A Prospective Diet-Wide Association Study For Risk of Colorectal Cancer in EPIC

Clinical Gastroenterology and Hepatology 2021;-:-–-
A Prospective Diet-Wide Association Study for Risk of

Colorectal Cancer in EPIC
Nikos Papadimitriou,*,‡,a Emmanouil Bouras,*,a Piet A. van den Brandt,§
David C. Muller,jj Areti Papadopoulou,* Alicia K. Heath,jj Elena Critselis,¶
Marc J. Gunter,‡ Paolo Vineis,jj Pietro Ferrari,‡ Elisabete Weiderpass,‡
Heiner Boeing,# Nadia Bastide,** Melissa A. Merritt,‡‡ David S. Lopez,§§,jjjj
Manuela M. Bergmann,¶¶ Aurora Perez-Cornago,## Matthias Schulze,***,‡‡‡
Guri Skeie,§§§ Bernard Srour,jjjjjj Anne Kirstine Eriksen,¶¶¶ Stina Boden,###
Ingegerd Johansson,**** Therese Haugdahl Nøst,§§§ Marco Lukic,§§§
Fulvio Ricceri,‡‡‡‡,§§§§ Ulrika Ericson,jjjjjjjj José María Huerta,¶¶¶¶,####
Christina C. Dahm,***** Claudia Agnoli,‡‡‡‡‡ Pilar Exezarreta Amiano,####,jjjjjjjjjj
Anne Tjønneland,¶¶¶,¶¶¶¶¶ Aurelio Barricarte Gurrea,#####
Bas Bueno-de-Mesquita,****** Eva Ardanaz,####,#####,‡‡‡‡‡‡ Jonna Berntsson,jjjjjjjjjjjj
Maria-Jose Sánchez,####,¶¶¶¶¶¶,######,******* Rosario Tumino,‡‡‡‡‡‡‡
Salvatore Panico,§§§§§§§ Verena Katzke,jjjjjj Paula Jakszyn,jjjjjjjjjjjjjj,¶¶¶¶¶¶¶
Giovanna Masala,####### Jeroen W. G. Derksen,******** J. Ramón Quirós,‡‡‡‡‡‡‡‡
Gianluca Severi,§§§§§§§§,jjjjjjjjjjjjjjjj Amanda J. Cross,jj Ellio Riboli,jj Ioanna Tzoulaki,*,jj
and Konstantinos K. Tsilidis*,jj
*Department of Hygiene and Epidemiology, University of Ioannina School of Medicine, Ioannina, Greece; ‡International Agency
for Research on Cancer, Lyon, France; §Department of Epidemiology, GROW School for Oncology and Developmental Biology,
Care and Public Health Research Institute, Maastricht University, Maastricht, Netherlands; jjDepartment of Epidemiology and
Biostatistics, School of Public Health, Imperial College London, London, United Kingdom; ¶Biomedical Research Foundation of
the Academy of Athens, Athens, Greece; #Department of Epidemiology, German Institute of Human Nutrition Potsdam-
Rehbrücke, Bergholz-Rehbrücke, Germany; **U1018, Nutrition, Hormones and Women’s Health Team, Centre for Research in
Epidemiology and Population Health, Inserm, Villejuif, France; ‡‡University of Hawai’i Cancer Center, Honolulu, Hawaii;
§§
Department of Preventive Medicine and Population Health, University of Texas Medical Branch School of Medicine,
Galveston, Texas; jjjjDivision of Urology, McGovern Medical School, University of Texas Health Science Center at Houston,
Houston, Texas; ¶¶German Institute of Human Nutrition Potsdam-Rehbrücke, Bergholz-Rehbrücke, Germany; ##Cancer
Epidemiology Unit, Nuffield Department of Population Health, University of Oxford, Oxford, United Kingdom; ***Department of
Molecular Epidemiology, German Institute of Human Nutrition Potsdam-Rehbruecke, Nuthetal, Germany; ‡‡‡Institute of
Nutrition Science, University of Potsdam, Nuthetal, Germany; §§§Department of Community Medicine, Faculty of Health
Sciences, University of Tromsø-The Arctic University of Norway, Tromsø, Norway; jjjjjjDivision of Cancer Epidemiology, German
Cancer Research Center (DKFZ), Heidelberg, Germany; ¶¶¶Diet, Genes and Environment, Danish Cancer Society Research
Center, Copenhagen, Denmark; ###Department of Radiation Sciences-Oncology, Umeå University, Umeå, Sweden;
****Department of Odontology, Umeå University, Umeå, Sweden; ‡‡‡‡Department of Clinical and Biological Sciences, University
of Turin, Turin, Italy; §§§§Unit of Epidemiology, Regional Health Service ASL TO3, Grugliasco, Italy; jjjjjjjjDepartment of Clinical
Sciences in Malmö, Lund University, Malmö, Sweden; ¶¶¶¶Murcia Regional Health Council, IMIB-Arrixaca, Murcia, Spain;
####
CIBER of Epidemiology and Public Health, Instituto de Salud Carlos III, Madrid, Spain; *****Department of Public Health,
Aarhus University, Aarhus, Denmark; ‡‡‡‡‡Epidemiology and Prevention Unit, Fondazione IRCCS Istituto Nazionale dei Tumori,
Milan, Italy; jjjjjjjjjjPublic Health Division of Gipuzkoa, Biodonostia Research Institute, San Sebastián, Spain; ¶¶¶¶¶Department of
Public Health, University of Copenhagen, Copenhagen, Denmark; #####Instituto de Salud Pública de Navarra, Pamplona, Spain;
******Department for Determinants of Chronic Diseases, National Institute for Public Health and the Environment, Bilthoven, the
Netherlands; ‡‡‡‡‡‡Navarra Institute for Health Research (IdiSNA), Pamplona, Spain; jjjjjjjjjjjjOncology and Pathology,
Department of Clinical Sciences Lund, Lund University, Lund, Sweden; ¶¶¶¶¶¶Escuela Andaluza de Salud Pública, Granada,
Spain; ######Instituto de Investigación Biosanitaria (ibs.GRANADA), Granada, Spain; *******Department of Preventive Medicine
and Public Health, University of Granada, Granada, Spain; ‡‡‡‡‡‡‡Cancer Registry and Histopathology Department, Provincial
Health Authority, Ragusa, Italy; §§§§§§§Dipartimento di Medicina Clinica e Chirurgia, University of Naples Federico II, Naples,
Italy; jjjjjjjjjjjjjjUnit of Nutrition and Cancer, Cancer Epidemiology Research Programme, Catalan Institute of Oncology, Barcelona,
Spain; ¶¶¶¶¶¶¶Blanquerna School of Health Sciences, Ramon Llull University, Barcelona, Spain; #######Cancer Risk Factors and
a
Authors share co-first authorship.
Abbreviations used in this paper: CI, confidence interval; CRC, colorectal
cancer; DWAS, diet-wide association study; EPIC, European Prospective
Investigation into Cancer and Nutrition; FDR, false discovery rate; HR, © 2021 by the AGA Institute
hazard ratio; MR, Mendelian randomization; NLCS, Netherlands Cohort 1542-3565/$36.00
Study; WCRF, World Cancer Research Fund. https://doi.org/10.1016/j.cgh.2021.04.028
2 Papadimitriou et al Clinical Gastroenterology and Hepatology Vol. -, No. -
Life-Style Epidemiology Unit, Institute for Cancer Research, Prevention and Clinical Network, Florence, Italy; ********Julius
Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht University, Utrecht, the Netherlands;
‡‡‡‡‡‡‡‡
Public Health Directorate, Asturias, Spain; §§§§§§§§CESP UMR1018, Gustave Roussy, Équipe “Exposome et Hérédité,”
Inserm-UVSQ, Université Paris-Saclay, Villejuif, France; and jjjjjjjjjjjjjjjjDepartment of Statistics, Computer Science and
Applications, University of Florence, Florence, Italy
BACKGROUND & AIMS: Evidence regarding the association of dietary exposures with colorectal cancer (CRC) risk is not
consistent with a few exceptions. Therefore, we conducted a diet-wide association study
(DWAS) in the European Prospective Investigation into Cancer and Nutrition (EPIC) to evaluate
the associations between several dietary exposures with CRC risk.
METHODS: The association of 92 food and nutrient intakes with CRC risk was assessed in 386,792 par-
ticipants, 5069 of whom developed incident CRC. Correction for multiple comparisons was
performed using the false discovery rate, and emerging associations were examined in the
Netherlands Cohort Study (NLCS). Multiplicative gene-nutrient interactions were also tested in
EPIC based on known CRC-associated loci.
RESULTS: In EPIC, alcohol, liquor/spirits, wine, beer/cider, soft drinks, and pork were positively associ-
ated with CRC, whereas milk, cheese, calcium, phosphorus, magnesium, potassium, riboflavin,
vitamin B6, beta carotene, fruit, fiber, nonwhite bread, banana, and total protein intakes were
inversely associated. Of these 20 associations, 13 were replicated in the NLCS, for which a meta-
analysis was performed, namely alcohol (summary hazard ratio [HR] per 1-SD increment in
intake: 1.07; 95% confidence interval [CI], 1.04–1.09), liquor/spirits (HR per 1-SD increment in
intake, 1.04; 95% CI, 1.02–1.06), wine (HR per 1-SD increment in intake, 1.04; 95% CI, 1.02–
1.07), beer/cider (HR per 1-SD increment in intake, 1.06; 95% CI, 1.04–1.08), milk (HR per 1-SD
increment in intake, 0.95; 95% CI, 0.93–0.98), cheese (HR per 1-SD increment in intake, 0.96;
95% CI, 0.94–0.99), calcium (HR per 1-SD increment in intake, 0.93; 95% CI, 0.90–0.95),
phosphorus (HR per 1-SD increment in intake, 0.92; 95% CI, 0.90–0.95), magnesium (HR per 1-
SD increment in intake, 0.95; 95% CI, 0.92–0.98), potassium (HR per 1-SD increment in intake,
0.96; 95% CI, 0.94–0.99), riboflavin (HR per 1-SD increment in intake, 0.94; 95% CI, 0.92–0.97),
beta carotene (HR per 1-SD increment in intake, 0.96; 95% CI, 0.93–0.98), and total protein (HR
per 1-SD increment in intake, 0.94; 95% CI, 0.92–0.97). None of the gene-nutrient interactions
were significant after adjustment for multiple comparisons.
CONCLUSIONS: Our findings confirm a positive association for alcohol and an inverse association for dairy
products and calcium with CRC risk, and also suggest a lower risk at higher dietary intakes of
phosphorus, magnesium, potassium, riboflavin, beta carotene, and total protein.
Keywords: nutrition; cohort study; colorectal cancer; epidemiology.

- 2021 Diet-Wide Association Study for Risk of CRC 3
olorectal cancer (CRC) is the third most common

C type of cancer worldwide with over 1.8 million
new cases and over 800,000 deaths in 2018.1 The inci-
What You Need to Know
Background
dence rates are higher in high income countries, but
there has been a recent large increase in the rates in low- Obesity and lack of physical activity are well-
and middle-income countries potentially due to the established risk factors of colorectal cancer (CRC)
“Westernization” of these societies.1 Several aspects of risk but evidence regarding the association of spe-
the Western lifestyle such as obesity and lack of physical cific foods and nutrients with CRC is not consistent,
activity are well-established risk factors of CRC,2,3 but ev- with a few exceptions.
idence regarding diet, and in particular the association of Findings
specific foods and nutrients with CRC, is not consistent, In an analysis of 92 foods and nutrients from over
with a few exceptions.4 The World Cancer Research 350,000 participants in the European Prospective
Fund (WCRF) Third Expert Report identified strong evi- Investigation into Cancer and Nutrition cohort and
dence that consuming processed meat, red meat, and replication analyses in the Netherlands Cohort Study,
alcohol increases risk of CRC, whereas consumption of we demonstrate positive associations for alcohol and
whole grains, foods containing dietary fiber, and dairy inverse associations for dairy and calcium intake.
products lowers CRC risk.4 Associations for other foods Additionally, lower CRC risk is observed at higher
and nutrients and CRC risk exist but are inconsistent dietary intakes of phosphorus, magnesium, potas-
and currently provide limited evidence according to sium, riboflavin, beta carotene, and total protein.
WCRF.4
The aim of this study was to systematically examine Implications for patient care
the associations between a wide set of dietary factors Consumption of alcohol increases the risk of CRC and
and risk of CRC in the European Prospective Investiga- should be discouraged, while dairy and calcium
tion into Cancer and Nutrition (EPIC) and the intake seems to reduce the risk of CRC and should be
Netherlands Cohort Study (NLCS), by conducting a diet- encouraged along with other nutritious dietary
wide association study (DWAS).5 The DWAS takes an choices.
analogous strategy to that of a genome-wide association
study by separately estimating associations for each food
and nutrient, using adjustments for multiple compari- factors (63 foods and 29 nutrients) were included in the
sons, and replicating promising associations in an inde- current analysis.
pendent study. In the NLCS, information on dietary intake over the
preceding 12 months was assessed at baseline using a
semi-quantitative 150-item food frequency question-
Materials and Methods naire, which has been validated and tested for repro-
ducibility (Supplementary Methods).8
Study Populations
Identification of CRC Cases
EPIC is a large European multicenter prospective
In EPIC and NLCS, incident CRC cases were identified
cohort that consists of 521,324 participants, mostly be-
by record linkage with population-based cancer regis-
tween 35 and 70 years of age, recruited between 1992
tries or a combination of registries, insurance records,
and 2000 from 23 centers across 10 European coun-
and active follow up of the study participants or their
tries.6 A total of 386,792 participants (71% women)
relatives. More details are provided in the
were included in the present analysis after pertinent
Supplementary Methods.
exclusions (Supplementary Methods).
NLCS is a prospective cohort study of 120,852 par-
ticipants, between 55 and 69 years of age and recruited Statistical Analyses
in 1986 from 204 computerized population registries
across the Netherlands that uses a case-cohort In EPIC, separate Cox proportional hazards regression
approach.7 Of the 5000 subcohort participants, 3893 models were used to investigate the associations be-
were included in the current analysis after pertinent tween each of the dietary factors with CRC risk. In the
exclusions (Supplementary Methods). NLCS, given the case-cohort design, Prentice-weighted
Cox proportional hazards regression models with
robust SE estimation were implemented.9 All of the
Assessment of Dietary Factors models were adjusted for total energy intake, smoking,
body mass index, physical activity, diabetes history, level
In EPIC, consumption of foods over the last 12 of education, and family history of CRC (in the NLCS
months was assessed at baseline using validated only) and further stratified by sex, age, and in EPIC also
country-specific food questionnaires.6 In total, 92 dietary by center.
To account for multiple comparisons, the false dis- approximately 30% of the participants were men, and
covery rate (FDR)–adjusted P values (or q values) were 47% were overweight or obese. About 50% of the
estimated for each association analysed.10 The dietary participants were never smokers, and 47% were
factors with an FDR <0.05 were subsequently selected physically active. More than half of the NLCS subcohort
for replication in the NLCS, and fixed-effects meta-anal- participants were male (54%), 47% were overweight or
ysis was performed to combine the results from the 2 obese, one-third (33%) were never smokers, and 48%
cohorts when heterogeneity was low or moderate (P spent more than 60 min/d on nonoccupational physical
value for heterogeneity > .1 and/or I2 50%). To activities.
further investigate the robustness of the associations
that were replicated in the NLCS, a mutual adjustment
DWAS in EPIC
model was used. Presence of nonlinear associations was
investigated using restricted cubic spline models. More
Of the 92 dietary factors that were examined in EPIC,
details on the statistical analyses methods are provided
20 were associated with CRC risk (FDR <0.05) (Figure 1;
in the Supplementary Methods.
Supplementary Table 2). Higher intakes of alcohol, li-
quor/spirits, wine, beer/cider, soft drinks, and pork were
Results positively associated with CRC, whereas higher milk,
cheese, calcium, phosphorus, magnesium, potassium,
Study Characteristics riboflavin, vitamin B6, beta carotene, fruit, fiber,
nonwhite bread, banana, and total protein intakes were
After a mean follow-up of 14.1 3.9 years, a total of associated with a lower CRC risk.
5069 (56.8% in women) incident malignant CRC cases After conducting the analysis by tumor subsite, evi-
were identified among the 386,792 included EPIC par- dence of heterogeneity between colon and rectal cancer
ticipants, of which 3143 were identified as colon (1495 was observed for intakes of magnesium, potassium,
proximal, 1435 distal, 213 unspecified) and 1715 as vitamin B6, and banana, with associations being inverse
rectal cancers. In the NLCS, 3765 cases (42.8% female) for colon cancer and null for rectal cancer
with incident and microscopically confirmed CRC were (Supplementary Table 3). Regarding proximal vs distal
included in the present analysis, of which 2612 were colon subsites, only total alcohol and wine had hetero-
colon (1348 proximal, 1187 distal) and 801 were rectal geneous results, whereby the associations were positive
cancers. only for distal colon cancer (Supplementary Table 4).
The main baseline characteristics of the study par- Additionally, heterogeneous associations were observed
ticipants are shown in Supplementary Table 1. In EPIC, by sex, for total alcohol and spirits, magnesium, fiber, and
Figure 1. Volcano plot showing results from the DWAS regarding the association between 92 dietary factors and CRC risk in
EPIC. The y-axis shows the FDR adjusted P values in –log10 scale from the Cox proportional hazards models for each dietary
factor. The x-axis shows the estimated HR for each dietary factor per 1-SD increase in daily consumption. The dashed
horizontal line represents the level of significance corresponding to FDR of 5%.
nonwhite bread, in which the associations were only Meta-analysis of EPIC and NLCS
observed in men (Supplementary Table 5). When we
The associations for most of the 20 dietary variables
investigated the association of red and processed meat
with CRC risk were homogeneous between EPIC and
with CRC risk by follow-up duration, a trend toward
NLCS, except for soft drinks, vitamin B6, fruit, fiber,
smaller HRs was observed as follow-up increased
nonwhite bread, banana, and pork (P value for hetero-
(Supplementary Figure 1). There was some evidence for
geneity < .1 and/or I2 > 50%), in which the associations
nonlinearity (P ¼ .028) in the association of alcohol
were null in the NLCS and therefore a meta-analysis was
intake and CRC risk (Supplementary Figure 2).
not performed. The remaining 13 associations yielded a
nominally significant summary finding: alcohol (HR per
Replication Analysis in the NLCS 1-SD increment in intake per day, 1.07; 95% CI,
1.04–1.09), liquor/spirits (HR per 1-SD increment in
Of the 20 associations with an FDR <0.05 in EPIC, 4 intake per day, 1.04; 95% CI, 1.02–1.06), wine (HR per 1-
associations reached nominal statistical significance (P < SD increment in intake per day, 1.04; 95% CI, 1.02–1.07),
.05) in the NLCS cohort in the analysis for CRC (Figure 2; beer/cider (HR per 1-SD increment in intake per day,
Supplementary Table 6), namely alcohol and liquor/ 1.06; 95% CI, 1.04–1.08), milk (HR per 1-SD increment in
spirits (positively) and milk and calcium intake intake per day, 0.95; 95% CI, 0.93–0.98), cheese (HR per
(inversely). An additional 4 associations, namely phos- 1-SD increment in intake per day, 0.96; 95% CI,
phorus, magnesium, riboflavin, and total protein, had a 0.94–0.99), calcium (HR per 1-SD increment in intake per
borderline inverse association in the NLCS, and the point day, 0.93; 95% CI, 0.90–0.95), phosphorus (HR per 1-SD
estimates were almost identical to the ones calculated increment in intake per day, 0.92; 95% CI, 0.90–0.95),
in EPIC. magnesium (HR per 1-SD increment in intake per day,
In a separate analysis by tumor subsite in the NLCS, 0.95; 95% CI, 0.92–0.98), potassium (HR per 1-SD
we found that most associations were consistent across increment in intake per day, 0.96; 95% CI, 0.94–0.99),
the different subsites with heterogeneous associations riboflavin (HR per 1-SD increment in intake per day,
only evident for phosphorus, potassium, vitamin B6, beta 0.94; 95% CI, 0.92–0.97), beta carotene (HR per 1-SD
carotene, and total protein in the analysis for colon vs increment in intake per day, 0.96; 95% CI, 0.93–0.98),
rectal cancer (Supplementary Tables 7 and 8). Little and total protein (HR per 1-SD increment in intake per
heterogeneity was observed by sex for CRC risk day, 0.94; 95% CI, 0.92–0.97) (Figure 2; Supplementary
(Supplementary Table 9). Table 6).
Figure 2. Forest plot showing

the hazard ratios and 95%
confidence intervals for the 20
FDR significant associations
(FDR <5%), in EPIC (—) and
NLCS (∙∙∙), as well as the re-
sults from a meta-analysis (MA)
(—). The x-axis shows the esti-
mated HR for each dietary fac-
tor for 1-SD increase in daily
consumption. The diamond and
the solid line represent the
pooled HR and 95% CI of the
MA. MA was not performed
when heterogeneity was high.
Pairwise Correlations and Mutual-Adjustment Gene-Nutrient Interaction Analysis

Analysis
Of the 73 20 gene-nutrient multiplicative in-
Most of the pair-wise correlation coefficients for the teractions that were tested, using the Bonferroni-
20 FDR-significant foods/nutrients in EPIC were weak adjusted P value threshold of 3.4 10–5, no interaction
and ranged from –0.25 to 0.79 (Figure 3). remained significant (Supplementary Table 11).
When alcohol, milk, cheese, calcium, phosphorus,
magnesium, potassium, riboflavin, beta carotene, and
total protein were included in a single multivariable- Discussion
adjusted model in EPIC, only alcohol remained signifi-
cantly associated with CRC risk (HR, 1.05; 95% CI, We used the DWAS approach to systematically eval-
1.03–1.11) (Supplementary Table 10). uate the association between dietary intakes of 92 foods
Figure 3. Pairwise partial correlation coefficients (Spearman’s r) of the 20 FDR-significant foods/nutrients in EPIC, adjusting
for age, sex, and center.
and nutrients and risk of CRC in EPIC and NLCS. We genetically proxied calcium or phosphorus concentra-
confirmed well-described associations in the literature tions.18,19 Additionally, although previous randomized
for alcoholic beverages (positive) and milk and calcium controlled trials have showed null associations for cal-
(inverse) with risk of CRC. In addition, our analysis cium supplementation in relation to CRC risk, a 13%
showed that higher intakes of phosphorus, magnesium, decreased risk of colorectal adenoma recurrence has
potassium, riboflavin, beta carotene, and total protein been reported in a meta-analysis of 4 randomized
were associated with a lower risk of CRC. controlled trials, with daily doses of calcium ranging
Alcohol consumption was positively associated with from 1200 to 2000 mg.20
risk of CRC in EPIC and the NLCS, and this association Many studies have investigated the association be-
was not different between colon and rectal cancer sub- tween red meat or processed meat consumption and risk
sites or by type of alcoholic beverage. In agreement, the of CRC. A dose-response meta-analysis by the WCRF
WCRF Third Expert Report has graded the quality of this third Expert Report concluded that there is strong evi-
evidence as strong.4 Persons with higher total alcohol dence that consuming red meat (including beef, pork,
consumption had a higher risk of CRC (summary HR per lamb, and goat from domesticated animals) or processed
SD increment in intake/day, 1.07; 95% CI, 1.04–1.09), meat (meat preserved by smoking, curing, salting, or
colon, and rectal cancer in the meta-analysis of EPIC and addition of chemical preservatives) increases the risk of
the NLCS. When we evaluated this association by prox- CRC by 12% per 100-g/d increment for red meat and
imal vs. distal colon cancer and by sex, we found het- 16% per 50-g/d for processed meat.4 A combination of
erogeneous associations in EPIC, with associations only mechanisms may contribute to the higher risk of colo-
present for distal colon cancer and in men, but these rectal tumorigenesis among individuals consuming
findings were not confirmed in the NLCS. The majority of larger amounts of red or processed meat. Cooking meat
the literature agrees that the positive association of at high temperatures may lead to the formation of het-
alcohol consumption with CRC risk is consistent by erocyclic amines and polycyclic aromatic hydrocarbons,
anatomical subsite and sex.11,12 Acetaldehyde, as a which have been associated with colorectal carcinogen-
metabolite of ethanol oxidation, can be carcinogenic in esis in experimental studies.21 Red meat also contains
colonocytes.13 Mendelian randomization (MR) studies haem iron at high levels that may stimulate the endog-
have failed to demonstrate an association between enous formation of carcinogenic N-nitroso compounds,
genetically proxied alcohol consumption and CRC risk, which promote colorectal tumorigenesis.22 Additionally,
but this analysis was underpowered to detect relatively processed meat can be an exogenous source of N-nitroso
small effects.14 compounds. Although accumulated evidence supports
Our study also confirmed the inverse association be- that higher intakes of red or processed meat are asso-
tween intake of dairy products and calcium with risk of ciated with higher risk of CRC, these findings were not
CRC, where individuals with higher calcium consumption replicated in our analysis in EPIC (HR per 36.2 g of red
had a 7% lower risk of CRC per 334.5 mg increment in meat intake daily, 1.02; 95% CI, 0.98–1.05; HR per 31.5 g
intake/day. One of the most prominent mechanisms by of processed meat intake daily, 1.04; 95% CI, 1.00–1.08).
which calcium is thought to act to reduce CRC risk is by An earlier report from EPIC in 2005, with a mean follow-
its ability to bind unconjugated bile acids and free fatty up of 4.8 years and 1329 incident CRC cases, observed a
acids, diminishing their potential toxic effects on the positive association between red and processed meat
colorectum.15 Heterogeneity by anatomical subsite or sex consumption with CRC risk.23 A potential reason for this
was not observed, in agreement with the WCRF meta- discrepancy is that EPIC, as most other cohorts, has
analysis and a more recent publication in the Nurses’ assessed meat consumption only during recruitment in
Health Study.4,11 Dairy products are also a rich source of the 1990s; thus, the current analysis assumes that con-
phosphorus, which was also inversely associated with sumption has stayed stable over 2 decades. However, a
CRC risk in our study but has been infrequently studied notable decrease in bovine meat consumption between
in other publications. A previous analysis of nutrient 2000 and 2013 has been noticed in Europe, which was
patterns in EPIC identified a pattern characterized by accompanied by an analogous increase in cheese, fish,
total protein, riboflavin, phosphorus and calcium that dairy, and poultry consumption. In the current paper, we
was associated with a 4% decreased CRC risk.16 All these observed a trend toward smaller HRs in the association
nutrients were analyzed independently in our analysis of red and processed meat with CRC risk as follow-up
and yielded inverse associations in EPIC that were robust increased. A recent time-varying exposure analysis in
after correcting for multiple testing and were replicated the Nurses’ Health Study and the Health Professionals
in the NLCS. Since several of these nutrients share Follow-up Study showed that a decrease in red meat
common sources of intake, a correlation of approxi- consumption and simultaneous increases in healthy
mately 0.50–0.70 was observed in EPIC, which makes it alternative food choices over time were associated with a
challenging to distinguish their independent effects.17 lower risk of all-cause mortality.24
Evidence from MR studies suggests that genetically The current DWAS study observed an inverse asso-
proxied milk consumption is associated with a reduced ciation of magnesium intake with risk of CRC, which
CRC risk but failed to demonstrate an association for agreed with the results of a recent meta-analysis of 7
observational studies.25 One purported mechanism by and overall dietary patterns were not accounted for. In-
which magnesium may be implicated in lower CRC risk is tercorrelations between dietary exposures may have led
by its potential to inhibit c-Myc oncogene expression in to low precision of the estimates, even though variance
colon cancer cells.26 Furthermore, magnesium has been inflation factors were relatively small, which might
shown to improve insulin sensitivity and lower plasma explain that none of the dietary factors remained in the
insulin concentrations, which may have an impact on multivariable adjusted model. Furthermore, it is possible
CRC development.27 No association of genetically proxied that there might be an association for foods or nutrients
circulating concentrations of magnesium was found in a that were not included in this analysis. Additionally, the
recent MR study.19 data derived from the Dutch food composition table were
We also observed an inverse association between not checked against the use of ENDB for nutrient calcu-
intake of beta carotene and risk of CRC, but few other lation, so discrepancies may have occurred, hence it is
studies have investigated this association.28 Our findings possible that some of the discrepancies observed be-
agree with a previous report from EPIC in 2014.28 tween the 2 cohorts for some dietary exposures are in
However, a cohort analysis in the ATBC (Alpha-Tocoph- part due to poor reproducibility in measurements.
erol, Beta-Carotene Cancer Prevention) trial, comprising Among the exposures for which heterogeneity was
26,951 middle-aged male smokers, showed no associa- observed between EPIC and NLCS, correlation between
tion between dietary beta carotene and risk of CRC.29 No the baseline food-frequency questionnaires and 24-hour
evidence of association between genetically proxied diet recalls has been reported to be good for fruit, fiber,
circulating concentrations of beta carotene and CRC have vitamin B6 and beverage consumption in the NLCS and
been reported in MR studies.19 fairly good for fiber and fruit across most EPIC centers,
Vitamins B2 and B6 are among the micronutrients but for exposures like nonwhite bread or vitamin B6 no
that play a pivotal role in one-carbon metabolism, which information was available.8,34 Finally, we cannot exclude
has been related to carcinogenesis because of its the possibility of residual confounding, although we
involvement in the synthesis of purines and pyrimidines adjusted for several potential confounders.
for subsequent DNA synthesis and in the synthesis of In conclusion, our study confirmed the well-
methionine for DNA methylation.30 Inverse associations established positive association for alcohol consump-
between riboflavin (vitamin B2) and vitamin B6 intake tion and inverse association for dairy products and
and CRC risk were observed in EPIC, but only the asso- calcium intake with CRC risk. The study further sug-
ciation with riboflavin was replicated in the NLCS. Pre- gested that higher intakes of magnesium, phosphorus,
vious studies on the association between riboflavin potassium, riboflavin, beta carotene, and total protein are
intake and CRC risk are scarce.31 Results from the associated with lower CRC risk.
Women’s Health Initiative Observational Study indicated
a 25% decreased CRC risk for the highest compared with
the lowest quartile of total riboflavin intake but was not Supplementary Material
statistically significant when only dietary intake of
riboflavin was considered.31 A meta-analysis of 8 studies Note: To access the supplementary material accom-
did not show an association between vitamin B6 intake panying this article, visit the online version of Clinical
and CRC risk, but blood levels of its active form, pyri- Gastroenterology and Hepatology at www.cghjournal.org,
doxal 5-phosphate, were associated with lower CRC and at https://doi.org/10.1016/j.cgh.2021.04.028.
risk.32
Little is known on the role that potassium may play in References
relation to CRC risk, and epidemiological evidence thus 1. Bray F, Ferlay J, Soerjomataram I, et al. Global cancer statistics
far is limited.33 We cannot rule out the possibility that 2018: GLOBOCAN estimates of incidence and mortality world-
the inverse association observed in our study may mirror wide for 36 cancers in 185 countries. CA Cancer J Clin 2018;
the effect of other nutrients, such as vitamin B6 or di- 68:394–424.
etary fiber, which share common dietary sources with 2. Kyrgiou M, Kalliala I, Markozannes G, et al. Adiposity and cancer
potassium. at major anatomical sites: umbrella review of the literature. BMJ
Strengths of this study include its large size and long 2017;356:j477.
follow-up duration and the DWAS approach that 3. Rezende LFM, Sa TH, Markozannes G, et al. Physical activity
and cancer: an umbrella review of the literature including 22
involved a comprehensive assessment of foods and nu-
major anatomical sites and 770 000 cancer cases. Br J Sports
trients while accounting for multiplicity of tests and
Med 2018;52:826–833.
replication of findings in an independent cohort. Another
4. World Cancer Research Fund, American Institute for Cancer
strength was the ability to explore associations according Research. Diet, Nutrition, Physical Activity, and the Prevention
to different anatomical subsites as well as by sex. The of Colorectal Cancer. Continuous Update Project Report. 2018.
primary limitation was that the analysis relied on a sin- Available at: https://www.wcrf.org/wp-content/uploads/2021/
gle dietary assessment at recruitment, not allowing to 02/Colorectal-cancer-report.pdf. Accessed December 1, 2020.
capture potential changes in dietary habits over time. In 5. Papadimitriou N, Muller D, van den Brandt PA, et al. A nutrient-
addition, intercorrelations between dietary exposures wide association study for risk of prostate cancer in the
European Prospective Investigation into Cancer and Nutrition 22. Santarelli RL, Pierre F, Corpet DE. Processed meat and colo-
and the Netherlands Cohort Study. Eur J Nutr 2020; rectal cancer: a review of epidemiologic and experimental evi-
59:2929–2937. dence. Nutr Cancer 2008;60:131–144.
6. Riboli E, Hunt KJ, Slimani N, et al. European Prospective 23. Norat T, Bingham S, Ferrari P, et al. Meat, fish, and colorectal
Investigation into Cancer and Nutrition (EPIC): study pop- cancer risk: the European Prospective Investigation into cancer
ulations and data collection. Public Health Nutr 2002; and nutrition. J Natl Cancer Inst 2005;97:906–916.
5:1113–1124. 24. Zheng Y, Li Y, Satija A, et al. Association of changes in red meat
7. van den Brandt PA, Goldbohm RA, van ’t Veer P, et al. A large- consumption with total and cause specific mortality among US
scale prospective cohort study on diet and cancer in the women and men: two prospective cohort studies. BMJ 2019;
Netherlands. J Clin Epidemiol 1990;43:285–295. 365:l2110.
8. Goldbohm RA, van den Brandt PA, Brants HA, et al. Validation 25. Meng Y, Sun J, Yu J, et al. Dietary intakes of calcium, iron,
of a dietary questionnaire used in a large-scale prospective magnesium, and potassium elements and the risk of colorectal
cohort study on diet and cancer. Eur J Clin Nutr 1994; cancer: a meta-analysis. Biol Trace Elem Res 2019;
48:253–265. 189:325–335.
9. Prentice RL. A case-cohort design for epidemiologic cohort 26. Mori H, Morishita Y, Shinoda T, et al. Preventive effect of
studies and disease prevention trials. Biometrika 1986; magnesium hydroxide on carcinogen-induced large bowel
73:1–11. carcinogenesis in rats. Basic Life Sci 1993;61:111–118.
10. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a 27. van den Brandt PA, Smits KM, Goldbohm RA, et al. Magnesium
practical and powerful approach to multiple testing. J R Stat Soc intake and colorectal cancer risk in the Netherlands Cohort
Series B Stat Methodol 1995;57:289–300. Study. Br J Cancer 2007;96:510–513.
11. Wei EK, Colditz GA, Giovannucci EL, et al. A comprehensive 28. Leenders M, Leufkens AM, Siersema PD, et al. Plasma and di-
model of colorectal cancer by risk factor status and subsite etary carotenoids and vitamins A, C and E and risk of colon and
using data from the Nurses’ Health Study. Am J Epidemiol 2017; rectal cancer in the European Prospective Investigation into
185:224–237. Cancer and Nutrition. Int J Cancer 2014;135:2930–2939.
12. Murphy N, Ward HA, Jenab M, et al. Heterogeneity of colorectal 29. Malila N, Virtamo J, Virtanen M, et al. Dietary and serum alpha-
cancer risk factors by anatomical subsite in 10 European tocopherol, beta carotene and retinol, and risk for colorectal
countries: a multinational cohort study. Clin Gastroenterol cancer in male smokers. Eur J Clin Nutr 2002;56:615–621.
Hepatol 2019;17:1323–1331.e6. 30. Newman AC, Maddocks ODK. One-carbon metabolism in can-
13. Seitz HK, Stickel F. Molecular mechanisms of alcohol-mediated cer. Br J Cancer 2017;116:1499–1504.
carcinogenesis. Nat Rev Cancer 2007;7:599–612. 31. Zschäbitz S, Cheng T-YD, Neuhouser ML, et al. B vitamin in-
14. Cornish AJ, Law PJ, Timofeeva M, et al. Modifiable pathways for takes and incidence of colorectal cancer: results from the
colorectal cancer: a Mendelian randomisation analysis. Lancet Women’s Health Initiative Observational Study cohort. Am J Clin
Gastroenterol Hepatol 2020;5:55–62. Nutr 2013;97:332–343.
15. Newmark HL, Wargovich MJ, Bruce WR. Colon cancer and di- 32. Larsson SC, Orsini N, Wolk A. Vitamin B6 and risk of colorectal
etary fat, phosphate, and calcium: a hypothesis. J Natl Cancer cancer: a meta-analysis of prospective studies. JAMA 2010;
Inst 1984;72:1323–1325. 303:1077–1083.
16. Moskal A, Freisling H, Byrnes G, et al. Main nutrient patterns and 33. Key TJ, Appleby PN, Masset G, et al. Vitamins, minerals,
colorectal cancer risk in the European Prospective Investigation essential fatty acids and colorectal cancer risk in the United
into Cancer and Nutrition study. Br J Cancer 2016; Kingdom Dietary Cohort Consortium. Int J Cancer 2012;
115:1430–1440. 131:E320–E325.
17. Welch AA, Fransen H, Jenab M, et al. Variation in intakes of 34. Kaaks R, Slimani N, Riboli E. Pilot phase studies on the accu-
calcium, phosphorus, magnesium, iron and potassium in 10 racy of dietary intake measurements in the EPIC project: overall
countries in the European Prospective Investigation into Cancer evaluation of results. European Prospective Investigation into
and Nutrition study. Eur J Clin Nutr 2009;63(Suppl Cancer and Nutrition. Int J Epidemiol 1997;26(Suppl 1):
4):S101–S121. S26–S36.
18. Larsson SC, Mason AM, Kar S, et al. Genetically proxied milk
consumption and risk of colorectal, bladder, breast, and pros- Reprint requests
tate cancer: a two-sample Mendelian randomization study. BMC Address requests for reprints to: Konstantinos K. Tsilidis, PhD, Department of
Med 2020;18:370. Epidemiology and Biostatistics, Imperial College London, St Mary’s Campus,
London, W2 1PG, United Kingdom. e-mail: k.tsilidis@imperial.ac.uk; fax: (þ30)
19. Tsilidis KK, Papadimitriou N, Dimou N, et al. Genetically pre- 26510 07853.
dicted circulating concentrations of micronutrients and risk of
colorectal cancer among individuals of European descent: a Acknowledgments
The authors are grateful to all the participants who have been part of the project
Mendelian randomization study. Am J Clin Nutr 2021 Mar 19 [E- and to the EPIC study team at the University of Cambridge who has enabled
pub ahead of print]. this research. The EPIC study data can be accessed via an application to the
EPIC Steering Committee (https://epic.iarc.fr/access/index.php). Further in-
20. Bonovas S, Fiorino G, Lytras T, et al. Calcium supplementation formation is available from the corresponding author upon request.
for the prevention of colorectal adenomas: a systematic review
and meta-analysis of randomized controlled trials. World J Conflicts of interest
Gastroenterol 2016;22:4594–4603. The authors disclose no conflicts.
21. Cross AJ, Sinha R. Meat-related mutagens/carcinogens in the
Funding
etiology of colorectal cancer. Environ Mol Mutagen 2004; This work was supported by the World Cancer Research Fund International
44:44–55. Regular Grant Programme (WCRF 2014/1180 to Konstantinos K. Tsilidis). The
coordination of EPIC is financially supported by International Agency for Funds, Dutch ZON (Zorg Onderzoek Nederland), World Cancer Research
Research on Cancer and also by the Department of Epidemiology and Fund (WCRF), Statistics Netherlands (The Netherlands); Health Research
Biostatistics, School of Public Health, Imperial College London which has Fund (FIS)–Instituto de Salud Carlos III (ISCIII), Regional Governments of
additional infrastructure support provided by the National Institute for Health Andalucía, Asturias, Basque Country, Murcia and Navarra, and the Catalan
Research Imperial Biomedical Research Centre. The national cohorts are Institute of Oncology (ICO) (Spain); Swedish Cancer Society, Swedish
supported by the Danish Cancer Society (Denmark); Ligue Contre le Cancer, Research Council and County Councils of Skåne and Västerbotten (Sweden);
Institut Gustave Roussy, Mutuelle Générale de l’Education Nationale, Institut and Cancer Research UK (14136 to EPIC-Norfolk; C8221/A29017 to EPIC-
National de la Santé et de la Recherche Médicale (INSERM) (France); German Oxford) and Medical Research Council (1000143 to EPIC-Norfolk; MR/
Cancer Aid, German Cancer Research Center (DKFZ), German Institute of M012190/1 to EPIC-Oxford) (United Kingdom). Where authors are identified
Human Nutrition Potsdam-Rehbruecke (DIfE), Federal Ministry of Education as personnel of the International Agency for Research on Cancer/World
and Research (BMBF) (Germany); Associazione Italiana per la Ricerca sul Health Organization, the authors alone are responsible for the views
Cancro-AIRC-Italy, Compagnia di SanPaolo and National Research Council expressed in this article and they do not necessarily represent the decisions,
(Italy); Dutch Ministry of Public Health, Welfare and Sports (VWS), policy, or views of the International Agency for Research on Cancer/World
Netherlands Cancer Registry (NKR), LK Research Funds, Dutch Prevention Health Organization.
- 2021 Diet-Wide Association Study for Risk of CRC 10.e1
Supplementary Methods was used to calculate standardized nutrient intakes.3 In

total, 92 dietary factors (63 foods and 29 nutrients) that
Study populations were available in at least 8 of the 9 countries were
included in the current analysis.
The European Prospective Investigation into Cancer In NLCS, information on dietary intake over the pre-
and Nutrition (EPIC) is a large European multicenter ceding 12 months was assessed at baseline using a semi-
prospective cohort that consists of 521,324 participants, quantitative 150-item food frequency questionnaire,
mostly between 35 and 70 years of age, recruited be- which has been validated and tested for reproduc-
tween 1992 and 2000 from 23 centers across 10 Euro- ibility.4,5 The Dutch food composition table was used for
pean countries, namely Denmark, France, Germany, the conversion of the data obtained from the food
Greece, Italy, the Netherlands, Norway, Spain, Sweden, questionnaires to nutrient intakes.6
and the United Kingdom.1 Of the 491,992 participants
with complete data on length of follow-up and without a Identification of colorectal cancer cases
cancer diagnosis before the baseline assessment, 6259
were excluded because they did not complete the life- In EPIC, incident colorectal cancer (CRC) cases were
style or dietary questionnaires at baseline, 9573 partic- identified by record linkage with population-based can-
ipants were excluded due to extreme values (top or cer registries in Denmark, Italy, the Netherlands, Nor-
bottom 1%) of the energy intake-to-energy requirement way, Spain, Sweden, and the United Kingdom, or a
ratio, and 64,671 were further excluded due to missing combination of registries, insurance records, and active
values in any of the covariates of interest (diabetes his- follow-up of the study participants or their relatives in
tory: 38,972; level of education: 16,931; smoking status: France, Germany, and Naples (Italy). The International
9678; physical activity: 8824). Data from Greece were Classification of Diseases–Tenth Revision and the second
also excluded from the current analysis, leaving 386,792 revision of the International Classification of Diseases for
participants (71% women) in the final analytical sample. Oncology were used to determine CRC cases (codes C18-
All participants gave written informed consent while C20).
approval for the study was obtained from the ethical In NLCS, incident CRC cases were identified by record
review boards of the International Agency for Research linkage to the Netherlands Cancer Registry and the Dutch
on Cancer and all local institutions in the participating National Pathology Registry record.7 CRC cases were
countries. classified according to International Classification of
The Netherlands Cohort Study (NLCS) is a prospec- Diseases for Oncology–Third Edition (codes C18–C20).
tive cohort study of 120,852 participants, between 55 In addition to overall CRC, we also examined associ-
and 69 years of age and recruited in 1986 from 204 ations for the following subsites: proximal colon
computerized population registries across the (C18.0–18.5), distal colon (C18.6–18.7), and rectum
Netherlands.2 The NLCS used a case-cohort approach for (C19–C20).
efficiency reasons, whereby a subcohort of 5000 partic-
ipants was selected at random immediately after base-
line.2 Of the 5000 participants, 3893 were included in the Statistical analyses
current analysis after excluding 226 with prevalent
cancer at recruitment, 690 with incomplete or inconsis- In EPIC, separate Cox proportional hazards regression
tent dietary data, and 191 participants with missing data models with age as the time scale were used to investi-
on confounders. NLCS was approved by the institutional gate the associations between each of the dietary factors
review boards of the Nederlandse Organisatie voor with CRC risk. Age at recruitment was set as the age at
Toegepast Natuurwetenschappehlijk Onderzoek Quality entry. Age at exit was defined either as the age at cancer
of Life research institute (Zeist, the Netherlands) and diagnosis or the age at death or age at the last follow-up,
Maastricht University (Maastricht, the Netherlands). whichever occurred first. In NLCS, given the case-cohort
design, Prentice-weighted Cox proportional hazards
regression models with robust standard error estimation
Assessment of dietary factors were implemented.8 In both EPIC and NLCS, the pro-
portionality of the hazard ratios was verified by exam-
In EPIC, consumption of foods over the last 12 ining the slope of the Schoenfeld residuals, and no
months was assessed at baseline using validated violations were found. Intakes of foods and nutrients
country-specific food questionnaires.1 In most countries were adjusted for energy intake using the residual
and centers the questionnaires were self-administered method and standardized prior to modeling.9 All of the
apart from Ragusa (Italy) and Spain, where in- models were adjusted for total energy intake (kcal,
terviewers were used. In Malmö (Sweden), a food record continuous); smoking status (never, former, current);
was used for cooked meals and a food frequency ques- body mass index (body mass index, <20, 20–22.9,
tionnaire was used for breakfast and foods consumed 23–24.9, 25–29.9, 30–34.9, 35 kg/m2); physical activity
between the main meals. The EPIC Nutrient Database (EPIC: Cambridge index [inactive, moderately inactive,
10.e2 Papadimitriou et al Clinical Gastroenterology and Hepatology Vol. -, No. -
moderately active, active], NLCS: nonoccupational phys- nucleotide polymorphisms or their proxies were avail-
ical activity [30, >30–60, >60–90, >90 min/d]); dia- able for 3361 participants. Nutrients were included in
betes history (no, yes); level of education (none/primary the interaction analyses as standardized continuous
school, technical/professional school, secondary school, variables and the same covariates as in the diet-wide
longer education), and family history of CRC (no, yes; in association study Cox proportional hazards regression
the NLCS only), and reflect associations per 1-SD in- models were used. P values were adjusted for multiple
crease in daily consumption. Additionally, all models comparisons using the Bonferroni correction based on
were further stratified by sex, age at recruitment (5-year the number of independent tests, with a corrected P
intervals), and in EPIC also by center in order to control value threshold at 3.4 10–5.
for center-specific differences like questionnaire design
and follow-up procedures.10
References
To account for multiple comparisons, the false dis- 1. Riboli E, Hunt KJ, Slimani N, et al. European Prospective
covery rate (FDR)–adjusted P values (or q values) were Investigation into Cancer and Nutrition (EPIC): study populations
estimated for each association analyzed using the and data collection. Public Health Nutr 2002;5:1113–1124.
sequential P-value approach proposed by Benjamini and 2. van den Brandt PA, Goldbohm RA, van ’t Veer P, et al. A large-
Hochberg.11 The dietary factors with an FDR <0.05 were scale prospective cohort study on diet and cancer in the
subsequently selected for replication in the NLCS, and Netherlands. J Clin Epidemiol 1990;43:285–295.
fixed effects meta-analysis was performed to combine 3. Slimani N, Deharveng G, Unwin I, et al. The EPIC nutrient
the results from the 2 cohorts when heterogeneity was database project (ENDB): a first attempt to standardize nutrient
low or moderate (P value for heterogeneity > .1 and/or databases across the 10 European countries participating in the
I2 50%). To further investigate the robustness of the EPIC study. Eur J Clin Nutr 2007;61:1037–1056.
associations that were replicated in the NLCS, a mutual 4. Goldbohm RA, van den Brandt PA, Brants HA, et al. Validation
adjustment model was used. Potential nonlinear associ- of a dietary questionnaire used in a large-scale prospective
ations were investigated using restricted cubic spline cohort study on diet and cancer. Eur J Clin Nutr 1994;
48:253–265.
models (5 knots). Nonlinearity was tested by using a
5. Goldbohm RA, van ’t Veer P, van den Brandt PA, et al. Repro-
likelihood ratio test comparing the model containing the
ducibility of a food frequency questionnaire and stability of di-
cubic spline terms with the model containing only the
etary habits determined from five annually repeated
linear term. measurements. Eur J Clin Nutr 1995;49:420–429.
Separate analyses for the FDR-significant dietary ex- 6. Nevo table: Dutch food composition table 1986-1987. The
posures were conducted in men and women and also by Hague, Netherlands: Voorlichtingsbureau voor de voeding,
anatomical subsite of CRC. For the FDR-significant foods 1986.
or nutrients in EPIC, the pairwise partial correlation 7. Van den Brandt PA, Schouten LJ, Goldbohm RA, et al. Devel-
coefficients were quantified, adjusting for age, sex and opment of a record linkage protocol for use in the Dutch Cancer
center, using Spearman’s rho (r). Additionally, the impact Registry for Epidemiological Research. Int J Epidemiol 1990;
of follow-up duration in the association of red and pro- 19:553–558.
cessed meat with CRC risk was investigated. All analyses 8. Prentice RL. A case-cohort design for epidemiologic cohort
were performed using R version 4.0.2 (R Foundation for studies and disease prevention trials. Biometrika 1986;73:1–11.
Statistical Computing, Vienna, Austria). 9. Willett W, Stampfer MJ. Total energy intake: implications for
epidemiologic analyses. Am J Epidemiol 1986;124:17–27.
10. Ferrari P, Day NE, Boshuizen HC, et al. The evaluation of the
Gene-Nutrient interactions
diet/disease relation in the EPIC study: considerations for the
calibration and the disease models. Int J Epidemiol 2008;
Potential multiplicative gene-nutrient interactions in 37:368–378.
EPIC were systematically investigated, between the food 11. Benjamini Y, Hochberg Y. Controlling the false discovery rate:
components that met the FDR threshold and known CRC- a practical and powerful approach to multiple testing. J R Stat
associated genetic variants from genome-wide associa- Soc Series B Stat Methodol 1995;57:289–300.
tion study.12 Of the approximately 100 genome-wide 12. Huyghe JR, Bien SA, Harrison TA, et al. Discovery of common
association study–identified single nucleotide poly- and rare genetic risk variants for colorectal cancer. Nat Genet
morphisms associated with CRC, data for 73 single 2019;51:76–87.
Supplementary Figure 1. Estimated hazard ratio of red meat (top panel) and processed meat (bottom panel) in relation to CRC
risk in EPIC, per cumulative year of follow-up. The y-axis shows the estimated HR for each dietary factor for 1-SD increase in
daily consumption. The models were adjusted for total energy intake (kcal, continuous); smoking status (never, former, cur-
rent); body mass index (<20, 20–22.9, 23–24.9, 25–29.9, 30–34.9, 35 kg/m2); physical activity (inactive, moderately inactive,
moderately active, active); diabetes history (no, yes); education status (none/primary, technical/professional, secondary,
longer); and stratified by sex, age at recruitment (5-year intervals), and center.
Supplementary Figure 2.
Comparison of 2 separate
modeling approaches.
Solid lines represent the
fitted regression lines and
the shaded area represent
the 95% confidence in-
tervals. The Cox regres-
sion using a linear
prediction of alcohol intake
is represented by the red
color and shade and the
Cox regression that further
includes cubic splines (at 5
nots) is represented by the
light blue color and shade,
conditioned at average
values of covariates and
confounders.
Supplementary Table 1. Baseline Demographic Characteristics in EPIC and the NLCS Subcohort
EPIC NLCS
Total Noncases Cases Total Noncases Cases
Total 386,792 381,723 5069 7496 3731 3765

Sex
Male 112,788 (29.2) 110,597 (29.0) 2191 (43.2) 4023 (53.7) 1871 (50.1) 2152 (57.2)
Female 274,004 (70.8) 271,126 (71.0) 2878 (56.8) 3473 (46.3) 1860 (49.9) 1613 (42.8)
Age at recruitment
<40 y 47,425 (12.3) 47,331 (12.4) 94 (1.9) — — —
40–44.9 y 52,795 (13.6) 52,548 (13.8) 247 (4.9) — — —
45–49.9 y 68,307 (17.7) 67,778 (17.8) 529 (10.4) — — —
50–54.9 y 88,025 (22.8) 86,807 (22.7) 1218 (24.0) — — —
55–59.9 y 64,757 (16.7) 63,557 (16.7) 1200 (23.7) 2718 (36.3) 1446 (38.8) 1272 (33.8)
60–64.9 y 49,840 (12.9) 48,519 (12.7) 1321 (26.1) 2658 (35.5) 1273 (34.1) 1385 (36.8)
65–69.9 y 12,218 (3.2) 11,884 (3.1) 334 (6.6) 2120 (28.3) 1012 (27.1) 1108 (29.4)
70–74.9 y 3011 (0.8) 2900 (0.8) 111 (2.2)
>75 y 414 (0.1) 399 (0.1) 15 (0.3)
Smoking status
Never 194,087 (50.2) 191,990 (50.3) 2097 (41.4) 2474 (33.0) 1303 (34.9) 1171 (31.1)
Former 103,942 (26.9) 102,268 (26.8) 1674 (33) 2991 (39.9) 1364 (36.6) 1627 (43.2)
Current 88,763 (22.9) 87,465 (22.9) 1298 (25.6) 2031 (27.1) 1064 (28.5) 967 (25.7)
Educationa
None/primary School 112,507 (29.1) 110,607 (29.0) 1900 (37.5) 2040 (27.2) 1038 (27.8) 1002 (26.6)
Technical/professional 87,563 (22.6) 86,290 (22.6) 1273 (25.1) 1599 (21.3) 798 (21.4) 801 (21.3)
school
Secondary school 86,072 (22.3) 85,224 (22.3) 848 (16.7) 2697 (36.0) 1349 (36.2) 1348 (35.8)
Longer education (incl. 100,650 (26.0) 99,602 (26.1) 1048 (20.7) 1160 (15.5) 546 (14.6) 614 (16.3)
university degree)
BMI
<20 kg/m2 26,550 (6.9) 26,385 (6.9) 165 (3.3) 243 (3.2) 139 (3.7) 104 (2.8)
20–22.9 kg/m2 99,036 (25.6) 98,100 (25.7) 936 (18.5) 1528 (20.4) 783 (21.0) 745 (19.8)
23–24.9 kg/m2 81,112 (21.0) 80,111 (21.0) 1001 (19.7) 2231 (29.8) 1129 (30.3) 1102 (29.3)
25–29.9 kg/m2 131,871 (34.1) 129,747 (34.0) 2124 (41.9) 3037 (40.5) 1445 (38.7) 1592 (42.3)
30–34.9 kg/m2 38,125 (9.9) 37,464 (9.8) 661 (13.0) 403 (5.4) 208 (5.6) 195 (5.2)
35 kg/m2 10,098 (2.6) 9916 (2.6) 182 (3.6) 54 (0.7) 27 (0.7) 27 (0.7)
Physical activityb
Inactive 72,301 (18.7) 71,167 (18.6) 1134 (22.4) 1546 (20.6) 765 (20.5) 781 (20.7)
Moderately inactive 132,369 (34.2) 130,641 (34.2) 1728 (34.1) 2350 (31.4) 1172 (31.4) 1178 (31.3)
Moderately active 106,613 (27.6) 105,417 (27.6) 1196 (23.6) 1623 (21.7) 798 (21.4) 825 (21.9)
Active 75,509 (19.5) 74,498 (19.5) 1011 (19.9) 1977 (26.4) 996 (26.7) 981 (26.1)
Diabetes
No 376,678 (97.4) 371,832 (97.4) 4846 (95.6) 7271 (97.0) 3608 (96.7) 3663 (97.3)
Yes 10,114 (2.6) 9891 (2.6) 223 (4.4) 225 (3.0) 123 (3.3) 102 (2.7)
Family history of CRC
No — — — 6935 (92.5) 3527 (94.5) 3408 (90.5)
Yes — — — 561 (7.5) 204 (5.5) 357 (9.5)
Values are n or n (%).

BMI, body mass index; CRC, colorectal cancer; EPIC, European Prospective Investigation into Cancer and Nutrition; NLCS, Netherlands Cohort Study.
a
The 4 educational level categories in the NLCS were formed as follows: primary school; lower vocational school; secondary, medium vocational; higher voca-
tional, university.
b
The 4 physical activity categories in the NLCS were based on nonoccupational physical activity and formed as follows: 30 min/d; >30 to 60 min/d; >60 to 90
min/d; >90 min/d.
Supplementary Table 2. HRs and 95% CIs for the Supplementary Table 2. Continued
Association of 92 Food and
Nutrient Intakes in Relation to Dietary Variable HR (95% CI)a P Value FDR SD
Colorectal Cancer Risk in the EPIC
Study Citrus fruits 0.98 (0.95–1.02) .302 0.497 62.5
e
Leafy vegetables 0.98 (0.94–1.02) .347 0.507 41.1
Dietary Variable HR (95% CI)a P Value FDR SD f
Cabbage 0.98 (0.94–1.02) .367 0.507 30.9
FDR-significant
Stalk vegetables, 1.02 (0.98–1.05) .355 0.507 12.8
associations
sproutse
Alcohol 1.07 (1.04–1.10) <.001 <0.001 17.8
Grain and pod 0.99 (0.94–1.03) .563 0.73 12.7
Spirits b
1.03 (1.01–1.06) .002 0.013 12.2 vegetablese
Wine 1.05 (1.02–1.08) .001 0.008 133 Grapesg 1.01 (0.98–1.04) .59 0.753 15.3
Beer, cider 1.07 (1.04–1.09) <.001 <0.001 244 Potatoes 1.01 (0.98–1.04) .607 0.755 74.8
h
Soft drinks 1.04 (1.02–1.07) .002 0.013 165.8 Onion, garlic 0.99 (0.95–1.03) .605 0.755 14.7
i
Milk 0.96 (0.93–0.99) .008 0.041 208.3 Fruiting vegetables 1.00 (0.97–1.04) .923 0.955 52.8
Cheese 0.95 (0.92–0.99) .007 0.041 34.2 Fruit and vegetables 1.00 (0.96–1.03) .935 0.956 115.3
juice
Calcium 0.92 (0.89–0.95) <.001 <0.001 334.5
Dairy products
Phosphorous 0.92 (0.89–0.94) <.001 <0.001 273.6
Yoghurt 0.98 (0.95–1.01) .176 0.371 89
Magnesium 0.95 (0.91–0.99) .009 0.044 82.3
Meat and meat products
Potassium 0.95 (0.92–0.98) .003 0.02 717.2
Processed meat 1.04 (1.00–1.08) .026 0.092 31.5
Riboflavin 0.94 (0.91–0.98) .001 0.011 0.6
j
Liver 1.02 (0.99–1.05) .259 0.482 4.7
Vitamin B6 0.95 (0.92–0.99) .006 0.035 0.4
Red meat 1.02 (0.98–1.05) .369 0.507 36.2
Beta carotene 0.95 (0.92–0.98) .002 0.015 2780.8
k
Beef 0.98 (0.95–1.02) .34 0.507 19.2
Fruits 0.96 (0.92–0.99) .008 0.041 178.1
i
Offal 1.01 (0.97–1.04) .664 0.793 6.2
Dietary fiber 0.93 (0.90–0.96) <.001 <0.001 6.2
Poultry 1.00 (0.97–1.03) .952 0.962 19.8
Nonwhite bread 0.93 (0.90–0.97) .001 0.008 72.9
l
Lamb 1.00 (0.97–1.04) .991 0.991 7.9
Banana 0.96 (0.93–0.99) .01 0.048 36.9
Fish and seafood
Pork 1.06 (1.03–1.09) <.001 0.001 17.5
Fish 0.96 (0.93–1.00) .033 0.109 31
Total proteins 0.94 (0.91–0.97) <.001 0.002 15.5
Fatty fishm 0.97 (0.94–1.00) .034 0.109 14.5
Non–FDR-significant associations
n
Fish products 0.97 (0.93–1.00) .045 0.137 8.7
Foods and food groups
o
Crustaceans 1.01 (0.98–1.05) .347 0.507 6.1
Breads and cereals
Lean fish p
0.99 (0.95–1.02) .451 0.592 23.4
White bread 1.05 (1.01–1.09) .012 0.052 73.6
Eggs
Breakfast cerealsc 0.97 (0.94–1.00) .065 0.181 42.8
Eggsq 1.04 (1.01–1.07) .018 0.064 17.5
Bread 0.98 (0.94–1.01) .23 0.46 79.7
Dietary fats
Crispbread, rusks 0.99 (0.96–1.03) .688 0.801 17.1
Mayonnaiser 1.02 (0.98–1.06) .34 0.507 5.7
Pasta, rice, other 1.01 (0.97–1.05) .679 0.801 65.5
grains Margarine 1.01 (0.97–1.05) .621 0.762 16.2
Salty biscuits, 1.00 (0.97–1.03) .919 0.955 6.4 Butter 1.00 (0.97–1.03) .92 0.955 8.6
crackers
Margarine 1.00 (0.97–1.03) .899 0.955 13.1
Fruits and vegetables (vegetables)
Root vegetables 0.96 (0.93–0.99) .016 0.062 30.2 Nuts, seeds, and legumes
Apple, pear 0.97 (0.95–1.00) .076 0.205 85.4 Legumess 0.94 (0.90–0.99) .015 0.061 26.1
d Nuts 0.98 (0.95–1.02) .265 0.482 8.4
Berries 0.97 (0.94–1.00) .085 0.216 12.5
e Confectionery
Stone fruits 0.98 (0.94–1.01) .214 0.437 45.6
e Ice cream 1.03 (1.00–1.06) .029 0.1 11.3
Mushrooms 1.02 (0.98–1.06) .267 0.482 9
Supplementary Table 2. Continued Supplementary Table 2. Continued

a
Dietary Variable HR (95% CI) P Value FDR SD Dietary Variable HR (95% CI)a P Value FDR SD
Confectionery 1.02 (0.99–1.05) .158 0.354 12.4 Vitamin E 0.98 (0.95–1.02) .375 0.508 4.4
(nonchocolate)t
Thiamine 0.98 (0.94–1.03) .421 0.561 0.4
Sugars (Sugar, 0.98 (0.95–1.01) .25 0.482 20.3
Vitamin D 0.99 (0.96–1.03) .645 0.78 3.5
honey,
jam, and syrup) Minerals
Cream puddings/ 0.98 (0.95–1.01) .29 0.494 23.3 Iron 0.98 (0.94–1.01) .12 0.291 2.6
dessertsu
Dry cakes, biscuitsf 0.98 (0.95–1.02) .313 0.506 12.3
CI, confidence interval; EPIC, European Prospective Investigation into Cancer
Cakes, sweets 0.98 (0.95–1.02) .344 0.507 38.5 and Nutrition; FDR, false discovery rate; HR, hazard ratio.
a
(non–milk All dietary factors entered the models as standardized continuous variables
and reflect associations per 1-SD increase in daily consumption. Nutrient in-
based)
takes were adjusted for total energy intake using the regression residual
Chocolate 1.00 (0.96–1.03) .818 0.907 13.6 method. The models were adjusted for total energy intake (kcal, continuous);
smoking status (never, former, current); body mass index <20, 20–22.9,
Beverages (nonalcoholic) 23–24.9, 25–29.9, 30–34.9, 35 kg/m2); physical activity (inactive, moderately
inactive, moderately active, active); diabetes history (no, yes); education status
Teai 1.00 (0.96–1.03) .809 0.907 304.1 (none/primary, technical/professional, secondary, longer [including university]).
They were further stratified by age at recruitment (<40, 40–44.9, 45–49.9,
Coffee 1.00 (0.96–1.03) .8 0.907 375.7
50–54.9, 55–59.9, 60–64.9, 65–69.9, 70–74.9, 75 years), sex, and recruitment
Alcohol and alcoholic center.
b
beverages Intake of spirits was missing for participants from Italy and Norway (9.2%
missing across EPIC).
Fortified winesv 1.00 (0.97–1.02) .768 0.884 15.8 c
Intake of breakfast cereals was missing for participants from Italy (10.2%
Combination foods d
Intake of berries was missing for participants from Norway and the United
Kingdom (16.6% missing across EPIC).
Soupw 1.02 (0.98–1.06) .296 0.495 79.3 e
Intake for mushrooms, leafy vegetables, stone fruits, stalk vegetables, pod
Condiments and sauces vegetables was missing for participants from Norway and Sweden (12.6%
Saucesx 1.00 (0.97–1.04) .865 0.948 18.9 f
Intake of cabbage and biscuits was missing for participants from Sweden
(6.1% missing across EPIC).
Nutrients g
Intake of grapes was missing for participants from Norway and Sweden
(26.1% missing across EPIC).
Carbohydrates h
Intake for onion and garlic was missing for participants from France, Norway,
Carbohydrates 0.97 (0.94–1.00) .061 0.176 36.9 and Sweden (28.4% missing across EPIC).
i
Intake of offal, tea, and fruiting vegetables was missing for participants from
Total sugars 0.98 (0.95–1.01) .151 0.348 32.3 Norway (6.4% missing across EPIC).
j
Intake of liver was missing for participants from the Netherlands, Norway, and
Starch 0.98 (0.95–1.01) .173 0.371 32.6 Sweden (20.7% missing across EPIC).
k
Intake of beef was missing for participants from Sweden (6.1% missing across
Dietary fats
EPIC).
l
Saturated fats 0.97 (0.94–1.00) .078 0.206 7.7 Intake of lamb was missing for participants from the Netherlands, Italy and
Sweden (22.9% missing across EPIC).
Total fats 0.98 (0.95–1.00) .09 0.223 13.5 m
Intake of fatty fish was missing for participants from Germany (6.6% missing
across EPIC).
Monounsaturated 0.98 (0.94–1.01) .177 0.371 7.3 n
Intake of fish products was missing for participants from France and Italy
fats (24.4% missing across EPIC).
o
Intake of crustaceans was missing for participants from Germany (12.5%
Fats (animal) 0.98 (0.95–1.01) .261 0.482 13
Cholesterol 1.02 (0.99–1.05) .28 0.486 115.8
p
Intake of lean fish was missing for participants from Germany, Italy and
q
Fats (plant) 0.98 (0.94–1.02) .359 0.507 13.1 Intake of egg was missing for participants from Sweden (6.1% missing across
EPIC).
Polyunsaturated 1.00 (0.97–1.03) .924 0.955 4.5 r
Intake of mayonnaise was missing for participants from Italy, Norway, and
fats Sweden (13.9% missing across EPIC).
s
Intake of legumes was missing for participants from Denmark and Norway
Proteins (20.0% missing across EPIC).
t
Protein (animal) 0.96 (0.93–0.99) .016 0.062 18.4 Intake of confectionary was missing for participants from Germany and Nor-
way (19.0% missing across EPIC).
u
Protein (plant) 0.96 (0.93–1.00) .055 0.163 7.8 Intake of cream puddings/desserts was missing for participants from Italy and
Vitamins v
Intake of fortified wines was missing for participants from Italy, Norway, and
Vitamin C 0.98 (0.95–1.01) .151 0.348 60.7 w
Intake of soup was missing for participants from Denmark, Italy, and Norway
Vitamin B12 0.98 (0.95–1.01) .28 0.486 3.6 (21.2% missing across EPIC).
x
Intake of sauces was missing for participants from Italy (1.3% missing across
Retinol, units 1.01 (0.99–1.04) .365 0.507 694.9 EPIC).
Supplementary Table 3. HRs and 95% CIs for the Association of the 20 Food and Nutrient Intakes With Colorectal Cancer
Risk by Tumor Location (Colon vs Rectal) in the EPIC Study
Dietary Variables Colon, HR (95% CI)a Rectum, HR (95% CI)a P Value for Heterogeneity
Alcohol 1.06 (1.02–1.10) 1.09 (1.04–1.14) .309

b
Spirits 1.02 (0.99–1.05) 1.06 (1.02–1.09) .113
Wine 1.04 (1.00–1.07) 1.07 (1.02–1.12) .278
Beer, cider 1.06 (1.02–1.09) 1.07 (1.04–1.11) .509
Soft drinks 1.04 (1.01–1.08) 1.04 (1.00–1.09) .988
Milk 0.96 (0.92–0.99) 0.96 (0.91–1.01) .946
Cheese 0.96 (0.92–1.01) 0.94 (0.88–1.00) .491
Calcium 0.93 (0.89–0.96) 0.91 (0.86–0.96) .552
Phosphorous 0.91 (0.87–0.95) 0.92 (0.87–0.97) .790
Magnesium 0.91 (0.87–0.96) 1.01 (0.95–1.08) .011
Potassium 0.92 (0.88–0.96) 1.01 (0.95–1.07) .008
Riboflavin 0.92 (0.88–0.97) 0.96 (0.90–1.02) .344
Vitamin B6 0.91 (0.87–0.95) 1.02 (0.96–1.08) .002
Beta carotene 0.95 (0.91–0.98) 0.96 (0.91–1.01) .602
Fruits 0.95 (0.91–0.99) 0.97 (0.92–1.03) .569
Dietary fiber 0.92 (0.88–0.95) 0.95 (0.90–1.00) .306
Nonwhite bread 0.93 (0.88–0.98) 0.95 (0.89–1.01) .669
Banana 0.94 (0.90–0.98) 1.00 (0.95–1.06) .041
Pork 1.06 (1.01–1.10) 1.07 (1.02–1.12) .686
Total proteins 0.93 (0.89–0.97) 0.95 (0.90–1.01) .409
CI, confidence interval; EPIC, European Prospective Investigation into Cancer and Nutrition; HR, hazard ratio.
a
All dietary factors entered the models as standardized continuous variables and reflect associations per 1-SD increase in daily consumption. Nutrient intakes
were adjusted for total energy intake using the regression residual method. The models were adjusted for total energy intake (kcal, continuous); smoking status
(never, former, current); body mass index <20, 20–22.9, 23–24.9, 25–29.9, 30–34.9, 35 kg/m2); physical activity (inactive, moderately inactive, moderately active,
active); diabetes history (no, yes); education status (none/primary, technical/professional, secondary, longer [including university]). They were further stratified by
age at recruitment (<40, 40–44.9, 45–49.9, 50–54.9, 55–59.9, 60–64.9, 65–69.9, 70–74.9, 75 years), sex, and recruitment center.
b
Intake of spirits was missing for participants from Italy and Norway (9.2% missing across EPIC).
Supplementary Table 4. HRsa and 95% CIs for the Association of the 20 Food and Nutrient Intakes With Colon Cancer Risk
by Tumor Location (Proximal vs Distal) in the EPIC Study
Dietary Variables Proximal, HR (95% CI)a Distal, HR (95% CI)a P Value for Heterogeneity
Alcohol 1.01 (0.96–1.07) 1.11 (1.05–1.16) .015

b
Spirits 1.02 (0.98–1.07) 1.00 (0.96–1.05) .564
Wine 1.00 (0.95–1.06) 1.07 (1.02–1.12) .087
Beer, cider 1.04 (0.99–1.09) 1.08 (1.03–1.12) .298
Soft drinks 1.02 (0.97–1.08) 1.06 (1.01–1.11) .311
Milk 0.97 (0.92–1.02) 0.96 (0.91–1.02) .931
Cheese 0.98 (0.92–1.05) 0.93 (0.87–0.99) .245
Calcium 0.94 (0.89–0.99) 0.91 (0.86–0.97) .432
Phosphorous 0.93 (0.87–0.98) 0.90 (0.85–0.96) .546
Magnesium 0.96 (0.89–1.03) 0.88 (0.82–0.95) .138
Potassium 0.94 (0.88–1.00) 0.92 (0.86–0.98) .599
Riboflavin 0.95 (0.89–1.02) 0.90 (0.84–0.97) .309
Vitamin B6 0.90 (0.85–0.96) 0.94 (0.88–1.01) .366
Beta carotene 0.94 (0.88–0.99) 0.97 (0.91–1.02) .431
Fruits 0.95 (0.89–1.01) 0.97 (0.91–1.03) .659
Dietary fiber 0.94 (0.88–0.99) 0.91 (0.86–0.96) .457
Nonwhite bread 0.95 (0.88–1.02) 0.93 (0.86–1.00) .643
Banana 0.91 (0.86–0.97) 0.98 (0.92–1.04) .128
Pork 1.03 (0.97–1.09) 1.08 (1.02–1.14) .263
Total proteins 0.92 (0.86–0.98) 0.93 (0.88–0.99) .702
a
b
Risk by Sex (Men vs Women) in the EPIC Study
Dietary Variables Men, HR (95% CI)a Women, HR (95% CI)a P Value for Heterogeneity
Alcohol 1.12 (1.08–1.16) 1.03 (0.99–1.07) .002

b
Spirits 1.05 (1.03–1.07) 0.98 (0.93–1.03) .010
Wine 1.04 (1.00–1.07) 1.06 (1.02–1.12) .386
Beer, cider 1.07 (1.05–1.10) 1.01 (0.93–1.10) .220
Soft drinks 1.03 (0.99–1.07) 1.06 (1.02–1.10) .376
Milk 0.96 (0.92–1.00) 0.97 (0.93–1.01) .777
Cheese 0.95 (0.90–1.00) 0.95 (0.91–1.00) .866
Calcium 0.91 (0.86–0.95) 0.93 (0.90–0.97) .407
Phosphorous 0.91 (0.86–0.95) 0.92 (0.89–0.96) .621
Magnesium 0.89 (0.84–0.96) 0.98 (0.93–1.03) .033
Potassium 0.92 (0.88–0.98) 0.97 (0.93–1.01) .170
Riboflavin 0.95 (0.89–1.01) 0.94 (0.90–0.98) .789
Vitamin B6 0.97 (0.92–1.02) 0.94 (0.90–0.98) .404
Beta carotene 0.94 (0.88–0.99) 0.96 (0.93–1.00) .434
Fruits 0.95 (0.90–1.00) 0.96 (0.92–1.00) .688
Dietary fiber 0.88 (0.84–0.93) 0.96 (0.93–1.01) .006
Nonwhite bread 0.89 (0.84–0.94) 0.99 (0.94–1.05) .008
Banana 0.97 (0.92–1.01) 0.95 (0.91–0.99) .653
Pork 1.05 (1.01–1.09) 1.08 (1.03–1.14) .303
Total proteins 0.94 (0.89–0.99) 0.94 (0.90–0.98) .909
a
b
Risk in the EPIC and the NLCS Study
Dietary Variables EPIC Study, HR (95% CI)a NLCS Study, HRb (95% CI)a P Value for Heterogeneity
Alcohol 1.07 (1.04–1.10) 1.06 (1.01–1.12) .704

c
Spirits 1.03 (1.01–1.06) 1.06 (1.01–1.11) .350
Wine 1.05 (1.02–1.08) 1.02 (0.97–1.08) .389
Beer, cider 1.07 (1.04–1.09) 1.03 (0.98–1.08) .192
Soft drinks 1.04 (1.02–1.07) 0.96 (0.91–1.02) .009
Milk 0.96 (0.93–0.99) 0.93 (0.89–0.98) .245
Cheese 0.95 (0.92–0.99) 0.99 (0.94–1.04) .221
Calcium 0.92 (0.89–0.95) 0.94 (0.90–0.99) .494
Phosphorus 0.92 (0.89–0.94) 0.95 (0.90–1.00) .237
Magnesium 0.95 (0.91–0.99) 0.95 (0.90–1.00) .986
Potassium 0.95 (0.92–0.98) 0.98 (0.94–1.03) .300
Riboflavin 0.94 (0.91–0.98) 0.95 (0.90–1.00) .768
Vitamin B6 0.95 (0.92–0.99) 1.01 (0.97–1.07) .053
Beta carotene 0.95 (0.92–0.98) 0.96 (0.92–1.01) .795
Fruit 0.96 (0.92–0.99) 1.00 (0.95–1.05) .142
Fiber 0.93 (0.90–0.96) 0.99 (0.94–1.03) .021
Nonwhite bread 0.93 (0.90–0.97) 1.00 (0.95–1.05) .035
Bananas 0.96 (0.93–0.99) 1.02 (0.97–1.07) .038
Pork 1.06 (1.03–1.09) 1.00 (0.95–1.05) .040
Total protein 0.94 (0.91–0.97) 0.95 (0.90–1.00) .692
a
b
Multivariable analyses were stratified for age at baseline (55–59, 60–64, 65–69 years), sex, and adjusted for smoking status (never, ex, current), body mass index
(<20, 20–<23, 23–<25, 25–<30, 30–<35, 35 kg/m2), nonoccupational physical activity (30, >30–60, >60–90, >90 min/d), highest level of education (primary
school or lower vocational, secondary or medium vocational, and higher vocational or university), family history of colorectal cancer (no, yes), history of diabetes,
energy intake (kcal, continuous).
c
Supplementary Table 7. HRs and 95% CIs for the Association of the 20 Food and Nutrients With Colorectal Cancer Risk by
Tumor Location (Colon vs Rectal) in the NLCS Study
Dietary Variables Colon, HR (95% CI)a Rectum, HR (95% CI)a P Value for Heterogeneity
Alcohol 1.03 (0.98–1.09) 1.11 (1.04–1.20) .100

Spirits 1.05 (0.99–1.10) 1.08 (1.00–1.16) .544
Wine 1.01 (0.95–1.06) 1.05 (0.97–1.14) .435
Beer, cider 0.99 (0.94–1.05) 1.06 (1.00–1.14) .118
Soft drinks 0.97 (0.92–1.03) 0.96 (0.87–1.06) .858
Milk 0.94 (0.89–0.99) 0.95 (0.88–1.03) .827
Cheese 0.98 (0.93–1.04) 1.04 (0.96–1.13) .239
Calcium 0.95 (0.90–1.00) 0.99 (0.91–1.07) .403
Phosphorus 0.93 (0.89–0.99) 1.04 (0.96–1.12) .019
Magnesium 0.94 (0.89–0.99) 1.00 (0.93–1.08) .186
Potassium 0.96 (0.91–1.01) 1.08 (1.00–1.17) .014
Riboflavin 0.94 (0.89–1.00) 1.00 (0.92–1.08) .221
Vitamin B6 0.98 (0.93–1.03) 1.12 (1.04–1.21) .004
Beta carotene 0.93 (0.88–0.98) 1.02 (0.94–1.10) .057
Fruits 0.99 (0.94–1.05) 1.02 (0.94–1.10) .543
Fiber 0.97 (0.92–1.03) 1.02 (0.95–1.10) .287
Nonwhite bread 0.98 (0.93–1.04) 1.06 (0.98–1.14) .102
Bananas 1.02 (0.96–1.08) 1.04 (0.96–1.12) .695
Pork 0.98 (0.93–1.03) 1.03 (0.95–1.12) .314
Total protein 0.93 (0.88–0.99) 1.02 (0.94–1.11) .076
a
Supplementary Table 8. HRs and 95% CIs for the Association of the 20 Food and Nutrient Intakes With Colon Cancer Risk by
Tumor Location (Proximal vs Distal) in the NLCS Study
Dietary Variables Proximal, HR (95% CI)a Distal, HR (95% CI)a P Value for Heterogeneity
Alcohol 1.04 (0.97–1.11) 1.03 (0.96–1.10) .843

Spirits 1.05 (0.99–1.12) 1.04 (0.97–1.11) .837
Wine 1.01 (0.94–1.08) 1.00 (0.93–1.07) .843
Beer, cider 0.99 (0.92–1.06) 1.00 (0.93–1.07) .843
Soft drinks 0.97 (0.90–1.05) 0.98 (0.90–1.06) .858
Milk 0.94 (0.87–1.00) 0.95 (0.89–1.02) .831
Cheese 1.00 (0.93–1.07) 0.98 (0.91–1.06) .702
Calcium 0.96 (0.90–1.03) 0.95 (0.89–1.02) .831
Phosphorus 0.95 (0.89–1.01) 0.94 (0.87–1.00) .825
Magnesium 0.95 (0.89–1.02) 0.93 (0.87–1.00) .669
Potassium 0.97 (0.90–1.04) 0.96 (0.90–1.03) .837
Riboflavin 0.94 (0.88–1.01) 0.96 (0.90–1.03) .669
Vitamin B6 0.99 (0.92–1.06) 0.97 (0.91–1.05) .691
Beta carotene 0.96 (0.89–1.02) 0.89 (0.82–0.96) .154
Fruits 1.01 (0.94–1.08) 0.98 (0.91–1.05) .553
Fiber 0.98 (0.92–1.05) 0.97 (0.91–1.04) .831
Nonwhite bread 0.97 (0.91–1.05) 1.00 (0.93–1.07) .551
Bananas 1.01 (0.94–1.08) 1.03 (0.95–1.11) .712
Pork 0.97 (0.91–1.04) 0.99 (0.92–1.06) .681
Total protein 0.93 (0.86–1.00) 0.94 (0.87–1.01) .843
a
Supplementary Table 9. HRsa and 95% CIs for the Association of the 20 Food and Nutrient Intakes With Colorectal Cancer
Risk by Sex (Men vs Women) in the NLCS Study
Dietary Variables Men, HR (95% CI) Women, HR (95% CI) P Value for Heterogeneity
Alcohol 1.07 (1.01–1.13) 1.06 (0.94–1.18) .848

Spirits 1.06 (1.01–1.12) 1.05 (0.90–1.23) .839
Wine 1.02 (0.95–1.09) 1.04 (0.96–1.12) .700
Beer, cider 1.02 (0.97–1.08) 0.99 (0.76–1.29) .557
Soft drinks 0.99 (0.91–1.07) 0.91 (0.83–1.01) .142
Milk 0.92 (0.86–0.98) 0.95 (0.88–1.03) .519
Cheese 1.01 (0.95–1.08) 0.95 (0.88–1.03) .247
Calcium 0.94 (0.88–1.01) 0.96 (0.89–1.03) .667
Phosphorus 0.96 (0.90–1.02) 0.94 (0.87–1.02) .661
Magnesium 0.95 (0.90–1.02) 0.95 (0.87–1.03) 1.000
Potassium 0.99 (0.92–1.05) 0.99 (0.92–1.07) 1.000
Riboflavin 0.94 (0.88–1.00) 0.97 (0.90–1.05) .523
Vitamin B6 1.02 (0.96–1.08) 1.02 (0.94–1.10) 1.000
Beta carotene 0.96 (0.90–1.02) 0.97 (0.90–1.04) .845
Fruits 1.01 (0.94–1.08) 0.99 (0.92–1.06) .694
Fiber 0.99 (0.94–1.05) 0.98 (0.91–1.06) .832
Nonwhite bread 1.00 (0.95–1.07) 1.00 (0.91–1.11) 1.000
Bananas 1.01 (0.94–1.08) 1.03 (0.96–1.11) .712
Pork 1.00 (0.94–1.07) 0.99 (0.92–1.07) .840
Total protein 0.96 (0.89–1.03) 0.94 (0.86–1.02) .697
a
Supplementary Table 10. Multivariable Analysis of Mutually Adjusted Foods and Nutrients
Variable Betaa SE HR Z Value P Value VIF
Alcohol 0.0530 0.0140 1.0544 3.7784 .0002 1.2

Milk 0.0052 0.0244 1.0052 0.2133 .8311 2.9
Cheese –0.0013 0.0260 0.9987 –0.0493 .9607 2.4
Calcium –0.0498 0.0380 0.9514 –1.3091 .1905 6.1
Phosphorous –0.0574 0.0450 0.9442 –1.2768 .2017 8.0
Magnesium –0.0084 0.0283 0.9916 –0.2982 .7655 2.1
Potassium 0.0047 0.0267 1.0047 0.1764 .8600 2.6
Riboflavin 0.0432 0.0328 1.0441 1.3155 .1884 3.1
Beta carotene –0.0328 0.0170 0.9677 –1.9263 .0541 1.2
Total proteins –0.0013 0.0288 0.9987 –0.0463 .9630 3.1
A value >10 is indicative of multicollinearity.

CI, confidence interval; EPIC, European Prospective Investigation into Cancer and Nutrition; HR, hazard ratio; VIF, variance inflation factor.
a
Also adjusted for total energy intake (kcal, continuous); smoking status (never, former, current); body mass index <20, 20–22.9, 23–24.9, 25–29.9, 30–34.9, 35
kg/m2); physical activity (inactive, moderately inactive, moderately active, active); diabetes history (no, yes); education status (none/primary, technical/profes-
sional, secondary, longer [including university]) and stratified by age at recruitment (<40, 40–44.9, 45–49.9, 50–54.9, 55–59.9, 60–64.9, 65–69.9, 70–74.9, 75
years), sex, and recruitment center.

A Prospective Diet-Wide Association Study For Risk of Colorectal Cancer in EPIC

Uploaded by

Copyright:

Available Formats

A Prospective Diet-Wide Association Study For Risk of Colorectal Cancer in EPIC

Uploaded by

Document Information

Original Description:

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

A Prospective Diet-Wide Association Study For Risk of Colorectal Cancer in EPIC

Uploaded by

Copyright:

Available Formats

Clinical Gastroenterology and Hepatology 2021;-:-–-

A Prospective Diet-Wide Association Study for Risk of

Keywords: nutrition; cohort study; colorectal cancer; epidemiology.

olorectal cancer (CRC) is the third most common

Figure 2. Forest plot showing

Pairwise Correlations and Mutual-Adjustment Gene-Nutrient Interaction Analysis

Supplementary Methods was used to calculate standardized nutrient intakes.3 In

Total Noncases Cases Total Noncases Cases

Total 386,792 381,723 5069 7496 3731 3765

Values are n or n (%).

Supplementary Table 2. Continued Supplementary Table 2. Continued

Alcohol 1.06 (1.02–1.10) 1.09 (1.04–1.14) .309

Alcohol 1.01 (0.96–1.07) 1.11 (1.05–1.16) .015

Alcohol 1.12 (1.08–1.16) 1.03 (0.99–1.07) .002

Alcohol 1.07 (1.04–1.10) 1.06 (1.01–1.12) .704

Alcohol 1.03 (0.98–1.09) 1.11 (1.04–1.20) .100

Alcohol 1.04 (0.97–1.11) 1.03 (0.96–1.10) .843

Alcohol 1.07 (1.01–1.13) 1.06 (0.94–1.18) .848

Variable Betaa SE HR Z Value P Value VIF

Alcohol 0.0530 0.0140 1.0544 3.7784 .0002 1.2

A value >10 is indicative of multicollinearity.

You might also like