See corresponding editorial on page 3.

Fiber intake and total and cause-specific mortality in the European

Prospective Investigation into Cancer and Nutrition cohort13
Shu-Chun Chuang, Teresa Norat, Neil Murphy, Anja Olsen, Anne Tjnneland, Kim Overvad,
Marie Christine Boutron-Ruault, Florence Perquier, Laureen Dartois, Rudolf Kaaks, Birgit Teucher,
Manuela M Bergmann, Heiner Boeing, Antonia Trichopoulou, Pagona Lagiou, Dimitrios Trichopoulos,
Sara Grioni, Carlotta Sacerdote, Salvatore Panico, Domenico Palli, Rosario Tumino, Petra HM Peeters,

Bas Bueno-de-Mesquita, Martine M Ros, Magritt Brustad, Lene Angell Asli, Guri Skeie, J Ramon Quiros,
Carlos A Gonzalez, Mara-Jose Sanchez, Carmen Navarro, Eva Ardanaz Aicua, Miren Dorronsoro, Isabel Drake,
Emily Sonestedt, Ingegerd Johansson, Goran Hallmans, Timothy Key, Francesca Crowe, Kay-Tee Khaw,
Nicholas Wareham, Pietro Ferrari, Nadia Slimani, Isabelle Romieu, Valentina Gallo, Elio Riboli, and Paolo Vineis

INTRODUCTION

Of the noncommunicable conditions, cardiovascular disease

(CVD)4, cancers, diabetes, and respiratory and digestive diseases accounted for .75% of deaths in high-income countries in
2004 (1). Current evidence indicates that a high dietary fiber
intake through regular consumption of whole-grain cereals, legumes, fruit, and vegetables has potential health benefits, particularly for preventing diabetes, CVD, and some cancers (2).

164

However, little is known about the association of dietary fiber

intake with specific causes of death other than CVD and cancer.
A higher intake of whole grains, a source of fiber and other potential beneficial nutrients, was found to be associated with a
reduced risk of developing non-CVD, noncancer inflammatory

1
From the School of Public Health, Imperial College London, London,
United Kingdom (S-CC, TN, NM, PHMP, VG, ER, and PV); the Institute of
Cancer Epidemiology, the Danish Cancer Society, Copenhagen, Denmark (AO
and A Tjnneland); the Department of Epidemiology, School of Public Health,
Aarhus University, Aarhus, Denmark (KO); INSERM, Center for Research in
Epidemiology and Public Health, Villejuif, France (MCB-R, FP, and LD);
Paris South University, Villejuif, France (MCB-R, FP, and LD); the Division
of Cancer Epidemiology, German Cancer Research Center, Germany (BT and
RK); the German Institute of Human Nutrition, Potsdam-Rehbruecke Department of Epidemiology, Nuthetal, Germany (MMB and HB); WHO Collaborating
Center for Food and Nutrition Policies, Department of Hygiene, Epidemiology
and Medical Statistics, University of Athens Medical School, Athens, Greece
(A Trichopoulou); the Hellenic Health Foundation, Athens, Greece (A Trichopoulou); the Department of Epidemiology, Harvard School of Public Health,
Boston, MA (PL and DT); the Bureau of Epidemiologic Research, Academy
of Athens, Athens, Greece (PL and DT); the Nutritional Epidemiology Unit,
Fondazione IRCCS Istituto Nazionale Tumori, Milano, Italy (SG); CPOPiemonte, Torino, Italy (CS); HuGeF Foundation, Torino, Italy (CS and
PV); the Department of Clinical and Experimental Medicine, Federico II
University, Naples, Italy (SP); the Molecular and Nutritional Epidemiology
Unit, Cancer Research and Prevention InstituteISPO, Florence, Italy (DP);
the Cancer Registry and Histopathology Unit, CivileM.P. Arezzo Hospital, ASP Ragusa, Italy (RT); the Julius Center, University Medical Center
Utrecht, Utrecht, Netherlands (PHMP); The National Institute for Public
Health and the Environment, Bilthoven, Netherlands (BB-d-M and MMR);
the Department of Gastroenterology and Hepatology, University Medical
Centre Utrecht, Utrecht, Netherlands (BB-d-M); the Department of Epidemiology, Biostatistics and HTA, Radboud University Nijmegen Medical
Center, Nijmegen, Netherlands (MMR); the Department of Community
Medicine, University of Troms, Troms, Norway (MB, LAA, and GS);
the Public Health and Health Planning Directorate, Asturias, Spain (JRQ);
the Cancer Epidemiology Research Programme, Catalan Institute of Oncology, Barcelona, Spain (CAG); the Andalusian School of Public Health,
Granada, Spain (M-JS); the Consortium for Biomedical Research in Epidemiology and Public Health, Spain (M-JS, CN, EAA, and MD); the Department of Epidemiology, Murcia Regional Health Authority, Murcia, Spain
(CN); Navarre Public Health Institute, Pamplona, Spain (EAA); the Public
Health Division of Gipuzkoa, Basque Regional Health Department, San
Sebastian, Spain (MD); the Department of Clinical Sciences in Malmo, Lund

Am J Clin Nutr 2012;96:16474. Printed in USA. 2012 American Society for Nutrition
Supplemental Material can be found at:
http://ajcn.nutrition.org/content/suppl/2012/06/29/ajcn.111.0
28415.DC1.html

Downloaded from ajcn.nutrition.org by guest on December 1, 2015

ABSTRACT
Background: Previous studies have shown that high fiber intake is
associated with lower mortality. However, little is known about the
association of dietary fiber with specific causes of death other than
cardiovascular disease (CVD).
Objective: The aim of this study was to assess the relation between
fiber intake, mortality, and cause-specific mortality in a large European prospective study of 452,717 men and women.
Design: HRs and 95% CIs were estimated by using Cox proportional hazards models, stratified by age, sex, and center and adjusted
for education, smoking, alcohol consumption, BMI, physical activity, total energy intake, and, in women, ever use of menopausal
hormone therapy.
Results: During a mean follow-up of 12.7 y, a total of 23,582 deaths
were recorded. Fiber intake was inversely associated with total mortality
(HRper 10-g/d increase: 0.90; 95% CI: 0.88, 0.92); with mortality from
circulatory (HRper 10-g/d increase: 0.90 and 0.88 for men and women,
respectively), digestive (HR: 0.61 and 0.64), respiratory (HR: 0.77
and 0.62), and non-CVD noncancer inflammatory (HR: 0.85 and 0.80)
diseases; and with smoking-related cancers (HR: 0.86 and 0.89) but
not with nonsmoking-related cancers (HR: 1.05 and 0.97). The associations were more evident for fiber from cereals and vegetables
than from fruit. The associations were similar across BMI and physical activity categories but were stronger in smokers and participants
who consumed .18 g alcohol/d.
Conclusions: Higher fiber intake is associated with lower mortality,
particularly from circulatory, digestive, and non-CVD noncancer inflammatory diseases. Our results support current recommendations of
high dietary fiber intake for health maintenance.
Am J Clin Nutr
2012;96:16474.

FIBER INTAKE AND MORTALITY IN EPIC

University, Sweden (ID and ES); the Departments of Odontology (IJ) and
Public Health and Clinical Medicine (GH), Umea University, Umea, Sweden; the Cancer Epidemiology Unit, University of Oxford, Oxford, United
Kingdom (TK and FC); the Department of Public Health and Primary Care,
University of Cambridge, Cambridge, United Kingdom (K-TK); the MRC
Epidemiology Unit, Cambridge, United Kingdom (NW); the International
Agency for Research on Cancer, Lyon, France (PF, NS, and IR); and the
Social and Environmental Health Research, London School of Hygiene and
Tropical Medicine, London, United Kingdom (VG).
2
The EPIC cohort is supported by the Europe Against Cancer Program of
the European Commission. The individual centers also received funding
from the following countries: Denmark (Danish Cancer Society), France
(Ligue Centre le Cancer, Institut Gustave Roussy, Mutuelle Generale de
lEducation Nationale, and Institut National de la Sante et de la Recherche
Medicale), Greece (the Hellenic Health Foundation, the Stavros Niarchos
Foundation, and the Hellenic Ministry of Health and Social Solidarity),
Germany (German Cancer Aid and Federal Ministry of Education and Research), Italy (Italian Association for Research on Cancer and the National
Research Council), Netherlands (Dutch Ministry of Public Health, Welfare
and Sports, Netherlands Cancer Registry, LK Research Funds, Dutch Prevention Funds, Dutch Zorg Onderzoek Nederland, World Cancer Research
Fund, and Statistics Netherlands), Norway (HelgaNordforsk Centre of Excellence in Food, Nutrition and Health, The Norwegian Extra Foundation for
Health and Rehabilitation, and The Norwegian Cancer Society), Spain
[Health Research Fund of the Spanish Ministry of Health (Exp 96/0032,
RETICC DR06/0020), the Spanish Regional Governments of Andalusia,
Asturias, Basque Country, Murcia (N0 6236), and the Navarra and the Catalan Institute of Oncology], Sweden (Swedish Cancer Society, Swedish Scientific Council, and regional governments of Skane and Vasterbotten), and
the United Kingdom (Cancer Research UK and Medical Research Council).
3
Address correspondence and reprint requests to T Norat, School of Public
Health, Imperial College London, St Marys Campus, Norfolk Place W2
1PG, London, United Kingdom. E-mail: t.norat@imperial.ac.uk.
4
Abbreviations used: CVD, cardiovascular disease; EPIC, European Prospective Investigation into Cancer and Nutrition; ICD-10, International Classification of Diseases, 10th Revision; SCFA, short-chain fatty acid.
Received October 11, 2011. Accepted for publication April 6, 2012.
First published online May 30, 2012; doi: 10.3945/ajcn.111.028415.

The aim of the current study was to assess the relation between
total dietary fiber intake and fiber from cereals, vegetables, and
fruit and total and cause-specific mortality within EPIC.
SUBJECTS AND METHODS

EPIC cohort
EPIC recruited 518,408 volunteers from 23 centers in 10
countries (Denmark, France, Germany, Greece, Italy, the Netherlands, Norway, Spain, Sweden, and the United Kingdom)
between 1992 and 2000. The cohort was described in detail
previously (16). In brief, the study population included volunteers
aged ;2570 y at the time of recruitment. Questionnaires on
diet, education, occupation, previous illnesses, alcohol, tobacco
consumption, and physical activity were completed by participants, and anthropometric measurements were collected. Individuals who did not complete the questionnaires or were at the
extreme ranking of the ratio of energy intake to the estimated
energy requirement (17) (top and bottom 1%) were excluded
from the analysis (n = 16,824). We further excluded 22,797
participants who had cancer at baseline; 387 participants with
missing date of death; and 25,683 participants who reported
having a history of at least one of the following diseases at
baseline (4370 of whom reported having more than one disease):
heart attack (n = 6242), angina (n = 7530), stroke (n = 3715),
and diabetes (n = 12,566). The number of participants included
in this analysis was 452,717. The study was approved by the
Institutional Review Board at the International Agency for Research on Cancer and local ethics committees. Informed consent
forms were filled out at each local center.
Diet assessment
Extensive self-administered quantitative dietary questionnaires were used in most centers except in Denmark, Naples,
Norway, Umea, and the United Kingdom, where semiquantitative
food-frequency questionnaires were used, and in Malmo, where
a modified diet-history method combining a quantitative foodfrequency questionnaire and 7-d menu book (food record) was
used (16). The method for estimating total dietary fiber intake was
described previously (9, 18). In brief, the gravimetric method for
estimating total dietary fiber (19) of the Association of Official
Analytic Chemists includes soluble and insoluble forms (including lignin) of nonstarch polysaccharides, and resistant starch
was used in all countries except Greece and the United Kingdom,
where total dietary fiber intake was estimated by using the
Englyst method, which includes only nonstarch polysaccharides.
Bread, fruit, and vegetables represented the largest food sources
of dietary fiber intake in EPIC, but food sources varied considerably between centers (9).
Assessment of endpoints
Mortality data were obtained at the regional or national level
(20). In Denmark, Italy, the Netherlands, Norway, Spain, Sweden,
and the United Kingdom, vital status and the causes and the dates
of death were ascertained by death indexes, record linkages with
cancer registries, and boards of health. Active follow-up by mail
or telephone with participants, municipal registries, regional

Downloaded from ajcn.nutrition.org by guest on December 1, 2015

diseases in the Iowa Womens Health Study (3), and a high dietary fiber intake was associated with a reduced risk of death from
respiratory and infectious diseases, in addition to CVD and cancer, in the NIH-AARP cohort (4). These results suggest that the
beneficial effects from dietary fiber might not be limited to CVD
and cancer.
High dietary fiber intake could promote overall health and be
associated with lower mortality through several mechanisms.
Fiber-rich foods may benefit weight control because they have
a low energy density; dietary fiber is fermented by intestinal bacteria to produce short-chain fatty acids (SCFAs), which influence
hepatic insulin sensitivity and lipid synthesis and modulate the
intestinal environment (eg, pH) and macronutrient absorption and
thus play a role in glycemic control; dietary fiber regulates gut
microflora populations and hence influence the metabolism of
bile salt, cholesterol, and fatty acids and the inflammatory response (58).
Previous studies on dietary fiber intake and cause-specific deaths
have been conducted in American populations. The European
Prospective Investigation into Cancer and Nutrition (EPIC) is a
large multicenter prospective cohort study with different dietary
patterns across European countries (9). Previous EPIC analyses
showed that plant-based diets rich in fiber were related to increased
survival in the elderly (10), total dietary fiber intake was associated
with reduced colorectal cancer risk (1113), and cereal fiber was
associated with decreased gastric cancer risk (14), whereas total
dietary fiber intake was not associated with prostate cancer (15).

165

166

CHUANG ET AL

Statistical analysis
The HRs and 95% CIs of the relations between total dietary
fiber intake and mortality were estimated by using multivariate
Cox proportional hazards models. Total dietary fiber and fiber
from cereals, fruit, and vegetables were categorized into 5 groups
according to the quintiles of the distribution among the whole
cohort. The lowest quintile was treated as the reference. The associations with legume fiber were also modeled, but as intakes
were low, the results were omitted from the study. The models were
adjusted for the following categorical variables: education (none or
primary school completed, technical or professional school, secondary school, above secondary school, or not specified), smoking
status (never; current 115, 1625, or 26 cigarettes/d; former
quit 10, 1119, or 20 y; current pipe or cigar occasional,
current/former missing number of cigarettes; or unknown status),
alcohol consumption (never or former; current 6, .618, .18
30, .3060, or .60 g/d; or missing), BMI (in kg/m2: ,18.5,
18.524.9, 2529.9, or 30), physical activity [based on physical
activity at work, cycling and sports (23): inactive, moderately
inactive, moderately active, active, or missing], total energy
intake (kcal/d, in continuous), and ever use of menopausal
hormone therapy for women. Models were stratified by age at
recruitment in 1-y categories to address the secular trends within
center and by study center to control for differences in dietary
and lifestyle questionnaires, follow-up procedures, and other
center-specific effects. Age was used as the primary time variable in the Cox proportional hazards models. Age at death or at
censoring date was used as time variable of end of the study.
Further adjustment for intakes of fish, red meat, and folate did
not change the associations materially, so these variables were
excluded from the final multivariate models. In analyses of fiber
from food sources, each fiber source was mutually adjusted by the

others. Pearson partial correlation coefficients adjusted for age at

recruitment, sex, and center were as follows: 0.11 between cereal
and fruit fibers, 0.06 between cereal and vegetable fibers, and 0.21
between fruit and vegetable fibers. Trend tests were performed by
modeling the median values of each fiber category as continuous
variables.
To improve the comparability of dietary data across study
centers and to partially correct for the effect of diet measurement
error in the risk estimates, dietary intakes from 24-h dietary recall
in a random sample of the cohort (24, 25) were regressed on the
dietary questionnaire intakes by using fixed-effects linear model
stratified by center and adjusted for the same list of covariates
mentioned above and additionally for the weekday and season of
recall (26, 27). Predicted values obtained from the calibration
model were used on a continuous scale in the risk model (28). A
bootstrap sampling procedure with a total of 100 iterations was
used to compute corrected SEs for the calibrated parameter
estimates (29). To reduce the probability of reverse causality,
the first 2 y of follow-up were excluded. To explore potential
effect modification, models were further stratified by smoking
(never, former, or current), alcohol consumption (never or former,
current 18 g/d, or current .18 g/d), BMI (,18.5, 18.524.9,
2529.9, or 30), and physical activity [inactive (combined
inactive and moderately inactive) and active (combined moderately active and active)]. Heterogeneity across categories was
assessed by using Cochrans Q test.
Because multiple outcomes were considered, we also modeled
the HR by using a competing risk model (deaths from cancer and
circulatory, respiratory, and digestive diseases) (3032) and
compared the risk coefficients and SEs in the subgroups of interest. To further explore the shape of the risk function, we fitted
a Cox proportional hazards model with restricted cubic spline
regression (33, 34). We specified 5 knot positions at medians of
each quintile and used a total dietary fiber intake of 25 g/d (35) as
the reference. The analyses were performed by using SAS 9.1
(SAS Institute Inc). All tests were 2 sided, and statistical significance was assessed at the level of 0.05.
RESULTS

The characteristics of the participants at baseline are shown

in Tables 1 and 2. The average age at recruitment was 50.8 6
9.8 y, and women accounted for 71% of participants. Among
both men and women, a higher total dietary fiber intake was
associated with higher education levels, being a never smoker,
and being more physically active.
The mean (6SD) follow-up was 12.7 6 2.4 y, with a total of
4,978,825 person-years. After the first 2 y of follow-up were
excluded (1415 deaths), 23,582 deaths were included in the
analysis. Total dietary fiber intake was inversely associated
with total mortality (HR28.5 vs ,16.4g/d: 0.76; 95% CI: 0.72,
0.80; P-trend , 0.001; HRper 10-g/d increase: 0.90; 95% CI: 0.88,
0.92; Table 3). Calibration strengthened the association
(HRper 10-g/d increase: 0.84; 95% CI: 0.80, 0.88). The associations
were similar in both men and women, across BMI strata and
physical activity categories, but were stronger in smokers and participants with alcohol intake .18 g/d, among both men and women.
The inverse association was observed in most countries, although
there was evidence of significant heterogeneity in the association
of dietary fiber and mortality between countries (P-heterogeneity

Downloaded from ajcn.nutrition.org by guest on December 1, 2015

health departments, physicians, and hospitals were adopted in

Germany, Greece, and France.
Causes of death were coded according to the International Classification of Diseases, 10th Revision (ICD-10). Because of differences across participating centers in time for reporting causes of
death, follow-up dates were truncated at the dates at which .80%
of the causes were known, ie, follow-up was truncated at 30 June
2005 for Cambridge; 31 December 2006 for France, Varese,
Turin, Naples, Granada, Murcia, Malmo, and Denmark; 31 December 2007 for Florence, San Sebastian, Umea, and Norway;
31 December 2008 for Ragusa, Asturias, Navarra, and the Netherlands; 30 June 2009 for Oxford; and the actual date of last contact for Greece and Germany.
The underlying causes of death were used to estimate the
following causes of death (1, 20): cancer (ICD-10: C00-D48),
circulatory (I00-I99), respiratory (J30-J98), digestive (K20-K92),
and non-CVD noncancer inflammatory diseases (3, 21). Cancer
deaths were further divided into smoking-related cancers [oral
cavity (C01-C06 and C08), oropharynx (C09, C10, and C12C14), nasopharynx (C11), esophagus (C15), stomach (C16),
colon and rectum (C18, C19, and C20), liver (C22), pancreas
(C25), nasal cavity and sinuses (C300 and C31), larynx (C32),
lung (C34), kidney (C64), bladder (C65 and C67), and myeloid
leukemia (C92)] (22) and nonsmoking-related cancers (all other
cancers). Mortality for external diseases (S00-Y98) was examined as negative control.

167

FIBER INTAKE AND MORTALITY IN EPIC

TABLE 1
Characteristics, in categorical scale, of study participants by categories of total dietary fiber intake1
Total dietary fiber intake (g/d)2
16.4 to ,20.1

20.1 to ,23.6

23.6 to ,28.5

28.5

21,360 (23.6)
264,381

21,982 (24.3)
269,596

24,194 (26.7)
297,497

27,551 (30.4)
338,742

35,477 (39.3)
440,643

2226
862
3879
846
983
3751
6048
2765

(10.4)
(4.0)
(18.2)
(4.0)
(4.6)
(17.6)
(28.3)
(12.9)

2553
1448
3571
1226
1785
3914
4169
3316

(11.6)
(6.6)
(16.2)
(5.6)
(8.1)
(17.8)
(19.0)
(15.1)

2743
2165
3524
1675
2115
3974
3887
4111

(11.3)
(8.9)
(14.6)
(6.9)
(8.7)
(16.4)
(16.1)
(17.0)

2772
3388
3933
2281
2234
3962
3507
5474

(10.1)
(12.3)
(14.3)
(8.3)
(8.1)
(14.4)
(12.7)
(19.9)

3130
6105
5657
3244
2238
3624
3281
8198

(8.8)
(17.2)
(15.9)
(9.1)
(6.3)
(10.2)
(9.2)
(23.1)

7155
5392
3193
4981
639

(33.5)
(25.2)
(14.9)
(23.3)
(3.0)

7173
5503
3013
5713
580

(32.6)
(25.0)
(13.7)
(26.0)
(2.6)

7798
5836
3305
6701
554

(32.2)
(24.1)
(13.7)
(27.7)
(2.3)

8897
6650
3576
7734
694

(32.3)
(24.1)
(13.0)
(28.1)
(2.5)

11,525
8246
4484
10,356
866

(32.5)
(23.2)
(12.6)
(29.2)
(2.4)

6499
7049
7472
340

(30.4)
(33.0)
(35.0)
(1.6)

7053
7832
6721
376

(32.1)
(35.6)
(30.6)
(1.7)

8017
8679
7140
358

(33.1)
(35.9)
(29.5)
(1.5)

9495
9684
7979
893

(34.5)
(35.1)
(29.0)
(3.2)

13,224
12,447
9422
384

(37.3)
(35.1)
(26.6)
(1.1)

4894 (22.9)
7174 (33.6)
4845 (22.7)
4060 (19.0)
387 (1.8)
69,084 (76.4)
897,533

4372 (19.9)
7198 (32.7)
5398 (24.6)
4597 (20.9)
417 (1.9)
68,657 (75.7)
87,892

4539 (18.8)
7773 (32.1)
5911 (24.4)
5442 (22.5)
529 (2.2)
66,572 (73.3)
855,872

4372 (15.9)
8339 (30.3)
7100 (25.8)
7060 (25.6)
680 (2.5)
63,034 (69.6)
815,213

4635 (13.1)
9514 (26.8)
9040 (25.5)
11,350 (32.0)
938 (2.6)
54,806 (60.7)
711,456

11,651
7108
4837
8402
3960
3264
6840
9996
4286
8740

(16.9)
(10.3)
(7.0)
(12.2)
(5.7)
(4.7)
(9.9)
(14.5)
(6.2)
(12.7)

13,473
6435
4831
8585
6039
3701
6418
6244
4816
8115

(19.6)
(9.4)
(7.0)
(12.5)
(8.8)
(5.4)
(9.3)
(9.1)
(7.0)
(11.8)

14,337
5740
4856
8998
6679
3051
5520
4375
5064
7952

(21.5)
(8.6)
(7.3)
(13.5)
(10.0)
(4.6)
(8.3)
(6.6)
(7.6)
(11.9)

14,372
5463
4607
10,430
6008
2266
4527
3059
5900
6402

(22.8)
(8.7)
(7.3)
(16.5)
(9.5)
(3.6)
(7.2)
(4.9)
(9.4)
(10.2)

11,772
4771
4231
14,096
3081
1364
2983
1935
7265
3308

22,518
15,628
14,848
13,817
2273

(32.6)
(22.6)
(21.5)
(20.0)
(3.3)

19,972
15,090
16,319
14,996
2280

(29.1)
(22.0)
(23.8)
(21.8)
(3.3)

17,962 (27.0)
14,317 (21.5)
16,576 (24.9)
15,449 23.2)
2268 (3.4)

15,779
13,372
16,063
15,356
2464

(25.0)
(21.2)
(25.5)
(24.4)
(3.9)

12,512 (22.8)
11,311 (20.6)
13,194 (24.1)
14,791 27.0)
2998 (5.5)

34,160
14,319
19,063
1542

(49.4)
(20.7)
(27.6)
(2.2)

37,214
15,291
14,434
1717

(54.2)
(22.3)
(21.0)
(2.5)

37,561
15,179
12,229
1603

(56.4)
(22.8)
(18.4)
(2.4)

36,677
14,683
10,154
1520

(58.2)
(23.3)
(16.1)
(2.4)

32,998
13,127
7510
1171

(60.2)
(24.0)
(13.7)
(2.1)

16,333
22,478
13,396
7509
9368

(23.6)
(32.5)
(19.4)
(10.9)
(13.6)

14,514
22,261
14,435
8519
8928

(21.1)
(32.4)
(21.0)
(12.4)
(13.0)

13,179
20,975
14,486
9131
8801

(19.8)
(31.5)
(21.8)
(13.7)
(13.2)

12,022
19,538
14,443
9822
7209

(19.1)
(31.0)
(22.9)
(15.6)
(11.4)

10,438
16,924
12,962
10,584
3898

(19.0)
(30.9)
(23.7)
(19.3)
(7.1)

(21.5)
(8.7)
(7.7)
(25.7)
(5.6)
(2.5)
(5.4)
(3.5)
(13.3)
(6.0)

1
Participants who reported having a history of heart attack, angina, stroke, or diabetes at baseline were excluded; thus, the percentages were not exactly
20% when the data for men and women were combined.
2
Fiber intakes were rounded to one decimal place.
3
The French and Norwegian cohorts recruited only women.

Downloaded from ajcn.nutrition.org by guest on December 1, 2015

Men [n (%)]
Person-years
Country [n (%)]3
Italy
Spain
United Kingdom
Netherlands
Greece
Germany
Sweden
Denmark
Education [n (%)]
None or primary school completed
Technical/professional school
Secondary school
Longer education (including university degree)
Not specified
Smoking status [n (%)]
Never
Former
Current
Unknown
Physical activity [n (%)]
Inactive
Moderately inactive
Moderately active
Active
Missing
Women [n (%)]
Person-years
Country [n (%)]3
France
Italy
Spain
United Kingdom
Netherlands
Greece
Germany
Sweden
Denmark
Norway
Education [n (%)]
None or primary school completed
Technical/professional school
Secondary school
Longer education (including university degree)
Not specified
Smoking status [n (%)]
Never
Former
Current
Unknown
Physical activity [n (%)]
Inactive
Moderately inactive
Moderately active
Active
Missing

,16.4

168

CHUANG ET AL
TABLE 2
Characteristics, in continuous scale, of study participants by categories of total dietary fiber intake
Total dietary fiber intake (g/d)1
,16.4

16.4 to ,20.1

20.1 to ,23.6

23.6 to ,28.5

28.5

1
2

Fiber intakes were rounded to one decimal place.

Mean 6 SD (all such values).

, 0.001; see Supplemental Figure 1 under Supplemental data in

the online issue). Spline regression showed that the change in total
mortality was more evident when the total dietary fiber intake was
,25 g/d (Figure 1).
In a cause-specific analysis (Table 4), we observed an inverse
association between total dietary fiber intake and risk of death
from smoking-related cancers and circulatory, respiratory, digestive, and inflammatory diseases in both men and women. No
associations were observed for deaths from external causes and
nonsmoking-related cancers. The competing risk analysis that
considered deaths from cancer and circulatory, respiratory, and
digestive diseases confirmed our results (data not shown). There
was evidence of effect modification (see Supplemental Tables
16 under Supplemental data in the online issue) by alcohol
drinking for the associations of total dietary fiber intake and
mortality from nonsmoking-related cancers, by smoking status
for mortality from respiratory diseases in both men and women,
and by smoking status and alcohol drinking for mortality from
non-CVD noncancer inflammatory diseases in men.
The association of fiber from cereals, fruit, and vegetables and
all-cause and selected cause-specific mortality is shown in
Figure 2. Cereal fiber was inversely associated with total mortality and with mortality from smoking-related cancers and
circulatory, digestive, and non-CVD noncancer inflammatory
diseases in men and women. The inverse associations with fiber

from vegetables were of similar magnitude, although not always

statistically significant, whereas the associations of fiber from
fruit were more evident in women than in men.
DISCUSSION

In this large cohort analysis, higher total dietary fiber intakes

were associated with lower mortality. Participants with a total
dietary fiber intake 28.5 g/d (the highest quintile) had a 24%
lower mortality than did those with an intake ,16.4 g/d. In the
cause-specific analyses, inverse associations were observed for
mortality from smoking-related cancers and circulatory, respiratory, digestive, and non-CVD noncancer inflammatory diseases. The associations between specific sources of fiber and
total and cause-specific mortality were generally of similar magnitude for cereal and vegetable fiber in both men and women. No
association with fiber from fruit was detected in men but in
women fiber from fruit was inversely related with mortality for
respiratory diseases and non-CVD noncancer inflammatory
diseases.
The wide range of total dietary fiber intake in the EPIC cohort
gave us the opportunity to explore the association between total
dietary fiber intake and mortality, and separately by fiber sources,
and to examine the dose-response relation across a large range of
intakes. The large study size provides statistical power to

Downloaded from ajcn.nutrition.org by guest on December 1, 2015

Person-years
264,381
269,596
297,497
338,742
440,643
Men
52.3 6 10.0
51.8 6 9.9
51.5 6 9.9
50.6 6 10.1
Age at recruitment (y)
52.0 6 10.12
Age at death (y)
64.3 6 10.1
64.5 6 10.0
64.1 6 10.0
63.8 6 10.0
63.0 6 10.1
Duration of follow-up (y)
12.4 6 2.9
12.3 6 2.8
12.3 6 2.7
12.3 6 2.6
12.4 6 2.5
Alcohol intake (g/d)
22.0 6 26.1
21.0 6 23.2
20.3 6 22.2
20.4 6 22.2
19.5 6 21.8
BMI (kg/m2)
26.3 6 3.6
26.5 6 3.6
26.5 6 3.6
26.4 6 3.6
26.3 6 3.7
Vegetable intake (g/d)
105.7 6 76.3
150.4 6 103.3 177.1 6 121.1 204.2 6 139.7 265.7 6 187.1
Fruit intake (g/d)
97.4 6 80.9
152.3 6 111.5 186.4 6 131.9 224.8 6 156.0 332.4 6 249.6
Red meat intake (g/d)
46.2 6 37.7
52.0 6 38.7
55.3 6 40.2
58.7 6 42.9
60.7 6 47.2
Total energy intake (kcal/d) 1854.0 6 478.0 2129.0 6 75.0 2330.0 6 494
2546.0 6 538
2933.0 6 648
Cereal fiber (%)
41.4 6 15.9
42.1 6 15.7
43.2 6 15.7
44.0 6 16.1
44.1 6 16.8
Fruit fiber (%)
13.7 6 9.9
15.4 6 10.2
15.9 6 10.2
16.2 6 10.4
17.9 6 11.5
Vegetable fiber (%)
17.1 6 10.8
17.8 6 10.7
17.5 6 10.4
17.1 6 10.1
16.7 6 9.8
Other fiber (%)
28.1 6 12.7
24.8 6 11.8
23.5 6 11.5
22.7 6 11.2
21.4 6 11.4
Women
Age at recruitment (y)
50.4 6 9.7
50.6 6 9.5
50.7 6 9.5
50.5 6 9.6
50.3 6 10.3
Age at death (y)
63.2 6 10.0
63.4 6 9.8
63.5 6 9.8
63.4 6 9.9
63.3 6 10.5
Duration of follow-up (y)
12.7 6 2.4
12.8 6 2.3
12.9 6 2.3
12.9 6 2.3
13.0 6 2.2
Alcohol (g/d)
8.4 6 12.9
8.0 6 11.6
7.8 6 11.2
7.8 6 11.0
7.7 6 10.8
BMI (kg/m2)
24.9 6 4.4
25.0 6 4.4
24.9 6 4.3
24.8 6 4.3
24.7 6 4.4
Fruit intake (g/d)
134.0 6 82.4
182.0 6 102.2 213.3 6 117.1 251.3 6 133.0 335.1 6 186.1
Vegetable intake (g/d)
134.6 6 93.6
197.5 6 114.4 242.0 6 132.7 295.1 6 155.6 417.7 6 251.5
Red meat intake (g/d)
33.6 6 28.2
37.7 6 29.8
39.3 6 30.8
40.4 6 32.4
38.1 6 34.9
Total energy intake (kcal/d) 1493.0 6 372
1764.0 6 386
1949.0 6 415
2148.0 6 458
2447.0 6 549
Cereal fiber (%)
36.0 6 15.3
35.9 6 14.8
36.1 6 14.9
36.1 6 14.7
36.1 6 15.3
Fruit fiber (%)
19.0 6 11.6
20.6 6 11.1
21.3 6 11.0
22.1 6 11.0
23.3 6 12.0
Vegetable fiber (%)
21.8 6 11.9
22.0 6 11.1
21.9 6 10.7
22.0 6 10.4
22.0 6 10.3
Other fiber (%)
23.4 6 11.3
21.5 6 10.7
20.7 6 9.7
19.9 6 9.5
18.6 6 9.5

1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00

23,582
10,366

2189
3794
4228

718
5247
4025

94
3534
4834
1904

6133
3979
13,216

6486
2916
3533

1878
8678
1957

337
6657
4235
1987

8557
3780

1.28
0.78
0.79
0.78

(0.25,
(0.69,
(0.71,
(0.66,

6.56)
0.88)
0.88)
0.92)

0.51
0.75
0.75
0.74

(0.08,
(0.65,
(0.66,
(0.61,

3.28)
0.672
0.86)
0.001
0.84) ,0.001
0.90)
0.005

1.20
0.91
0.89
0.90

(0.60,
(0.85,
(0.84,
(0.82,

2.40)
0.96)
0.94)
0.98)

(0.52,
(0.82,
(0.77,
(0.76,

1.14)
0.96)
0.94)
1.02)

0.86
0.82
0.82
0.90

(0.56,
(0.75,
(0.73,
(0.77,

1.32)
0.90)
0.91)
1.05)

0.58
0.82
0.83
0.81

(0.35,
(0.74,
(0.74,
(0.68,

0.96)
0.90)
0.94)
0.97)

0.69
0.75
0.79
0.82

(0.41,
(0.67,
(0.69,
(0.66,

1.19)
0.138
0.84) ,0.001
0.91)
0.003
1.00)
0.043

0.87 (0.81, 0.93) 0.83 (0.77, 0.89) 0.81 (0.74, 0.88) 0.78 (0.71, 0.86) ,0.001
0.87 (0.78, 0.97) 0.82 (0.73, 0.92) 0.80 (0.70, 0.90) 0.72 (0.62, 0.83) ,0.001

0.77
0.89
0.85
0.88

(0.64,
(0.86,
(0.86,
(0.80,

1.05)
0.95)
0.98)
0.97)
0.91 (0.87, 0.95)
0.87 (0.82, 0.93)

0.82
0.90
0.92
0.88

0.91 (0.84, 1.00)

0.91 (0.88, 0.95)
0.85 (0.77, 0.93)

3.86)
0.90)
0.82)
0.94)

0.85 (0.74, 0.98) 0.92 (0.79, 1.07) 0.83 (0.70, 0.98) 0.77 (0.63, 0.94)
0.014
0.91 (0.85, 0.97) 0.83 (0.77, 0.89) 0.83 (0.77, 0.90) 0.78 (0.71, 0.85) ,0.001
0.74 (0.64, 0.86) 0.73 (0.63, 0.86) 0.74 (0.62, 0.87) 0.75 (0.61, 0.92)
0.008

(0.10,
(0.71,
(0.67,
(0.68,

0.93 (0.89, 0.98)

0.89 (0.83, 0.95)
0.82 (0.76, 0.89)

0.62
0.80
0.74
0.80

0.87 (0.80, 0.95) 0.86 (0.79, 0.94) 0.85 (0.77, 0.94) 0.80 (0.72, 0.90)
0.001
0.88 (0.78, 1.00) 0.77 (0.68, 0.88) 0.76 (0.66, 0.87) 0.75 (0.64, 0.88) ,0.001
0.80 (0.72, 0.89) 0.77 (0.69, 0.87) 0.75 (0.66, 0.86) 0.66 (0.56, 0.78) ,0.001

3.94)
0.87)
0.93)
1.00)
0.89 (0.85, 0.94)
0.90 (0.85, 0.95)
0.90 (0.87, 0.93)

(0.16,
(0.70,
(0.77,
(0.74,

0.87 (0.80, 0.94) 0.78 (0.71, 0.85) 0.80 (0.73, 0.88) 0.78 (0.70, 0.86) ,0.001
0.76 (0.68, 0.85) 0.76 (0.68, 0.85) 0.75 (0.67, 0.84) 0.68 (0.60, 0.78) ,0.001
0.90 (0.82, 0.92) 0.80 (0.76, 0.87) 0.80 (0.76, 0.87) 0.80 (0.71, 0.82) ,0.001

0.80
0.78
0.85
0.86

0.96 (0.85, 1.09)

0.94 (0.90, 0.99)
0.78 (0.74, 0.83)

0.91 (0.70, 1.18) 1.03 (0.79, 1.34) 0.89 (0.67, 1.18) 0.84 (0.62, 1.14)
0.259
0.90 (0.82, 0.98) 0.80 (0.73, 0.89) 0.84 (0.76, 0.93) 0.83 (0.75, 0.93)
0.005
0.72 (0.65, 0.80) 0.66 (0.60, 0.74) 0.68 (0.61, 0.76) 0.57 (0.50, 0.65) ,0.001

0.249

0.748

0.394

0.024

0.771

0.872

,0.001

,0.001

0.6442

(0.37,
(0.77,
(0.78,
(0.74,

5.00)
0.93)
0.93)
0.98)

(0.43,
(0.75,
(0.72,
(0.65,

1.14)
0.91)
0.92)
0.93)
0.81 (0.74, 0.89)
0.78 (0.70, 0.87)

0.70
0.83
0.82
0.78

0.81 (0.67, 0.97)

0.84 (0.78, 0.92)
0.73 (0.61, 0.87)

0.86 (0.78, 0.96)

0.77 (0.67, 0.88)
0.72 (0.64, 0.82)

0.85 (0.78, 0.92)

0.85 (0.78, 0.92)
0.82 (0.77, 0.88)

1.35
0.85
0.85
0.85

0.96 (0.79, 1.18)

0.91 (0.85, 0.99)
0.69 (0.63, 0.77)

0.97 (0.86, 1.09)

0.87 (0.79, 0.95)
0.71 (0.64, 0.98)

0.84 (0.80, 0.88)

0.84 (0.86, 0.92)

P-trend Per 10-g increase Heterogeneity Per 10-g increase, calibrated

0.98 (0.91, 1.06)

0.89 (0.85, 0.95)
0.79 (0.74, 0.84)

28.5

1.00 (0.85, 1.16) 0.97 (0.83, 1.13) 0.79 (0.67, 0.93) 0.91 (0.76, 1.09)
0.131
0.80 (0.72, 0.89) 0.69 (0.62, 0.78) 0.81 (0.72, 0.91) 0.71 (0.63, 0.81) ,0.001
0.74 (0.67, 0.82) 0.70 (0.63, 0.77) 0.70 (0.62, 0.78) 0.61 (0.53, 0.69) ,0.001

23.6 to ,28.5
0.90 (0.88, 0.92)
0.89 (0.86, 0.92)

20.1 to ,23.6

0.85 (0.82, 0.89) 0.80 (0.76, 0.84) 0.81 (0.77, 0.84) 0.76 (0.72, 0.80) ,0.001
0.80 (0.77, 0.88) 0.80 (0.72, 0.82) 0.80 (0.73, 0.84) 0.70 (0.68,0.79) ,0.001

16.4 to ,20.1

Downloaded from ajcn.nutrition.org by guest on December 1, 2015

1
The HRs were estimated by using a Cox proportional hazard model. Age was used as the primary time variable in the Cox proportional hazard models. Age at death or at the censoring date was used as the
time variable for the end of the study. The models excluded the first 2 y of follow-up and stratified by age at recruitment, sex, and center and adjusted for education (none or primary school completed, technical
or professional school, secondary school, longer education, and not specified), smoking (never; current 115, 1625, or 26 cigarettes/d; former quit 10, 1120, or 20 y; current pipe or cigar occasional,
current/former missing, or unknown), alcohol consumption (never or former; current 6, .618, .1830, .3060, or .60 g/d; or missing), BMI (in kg/m2: ,18.5, 18.524.9, 2529.9, or 30), physical
activity (based on the Cambridge/Bilthoven Physical Activity Index: inactive, moderately inactive, moderately active, active, or missing), and total energy intake (kcal/d).
2
Heterogeneity between men and women.
3
Excludes smoking status unknown.
4
Excludes drinking intensity missing.
5
Excludes physical activity missing. Inactive: combination of inactive and moderately inactive; active: combination of moderately active and active.
6
Further adjusted for ever use of menopausal hormone therapy.

Overall
Men
Smoking status3
Never smoker
Former smoker
Current smoker
Alcohol consumption4
Never or former drinkers
Current drinker, 18 g/d
Current drinker, .18 g/d
BMI
,18.5 kg/m2
18.524.9 kg/m2
2529.9 kg/m2
30 kg/m2
Physical activity5
Inactive
Active
Women6
Smoking status3
Never smoker
Former smoker
Current smoker
Alcohol consumption4
Never or former drinkers
Current drinker, 18 g/d
Current drinker, .18 g/d
BMI
,18.5 kg/m2
18.524.9 kg/m2
2529.9 kg/m2
30 kg/m2
Physical activity5
Inactive
Active

No. of deaths ,16.4

Total dietary fiber intake (g/d)

TABLE 3
HRs of death (and 95% CIs) by total dietary fiber intake and per 10-g increase in intake1

FIBER INTAKE AND MORTALITY IN EPIC

169

170

CHUANG ET AL

investigate cause-specific mortality adjusted for the main potential confounders. The prospective design of the study rules out
recall bias.
One of the limitations of our study was dietary measurement
error. The observed associations were strengthened after calibration for partial correction for measurement error. Previous
studies suggested that women and obese subjects tend to underreport energy and protein intakes (36), but this differential
reporting does not seem to have influenced our results. The
strength of the association between total dietary fiber intake and
total mortality was similar in men and women and in different
strata of BMI. We excluded the first 2 y of follow-up to reduce the
effect of reverse causation (changes in lifestyle after preclinical
symptoms). In addition, we excluded participants with selfreported chronic diseases at baseline. As we considered total
dietary fiber intake in relation to several causes of death, some of
the observed associations may be due to chance.
Our results on total mortality are consistent with previous
reports (4, 3739). We observed a 10% lower risk of death per
10-g/d increase in total dietary fiber intake, compared with a 9%
lower risk observed in the Zutphen study (37) and a 12% and 15%
lower risk among men and women, respectively, in the NIHAARP cohort (4). The associations did not differ across categories of BMI or physical activity, but they were stronger in
smokers and participants in the highest category of alcohol intake
(.18 g/d). In our analyses, we had evidence of heterogeneity
of results across countries. Differences in questionnaires and
methods for estimating fiber intake do not appear to explain the
observed heterogeneity. A possible explanation might be that the
main sources of total dietary fiber intake differed across coun-

tries: the strongest associations were observed in the Danish and

Greek cohorts, which also had the highest percentage of fiber
from cereals and vegetables (56% and 16% in Denmark and
29% from cereals and 36% from vegetables, respectively, in
Greece) in EPIC.
Fiber intake was related to mortality from smoking-related cancers, but was not related to mortality from nonsmoking-related
cancers. Adjustment for smoking attenuated the association with
smoking-related cancers by .10% (see Supplemental Table 1 under
Supplemental data in the online issue). Although total dietary
fiber may have different effects on different sites of cancer (eg,
colorectal cancer, included in the smoking-related cancer), it is
possible that the observed relation is influenced, at least partially, by
residual confounding. Similar confounding and effect modification
by smoking were observed in the association between total dietary
fiber intake and deaths from respiratory diseases (see Supplemental
Table 4 under Supplemental data in the online issue). No association was found between total dietary fiber intake and deaths from
respiratory infections (J00-J06, J10-J18, J20-J22, and H65-H66; n =
28 cases; HRper 10-g/d increase: 0.85; 95% CI: 0.40, 1.82).
In agreement with previous studies, we observed an inverse
association between total dietary fiber intake and mortality from
circulatory diseases (4, 37, 40, 41). This finding is consistent with
the results of clinical trials showing a decrease in LDL-cholesterol
concentrationsa major risk factor for circulatory diseasesand
a reduction in coronary artery disease incidence by increasing
soluble fiber intake (42). However, such effects were not observed
for insoluble fibers in clinical trials (43).
We also observed an inverse association between total dietary
fiber intake and deaths from digestive diseases. The most common

Downloaded from ajcn.nutrition.org by guest on December 1, 2015

FIGURE 1. HRs of death according to total dietary fiber intake. The HRs were estimated by using a Cox proportional hazard model. The solid line indicates
HRs, and the dashed lines indicate 95% CIs derived from a restricted cubic spline regression, with knots placed at the medians of each quintile of the
distribution of total dietary fiber intake. The reference point for total dietary fiber intake is 25 g/d. Age was used as the primary time variable in the Cox
proportional hazard models. Age at death or at censoring date was used as the time variable for the end of the study. The models excluded the first 2 y of
follow-up and stratified by age at recruitment, sex, and center and adjusted for education (none or primary school completed, technical or professional school,
secondary school, longer education, and not specified), smoking (never; current 115, 1625, or 26 cigarettes/d; former quit 10, 1120, or 20 y; current
pipe or cigar occasional, current/former missing, or unknown), alcohol consumption (never or former; current 6, .618, .1830, .3060, or .60 g/d; or
missing), BMI (in kg/m2: ,18.5, 18.524.9, 2529.9, or 30), physical activity (based on the Cambridge/Bilthoven Physical Activity Index: inactive,
moderately inactive, moderately active, active, or missing), and total energy intake (kcal/d).

1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00

5575
2478
3097
2115
397
311
1110
418

,16.4

4039
2640
1399
2489
323
310
909
489

No. of deaths

0.88
0.87
1.06
0.83
0.80
0.81
0.80
0.94

0.93
0.90
0.99
0.91
0.77
0.54
0.66
0.62
(0.81,
(0.77,
(0.93,
(0.72,
(0.59,
(0.56,
(0.66,
(0.68,

(0.84,
(0.79,
(0.85,
(0.79,
(0.55,
(0.38,
(0.53,
(0.44,
0.95)
0.99)
1.20)
0.95)
1.09)
1.16)
0.97)
1.29)

1.03)
1.03)
1.16)
1.03)
1.09)
0.76)
0.81)
0.86)

16.4 to ,20.1

0.92
0.92
1.01
0.80
0.51
0.72
0.65
0.79

0.84
0.80
0.94
0.82
0.56
0.53
0.52
0.89
(0.84,
(0.80,
(0.89,
(0.69,
(0.35,
(0.49,
(0.52,
(0.55,

(0.75,
(0.69,
(0.79,
(0.71,
(0.38,
(0.37,
(0.41,
(0.66,
1.00)
1.05)
1.14)
0.93)
0.73)
1.05)
0.80)
1.13)

0.93)
0.91)
1.11)
0.94)
0.83)
0.76)
0.66)
1.22)

20.1 to ,23.6

0.90
0.87
0.93
0.73
0.67
0.58
0.65
0.85

0.92
0.86
0.96
0.81
0.57
0.42
0.63
0.64
(0.81,
(0.75,
(0.82,
(0.61,
(0.46,
(0.38,
(0.52,
(0.59,

(0.82,
(0.75,
(0.80,
(0.70,
(0.38,
(0.28,
(0.50,
(0.46,
0.99)
1.01)
1.06)
0.86)
0.98)
0.89)
0.82)
1.24)

1.03)
0.99)
1.15)
0.94)
0.85)
0.63)
0.79)
0.90)

23.6 to ,28.5

Total dietary fiber intake (g/d)

(0.72,
(0.63,
(0.96,
(0.71,
(0.30,
(0.18,
(0.41,
(0.57,
(0.73,
(0.63,
(0.87,
(0.55,
(0.25,
(0.25,
(0.47,
(0.61,

0.82
0.75
1.00
0.67
0.40
0.42
0.62
0.94

28.5
0.82
0.75
1.16
0.83
0.47
0.29
0.54
0.81

Downloaded from ajcn.nutrition.org by guest on December 1, 2015

0.92)
0.89)
1.14)
0.82)
0.64)
0.70)
0.80)
1.43)

0.93)
0.87)
1.40)
0.98)
0.75)
0.46)
0.70)
1.17)

0.001
0.004
0.258
,0.001
,0.001
,0.001
,0.001
0.771

0.009
,0.001
0.884
0.032
0.001
,0.001
,0.001
0.559

P-trend

0.91
0.89
0.97
0.88
0.62
0.64
0.80
1.06

0.93
0.86
1.05
0.90
0.77
0.61
0.85
0.99

(0.86,
(0.82,
(0.90,
(0.81,
(0.50,
(0.51,
(0.71,
(0.88,

(0.88,
(0.80,
(0.96,
(0.84,
(0.62,
(0.49,
(0.76,
(0.85,

0.96)
0.96)
1.03)
0.97)
0.76)
0.80)
0.91)
1.27)

0.98)
0.92)
1.15)
0.97)
0.94)
0.76)
0.96)
1.15)

Per 10-g increase

0.86
0.80
0.92
0.79
0.62
0.42
0.66
1.02

0.91
0.89
1.06
0.84
0.67
0.44
0.76
1.00

(0.78,
(0.70,
(0.80,
(0.67,
(0.43,
(0.27,
(0.53,
(0.71,

(0.84,
(0.80,
(0.92,
(0.74,
(0.48,
(0.31,
(0.62,
(0.79,

0.95)
0.93)
1.05)
0.94)
0.89)
0.66)
0.83)
1.48)

0.99)
0.99)
1.22)
0.94)
0.94)
0.62)
0.92)
1.28)

Per 10-g
increase,
calibrated

1
The HRs were estimated by using a Cox proportional hazard model. Age was used as the primary time variable in the Cox proportional hazard models. Age at death or at the censoring date was used as the
time variable for the end of the study. The models excluded the first 2 y of follow-up and stratified by age at recruitment, sex, and center and adjusted for education (none or primary school completed, technical
or professional school, secondary school, longer education, and not specified), smoking (never; current, 115, 1625, or 26 cigarettes/d; former quit 10, 1120, or 20 y; current pipe or cigar, occasional,
current/former missing; or unknown), alcohol consumption (never or former; current 6, .618, .1830, .3060, or .60 g/d; or missing), BMI (in kg/m2: ,18.5, 18.524.9, 2529.9, or 30), physical
activity (based on the Cambridge/Bilthoven Physical Activity Index: inactive, moderately inactive, moderately active, active, or missing), and total energy intake (kcal/d). ICD-10, International Classification of
Diseases, 10th Revision.
2
Smoking-related cancers (22) include cancers in the oral cavity (C01-C06, C08), oropharynx (C09, C10, and C12-C14), nasopharynx (C11), esophagus (C15), stomach (C16), colon and rectum (C18, C19,
and C20), liver (C22), pancreas (C25), nasal cavity and sinuses (C300, C31), larynx (C32), lung (C34), kidney (C64), bladder (C65, C67), and myeloid leukemia (C92).
3
Nonsmoking-related cancers include all other cancers not included in the smoking-related cancers.
4
Further adjusted for ever use of menopausal hormone therapy.

Men
Cancer (C00-D48)
Smoking-related cancers2
Non-smoking-related cancers3
Circulatory diseases (I00-I99)
Respiratory diseases (J30-J98)
Digestive diseases (K20-K92)
Non-CVD noncancer inflammatory diseases
External causes (S00-Y98)
Women4
Cancer (C00-D48)
Smoking-related cancers2
Non-smoking-related cancers3
Circulatory diseases (I00-I99)
Respiratory diseases (J30-J98)
Digestive diseases (K20-K92)
Non-CVD noncancer inflammatory diseases
External causes (S00-Y98)

Cause of death (ICD-10)

TABLE 4
Hazard ratios of cause-specific death (and 95% CIs) by quintile of total dietary fiber intake1

FIBER INTAKE AND MORTALITY IN EPIC

171

172

CHUANG ET AL

cause of deaths from digestive diseases in the current analysis was

diseases of the liver (46%). Confounding from alcohol consumption was not apparent (not adjusted for alcohol drinking
HRper 10-g/d increase: 0.58; 95% CI: 0.50, 0.68). The inverse association was observed in both men and women who reported
moderate drinking (18 g/d) at recruitment (see Supplemental
Table 5 under Supplemental data in the online issue).
Previous studies on whole grain (3) or fiber (21) intake have
reported stronger associations with deaths from non-CVD noncancer inflammatory diseases than from cancer or CVD deaths (3),
similar to the results reported herein (Figure 2). This finding is
consistent with results of cross-sectional analyses, in which total
dietary fiber intake was found to be inversely associated with
several inflammatory markers in blood, such as C-reactive protein,
interleukin-6, and tumor necrosis factor-a (4446). One of the
proposed mechanisms that may underlie the associations is the
production of SCFAs from the fermentation of dietary fiber in the
colon (47). It is unknown how SCFAs influence systemic inflammation; animal studies have indicated that the changes in gut
microbiota after a high-fat diet may increase gut permeability and
plasma endotoxin concentrations, which promote a state of
chronic systemic inflammation. Dietary fiber supplementation
was also found to reduce endotoxin concentrations, which supports the idea that dietary fiber may affect gut microbiota and as
a consequence reduce systemic inflammation (4750).

Because the association of dietary fiber intake with mortality

was stronger in smokers and participants in the highest category
of alcohol intake (.18 g/d), it is possible that part of the association is not explained by an effect of dietary fiber itself but
that dietary fiber acts as a marker of other exposures related to
health. For instance, whole grain is a main source of cereal fiber,
but may also provide compounds with health benefits, such as
magnesium or zinc. It is possible that these compounds are the
major contributors in inflammation modulation instead of cereal
fiber (51). The same argument could apply to fiber from vegetables. In our analyses, further adjustment for magnesium or vegetable intakes did not change the associations. Smoking can be
a confounder for the association between total dietary fiber intake
and mortality from smoking-related cancers and respiratory diseases. However, smoking-adjusted and -unadjusted risk estimates
were similar in the association between total dietary fiber intake
and mortality from digestive diseases, and an inverse association
was also observed in never smokers. Total dietary fiber intake
could be a marker of overall dietary patterns or lifestyles (5255);
however, because we carefully adjusted for the main potential
lifestyle confounders it is unlikely that the observed associations
were entirely the result of residual confounding.
In conclusion, although we cannot rule out residual confounding from smoking on the smoking-related cancer and respiratory disease deaths, our results are generally in line with

Downloaded from ajcn.nutrition.org by guest on December 1, 2015

FIGURE 2. HRs and 95% CIs of total death and cause-specific deaths per 5-g/d increase in fiber intake in men (A) and women (B). The HRs were
estimated by using a Cox proportional hazard model. Age was used as the primary time variable in the Cox proportional hazard models. Age at death or at
censoring date was used as time variable of end of the study. The models excluded the first 2 y of follow-up and stratified by age at recruitment, sex, and center
and adjusted for education (none or primary school completed, technical or professional school, secondary school, longer education, and not specified),
smoking (never; current 115, 1625, or 26 cigarettes/d; former quit 10, 1120, or 20 y; current pipe or cigar occasional, current/former missing, or
unknown), alcohol consumption (never or former; current 6, .618, .1830, .3060, or .60 g/d; or missing), BMI (in kg/m2: ,18.5, 18.524.9, 2529.9,
or 30), physical activity (based on the Cambridge/Bilthoven Physical Activity Index: inactive, moderately inactive, moderately active, active, or missing),
total energy intake (kcal/d), other sources of fiber intake, and ever use of menopausal hormone therapy for women. aFurther adjusted for ever use of
menopausal hormone therapy. The x axis refers to HRs.

FIBER INTAKE AND MORTALITY IN EPIC

previous studies (3, 4, 3741). We observed inverse associations

between total dietary fiber intake and mortality, and specifically
mortality from circulatory, digestive, and non-CVD noncancer
inflammatory diseases. These results show that high fiber intake,
mainly from cereals and vegetables, may reduce the risk of death
from these diseases.
The authors responsibilities were as followsS-CC, TN, NM, and PV:
participated in the data analysis, manuscript writing, and interpretation of
the results; and AO, A Tjnneland, KO, MCB-R, FP, LD, RK, BT, MMB,
HB, A Trichopoulou, PL, DT, SG, CS, SP, DP, RT, PHMP, BB-d-M,
MMR, MB, LAA, GS, JRQ, CAG, M-JS, CN, EAA, MD, ID, ES, IJ, GH,
TK, FC, K-TK, NW, PF, NS, IR, VG, ER, and PV: were involved in data collection and interpretation of the results. None of the authors declared a conflict
of interest. The funders played no role in designing or conducting the study or
in the collection, management, analysis, or interpretation of the data and had
no input on the preparation, review, or approval of this manuscript.

REFERENCES

16. Riboli E, Hunt KJ, Slimani N, Ferrari P, Norat T, Fahey M, Charrondiere

UR, Hemon B, Casagrande C, Vignat J, et al. European Prospective
Investigation into Cancer and Nutrition (EPIC): study populations and
data collection. Public Health Nutr 2002;5:111324.
17. Bingham SA, Gill C, Welch A, Day K, Cassidy A, Khaw KT, Sneyd MJ,
Key TJ, Roe L, Day NE. Comparison of dietary assessment methods in
nutritional epidemiology: weighed records v. 24 h recalls, food-frequency
questionnaires and estimated-diet records. Br J Nutr 1994;72:61943.
18. Slimani N, Deharveng G, Unwin I, Southgate DA, Vignat J, Skeie G,
Salvini S, Parpinel M, Moller A, Ireland J, et al. The EPIC nutrient
database project (ENDB): a first attempt to standardize nutrient databases across the 10 European countries participating in the EPIC study.
Eur J Clin Nutr 2007;61:103756.
19. DeVries JW, Rader JI. Historical perspective as a guide for identifying
and developing applicable methods for dietary fiber. J AOAC Int 2005;
88:134966.
20. Pischon T, Boeing H, Hoffmann K, Bergmann M, Schulze MB,
Overvad K, van der Schouw YT, Spencer E, Moons KG, Tjonneland A,
et al. General and abdominal adiposity and risk of death in Europe.
N Engl J Med 2008;359:210520.
21. Buyken AE, Flood V, Empson M, Rochtchina E, Barclay AW,
Brand-Miller J, Mitchell P. Carbohydrate nutrition and inflammatory
disease mortality in older adults. Am J Clin Nutr 2010;92:63443.
22. IARC Working Group. Tobacco smoke and involuntary smoking. 2004:
11438.
23. Wareham NJ, Jakes RW, Rennie KL, Mitchell J, Hennings S, Day NE.
Validity and repeatability of the EPIC-Norfolk Physical Activity
Questionnaire. Int J Epidemiol 2002;31:16874.
24. Slimani N, Kaaks R, Ferrari P, Casagrande C, Clavel-Chapelon F,
Lotze G, Kroke A, Trichopoulos D, Trichopoulou A, Lauria C, et al.
European Prospective Investigation into Cancer and Nutrition (EPIC)
calibration study: rationale, design and population characteristics.
Public Health Nutr 2002;5:112545.
25. Slimani N, Ferrari P, Ocke M, Welch A, Boeing H, Liere M, Pala V,
Amiano P, Lagiou A, Mattisson I, et al. Standardization of the 24-hour
diet recall calibration method used in the european prospective investigation into cancer and nutrition (EPIC): general concepts and
preliminary results. Eur J Clin Nutr 2000;54:90017.
26. Ferrari P, Kaaks R, Fahey MT, Slimani N, Day NE, Pera G, Boshuizen
HC, Roddam A, Boeing H, Nagel G, et al. Within- and between-cohort
variation in measured macronutrient intakes, taking account of measurement errors, in the European Prospective Investigation into Cancer
and Nutrition study. Am J Epidemiol 2004;160:81422.
27. Ferrari P, Day NE, Boshuizen HC, Roddam A, Hoffmann K, Thiebaut
A, Pera G, Overvad K, Lund E, Trichopoulou A, et al. The evaluation
of the diet/disease relation in the EPIC study: considerations for the
calibration and the disease models. Int J Epidemiol 2008;37:36878.
28. Rosner B, Willett WC, Spiegelman D. Correction of logistic regression
relative risk estimates and confidence intervals for systematic within-person measurement error. Stat Med 1989;8:105169.
29. Rosner B, Gore R. Measurement error correction in nutritional epidemiology based on individual foods, with application to the relation of
diet to breast cancer. Am J Epidemiol 2001;154:82735.
30. Smith-Warner SA, Spiegelman D, Ritz J, Albanes D, Beeson WL,
Bernstein L, Berrino F, van den Brandt PA, Buring JE, Cho E, et al.
Methods for pooling results of epidemiologic studies: the Pooling
Project of Prospective Studies of Diet and Cancer. Am J Epidemiol
2006;163:105364.
31. Rosthj S, Andersen PK, Abildstrom SZ. SAS macros for estimation of
the cumulative incidence functions based on a Cox regression model
for competing risks survival data. Comput Methods Programs Biomed
2004;74:6975.
32. Baer HJ, Glynn RJ, Hu FB, Hankinson SE, Willett WC, Colditz GA,
Stampfer M, Rosner B. Risk factors for mortality in the nurses health
study: a competing risks analysis. Am J Epidemiol 2011;173:31929.
33. Durrleman S, Simon R. Flexible regression models with cubic splines.
Stat Med 1989;8:55161.
34. Heinzl H, Kaider A. Gaining more flexibility in Cox proportional
hazards regression models with cubic spline functions. Comput
Methods Programs Biomed 1997;54:2018.
35. Krauss RM, Eckel RH, Howard B, Appel LJ, Daniels SR, Deckelbaum
RJ, Erdman JW Jr, Kris-Etherton P, Goldberg IJ, Kotchen TA, et al.
AHA dietary guidelines: revision 2000: A statement for healthcare

Downloaded from ajcn.nutrition.org by guest on December 1, 2015

1. The global burden of disease. 2004 update. Geneva, Switzerland:

World Health Organization, 2008.
2. Joint WHO/FAO Expert Consultation. Diet, Nutrition and the prevention of chronic diseases. Geneva, Switzerland: World Health Organization, 2003.
3. Jacobs DR Jr, Andersen LF, Blomhoff R. Whole-grain consumption is
associated with a reduced risk of noncardiovascular, noncancer death
attributed to inflammatory diseases in the Iowa Womens Health Study.
Am J Clin Nutr 2007;85:160614.
4. Park Y, Subar AF, Hollenbeck A, Schatzkin A. Dietary fiber intake and
mortality in the NIH-AARP Diet and Health Study. Arch Intern Med
2011;171:10618.
5. Babio N, Balanza R, Basulto J, Bullo M, Salas-Salvado J. Dietary fibre:
influence on body weight, glycemic control and plasma cholesterol
profile. Nutr Hosp 2010;25:32740.
6. Young GP, Hu Y, Le Leu RK, Nyskohus L. Dietary fibre and colorectal
cancer: a model for environmentgene interactions. Mol Nutr Food Res
2005;49:57184.
7. King DE. Dietary fiber, inflammation, and cardiovascular disease. Mol
Nutr Food Res 2005;49:594600.
8. Galisteo M, Duarte J, Zarzuelo A. Effects of dietary fibers on disturbances clustered in the metabolic syndrome. J Nutr Biochem 2008;19:
7184.
9. Cust AE, Skilton MR, van Bakel MM, Halkjaer J, Olsen A, Agnoli C,
Psaltopoulou T, Buurma E, Sonestedt E, Chirlaque MD, et al. Total
dietary carbohydrate, sugar, starch and fibre intakes in the European
Prospective Investigation into Cancer and Nutrition. Eur J Clin Nutr
2009;63(Suppl 4):S3760.
10. Bamia C, Trichopoulos D, Ferrari P, Overvad K, Bjerregaard L,
Tjonneland A, Halkjaer J, Clavel-Chapelon F, Kesse E, Boutron-Ruault
MC, et al. Dietary patterns and survival of older Europeans: the EPICElderly Study (European Prospective Investigation into Cancer and
Nutrition). Public Health Nutr 2007;10:5908.
11. Bingham SA, Day NE, Luben R, Ferrari P, Slimani N, Norat T,
Clavel-Chapelon F, Kesse E, Nieters A, Boeing H, et al. Dietary fibre in
food and protection against colorectal cancer in the European Prospective Investigation into Cancer and Nutrition (EPIC): an observational study. Lancet 2003;361:1496501.
12. Bingham SA, Norat T, Moskal A, Ferrari P, Slimani N, ClavelChapelon F, Kesse E, Nieters A, Boeing H, Tjonneland A, et al. Is the
association with fiber from foods in colorectal cancer confounded by
folate intake? Cancer Epidemiol Biomarkers Prev 2005;14:15526.
13. Bingham S. The fibre-folate debate in colo-rectal cancer. Proc Nutr Soc
2006;65:1923.
14. Mendez M, Guillem P, Agudo A, Bueno-de-Mesquita HB, Palli D,
Boeing H, Carneiro F, Berrino F, Sacerdote C, Tumino R, et al. Cereal
fiber intake may reduce risk of gastric adenocarcinomas: the EPICEURGAST study. Int J Cancer 2007;121:161823.
15. Suzuki R, Allen NE, Key TJ, Appleby PN, Tjonneland A, Johnsen NF,
Jensen MK, Overvad K, Boeing H, Pischon T, et al. A prospective
analysis of the association between dietary fiber intake and prostate
cancer risk in EPIC. Int J Cancer 2009;124:2459.

173

174

36.
37.
38.

39.
40.

41.

42.

44.

45.

professionals from the Nutrition Committee of the American Heart

Association. Circulation 2000;102:228499.
Lissner L, Troiano RP, Midthune D, Heitmann BL, Kipnis V, Subar AF,
Potischman N. OPEN about obesity: recovery biomarkers, dietary reporting errors and BMI. Int J Obes (Lond) 2007;31:95661.
Streppel MT, Ocke MC, Boshuizen HC, Kok FJ, Kromhout D. Dietary
fiber intake in relation to coronary heart disease and all-cause mortality
over 40 y: the Zutphen Study. Am J Clin Nutr 2008;88:111925.
Todd S, Woodward M, Tunstall-Pedoe H, Bolton-Smith C. Dietary
antioxidant vitamins and fiber in the etiology of cardiovascular disease
and all-causes mortality: results from the Scottish Heart Health Study.
Am J Epidemiol 1999;150:107380.
Lubin F, Lusky A, Chetrit A, Dankner R. Lifestyle and ethnicity play
a role in all-cause mortality. J Nutr 2003;133:11805.
Eshak ES, Iso H, Date C, Kikuchi S, Watanabe Y, Wada Y, Wakai K,
Tamakoshi A. Dietary fiber intake is associated with reduced risk of
mortality from cardiovascular disease among Japanese men and
women. J Nutr 2010;140:144553.
Pereira MA, OReilly E, Augustsson K, Fraser GE, Goldbourt U,
Heitmann BL, Hallmans G, Knekt P, Liu S, Pietinen P, et al. Dietary
fiber and risk of coronary heart disease: a pooled analysis of cohort
studies. Arch Intern Med 2004;164:3706.
Brown L, Rosner B, Willett WW, Sacks FM. Cholesterol-lowering
effects of dietary fiber: a meta-analysis. Am J Clin Nutr 1999;69:
3042.
Smith CE, Tucker KL. Health benefits of cereal fibre: a review of
clinical trials. Nutr Res Rev 2011;Feb 15:114.
Ma Y, Griffith JA, Chasan-Taber L, Olendzki BC, Jackson E, Stanek EJ
III, Li W, Pagoto SL, Hafner AR, Ockene IS. Association between
dietary fiber and serum C-reactive protein. Am J Clin Nutr 2006;83:
7606.
Ma Y, Hebert JR, Li W, Bertone-Johnson ER, Olendzki B, Pagoto SL,
Tinker L, Rosal MC, Ockene IS, Ockene JK, et al. Association between

46.

47.

48.
49.

50.

51.

52.
53.

54.

55.

dietary fiber and markers of systemic inflammation in the Womens

Health Initiative Observational Study. Nutrition 2008;24:9419.
Wannamethee SG, Whincup PH, Thomas MC, Sattar N. Associations
between dietary fiber and inflammation, hepatic function, and risk of
type 2 diabetes in older men: potential mechanisms for the benefits of
fiber on diabetes risk. Diabetes Care 2009;32:18235.
Meijer K, de Vos P, Priebe MG. Butyrate and other short-chain fatty
acids as modulators of immunity: what relevance for health? Curr Opin
Clin Nutr Metab Care 2010;13:71521.
Son G, Kremer M, Hines IN. Contribution of gut bacteria to liver
pathobiology. Gastroenterol Res Pract 2010;2010:pii: 453563.
Cani PD, Delzenne NM. Interplay between obesity and associated
metabolic disorders: new insights into the gut microbiota. Curr Opin
Pharmacol 2009;9:73743.
Conterno L, Fava F, Viola R, Tuohy KM. Obesity and the gut microbiota: does up-regulating colonic fermentation protect against obesity
and metabolic disease? Genes Nutr 2011;6:24160.
Fardet A. New hypotheses for the health-protective mechanisms of
whole-grain cereals: what is beyond fibre? Nutr Res Rev 2010;23:
65134.
Lang R, Jebb SA. Who consumes whole grains, and how much? Proc
Nutr Soc 2003;62:1237.
Thane CW, Jones AR, Stephen AM, Seal CJ, Jebb SA. Comparative
whole-grain intake of British adults in 1986-7 and 2000-1. Br J Nutr
2007;97:98792.
Hulshof KF, Brussaard JH, Kruizinga AG, Telman J, Lowik MR.
Socio-economic status, dietary intake and 10 y trends: the Dutch National Food Consumption Survey. Eur J Clin Nutr 2003;57:12837.
Kyr C, Skeie G, Dragsted LO, Christensen J, Overvad K, Hallmans G,
Johansson I, Lund E, Slimani N, Johnsen NF, et al. Intake of whole
grains in Scandinavia is associated with healthy lifestyle, socio-economic
and dietary factors. Public Health Nutr 2011;14:178795.

Downloaded from ajcn.nutrition.org by guest on December 1, 2015

43.

CHUANG ET AL