Textbook Interpreting Epidemiologic Evidence Connecting Research To Applications Second Edition Savitz Ebook All Chapter PDF

Interpreting epidemiologic evidence:
connecting research to applications

Second Edition Savitz
Visit to download the full and correct content document:
https://textbookfull.com/product/interpreting-epidemiologic-evidence-connecting-resea
rch-to-applications-second-edition-savitz/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Interpreting Complex Forensic DNA Evidence 1st Edition

Jane Moira Taupin
https://textbookfull.com/product/interpreting-complex-forensic-
dna-evidence-1st-edition-jane-moira-taupin/
Foundations of Clinical Research Applications to

Evidence Based Practice 4th Edition Leslie G. Portney
Dpt Phd Fapta
https://textbookfull.com/product/foundations-of-clinical-
research-applications-to-evidence-based-practice-4th-edition-
leslie-g-portney-dpt-phd-fapta/
Interpreting and Using Statistics in Psychological

Research Andrew (Drew) N Christopher
https://textbookfull.com/product/interpreting-and-using-
statistics-in-psychological-research-andrew-drew-n-christopher/
Intermediate Algebra: Connecting Concepts through

Applications 2nd Edition Mark Clark
https://textbookfull.com/product/intermediate-algebra-connecting-
concepts-through-applications-2nd-edition-mark-clark/
Entropy and Information Optics : Connecting Information
and Time, Second Edition Yu
https://textbookfull.com/product/entropy-and-information-optics-
connecting-information-and-time-second-edition-yu/
Applications of Social Research Methods to Questions in

Information and Library Science Second Edition Barbara
M. Wildemuth
https://textbookfull.com/product/applications-of-social-research-
methods-to-questions-in-information-and-library-science-second-
edition-barbara-m-wildemuth/
Cognitive Psychology: Connecting Mind, Research, and

Everyday Experience 5th Edition E. Bruce Goldstein
https://textbookfull.com/product/cognitive-psychology-connecting-
mind-research-and-everyday-experience-5th-edition-e-bruce-
goldstein/
Second Language Pragmatics From Theory to Research 1st

Edition Culpeper
https://textbookfull.com/product/second-language-pragmatics-from-
theory-to-research-1st-edition-culpeper/
Applied Epidemiologic Principles and Concepts:

Clinicians' Guide to Study Design and Conduct 2nd
Edition Laurens Holmes Jr
https://textbookfull.com/product/applied-epidemiologic-
principles-and-concepts-clinicians-guide-to-study-design-and-
conduct-2nd-edition-laurens-holmes-jr/
i
Interpreting Epidemiologic
Evidence
ii
iii
Interpreting Epidemiologic
Evidence
Connecting Research to Applications
Second Edition
David A. Savitz
and
Gregory A. Wellenius
1
iv
1
Oxford University Press is a department of the University of Oxford. It furthers
the University’s objective of excellence in research, scholarship, and education
by publishing worldwide. Oxford is a registered trade mark of Oxford University
Press in the UK and certain other countries.
Published in the United States of America by Oxford University Press

198 Madison Avenue, New York, NY 10016, United States of America.
© Oxford University Press 2016
First Edition published in 2003

Second Edition published in 2016
All rights reserved. No part of this publication may be reproduced, stored in

a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by license, or under terms agreed with the appropriate reproduction
rights organization. Inquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above.
You must not circulate this work in any other form

and you must impose this same condition on any acquirer.
Library of Congress Cataloging-in-Publication Data

Names: Savitz, David A., author. | Wellenius, Gregory A., author.
Title: Interpreting epidemiologic evidence : connecting research to applications /
David A. Savitz and Gregory A. Wellenius.
Description: 2nd edition. | Oxford ; New York : Oxford University Press, [2016]
Identifiers: LCCN 2016002798 (print) | LCCN 2016003547 (ebook) | ISBN 9780190243777
(pbk. : alk. paper) | ISBN 9780190243784 (ebook) | ISBN 9780190243791 (ebook)
Subjects: | MESH: Epidemiologic Research Design | Bias (Epidemiology)
Classification: LCC RA652.4 (print) | LCC RA652.4 (ebook) | NLM WA 950 | DDC 614.4—dc23
LC record available at http://lccn.loc.gov/2016002798
This material is not intended to be, and should not be considered, a substitute
for medical or other professional advice. Treatment for the conditions described
in this material is highly dependent on the individual circumstances. And, while this
material is designed to offer accurate information with respect to the subject matter
covered and to be current as of the time it was written, research and knowledge about
medical and health issues is constantly evolving and dose schedules for medications are
being revised continually, with new side effects recognized and accounted for regularly.
Readers must therefore always check the product information and clinical procedures
with the most up- to- date published product information and data sheets prov ided by the
manufacturers and the most recent codes of conduct and safety regulation. The publisher
and the authors make no representations or warranties to readers, express or implied,
as to the accuracy or completeness of this material. Without limiting the foregoing,
the publisher and the authors make no representations or warranties as to the accuracy
or efficacy of the drug dosages mentioned in the material. The authors and the publisher
do not accept, and expressly disclaim, any responsibility for any liability, loss, or risk that
may be claimed or incurred as a consequence of the use and/ or application
of any of the contents of this material
1 3 5 7 9 8 6 4 2
Printed by WebCom, Inc., Canada
v
Contents
Preface xi
1. Introduction 1
Synopsis 1
Learning Objectives 1
Perspective 1
Approach to the Evaluation of Evidence 3
Organization of Book 4
2. The Nature of Epidemiologic Evidence 7

Synopsis 7
Goals of Epidemiologic Research 7
Measurement of Causal Relations Between Exposure and Disease 11
Applications of Epidemiologic Research 14
Framework for Examining Epidemiologic Evidence 15
Relationship of Epidemiology to Health Policy 16
Exercise: Critical Assessment of Study Methods, Results, and Applications 19
3. Causal Diagrams for Epidemiologic Inference 21

Synopsis 21
Introduction 21
Causal Diagrams in Epidemiology 23
Purpose and Terminology 23
Directed Acyclic Graphs Encode Our Assumptions 24
Statistical Associations 26
Connection to Data Analyses 30
Depicting Passage of Time 32
Direct Versus Indirect Effects 32
Concluding Thoughts 33
Recommended Additional Readings 33
Exercise: Application of Causal Diagrams for Epidemiologic Inference 34
vi
vi Contents
4. Strategy for Drawing Inferences from Epidemiologic Evidence 35

Synopsis 35
Conceptual Framework for the Evaluation of Error 35
Estimation of Measures of Association 37
Systematic Evaluation of Sources of Error 38
Objective Evaluation of Sources of Error 39
Identifying the Most Important Sources of Error 40
Specifying Bias Scenarios 42
Exercise: Specifying Scenarios of Bias 44
5. Confounding I: Theoretical Considerations 45

Synopsis 45
Definition 46
Identifying Potential Confounders 47
Traditional Approach to Assessing Confounding 48
Modern Approach to Assessing Confounding 49
Inappropriate Adjustments 51
Assessing the Direction and Magnitude of Potential Confounding 53
Methods of Controlling Confounding 56
Randomization 57
Selection of Study Setting Free of Confounding 57
Restrict Study Groups to Enhance Comparability 58
Statistical Adjustment for Confounding 59
Recommended Additional Readings 60
Exercise: Conceptual Basis of Confounding 61
6. Confounding II: Practical Considerations 63

Synopsis 63
Evaluating the Presence and Impact of Confounding 64
Specifying Scenarios of Confounding 64
Assessing Whether Confounding Is Present 65
Consider Potential for Complete Confounding 65
Assess Consequences of Inaccurate Confounder Measurement 66
Applying Knowledge of Confounding Based on Other Studies 68
Assessing Confounding When Risk Factors are Unknown 70
Dose-Response Gradients and Potential for Confounding 71
Integrated Assessment of Potential Confounding 72
Exercise: Connecting Conceptual and Statistical Assessment of Confounding 74
vii
v ii Contents
7. Selection Bias and Confounding Resulting from Selection in Cohort Studies 77

Synopsis 77
Study Designs 78
Definition and Examples of Selection Bias 78
Selection Bias Versus Confounding 80
Evaluation of Bias in Cohort Studies 82
Compare Those Included to Those Not Included 82
Compare Disease Rates Among Unexposed to External Populations 83
Assess Whether Expected Patterns of Disease Are Present 83
Assess Pattern of Results Related to Participant Selection 84
Assess Rates for Diseases Known Not to Be Affected by the Exposure 85
Integrated Assessment of Potential for Bias in Cohort Studies 86
Exercise: Assessment of Bias Due to Selection in Cohort Studies 90
8. Selection Bias in Case-Control Studies 93

Synopsis 93
Control Selection 94
Participant Selection in Case-Control and Cohort Studies 94
Selection of Controls from the Source Population 96
Coherence of Cases and Controls 98
Evaluation of Selection Bias in Case-Control Studies 100
Temporal Coherence of Cases and Controls 100
Discretionary Healthcare of Cases and Controls 102
Compare Exposure Prevalence in Controls to an External Population 104
Determine Whether Exposure Prevalence Varies as Expected Among Controls 104
Examine Markers of Potential Selection Bias in Relation
to Measures of Association 105
Adjust Measures of Association for Known Sources of Noncomparability 106
Determine Whether Established Associations Can Be Confirmed 107
Integrated Assessment of Potential for Selection Bias in Case-Control Studies 109
Exercise: Assessing Selection Bias in Case-Control Studies 111
9. Bias Due to Loss of Study Participants 113

Synopsis 113
Conceptual Framework for Examining Bias Due to Loss of Study Participants 113
Evaluation of Bias Due to Loss of Study Participants 118
Characterize Nonparticipants 118
Consider Gradient of Difficulty in Recruitment 119
viii
viii Contents
Stratify Study Base by Markers of Participation 122

Impute Information for Nonparticipants 123
Integrated Assessment of Potential for Bias Due to Loss of Study Participants 124
Exercise: Examining Implications of Nonparticipation 126
10. Measurement and Classification of Exposure 127

Synopsis 127
Introduction 128
Ideal Versus Operational Measures of Exposure 128
Biologically Relevant Exposure 129
Temporally Relevant Exposure 132
Optimal Level of Exposure Aggregation 133
Comparison of Optimal to Operational Measures of Exposure 134
Does Exposure Misclassification Differ by Disease Status? 135
Definitions 135
Mechanisms of Differential Exposure Misclassification 135
Evaluation of Exposure Misclassification 136
Compare Routine Measure to Superior Measures 137
Examine Multiple Indicators of Exposure 138
Examine Subsets of the Population with Differing Exposure Data Quality 139
Evaluate Known Predictors of Exposure 140
Evaluate Known Consequences of Exposure 141
Examine Dose-Response Gradients 141
Evaluate Whether Exposure Misclassification Differs by Disease Status 143
Identification of Subgroups with Nondifferential Exposure Misclassification 145
Integrated Assessment of Bias Due to Exposure Misclassification 146
Exercise: Assessing the Presence and Impact of Exposure Misclassification 147
11. Measurement and Classification of Disease 149

Synopsis 149
Framework for Evaluating Disease Misclassification 150
Sources of Disease Misclassification 151
Impact of Differential and Nondifferential Disease Misclassification 154
Evaluation of Disease Misclassification 157
Verify Diagnostic Accuracy for Subset of Study Participants 157
Examine Results Across Levels of Diagnostic Certainty 159
Evaluate Alternate Methods of Disease Grouping 162
Determine Whether Misclassification Is Differential by Exposure Status 163
Create Subgroups with Accurate Ascertainment or Nondifferential
Underascertainment 165
ix
ix Contents
Restrict Inference to Disease Outcome That Can Be Ascertained Accurately 165

Integrated Assessment of Potential for Bias Due to Disease Misclassification 166
Exercise: Assessing the Presence and Impact of Disease Misclassification 168
12. Random Error 171

Synopsis 171
Nature of Random Variation 171
Sequential Approach to Considering Random and Systematic Error 172
Special Considerations in Evaluating Random Error in Observational Studies 173
Statistical Significance Testing 174
Interpretation of Confidence Intervals 177
Multiple Comparisons and Related Issues 179
Integrated Assessment of Random Error 181
Exercise: Assessing Random Error 183
13. Integration of Evidence Across Studies 185

Synopsis 185
Introduction 186
Systematic Evidence Reviews 186
Data Pooling and Comparative Analyses 186
Meta-Analysis 189
Interpreting Consistency and Inconsistency Among Studies 191
Inconsistent Findings 191
Consistent Findings 194
Evolution of Epidemiologic Research 194
Integrated Assessment from Combining Evidence Across Studies 195
Exercise: Interpreting Evidence from a Collection of Studies 197
14. Characterization and Communication of Conclusions 199

Synopsis 199
Presenting Clear, Objective, and Informed Conclusions 200
Applications of Epidemiology 201
Integration of Epidemiologic Evidence with Other Information 202
Identification of Key Concerns 204
Controversy over Interpretation 205
The Case Against Algorithms for Interpreting Epidemiologic Evidence 206
Exercise: Communicating Summary Assessment of Epidemiologic Evidence 209
Index 211
x
xi
Preface
The expectations for epidemiology as a tool to help guide our individual and collec-
tive choices to improve human health have markedly increased since the first edi-
tion of this book was published in 2003. Epidemiologic methods have continued to
improve, with more sophisticated and formal approaches to assessing causality, and
the realms of application have expanded, with a much greater emphasis on using epi-
demiology to inform not just behaviors and environmental and public health policy
but also health services and clinical medicine.
The fundamental goal for the book remains unchanged: provide guidance for
making sound judgments regarding the inferences that are warranted by the prod-
ucts of epidemiologic research. This judgment requires a balance in recognizing the
considerable power of epidemiology, the study of human experience in the real world,
with its many pitfalls, calling for careful, objective reasoning. The intent of this book
is to provide guidance in the art of interpretation, seeking a methodical and objec-
tive assessment of epidemiologic evidence. One of the most difficult connections for
those new to the field (and even for experienced investigators) is reconciling the tech-
nical aspects of research methods with their application to real-world challenges.
In the course of training and the ever-growing depth and breadth of epidemiologic
technology, we need to reemerge at some point to apply what we have learned in order
to extract the appropriate message to relay to the consumers of our research as well
as to colleagues who are less familiar with the topic. The use of this book requires
grounding in fundamental epidemiologic methods, at least a solid introductory level
course, but not necessarily command of advanced quantitative methods.
The new edition was revised with several specific improvements in mind. We
strived to make it more useful to a broader audience, including public health prac-
titioners, physicians and other health professionals concerned with evidence-based
practice, and experts in health policy and healthcare, for whom a second course fol-
lowing introductory epidemiology may well be their final one. Many of our readers
may never be expected to independently design or conduct epidemiologic studies,
yet they will clearly be making use of epidemiologic evidence quite extensively in the
course of their professions.
For those pursuing doctoral degrees in epidemiology and likely to go on to design
and conduct studies as well as use the information obtained, the challenges of in-
terpretation and application should be introduced periodically through training to
complement courses in quantitative and conceptual methods at the intermediate or
advanced level. Delving deeply into quantitative methods for several years without
resurfacing to assess what it all means can perpetuate an unhealthy separation of the
xi
xii
xii Preface
technology from its purpose. Including the material in this book at some point in the
quantitative methods course sequence may help to prevent methodologic knowledge
from becoming untethered to its purpose. We have had some experience in present-
ing this material to both audiences—students that have completed the full methods
course sequence and those who were in a second course following an introduction to
the field—and this second edition is intended to continue and improve our ability to
serve these two audiences.
We have made a number of significant changes to make the book more useful as
a course text, not just a reference, providing a synopsis and learning objectives at the
beginning of the chapters and discussion questions for consideration at the end of
the chapters. The intent is for each chapter to provide an orientation to the distinctive
issues that should be considered in addressing a particular challenge in interpreta-
tion, for example, assessing the impact of exposure measurement error, and use ex-
amples from the literature to draw out thoughtful discussion of the issue. In that way,
those who are interested in a particular application of epidemiology, for example, for
environmental regulation or clinical guidelines, could choose articles for the discus-
sion that are salient to that specific audience.
Another important feature of the revision is to draw on important and useful
methodologic developments that have occurred over the years since the first edition
was published. The intent is not to provide a menu of intellectually exciting new con-
cepts and tools, but to selectively zero in on those that have the most immediate
value in refining interpretation and application. The repertoire of quantitative and
conceptual tools in epidemiology is vast and well reflected in excellent texts (par-
ticularly Modern Epidemiology, 3rd edition; Rothman et al., 2008), but not all of
those tools are commonly needed to answer the question, “What have we learned
about this issue from epidemiology?” or “How credible are the findings?” We drew
selectively on the technology based on our assessment of the frequency with which
the tool will be useful and the ability to internalize the methodological principle for
routine application.
A dimension of the book that has been strengthened is the relationship between
the attempt to apply epidemiologic findings and the need for new research to expand
knowledge. When we try to use epidemiology to answer an important practical ques-
tion and understand the specific ways in which it falls short, we have defined the
frontier for new studies. The more effectively we can pinpoint why the current evi-
dence falls short, the closer we are to being able to specify the studies that would fill
the gap. There is a growing impatience with epidemiology, a desire to more quickly
inform critical decisions, and we have emphasized the connection between applica-
tions, identification of shortcomings, and the research agenda. This strategy should
be useful to those who set research agendas at the institutional level as well as indi-
vidual investigators who need to argue ever more persuasively in grant proposals for
the importance of initiating new studies.
A component of “application” that has been given more emphasis and a new chap-
ter concerns communication with a variety of nontechnical audiences. That would
include policy makers, the media, attorneys, and more generally, the educated public.
xiii
x iii Preface
Refining our ability to be the bridge between the arcane world of epidemiologic re-
search and the needs and even just curiosity of others would serve the discipline
of epidemiology and society well. Perhaps the most demanding and complete level
of understanding is the ability to accurately convey the state of the science to a lay
audience, and the tools and concepts provided here are intended to strengthen that
capability.
The overarching goal is to draw on and internalize epidemiologic methods to the
point that they become part of our instinctive approach to research—developing epi-
demiologic “common sense.” The separation between the world of complex tools and
jargon, on the one hand, and the way thoughtful people function in the real world,
on the other, can be substantial. When we are talking with a trusted colleague or an
informed friend and they ask, “What’s really going on with X?” or “Should I make
changes in my life based on the research?” we distill a large body of evidence to come
to the bottom line. This book is intended to make that distillation as thoughtful,
informed, transparent, and valid as possible. To put it in its most ambitious form,
the goal is to help readers to develop wisdom in evaluating epidemiologic research.
Reference
Rothman KJ, Lash TL, Greenland S. Modern Epidemiology, 3rd edition. Philadelphia: Lippin
cott, Williams & Wilkins, 2008.
xiv
1
1
Introduction
SYNOPSIS
This chapter provides an introduction to this book’s approach to examining
and interpreting epidemiologic evidence. A distinction is drawn between an
approach that uses formal rules or guidelines and a more nuanced interpreta-
tion based on specific methodologic concerns bearing on a particular study. It
previews the themes the book addresses to assess the validity of epidemiologic
studies for drawing causal inferences, and describes the way the book has been
organized for that purpose.
LEARNING OBJECTIVES
• Become oriented to the art and process of applying methodological principles
to the evaluation of the quality and implications of epidemiologic evidence.
• Appreciate the value and the challenges in using the full array of methodo-
logic considerations and subject matter knowledge to reach an integrated
judgment.
PERSPECTIVE
This book was written both for those who generate epidemiologic research (the
producers) as for those who make use of it (the consumers). In either case, a basic
understanding of epidemiologic principles is necessary at the outset. We will not
introduce novel methods or assume an extensive background in advanced quantita-
tive techniques. The material is intended for those who wish to make the transition
from viewing epidemiology as a purely academic exercise to a means of making wise
decisions about health drawing on epidemiologic evidence, including public health
practitioners, healthcare providers, risk assessors, policy analysts, and laboratory sci-
entists in such disciplines as toxicology or microbiology. The ability to intelligently
characterize the certainty of our knowledge or the strength of the evidence for or
against a specific hypothesis can be of great value to those who must make judgments
among policy alternatives. Where the scientific evidence is incomplete, as it almost
always is, applying the principles presented in this book will not bring certainty or
end controversy, but the approaches we describe will yield a more informed, objec-
tive assessment of the underlying reasons for ambiguity and convey a sense of the
1
2
2 Interpreting Epidemiologic Evidence
solidity or fragility of the available scientific evidence. By pinpointing why or where

the evidence falls short of certainty, we can give questions a sharper focus, leading
to a clearer description of the state of knowledge at any point in time and helping to
pinpoint what research would reduce the remaining uncertainty.
The perspective offered here runs counter to algorithmic approaches to assessing
epidemiologic evidence. The crudest approaches rely on an uncritical acceptance of
the data from a single study or small number of studies. If an association is found,
regardless of the quality of the research that generated it, a causal connection is in-
ferred. While this is obviously wrong, it is not uncommon to find such reasoning
applied in the media and by advocates with a vested interest. The most widely cited
framework for assessing the causal implications of associations is that of Bradford
Hill (1965) (Box 1.1), which continues to be used widely by those who evaluate re-
search findings in epidemiology. Hill’s criteria serve as a set of considerations for
interpreting positive associations, neglecting the need to evaluate the credibility of
an observed absence of association. While Hill notes that these are “considerations”
not “criteria,” they tend to be distilled into a checklist for assessing causality. There
are many parallels between the Hill criteria and the reasoning presented here, but
we attempt to avoid even the illusion of a checklist in how to approach the question
of causality. Interpretation requires consideration of the full spectrum of issues and
the relative importance of those considerations varies markedly from topic to topic.
The significant advances in quantitative and conceptual methods in epidemiology
reflected in recent texts (Rothman et al., 2008; Hernan & Robins, 2015) are notewor-
thy and extremely useful for researchers in the field. For the few experts who have
Box 1.1 Bradford Hill Criteria for Assessing Causality (Hill, 1965)
Strength of association: The larger the association, the more likely that it is to be
causal.
Consistency: Observing the same association in different populations and settings
makes it more likely the association is causal.
Specificity: If there is a single disease associated with the exposure it is more likely to
be causal than if many, unrelated diseases are associated with exposure.
Temporality: The exposure must precede the disease for it to be causal.
Biological gradient: A larger amount of should cause a higher incidence of disease.
Plausibility: A biological pathway that could explain a link between exposure and
disease supports causality.
Coherence: Convergent evidence from both epidemiological and laboratory findings
increases the likelihood of causality.
Experiment: Where feasible, experimental evidence linking exposure to disease sup-
ports causality.
Analogy: A causal effect of exposure is supported by evidence that similar exposures
cause the disease.
3
3 Introduction
full command of these challenging quantitative methods, their application to the

interpretation of epidemiologic evidence to draw causal inferences may be straight-
forward. However, for most epidemiologists (the present authors included), moving
back and forth between the mathematics and architecture of epidemiology to real-
world questions and judgments is challenging.
Between simplistic and often erroneous approaches to interpreting epidemio-
logic evidence on the one end and highly sophisticated but potentially inaccessible
approaches on the other, there is informed judgment. Informed judgment requires
drawing on advanced methods, but doing so selectively rather than comprehensively,
distilling the essence of their implications, and connecting them tightly to applica-
tions. The ability to summarize, interpret, and judge evidence in a manner that can
be transmitted to the educated public not trained in epidemiology is a useful skill
and benchmark, whether or not we in fact have this assignment.
Interest in understanding and using epidemiology has grown considerably, re-
flected in media attention, courtroom applications, and interactions with scientists
in other disciplines. Many outside of epidemiology have one of two extreme reactions
to the evidence we generate. On one extreme some may be so impressed with our
findings based on free living human beings exposed to circumstances that may cause
disease that observed associations are taken as direct reflections of causal effects with
little scrutiny or caution. At the other extreme some may be more impressed with
the many potential sources of error, ubiquitous candidate confounders, and the pre-
dictable controversy among epidemiologists, and therefore come to believe that all
our observations are hopelessly flawed and cannot be trusted as indicators of causal
relations. Students often start with a naive, optimistic view of the power of the epide-
miologic approach, become dismayed with the many sources of potential error, and
then (hopefully) emerge with a sophistication that intelligently balances the promise
and the pitfalls. More thoughtful epidemiologists appreciate that the truth generally
lies somewhere between these two extremes. Even for those who are familiar with
the tools needed to evaluate evidence, however, the integration of that evidence into
a global assessment is typically done subjectively.
APPROACH TO THE EVALUATION OF EVIDENCE

This book is not a step-by-step manual for interpreting epidemiologic data that guar-
antees drawing the correct conclusion; the evaluation of scientific evidence cannot
be reduced to an algorithm for drawing valid inferences. A more attainable goal is
to elucidate the underlying issues involved in the interpretation of evidence so that
unbiased, knowledgeable epidemiologists can reach agreement or identify precisely
where and why they disagree. In doing so, the same issues that need to be explained
to the consumers of epidemiologic evidence will be identified, and the nature of the
disagreement explained in a way that educates and informs those who are asking the
question.
In this book, we have tried to develop in some detail the array of considerations
that should be taken into account to characterize epidemiologic evidence, suggest
4
how to identify the key considerations, and most importantly, offer a variety of strat-
egies to determine whether a potential methodologic problem is likely to be influ-
ential, and if so, what magnitude and direction of influence it may have. The critical
distinction is between what might occur and a judgment about how likely it is to have
affected the results in a material way. The methodologic literature, particularly the
recent synthesis by Rothman et al. (2008), provides the starting point for that evalua-
tion. This book applies methodological principles in specific and practical ways to the
assessment of research findings in an effort to help reach sound judgments.
The fundamental questions about study validity and bias apply to individual re-
ports, so that the approach described in this book can be applied to ask the question
of how much confidence we should place in a particular finding, what the sources of
uncertainty are and how likely they are to have had a material effect on the results.
As we expand the source of information from a single study to a collection of studies,
we have the opportunity to learn from patterns of methods and results. The results
of one study can help to inform the interpretation of another, for example, where a
particular study addresses a potential source of confounding and the insights re-
garding confounding can improve our assessment of the potential for confounding
in another study. As discussed in detail in chapter 13, a set of studies pertaining to a
given topic provides an enhanced opportunity to assess a possible causal relationship
and biases.
ORGANIZATION OF BOOK
The book has been organized to the extent possible in the order that issues arise.
Chapter 2 sets the stage for evaluating epidemiologic evidence by clarifying the ex-
pected product of epidemiologic research, defining the benchmark against which
studies can be judged. A new chapter addresses causal modeling as a means of clear
expression of hypothesized causal relationships to be evaluated through epidemio-
logic research, focusing particularly on directed acyclic graphs (DAGs) (chapter 3).
Next, we propose an overall strategy and philosophy for considering the quality of
epidemiologic research findings (chapter 4).
Confounding is fundamental to interpreting epidemiology; c hapter 5 provides a
conceptual orientation and c hapter 6 focuses on practical approaches to evaluating
confounding. Selection bias has distinctive manifestations in cohort (chapter 7) and
case-control (chapter 8) studies. The specific pathway of selection bias resulting from
loss of study participants is so ubiquitous and important that it calls for a separate
chapter (chapter 9). Measurement error in both exposure (chapter 10) and disease
(chapter 11) is addressed next. A discussion of random error follows the discussion
of systematic biases (chapter 12). Specific issues arising in the interpretation of a set
of studies are considered in chapter 13, and the final chapter addresses the descrip-
tion and communication of conclusions (chapter 14). The order of the chapters was
chosen with some care, moving from the framework to confounding as a central
principle and concern, through specific biases, and then back to a broader perspec-
tive. Nonetheless, the chapters are largely freestanding, referring to one another but
5
5 Introduction
not demanding that the reader retain knowledge from the earlier ones to be able to
understand the later ones.
References
Hernan MA, Robins JM. Causal Inference. http://w ww.hsph.harvard.edu/miguel-hernan/

causal-inference-book/2015.
Hill AB. The environment and disease: association or causation? Proc R Soc Med. 1965;
58:295–300.
Rothman KJ, Greenland S, Lash TL. Modern Epidemiology, 3rd edition. Philadelphia, PA:
Lippincott, Williams & Wilkins, 2008.
6
7
2
The Nature of Epidemiologic Evidence
SYNOPSIS
Assessing the contribution of epidemiologic studies starts with an appreciation
of the goal. In addition to a purely descriptive goal of characterizing disease in
the population, we are often interested in quantifying the causal relationship
between exposure and disease. Study methods and results are scrutinized to
assess the extent to which that goal has been attained. The causal question
can be somewhat challenging to define clearly but should involve some hypo-
thetically modifiable influence on disease to yield practically useful informa-
tion. While epidemiology can contribute to the ultimate goals of improving
health or guiding policy, this cannot be defined as the goal for epidemiologic
research, as it can only do so in combination with other lines of scientific evi-
dence, ultimately dependent on practical considerations as well.
LEARNING OBJECTIVES
• Recognize the connection between the goals of epidemiologic research,
quantifying the causal relation between exposure and disease, and the study
design, execution, results, and interpretation.
• Understand the role of epidemiology in guiding policy and improving
health in the context of other lines of scientific evidence and nonscientific
considerations.
• Be able to provide an evaluation of epidemiologic evidence tailored to differ-
ent audiences and varying applications.
• Recognize the strengths and limitations of epidemiology as a discipline rela-
tive to other scientific approaches to assessing causal relationships between
exposure and disease.
GOALS OF EPIDEMIOLOGIC RESEARCH

Accurate description of the magnitude and pattern of disease is a traditional and
relatively straightforward product of epidemiology, often the starting point for ad-
dressing etiologic hypotheses or planning public health interventions. Analytic epi-
demiologic studies, intended to examine causes of disease, include the measurement
goals of descriptive epidemiology but have additional requirements to be valid. To
7
8
Table 2.1 Criteria for Assessing Effectiveness of Epidemiologic Research

Criterion Assessment
Quality of study design Are the methods for constituting the study population and
assessing exposure, outcome, and covariates appropriate
to generate valid results?
Effectiveness of study Has the design been implemented in an appropriate way
execution to generate valid results?
Clarity of results Do the results provide a clear, internally consistent body
of information that is readily interpreted and explained?
Value of study for Do the study findings provide useful information for
guiding policy making policy decisions that will ultimately be beneficial
to public health?
evaluate the quality or strength of epidemiologic evidence, we need to be clear on

the intended product for a given study or body of research (Table 2.1). The effective-
ness of epidemiologic research must be defined in relation to attainable, specific
benchmarks in order to make judgments about how close the evidence comes to
fulfilling its goals.
Once the goals have been defined, we can assess how the study measures up
against that benchmark, considering specific features of the study methods and prod-
ucts: (1) Study design, whether the study was conceived and planned so that it is capable
of generating valid results; (2) Study execution, whether the data were successfully col-
lected and analyzed in a manner that fulfills the promise of the design; (3) Clarity of
results, whether the information that is generated provides a strong and clear set of find-
ings, not a feature that the investigator controls but a major determinant of study utility;
and (4) Contribution to policy, addressing the value of the knowledge gained for inform-
ing action and ultimately improving health outcomes. In sequence, each is necessary but
not sufficient for success of the next. The big picture view of research is presented in this
chapter, with more refined, extended discussion in the following chapters.
Study design refers to the plan and architecture of the research, whether the study
groups, variables, and general plan for the study are soundly conceived. This requires
bringing together subject matter expertise on the substantive questions with expertise
in research methods. In this way, the work will be informed by relevant studies that
have come before, advancing knowledge through new and improved features. In the
design phase, we determine whether the study will be randomized or observational,
whether the structure will be a cohort study or employ case-control sampling, and
what the key exposures, outcomes, and covariates will be and how they will be as-
certained. To a large extent, the fate of the study is sealed, at least in the negative
direction, if these decisions are not well suited to the goal of the research. There are
still abundant opportunities for a well-conceived study to go awry but no way to
rescue one that is fundamentally flawed from the outset. Thus the first criterion for
evaluation of epidemiologic research is to consider whether the chosen study design
provides the capacity to generate informative, useful results.
9
9 The Nature of Epidemiologic Evidence
Study execution involves the implementation of the design that was selected. This
includes constituting the study population, which may involve recruitment, ascertain-
ment through records, sampling, or other means of engaging the study participants.
Information may be ascertained on those study participants on the exposures, out-
comes, and covariates (potential confounders or effect modifiers) through a variety of
means such as records (paper or electronic), interviews, clinical examinations, biological
assessments, or environmental monitoring. The effectiveness at this stage is combined
with the soundness of the design to determine the validity of the results that are gener-
ated. Even with an optimal design, if the study execution falls short through such famil-
iar pitfalls as unsuccessful recruitment of participants or poor quality measurement of
key information, the validity of the results will suffer. On the other hand, an elegantly
executed but poorly conceived study will also suffer from loss of validity. An apt but
unoriginal guiding principle is that “If it’s not worth doing, it’s not worth doing well.”
Clarity of results refers to the degree to which the results generate a simple story,
internally consistent and easily summarized. This would be the case when a study is
clearly positive, with strong associations, dose-response gradients, demonstrably free
of confounding, and so forth. Or it may mean that the results are unambiguously null.
This is not under the control of the investigator, and the temptation to present mixed
or ambiguous results as clear and convincing should be avoided. A good design and
execution of the study may assure validity but does not guarantee that the data will
cooperate and yield clear information, nor does a flawed study design or execution
mean that the results will be ambiguous in appearance. Much of this book pertains to
thoughtful, informative approaches to characterizing ambiguous evidence.
The final step determining the success of epidemiologic research is to ask whether
it contributes meaningfully to applications that ultimately yield improvements to
public health. Strong research methods and even clear results do not guarantee that
the information will yield benefit to policy makers that translates into improved
health outcomes. Falling short in the earlier steps clearly precludes that possibility,
however. One of the themes in this book is that with a sound study design and effec-
tive study execution, the results, even if inconclusive, should still be contributory to
policy applications. Ambiguous, even internally inconsistent findings, if based on ap-
propriate methods, provide evidence that policy makers should take into account. By
delving more deeply into the evidence to understand it and communicate it clearly,
policy decisions are certain to be wiser.
Epidemiology, like all scientific disciplines, benefits from scrutiny and criticism.
That provides the basis for improvements in the methods of our discipline and
strengthens the quality of the research that results. However, there are reasons to
resist looking to public health impact as the definitive criterion for evaluating re-
search, particularly individual studies. This may seem defensive—particularly as the
ultimate goal of epidemiologic research is in fact public health benefit—but we have
to focus on measurable, attainable indicators of success as the building blocks that
lead to improved health. Generating valid results informing important substantive
questions is sufficiently ambitious and a more honest representation of what epide-
miology has to offer. Overreaching into claims that such research “improves health”
10
not only sets the bar beyond what is attainable but also has some unintended adverse
consequences for both the field of epidemiology and policymaking in public health.
While epidemiology is a powerful approach to understanding and improving
health, we must guard against epidemiologic chauvinism that undervalues other re-
search approaches that inform policy as well as important considerations unrelated
to research such as ethics and politics. Focusing on “improving health” as the goal
of epidemiologic research may tempt investigators to overinterpret or underinterpret
their data rather than seeking to provide the clearest, most unbiased interpretation at-
tainable. Beyond the unquestionable desire for researchers to exaggerate the certainty
of their findings to elevate the stature of the study, upgrade the journal in which it
appears, and thus advance their careers, the ostensibly altruistic motive of improv-
ing public health can also intrude on objectivity. Researchers need to focus single-
mindedly on generating valid information and communicating that information
clearly, not exaggerating or downplaying results to “help” reach the “right” decision.
The most profound, lasting contributions from epidemiology come from carefully de-
signed and executed studies addressing well-chosen topics, properly analyzed, and
explained to researchers and policy makers in an articulate and unbiased manner.
At the other extreme, the bar for successful epidemiologic research can be set
so low that success is guaranteed even if there is little actual value to the product.
We could define the goal of epidemiology as the mechanical process of gathering
and analyzing data and generating statistical results, such as odds ratios or regres-
sion coefficients, divorced from potential inferences and applications. Theoretical
and logistical challenges disappear one by one as the benchmark is lowered suc-
cessively. If a study’s intent is defined as assessment of the association between
the boxes checked on a questionnaire and the reading on the dial of a machine for
those individuals who are willing to provide the information, then success can be
guaranteed. We can undoubtedly locate pencils, get some people to check boxes,
find a machine that will give a reading, and calculate measures of association.
Focusing on the mechanical process of the research is conservative and modest,
traits valued by scientists, and averts the criticism that comes when we attempt to
make broader inferences from the data. While in no way denigrating the impor-
tance of study execution (sharp pencils may actually help to reduce errors in coding
and data entry!), these mechanical components are only a means to the more in-
teresting and challenging end of extending knowledge that has the potential for
biomedical and societal benefit.
At a slightly less ridiculously low level of aspiration, expectations for epidemi-
ology are sometimes couched in such vague terms as “measuring statistical asso-
ciations” or “producing leads,” recognizing that scientific advances typically require
corroborative research from other disciplines, often basic biomedical or clinical sci-
ences. Convergent evidence from multiple disciplines is essential in many cases and
always valuable, but the suggestion that epidemiology only sets the stage for others
is mistaken. Appropriately ambitious epidemiologic studies should be pursued and
held to a high standard, not evading our responsibility through lowering expecta-
tions. Epidemiologists can and do go well beyond making agnostic statements about
11
associations (ignoring causality) or generating hypotheses for other scientists to

pursue. Epidemiology produces evidence, like other scientific disciplines, that con-
tributes to causal inferences about the etiology and prevention of disease in human
populations.
For the purposes of this book, we define the goal for epidemiologic research as
the quantification of the causal relation between exposure and disease. Although the
research itself generates only statistical estimates of association as an initial output,
the utility of those estimated associations in advancing science and ultimately public
health generally depends on the extent to which they provide meaningful informa-
tion on the underlying causal relations. The ideal study yields a quantitative measure
of association that reflects the causal influence of exposure on disease. Methodologic
problems and errors cause a deviation between the study results and this ideal mea-
sure, and improvements in research bring the study results closer to an accurate mea-
sure of the causal effect.
MEASUREMENT OF CAUSAL RELATIONS

BETWEEN EXPOSURE AND DISEASE
Estimation of causal effects as the focus of epidemiology was initially emphasized
by Rothman (1986). This is a straightforward, ambitious goal: quantify the causal
impact of exposure on disease. We are not designing studies just to measure associa-
tions, but as the means to an end, which is to quantify causal effects. Some would
argue that the focus of epidemiology is on guiding or improving public health, and
those aspirations are quite relevant to our choice of topics as well as the audience for
the knowledge we generate. However, the fundamental building block contributed by
epidemiology is information on cause and effect relationships.
Epidemiologists and those who use epidemiologic evidence sometimes shy away
from such an explicitly high aspiration, measuring causal relationships. One inhib-
iting factor is that identification of such universal truths can only be inferred with
varying degrees of confidence, never proven, so our goals are never fully attained. We
cannot prove unequivocally that a particular exposure caused disease in an individ-
ual because that would require knowing with certainty what would have happened
had the exposure been different. As discussed in detail in the context of confounding,
inferring causality is equivalent to inferring what would have happened under other
circumstances that did not in fact occur. Causal inference is inherently conjectural,
and the methods of epidemiology are intended to refine our ability to make informed
conjecture, never proving causality.
A related criticism of stipulating the quantification of causal effects as the goal of
epidemiology is the mistaken view that causality can only be demonstrated in stud-
ies in which exposure is randomized. This is incorrect on two levels: First, it must
be recognized that causality is never proven, remaining an informed judgment no
matter what scientific tools are used, including randomization. Second, while ran-
domization is a powerful tool for reducing confounding, it is not a magic bullet that
automatically puts all other concerns to rest. There are a number of methodological
12
challenges to inferring causality, and focusing on only one (albeit an important one)
distorts the evidence. The cliché that “epidemiologists study associations, experi-
mental sciences study causation” is wrong on both accounts.
With measurement of causal relations as the goal, assessment of epidemiologic
evidence focuses on the aspects of study design, conduct, and analysis that may in-
troduce distortion or enhance the accuracy of measurement. We work to identify
and eliminate, where possible, causes of deviation between what we measure and
the causal relationship of interest. Error, a deviation between the measured result
and the true causal relation between exposure and disease, arises from both random
and systematic processes. There is no fundamental distinction between accurately
measuring a null association and any other association, despite the framework of
statistical hypothesis testing that focuses on the deviation (or lack thereof) between
the study results and those predicted under the null hypothesis. The null hypothesis
or lack of association is just another possible state of nature that a valid study seeks to
identify. Measurement of a causal relation between exposure and disease focuses on
the quantitative index that characterizes the strength of association, which can be a
ratio or difference measure, or a regression coefficient in which disease is the depen-
dent variable, but a p-value is not a quantitative measure of a potential causal effect.
Causal inference would be easier if there were a simple algorithm for doing so
accurately, but there is no methodical process that generates a definitive (or even
a probabilistic) conclusion. Despite not having been designed for that purpose, the
Bradford Hill considerations for inferring causality are sometimes used in that
manner (Hill, 1965). Neither those nor any other algorithmic approach will lead to
an unambiguous conclusion regarding the extent to which the measured association
accurately reflects the magnitude of the causal relationship of interest. In practice, a
series of uncertainties preclude doing so with great confidence, with the list of alter-
native explanations limited only by the imagination of critics. Judgment regarding
whether a particular alternative explanation has truly been eliminated (or confirmed)
is itself subjective. Hypotheses of bias may be more directly testable than the hypoth-
esis of causality, but they remain challenging to definitively prove or disprove. The
culmination of the examination of candidate contributors to bias is a judgment of
how plausible or strong the distortion is likely to be and how confidently such an as-
sertion can be made rather than a simple dichotomy of present or absent. Thus, the
answer to the ultimate question of whether the reported association correctly mea-
sures the etiologic relationship will at best be “maybe,” with the goal of accurately
characterizing where the evidence fits within that spectrum.
Epidemiology is well suited to address a wide range of exposures and diseases,
not just the prototypic chemical or drug causing a well-defined disease. Exposure
includes any potential disease determinant, encompassing age, gender, time period,
social conditions, geographic location, and healthcare in addition to more conven-
tional individual exposures such as diet, stress, or exposure to chemical pollutants.
Conceptually, there is an important difference between potential disease determi-
nants that are clearly modifiable at least in principle (exposure to exogenous agents
such as chemicals or microorganisms, behaviors such as tobacco use or physical
13
activity, social conditions such as poverty or discrimination) and those that are
not. Nonmodifiable characteristics include traits such as sex, age, and ethnicity as
well as geographic setting and time period. Conceptually it may be of interest to
contemplate the causal effect of such nonmodifiable attributes in the abstract (e.g.,
“What if the Hispanics in our study had instead been Anglo?”) to understand the
etiologic process, but the public health value may not always be so clear. Observing
patterns of risk in relation to nonmodifiable factors can, of course, suggest etiologic
hypotheses regarding modifiable risk factors. Insofar as we find differences associ-
ated with Hispanic ethnicity, for example, we might ask about an etiologic impact of
discrimination, cultural practices, healthcare, or other modifiable determinants of
the observed pattern. Anticipating the application of what we learn can help to refine
the study goals so we are not just examining the effect of “sex” or “race” on disease
but rather seeking to understand the effect of levels of steroid hormones or racism,
potentially modifiable influences.
More subtle is the distinction between endogenous attributes, which may be inter-
mediate on the pathway from exposure to disease, and their exogenous determinants.
Body mass index (BMI) is modifiable, of course, but we cannot envision a hypothetical
experiment in which BMI is assigned randomly. We can readily conceive of studies in
which diet and physical activity are randomly assigned, with consequences for BMI,
but not BMI itself. Epidemiologic studies of theoretically modifiable determinants are
conceptually clearer and of greater public health value than those in which the expo-
sure is only indirectly related to a potential cause of disease. Focusing on causal effects
helps to anchor the research on questions of practical value in improving health.
Similarly, disease is used as shorthand for all health conditions of interest, includ-
ing clinical disease, disability, physiologic alterations, and social disorder. To fit within
the framework of epidemiologic inquiry applied in this book, the health measure
should be of some ultimate clinical or public health relevance. We would probably ex-
clude from the scope of epidemiology efforts to predict cigarette brand preference, for
example, even though it is important to public health, or voting patterns or migration,
for example, even though the tools used by marketing researchers, political scientists,
and sociologists are very similar to those of epidemiologists. Once the realm of health
is defined, exposure constitutes everything that potentially influences it.
The exposures and diseases we wish to study are often abstract constructs that
cannot be directly measured. Thus, the data that are collected for study are not direct
assessments of the exposure and disease but only operational measures based on
available tools such as questionnaires, biological measurements, and findings from
physical examination. Some such measures come closer than others to capturing the
condition or event of ultimate interest. Nevertheless, it is important to keep in mind
that the operational measures are not the entities of interest themselves (e.g., deposi-
tion of graphite on a form is not dietary intake, a peak on a mass spectrometer print-
out is not DDT exposure, an income level is not social class), but serve as indirect
indicators of broader, often more abstract, constructs.
A key issue in evaluating epidemiologic evidence is how effectively the opera-
tional definitions approximate the constructs of ultimate interest. The concept of
14
misclassification applies to all the sources of error between the operational measure
and the constructs of interest. The most obvious and easily handled sources of mis-
classification are clerical error or faulty instrumentation, whereas failure to properly
define the relevant constructs, failure to elicit the necessary data to reflect those con-
structs, and assessment of exposure or disease in the wrong time period illustrate
the more subtle and often more important sources of misclassification. Studies can
only yield a measure of association between an operational measure of exposure and
disease, and the nature and magnitude of our ultimate interests and what we actually
achieved calls for careful scrutiny.
APPLICATIONS OF EPIDEMIOLOGIC RESEARCH

If accurate estimation of causal relations is the goal of epidemiologic studies, then suc-
cess has been attained when the measure of association accurately quantifies the causal
impact of exposure on disease. While the observation can only come from a particular
study population, setting, and time period, the inference is intended to be generaliz-
able. Extrapolation of findings to previously unstudied populations, by definition, goes
beyond the available data, and is thus vulnerable to error in addition to whatever error
is contained in the studies that provide the basis for the extrapolation. Postulating uni-
versal causal relations (“smoking causes lung cancer”) reflects the ultimate extrapola-
tion, synthesizing a series of individual studies into the untestable assertion about what
exposure would do to disease risk in all possible past, present, and future populations.
Nonetheless, when we apply epidemiologic evidence to guide decisions about individ-
ual behavior and public policy, we are implicitly extrapolating a set of research observa-
tions to just such new and previously untested situations and populations. Causality is
assessed based on judgments about the validity of a set of studies, but the use of such
information to guide future policy and behavior reflects a further extrapolation into
new time periods and populations that have not been studied before.
Application of epidemiologic evidence to other populations, to individual
decision-making, or to public health policy requires caution. Even if the quantifica-
tion of causal effects in a particular study population is valid, the potential for effect
modification arises whenever the population of interest changes. Causes of disease in
a socioeconomically advanced society (well nourished, access to medical care, etc.)
may not apply in the same way to populations in resource-poor settings, for example.
Extrapolation of quantitative evidence on causality is subject to modification by the
prevalence of other causes of disease (Pearce, 2011). The dominant causes in one so-
ciety versus another may truly differ even with complete and accurate knowledge of
causal influences.
There are pragmatic issues to confront in applying epidemiologic evidence to
guide policy. Feasibility and cost of modifying exposure through behavior change,
clinical guidelines, or regulation has to be incorporated into decision-making, along
with unintended consequences of an intervention. Ethical considerations regarding
the desired actions need to be taken into account. The application of epidemiology
to policy or other societal or individual actions lies outside the scientific discipline
Another random document with
no related content on Scribd:
"schizogamy" makes its appearance in certain Syllidae, resulting in the formation
of two morphologically and physiologically distinct individuals which lead
independent lives. The appearance of a head and of a zone of proliferation leading
to the formation of a chain of sexual zooids is accompanied by a delay in the
appearance of the genital organs, for in Autolytus these arise during the formation
of the new individuals, as part of the general process of new formation; whilst in
Myrianida the delay is prolonged, and the generative elements do not make their
appearance till after the new individuals have reached some size.
More simple cases of the separation of the body into two parts, sexual and
asexual, occur also in some of the Serpulidae. Thus in Filigrana and Salmacina
the generative elements make their appearance in the hinder segments, as they
do throughout the Sabelliformia; and this hinder part of the body separates from
the anterior region after the formation of a new head between the two regions.[336]
Another modification of the process of budding and fission is exhibited by Syllis

ramosa, one of the most interesting forms of animal life which was obtained by the
"Challenger." This worm consists of a main stem, whence arise a number of lateral
branches, which may also branch so as to give rise to an arborescent colony (Fig.
150). The branches of the first and second and higher orders arise by budding
from the sides of the original form or branches of lower order; and some of these
branches develop generative products, and bud forth a head near the point of
attachment. These sexual branches, no doubt, separate from the colony and
distribute the ova. The worm lives in a Hexactinellid sponge, Crateromorpha
meyeri, living in depths of 95 to 140 fathoms in the Eastern seas.[337]
Fig. 150.—Portion of Syllis ramosa. (Reduced from M‘Intosh.)
Regeneration of lost Parts.—The process of budding and fission of the

worm into two parts is merely an extension of that resulting in the formation of
new segments when the worm is injured. In most of the Nereidiform
Polychaetes the number of segments forming the body continues to increase
throughout life by the formation of new segments between the anal segment
and the one in front of it; that is to say, there is normally a process of budding
taking place at this point. Now in many of the longer worms it may be noticed
that the segments of the hinder end suddenly become smaller than the rest;
these are segments newly formed to replace those lost by the worm. But this
"regeneration," though the same in principle as ordinary growth at the
penultimate segment, is due to activity in a segment (any segment) further
forwards; in other words, in the less modified worms every segment has the
power of forming new tissues, just as each of the joints of a crab's leg has the
power of forming the remaining joints when injured. It is not therefore
surprising that a "zone of budding" arises in an uninjured worm at certain
seasons, viz. that of reproduction; it is a property that each worm possesses,
though generally it remains latent till injury provides the stimulus.
Moreover, not only can new segments arise at the hinder end, but a new head
can be formed at the anterior end, as has been observed in worms belonging
to many families—in the less modified Syllidae,[338] in others of the
Nereidiformia, and even in Sabellids, where the greatly specialised gill
filaments can be reproduced. Thus Sir J. Dalyell[339] noted in Dasychone that
the crown of branchiae was regenerated in about a month in springtime, while
in winter the process occupied 116 days. He cut a Dasychone into three
pieces; the hindermost produced a head, the anterior piece developed an
anus, and the middle portion formed both a head and tail!
These regenerated heads are of course at first smaller than the rest of the
body, but soon grow to a normal size. Naturally this extensive power of
regeneration is of extreme value to the Polychaetes, for if a fish or other
enemy bites the head off a worm, a new one can form; and it is not difficult to
see in this the origin of the reproduction by fission as a normal process.
CHAPTER XI
NATURAL HISTORY OF POLYCHAETES—GENERAL HABITS—CHARACTER OF TUBE

AND ITS FORMATION—COLOURING—PROTECTIVE AND MIMETIC DEVICES—
PHOSPHORESCENCE—FOOD—USES—ASSOCIATED WORMS—WORMS AS HOSTS—
DISTRIBUTION—FOSSIL REMAINS.
All the many hundreds of species of Polychaetes are marine, with a very few
exceptions, which have been in recent years recorded from fresh (i.e.
drinkable) water, viz. a species of Nereis from a lake in Mingrelia, another
Nereis and a Lumbriconereis from running water in Trinidad,[340] a Sabellid,
Manayunkia speciosa,[341] from Philadelphia; and another Sabellid,
Coabangia,[342] from fresh water at Tonquin, which lives in borings in shells of
Melania; and it is by no means improbable that other fresh-water Polychaetes
exist in Lake Tanganyika in Africa, where a Medusa has recently been
discovered.
In brackish water of various densities many Polychaetes live; Arenicola

especially is regardless of the character of the medium, and Nereis
diversicolor appears to withstand considerable admixture of fresh water.
The majority of the Polychaetes occur "inshore," that is, between tide-marks
and in shallow water down to 20 fathoms; but they occur at all depths more or
less abundantly, and some have been dredged from depths of more than
3000 fathoms.
The nature of the soil composing the shore has a good deal to do with the
number of worms to be found there; thus in calcareous districts they are fewer
than in places where harder rocks, such as granite, form the shore line, for
the chalk or limestone wears away more quickly, and exposes to destruction
the worms which may have sheltered in its crevices: further, it does not give
so permanent a place of attachment to seaweeds, on which many
Polychaetes feed. The calcareous rocks, too, are more likely to be traversed
by springs of fresh water, which is not to the taste of the worms. The sand
resulting from the destruction of the rocks, whether hard or soft, is of itself
unsuitable to the majority of worms, which are most abundant where mud
containing decaying vegetable matter is mixed with the sand: this, which
gives a firmer consistency to the soil, so that the burrows retain their form
better, supplies food for the burrowers.
General Habits.—The division of the Polychaetes into the "Errantia" or free-
swimming and wandering forms, and "Sedentaria" or tubicolous and
sedentary forms, is a misleading mode of classification, for as a matter of fact
only a comparatively few forms are really free-swimming throughout life; the
majority, even if they do not form definite tubes, burrow galleries for
themselves in the soil, and these burrows are in many cases only rarely left;
this is true of both groups. Amongst the "errant" Polychaetes nearly all the
Eunicidae secrete a parchment-like tube, and some Polynoids form mud
tubes. Among the "sedentary worms" there are forms which merely burrow;
while Myxicola readily leaves its gelatinous tube and swims freely; Pectinaria
carries its house with it as it moves about, and Polycirrus, a Terebellid, does
not form any tube at all.
Owing to their sedentary habits, quite a representative collection of genera

may be made, especially at a spring tide, at any seaside place which is
provided with a sandy shore, and with rocks and seaweed. The larger
species, however, require to be dredged, and the best time is at night, for
then many forms which during the day are concealed in their burrows, will be
issuing forth to obtain food.
It may be useful to give instances of worms occurring in various situations

between tide-marks. Throughout pretty well the whole of the area left
uncovered by the tide, even up to nearly high-water mark in many parts of the
coast, the cylindrical "castings" of sand and mud, forming little heaps, indicate
the burrows of Arenicola, the common "lug-worm"; these "castings" have
passed through the worm's body, having been swallowed during the process
of burrowing as well as for the purpose of obtaining food, as in the case of the
earthworms. Rather nearer the water may be seen little tufts of sand-threads,
about an inch high, springing from a short piece of cylindrical, sandy tube
rising up out of the sand; this is the head end of the tube of Terebella
conchilega (Fig. 153).
Amongst the rocks may be found loose stones of different sizes; on lifting
them up, various kinds of worms may be brought to light, according to the
locality, the time of year, the position with respect to the sea, and so on.
Polynoë is pretty sure to be present somewhere near low-tide mark; the
number of species is considerable, and their colouring very varied: but as the
worms have a habit of remaining still on the under surface of the uplifted
stone, the observer may easily overlook them.
Other worms occur below the stones, more or less buried in the sand or mud;
for instance, a small Nereis may be lying in its temporary burrow immediately
underneath, and will at once withdraw from the now injured part of the burrow;
while deeper in the mud or sand, especially in rather highly-smelling mud,
little red worms are abundant, such as Scoloplos, Nerine, Capitella, and
others. By digging near low water one may find Nephthys, Glycera, and
others burrowing or hiding in the soil.
In rock pools, or sandy stretches amongst rocks kept moist and cool by
abundant Fucus, one may see under stones the red or yellow gill filaments of
Cirratulus and of Terebellids protruding from their burrows and tubes, while
other worms are to be met with in clefts of the rocks, and amongst the roots of
Laminaria.
Still farther out, below low-water mark, where one must wade, can be seen
the beautiful branchial crowns of various Sabellids protruding from their tubes;
but care is necessary on approaching these worms, as eyes are, in many
cases, present on the branchiae and a shadow is readily perceived; then the
brightly-coloured tuft disappears, and only a piece of sandy or muddy
cylindrical tubing remains to tell where the Sabella has withdrawn. In order to
obtain the worms one must dig quickly and deeply before they have been
disturbed; for the tube is of considerable length, and the inhabitant withdraws
to the bottom of it. Some of these soft-skinned worms have the power of
boring into hard rocks,[343] though by what means they do so is uncertain.
[344] Polydora ciliata makes a tube of mud projecting from the mouth of U-
shaped galleries in chalk, limestone, shells, and even shale; it has no hard
jaws or other structures sufficient to account for the holes, but it is possible
that the specially strong chaetae on the sixth segment may be of some use in
this work. Other lithodomous worms are Sabella saxicava and Dodecaceria
concharum, which is a common little borer, forming galleries in oyster-shells,
etc.
The tubes formed by these Polychaetes are very varied in constitution.[345] In

some cases a mucus, which hardens to form a firm protective envelope, is
secreted from special parts (e.g. the ventral gland shields of Terebellids and
Sabelliformia), or from the greater part of the general surface of the body; in
other cases the secretion serves to stick together particles of mud or sand, or
shelly fragments, so as to form a more or less cylindrical tube (rarely
branched), which is lined internally by the hardened "mucus," having the
appearance of silk.
Fig. 151.—Clymene ebiensis in its tube (t) (from Règne Animal). a, Anterior, p,
posterior end, which is, however, injured.
But the process of tube-making is not a simple one, for in many cases, at
least, the worms exhibit definite powers of choice. Thus some species of
Sabella choose only the very finest particles of mud; Terebella conchilega
chooses fragments of shell and grains of sand; Onuphis conchylega employs
small stones more or less of a size; Sabellaria makes use only of sand grains.
Whilst some worms, like Terebella, Nicomache, and others, make a very
irregular tube, Pectinaria builds a most remarkably neat house, open at each
end, which it carries about with it, the narrow end uppermost (Fig. 152); the
grains of sand are nearly all of the same size and only one layer in thickness,
embedded in abundant "mucus," and with the outer surface quite smooth.
Sir J. Dalyell[346] made some most interesting observations on the method

followed by sundry tube-formers in the building of their tenements, and these
observations, though made nearly half a century ago, have required very little
addition or correction in modern times. In speaking of Sabella, he writes as
follows:—
Fig. 152.—The tube of Pectinaria auricoma. × 3. (From M‘Intosh.) This is its
natural position as carried about by the animal.
Fig. 153.—The upper end of the tube of Terebella conchilega. Slightly enlarged.
(From M‘Intosh.)
"Let a tall and ample crystal jar containing a Sabella be emptied of its
contents and speedily replenished with sea-water; the animal, if in view, has
retreated during the short interval; the orifice of the tube is closed, all is at
rest. But soon after replenishment it rises, to display its branchial plume still
more vigorously than before, and remains stationary, as if enjoying the
freshness of the renovated element, always so grateful—the harbinger of
health and strength to those whose dwelling is there. The passing spectator
would conclude that he now beholds only a beautiful flower, completely
expanded, inclining towards the light like some of those ornaments of nature
decorating our gardens. He pauses in admiration. But if a drop of liquid mud
falls amidst the element from above, disturbing its purity, then, while the
plume unfolds to its utmost capacity, does the animal commence a slow
revolution, the body also passing around within the tube. Now are the
thousands of cilia fringing the ribs [i.e. the secondary filaments] of the
branchiae discovered to be in vigorous activity, and their office to be
wondrous. A loose muddy mass is soon afterwards visibly accumulating in the
bottom of the funnel; meantime the neck or first segment of the body, rising
unusually high above the orifice of the tube, exhibits two trowels beating down
the thin edge as they fold and clasp over the margin, like our fingers pressing
a flattened cake against the palm of the hand. [This refers to the lappets of
the peristomial collar.] During these operations muddy collections are seen
descending between the roots of the fans [right and left gills] towards the
trowels, while another organ, perhaps the mouth, is also occupied, it may be,
in compounding the preparation with adhesive matter. Still does the partial or
complete revolution of the plume above, and of the body within the tube,
continue; the bulk of the muddy mass diminishes, activity abates; it is
succeeded by repose, when the tube is found to have received evident
prolongation."
Fig. 154.—Terebella conchilega Pall. Upper end of the tube (s) showing the
anterior end of the worm. h, Its head; t, tentacles collecting sand grains (y)
in their grooves; x, sand grains in mouth of worm; f, filamentous fringe of
tube. (After Watson.)
The Terebellids use their numerous tentacles in searching for particles of

sand, etc.; each tentacle is grooved along its ventral surface, and the particle
is conveyed along the furrow to the mouth. These particles are actually taken
into the mouth, and mixed with some sort of secretion; on ejection again,
each particle is placed by another tentacle in its position at the edge of the
tube, and by means of its lower lip the Terebellid works it into place.[347]
But whereas the greater number of tubicolous worms make use of

adventitious material wherewith to strengthen the wall of their tube, the
Serpulidae secrete carbonate of lime from their tube-glands, and mould a
tube of this substance. Amongst the Eunicidae the secreted substance is of
itself strong enough to protect the animal; for in Hyalinoecia and species of
Eunice the tube consists of a translucent, tough, parchment-like material.
Chemical analysis has been employed in a few cases to determine the

substance composing the tube. In the case of Hyalinoecia (sometimes
erroneously called Onuphis) the material consists of a phosphoric salt
containing magnesia and a characteristic organic substance "onuphin"[348]; in
Spirographis, a Sabellid, the name "spirographin" is given to its special
secretion, whilst in Serpulids the organic base of the calcareous tube is
"conchiolin."
Fig. 155.—Eunice tibiana Pourt. × ½. The branching tube (t) with the worm (w)
protruding its head through one of several openings. (From Ehlers.)
The majority of worms are solitary, but there are a few instances of social
worms—not that there is any co-operation or distribution of labour amongst
the individuals, but they merely occur together in quantities; thus the sandy
tubes of Sabellaria may form compact masses of several cubic feet, which,
left uncovered by the receding tide, look like rocks upon the shore; as, for
instance, at Paignton and Torquay. Filigrana implexa and Serpula uncinata
similarly intertwine their calcareous tubes to form masses.
Whereas most worms live at the bottom of the sea, at various depths, a few
are to be found at the surface. Purely pelagic habits are confined to a few
families, viz. Tomopteridae, Typhloscolecidae, and the Alciopids and others
amongst the Phyllodocidae; though Nectochaeta, one of the Polynoidae, and
Ophryotrocha, one of the Eunicidae, are modified for this mode of life.[349]
Several genera become pelagic during the breeding season. All these forms
are excellent swimmers, and many of them are transparent.
The Colouring of Polychaetes.—The majority of Polychaetes quickly lose

their colour in spirits, and become uniformly dull or light brown in museums.
There are a few, however, which retain their brilliancy, like Aphrodite and
Chloeia, but in both cases the coloration is due to the beautiful hair-like
bristles ranged along each side of the animal; in the former the colours of the
rainbow flash from specimens which have been kept in spirit for any length of
time. The Polynoids, too, with their golden chaetae and pigmented scales,
retain to some extent their characteristic colouring. But the colours of most
Annelids are due to pigments in the skin, together with the haemoglobin of the
blood, which are soluble, or otherwise changed, in alcohol; for instance, the
bright greenish-blue tint of the common Phyllodoce of our coasts is changed
to a rich chocolate brown; but such cases are rare, most worms becoming
more or less decolorised.
The varied colouring in the Polychaetes, as in other animals, is due to a
variety of causes. The red is in many cases due to haemoglobin of the
vascular system showing through the transparent body; the green of the
tentacles of the Sabellids and Chlorhaemids is similarly due to chlorocruorin.
In other cases the contents of the intestine or the tint of the coelomic fluid
may affect the colour of the worm. In Capitella the coloured excretory
products are regained in the skin; in an Eunicid living in a yellow sponge, on
which it feeds, the colouring matter is extracted and stored in the skin; in the
same kind of way green caterpillars may owe their tint to feeding on green
leaves. But many of the Polychaetes possess distinct pigments in the skin;
thus in Arenicola the dark pigment melanin has been recognised; in Cirratulus
and Nereis certain lipochromes; whilst Eulalia viridis contains a pigment allied
to bonellein. These various pigments yield different absorption bands when a
solution is examined with the spectroscope; others, however, give no bands,
but are distinguished by different chemical reactions.[350] The colour of the
intestine of Chaetopterus has been stated to be due to "modified chlorophyll,"
but it is quite a different substance.
When seen in the living and healthy condition, however, these Polychaete
worms vie with the very butterflies in their brilliant and beautiful colourings,
and though our own worms are not lacking in beauty, many tropical and
southern forms exceed them in gayness of tint. Bright reds, orange, yellows,
greens, blues, rich violets, and sombre browns are all displayed.[351]
The handsome Terebella nebulosa of our own coasts is coloured bright red,
sprinkled with white spots. Nicomache lumbricalis is pink, with red girdles.
Eunicids are frequently red or brown, and the red gills along each side,
together with a brilliant iridescence, render these worms very beautiful.
Nereids present a great range of coloration, from light green to sundry tints of
brown and red in various combinations. Amongst the Serpulids our common
S. vermicularis is a very showy little worm, with its orange body, its red gills
splashed with orange, and its orange operculum streaked with red; and a
Southern form, Placostegus coeruleus, occurring at the Cape of Good Hope,
is provided with beautiful lavender-blue gills. Our own Sabellids present
examples of beautiful markings on the gills, in different colours or in different
shades of the same colour. Amongst Polynoids, P. leucohyba, from the
Antilles, has blue elytra; Hemilepidia erythrotaenia, a long worm from the
Cape of Good Hope, has the anterior end of its body covered with light blue
elytra, whilst the uncovered part is orange, with a broad magenta-red band
along the dorsal surface.
The Phyllodocids are mostly very brightly coloured. The common P.
lamelligera of our coast has a bluish-green body, with olive-green parapodia;
but Lopadorhynchus erythrophyllum, from Jamaica, has a blue body with red
parapodia; whilst Notophyllum myriacyclum has a brown body with
longitudinal dark-brown stripes and yellow parapodia. Both these worms live
in coral reefs, where brilliancy of colour is one of the characteristic features of
the fauna. Other worms are of various shades of green: the dark green
Arenicola with red gills; the bright green Eulalia viridis; the deep green
Amphinome smaragdina, from Jamaica; Gnathosyllis diplodonta, with its
green and yellow body, serve as examples.
Patterns or "markings" may be exemplified by Lepidasthenia elegans (Fig.

156), and Myrianida fasciata, which has a bright red band on each segment
(Fig. 149, p. 280). From this brief list of examples it will be seen that beautiful,
and even brilliant, coloration is not confined to any particular mode of life;
many of the most typically tubicolous forms, like the Terebellids and
Serpulids, are as brilliantly coloured as the most typically free-swimming
genera, like the Phyllodocids. Carnivorous forms like Amphinomids and
Syllids present as wide a range of tint as the limivorous forms like Cirratulus,
Sabella, or Maldanids. Shore-lovers, and deep-sea dwellers, and surface-
swimmers, all exhibit equally bright or equally sombre tints; it is therefore
difficult and rash to dogmatise on the "use" of these colourings to these
animals, or to point to this worm as being protectively, to the other as being
warningly, coloured; for we are too ignorant as to the habits of the worms.
Fig. 156.—Lepidasthenia elegans Gr., × 2, to illustrate colour-markings: the
dark bands in the anterior part of the body occupy two elytriferous, and the
intermediate segments. In the hinder region, where the elytra are in every
third segment, this one is dark. el.12, The twelfth elytron.
Protective and Mimetic Devices.—From the point of view of "protection" in

the evolutionist's sense of the word, we can say but little. Protective
resemblance there is undoubtedly amongst the Polynoids, for the scales of
these forms resemble more or less closely the stones or sand amongst which
they live; in the same species there is great variety in coloration. This
protective habit is carried still further in the case of Psammolyce by the
attachment of sand grains to little cups on the elytra, so that the back of the
animal is concealed. Certain commensals, such as Polynoë arenicolae, P.
pentactes, are coloured so as to resemble their associates. In a few cases it
is possible that the gills of Sabelliformia are protectively coloured; for in
Sabella pavonia they vary from a light yellowish tint to a deep violet-brown,
and the dark markings on them are therefore more or less distinct. Spread out
as the gills are in life, they are in many cases difficult to recognise; it is rather
their movement as they are withdrawn that attracts one's attention to them, as
the tubes of these worms frequently serve for the attachment of brownish
seaweeds, to which the gills bear resemblance. But, as a matter of fact, little
work has been done in this direction, and speculation on the matter without
evidence is worthless. Many pelagic forms, being transparent, such as
Tomopteris and Alciopids, are no doubt protected by their lack of colour; yet
these forms present brightly-coloured spots,—the light-producing organs in
the parapodia of the former, and the large dark eyes of the latter.
Semper[352] mentions a case of possible mimicry in a species of Myxicola

which lives in the clefts of a coral, Cladocora. The branchial funnel, when
expanded, resembles very closely the expanded coral in size, colour, etc.; but
he points out that the species occurs in other situations, where its colouring is
not protective. Probably the "mimicry" is in other instances merely accidental.
No doubt many Polychaetes may be "warningly coloured," but experimental

evidence is incomplete. Polycirrus aurantiacus is bright red, with orange
tentacles; these worms were rejected by certain fish.[353] The animal has
given up living in tubes as all its allies do, and it is the tentacles which appear
to be distasteful to its enemies, for when irritated it coils itself up and wraps
itself round with its tentacles. Moreover, when the tentacles were cut off the
fish did not reject the body of the worm. The tentacles are thus coloured in
such a way that fish recognise them, and associate with the colour some
distasteful property.
Fig. 157.—Chaetopterus variopedatus Ren. × ½. On the left the entire animal,

with the three regions A, B, C. c, Peristomial cirrus; d, "sucker"; e, the
great "wings"; f, "fan"; m, mouth. On the right the animal is represented in
the dark, under stimulation, so as to exhibit the phosphorescent portions of
the body. (From Panceri.)
Phosphorescence.—Many worms of very different habits have the power of

emitting a light from some parts of the body, and they are then said to be
"phosphorescent."[354] Probably Chaetopterus is most eminently photogenic;
the base of the great "wings," the "fans," and other parts emit, on stimulation,
an azure blue to greenish light, so bright that one may read one's watch by it.
Several species of Polynoë exhibit a similar phenomenon, each elytron, with
the exception of the area of attachment, being brilliantly illuminated. In these
species the phosphorescent elytra are frequently thrown off by the animal, so
that possibly they deceive enemies. Polycirrus aurantiacus produces a
beautiful violet phosphorescence; usually its many tentacles alone show the
light, but under strong stimulation the entire body takes part in the display,
and no doubt the phosphorescence has, like the colour, a "warning" purpose.
The production of the light in these various forms is apparently due to two
different processes. In some cases, e.g. Chaetopterus, Syllids, Terebellids, it
appears to be due to the oxidation of certain cell contents which are
discharged more or less freely on irritation of the nerves; whilst in Polynoids
the phenomenon is due to some purely nervous process, for the elytra have
no glands, but are provided with ganglia and a nervous network.
In other worms, however, there are definite light-producing organs. In

Tomopteris there is on each parapodium, above and below, a brightly-
coloured spherical organ, which for a long time was regarded as an eye, but
from its structure appears to be a "photogen" (Fig. 167, p. 315). The same is
very likely the true explanation of the segmental "eyes" of Polyophthalmus, for
their structure recalls that of the light-organs of deep-sea fishes.
As many of the phosphorescent Polynoids are commensals, while

Chaetopterus inhabits tubes, and close allies of other phosphorescent worms
have no power of emitting light, it is impossible to apply the same explanation
of its purpose to all cases alike; in some it may be "accidental," though in
others it may be of definite use in warning enemies or in attracting prey.
The Food of Worms.—The Nereidiformia are mostly carnivorous, and feed

on small Crustacea, Mollusca, sponges, and other animals; and Polynoids are
even said to eat one another. Many worms do not disdain various seaweeds,
whilst the Spioniformia and Scoleciformia, which burrow in mud and sand,
and are without biting organs, swallow the mud and digest what animal or
vegetable débris it may contain. The Terebellids and Cryptocephala depend
on minute organisms which may be driven into the mouth by the action of the
cilia of the gills or tentacles.
In the case of deep-sea forms, it is an interesting fact that the intestines are
not unfrequently crammed with Radiolaria and Foraminifera in a fairly fresh,
uninjured condition, indicating that these Rhizopods do not merely sink to the
bottom, but must actually live there.[355]
The economic purposes to which Polychaetes are put are few; they are
used either as bait for fishes or as food for man.
One of the commonest baits used for certain fish, as all who have done any
sea-fishing off the piers of our coasts know, is the common lug-worm
(Arenicola marina), whilst Nephthys caeca and Nereis fucata are also used in
some places; and for whiting Nereis cultrifera and N. diversicolor. Marphysa
sanguinea, known to the fishermen in some parts as "varme," is less
frequently used.
A peculiar worm—Palolo viridis—is used as food by the natives of Samoa and

Fiji. The worm is similar to our Eunicid Lysidice ninetta, and lives in fissures
among corals on the reefs, at a depth of about two fathoms. At certain days in
October and November they leave the reefs and swim to the shores of the
above islands, probably to spawn; and this occurs on two days in each of the
above months—the day on which the moon is in her last quarter, and the day
before. The natives, who call the worm "Mbalolo," give the name "Mbalolo
lailai" (little) to October, and "Mbalolo levu" (large) to November, thereby
indicating the relative abundance of the worms in these two months. The
natives eat them either alive or baked, tied up in leaves; and they are
esteemed so great a delicacy that presents of them are sent by the chiefs
who live on shore to those living inland. A dark green-blue Phyllodocid, which
is called "A'oon," occurs in abundance off Mota Island, amongst the New
Hebrides, has similar habits, and is also eaten.[356]
Associated Worms.—A considerable number of worms live in association

with other animals, either as commensals or as parasites, and it is not in
every case possible to decide in what relation the two animals stand.
Labrorostratus parasiticus, a Eunicid, is parasitic in the body-cavity of
Odontosyllis ctenostomatus (Fig. 158); such an association between two
members of the same group of animals is peculiar; but still more exceptional
is the occurrence of Haematocleptes terebellides, as a parasite in Marphysa
sanguinea, for both parasite and host are members of the same family, the
Eunicidae. Another Eunicid, Oligognathus bonelliae, occurs in the body-cavity
of the Gephyrean Bonellia.
The Polynoid Acholoe astericola and the Hesionid Ophiodromus flexuosus

occur as ectoparasites (or perhaps commensals) in the ambulacral grooves of
the starfish Astropecten aurantiacus. An Amphinomid is stated to live in the
branchial chamber of the barnacle, Lepas anatifera. Alciopina parasitica lives,
during the early stages of its life-history, within Cydippe, and it is possible that
most of the Alciopids thus make use of Ctenophores as their nurseries.
A considerable number of the Polynoids are ectoparasitic: P. castanea lodges

in the peri-oral region of Spatangus purpureus, and in the ambulacral grooves
of Astropecten; P. (Halosydna) bairdi lives between the mantle and foot of the
mollusc Fissurella cratitia; P. pentactes is found on the body of the
Holothurian Cucumaria pentactes, and appears to be protectively coloured. P.
(Antinoë) parasitica lives under the elytra of another Polynoid, and P.
acanellae on the coral Acanella normani.[357]
Fig. 158.—Odontosyllis ctenostomatus, with (L) Labrorostratus parasiticus in its

body-cavity. The parapodia and cirri are omitted from the greater part of
the body. (After St. Joseph.) × 4.
As commensals there may be mentioned Nereis fucata, which lives in the

upper coil of whelk-shells which are inhabited by a hermit crab. The same
shell usually bears a particular sea-anemone, so that there are three animals
living together in or upon the cast-off house of a fourth. Siphonostoma is
found in the "nests" made by the mollusc Lima. A Eunice is constantly
associated with the coral Lophohelia prolifera, amongst the branches of which
the worm twines its tube; whilst another Polychaete inhabits a tube formed by
the interweaving of the fine branches of the coral Antipathes filix,[358] found in
the West Indian seas. A species of Polydora forms its tube in Heliopora. The
Polynoids present many instances of commensalism, a few of which may be
here mentioned. P. johnstoni Marenz. is only found in the tubes of Terebella
nebulosa; other species occur in the tubes of other Terebellids. P. marphysae
lives in tubes of the Eunicid Marphysa sanguinea. Two species live in the
tubes of Chaetopterus. P. extenuata has been found in tubes of Serpula
vermicularis, while P. arenicolae occurs on the body of the common lug-worm,
with the colouring of which it closely harmonises.
Worms as Hosts.—The Polychaeta serve not only as food for fishes,

Crustacea, and other predatory animals of larger size, but are also liable to be
the hosts of parasites[359] such as Gregarines, and even, as we have seen, of
other members of their own group. Sundry ectoparasitic Copepoda have been
found attached to worms between the parapodia or to the sides of the feet,
and an unnamed Copepod occurs attached, sometimes in considerable
numbers, to the sides of Nereis cultrifera. The Polychaeta also act as
protectors to other animals, for on the under surface of elytra of sundry
Polynoids may very frequently be found specimens of Loxosoma, which may
also be attached to gills of Eunicids; whilst below those of Aphrodite echidna
and Hermadion pellucidum, Pedicellina belgica occurs. Under the felt of A.
aculeata the Sabellid Branchiomma vigilans forms its tube, and Vorticellids
may be found on chaetae, gills, or other parts of the body of sundry worms.
Distribution.—Very little can be said in a brief way of the geographical

distribution of these worms, for many of the genera are cosmopolitan,
although only a few species occur in all the great oceans, e.g. Polynoë
imbricata, Hyalinoecia tubicola, Nerine (Scolecolepis) cirrata, and Terebellides
stroemi.
As for species, it can be said generally that the different oceanic areas and
even different coasts present different species, but we know practically
nothing of variation amongst Polychaeta, and many so called species may be
mere local varieties, for frequently the descriptions of "new species" are
scarcely intelligible. At any rate we know that certain species occur at widely
separated localities, for two or three species of Polynoids occur in Japan, and
again at Dinard on the French coast. A considerable number of species are
common to both sides of the North Atlantic ocean, having been obtained off
Norway and in the Gulf of the St. Lawrence. A few of these which are
common on our coasts may be enumerated:—Nereis pelagica, Nicomache
lumbricalis, Glycera capitata, Thelepus cincinnatus, Scoloplos armiger,
Sabella pavonia, Ophelia limacina, Aphrodite aculeata, Trophonia plumosa,
Polynoë squamata, Capitella capitata, Sthenelais limicola.
As for bathymetrical distribution,[360] many genera occur at all depths, though

Polychaetes appear to be most abundant, as far as we know at present, in
"shallow water"—that is, down to twenty fathoms or so; but this may be due to
the greater facility of collection on shore and in these slight depths, for the
"Challenger" obtained considerable numbers of new species at greater
depths.
The "deep-sea" forms are chiefly tubicolous, and since these tubes are fixed
and partially embedded in the bottom, probably comparatively few are
brought up. Some genera occur at very great depths; thus the Terebellid
Leaena abyssorum and the Serpulid Placostegus benthalianus were brought
up from 3125 fathoms—the greatest depth from which Polychaetes were
obtained by H.M.S. "Challenger"; and it is interesting to note that species of
each of these two genera occur in shallow water, the Serpulid being
represented in our own coast fauna by P. tricuspidata.
Amongst our own fauna, a few examples may be given of the "replacement of
species."[361] The littoral Sthenelais boa is represented by S. limicola in
deeper water; Sabellaria alveolata by S. spinulosa; Polynoë imbricata by
several deep-water species. Similarly with genera: the littoral Pomatoceros is
replaced by Serpula in deeper water; and the Hesionid Psamathe by Castalia.
The limitation of species to certain regions, or to certain depths of an ocean,

may appear at first sight peculiar, in view of the unrestricted communication
between all its parts; but there are as efficient "barriers" there as on land, for
generally a particular worm can live only in a certain temperature and at a
certain pressure, and is dependent for its food on particular organisms, which
in their turn depend on the depth and its accompaniments. It is, in fact, so
much the more peculiar that certain species are more or less cosmopolitan, or
occur at widely distant points. It is less peculiar, of course, to find different
species of the same genus at different depths or in different areas, for any
slight variation in a species advantageous to new conditions would readily be
fixed, and give rise to a new species.
The distribution of the Polychaeta depends probably on the pelagic larvae,

which are carried by currents from one part of an ocean to another. There can
be little doubt that many Polychaetes are very "plastic," and can adapt
themselves to changed conditions of life with considerable ease; for Nereis
diversicolor, Arenicola marina, and others live equally well in water of very
different densities, and with a different food supply. The great variety in the
"habitats," and presumably therefore in their food supply, etc., exhibited by
many Polychaetes, as well as the great variation observable in some species
of Polynoina, and the close affinity of the species and genera of this sub-
family, lead us to the same conclusion.
Extinct Polychaetes.—The most numerous fossil records of the Polychaetes

are calcareous tubes of various shapes and sizes; they are irregularly or
spirally curved, and are very usually attached at one end, or by one surface,
to stones or to fossils. These tubes belong to the Serpulidae, and are referred
to the genera Serpula, Spirorbis, Ditrupa, and others.[362]
Spirorbis is the oldest unequivocal representative of the Polychaetes, as its

tubes are found more or less abundantly in the Silurian and other Palaeozoic
strata. In Palaeozoic times Serpula was rare, as it was too in the Trias and
Lias, but in the Jurassic strata it becomes abundant. In the chalk, S. socialis
may occur in masses like S. uncinata of the present day, forming "Serpulite
chalk." In the older tertiaries the genus is represented by Spirulaea.
Terebella lapilloides occurs in the Lias as a cylindrical, more or less curved

tube of sand-grains.
Fig. 159.—Eunicites avitus Ehl. A fossil worm from the lithographic slate of
Solenhofen: the jaws are seen in front, and the acicula along each side.
(From Ehlers.) Natural size.
Amongst the Nereidiformia the remains are fewer, but the acicula and the
hard jaws are preserved in certain rocks, and can be referred to existing
families. Eunicites avitus[363] is represented by a double series of acicula,
indicating the parapodia of the two sides; and by remains of both upper and
lower jaws (Fig. 159). Four different species of the worm have been described
from the lithographic slate of Bavaria, of Jurassic age; and several upper jaws
of other Eunicids have been discovered in the Palaeozoic beds of Canada
and Scotland, and have received the names Lumbriconereites, Oenonites,
and Arabellites, in reference to their nearest allies amongst living genera.
There are, however, numerous remains, in the forms of tracks or casts, in the
earlier rocks, which have been referred to the Polychaeta. The names
Crossopodia, Myrianites, Nereites, Phyllodocites, have been given to some of
these traces, though they are open to numerous other interpretations. Some
of the "tracks" are similar to those made by living Crustacea in walking over
wet sand; others appear to be the casts of some animals. Tubular burrows in
rocks or fossils, some straight, others U-shaped, have received such names
as Arenicolites, Scolithus, Histioderma; whilst under the name Lumbricaria
certain cylindrical, coiled structures, resembling worm "castings," are met with
in this same lithographic stone of Solenhofen. Many of the tubes referred to
Polychaetes by the earlier palaeontologists have been transferred to other
groups; thus Cornulites is now believed to be a Pteropod shell.
This very meagre geological record is quite insufficient to form any basis for a
phylogeny of the group. And this poor supply of remains is not surprising,
when we consider the soft nature of the tissues, the absence, in the majority
of families, of skeleton and of other parts which could have been fossilised;
yet we might have expected a greater abundance of fossilised jaws than is
represented at present. But it must be borne in mind that the conditions of life
of these soft-bodied animals are not conducive to their leaving abundant
fossilised remains.
CHAPTER XII
CHARACTERS OF THE SUB-ORDERS OF POLYCHAETES—CHARACTERS OF THE

FAMILIES—DESCRIPTION OF BRITISH GENERA AND SPECIES—THE MYZOSTOMARIA.
Systematic.—The Order Polychaeta may be divided into two branches, in

one of which, the Phanerocephala, the prostomium retains its ancestral
condition as a lobe overhanging the mouth, and frequently carries, in addition
to paired eyes, certain sensory processes of a simple structure, the tentacles

Textbook Interpreting Epidemiologic Evidence Connecting Research To Applications Second Edition Savitz Ebook All Chapter PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Textbook Interpreting Epidemiologic Evidence Connecting Research To Applications Second Edition Savitz Ebook All Chapter PDF

Uploaded by

Copyright:

Available Formats

Interpreting epidemiologic evidence:

connecting research to applications

Interpreting Complex Forensic DNA Evidence 1st Edition

Foundations of Clinical Research Applications to

Interpreting and Using Statistics in Psychological

Intermediate Algebra: Connecting Concepts through

Applications of Social Research Methods to Questions in

Cognitive Psychology: Connecting Mind, Research, and

Second Language Pragmatics From Theory to Research 1st

Applied Epidemiologic Principles and Concepts:

Published in the United States of America by Oxford University Press

© Oxford University Press 2016

First Edition published in 2003

All rights reserved. No part of this publication may be reproduced, stored in

You must not circulate this work in any other form

Library of Congress Cataloging-​in-​Publication Data

2. The Nature of Epidemiologic Evidence 7

3. Causal Diagrams for Epidemiologic Inference 21

4. Strategy for Drawing Inferences from Epidemiologic Evidence 35

5. Confounding I: Theoretical Considerations 45

6. Confounding II: Practical Considerations 63

7. Selection Bias and Confounding Resulting from Selection in Cohort Studies 77

8. Selection Bias in Case-​Control Studies 93

9. Bias Due to Loss of Study Participants 113

Stratify Study Base by Markers of Participation 122

10. Measurement and Classification of Exposure 127

11. Measurement and Classification of Disease 149

Restrict Inference to Disease Outcome That Can Be Ascertained Accurately 165

12. Random Error 171

13. Integration of Evidence Across Studies 185

14. Characterization and Communication of Conclusions 199

2 Interpreting Epidemiologic Evidence

solidity or fragility of the available scientific evidence. By pinpointing why or where

full command of these challenging quantitative methods, their application to the

APPROACH TO THE EVALUATION OF EVIDENCE

4 Interpreting Epidemiologic Evidence

Hernan MA, Robins JM. Causal Inference. http://​w ww.hsph.harvard.edu/​miguel-​hernan/​

GOALS OF EPIDEMIOLOGIC RESEARCH

8 Interpreting Epidemiologic Evidence

Table 2.1 Criteria for Assessing Effectiveness of Epidemiologic Research

evaluate the quality or strength of epidemiologic evidence, we need to be clear on

9 The Nature of Epidemiologic Evidence

10 Interpreting Epidemiologic Evidence

11 The Nature of Epidemiologic Evidence

associations (ignoring causality) or generating hypotheses for other scientists to

MEASUREMENT OF CAUSAL RELATIONS

12 Interpreting Epidemiologic Evidence

13 The Nature of Epidemiologic Evidence

14 Interpreting Epidemiologic Evidence

APPLICATIONS OF EPIDEMIOLOGIC RESEARCH

Another modification of the process of budding and fission is exhibited by Syllis

Regeneration of lost Parts.—The process of budding and fission of the

NATURAL HISTORY OF POLYCHAETES—GENERAL HABITS—CHARACTER OF TUBE

In brackish water of various densities many Polychaetes live; Arenicola

Owing to their sedentary habits, quite a representative collection of genera

It may be useful to give instances of worms occurring in various situations

The tubes formed by these Polychaetes are very varied in constitution.[345] In

Sir J. Dalyell[346] made some most interesting observations on the method

The Terebellids use their numerous tentacles in searching for particles of

But whereas the greater number of tubicolous worms make use of

Chemical analysis has been employed in a few cases to determine the

The Colouring of Polychaetes.—The majority of Polychaetes quickly lose

Patterns or "markings" may be exemplified by Lepidasthenia elegans (Fig.

Library of Congress Cataloging-in-Publication Data

8. Selection Bias in Case-Control Studies 93

Hernan MA, Robins JM. Causal Inference. http://w ww.hsph.harvard.edu/miguel-hernan/