See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/51673941

Dietary patterns and colorectal cancer: Systematic review and meta-analysis

Article  in  European journal of cancer prevention: the official journal of the European Cancer Prevention Organisation (ECP) · September 2011
DOI: 10.1097/CEJ.0b013e3283472241 · Source: PubMed

CITATIONS READS

184 1,040

3 authors:

Bruno Magalhães Bárbara Peleteiro

Instituto Português de Oncologia Institute of Public Health, University of Porto
29 PUBLICATIONS   230 CITATIONS    100 PUBLICATIONS   2,871 CITATIONS   

SEE PROFILE SEE PROFILE

Nuno Lunet
University of Porto
411 PUBLICATIONS   13,728 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Defining predictive models of the variation in esophageal cancer incidence View project

Famílias em época de Pandemia View project

All content following this page was uploaded by Bruno Magalhães on 21 July 2020.

The user has requested enhancement of the downloaded file.


Dietary patterns and colorectal cancer: systematic review
and meta-analysis
Bruno Magalhãesa,b, Bárbara Peleteirob,c and Nuno Lunetb,c
Studies on the association between single foods or
nutrients and colorectal cancer have provided inconsistent
results. Previous reviews did not conduct a quantitative
synthesis of the relation with dietary patterns. We
conducted a systematic review and meta-analysis of
studies addressing the association between dietary
patterns and colorectal cancer. Studies quantifying the
association between dietary patterns (defined a posteriori)
and colorectal cancer were identified in PubMed (until
01.08.2010) and through backward and forward citation
tracking (ISI Web of Science and Scopus). Summary
relative risk (RR) estimates and 95% confidence intervals
c 2011 Wolters Kluwer Health | Lippincott Williams &
(95% CI) were computed for highest versus lowest levels
of exposure, for colon cancer (CC) and rectal cancer (RC),
and for proximal and distal CC, by random effects
meta-analysis. Heterogeneity was quantified using the
I2 statistic. Eight cohort and eight case–control studies
defining patterns through principal components and
factor analyses were included in the systematic review.
Meta-analyses were conducted for three patterns:
(i) ‘drinker,’ characterized by high alcohol consumption
(CC: RRcombined = 0.96, 95% CI: 0.82–1.12, I2 = 0.6%; RC:
RRcombined = 0.83, 95% CI: 0.47–1.45, I2 = 65.1%);
(ii) ‘healthy,’ characterized by high fruit/vegetables
Introduction
Cancers of the colon and rectum are responsible for a large
proportion of the global burden of cancer morbidity and
mortality, particularly in the more developed countries
where these malignancies rank third in incidence and
second in mortality (Ferlay et al., 2008). Their frequency
varies widely across countries, by as much as 10-fold,
between low-incidence areas in Asia and Africa and northern
Europe or the United States (Peto, 2001), and among
populations moving from low-risk to high-risk areas the
rates rapidly approach those observed in the host regions
(Nilsson et al., 1993; McCredie, 1998; Monroe et al., 2003).
Lifestyles, namely dietary exposures (Shike, 1999; Slattery,
2000; Chan and Giovannucci, 2010) are likely to play an
important role in colorectal cancer etiology, but studies
examining the analyses of the independent effects of
specific nutrients or foods have conceptual and methodo-
logical limitations and have yielded inconsistent or conflict-
ing findings (World Cancer Research Fund, 1997; World
Cancer Research Fund and American Institute for Cancer
All supplemental digital content is available directly from the corresponding
author
0959-8278
c 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Review article 1
consumption (CC: RRcombined = 0.80, 95% CI: 0.70–0.90,
I2 = 55.1%; RC: RRcombined = 1.02, 95% CI: 0.89–1.17,
I2 = 10.8%); (iii) ‘western,’ characterized by high red/
processed meat consumption (CC: RRcombined = 1.29,
95% CI: 1.13–1.48, I2 = 31.7%; RC: RRcombined = 1.13, 95%
CI: 0.92–1.39, I2 = 40.6%). Summary estimates for proximal
and distal CC were similar. The risk of CC was increased
with patterns characterized by high intake of red and
processed meat and decreased with those labelled as
‘healthy.’ No significant associations were observed
for RC. European Journal of Cancer Prevention 00:000–000
Wilkins.
European Journal of Cancer Prevention 2011, 00:000–000
Keywords: colon cancer, colorectal cancer, dietary patterns, eating patterns,
foods, rectal cancer
a
Department of Oncologic Surgery, Portuguese Oncology Institute – Porto
(IPO-Porto), bDepartment of Hygiene and Epidemiology, University of Porto Medical
School and cInstitute of Public Health – University of Porto (ISPUP), Portugal
Correspondence to Nuno Lunet, MPH, PhD, Serviço de Higiene e Epidemiologia,
Faculdade de Medicina do Porto, 4200–319 Porto, Portugal
Tel: +351 225513652; fax: +351 225513653;
e-mail: nlunet@med.up.pt
Received 14 March 2011 Accepted 15 March 2011
Research, 2007). As people are exposed to a variety of foods
with complex combinations of nutrients, dietary patterns
are more suited to capture the overall effects of dietary
exposures than isolated nutrient or food item analyses.
Furthermore, the effects of single nutrients may be too
small to detect, and the analyses assessing the effect of a
large number of nutrients or food items may produce
statistically significant associations simply by chance.
Two systematic reviews on dietary patterns and colorectal
cancer were recently published (Miller et al., 2010a;
Randi et al., 2010), but no quantitative estimates were
computed. Therefore, we conducted a systematic review
and meta-analysis of studies addressing the association
between dietary patterns (defined a priori) and colorectal
cancer, to achieve a quantitative synthesis of the available
evidence on the association between dietary patterns and
the risk of colorectal cancer, by subsite.
Methods
Published cohort and case–control studies presenting results
on the association between dietary patterns and colorectal
cancer were identified through PubMed (http://www.ncbi.
nlm.nih.gov/entrez/) search, from inception to 1 August
DOI: 10.1097/CEJ.0b013e3283472241
 2 European Journal of Cancer Prevention 2011, Vol 00 No 00

Fig. 1

155 potentially relevant articles were identified

through Pubmed search
121 articles were excluded after evaluation of the
title/abstract because they referred to animal studies, were
review articles or did not convey information on the
association between colorectal cancer and
environmental factors/diet

The full text of 34 articles was retrieved for more

detailed assessment
12 articles were excluded after evaluation of the full text:
4 assessed the effect of interventions
4 had adenomas as the outcome
1 addressed the recurrence of colorectal cancer
2 defined dietary patterns using an a priori method
1 combined data on diet patterns with family history and
genetic polymorphisms

22 articles included in the systematic review

420 new potentially relevant articles were identified

2 reports referred to
through forward and backward citation tracking (ISI
systematic reviews and
web of knowledge and Scopus) of the articles
were used only for
included in the systematic review
identification of original
studies by search of the
reference lists
413 articles were excluded after evaluation of
1 report compared three the title/abstract because they reported animal
methods to define dietary studies, were review articles or did not
patterns using data convey information on the association
presented in two studies between colorectal cancer and environmental
already included in the factors/diet
review

7 articles potentially eligible for the systematic review

were identified and full texts retrieved for more detailed

7 articles were excluded after the

evaluation of the full text because none
addressed the relation between dietary
patterns and colorectal cancer.

19 articles included in the systematic review

(9 cohort and 10 case – control studies)

3 studies were excluded from meta-analysis because:

1 defined dietary patterns using data on nutrient’ intake
2 relied on cluster analysis to define dietary patterns

16 articles included in the meta-analysis

(8 cohort and 8 case – control studies)

Provided estimates Provided estimates Provided estimates Provided estimates Provided estimates
on the risk of on the risk of colon on the risk of on the risk of on the risk of
colorectal cancer cancer proximal colon proximal colon rectum cancer
cancer cancer
9 articles 11 articles 6 articles 6 articles 8 articles

Systematic review flow-chart.

2010, under the following expression: (((colon OR rectum
OR colorectal OR rectal OR colonic) AND cancer) OR
(colorectal cancer OR colon cancer OR rectum cancer OR
rectal cancer)) AND (‘dietary pattern’ OR ‘dietary patterns’
OR ‘eating pattern’ OR ‘eating patterns’ OR ‘food pattern’
OR ‘food patterns’). Backward and forward citation tracking,
both in ISI Web of Science (http://apps.isiknowledge.com)
and Scopus (http://www.scopus.com), were used to identify
articles cited by those selected in previous steps of the
search strategy, or citing them, respectively. The systematic
review flow-chart is presented in Fig. 1. All the identified
references referred to publications in English and therefore
no further language restrictions were applied.
In the studies identified, two approaches were used for
developing a general descriptor for the dietary pattern.
The first, a priori, is based on previous knowledge of the
favorable or unfavorable effects of various constituents of

Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

the diet. The alternative approach is data driven or a
posteriori, based on dietary data obtained directly from the
studied population (Michels and Schulze, 2005). Only
the studies evaluating the association between dietary
patterns defined a posteriori and colorectal cancer were
considered eligible for the present systematic review.
Data extraction was independently conducted by two
researchers (B. M. and B. P.), according to a predefined
protocol that covered the following items: publication
year; country/region of the studied subjects; study design
(cohort, case-cohort, nested case–control, population-
based case–control, hospital-based case–control); number
of participants and their characteristics (selection proce-
dure, age and sex); exposure assessment (method used
for the evaluation of dietary intake); strategy used to
define dietary patterns [data driven or a posteriori, in
which statistical methods such as factor analysis and
cluster analysis are used to generate dietary patterns
empirically from the study population (Michels and
Schulze, 2005)]; measures of the association between
dietary patterns and colorectal cancer (odds ratio, relative
risk or hazard ratio), and respective precision estimates;
control for potential confounding factors. Discrepancies
in data extraction were resolved by consensus or by
involving a third researcher (N. L.).
In addition, relative risk estimates referring to colorectal
cancer, site-specific data for cancers of the colon (all
locations, proximal and distal) and rectum, were extracted
whenever available. When a study provided several relative
risk estimates with different degrees of adjustment for
confounding, we opted for the one adjusting for the largest
number of factors. When studies presented results includ-
ing and excluding events in the initial years of follow-up we
opted for the data with all participants, because they were
most frequently presented in the original papers.
To summarize the results, we only considered the
patterns observed more frequently in the different
studies. As the labeling of the patterns is somewhat
arbitrary and the dietary patterns are locale-specific, we
considered that those similar regarding factor loadings for
the foods usually associated with the risk of colorectal
cancer (World Cancer Research Fund and American
Institute for Cancer Research, 2007) corresponded to
equivalent patterns for the purpose of summarizing the
evidence on this topic, as described in Table 1. Studies
that identified dietary patterns labeled as ‘western,’ ‘pork
and processed meat,’ ‘pork and processed meat, and
potatoes,’ ‘meat, potatoes and refined grains’ and ‘high
fat, meat and potatoes’ tended to have high loadings for
red and/or processed meats, refined grains, potatoes,
sweets, and high-fat dairy, and therefore were grouped
under the label ‘western.’ ‘Prudent,’ ‘healthy,’ ‘vegeta-
bles,’ and ‘fruit and vegetables’ dietary patterns tended to
have high loadings of foods such as fruit, vegetables,
poultry, fish, low-fat dairy, and whole grains, and these
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Dietary patterns and colorectal cancer Magalhães et al. 3
patterns were analyzed together and labeled as ‘healthy.’
‘Drinker’ dietary patterns tended to have high loadings of
alcoholic beverages.
For meta-analysis, cumulative incidence ratios or inci-
dence density ratios, hazard ratios, and odds ratios were
treated the same and are referred to as RR estimates.
Estimates from publications providing sex-specific and/or
race-specific results were used in the final analysis as
obtained from different studies. We computed combined
RR estimates (RRcombined) for the highest versus lowest
category of exposure to the factor that characterizes each
pattern, and 95% confidence intervals (95% CI) using a
random effects model (Jackson et al., 2010), for groups of
more than two studies considered methodologically
homogeneous by the authors, referring to each of the
following outcomes: rectum, colon, proximal colon, distal
colon, and colorectal cancers. The same study could
provide risk estimates for more than one of the outcomes
of interest.
Statistical tests for homogeneity (Petiti, 2000) were
carried out, and heterogeneity was quantified using the I2
statistic (Higgins and Thompson, 2002). Publication bias
was examined through visual inspection of the funnel
plots, the Begg adjusted rank correlation test (Begg and
Mazumdar, 1994) and the Egger regression asymmetry
test (Egger et al., 1997). A 0.1 level of significance was
used in the statistical tests to increase the sensitivity in
detecting asymmetry. All statistical analyses were con-
ducted with STATA, version 9.2.
Results
The systematic review flow-chart is presented in Fig. 1.
The literature search yielded 19 eligible articles,
published from 1975 to 2010, described in detail in
Annexes 1 and 2. Ten were conducted in North America
(Randall et al., 1992; Slattery et al., 1998; Fung et al.,
2003; Wu et al., 2004; Nkondjock and Ghadirian
2005; Flood et al., 2008; Satia et al., 2009; Williams et al.,
2009; Wirfalt et al., 2009; Miller et al., 2010b), five in
Europe (Terry et al., 2001; Dixon et al., 2004; Rouillier
et al., 2005; Kesse et al., 2006; Bravi et al., 2010), three in
Asia (Kim et al., 2005; Butler et al., 2008; Kurotani et al.,
2010), and one in South America (De Stefani et al., 2009).
Most of these studies used validated food frequency
questionnaires to assess dietary intake but in one study,
this information was not specified (Terry et al., 2001) and
in another, the questionnaire was not validated (De
Stefani et al., 2009), although reproducibility had been
evaluated. The number of items in the food frequency
questionnaire ranged between 64 (De Stefani et al., 2009)
and more than 700 (Slattery et al., 1998). In general, the
assessment of food consumption referred to the year
preceding diagnosis for cases and interview for controls,
except five studies that evaluated dietary intake in
the previous month (Kim et al., 2005), in the previous
 4 European Journal of Cancer Prevention 2011, Vol 00 No 00

Table 1 Dietary patterns identified in the original studies and considered for meta-analysis
Dietary patterns as labeled in Dietary patterns considered for meta-analysis
Reference and country the original articles and respective labels

Kurotani et al., 2010 Japan ‘Prudent’ ‘Healthy’

‘High fat’ Not used in meta-analysis
‘Light-meal’ Not used in meta-analysis
Miller et al., 2010b USA ‘Fruits and vegetables’ (M, F) ‘Healthy’
‘Meat, potatoes, and refined grain’ (M, F) ‘Western’
‘Alcohol and sweetened beverages’ (M) ‘Drinker’
Williams et al., 2009 USA ‘High fat, meat, and potatoes’ (B, W) ‘Western’
‘Vegetable, fish, and poultry’ (W) ‘Healthy’
‘Fruit and vegetables’ (B) ‘Healthy’
‘Fruit, whole grain, and dairy’ (W) Not used in meta-analysis
‘Fruit and dairy’ (B) Not used in meta-analysis
De Stefani et al., 2009 Uruguay ‘Prudent’ ‘Healthy’
‘Traditional’ Not used in meta-analysis
‘Western’ ‘Western’
‘Drinker’ ‘Drinker’
Satia et al., 2009 USA ‘Western-Southern’ ‘Western’
‘Fruit-vegetable’ ‘Healthy’
‘Metropolitan’ Not used in meta-analysis
Butler et al., 2008 China ‘Vegetable-fruit-soy’ Not used in meta-analysis
‘Meat-dim sum’ Not used in meta-analysis
Flood et al., 2008 USA ‘Fruit and vegetables’ ‘Healthy’
‘Fat reduced and diet foods’ Not used in meta-analysis
‘Red meat and potatoes’ ‘Western’
Kesse et al., 2006 France ‘Healthy’ ‘Healthy’
‘Western’ ‘Western’
‘Drinker ‘Drinker’
‘Meat eaters’ ‘Western’
Kim et al., 2005 Japan ‘Healthy’ ‘Healthy’
‘Traditional’ Not used in meta-analysis
‘Western’ ‘Western’
Nkondjock and Ghadirian 2005 Canada ‘Chocolate–cereal’ Not used in meta-analysis
‘Pork and processed meat’ ‘Western’
‘Drinker’ ‘Drinker’
Wu et al., 2004 USA ‘Prudent’ ‘Healthy’
‘Western’ ‘Western’
Dixon et al., 2004 Finland, Netherlands, Sweden ‘Vegetables’ ‘Healthy’
‘Pork, processed meat, potatoes’ ‘Western’
Fung et al., 2003 USA ‘Prudent’ ‘Healthy’
‘Western’ ‘Western’
Terry et al., 2001 Sweden ‘Healthy’ ‘Healthy’
‘Western’ ‘Western’
‘Drinker’ ‘Drinker’
Slattery et al., 1998 USA ‘Western’ ‘Western’
‘Prudent’ ‘Healthy’
‘Drinker’ ‘Drinker’
‘Substituters’ Not used in meta-analysis
Randall et al., 1992 USA ‘Salad’ Not used in meta-analysis
‘Fruit’ Not used in meta-analysis
‘Healthful’ ‘Healthy’
‘High fat’ Not used in meta-analysis
‘Whole grain’ Not used in meta-analysis
‘Traditional’ Not used in meta-analysis
‘Low cost’ Not used in meta-analysis
‘Snacks’ (M) Not used in meta-analysis
‘Light’ (F) Not used in meta-analysis

B, blacks; F, females; M, males; W, whites.

6 months (Terry et al., 2001), or in the previous 2 years
(Slattery et al., 1998; Nkondjock and Ghadirian 2005; Bravi
et al., 2010).
From the 19 studies included in the systematic review,
three were not considered for meta-analysis because the
strategies of data analysis yielded results not directly
comparable with those from the remaining studies that
used principal components/factor analysis to define
patterns of intake of different food items. Wirfalt et al.
(2009) relied on cluster analysis and Rouillier et al. (2005)
on principal component analysis to select the most appro-
priate food groups, and then on a hierarchical agglom-
erative clustering method, to define the dietary patterns.
Bravi et al. (2010) defined patterns of nutrient intake.
Thus, 16 studies [eight cohort (Terry et al., 2001; Fung
et al., 2003; Dixon et al., 2004; Wu et al., 2004; Kim et al.,
2005; Kesse et al., 2006; Butler et al., 2008; Flood et al.,
2008), seven population-based case–control (Randall
et al., 1992; Slattery et al., 1998; Nkondjock and Ghadirian
2005; Satia et al., 2009; Williams et al., 2009; Kurotani
et al., 2010; Miller et al., 2010b), and one hospital-based

Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

case–control (De Stefani et al., 2009)] were considered for
meta-analysis.
Regarding the outcome evaluated, RR estimates for
colorectal cancer were obtained from nine articles (Terry
et al., 2001; Dixon et al., 2004; Kim et al., 2005; Kesse et al.,
2006; Butler et al., 2008; Flood et al., 2008; De Stefani
et al., 2009; Kurotani et al., 2010; Miller et al., 2010b), 11
articles provided estimates for colon cancer (Randall
et al., 1992; Slattery et al., 1998; Terry et al., 2001; Fung
et al., 2003; Dixon et al., 2004; Wu et al., 2004; Kim et al.,
2005; Nkondjock and Ghadirian 2005; Butler et al.,
2008; De Stefani et al., 2009; Satia et al., 2009), six for
both proximal and distal colon cancers (Slattery et al.,
1998; Terry et al., 2001; Wu et al., 2004; Kim et al.,
2005; Flood et al., 2008; Kurotani et al., 2010), and eight
for rectal cancer (Terry et al., 2001; Dixon et al., 2004;
Kim et al., 2005; Butler et al., 2008; Flood et al., 2008; De
Stefani et al., 2009; Williams et al., 2009; Kurotani et al.,
2010) (Fig. 1).
Although dietary patterns observed across populations
from different geographic locations and cultural back-
grounds are locale-specific, it was possible to identify
three dietary patterns with similar characteristics that
were common to the majority of the studies identified,
which were labeled as ‘drinker,’ 20 RR estimates (by
cancer subsite and sex strata) available from six studies
(Slattery et al., 1998; Terry et al., 2001; Nkondjock and
Ghadirian 2005; Kesse et al., 2006; De Stefani et al.,
2009; Miller et al., 2010b); ‘healthy,’ 63 estimates in 14
studies (Randall et al., 1992; Slattery et al., 1998; Terry
et al., 2001; Fung et al., 2003; Dixon et al., 2004; Wu et al.,
2004; Kim et al., 2005; Kesse et al., 2006; Flood et al.,
2008; De Stefani et al., 2009; Satia et al., 2009; Williams
et al., 2009; Kurotani et al., 2010; Miller et al., 2010b);
‘western,’ 60 estimates in 13 studies (Slattery et al.,
1998; Terry et al., 2001; Fung et al., 2003; Dixon et al.,
2004; Wu et al., 2004; Kim et al., 2005; Nkondjock and
Ghadirian, 2005; Kesse et al., 2006; Flood et al., 2008; De
Stefani et al., 2009; Satia et al., 2009; Williams et al.,
2009; Miller et al., 2010b).
The risk of colorectal cancer was lower among subjects
with a ‘healthy’ dietary pattern (RRcombined = 0.83, 95%
CI: 0.73–0.94, I2 = 53.1%) and higher among those with a
‘western’ pattern (RRcombined = 1.19, 95% CI: 1.04–1.37,
I2 = 52.6%) but no significant associations were observed
for the ‘drinker’ pattern (RRcombined = 0.98, 95% CI:
0.77–1.26, I2 = 40.8%).
The risk of colon cancer was approximately 30% higher
among subjects with a ‘western’ pattern and 20% lower
for the ‘healthy’ pattern (Fig. 2). No meaningful
differences were observed between the combined esti-
mates for proximal and distal colon cancers. The ‘healthy’
pattern was significantly associated with a lower risk
of both proximal (RRcombined = 0.80, 95% CI: 0.66–0.96,
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Dietary patterns and colorectal cancer Magalhães et al. 5
I2 = 41.0%) and distal (RRcombined = 0.82, 95% CI:
0.71–0.95, I2 = 15.2%) colon cancers. For the ‘western’
pattern, the summary estimates were 1.11 (95% CI:
0.93–1.32, I2 = 22.6%) and 1.32 (95% CI: 0.99–1.77,
I2 = 65.3%) for proximal and distal cancer, respectively.
Virtually no association was observed between the
‘drinker’ pattern and proximal (RRcombined = 0.94, 95%
CI: 0.75–1.19, I2 = 0%) or distal (RRcombined = 1.08,
95% CI: 0.83–1.41, I2 = 21.2%) cancers.
For rectal cancer, no significant associations were observed
for any of the dietary patterns evaluated (Fig. 3).
Funnel plots revealed little evidence of asymmetry, and
the corresponding statistical tests did not show publica-
tion bias, except for the drinker pattern, regarding both
colon and rectum cancers, but the number of studies is
too small for a robust interpretation of these results
(Fig. 4).
Discussion
The risk of colon cancer is higher among subjects with
dietary patterns characterized by high consumption of red
meat and lower when high consumption of fruits and
vegetables defines the predominant pattern. No signifi-
cant associations were observed for the patterns char-
acterized by alcohol consumption. Results were similar
for proximal and distal colon tumors, but none of the
patterns was significantly associated with the risk of
rectal cancer, although a positive relation is suggested for
the ‘western’ pattern.
This study is limited by the need for methodological
homogeneity assumptions across a set of studies with
important differences in population characteristics, study
design, and methods used for measuring exposure and for
characterizing dietary patterns. However, three patterns
that were considered equivalent for the purpose of the
present analysis emerged from most studies, and the
heterogeneity of the results was low to moderate for most
summary estimates, except when analyzing the patterns
defined by a high consumption of alcohol. The observed
homogeneity results from the grouping of results for
meta-analysis based on the main components of the
dietary patterns instead of their arbitrarily attributed
labels. For example, a Chinese study (Butler et al., 2008)
identified two dietary patterns whose labels (‘vegetable-
fruit-soy’ and ‘meat-dim sum’) suggested them to be
similar to the ‘healthy’ and the ‘western,’ respectively,
but the results were not considered for meta-analysis
because their main components did not resemble those
from the patterns described in the remaining studies.
The observed homogeneity in the results also reflects
essentially common dietary habits across populations, at
least in what concerns the aggregation of individual
exposures known to be associated with colorectal cancer.
 6 European Journal of Cancer Prevention 2011, Vol 00 No 00

Fig. 2

Study Relative %
ID risk (95% CI) weight

Healthy
Stefani, 2009 (M) 0.48 (0.26, 0.88) 3.15
Stefani, 2009 (F) 0.62 (0.35, 1.09) 3.44
Satia, 2009 (W) 0.40 (0.28, 0.57) 6.20
Satia, 2009 (B) 1.10 (0.73, 1.66) 5.16
Kim, 2005 (M) 0.83 (0.49, 1.41) 3.80
Kim, 2005 (F) 0.76 (0.39, 1.49) 2.67
Wu, 2004 (M) 0.84 (0.64, 1.10) 7.62
Dixon, 2004 (M - Finland) 1.05 (0.66, 1.67) 4.50
Dixon, 2004 (M - the Netherlands) 0.93 (0.65, 1.33) 6.07
Dixon, 2004 (F - the Netherlands) 0.78 (0.53, 1.14) 5.69
Dixon, 2004 (F - Sweden) 0.96 (0.71, 1.30) 7.00
Fung, 2003 (F) 0.71 (0.50, 1.00) 6.20
Terry, 2001 (F) 0.83 (0.54, 1.27) 5.02
Slattery, 1998 (M) 0.66 (0.50, 0.87) 7.61
Slattery, 1998 (F) 0.73 (0.55, 0.97) 7.36
Randall, 1992 (M) 0.96 (0.79, 1.17) 9.17
Randall, 1992 (F) 1.00 (0.83, 1.21) 9.34
Subtotal (i 2 = 55.1%, P = 0.003) 0.80 (0.70, 0.90) 100.00

Western
Stefani, 2009 (M) 2.19 (1.20, 3.98) 4.09
Stefani, 2009 (F) 1.31 (0.71, 2.42) 3.93
Satia, 2009 (W) 1.10 (0.71, 1.71) 6.39
Satia, 2009 (B) 0.90 (0.52, 1.56) 4.68
Kim, 2005 (M) 1.05 (0.63, 1.75) 5.22
Kim, 2005 (F) 2.21 (1.10, 4.44) 3.16
Wu, 2004 (M) 1.21 (0.91, 1.60) 10.85
Dixon, 2004 (M - Finland) 1.12 (0.60, 2.09) 3.82
Dixon, 2004 (M - the Netherlands) 0.96 (0.64, 1.43) 7.40
Dixon, 2004 (F - the Netherlands) 0.92 (0.63, 1.35) 7.80
Dixon, 2004 (F - Sweden) 1.62 (1.12, 2.34) 8.13
Fung, 2003 (F) 1.46 (0.97, 2.19) 7.17
Terry, 2001 (F) 0.93 (0.57, 1.52) 5.49
Nkondjock, 2005 (F) 1.60 (0.90, 2.85) 4.34
Slattery, 1998 (M) 1.80 (1.28, 2.53) 8.92
Slattery, 1998 (F) 1.49 (1.05, 2.12) 8.61
2
Subtotal (i = 31.7%, P = 0.109) 1.29 (1.13, 1.48) 100.00

Drinker
Stefani, 2009 (M) 0.60 (0.31, 1.15) 5.72
Stefani, 2009 (F) 0.79 (0.44, 1.42) 6.96
Terry, 2001 (F) 1.14 (0.78, 1.66) 16.69
Nkondjock, 2005 (F) 0.76 (0.45, 1.29) 8.61
Slattery, 1998 (M) 1.09 (0.84, 1.42) 34.28
Slattery, 1998 (F) 0.91 (0.68, 1.22) 27.73
2
Subtotal (i = 0.6%, P = 0.412) 0.96 (0.82, 1.12) 100.00

0.2 0.5 1 2 5

Meta-analyses of studies quantifying the association between dietary patterns and colon cancer. B, blacks; CI, confidence interval; F, females;
M, males; W, whites.

Unlike the previous systematic reviews on this topic
(Miller et al., 2010a; Randi et al., 2010) our study provides
summary estimates that may be interpreted as an average
effect of locale-specific dietary patterns having important
similarities in their main components.
It is possible that other dietary patterns, probably more
setting-specific than the ones considered in the present
meta-analysis, are also relevant to colon or rectal cancer
risk, but more research on this topic is needed for a larger
number of studies depicting similarities regarding other
patterns to allow the computation of summary estimates.
Studies assessing dietary patterns through cluster analysis
(Rouillier et al., 2005; Wirfalt et al., 2009) were not
considered for quantitative synthesis because a direct
comparison between studies is not possible without
additional assumptions, which can hardly be made,
regarding the homogeneity of the reference groups for
the computation of RR estimates. In fact, the RR
estimates derived from these studies reflect the contrast
between subjects with different patterns (with both
these patterns and the choice of the reference group
varying across populations and studies) instead of a
comparison of subjects with different degrees of adoption
of a specific pattern (as when the highest versus the
lowest exposures to a certain factor are compared). This
is reflected in the larger heterogeneity of results across
the studies excluded. Wirfalt et al. (2009) showed that
high intakes of ‘vegetables and fruits’ were associated
with 15 and 26% reductions in the risk of colorectal and
rectal cancers among men, respectively, when compared
with less salutary food choices, and no association was

Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
 Dietary patterns and colorectal cancer Magalhães et al. 7

Fig. 3

Study Relative %
ID risk (95% CI) weight

Healthy
Kurotani, 2010 (M, F) 0.95 (0.64, 1.41) 10.71
Williams, 2009 (B, M, F) 1.50 (0.71, 3.17) 3.33
Stefani, 2009 (M) 0.72 (0.45, 1.16) 7.68
Stefani, 2009 (F) 0.88 (0.46, 1.69) 4.32
Flood, 2008 (M) 0.81 (0.62, 1.05) 20.30
Flood, 2008 (F) 1.26 (0.85, 1.87) 10.74
Kim, 2005 (M) 0.76 (0.37, 1.57) 3.54
Kim, 2005 (F) 1.43 (0.62, 3.29) 2.72
Dixon, 2004 (M - Finland) 1.48 (0.88, 2.49) 6.59
Dixon, 2004 (M - the Netherlands) 1.23 (0.83, 1.83) 10.69
Dixon, 2004 (F - the Netherlands) 1.33 (0.76, 2.34) 5.68
Dixon, 2004 (F - Sweden) 1.12 (0.70, 1.79) 7.93
Terry, 2001 (F) 0.77 (0.44, 1.35) 5.75
Subtotal (i 2 = 10.8% P = 0.337) 1.02 (0.88, 1.17) 100.00

Western
Williams, 2009 (W, M, F) 1.84 (1.08, 3.14) 8.61
Williams, 2009 (B, M, F) 0.89 (0.27, 2.97) 2.53
Stefani, 2009 (M) 1.76 (1.06, 2.93) 9.13
Stefani, 2009 (F) 1.64 (0.76, 3.55) 5.23
Flood, 2008 (M) 1.26 (0.97, 1.64) 15.62
Flood, 2008 (F) 1.00 (0.69, 1.45) 12.32
Kim, 2005 (M) 0.73 (0.36, 1.47) 6.06
Kim, 2005 (F) 0.77 (0.32, 1.84) 4.35
Dixon, 2004 (M - Finland) 2.21 (1.07, 4.57) 5.75
Dixon, 2004 (M - the Netherlands) 0.83 (0.51, 1.35) 9.67
Dixon, 2004 (F - the Netherlands) 0.82 (0.48, 1.40) 8.67
Dixon, 2004 (F - Sweden) 0.56 (0.27, 1.17) 5.67
Terry, 2001 (F) 1.20 (0.61, 2.36) 6.39
Subtotal (i 2 = 40.6%, P = 0.063) 1.13 (0.92, 1.39) 100.00

Drinker
Stefani, 2009 (M) 1.14 (0.70, 1.85) 37.47
Stefani, 2009 (F) 0.40 (0.19, 0.84) 27.25
Terry, 2001 (F) 1.03 (0.60, 1.76) 35.27
Subtotal (i 2 = 65.1%, P = 0.057) 0.83 (0.47, 1.45) 100.00

0.2 0.5 1 2 5

Meta-analyses of studies quantifying the association between dietary patterns and rectal cancer. B, blacks; CI, confidence interval; M, males;
F, females; W, whites.

found for colon cancer or among women. Rouillier et al.
(2005) defined five food patterns: ‘low-energy’ (used as
the reference group in this analysis), ‘high-starch, high
fat, and low-fruit,’ ‘high-processed meat, energy, alcohol,
and starchy foods,’ ‘high-fish, cereals, honey, olive oil,
fruit and vegetables,’ and ‘high-flour, sugar, chocolate,
animal fats, and eggs.’ The low energy diet seemed as
protective all along the adenoma-carcinoma sequence,
contrary to a high-energy, high-processed meat and
animal fat diet, but none of the results were statistically
significant for colorectal cancer.
The study by Bravi et al. 2010 was also excluded because
it identified patterns of nutrient intake that were labeled
as ‘animal products,’ ‘starch-rich,’ ‘vitamins and fiber,’ and
‘unsaturated fats’ (animal or vegetable source). The
starch-rich pattern was associated with an increased risk
of both colon and rectal cancer, whereas the vitamins and
fiber pattern was associated with a reduced risk of rectal
cancer, and the unsaturated fats pattern with a reduced
risk only for colon cancer. The interpretation of these
findings is not substantially different from the conclu-
sions of our meta-analysis.
Our results also need to be discussed in the context of
the available evidence on the relation between individual
food items and colorectal cancer. The 2007 World Cancer
Research Fund/American Institute for Cancer Research
report concluded that there was convincing evidence of a
causal role of red and processed meat, obesity, and alcohol
(in men; probable risk in women) in the etiology of
colorectal cancer, and foods containing dietary fiber,
garlic, milk, and calcium were considered to be probable
protective factors. The relation between dietary patterns
and colorectal cancer, however, was not formally assessed
in the study, although only seven of the 16 studies
included in our meta-analyses (Butler et al., 2008; Flood
et al., 2008; De Stefani et al., 2009; Satia et al.,
2009; Williams et al., 2009; Kurotani et al., 2010; Miller
et al., 2010b) were published after 2007.

Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
 8 European Journal of Cancer Prevention 2011, Vol 00 No 00

Fig. 4

Healthy Western Drinker

Funnel plot with pseudo 95% Funnel plot with pseudo 95% Funnel plot with pseudo 95%
Colon cancer confidence limits confidence limits confidence limits
0 0 0

Standard error of RR

Standard error of RR
Standard error of RR
Healthy (Begg´s: P = 0.19;
Egger´s: P = 0.15) 0.1 0.1 0.1

Western (Begg´s: P = 0.72; 0.2 0.2 0.2

Egger´s: P = 0.88)
0.3 0.3 0.3
Drinker (Begg´s: P = 0.04;
Egger´s: P = 0.07) 0.4 0.4 0.4
0.5 1 1 0.5 1
RR RR RR

Funnel plot with pseudo 95% Funnel plot with pseudo 95% Funnel plot with pseudo 95%
Rectal cancer confidence limits confidence limits confidence limits
0 0 0
Standard error of RR

Standard error of RR

Standard error of RR
Healthy (Begg´s: P = 0.71;
0.1 0.1
Egger´s: P = 0.25) 0.2
0.2 0.2
Western (Begg´s: P = 0.90;
Egger´s: P = 0.61) 0.3 0.4
0.3
Drinker (Begg´s: P = 0.12; 0.4 0.6 0.4
Egger´s: P = 0.08)
0.5 1 0.5 1 0.5 1
RR RR RR

Funnel plots of studies evaluating the association between dietary patterns and colon or rectum cancers. Begg’s adjusted rank correlation test and
Egger’s regression asymmetry test.

A recent review of 10 independent meta-analyses con- knowledge an updated quantification of dietary determi-
cluded that alcohol was responsible for 53% and 69% nants of cancer that goes beyond single nutrients and foods.
increased risk of tumors of the colon and rectum, No significant association was observed between any of the
respectively (Huxley et al., 2009). The small number of most frequently observed dietary patterns and rectal
studies included in our review that defined dietary cancer, but these results support the current guidelines
patterns labeled as ‘drinker’ may reflect a lower homo- and recommendations from all the major scientific associa-
geneity in the patterns associated with alcohol drinking tions that encourage a high consumption of fruit and
and precludes a more robust assessment of these vegetables and a low consumption of red and processed
exposures. In the review referred to above (Huxley meat for primary prevention of colon cancer.
et al., 2009), red and processed meats were identified as
risk factors for colon and rectal cancers, although for
processed meats in relation to rectal cancers, the risk Acknowledgements
increased but the association was not statistically Conflicts of interest
significant, which is in accordance with our findings of a None declared.
nonsignificant association between the ‘western’ pattern
and rectal cancer. The review of the meta-analyses
(Huxley et al., 2009) found no associations for fruit, References
Begg CB, Mazumdar M (1994). Operating characteristics of a rank correlation
vegetables, fish, and poultry for colon or rectal cancer test for publication bias. Biometrics 50:1088–1101.
(Huxley et al., 2009), in accordance with the conclusions Bravi F, Edefonti V, Bosetti C, Talamini R, Montella M, Giacosa A, et al. ( 2010).
of the World Cancer Research Fund/American Institute Nutrient dietary patterns and the risk of colorectal cancer: a case–control
study from Italy. Cancer Causes Control 21:1911–8.
for Cancer Research, but contrasting with our finding Butler LM, Wang R, Koh WP, Yu MC (2008). Prospective study of dietary patterns
regarding the ‘healthy’ patterns. The latter, however, are and colorectal cancer among Singapore Chinese. Br J Cancer 99:
characterized by high intake of potential protective 1511–1516.
Chan AT, Giovannucci EL (2010). Primary prevention of colorectal cancer.
factors other than fruit and vegetables. Gastroenterology 138:2029–2043e10.
De Stefani E, Deneo-Pellegrini H, Boffetta P, Ronco AL, Aune D, Acosta G
Our analysis also adds the assessment of the relation (2009). Dietary patterns and risk of cancer: a factor analysis in Uruguay.
between dietary exposures and proximal and distal Int J Cancer 124:1391–1397.
cancers of the colon to previous research on this topic. Dixon LB, Balder HF, Virtanen MJ, Rashidkhani B, Mannisto S, Krogh V, et al.
(2004). Dietary patterns associated with colon and rectal cancer: results from
The present systematic review and meta-analysis provides the dietary patterns and cancer (DIETSCAN) project. Am J Clin Nutr
80:1003–1011.
a comprehensive assessment of the association between Egger M, Davey Smith G, Schneider M, Minder C (1997). Bias in meta-analysis
dietary patterns and colorectal cancer, adding to previous detected by a simple, graphical test. BMJ 315:629–634.

Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Ferlay J, Shin H, Bray F, Forman D, Mathers C, Parkin D, et al. GLOBOCAN. 2008
Cancer incidence and mortality worldwide. IARC CancerBase No. 10 [Internet]
2010; Available at: http://globocan.iarc.fr. [Accessed 1 March 2011].
Flood A, Rastogi T, Wirfalt E, Mitrou PN, Reedy J, Subar AF, et al. (2008). Dietary
patterns as identified by factor analysis and colorectal cancer among middle-
aged Americans. Am J Clin Nutr 88:176–184.
Fung T, Hu FB, Fuchs C, Giovannucci E, Hunter DJ, Stampfer MJ, et al. (2003).
Major dietary patterns and the risk of colorectal cancer in women. Arch Intern
Med 163:309–314.
Higgins J, Thompson S (2002). Quantifying heterogeneity in a meta-analysis. Stat
Med 21:1539–1558.
Huxley RR, Ansary-Moghaddam A, Clifton P, Czernichow S, Parr CL, Woodward
M (2009). The impact of dietary and lifestyle risk factors on risk of colorectal
cancer: a quantitative overview of the epidemiological evidence. Int J Cancer
125:171–180.
Jackson D, White IR, Thompson SG (2010). Extending Dersimonian and Laird’s
methodology to perform multivariate random effects meta-analyses. Stat Med
29:1282–1297.
Kesse E, Clavel-Chapelon F, Boutron-Ruault MC (2006). Dietary patterns and risk
of colorectal tumors: a cohort of French women of the national education
system (E3N). Am J Epidemiol 164:1085–1093.
Kim MK, Sasaki S, Otani T, Tsugane S (2005). Dietary patterns and subsequent
colorectal cancer risk by subsite: a prospective cohort study. Int J Cancer
115:790–798.
Kurotani K, Budhathoki S, Man Joshi A, Yin G, Toyomura K, Kono S, et al. (2010).
Dietary patterns and colorectal cancer in a Japanese population: the Fukuoka
colorectal cancer study. Br J Nutr1–9.
McCredie M (1998). Cancer epidemiology in migrant populations. Recent
Results Cancer Res 154:298–305.
Michels KB, Schulze M (2005). Can dietary patterns help us detect diet–disease
associations? Nutrition Research Reviews 18:241–248.
Miller PE, Lesko SM, Muscat JE, Lazarus P, Hartman TJ (2010a). Dietary patterns
and colorectal adenoma and cancer risk: a review of the epidemiological
evidence. Nutr Cancer 62:413–424.
Miller PE, Lazarus P, Lesko SM, Muscat JE, Harper G, Cross AJ, et al. (2010b).
Diet index-based and empirically derived dietary patterns are associated with
colorectal cancer risk. J Nutr 140:1267–1273.
Monroe KR, Hankin JH, Pike MC, Henderson BE, Stram DO, Park S, et al. (2003).
Correlation of dietary intake and colorectal cancer incidence among Mexican-
American migrants: the multiethnic cohort study. Nutr Cancer 45:133–147.
Nilsson B, Gustavson-Kadaka E, Rotstein S, Hakulinen T, Rahu M, Aareleid T,
et al. (1993). Cancer incidence in Estonian migrants to Sweden. Int J Cancer
55:190–195.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
View publication stats
Dietary patterns and colorectal cancer Magalhães et al. 9
Nkondjock A, Ghadirian P (2005). Associated nutritional risk of breast and colon
cancers: a population-based case–control study in Montreal, Canada.
Cancer Lett 223:85–91.
Petiti DB (2000). Meta-analysis, decision analysis, and cost effectiveness
analysis: methods for quantitative synthesis in medicine. 2nd ed. Oxford:
Oxford University Press.
Peto J (2001). Cancer epidemiology in the last century and the next decade.
Nature 411:390–395.
Randall E, Marshall JR, Brasure J, Graham S (1992). Dietary patterns and colon
cancer in western New York. Nutr Cancer 18:265–276.
Randi G, Edefonti V, Ferraroni M, La Vecchia C, Decarli A (2010). Dietary
patterns and the risk of colorectal cancer and adenomas. Nutr Rev 68:
389–408.
Rouillier P, Senesse P, Cottet V, Valleau A, Faivre J, Boutron-Ruault MC (2005).
Dietary patterns and the adenomacarcinoma sequence of colorectal cancer.
Eur J Nutr 44:311–318.
Satia JA, Tseng M, Galanko JA, Martin C, Sandler RS (2009). Dietary patterns
and colon cancer risk in whites and African Americans in the North Carolina
colon cancer study. Nutr Cancer 61:179–193.
Shike M. Diet and lifestyle in the prevention of colorectal cancer: an overview. Am
J Med 1999; 106:11S–15S, discussion 50S-51S.
Slattery ML, Boucher KM, Caan BJ, Potter JD, Ma KN (1998). Eating patterns
and risk of colon cancer. Am J Epidemiol 148:4–16.
Slattery ML (2000). Diet, lifestyle, and colon cancer. Semin Gastrointest Dis
11:142–146.
Terry P, Hu FB, Hansen H, Wolk A (2001). Prospective study of major dietary
patterns and colorectal cancer risk in women. Am J Epidemiol 154:
1143–1149.
Williams CD, Satia JA, Adair LS, Stevens J, Galanko J, Keku TO, et al. (2009).
Dietary patterns, food groups, and rectal cancer risk in whites and African-
Americans. Cancer Epidemiol Biomarkers Prev 18:1552–1561.
Wirfalt E, Midthune D, Reedy J, Mitrou P, Flood A, Subar AF, et al. (2009).
Associations between food patterns defined by cluster analysis and
colorectal cancer incidence in the NIH-AARP diet and health study. Eur J
Clin Nutr 63:707–717.
World Cancer Research Fund. (1997). Food, nutrition and prevention of cancer: a
global perspective. Washington DC: American Institute for Cancer Research.
World Cancer Research Fund and American Institute for Cancer Research.
(2007). Food, Nutrition, Physical Activity, and the Prevention of Cancer: a
Global Perspective. Washington DC: AICR.
Wu K, Hu FB, Fuchs C, Rimm EB, Willett WC, Giovannucci E (2004). Dietary
patterns and risk of colon cancer and adenoma in a cohort of men (United
States). Cancer Causes Control 15:853–862.