and
(Västerbotten), Exposure assessment figures regarding the number of
≥4 14 0.52 (0.30–0.88)
1992–2007 method: semi- women and BC cases
(enrolment), quantitative FFQ Trend test P value, 0.247 Strengths: country with very
follow-up until Total/boiled/brewed coffee intakes (occasions/ high consumptions of coffee,
2007 (median day) method of coffee preparation
follow-up < 1 58 1.00 considered, case ascertainment
6.6 yr) 1–3 367 1.06 (0.80–1.40) through high-quality national
≥4 163 0.92 (0.68–1.25) cancer registry
Limitations: low participation
Filtered coffee intake (occasions/day)
rates (57% and 67%), but minimal
< 1 159 1.00 evidence of systematic differences
1–3 328 1.00 (0.83–1.21) in the social and demographic
≥4 101 1.01 (0.79–1.31) characteristics of participants
and non-participants, age used as
a proxy marker for menopausal
status
Table 2.7 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, site category or cases/ (95% CI)
enrolment/ exposure assessment level deaths
follow-up method
period
Iwasaki et al. 53 793 women (581 Breast Coffee consumption Age, area, age at Green tea consumption was the
(2010) cases) aged 40–69 yr, < 1 cup/wk 161 1.00 menarche, menopausal main exposure
Japan, Cohort participants in JPHC 1–4 cups/wk 180 1.15 (0.91–1.46) status at baseline, Strengths: population-
I enrolled in Study age at menopause for based, comprehensive case
1–2 cups/day 173 1.12 (0.87–1.43)
1990, Cohort Exposure postmenopausal women, ascertainment
II enrolled in assessment method: ≥ 3 cups/day 63 1.22 (0.87–1.71) number of births, age at Limitations: relatively low
1993, follow- questionnaire, Trend test P value, 0.26 first birth, height, BMI, consumption of coffee in this
up until assessments at alcohol intake among population, relatively small
31/12/2006 baseline and after regular drinkers, smoking, number of cases, unusual
(average 5 years (1995–1998) leisure time physical analysis, not particularly detailed
13.6 yr) activity, exogenous analysis of coffee
hormone use, family
history of BC, intakes of
green tea, oolong tea, and
black tea
Fagherazzi 67 703 (2868 BC Breast Coffee consumption (cups/day) Age, baseline variables Strengths: substantial number
et al. (2011) cases) French women Non- 410 1.00 (total energy intake, ever of cases, case ascertainment
France, 1990 aged 40–65 yr at consumer use of oral contraceptives, through pathology reports
(enrolment), recruitment, insured ≤ 1 491 1.02 (0.91–1.15) age at menarche, age Limitations: selection of teachers
follow-up until by the national at menopause, number may reduce generalizability of
1.1–3 1133 0.98 (0.85–1.11)
June 2005 health insurance of children, age at first results, lack of repeated measures
(median 11 yr) system > 3 834 1.02 (0.9–1.16) pregnancy, history of for coffee consumption
Exposure assessment Trend test P value, 0.79 BC in the family and
method: self- years of schooling),
administered time-dependent
questionnaire variables (current use of
assessing using diet postmenopausal hormone
over previous year therapy, postmenopausal
women only), personal
history of benign breast

Drinking coffee
disease, menopausal
status, BMI
213
214

IARC MONOGRAPHS – 116
Table 2.7 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, site category or cases/ (95% CI)
enrolment/ exposure assessment level deaths
follow-up method
period
Gierach et al. 198 404 (9915 cases) Breast Coffee consumption Age at entry, race/ Results did not vary by BMI
(2012) female residents Never 1138 1.00 ethnicity, education, or history of benign breast
USA, of eight US states ≤ 2 cups/wk 1114 1.06 (0.97–1.15) BMI, smoking status and biopsy, or by clinical features
1995–1996 aged 50–71 yr when dose, alcohol, proportion of the tumour. No evidence of
3–6 cups/wk 662 1.00 (0.91–1.10)
(enrolment), recruited in NIH- of total energy from fat, an association between breast
follow-up until AARP 1 cup/day 1833 1.02 (0.94–1.09) age at first live birth, cancer risk and either caffeinated
2006 Exposure assessment 2–3 cups/day 3951 1.02 (0.95–1.09) menopausal HRT use, or decaffeinated coffee
method: 124-item ≥ 4 cups/day 1217 0.98 (0.91–1.07) history of breast biopsy, Strengths: large size, availability
food FFQ assessing Trend test P value, 0.38 family history of breast of extensive information on
diet over the past cancer in a first-degree potential confounding factors,
year relative examination of associations for
many clinical features of breast
tumours
Limitations: coffee was assessed
only at baseline
Oh et al. (2015) 42 099 (1395 BC Breast Coffee consumption (cups/day) Age, BMI, duration of Similar patterns of associations
Sweden, cases) women aged 0 99 0.86 (0.69–1.08) breastfeeding, alcohol were observed for pre- and
1991–1992 30–49 yr in the 1–2 338 1.00 consumption, smoking postmenopausal BC
(enrolment), Swedish WLH study, status, education, physical Strengths: population-based
3–4 537 0.87 (0.76–1.00)
follow-up until a random sample of activity sample, extended follow-up,
2012 (856 529 women residing in ≥5 421 0.81 (0.70–0.94) examination of the studied
person-years) the Uppsala Health Per 1 cup/day 1395 0.97 (0.94–0.99) association by ER/PR status
Care Region in increment Limitations: coffee assessed only
Sweden Trend test P value, 0.009 at baseline
Exposure assessment
method: validated
FFQ for coffee/tea
intakes, diet during
previous year
assessed
Table 2.7 (continued)

Reference, Population size, Organ Exposure Exposed Risk estimate Covariates controlled Comments
location, description, site category or cases/ (95% CI)
enrolment/ exposure assessment level deaths
follow-up method
period
Bhoo-Pathy 335 060 (10 198 Breast Coffee consumption (total, caffeinated, Age at menarche, ever use Strengths: substantial
et al. (2015) BC cases) female decaffeinated): postmenopausal of oral contraceptives, numbers of BC cases (even for
10 European participants aged No 732 1.02 (0.94–1.12) age at first delivery, ever premenopausal BC), multi-
countries, 25–70 yr in the EPIC Low 2296 1.00 breastfeeding, smoking country design ensuring
1992–2000 cohort study status, education, physical variation in coffee consumption,
Moderately 1979 0.97 (0.91–1.03)
(enrolment), Exposure assessment activity, alcohol, height, comprehensive statistical analysis
low
follow-up until method: self- or weight, energy intake from Limitations: selected cohorts
2010 interviewer- Moderately 2267 0.97 (0.92–1.03) fat and non-fat sources, (volunteers in most countries),
administered high total saturated fat and fibre lack of repeated assessments of
validated High 1860 0.95 (0.89–1.01) intakes, tea intake, ever coffee consumption (possibly
country-specific Per 100 9134 0.99 (0.98–0.99) use of postmenopausal important after 10-year follow-up)
questionnaires mL/day hormones
(usually FFQs) increment
Trend test P value, 0.055
Coffee consumption (total, caffeinated, Age at menarche, ever use
decaffeinated): premenopausal of oral contraceptives,