Prospective cohort studies have consistently found no important link between fiber intake and risk of colorectal cancer. The recent large, prospective European Prospective Investigation into Cancer and Nutrition has challenged this paradigm by suggesting significant protection by high fiber intake. We prospectively investigated the association of fiber intake with the incidence of colon and rectal cancers in two large cohorts: the Nurses' Health Study (76,947 women) and the Health Professionals Follow-up Study (47,279 men). Diet was assessed repeatedly in 1984, 1986, 1990, and 1994 among women and in 1986, 1990, and 1994 among men. The incidence of cancer of the colon and rectum was ascertained up to the year 2000. Relative risk estimates were calculated using a Cox proportional hazards model simultaneously controlling for potential confounding variables. During follow-up including 1.8 million person-years and 1,596 cases of colorectal cancer, we found little association with fiber intake after controlling for confounding variables. The hazard ratio for a 5-g/d increase in fiber intake was 0.91 (95% confidence interval, 0.87-0.95) after adjusting for covariates used in the European Prospective Investigation into Cancer and Nutrition study and 0.99 (95% confidence interval, 0.95-1.04) after adjusting for additional confounding variables. Our data from two large prospective cohorts with long follow-up and repeated assessment of fiber intake and of a large number of potential confounding variables do not indicate an important association between fiber intake and colorectal cancer but reveal considerable confounding by other dietary and lifestyle factors.

The potential importance of dietary fiber intake for prevention of colorectal cancer was suggested more than three decades ago by Burkitt, who related the low incidence of colon cancer in parts of Africa to the local diets, which were high in plant foods and dietary fiber (1). Burkitt hypothesized that high fiber intake results in a high stool bulk, which reduces stool transit time and thus lowers exposure to potential carcinogens. Other mechanisms for a reduced risk of colorectal cancer have been proposed, including the capacity of fiber to bind bile acids and lower pH (2).

Burkitt's initial observation was followed by numerous case-control studies. A combined analysis of 13 case-control studies (3) as well as a meta-analysis of 16 case-control studies (4) indicated an inverse association between fiber intake and colorectal cancer. Inclusion of studies was selective, however, and effect estimates unadjusted for potential confounders were used for most studies. Moreover, recall bias is a severe threat to the validity of retrospective case-control studies of fiber intake and any disease outcome.

In contrast, 10 prospective studies (which avoid the potential for recall and control selection bias; refs. 5-14), including initial analyses from the Nurses' Health Study (NHS, ref. 13) and the Health Professionals' Follow-up Study (HPFS; ref. 14), have largely failed to support this association. Most recently, a report from the European Prospective Investigation into Cancer and Nutrition (EPIC) including 1,065 incident cases of colorectal cancer among 519,978 individuals followed for an average of 4.5 years described an inverse link between fiber intake and colorectal cancer incidence (15). This study included various populations, ranging from Scandinavia to the Mediterranean, with diverse dietary habits. Two different methods were used to calculate fiber intake from the food frequency questionnaire (FFQ). Folate intake and other potential confounders, however, were not controlled for in the EPIC analysis.

We conducted an updated analysis of the association between fiber intake and colorectal cancer in the NHS and the HPFS using both methods employed in EPIC to derive fiber intake from repeatedly given dietary questionnaires over 14 to 16 years of follow-up.

Study Cohorts

The NHS was initiated in 1976, when 121,700 female registered nurses ages 30 to 55 years completed a self-administered questionnaire providing information on demographics, lifestyle, and medical history. Similarly, the HPFS comprises 51,529 male health professionals, including dentists, veterinarians, pharmacists, optometrists, osteopaths, and podiatrists, who were 40 to 75 years of age upon enrollment in 1986. Participants in both cohorts have been followed through self-administered biennial questionnaires that serve to update information on lifestyle factors and disease. The NHS has been approved by the Institutional Review Board of the Brigham and Women's Hospital in Boston and the HPFS has Institutional Review Board approval of the Harvard School of Public Health, Boston.

The study populations for the present analyses consisted of all women free of cancer in 1984 who completed the 1984 FFQ and who reported a total caloric intake between 500 and 3,500 cal/d, as well as all men free of cancer in 1986 who completed the 1986 FFQ and who reported a total caloric intake between 800 and 4,200 cal/d.

Dietary Assessment

Dietary intake data were collected repeatedly in both cohorts by self-administered semiquantitative FFQs (16). In the NHS, diet was assessed in 1980, 1984, 1986, 1990, and 1994; for the present analysis, dietary information was used starting in 1984, when an expanded FFQ including ∼130 food items with greater detail on fiber intake was introduced. In the HPFS, diet was assessed in 1986, 1990, and 1994, using an equivalent FFQ.

Nine mutually exclusive response categories were provided for the frequency of intake in both cohorts. The choices ranged from “almost never or less than once per month” to “six or more times per day”. Participants reported their average intake of a prespecified portion size for each food over the past year. Reproducibility and validity of the FFQ for both women and men have been reported previously (17, 18). The correlation coefficients (adjusted for random within-person variation) for the comparison of dietary fiber intake assessed with an FFQ administered twice one year apart and the means of four 1-week diet records were 0.51 and 0.58 among women (17) and for the comparison of dietary fiber intake assessed with an FFQ administered twice one year apart and the means of two 1-week diet records 0.63 and 0.68 among men (18).

Overall fiber intake was calculated using the Association of Official Analytical Chemists (AOAC) method (accepted by the U.S. Food and Drug Administration and the Food and Agriculture Organization of the WHO for nutrition labeling purposes; ref. 19) and Englyst nonstarch polysaccharides fiber (used primarily in the United Kingdom; refs. 20, 21). Whereas the AOAC derivation includes some starch as dietary fiber, the method by Englyst distinguishes nonstarch polysaccharides from starch.

Ascertainment of Cases

On each biennial questionnaire, we ask cohort participants whether cancer of the colon or rectum had been diagnosed during the previous 2 years. Deaths are reported to us primarily through family members; to identify fatalities among subjects who had not responded to questionnaires, we used the National Death Index and the U.S. Postal Service. We have estimated that >98% of deaths are ascertained (22).

When a participant (or next of kin for decedents) reported a diagnosis of cancer, we sought permission to obtain relevant medical records and pathology reports. A study physician blinded to all questionnaire data reviewed the medical records to extract information on the histologic type, the anatomic location, and the stage of the cancer. For this analysis, we included incident cases of colon and rectal cancer diagnosed between 1984 and 2000 for NHS and between 1986 and 2000 for HPFS, because follow-up of the self-reported cases was complete up to 2000. We included some colorectal cancer cases that were not clearly defined as colon nor rectal cancer in the colorectal cancer analyses. Only cases of invasive adenocarcinoma were included in this analysis; cases of carcinomas in situ were not considered.

Follow-up

Follow-up rates for the cohorts were calculated as the total number of person-years during which questionnaires were returned divided by the total number of possible person-years of follow-up for the cohort. The follow-up rate for the population studied in this analysis was 98.9% for the NHS and 97.0% for the HPFS.

Statistical Analysis

Dietary fiber was used in the analysis as intake in grams per 1,000 cal. The nutrient density of fiber intake per 1,000 cal was chosen to account for differences in absolute intake among women and men, which simplifies the combined presentation of results for women and men. Fiber was also grouped into five equally spaced categories to define exposure.

Absolute fiber intake was also used in the analysis as a continuous variable, and the hazard ratio (HR) of colorectal cancer was assessed per 5-g increase in absolute fiber intake per day. When using fiber intake as a continuous variable, fiber intake was truncated at 35 g/d for AOAC fiber and at 30 g/d for Englyst fiber to avoid undue influence of outliers or implausible values.

Incidence rates of colorectal cancer within each category of fiber intake were calculated by dividing the number of new cases of colorectal cancer by person-years of follow-up. Person-years of follow-up for each participant were calculated from the date of return of the 1984 questionnaire (NHS) or the 1986 questionnaire (HPFS) to the date of diagnosis of colon or rectal cancer, death, or the end of follow-up (with a cutoff date of June 1, 2000, for the NHS and January 31, 2000, for the HPFS), whichever occurred first. Participants with cancers other than nonmelanoma skin cancer at baseline were excluded from the analyses. Participants who reported Crohn's disease or ulcerative colitis were excluded at baseline, and follow-up was censored when these diseases were diagnosed after baseline. Analyses were carried out for all colorectal cancers and separately for colon and rectal cancer.

A Cox proportional hazards model was used to calculate the relative risk of developing invasive colorectal cancer (23). The proportional hazards model allowed us to adjust simultaneously for multiple potential time-dependent confounders of this association. We included only covariates in our models that were a priori possible risk factors for colorectal cancer to avoid potential overadjustment. The following dietary predictors of colorectal cancer were included in the final models: folate (24), red meat and processed meat (25), calcium (26), alcohol (2), methionine (27), and glycemic load (28).

Regression models were adjusted for age, time period, family history of colorectal cancer, history of sigmoidoscopy or colonoscopy, height, body mass index (weight/height2), physical activity, regular aspirin use, duration of aspirin use, pack-years of smoking in early adulthood, multivitamin supplement use, total caloric intake, alcohol consumption, dietary folate, red meat consumption, processed meat consumption, glycemic load, calcium intake, methionine intake, and (among women) menopausal status and postmenopausal hormone use. All covariates except for height were repeatedly assessed and updated in the analysis.

All nutrients included in the covariate-adjusted models were energy-adjusted. In addition, total caloric intake was included in the covariate-adjusted model to control for residual confounding by total energy intake and to minimize extraneous variation due to general underreporting or overreporting of food items on the FFQ (16). To represent long-term dietary patterns of individual subjects as accurately as possible and to reduce random within-person variation, we modeled the incidence of colorectal cancer in relation to the cumulative average fiber intake from all available dietary questionnaires up to the start of each 2-year follow-up interval (29). Among women, dietary data from the 1984 questionnaire were used to predict colorectal cancers diagnosed between 1984 and 1986; the average of the 1984 and 1986 dietary intake was used to predict outcomes between 1986 and 1990; the average of the 1984, 1986, and 1990 FFQs was used to predict colorectal cancer between 1990 and 1994; and the average of the 1984, 1986, 1990, and 1994 FFQs was used to predict colorectal cancers from 1994 to 2000. Among men, dietary data from the 1986 questionnaire were used to predict the outcomes between 1986 and 1990; the average of 1986 and 1990 dietary intake was used to predict outcomes between 1990 and 1994; and the average of 1986, 1990, and 1994 dietary intake was used to predict outcomes between 1994 and 2000. This method uses all self-reports of fiber intake to predict the incidence of colorectal cancer, rather than the most recent report, thus giving equal weight to recent diet and diet in the more distant past and reducing measurement error due random within-person variation.

We also analyzed our data following as closely as possible the methods and models used in the EPIC study (15). For these analyses, we used only baseline AOAC fiber information (1984 for NHS and 1986 for HPFS) in quintiles of intake. Regression models were adjusted for age, height, baseline weight, and baseline total caloric intake.

Because of the difference in gender, follow-up time, FFQs, and covariates in the two cohorts, analyses were done separately for each cohort, and the results were then combined using a fixed effects model weighting the two relative risk estimates by the inverse of the SE (30). Tests of heterogeneity were used to evaluate whether associations differed between women and men; results are shown separately whenever significant heterogeneity was seen.

The 76,947 women included in this analysis contributed 1,179,656 person-years of follow-up during 16 years; 744 new cases of colon cancer and 175 new cases of rectal cancer were documented during follow-up. The 47,279 men included in this analysis contributed 596,842 person-years of follow-up during 14 years; 458 new cases of colon cancer and 135 new cases of rectal cancer were documented during follow-up.

The Pearson correlation coefficient between AOAC and Englyst fiber was 0.99 in both cohorts. The mean intake of AOAC fiber was 19.5 g/d (SD = 6.8) among women, 21.9 g/d (SD = 8.2) among men, and 20.3 g/d (SD = 7.4) among women and men combined.

The distribution of predictors of colorectal cancer by frequency of fiber intake (per kcal) during the observation period in the NHS and the HPFS is presented in Table 1. Persons who reported higher fiber intake per calories tended to be older and to have had a higher prevalence of health-seeking behaviors, as indicated by higher rates of sigmoidoscopy, multivitamin supplement use, and physical activity; and lower rates of smoking and alcohol consumption. Participants with higher fiber density consumed less red meat and more dietary folate and calcium and had a higher glycemic load.

Table 1.

Age-standardized characteristics of the study population during follow-up according to fiber intake/1,000 cal (mean values unless indicated otherwise)

AOAC fiber intake/1,000 cal
<8.0 g8.0-10.0 g10.0-12.0 g12.0-14.0 g>14.0 g
Women      
Age 54.2 56.1 58.1 59.8 60.9 
Family history of colon cancer (%) 15.0 15.6 16.0 16.3 16.3 
Height (cm) 163.9 163.9 163.9 163.9 163.6 
Body mass index (kg/m225.3 25.6 25.6 25.3 24.8 
History of sigmoidoscopy (%) 25.1 28.7 31.8 33.2 34.3 
Physical activity (h/wk) 1.7 2.0 2.3 2.7 3.1 
Smoking (pack-years >35 y in the past) 2.2 1.8 1.7 1.6 1.5 
Alcohol intake (g/d) 10.6 6.3 4.7 3.9 3.0 
Total caloric intake/d 1,810 1,793 1,754 1,698 1,608 
Red meat consumption (servings/d)* 0.35 0.30 0.26 0.22 0.16 
Processed meat consumption (servings/d) 0.40 0.33 0.26 0.20 0.13 
Folate intake (μg/d) 219 260 295 328 377 
Glycemic load 9.3 9.8 10.2 10.6 11.2 
Calcium intake (mg/d)§ 857 913 979 1,051 1,138 
Methionine intake (mg/d)§ 1.8 1.9 1.9 1.9 1.8 
Multivitamin supplement use (%)      
    Never 62.0 59.0 55.2 52.3 49.1 
    Past 0.1 0.0 0.0 0.0 0.1 
    Current, <5 y 11.8 12.4 13.1 13.7 14.5 
    Current, ≥5 y 19.7 23.5 27.0 29.2 30.4 
Premenopausal (%) 13.7 15.5 15.2 14.1 13.2 
Postmenopausal hormone use (among postmenopausal women, %)      
    Never 26.5 25.0 23.8 22.6 22.3 
    Past 13.8 14.7 15.0 15.0 14.2 
    Current 21.9 25.9 28.7 29.8 28.5 
Aspirin use, d/mo (%)      
    Never 19.3 18.4 18.8 19.7 22.5 
    1-14 33.4 35.6 36.1 36.7 35.2 
    >14 17.1 16.5 16.3 16.4 15.9 
Duration of aspirin use, y (%),      
    1-4 14.4 13.6 14.0 14.9 16.1 
    5-10 14.7 15.5 16.0 15.9 15.7 
    >10 32.4 35.4 35.3 34.4 30.9 
Men      
Age 56.7 58.3 60.1 61.6 62.5 
Family history of colon cancer (%) 11.3 11.8 12.1 12.5 12.8 
Height (m) 178.2 178.3 178.2 178.1 177.8 
Body mass index (kg/m226.0 26.0 25.8 25.5 25.1 
History of sigmoidoscopy (%) 36.8 43.0 46.2 48.5 48.5 
Physical activity (METS) 24.0 28.6 32.0 34.5 38.4 
Smoking (pack-years before age 30) 8.0 6.5 5.7 5.3 5.0 
Alcohol intake (g/d) 18.5 12.5 9.7 8.1 6.1 
Total caloric intake/d 2,068 2,035 1,975 1,895 1,818 
Red meat consumption (servings/d)* 0.36 0.29 0.24 0.19 0.12 
Processed meat consumption (servings/d) 0.52 0.41 0.31 0.22 0.13 
Folate intake (μg/d) 281 331 372 405 458 
Glycemic load 11.2 12.1 12.8 13.5 14.7 
Calcium intake (mg/d)§ 845 868 899 943 1,005 
Methionine intake (mg/d)§ 2.1 2.2 2.1 2.1 2.0 
Multivitamin supplement use (%)      
    Never 39.5 34.9 31.6 28.9 25.7 
    Past, ≤5 y 0.9 0.7 0.7 0.7 0.8 
    Past, ≥5 y 0.3 0.2 0.3 0.3 0.3 
    Current, <5 y 10.9 12.8 13.6 14.3 14.7 
    Current, ≥5 y 17.7 20.0 22.5 24.6 26.4 
Aspirin use, d/mo (%)      
    Never 16.5 17.7 17.5 17.8 18.6 
    1-14 19.8 22.0 21.4 20.0 17.2 
    >14 23.3 25.5 27.0 28.1 26.8 
Duration of aspirin use, y (%),      
    1-4 24.0 24.2 24.5 24.2 23.1 
    5-10 18.4 21.6 22.1 22.0 19.5 
    >10 4.8 6.3 6.8 6.6 5.7 
AOAC fiber intake/1,000 cal
<8.0 g8.0-10.0 g10.0-12.0 g12.0-14.0 g>14.0 g
Women      
Age 54.2 56.1 58.1 59.8 60.9 
Family history of colon cancer (%) 15.0 15.6 16.0 16.3 16.3 
Height (cm) 163.9 163.9 163.9 163.9 163.6 
Body mass index (kg/m225.3 25.6 25.6 25.3 24.8 
History of sigmoidoscopy (%) 25.1 28.7 31.8 33.2 34.3 
Physical activity (h/wk) 1.7 2.0 2.3 2.7 3.1 
Smoking (pack-years >35 y in the past) 2.2 1.8 1.7 1.6 1.5 
Alcohol intake (g/d) 10.6 6.3 4.7 3.9 3.0 
Total caloric intake/d 1,810 1,793 1,754 1,698 1,608 
Red meat consumption (servings/d)* 0.35 0.30 0.26 0.22 0.16 
Processed meat consumption (servings/d) 0.40 0.33 0.26 0.20 0.13 
Folate intake (μg/d) 219 260 295 328 377 
Glycemic load 9.3 9.8 10.2 10.6 11.2 
Calcium intake (mg/d)§ 857 913 979 1,051 1,138 
Methionine intake (mg/d)§ 1.8 1.9 1.9 1.9 1.8 
Multivitamin supplement use (%)      
    Never 62.0 59.0 55.2 52.3 49.1 
    Past 0.1 0.0 0.0 0.0 0.1 
    Current, <5 y 11.8 12.4 13.1 13.7 14.5 
    Current, ≥5 y 19.7 23.5 27.0 29.2 30.4 
Premenopausal (%) 13.7 15.5 15.2 14.1 13.2 
Postmenopausal hormone use (among postmenopausal women, %)      
    Never 26.5 25.0 23.8 22.6 22.3 
    Past 13.8 14.7 15.0 15.0 14.2 
    Current 21.9 25.9 28.7 29.8 28.5 
Aspirin use, d/mo (%)      
    Never 19.3 18.4 18.8 19.7 22.5 
    1-14 33.4 35.6 36.1 36.7 35.2 
    >14 17.1 16.5 16.3 16.4 15.9 
Duration of aspirin use, y (%),      
    1-4 14.4 13.6 14.0 14.9 16.1 
    5-10 14.7 15.5 16.0 15.9 15.7 
    >10 32.4 35.4 35.3 34.4 30.9 
Men      
Age 56.7 58.3 60.1 61.6 62.5 
Family history of colon cancer (%) 11.3 11.8 12.1 12.5 12.8 
Height (m) 178.2 178.3 178.2 178.1 177.8 
Body mass index (kg/m226.0 26.0 25.8 25.5 25.1 
History of sigmoidoscopy (%) 36.8 43.0 46.2 48.5 48.5 
Physical activity (METS) 24.0 28.6 32.0 34.5 38.4 
Smoking (pack-years before age 30) 8.0 6.5 5.7 5.3 5.0 
Alcohol intake (g/d) 18.5 12.5 9.7 8.1 6.1 
Total caloric intake/d 2,068 2,035 1,975 1,895 1,818 
Red meat consumption (servings/d)* 0.36 0.29 0.24 0.19 0.12 
Processed meat consumption (servings/d) 0.52 0.41 0.31 0.22 0.13 
Folate intake (μg/d) 281 331 372 405 458 
Glycemic load 11.2 12.1 12.8 13.5 14.7 
Calcium intake (mg/d)§ 845 868 899 943 1,005 
Methionine intake (mg/d)§ 2.1 2.2 2.1 2.1 2.0 
Multivitamin supplement use (%)      
    Never 39.5 34.9 31.6 28.9 25.7 
    Past, ≤5 y 0.9 0.7 0.7 0.7 0.8 
    Past, ≥5 y 0.3 0.2 0.3 0.3 0.3 
    Current, <5 y 10.9 12.8 13.6 14.3 14.7 
    Current, ≥5 y 17.7 20.0 22.5 24.6 26.4 
Aspirin use, d/mo (%)      
    Never 16.5 17.7 17.5 17.8 18.6 
    1-14 19.8 22.0 21.4 20.0 17.2 
    >14 23.3 25.5 27.0 28.1 26.8 
Duration of aspirin use, y (%),      
    1-4 24.0 24.2 24.5 24.2 23.1 
    5-10 18.4 21.6 22.1 22.0 19.5 
    >10 4.8 6.3 6.8 6.6 5.7 
*

Beef, pork, or lamb as a main dish.

From dietary sources.

Calorie-adjusted.

§

From dietary sources and supplements.

Column percent do not add up to 100% because of missing values.

Among users.

Age-adjusted analyses indicated a significant inverse association between fiber intake and colorectal cancer, with a significant trend across categories of intake for both women and men (Table 2). This association was similar for AOAC and Englyst fiber. An increase of 5 g/d in fiber intake was associated with a HR of 0.90 [95% confidence interval (95% CI), 0.86-0.95] for AOAC fiber and of 0.89 (95% CI, 0.84-0.95) for Englyst fiber among the entire study population (Table 2).

Table 2.

A. Cumulative updated AOAC fiber intake and hazard ratio of colorectal cancer among women participating in the NHS (1984-2000) and men participating in the HPFS (1986-2000)

AOAC fiber intake per 1,000 cal per d
PtrendHR*for a 5-g increase in fiber intake
<8.0 g8.0-10.0 g10.0-12.0 g12.0-14.0 g>14.0 g
Age-adjusted HR        
    Women 1.00 (reference) 1.05 (0.84-1.31) 0.99 (0.80-1.24) 0.88 (0.69-1.13) 0.74 (0.56-0.96) <0.001 0.89 (0.83-0.96) 
    Men 1.00 (reference) 0.78 (0.62-0.99) 0.66 (0.52-0.84) 0.69 (0.53-0.89) 0.68 (0.52-0.87) <0.001 0.91 (0.86-0.97) 
    Pooled (95% CI) 1.00 (reference) 0.91 (0.78-1.07)  0.78 (0.66-0.93) 0.70 (0.59-0.85) NA 0.90 (0.86-0.95) 
Covariate-adjusted HR§        
    Women 1.00 (reference) 1.13 (0.90-1.43) 1.15 (0.90-1.49) 1.11 (0.83-1.48) 0.98 (0.70-1.37) 0.70 0.98 (0.88-1.08) 
    Men 1.00 (reference) 0.83 (0.65-1.06) 0.75 (0.57-0.99) 0.83 (0.61-1.14) 0.91 (0.65-1.28) 0.86 0.97 (0.89-1.05) 
    Pooled (95% CI) 1.00 (reference) 0.98 (0.83-1.16)  0.97 (0.78-1.20) 0.94 (0.74-1.20) NA 0.97 (0.91-1.04) 
Total no. cases 249 420 426 277 224  1,596 
Total person-years of follow-up
 
291,167
 
492,787
 
470,669
 
285,547
 
236,330
 

 
1,776,500
 
B. Cumulative updated Englyst fiber intake and HR of colorectal cancer among women participating in the NHS (1984-2000) and men participating in the HPFS (1986-2000)
 
       
 Englyst fiber intake per 1,000 cal per d
 
    Ptrend HR* for a 5-g increase in fiber intake 

 
<6.0 g
 
6.0–7.5 g
 
7.5-9.0 g
 
9.0-10.5 g
 
>10.5 g
 

 
 
Age-adjusted HR        
    Women 1.00 (reference) 0.97 (0.79-1.18) 0.98 (0.80-1.20) 0.85 (0.68-1.07) 0.76 (0.58-0.98) 0.02 0.88 (0.79-0.97) 
    Men 1.00 (reference) 0.82 (0.65-1.02) 0.72 (0.57-0.91) 0.71 (0.55-0.92) 0.72 (0.56-0.93) 0.01 0.90 (0.83-0.97) 
    Pooled (95% CI) 1.00 (reference) 0.90 (0.78-1.04) 0.86 (0.74-1.00) 0.79 (0.66-0.93) 0.74 (0.61-0.89) <0.001 0.89 (0.84-0.95) 
Covariate-adjusted HR§        
    Women 1.00 (reference) 1.06 (0.86-1.31) 1.17 (0.93-1.48) 1.09 (0.83-1.44) 1.04 (0.75-1.44) 0.75 1.00 (0.88-1.14) 
    Men 1.00 (reference) 0.89 (0.70-1.12) 0.83 (0.64-1.08) 0.88 (0.65-1.20) 0.96 (0.69-1.34) 0.93 0.98 (0.89-1.09) 
    Pooled (95% CI) 1.00 (reference) 0.98 (0.84-1.15)  0.99 (0.81-1.22) 1.00 (0.79-1.26) NA 0.99 (0.91-1.07) 
Total no. cases 309 432 414 243 198  1,596 
Total person-years of follow-up 372,343 511,454 441,137 251,082 200,482  1,776,498 
AOAC fiber intake per 1,000 cal per d
PtrendHR*for a 5-g increase in fiber intake
<8.0 g8.0-10.0 g10.0-12.0 g12.0-14.0 g>14.0 g
Age-adjusted HR        
    Women 1.00 (reference) 1.05 (0.84-1.31) 0.99 (0.80-1.24) 0.88 (0.69-1.13) 0.74 (0.56-0.96) <0.001 0.89 (0.83-0.96) 
    Men 1.00 (reference) 0.78 (0.62-0.99) 0.66 (0.52-0.84) 0.69 (0.53-0.89) 0.68 (0.52-0.87) <0.001 0.91 (0.86-0.97) 
    Pooled (95% CI) 1.00 (reference) 0.91 (0.78-1.07)  0.78 (0.66-0.93) 0.70 (0.59-0.85) NA 0.90 (0.86-0.95) 
Covariate-adjusted HR§        
    Women 1.00 (reference) 1.13 (0.90-1.43) 1.15 (0.90-1.49) 1.11 (0.83-1.48) 0.98 (0.70-1.37) 0.70 0.98 (0.88-1.08) 
    Men 1.00 (reference) 0.83 (0.65-1.06) 0.75 (0.57-0.99) 0.83 (0.61-1.14) 0.91 (0.65-1.28) 0.86 0.97 (0.89-1.05) 
    Pooled (95% CI) 1.00 (reference) 0.98 (0.83-1.16)  0.97 (0.78-1.20) 0.94 (0.74-1.20) NA 0.97 (0.91-1.04) 
Total no. cases 249 420 426 277 224  1,596 
Total person-years of follow-up
 
291,167
 
492,787
 
470,669
 
285,547
 
236,330
 

 
1,776,500
 
B. Cumulative updated Englyst fiber intake and HR of colorectal cancer among women participating in the NHS (1984-2000) and men participating in the HPFS (1986-2000)
 
       
 Englyst fiber intake per 1,000 cal per d
 
    Ptrend HR* for a 5-g increase in fiber intake 

 
<6.0 g
 
6.0–7.5 g
 
7.5-9.0 g
 
9.0-10.5 g
 
>10.5 g
 

 
 
Age-adjusted HR        
    Women 1.00 (reference) 0.97 (0.79-1.18) 0.98 (0.80-1.20) 0.85 (0.68-1.07) 0.76 (0.58-0.98) 0.02 0.88 (0.79-0.97) 
    Men 1.00 (reference) 0.82 (0.65-1.02) 0.72 (0.57-0.91) 0.71 (0.55-0.92) 0.72 (0.56-0.93) 0.01 0.90 (0.83-0.97) 
    Pooled (95% CI) 1.00 (reference) 0.90 (0.78-1.04) 0.86 (0.74-1.00) 0.79 (0.66-0.93) 0.74 (0.61-0.89) <0.001 0.89 (0.84-0.95) 
Covariate-adjusted HR§        
    Women 1.00 (reference) 1.06 (0.86-1.31) 1.17 (0.93-1.48) 1.09 (0.83-1.44) 1.04 (0.75-1.44) 0.75 1.00 (0.88-1.14) 
    Men 1.00 (reference) 0.89 (0.70-1.12) 0.83 (0.64-1.08) 0.88 (0.65-1.20) 0.96 (0.69-1.34) 0.93 0.98 (0.89-1.09) 
    Pooled (95% CI) 1.00 (reference) 0.98 (0.84-1.15)  0.99 (0.81-1.22) 1.00 (0.79-1.26) NA 0.99 (0.91-1.07) 
Total no. cases 309 432 414 243 198  1,596 
Total person-years of follow-up 372,343 511,454 441,137 251,082 200,482  1,776,498 
*

HR and 95% CI.

Data were not combined because of statistically significant heterogeneity (at P < .05) in HR estimates from women and men.

Not applicable.

§

HR and 95% CI adjusted for age (1-y intervals), time period, family history of colorectal cancer (binary), history of sigmoidoscopy or colonoscopy (using an indicator for each 2-y questionnaire cycle to allow for multiple sigmoidoscopies or colonoscopies), height (continuous), body mass index (weight/height2, continuous), physical activity (women: 1-2, 2-4, 4-7, >7 h/wk; men: in METS/wk, quintiles), regular aspirin use (none, 1-14, and ≥15 d/mo), duration of aspirin use (<1, 1-4, 5-10, >10 y), pack-years of early-onset smoking (women: ≥35 y in the past: none, <10, and ≥10 pack-years; men: before age 30: <10 and >10 pack-years), multivitamin supplement use (never, past <5 y, past ≥5 y, current <5 y, current ≥5 y), total caloric intake (continuous), alcohol consumption (women: none, 0.1-4.9, 5.0-14.9, ≥15 g/d; men: none, <10, 10-19.9, 20-29.9, ≥30 g/d), dietary folate (quintiles), red meat consumption (never, <1/mo, 1-3/mo, 2-4/wk, ≥5/wk), processed meat (never, <1/mo, 1-3/mo, 2-4/wk, ≥5/wk), glycemic load (continuous), calcium (<700 and ≥700 mg), methionine (continuous) and (among women) menopausal status and postmenopausal hormone use (never, current, and past).

When potential confounders of the association between fiber intake and colorectal cancer assessed in our studies were controlled in the analysis, the association largely disappeared (Table 2). The most important confounding factors were intake of folate, red meat consumption, processed meat consumption, and glycemic load. Overall, a 5-g increase in fiber intake per day was associated with a fully covariate-adjusted HR of 0.97 (95% CI, 0.91-1.04) for AOAC fiber and of 0.99 (95% CI, 0.91-1.07) for Englyst fiber among the entire study population (Table 2).

Following the analytic model used in EPIC, we observed results similar to those found in EPIC (Table 3). The EPIC investigators reported an HR of 0.75 (95% CI, 0.59-0.95) for the highest quintile of fiber intake compared with the lowest quintile, controlling for age, weight, height, and total caloric intake (15). The respective HR value from our population with the same covariate adjustments was 0.69 (95% CI, 0.58-0.82; Table 3). In these analyses, fiber intake assessed at baseline was used to most closely mimic the analytic model used in EPIC. After fully adjusting the model for all covariates used in our analyses, this value changed to 0.85 (95% CI, 0.70-1.04) in our data when using fiber assessed at baseline and 0.94 (95% CI, 0.76-1.17) when using fiber that was cumulatively updated.

Table 3.

Baseline AOAC fiber intake and HR of colorectal cancer among women participating in the NHS (1984-2000) and men participating in the HPFS (1986-2000) using the same categorization for fiber intake and covariate adjustment as used in the EPIC study (age-adjusted and limited covariate adjusted models) and using the categorization as used in the EPIC study but fully adjusting for potential confounding variables (fully covariate-adjusted models)

Quintiles of fiber intake
PtrendHR*for a 5-g increase in fiber intake
Q1Q2Q3Q4Q5
Mean baseline fiber intake/1,000 calories (g)        
    Women (range) 6.5 (0.5-7.7) 8.4 (7.7-9.1) 9.8 (9.1-10.5) 11.4 (10.5-12.4) 14.9 (12.4-35.0)   
    Men (range) 6.5 (0.2-7.8) 8.6 (7.8-9.3) 10.1 (9.3-10.9) 11.9 (10.9-13.1) 16.1 (13.1-35.0)   
Age-adjusted HR*        
    Women 1.00 (reference) 0.97 (0.78-1.20) 1.01 (0.82-1.25) 1.01 (0.83-1.24) 0.74 (0.60-0.92) 0.01 0.95 (0.91-1.00) 
    Men 1.00 (reference) 0.74 (0.58-0.94) 0.64 (0.50-0.82) 0.73 (0.58-0.92) 0.63 (0.49-0.80) <0.001 0.96 (0.92-1.01) 
    Pooled (95% CI) 1.00 (reference) 0.86 (0.73-1.01)   0.69 (0.59-0.81) NA 0.96 (0.93-0.99) 
Limited covariate-adjusted HR§        
    Women 1.00 (reference) 0.97 (0.78-1.19) 1.01 (0.82-1.25) 1.01 (0.82-1.24) 0.74 (0.59-0.93) 0.02 0.92 (0.86-0.99) 
    Men 1.00 (reference) 0.74 (0.58-0.94) 0.65 (0.51-0.83) 0.73 (0.57-0.92) 0.62 (0.48-0.81) <0.001 0.90 (0.85-0.96) 
    Pooled (95% CI) 1.00 (reference) 0.86 (0.73-1.01)   0.69 (0.58-0.82) NA 0.91 (0.87-0.95) 
    Total no. cases 317 299 304 357 299  1,576 
Total person-years of follow-up 353,220 354,742 353,944 351,729 348,570  1,762,205 
Fully covariate-adjusted HR        
    Baseline fiber        
    Women 1.00 (reference) 1.01 (0.81-1.26) 1.11 (0.89-1.38) 1.17 (0.93-1.46) 0.90 (0.70-1.17) 0.57 0.98 (0.92-1.04) 
    Men 1.00 (reference) 0.79 (0.61-1.01) 0.71 (0.54-0.93) 0.85 (0.65-1.12) 0.79 (0.59-1.07) 0.36 1.00 (0.94-1.06) 
    Pooled 1.00 (reference) 0.91 (0.77-1.07)  1.02 (0.86-1.22) 0.85 (0.70-1.04) NA 0.99 (0.95-1.04) 
    Total no. cases 317 299 304 357 299  1,576 
    Total person-years of follow-up 353,220 354,742 353,944 351,729 348,570  1,762,205 
Updated fiber        
    Women 1.00 (reference) 1.13 (0.90-1.40) 1.06 (0.84-1.34) 1.17 (0.92-1.50) 0.94 (0.70-1.25) 0.61 0.98 (0.90-1.07) 
    Men 1.00 (reference) 0.83 (0.64-1.08) 0.77 (0.58-1.02) 0.84 (0.63-1.12) 0.95 (0.69-1.32) 0.90 1.00 (0.92-1.07) 
    Pooled 1.00 (reference) 0.99 (0.84-1.17) 0.93 (0.77-1.11) 1.02 (0.84-1.23) 0.94 (0.76-1.17) 0.78 0.99 (0.93-1.05) 
    Total no. cases 306 310 304 348 308  1,576 
    Total person-years of follow-up 351,277 359,143 357,988 353,834 339,963  1,762,205 
Quintiles of fiber intake
PtrendHR*for a 5-g increase in fiber intake
Q1Q2Q3Q4Q5
Mean baseline fiber intake/1,000 calories (g)        
    Women (range) 6.5 (0.5-7.7) 8.4 (7.7-9.1) 9.8 (9.1-10.5) 11.4 (10.5-12.4) 14.9 (12.4-35.0)   
    Men (range) 6.5 (0.2-7.8) 8.6 (7.8-9.3) 10.1 (9.3-10.9) 11.9 (10.9-13.1) 16.1 (13.1-35.0)   
Age-adjusted HR*        
    Women 1.00 (reference) 0.97 (0.78-1.20) 1.01 (0.82-1.25) 1.01 (0.83-1.24) 0.74 (0.60-0.92) 0.01 0.95 (0.91-1.00) 
    Men 1.00 (reference) 0.74 (0.58-0.94) 0.64 (0.50-0.82) 0.73 (0.58-0.92) 0.63 (0.49-0.80) <0.001 0.96 (0.92-1.01) 
    Pooled (95% CI) 1.00 (reference) 0.86 (0.73-1.01)   0.69 (0.59-0.81) NA 0.96 (0.93-0.99) 
Limited covariate-adjusted HR§        
    Women 1.00 (reference) 0.97 (0.78-1.19) 1.01 (0.82-1.25) 1.01 (0.82-1.24) 0.74 (0.59-0.93) 0.02 0.92 (0.86-0.99) 
    Men 1.00 (reference) 0.74 (0.58-0.94) 0.65 (0.51-0.83) 0.73 (0.57-0.92) 0.62 (0.48-0.81) <0.001 0.90 (0.85-0.96) 
    Pooled (95% CI) 1.00 (reference) 0.86 (0.73-1.01)   0.69 (0.58-0.82) NA 0.91 (0.87-0.95) 
    Total no. cases 317 299 304 357 299  1,576 
Total person-years of follow-up 353,220 354,742 353,944 351,729 348,570  1,762,205 
Fully covariate-adjusted HR        
    Baseline fiber        
    Women 1.00 (reference) 1.01 (0.81-1.26) 1.11 (0.89-1.38) 1.17 (0.93-1.46) 0.90 (0.70-1.17) 0.57 0.98 (0.92-1.04) 
    Men 1.00 (reference) 0.79 (0.61-1.01) 0.71 (0.54-0.93) 0.85 (0.65-1.12) 0.79 (0.59-1.07) 0.36 1.00 (0.94-1.06) 
    Pooled 1.00 (reference) 0.91 (0.77-1.07)  1.02 (0.86-1.22) 0.85 (0.70-1.04) NA 0.99 (0.95-1.04) 
    Total no. cases 317 299 304 357 299  1,576 
    Total person-years of follow-up 353,220 354,742 353,944 351,729 348,570  1,762,205 
Updated fiber        
    Women 1.00 (reference) 1.13 (0.90-1.40) 1.06 (0.84-1.34) 1.17 (0.92-1.50) 0.94 (0.70-1.25) 0.61 0.98 (0.90-1.07) 
    Men 1.00 (reference) 0.83 (0.64-1.08) 0.77 (0.58-1.02) 0.84 (0.63-1.12) 0.95 (0.69-1.32) 0.90 1.00 (0.92-1.07) 
    Pooled 1.00 (reference) 0.99 (0.84-1.17) 0.93 (0.77-1.11) 1.02 (0.84-1.23) 0.94 (0.76-1.17) 0.78 0.99 (0.93-1.05) 
    Total no. cases 306 310 304 348 308  1,576 
    Total person-years of follow-up 351,277 359,143 357,988 353,834 339,963  1,762,205 
*

HR and 95% CI.

Data were not combined because of statistically significant heterogeneity (at P < .05) in HR estimates from women and men.

Not applicable.

§

HR and 95% CI adjusted for age, height, weight, and caloric intake from fat and caloric intake from sources other than fat.

HR and 95% CI adjusted for age, time period, family history of colorectal cancer, history of sigmoidoscopy or colonoscopy, height, BMI , physical activity, regular aspirin use, duration of aspirin use, pack-years of early-onset smoking, multivitamin supplement use, caloric intake from fat and caloric intake from sources other than fat, alcohol consumption dietary folate, red meat consumption, processed meat, glycemic load, calcium, methionine, and (among women) menopausal status and postmenopausal hormone use.

Baseline fiber: 1984 for NHS and 1986 for HPFS. Updated fiber: cumulatively updated fiber 1984-1986-1990-1994 for NHS and 1986-1990-1994 for HPFS.

Using our fully adjusted analytic model, results for colon and rectal cancers considered separately were comparable to those for colorectal cancer (Table 4). Cereal, fruit, and vegetable fiber were also considered separately. Fruit fiber was inversely related to colorectal cancer, and the association was of borderline statistical significance (covariate-adjusted HR, 0.89; 95% CI, 0.80-1.00 for a 5-g increase in intake), but no significant trend across quintiles emerged (P = 0.20; Table 5). Neither cereal nor vegetable fiber was notably related to colorectal cancer among women or men.

Table 4.

Cumulative updated AOAC fiber intake and hazard ratio of colon and rectal cancer among women participating in the NHS (1984-2000) and men participating in the HPFS (1986-2000)

AOAC fiber intake per 1,000 cal per d
PtrendHR* for a 5-g increase in fiber intake
<8.0 g8.0–10.0 g10.0-12.0 g12.0-14.0 g>14.0 g
Colon cancer        
Covariate-adjusted HR        
    Women 1.00 (reference) 1.18 (0.91-1.53) 1.24 (0.93-1.64) 1.18 (0.86-1.63) 0.95 (0.65-1.39) 0.63 0.98 (0.88-1.10) 
    Men 1.00 (reference) 0.85 (0.63-1.15) 0.78 (0.56-1.09) 0.94 (0.65-1.36) 0.85 (0.56-1.30) 0.76 0.96 (0.87-1.06) 
    Pooled (95% CI) 1.00 (reference) 1.03 (0.84-1.25)  1.07 (0.84-1.37) 0.91 (0.68-1.20) NA§ 0.97 (0.90-1.05) 
Total no. cases 180 316 332 220 154  1,202 
Total person-years of follow-up 291,306 493,094 470,957 285,667 236,520  1,777,544 
Rectal cancer        
Covariate-adjusted HR        
    Women 1.00 (reference) 0.95 (0.57-1.59) 0.85 (0.48-1.50) 0.84 (0.44-1.64) 1.10 (0.52-2.29) 0.91 0.95 (0.75-1.20) 
    Men 1.00 (reference) 1.16 (0.67-2.03) 0.79 (0.41-1.51) 0.89 (0.43-1.85) 1.34 (0.62-2.89) 0.58 0.99 (0.83-1.18) 
    Pooled (95% CI) 1.00 (reference) 1.04 (0.72-1.52) 0.82 (0.54-1.26) 0.87 (0.53-1.42) 1.21 (0.71-2.05) 0.63 0.98 (0.85-1.12) 
Total no. cases 49 89 72 46 54  310 
Total person-years of follow-up 291,711 493,901 471,996 286,411 236,916  1,780,935 
AOAC fiber intake per 1,000 cal per d
PtrendHR* for a 5-g increase in fiber intake
<8.0 g8.0–10.0 g10.0-12.0 g12.0-14.0 g>14.0 g
Colon cancer        
Covariate-adjusted HR        
    Women 1.00 (reference) 1.18 (0.91-1.53) 1.24 (0.93-1.64) 1.18 (0.86-1.63) 0.95 (0.65-1.39) 0.63 0.98 (0.88-1.10) 
    Men 1.00 (reference) 0.85 (0.63-1.15) 0.78 (0.56-1.09) 0.94 (0.65-1.36) 0.85 (0.56-1.30) 0.76 0.96 (0.87-1.06) 
    Pooled (95% CI) 1.00 (reference) 1.03 (0.84-1.25)  1.07 (0.84-1.37) 0.91 (0.68-1.20) NA§ 0.97 (0.90-1.05) 
Total no. cases 180 316 332 220 154  1,202 
Total person-years of follow-up 291,306 493,094 470,957 285,667 236,520  1,777,544 
Rectal cancer        
Covariate-adjusted HR        
    Women 1.00 (reference) 0.95 (0.57-1.59) 0.85 (0.48-1.50) 0.84 (0.44-1.64) 1.10 (0.52-2.29) 0.91 0.95 (0.75-1.20) 
    Men 1.00 (reference) 1.16 (0.67-2.03) 0.79 (0.41-1.51) 0.89 (0.43-1.85) 1.34 (0.62-2.89) 0.58 0.99 (0.83-1.18) 
    Pooled (95% CI) 1.00 (reference) 1.04 (0.72-1.52) 0.82 (0.54-1.26) 0.87 (0.53-1.42) 1.21 (0.71-2.05) 0.63 0.98 (0.85-1.12) 
Total no. cases 49 89 72 46 54  310 
Total person-years of follow-up 291,711 493,901 471,996 286,411 236,916  1,780,935 
*

HR and 95% CI.

HR and 95% CI adjusted for age, time period, family history of colorectal cancer, history of sigmoidoscopy or colonoscopy, height, body mass index, physical activity, regular aspirin use, duration of aspirin use, pack-years of early-onset smoking, multivitamin supplement use, total caloric intake, alcohol consumption dietary folate, red meat consumption, processed meat, glycemic load, calcium, methionine, and (among women) menopausal status and postmenopausal hormone use.

Data were not combined because of statistically significant heterogeneity (at P < .05) in HR estimates from women and men.

§

Not applicable.

Table 5.

Cumulative updated cereal, fruit, and vegetable fiber intake and hazard ratio of colon and rectal cancer among women participating in the NHS (1984-2000) and men participating in the HPFS (1986-2000)

Quintiles of fiber intake
PtrendHR* for a 5-g increase in fiber intake
Q1Q2Q3Q4Q5
Cereal fiber        
Mean intake/1,000 cal (g)        
    Women 2.3 3.3 4.1 5.2 8.0   
    Men 2.8 4.3 5.6 7.3 11.45   
Covariate-adjusted HR        
    Women 1.00 (reference) 0.81 (0.66-1.00) 0.88 (0.71-1.08) 0.86 (0.69-1.07) 0.89 (0.71-1.12) 0.63 0.94 (0.79-1.11) 
    Men 1.00 (reference) 0.81 (0.63-1.04) 0.98 (0.76-1.25) 0.89 (0.69-1.16) 0.79 (0.60-1.05) 0.19 0.93 (0.81-1.06) 
    Pooled (95% CI) 1.00 (reference) 0.81 (0.69-0.95) 0.92 (0.78-1.07) 0.87 (0.74-1.03) 0.85 (0.71-1.02) 0.21 0.93 (0.84-1.03) 
Total number of cases 360 289 327 316 304  1,596 
Total person-years of follow-up 343,829 360,644 362,529 359,991 349,169  1,776,162 
        
Fruit fiber        
Mean intake/1,000 cal (g)        
    Women 1.4 2.5 3.5 4.7 7.3   
    Men 1.4 2.8 4.1 5.6 9.3   
Covariate-adjusted HR        
    Women 1.00 (reference) 0.94 (0.75-1.16) 1.05 (0.84-1.31) 0.85 (0.67-1.07) 0.88 (0.68-1.13) 0.20 0.87 (0.73-1.04) 
    Men 1.00 (reference) 1.02 (0.79-1.33) 0.91 (0.70-1.19) 1.08 (0.82-1.41) 0.92 (0.68-1.23) 0.62 0.91 (0.79-1.06) 
    Pooled (95% CI) 1.00 (reference) 0.97 (0.82-1.15) 0.99 (0.84-1.18) 0.94 (0.79-1.13) 0.89 (0.74-1.08) 0.20 0.89 (0.80-1.00) 
Total no. cases 290 307 334 333 326  1,590 
Total person-years of follow-up 348,760 360,786 360,417 357,803 344,148  1,771,914 
        
Vegetable fiber        
Mean intake/1,000 cal (g)        
    Women 3.6 5.0 6.0 7.2 10.0   
    Men 3.6 5.3 6.6 8.3 12.2   
Covariate-adjusted HR        
    Women 1.00 (reference) 1.12 (0.90-1.39) 1.07 (0.85-1.34) 1.26 (1.01-1.59) 1.20 (0.94-1.56) 0.11 1.05 (0.90-1.23) 
    Men 1.00 (reference) 1.04 (0.82-1.32) 0.84 (0.65-1.08) 0.93 (0.72-1.20) 1.09 (0.83-1.42) 0.57 1.04 (0.91-1.19) 
    Pooled (95% CI) 1.00 (reference) 1.08 (0.92-1.27) 0.96 (0.81-1.14) 1.11 (0.93-1.31) 1.15 (0.96-1.38) 0.12 1.05 (0.95-1.16) 
Total no. cases 301 328 294 338 335  1,596 
Total person-years of follow-up 347,891 360,862 363,497 358,089 345,803  1,776,142 
Quintiles of fiber intake
PtrendHR* for a 5-g increase in fiber intake
Q1Q2Q3Q4Q5
Cereal fiber        
Mean intake/1,000 cal (g)        
    Women 2.3 3.3 4.1 5.2 8.0   
    Men 2.8 4.3 5.6 7.3 11.45   
Covariate-adjusted HR        
    Women 1.00 (reference) 0.81 (0.66-1.00) 0.88 (0.71-1.08) 0.86 (0.69-1.07) 0.89 (0.71-1.12) 0.63 0.94 (0.79-1.11) 
    Men 1.00 (reference) 0.81 (0.63-1.04) 0.98 (0.76-1.25) 0.89 (0.69-1.16) 0.79 (0.60-1.05) 0.19 0.93 (0.81-1.06) 
    Pooled (95% CI) 1.00 (reference) 0.81 (0.69-0.95) 0.92 (0.78-1.07) 0.87 (0.74-1.03) 0.85 (0.71-1.02) 0.21 0.93 (0.84-1.03) 
Total number of cases 360 289 327 316 304  1,596 
Total person-years of follow-up 343,829 360,644 362,529 359,991 349,169  1,776,162 
        
Fruit fiber        
Mean intake/1,000 cal (g)        
    Women 1.4 2.5 3.5 4.7 7.3   
    Men 1.4 2.8 4.1 5.6 9.3   
Covariate-adjusted HR        
    Women 1.00 (reference) 0.94 (0.75-1.16) 1.05 (0.84-1.31) 0.85 (0.67-1.07) 0.88 (0.68-1.13) 0.20 0.87 (0.73-1.04) 
    Men 1.00 (reference) 1.02 (0.79-1.33) 0.91 (0.70-1.19) 1.08 (0.82-1.41) 0.92 (0.68-1.23) 0.62 0.91 (0.79-1.06) 
    Pooled (95% CI) 1.00 (reference) 0.97 (0.82-1.15) 0.99 (0.84-1.18) 0.94 (0.79-1.13) 0.89 (0.74-1.08) 0.20 0.89 (0.80-1.00) 
Total no. cases 290 307 334 333 326  1,590 
Total person-years of follow-up 348,760 360,786 360,417 357,803 344,148  1,771,914 
        
Vegetable fiber        
Mean intake/1,000 cal (g)        
    Women 3.6 5.0 6.0 7.2 10.0   
    Men 3.6 5.3 6.6 8.3 12.2   
Covariate-adjusted HR        
    Women 1.00 (reference) 1.12 (0.90-1.39) 1.07 (0.85-1.34) 1.26 (1.01-1.59) 1.20 (0.94-1.56) 0.11 1.05 (0.90-1.23) 
    Men 1.00 (reference) 1.04 (0.82-1.32) 0.84 (0.65-1.08) 0.93 (0.72-1.20) 1.09 (0.83-1.42) 0.57 1.04 (0.91-1.19) 
    Pooled (95% CI) 1.00 (reference) 1.08 (0.92-1.27) 0.96 (0.81-1.14) 1.11 (0.93-1.31) 1.15 (0.96-1.38) 0.12 1.05 (0.95-1.16) 
Total no. cases 301 328 294 338 335  1,596 
Total person-years of follow-up 347,891 360,862 363,497 358,089 345,803  1,776,142 
*

HR and 95% CI.

HR and 95% CI adjusted for age, time period, family history of colorectal cancer, history of sigmoidoscopy or colonoscopy, height, body mass index, physical activity, regular aspirin use, duration of aspirin use, pack-years of early-onset smoking, multivitamin supplement use, total caloric intake, alcohol consumption dietary folate, red meat consumption, processed meat, glycemic load, calcium, methionine, and (among women) menopausal status and postmenopausal hormone use.

We also conducted analyses among a restricted population excluding individuals who reported ever taking multivitamins or consuming breakfast cereal at least once a week to minimize confounding by folate intake (n = 1,071 cases). Associations were further weakened among this restricted cohort. The covariate-adjusted HR for an increase in fiber intake by 5 g was 0.98 (95% CI, 0.91-1.06) for AOAC fiber among women and men combined.

In these two large prospective cohorts spanning 1.8 million person-years of observations and including 1,596 incident cases of colorectal cancer, we found no important overall association between fiber intake and colorectal cancer after controlling for confounding variables. Age-adjusted analyses as well as analyses controlling for age, height, weight, and total caloric intake as done in the EPIC study indicated significant inverse associations. We observed considerable confounding by dietary and nondietary variables when relating fiber intake and colorectal cancer; when we adjusted for confounding, the associations disappeared. Our cohorts included total person-time of follow-up similar to that of the EPIC study and one third more cases of colorectal cancer than EPIC; thus, the lack of association in our study was not due to lower statistical power. The NHS and HPFS had a mean follow-up time of 15 years; the mean follow-up in EPIC was 4.5 years.

Few cohort studies have assessed as much covariate information as the NHS and the HPFS. The recently published analyses of the EPIC cohort adjusted for age, weight, height, sex, nonfat energy, energy from fat, and study center (15). The authors further stated that smoking and physical activity had no significant effect and therefore did not include them in the analytic model. Physical activity has been relatively consistently related to lower risk of colorectal cancer in other studies (31). Hence, it is possible that physical activity was not captured well with the EPIC questionnaire or the range of physical activity was too limited to show any association. It is possible that residual confounding remains in the data from the EPIC study because some of the most important confounders in our analyses, such as folate (32), calcium (33), and glycemic load, were not controlled in the EPIC analyses (the EPIC authors state that adjustment for red meat and processed meat did not affect their results). In trying to pinpoint nutrients associated with colorectal cancer, it seems particularly important to control for folate intake because intake of fiber and folate, especially in populations with low use of supplements and fortified foods, will generally be from similar foods, namely fruits, vegetables, and whole grains.

Potential confounding variables were included in our analytic model if they were likely a priori risk factors for colorectal cancer and not in the causal pathway between fiber intake and cancer outcomes. Dietary variables included in our multivariate models other than fiber were generally associated with colorectal cancer and their correlation with fiber was not very strong. Thus, overadjustment is not a likely explanation for the different results observed after adjusting for several covariates (34).

In our previous analyses from these cohorts with shorter follow-up, we had observed similar results (13, 14). These analyses were conducted similar to the analyses carried out in the EPIC study: fiber intake and potential confounding variables were used only from the baseline assessment and fiber intake was categorized in quintiles. A considerable number of potential confounding variables were considered in our prior analyses, however.

The similarity in the results of our study and EPIC when adjustments were made for the same covariates (age, height, weight, caloric intake from fat, and caloric intake from sources other than fat) suggests that we were able to measure fiber intake adequately. Observed associations between fiber intake and diverticular disease (35), constipation (36), diabetes (37, 38), and cardiovascular disease (39) in our cohorts lends further support to our ability to assess fiber intake with sufficient precision; in these analyses, we also adjusted for several covariates and associations persisted.

Despite the restriction to a U.S. population, our cohorts displayed considerable variation in fiber intake. Across both cohorts, the mean daily AOAC fiber intake was 20.36 g with a SD of 7.36. In the EPIC cohort, the respective values were 21.97 ±7.38 g. Thus, variations in intake in our cohorts and in EPIC do not differ substantially and are not likely to be a sufficient explanation for the difference in results. Our highest category comprised mean intakes of ∼30 g/d compared with about 12 g/d in the lowest category. The respective comparison made in EPIC was 32 and 13 g. Hence, the absolute intake in the two populations seemed comparable.

We calculated associations using two different methods to calculate fiber intake, AOAC and Englyst fiber, and we obtained similar results with the two methods. In the EPIC cohort, the British sites calculated fiber intake by the Englyst method, whereas all other centers followed the AOAC method. Overall results from EPIC combined these estimates.

Fiber sources may differ across these populations. Europeans may eat different grain sources (e.g., rye bread or flaxseed) than do individuals in the United States, and different fibers (e.g., lignans or wheat bran) may have different associations with disease. Similarly, different types of fruits may be consumed (e.g., lignan content varies by fruit type). It is conceivable that some of the differences across studies are due to different proportions of fiber subcomponents.

Despite the lack of association between fiber intake and colorectal cancer in our cohorts after appropriate adjustment for confounding variables, we can explain considerable variation in colorectal cancer incidence with other modifiable risk factors. Variation in obesity, physical inactivity, alcohol consumption, smoking during early adulthood, red meat consumption, and low intake of folic acid from supplements accounts for 71% of the colon cancers observed among the men in our cohort (40).

Data collected in observational studies, whether dietary or nondietary, are assessed with error (41). Covariates in multivariate models are measured with error and those errors may be correlated across different covariates (42). Error correlations are particularly high among dietary variables; hence, models including several nutrients may create spurious associations in unpredictable direction (43). Using nutrient residuals or nutrient densities as was done in our analyses reduces distortion of estimates by correlated measurement error (43). Furthermore, it is likely that fiber intake was assessed with less measurement error than other dietary covariates, except perhaps alcohol consumption.

The EPIC investigators attempted to correct for measurement error of fiber intake using a 24-hour recall as a “gold standard” (15). Currently available measurement error correction models assume independence of errors of the two methods. Errors in the FFQ and the 24-hour recall, however, are likely correlated (44). Thus, application of traditional measurement error correction models may adjust estimates in an unpredictable direction (43-46).

We did not formally correct for measurement error as appropriate methods to correct for measurement error of repeatedly assessed and averaged dietary data have not yet been developed. Repeated assessments of diet were used in the NHS and the HPFS, which provide a better measure of long-term intake than does our baseline diet although associations with colorectal cancer were not different for baseline or updated diet. The use of cumulatively updated dietary data reduces random (but not systematic) within-person measurement error (16).

In conclusion, prospective cohort studies have fairly consistently indicated no important association between fiber intake and colorectal cancer. Our updated analysis from NHS and HPFS produced results similar to those of the EPIC study when we adjusted for the same covariates; however, when we controlled for additional confounders, the inverse association between fiber intake and colorectal cancer vanished. Nevertheless, because high fiber intake has been associated with other important health outcomes, recommendations to consume generous amounts of whole grains, fruits, and vegetables are well supported.

Grant Support: Department of Health and Human Services, National Cancer Institute, NIH grants CA87969 and CA55075 (Nurses' Health Study and Health Professionals' Follow-up Study) and Department of Health and Human Services, National Institute of Diabetes and Digestive and Kidney Diseases, NIH grant R01 DK54900 (K.B. Michels).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Christopher Staskewicz, MS, for expert programming assistance.

1
Burkitt DP. Related disease: related cause?
Lancet
1969
;
2
:
1229
–31.
2
World Cancer Research Fund/American Institute for Cancer Research. Food, nutrition and prevention of cancer: a global perspective. Washington: American Institute for Cancer Research; 1997. p. 216–51.
3
Howe GR, Benito E, Castelleto R, et al. Dietary intake of fiber and decreased risk of cancers of the colon and rectum: evidence from the combined analysis of 13 case-control studies.
J Nat Cancer Inst
1992
;
84
:
1887
–96.
4
Trock B, Lanza E, Greenwald P. Dietary fiber, vegetables, and colon cancer: critical review and meta-analyses of the epidemiologic evidence.
J Natl Cancer Inst
1990
;
82
:
650
–61.
5
Heilbrun LK, Nomura A, Hankin JH, Stemmermann GN. Diet and colorectal cancer with special reference to fiber intake.
Int J Cancer
1989
;
44
:
1
–6.
6
Steinmetz KA, Kushi LH, Bostick RM, Folsom AR, Potter JD. Vegetables, fruit, and colon cancer in the Iowa Women's Health Study.
Am J Epidemiol
1994
;
139
:
1
–15.
7
Kato I, Akhmedkhanov A, Koenig K, et al. Prospective study of diet and female colorectal cancer: the New York University Women's Health Study.
Nutr Cancer
1997
;
28
:
276
–81.
8
Sellers TA, Bazyk AE, Bostick RM, et al. Diet and risk of colon cancer in a large prospective study of older women: an analysis stratified on family history (Iowa, United States).
Cancer Causes Control
1998
;
9
:
357
–67.
9
Pietinen P, Malila N, Virtanen M, et al. Diet and the risk of colorectal cancer in a cohort of Finnish men.
Cancer Causes Control
1999
;
10
:
387
–96.
10
Terry P, Giovannucci E, Michels KB, et al. Fruit, vegetable, dietary fiber, and risk of colorectal cancer.
J Natl Cancer Inst
2001
;
93
:
525
–33.
11
Mai V, Flood A, Peters U, Lacey JV, Schairer C, Schatzkin A. Dietary fibre and risk of colorectal cancer in the Breast Cancer Detection Demonstration Project (BCDDP) follow-up cohort.
Int J Epidemiol
2003
;
32
:
234
–9.
12
McCullough ML, Robertson AS, Chao A, et al. A prospective study of whole grains, fruits, vegetables and colon cancer risk.
Cancer Causes Control
2003
;
14
:
959
–70.
13
Fuchs CS, Giovannucci EL, Colditz GA, et al. Dietary fiber and the risk of colorectal cancer and adenoma in women.
N Engl J Med
1999
;
340
:
169
–76.
14
Giovannucci E, Rimm EB, Stampfer MJ, Colditz GA, Ascherio A, Willett WC. Intake of fat, meat, and fiber in relation to risk of colon cancer in men.
Cancer Res
1994
;
54
:
2390
–7.
15
Bingham SA, Day NE, Luben R, 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
:
1496
–501.
16
Willett WC. Nutritional epidemiology. 2nd ed. New York: Oxford University Press; 1998.
17
Willett WC, Sampson L, Stampfer MJ, et al. Reproducibility and validity of a semiquantitative food frequency questionnaire.
Am J Epidemiol
1985
;
122
:
51
–65.
18
Rimm EB, Giovannucci EL, Stampfer MJ, Colditz GA, Litin LB, Willett WC. Reproducibility and validity of an expanded self-administered semiquantitative food frequency questionnaire among male health professionals.
Am J Epidemiol
1992
;
135
:
1114
–26.
19
Prosky L, Asp NG, Furda I, DeVries JW, Schweizer TF, Harland BF. Determination of total dietary fiber in foods and food products: collaborative study.
J Assoc Off Anal Chem
1985
;
68
:
677
–9.
20
Englyst HN, Bingham SA, Runswick SA, Collinson E, Cummings JH. Dietary fibre (non-starch polysaccharides) in fruit, vegetables and nuts.
J Hum Nutr Diet
1988
;
1
:
247
–86.
21
Englyst HN, Bingham SA, Runswick SA, Collinson E, Cummings JH. Dietary fibre (non-starch polysaccharides) in cereal.
J Hum Nutr Diet
1989
;
2
:
253
–71.
22
Stampfer MJ, Willett WC, Speizer FE, et al. Test of the National Death Index.
Am J Epidemiol
1984
;
119
:
837
–9.
23
Cox DR. Regression models and life tables.
J R Stat Soc
1972
;
32
:
187
–220.
24
Choi SW, Mason JB. Folate status: effects on pathways of colorectal carcinogenesis.
J Nutr
2002
;
132
:
2413
–8S.
25
Sandhu MS, White IR, McPherson K. Systematic review of the prospective cohort studies on meat consumption and colorectal cancer risk: a meta-analytical approach.
Cancer Epidemiol Biomarkers Prev
2001
;
10
:
439
–46.
26
Norat T, Riboli E. Dairy products and colorectal cancer. A review of possible mechanisms and epidemiological evidence.
Eur J Clin Nutr
2003
;
57
:
1
–17.
27
Harnack L, Jacobs DR Jr, Nicodemus K, Lazovich D, Anderson K, Folsom AR. Relationship of folate, vitamin B-6, vitamin B-12, and methionine intake to incidence of colorectal cancers.
Nutr Cancer
2002
;
43
:
152
–8.
28
McKeown-Eyssen G. Epidemiology of colorectal cancer revisited: are serum triglycerides and/or plasma glucose associated with risk?
Cancer Epidemiol Biomarkers Prev
1994
;
3
:
687
–95.
29
Hu FB, Stampfer MJ, Rimm E, et al. Dietary fat and coronary heart disease: a comparison of approaches for adjusting for total energy intake and modeling repeated dietary measurements.
Am J Epidemiol
1999
;
149
:
531
–40.
30
Hedges LV, Olkin I. Statistical methods for meta-analysis. New York: Academic Press; 1985.
31
Slattery ML, Edwards S, Curtin K. Physical activity and colorectal cancer.
Am J Epidemiol
2003
;
158
:
214
–24.
32
Kim YI. Role of folate in colon cancer development and progression.
J Nutr
2003
;
133
:
3731
–9S.
33
Wu K, Willett WC, Fuchs CS, Colditz GA, Giovannucci EL. Calcium intake and risk of colon cancer in women and men.
J Natl Cancer Inst
2002
;
94
:
437
–46.
34
Day NE, Byar DP, Green SB. Overadjustment in case-control studies.
Am J Epidemiol
1980
;
112
:
696
–706.
35
Aldoori WH, Giovannucci EL, Rockett HR, Sampson L, Rimm EB, Willett WC. A prospective study of dietary fiber types and symptomatic diverticular disease in men.
J Nutr
1998
;
128
:
714
–9.
36
Dukas L, Willett WC, Giovannucci EL. Association between physical activity, fiber intake, and other lifestyle variables and constipation in a study of women.
Am J Gastroenterol
2003
;
98
:
1790
–6.
37
Salmeron J, Manson JE, Stampfer MJ, Colditz GA, Wing AL, Willett WC. Dietary fiber, glycemic load, and risk of non-insulin-dependent diabetes mellitus in women.
JAMA
1997
;
277
:
472
–7.
38
Salmeron J, Ascherio A, Rimm EB, et al. Dietary fiber, glycemic load, and risk of NIDDM in men.
Diabetes Care
1997
;
20
:
545
–50.
39
Wolk A, Manson JE, Stampfer MJ, et al. Long-term intake of dietary fiber and decreased risk of coronary heart disease among women.
JAMA
1999
;
281
:
1998
–2004.
40
Platz EA, Willett WC, Colditz GA, Rimm EB, Spiegelman D, Giovannucci E. Proportion of colon cancer risk that might be preventable in a cohort of middle-aged US men.
Cancer Causes Control
2000
;
11
:
579
–88.
41
Michels KB. A renaissance for measurement error.
Int J Epidemiol
2001
;
30
:
421
–2.
42
Marshall JR, Hastrup JL, Ross JS. Mismeasurement and the resonance of strong confounders: correlated errors.
Am J Epidemiol
1999
;
150
:
88
–96.
43
Michels KB, Bingham SA, Luben R, Welch AA, Day NE. The effect of correlated measurement error in multivariate models of diet.
Am J Epidemiol
2004
;
160
:
59
–67.
44
Kipnis V, Midthune D, Freedman LS, et al. Empirical evidence of correlated biases in dietary assessment instruments and its implications.
Am J Epidemiol
2001
;
153
:
394
–403.
45
Plummer M, Clayton D. Measurement error in dietary assessment: an investigation using covariance structure models. Part I.
Stat Med
1993
;
12
:
925
–35.
46
Plummer M, Clayton D. Measurement error in dietary assessment: an investigation using covariance structure models. Part II.
Stat Med
1993
;
12
:
937
–48.