Intake of dietary fiber, especially from cereal foods, is associated with lower incidence of colon cancer in the HELGA cohort
The role of dietary fiber on the risk of colon and rectal cancer has been investigated in numerous studies, but findings have been inconsistent. The purpose of this study was to examine associations between intake of dietary fiber and risk of incident colon (including distal and proximal colon) and rectal cancer in the prospective Scandinavian HELGA cohort and to determine if fiber source (vegetables, fruits, potatoes, cereals) impacted the association. We included 1,168 incident cases (691 colon, 477 rectal cancer), diagnosed during a median of 11.3 years, among 108,081 cohort members. Sex-specific incidence rate ratios (IRRs) of colon and rectal cancer were related to intake of total or specific fiber source using Cox proportional hazards models. For men, an inverse association was observed between intake of total fiber and the risk of colon cancer per an incremental increase of 10 g day−1, IRR (95% CI): 0.74 (0.64–0.86). Intake of cereal fiber per 2 g day−1 was associated with an IRR of 0.94 (0.91–0.98), which was also seen for intake of cereal fiber from foods with high fiber content (≥5 g per 100 g product), where the IRR per 2 g day−1 was 0.94 (0.90–0.98). In women, intake of cereal fiber per 2 g day−1 was also associated with lower risk of colon cancer, 0.97 (0.93–1.00). No clear associations were seen for rectal cancer. Our data indicate a protective role of total and cereal fiber intake, particularly from cereal foods with high fiber content, in the prevention of colon cancer.
Cancers of the colon and rectum are among the most frequently diagnosed in the Scandinavian countries as well as globally,1 and it is believed that a substantial share of these could be prevented through dietary modifications.2 Intake of dietary fiber is hypothesized to lower colon and rectal cancer risk through biological mechanisms that include dilution of fecal carcinogens, shortening of bowel transit time, changes in the gut microbiota, production of short chain fatty acids with resulting decrease in pH, or adsorption of cytotoxic bile acids.3 However, while numerous studies have been published on the subject, the evidence remains inconclusive; case-control studies have to some extent shown a protective association,4, 5 but the same has not convincingly been observed in either intervention6 or prospective cohort studies.7–24 However, some cohort studies indicate a protective role,25–31 and possible reasons for these discrepancies could include limited statistical power, low fiber intake or variation in fiber intake, or inadequately detailed data. Furthermore, most studies have investigated colorectal cancer as one disease or did not differentiate between colon cancer subtypes, and only few considered the impact of different fiber sources. Especially fiber source may be important, as dietary fiber is a term encompassing a heterogeneous group of fiber types, which may act differently physiologically, and the ratio of these fiber types in a food will depend on the dietary source.32, 33
While no clear tendency is apparent when considering the evidence from individual prospective cohort studies, the report published by the World Cancer Research Fund in 2007 summarizing the current available evidence on dietary fiber intake and risk of colorectal cancer concluded that there appears to be a dose–response relationship based on evidence from cohort studies (relative risk per 10 g day−1 = 0.90, 95% CI = 0.84–0.97), and that cereal foods containing dietary fiber probably protect against colorectal cancer.34 Furthermore, a previous study on the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort, where dietary fiber intake overall but also fiber sources were related to colorectal cancer risk, found a borderline significant inverse association with intake of cereal fiber (IRR per quintile: 0.95, 95% CI: 0.90–1.00).25
The aim of the present study was to add to the existing, and somewhat inconsistent, body of evidence regarding the association between dietary fiber intake and incidence of colon and rectal cancer while focusing specifically on different sources of fibers (vegetables, fruits, potatoes and cereals) and specific tumor location (proximal and distal colon, and rectum). The study population used for the present study was the HELGA cohort, which consists of participants from three prospective cohorts in Denmark, Norway and Sweden. As cereal foods are an important part of the Scandinavian diet, this cohort provides an opportune setting for studying the association between intake of cereal fiber and colon and rectal cancer; furthermore, a wide intake range of cereal foods in this cohort enabled us to differentiate between cereal foods with either high- or low fiber content.
Material and Methods
The HELGA cohort consists of persons from three prospective Scandinavian cohorts: The Norwegian Women and Cancer study (NOWAC), The Northern Sweden Health and Disease Study (NSHDS), and the Danish Diet, Cancer and Health study (DCH). The three cohorts are part of the EPIC cohort and are described in detail elsewhere.35–37 In brief, the participants were recruited from the general population using a unique personal identification number and were aged between 30 and 64 years at baseline. The Norwegian cohort only recruited women, and the Swedish and Danish cohort only included persons from certain areas in the country.
The participants filled in validated country-specific food frequency questionnaires (FFQs)38–40 regarding their habitual food intake as well as extensive lifestyle questionnaires at baseline. Weight and height measurements, from which body mass index (BMI; kg m−2) was calculated, were done directly at the study centers in Sweden and Denmark, while they were self-reported in Norway.
In all three cohorts the dietary fiber determination was based on the Official method of analysis of the Association of Official Analytical Chemists (AOAC method) 985.29.41 Daily fiber intakes were calculated using country-specific food composition tables and existing or specifically developed standardized recipes and portion sizes. The dietary variables of interest in this study were total dietary fiber and fiber from four sources: (i) vegetables, (ii) fruits, (iii) potatoes and (iv) cereals and cereal foods. We also subdivided fiber from cereal foods into two groups of either low (<5 g per 100 g) or high (≥5 g per 100 g) fiber content; this cut-off was chosen based on product recommendations from the Keyhole symbol (used in Denmark, Norway and Sweden to identify the healthier food option)42 as well as the Danish Whole Grain Campaign.43 Examples of foods in these two groupings included white bread, pasta and biscuits for the group with low dietary fiber content, while nonwhite bread, breakfast cereals and crisp bread were included in the group with high dietary fiber content, and the groups could be considered as indicators of “refined grains” and “whole grains,” respectively.
The initial study population consisted of 119,913 persons. Participants were excluded due to prevalent cancer (n = 2,597) or missing information on potential confounders from the lifestyle questionnaire (n = 9,235), leaving 108,081 persons [38,852 (36%) men and 69,229 (64% women)] for analysis. The distribution of participants by country was as follows: Sweden, 12,013 men and 7,920 women; Denmark, 26,839 men and 28,948 women; Norway, 32,361 women.
The HELGA cohort was followed for cancer incidence using information from national cancer registries, patient registries and cause of death registries until December 31, 2007 (Denmark), December 31, 2008 (Sweden and Norway). Cancer incidence data were coded according to the 10th revision of the International Statistical Classification of Diseases, Injuries and Causes of Death (ICD-10). Proximal (right) colon tumors, located in the cecum, appendix, ascending colon, hepatic flexure, transverse colon and splenic flexure (C18.0–18.5) as well as distal (left) colon tumors, located in the descending and sigmoid colon (C18.6–18.7), were included. Cancers of the colon overall included the mentioned proximal and distal colon cancers as well as overlapping lesion of colon (C18.8) and colon cancer, not otherwise specified (NOS) (C18.9). Cancers of the rectum included tumors in the rectosigmoid junction (C19.9) and rectum (C20.9). Anal tumors were not included in cancers of the rectum.
For the baseline characteristics the continuous variables are presented as medians with corresponding percentiles (5th and 95th), while the categorical variables are presented as percentages. Analyses of the associations between dietary fiber variables and the incidence rate ratios (IRRs) of colon (including the subtypes proximal and distal colon) and rectal cancer were based on Cox proportional hazards models (including time-dependent variables) using age as the time axis to ensure that the estimation procedure was based on comparisons of individuals at the same age. Time under study was included as the time-dependent variable and was modeled by a linear spline with a boundary at 1, 2 and 3 years after entry into the cohort study to allow the rate to change with time.44 Two-sided 95% confidence intervals for the IRR were calculated on the basis of Wald's test of the Cox regression parameter, i.e. on the log rate ratio scale. Tests for interaction between intake of total dietary fiber and sex or cohort country, respectively, were performed.
All quantitative variables were entered linearly into Cox models after linear spline testing, as this is biologically more plausible than the step functions corresponding to categorization; furthermore, this approach increases the power of the analysis.45 The IRRs were expressed per 10 g day−1 increment for total dietary fiber intake and 2 g day−1 for vegetable, fruit, potato and cereal fiber intake (including high and low cereal fiber content), respectively. The units were selected based on the interquartile range for each exposure variable and were the primary analyses of this study. The continuous analyses were for overall colon cancer and rectal cancer further supplemented by categorical analyses based on sex-specific quartile intakes in the cohort for each of the fiber variables.
Two models are presented: an age-adjusted (crude) and a multivariable model, which was mutually adjusted for the fiber subtype analyses and adjusted for common colon and rectal cancer risk factors assessed from the FFQs and lifestyle questionnaires. These included: BMI (kg m−2; continuous), alcohol intake among users (g day−1; continuous), alcohol abstainers, intake of red and processed meat (g day−1; continuous), highest school level completed (primary, technical/professional, secondary, longer incl. university), smoking status (never, former, current), and for women, use of hormone replacement therapy (HRT).
Multiple testing was not handled in the statistical models, and the results should be interpreted with this in consideration.
The PHREG procedure in SAS (release 9.1, SAS Institute, Cary, NC) on a TextPad platform was used for the statistical analyses.
During a median follow-up of 11.3 years, 1,168 incident cases of colon and rectal cancer were diagnosed in the HELGA cohort; of these, 157 were from the Swedish cohort, 818 were from the Danish cohort while 193 were from the Norwegian cohort. Among men, 312 of included cancer cases were colon cancers (135 proximal, 154 distal, 1 overlapping lesion and 22 NOS) and 257 were rectal cancers, and the corresponding figures for women were 379 colon cancers (178 proximal, 172 distal, 3 overlapping lesions and 26 NOS) and 220 rectal cancers.
The test for interaction between sex and total fiber intake revealed a significant interaction (p < 0.001), and consequently, data were analyzed separately by sex. The test for interaction between cohort country and total fiber intake was not significant, and data were analyzed for the combined cohort with an underlying stratification for country to account for possible differences resulting from different inclusion criteria or questionnaire design in the three cohorts. Baseline characteristics of the total cohort and of colon and rectal cancer cases, respectively, are presented separately for men and women in Table 1. Compared with the total male cohort, male colon and rectal cancer cases had a higher median intake of alcohol and of red and processed meat. Furthermore, more were current smokers. Female colon and rectal cancer cases had shorter educations and higher median intakes of red and processed meat compared with the total female cohort. For rectal cancer cases, a larger proportion were current smokers, while for colon cancer cases, more were current HRT users. No apparent differences were seen between cases and the cohort for median intakes of total fiber or fiber from different sources for either men or women.
Table 1. Baseline characteristics and fiber consumption among the total HELGA cohort and colon and rectal cancer cases, presented separately for men and women
The main sources of fiber intake in the HELGA cohort were cereals (56%), followed by fiber from vegetables (17%), fruits (16%) and potatoes (11%; results not shown in table). The corresponding figures for the individual cohorts were very similar, and fiber from cereals dominated in all with a mean intake of 58%, 57% and 54% in Sweden, Denmark and Norway, respectively. Fiber from potatoes was a larger contributor to total fiber intake in Sweden than in the two other countries (16% vs. 10% in both Denmark and Norway), while a lower intake of fiber from vegetables in Sweden was seen (10% vs. 18% and 19% in Denmark and Norway, respectively).
In the following, only the multivariable adjusted results will be mentioned for all exposures and outcomes. Table 2 presents the IRRs (95% CI) for colon cancer, presented separately for men and women, in association with intake of total and of different sources of dietary fiber. Total fiber intake was associated with a significantly lower risk of colon cancer among men when comparing the highest to the lowest quartile of intake (IRR: 0.55, 95% CI: 0.38–0.79). This inverse association was also seen in the continuous analysis per an incremental increase of 10 g day−1 (IRR: 0.74, 95% CI: 0.64–0.86). Furthermore, intake of cereal fiber overall and cereal fiber from foods with high fiber content was associated with a significantly lower risk of colon cancer among men, both when comparing the highest to the lowest quartile of intake (IRR for cereal fiber overall: 0.66, 95% CI: 0.46–0.93; IRR for cereal fiber from high fiber foods: 0.55, 95% CI: 0.38–0.80) and in the continuous analyses per 2 g day−1 (IRR for cereal fiber overall: 0.94, 95% CI: 0.91–0.98; IRR for cereal fiber from high fiber foods: 0.94, 95% CI: 0.90–0.98). For women, an increase of 2 g day−1 of cereal fiber intake was associated with a borderline significantly lower risk of colon cancer (IRR: 0.97, 95% CI: 0.93–1.00). Potato fiber intake was inversely related to colon cancer risk for men (IRR: 0.65, 95% CI: 0.47–0.92 for a 2 g day−1 increase in intake), whereas higher risk was seen for women but only when comparing the highest to the lowest quartile of potato fiber intake (IRR: 1.59, 95% CI: 1.12–2.26).
Table 2. Incidence rate ratios and 95% CI of colon cancer in men and women in the HELGA cohort according to daily intake of total fiber per 10 g day−1, of individual sources of fiber, each per 2 g day−1, or by quartile of intake of the fiber types
Analyses using continuous variables of fiber intake were also performed for proximal and distal colon cancer (results not shown in table for both colon cancer subtypes). An inverse association for distal colon cancer per an incremental increase of 10 g day−1 of total fiber was seen for men (IRR: 0.66, 95% CI: 0.53–0.83). For both proximal and distal colon cancer, lower risk was for men seen for intake of cereal fiber from foods with high fiber content per an incremental increase of 2 g day−1 (IRR for both: 0.94, 95% CI: 0.89–1.00). Similarly for women, intake of cereal fiber overall and cereal fiber from foods with high fiber content was associated with lower risk of proximal colon cancer (IRR for cereal fiber overall: 0.93, 95% CI: 0.87–1.00; IRR for cereal fiber from high fiber foods: 0.93, 95% CI: 0.87–0.99). Significantly higher risk of proximal colon cancer was seen in association with intake of potato fiber per 2 g day−1 for women (IRR: 1.30, 95% CI: 1.06–1.61). Conversely, potato fiber intake was for men associated with a lower risk of distal colon cancer (IRR: 0.75, 95% CI: 0.60–0.95).
For rectal cancer, no associations were observed for men or women in the continuous analyses (Table 3). When comparing the highest to the lowest quartile of intake of cereal fiber from foods with low fiber content, higher risk of rectal cancer was seen for men (IRR: 1.44, 95% CI: 1.01–2.06). For women, lower risk of rectal cancer was seen when comparing the highest to the lowest quartile of intake of cereal fiber overall (IRR: 0.64, 95% CI: 0.44–0.93) and of cereal fiber from foods with high fiber content (IRR: 0.69, 95% CI: 0.47–1.00).
Table 3. Incidence rate ratios and 95% CI of rectal cancer in men and women in the HELGA cohort according to daily intake of total fiber per 10 g day−1, of individual sources of fiber, each per 2 g day−1, or by quartile of intake of the fiber types
A model adjusting additionally for energy was also prepared, but the adjustment had a negligible effect on the estimates (results not shown).
In this prospective study in a Scandinavian cohort, we found an inverse association between intake of total dietary fiber and the risk of colon cancer in men, which was also seen for the subsite distal colon cancer. Furthermore, intake of cereal fiber, in particular cereal fiber from foods with high fiber content, was associated with a lower risk of overall colon cancer in men and of the sub-sites proximal and distal colon cancer. A protective association between intake of cereal fiber and proximal colon cancer risk was also seen for women. Higher intake of potato fiber appeared to be inversely associated with colon cancer risk in men, particularly distal colon cancer, while higher risk of proximal colon cancer was seen for women. No associations were found for rectal cancer in the analyses using continuous variables.
Use of data from this combined Scandinavian cohort has a number of advantages. First, all three cohorts included in HELGA are prospective, population-based cohorts that use linkage to national cancer registries to identify incident cancer cases, consequently ensuring complete and valid identification of cancer cases, and all three cohorts have near complete follow-up with regard to incidence of cancer, death, and emigration. Second, the inclusion of a large number of cases allowed us to examine colon cancer overall as well as the two subsites, proximal and distal colon cancer, which are presumed to be different classes of colon cancer.46 Third, we were able to perform analyses using dietary fiber from different dietary sources as the exposure, enabling us to differentiate between cereal foods with high and low cereal fiber content. The HELGA cohort is highly appropriate for studying the association between intake of cereal fiber and colon and rectal cancer, as more than 50% of fiber intake for all cohorts came from intake of cereals and cereal foods. We analyzed cereal fiber overall as well as in two subgroups, cereal fiber from foods with low fiber content (<5 g per 100 g) and cereal fiber from foods with high fiber content (≥5 g per 100 g) according to criteria as stated by the Keyhole symbol and the Danish Whole Grain campaign. To our knowledge, a sub-division of fiber types in this manner has not been attempted in any previous studies. Fourth, all three cohorts have based their dietary fiber determination on AOAC method 985.29,41 thus ensuring comparability of fiber data.
The relatively small number of proximal and distal colon cases is a limitation of this study; for this reason, we only performed continuous analyses for these subsites to maximize statistical power. Furthermore, although the FFQs have been validated for each of the three cohorts,38–40 some measurement error is likely to be present. In addition, although the three cohorts have based their dietary fiber determination on the same method, the data stems from self-reported FFQs, which are inherently prone to some degree of misclassification, albeit presumably nondifferential due to the prospective nature of the study designs. Also, besides a possible linkage with a healthy lifestyle, intake of dietary fiber is linked with other components of plant-based foods, and it may be difficult to assess the specific impact of fiber intake isolated from other lifestyle factors. It may in fact be more appropriate to examine fiber intake as part of a food pattern or index.47 Finally, we were unable to adjust for physical inactivity, a risk factor for colon and rectal cancer, due to inadequate questionnaire data, and we cannot rule out the possibility of residual confounding despite attempts to adjust for other plausible colon and rectal cancer risk factors.
We initially adjusted for folate intake as part of our adjustment strategy; confounding by folate is often mentioned as a weak point of studies on fiber intake that are unable to do this.48 However, we observed a higher risk of rectal cancer in men in association with intake of fruit fiber when we adjusted for folate as a covariate. Fruit is a rich source of folate, and the higher risk seen after adjustment for folate may be an expression of less favorable components of fruit rather than of dietary fiber. As estimates regarding total fiber and cereal fiber intake were essentially unchanged without folate adjustment, we removed folate from our multivariable model.
Overall, most of the associations in the present study were found for men, which is congruent with two other studies that differentiated between men and women28, 31; the remaining cohort studies that performed sex-specific analyses did not find any overall associations.8, 15–17, 24, 30 The differences seen between men and women are difficult to explain. Hormonal factors are, however, known to play a role in the etiology of colorectal cancer,49 and it is possible that other risk and/or preventive factors consequently have different effects in men and women. This has, e.g., been seen with regard to obesity.50 In addition, associations were primarily seen for colon cancer rather than rectal cancer, which is to some degree expected, as any beneficial effects of dietary fibers would presumably be reduced due to the short retention time in the rectum.51 Results for distal and proximal colon cancer were generally similar, indicating an inverse association of total and cereal fiber intake on both subsites of colon cancer. Some differences were seen, but no clear pattern emerged from the subsite analyses.
Our results for total dietary fiber and risk of colon cancer are similar to those observed in a previous EPIC study, where only the Danish and Swedish part of the HELGA cohort were included due to a low number of Norwegian cases at the time of analysis.25 The present study included more cancer cases from updated registries and had a specific Scandinavian focus with particular emphasis on cereal fiber, especially from high fiber cereal foods. Besides the EPIC study, our results are congruent with four other prospective studies17, 28, 30, 31 and with a recent nested case-control study, nested within seven UK cohort studies, that found that intake of total dietary fiber was inversely associated with colorectal cancer risk (when using data from food diaries).26
Only two other studies have found results similar to ours with regard to intake of cereal fiber;25, 30 however, neither attempted to subdivide cereal foods into high- and low fiber cereal foods. The association for cereal foods in this study appeared to be driven by the high fiber cereal foods, as no inverse associations where observed with intake of low fiber cereal foods. Foods included in the grouping high fiber cereal foods can be seen as a measure of whole grain, as examples of the foods include breakfast cereals (primarily oats), nonwhite bread (primarily rye and wholegrain bread) and crisp bread (often wholegrain). Thus, the protective effect of high fiber cereal foods observed in the present study can be considered as an effect of whole grain food intake. This observation is supported by the findings of Egeberg et al., who found an inverse association of whole grain food intake on the risk of colon cancer in the DCH cohort, which is part of the HELGA cohort.52
Intake of fiber from potatoes was surprisingly associated with a higher risk of proximal colon cancer in women but a lower risk of colon cancer (including distal colon cancer) in men in the present study. As for all results, chance cannot be ruled out, especially considering the multiple comparisons performed in this study. The finding was unexpected, as potatoes are not a dietary component suspected to increase the risk of colon cancer, but perhaps it could be associated with the high starch content of potatoes or with some other component found therein. In other populations, where potatoes are mainly consumed fried, the carcinogenic substance acrylamide that is formed when starch is cooked at high temperatures could be a possible explanation.53 The current study is, however, based on populations where potatoes are mainly processed by boiling, which does not lead to acrylamide formation. However, none of the previous cohort studies have looked at fiber from potatoes separately, and we are therefore unable to judge the validity of the finding. It is also unclear why the findings are in opposite directions for men and women, and it deserves further attention in future studies.
Besides the abovementioned findings for total and cereal fiber, only few associations were found in the present study. Part of the reason for this may be that dietary fiber comprises chemically and physico-chemically a very diverse group of components, which can be grouped based on their chemistry or on their physiological effects. Dietary fiber may include pectins, lignin, cellulose, betaglucans, arabinoxylans, fructans, and several other components. Lignins are complex polyphenols and should, if possible, be treated separately in investigations on health effects of fiber. The physiochemical and physiological effects of dietary fiber, such as gel-forming capacity, solubility, and fermentability, may vary substantially between different types of dietary fibers.32, 33 As foods may contain a combination of different fibers, it would have been interesting to examine dietary fiber according to their physiochemical and putative physiological effects, thus combining fiber types across fiber sources. This would contribute with a better understanding of which types of fibers are important in cancer prevention. Unfortunately, we were not able to extract data on specific dietary fiber types, as the country-specific food composition tables used for dietary fiber estimations in their present form do not include information on specific fiber types; we suggest future studies to consider this aspect if possible.
In this study we intended to discriminate between fiber types by dividing cereal fibers into groups based on fiber content. In doing so, we are naturally also discriminating by other factors defining cereals with different fiber contents, such as micronutrients from bran and germ and the different contents of starch. Different cereals contain different amounts of soluble and insoluble fibers, and it is possible that the physico-chemical properties of these may explain our findings showing a stronger protective effect of dietary fiber from foods with high fiber content. As discussed previously, this food group consists mainly of oat and rye-based cereals. Rye and oat contain soluble fiber to a higher extent than wheat54; in the gut soluble fibers are more readily fermented to short-chain fatty acids including butyrate, which may have favorable effects on colonic health.55
Lastly, fibers are primarily ingested as parts of specific foods, mainly various grains, cereals, vegetables and fruits, and are in fact structural components of the plant cell wall. Dietary fiber exists with a range of other compounds in complex three-dimensional structures that may have functional effects which are altered, diminished or even lost once the fiber has been isolated or the food has been processed. For this reason it may also be worthwhile to analyze fiber from unprocessed and processed products in future epidemiological investigations.
It is important to remember that the study populations included in the present study have relatively high fiber intakes, especially of cereal fibers. Though the associations between fiber intake and incidence of colon cancer were all tested to be linear, the categorical analyses for cereal fiber among men indicated that the lower risk was mainly among those with the highest intake (>17.8 g day−1). This should be considered if results from the present study are compared to studies based on populations with lower intakes of cereal fibers.
In conclusion, intake of total dietary fiber and of cereal fiber was inversely associated with risk of colon cancer in men in this prospective Scandinavian cohort. Intake of cereal fiber from foods with high fiber content, in particular, appeared to be associated with lower risk, and this may support a protective role of whole grain foods in colon cancer prevention. It may in future studies be appropriate to distinguish further between different sources of fiber (including fibers from cereal products with high and low fiber content) and subgroups of fiber based on their physiochemical properties or to investigate the intake of dietary fiber as part of a food pattern.
The authors acknowledge Knut Hansen for assistance with data preparation. The views and opinions expressed in this publication are purely those of the writers and may not in any circumstances be regarded as stating an official position of the European Commission. There are no conflicts of interest.