In Europe and most of the industrialized world, colorectal cancer (CRC) is the third most common cancer in men after lung and prostate cancer and the second most common in women after breast cancer.1 There is experimental evidence suggesting that folate, a vitamin found mostly in vegetables and cereals, may reduce the risk of CRC.2 An inadequate intake of folate may increase the risk of CRC by various mechanisms, including hypomethylation of DNA, which interferes with gene expression and impairs DNA repair, or by uracil misincorporation leading to DNA breaks and instability. The epidemiologic data, however, are not consistent; to date, 53, 4, 5, 6, 7 out of 73, 4, 5, 6, 7, 8, 9 prospective investigations and 710, 11, 12, 13, 14, 15, 16 out of 11 case-control studies10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20 have found a relatively low risk for CRC in association with a relatively high intake of folate. No intervention studies have investigated the effect of folate supplementation on CRC risk, although the one trial with adenomatous polyps as an endpoint found a nonsignificant reduction in colorectal adenoma recurrence after 2 years in those receiving a folate supplement compared to placebo.21
The aim of this study is to conduct a metaanalysis of published data to provide an overall estimate of the association of dietary and total folate (folate from food and supplements) with CRC risk.
Material and methods
We conducted electronic searches of the PubMed database, which includes articles back to the 1950s and up to January 2004, using both MeSH headings and text words including the terms “folate,” “folic acid,” “colorectal,” “colon,” “rectum,” “bowel,” “cancer” and exploded variants. All full articles that matched the inclusion criteria were retrieved and the reference lists in those articles were hand-searched for other relevant publications.
We sought to identify all published English-language articles that provided estimates of risk (with 95% confidence intervals) of cancer of the colon and/or rectum in relation to dietary or total folate intake. Two studies that did not show confidence intervals for the estimates were excluded from the meta-analysis but the results of those studies were qualitatively assessed.10, 16
Data extraction and classification
We abstracted relative risks (RRs; odds ratios reported in case-control studies are described as RRs in this article) for the highest intake of folate relative to the lowest, together with 95% confidence intervals (CIs). When several estimates were presented using different adjustments,3, 14, 22, 23 the fully adjusted ones were included. Where studies presented information for dietary and total folate intake separately, the estimates for both were abstracted.
Statistical methods and analyses
The studies mostly reported relative risks according to quintiles (cohort studies) or quartiles (case-control studies) of folate intake; therefore, for this analysis, they were transformed using an established method24 so that all the meta-analyses are based on the relative risk in the top vs. bottom quintiles of intake for cohort studies and top vs. bottom quartiles for case-control studies.
Summary estimates of the standardized RRs were derived using random- and fixed-effect models; both yielded similar results and only estimates from random-effect models are presented. The 95% CIs from the original publications were used to calculate the standard errors of the standardized RRs, and the weighted average of the RRs was calculated by giving each study a weight proportional to its precision (i.e., the inverse of the variance). Thus, larger studies, with more precise estimates and narrower confidence intervals, were given greater weight than smaller ones. We tested for heterogeneity using Cochran's Q-test to evaluate the consistency of findings. Separate analyses were performed for cohort and case-control studies.
A metaregression analysis25 was performed to investigate whether the association between folate intake and risk differed according to sex, endpoint (colon or rectum) and source of folate (from foods only or foods plus supplements).
The results are presented graphically, whereby squares represent study-specific estimates and diamonds represent pooled estimates. The area of each square is proportional to the inverse of the variance of the natural logarithm of the RR; 95% CIs for individual RRs are represented by horizontal lines and for the pooled estimates by the width of the diamonds. All the analyses were performed using the statistical package Stata 8.0 (StataCorp, College Station, TX).
Seven cohort studies3, 4, 5, 6, 7, 8, 9 and 9 case-control studies11, 12, 13, 14, 15,17, 18, 19, 20 were included in this meta-analysis. Four cohort studies7, 8, 9 and 3 case-control studies13,15,19 presented information on total folate intake as a measure of exposure (Table I). Two cohort studies3,4 and 4 case-control studies14,15,17,20 presented RRs adjusted for fiber while 4 cohort studies4, 5, 6,9 and 1 case-control study13 presented RRs adjusted for other vitamins. Folate intake ranged from less than 103 to more than 422 μg/day for dietary folate and from less than 188 to more than 2,430 μg/day for total folate. All but 2 studies3,17 categorized folate intake in quartiles or quintiles.
Table I. Characteristics of the Cohort and Case-Control Studies Included in the Metaanalysis
Measure of exposure
Categories of exposure
Range of exposure (μg/day)
Adjusted for fiber
Adjusted for vitamins
This article reports results for dietary and total folate intake; however, the results for total folate are cited, based on a longer follow-up, in Wei et al.7
Figure 1 shows the RRs, after standardization, of CRC for the highest vs. lowest category of dietary and total folate intake in cohort and case-control studies. Very similar results were found using the nonstandardized RRs as reported in the individual studies (results not shown).
For cohort studies, the summary estimate of risk for high dietary folate intake was 0.75 (95% CI = 0.64–0.89), with no significant heterogeneity between studies (chi-square = 4.96; 7 df; p = 0.67). The association of total folate intake with CRC risk in cohort studies was close to null (RR = 0.95; 95% CI = 0.81–1.11), with no significant heterogeneity between studies (chi-square = 4.57; 4 df; p = 0.33).
Among case-control studies, the overall RR for high vs. low dietary folate intake was 0.76 (95% CI = 0.60–0.96), although the test for heterogeneity was highly significant (chi-square = 23.10; 9 df; p < 0.01); for total folate from foods and supplements, the RR was 0.81 (95% CI = 0.62–1.05), with no heterogeneity between the 4 estimates from 3 studies (chi-square = 2.39; 3 df; p = 0.50).
The estimates from the 2 case-control studies excluded from the meta-analysis were compatible with these results. RRs in the highest vs. the lowest intakes of dietary folate were 0.56 (ptrend < 0.05) for both sexes in a study in Spain10 and 2.08 (ptrend > 0.05) and 0.73 (ptrend > 0.05) for men and women, respectively, in a study in Washington State.16
The metaregression analysis revealed that, within cohort studies, the association between folate intake and risk differed significantly to the endpoint considered, folate intake having a slightly stronger association with risk for colon compared to rectal cancer (p = 0.03). The association between folate intake and risk for CRC did not differ significantly according to source of folate intake (dietary or total, p = 0.06) or according to sex (p = 0.74). Within case-control studies, there were no significant differences in the association of folate intake and risk according to endpoint, source of folate, or sex.
Overall, this meta-analysis shows that, among cohort studies, there is a significant 25% lower risk of CRC among those in the highest category of dietary folate intake compared with those in the lowest category, with no evidence of heterogeneity between the study estimates. For total folate intake, there was a nonsignificant 5% lower risk of CRC. Although the pooled estimates for case-control studies also showed a reduced risk, there was strong evidence of heterogeneity between studies, which may not absolutely be due to differences in the study characteristics examined of date, endpoint, or sex, and thus the pooled estimates should be interpreted with caution.
Folate intake is generally correlated with intake of fiber and some vitamins. The apparent protective effect of folate may therefore have been confounded by other dietary factors that may themselves be protective for CRC. Where possible, the RRs included in this meta-analysis were adjusted for intake of other vitamins and fiber. However, the effect of such adjustment is not clear and in 23,20 of the 3 studies3,4,20 where results were shown unadjusted and further adjusted for fiber, the RR did not change appreciably. Giovannucci et al.3 found also that a significant association between folate and CRC remained after adjusting for other vitamins and minerals (vitamins A, C, D, E and calcium); conversely, after adjustment for folate intake, none of these dietary factors was significantly associated with CRC risk.
Food frequency and recall questionnaires do not estimate folate intake with great precision. In addition, estimated folate in food may not accurately reflect actual intake and absorption because folate is sensitive to heat, pH and oxidation; the availability of folate is also affected by the presence of certain inhibitors and folate conjugates found in some foods.26 Moreover, folate from supplements (folic acid or pteroyl-monoglutamic acid) has a greater bioavailability than polyglutamate folate from foods, which needs to be hydrolysed to monoglutamate in the intestine before absorption.28 All these factors can introduce some degree of measurement error, thought to be largely random, in the assessment of folate intake, which would be expected to bias the results toward the null; this type of measurement error is thus unlikely to account for the inverse association found between dietary folate intake and CRC risk.
The larger reduction in risk associated with folate from dietary sources alone compared with total folate intake was unexpected. Only 3 studies included in this metaanalysis presented estimates for both dietary and total folate intake separately. The Nurses' Health Study3 found a large reduction in risk among those taking multivitamin supplements for 15 years or more compared with those taking them for less than 15 years or never, but no significant reduction in risk was observed for high intakes of dietary folate. Another study15 found similar RRs for high intakes of both dietary and total folate and a third study8 found a stronger, although nonsignificant, association of risk with dietary folate compared to total folate. More data are needed to clarify whether dietary or total folate is more strongly related to risk, but a plausible explanation is that confounding by other dietary factors might be playing a bigger role when investigating the association between CRC and dietary folate than when looking at total folate.
It is unclear whether the association between folate intake and CRC risk is stronger for colon than rectum. From the metaregression analysis and within cohort studies, there was an apparent stronger association with colon cancer than with rectal cancer. However, this result is dominated by 2 studies reported in one publication7 and became nonsignificant after their exclusion. The most appropriate way to disentangle this would have been to pool the estimates for the 2 endpoints from each study separately; however, information on colon cancer (CC) and rectal cancer (RC) separately was only available for 3 cohort studies and 2 case-control studies. Therefore, the question remains open and further studies looking at CC and RC will be needed to clarify this.
It is possible that other factors, in combination with inadequate folate intake, are needed to provoke changes such as disturbances in DNA methylation and other processes related to CRC risk. It has been shown that a high alcohol intake might have an adverse effect on folate metabolism,28 and at least one epidemiologic study has observed that individuals with a diet low in folate (and methionine) and high in alcohol have a relatively high risk of CRC.29 Furthermore, it is possible that certain subgroups of the population, such as those with specific polymorphisms2 or with a family history of CRC, may receive a greater benefit from a diet high in folate.22
In conclusion, the results of this meta-analysis offer some support for the hypothesis that relatively high folate consumption is associated with a decrease in the risk of CRC, but confounding by other dietary factors cannot be ruled out. Further studies are needed to determine whether folate from foods alone or from supplements is associated with a greater reduction in risk.