Fruit and Vegetable Consumption and Colorectal Adenomas in the Nurses' Health Study Free

Colorectal adenomatous polyps (also called adenomas) are precursors of most colorectal carcinomas, the third most common type of cancer in men and women and the second most common cause of cancer death in the United States.⁸

The influence of dietary factors in the pathogenesis of colorectal cancer has been studied extensively (2). Less data are available on the role of diet in adenoma formation. Dietary factors may influence adenomas at several stages: incidence, recurrence, or progression of adenomas to cancer.

Consumption of fruits and vegetables may confer protection from adenomas through components, such as carotenoids, folate, vitamin C, flavonoids, organosulfides, isothiocyanates, and protease inhibitors, which may mitigate DNA damage and thus reduce mutations (2). Another possible dietary contribution is the fermentable fiber in fruits and vegetables, which decreases transit time, increases bulk, lowers pH, and produces potentially anticarcinogenic short-chain fatty acids (2). Furthermore, unknown constituents of fruits and vegetables may be responsible for the effects because the influence of fiber and antioxidant vitamin supplements on the recurrence of colon polyps remain equivocal (3).

Some observational studies have addressed the relation between fruit and vegetable consumption and colorectal adenomas. Six (3–8) of the nine (3–11) case-control studies available report significant inverse associations between fruit and vegetable consumption and the risk of adenomas.

The results from the Polyp Prevention Trial did not indicate any benefit for a diet high in fiber, fruits, and vegetables in preventing the recurrence of colorectal adenomas. However, this study was restricted to participants who had previous adenomas, and their period of intervention and follow-up was 4 years, which may be insufficient to influence the occurrence or recurrence of adenomas (12–14).

Our research group has previously found that frequent consumption of fruit was associated with a lower risk of colorectal adenomas among men in the Health Professionals' Follow-up Study (HPFS; ref. 15). We here present the relation of overall consumption of fruits and vegetables, and of subgroups of fruits and vegetables, to risk of colorectal adenomas in a large cohort of women, the Nurses' Health Study (NHS).

Study population. The NHS began in 1976 when 121,700 U.S. female registered nurses, ages 30 to 55 years, completed a questionnaire providing information on their demographic, reproductive, and lifestyle factors and disease history. Since then, we have mailed biennial questionnaires to update information and to identify newly diagnosed cases of cancer and other diseases. In 1980, a validated diet assessment was added to the questionnaire (16). We also asked participants about examination by colonoscopy or sigmoidoscopy and about indications for the procedure. To minimize detection and selection bias in the analysis of adenomatous polyps, we restricted the cohort to women who reported undergoing either colonoscopy or sigmoidoscopy during the study period. In addition, we excluded women who reported a baseline history of cancer (except nonmelanoma skin cancer), ulcerative colitis, or a familial polyposis syndrome and who reported a total caloric intake below 500 calories or above 3,500 calories. From 1980 through 1998, 34,467 women had undergone either colonoscopy or sigmoidoscopy and were eligible for this analysis. The study was approved by the Institutional Review Board at Brigham and Women's Hospital.

Ascertainment of cases. At each biennial follow-up, we inquired whether participants had undergone colonoscopy or sigmoidoscopy, the reasons for the procedure, and whether colon or rectal polyps had been diagnosed and, if so, the date of diagnosis. When a participant reported a diagnosis of colorectal adenoma, we sought her permission to obtain medical records and pathology reports. Study physicians, blinded to exposure data, reviewed all records, and extracted data on histologic type, anatomic location, and size of polyps. Because a substantial portion of respondents had sigmoidoscopies, which examine only the distal colon and rectum, we defined cases as one or more pathology-verified adenomas <60 cm from the anus (17, 18). If more than one adenoma was diagnosed, the case was classified by the largest and most histologically advanced. Women with adenomas proximal to the descending colon were classified as noncases.

Dietary assessment. Dietary intake data were collected in 1980, 1984, 1986, 1990, and 1994 using validated self-administered semiquantitative food frequency questionnaires (FFQ; ref. 16). In 1980, the FFQ used for the NHS consisted of 61 food items, including six questions on fruit consumption comprising 10 fruits and 11 questions on vegetable consumption comprising 14 vegetables. The 1984 FFQ was expanded to include 15 questions on fruit consumption comprising 20 fruits and 28 questions on vegetable consumption comprising 33 vegetables; the questionnaires used in 1986, 1990, and 1994 were similar to that used in 1984. Nine mutually exclusive response possibilities were provided for the frequency of intake. 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. Responses regarding individual food items were converted to average daily intake of each fruit and vegetable item for each participant. We combined the average daily intake figures for individual food items to compute total fruit and vegetable intake as well as intake of composite fruit and vegetable groups. Fruit and vegetable subgroups were defined a priori based on the criteria by Smith et al. (19); the groups were modified to conform to our questionnaires (20). The composite items and their individual food components are described in Appendix A. We also examined the consumption of individual fruits and vegetables in relation to colorectal adenoma risk.

The validity of our assessment of individual food items has been documented by comparisons with multiple weighed dietary records, correcting for within-person weekly variation in diet (21). The average correlation coefficient comparing responses for specific fruits and vegetables on the 1980 FFQ with intake from four 1-week dietary records corrected for within-person variation was about 0.54, ranging from 0.17 for spinach to 0.84 for orange juice. The average correlation coefficient for fruit consumption based on four (commonly consumed) fruits was 0.72 and for vegetable consumption based on nine vegetables was 0.46 (21). The validity of reporting was generally higher for commonly consumed foods than for less commonly consumed foods.

Statistical analysis. Consumption of fruits and consumption of vegetables were grouped in five categories: <1.5 servings per day (denoted as ≤1/d in the tables; reference category), 1.5 to 2.4 servings per day (2/d), 2.5 to 3.4 servings per day (3/d), 3.5 to 4.4 servings per day (4/d), ≥4.5 servings per day (≥5/d). Few participants consumed less than two servings of fruit and vegetables combined. Thus, the categories for combined fruit and vegetable consumption were <3.5 servings per day (≤3/d; reference category), 3.5 to 4.4 servings per day (4/d), 4.5 to 5.4 servings per day (5/d), 5.5 to 6.4 servings per day (6/d), and ≥6.5 servings per day (≥7/d). For composite fruit and vegetable groups (e.g., cruciferous vegetables), cut points of categories of consumption had to be chosen differently because of the lower frequency of intake. Daily consumption of fruits and vegetables was calculated from the frequencies prespecified on the FFQ. Odds ratios (ORs) for an increase in intake by one serving per day were also calculated as a continuous variable, which was truncated at 10 servings per day for fruit groups and vegetable groups (i.e., self-reported consumption of >10 servings per day was coded as 10 servings per day) and at 15 servings for combined fruit and vegetable intake to avoid undue influence of outliers. Fewer than 0.5% of participants reported consumption of >10 servings of fruits or vegetables per day.

Logistic regression models were used to adjust for age and multiple known or suspected adenoma risk factors simultaneously. We used ORs from these models defined as the prevalence of adenoma in a category of consumption of fruit and vegetables divided by the corresponding prevalence among participants with the lowest consumption of these foods to approximate the relative risk.

To represent long-term dietary patterns of individual subjects as accurately as possible and to reduce random within-person variation, the diagnosis of colorectal adenoma was modeled in relation to the cumulative average fruit and vegetable intake from all available dietary questionnaires up to the start of each 2-year follow-up interval (22). For example, dietary data from the 1980 questionnaire were used to predict colorectal adenomas diagnosed between 1980 and 1984. The average of the 1980 and 1984 dietary intake was used to predict outcomes between 1984 and 1986. The average of the 1980, 1984, and 1986 FFQs was used to predict colorectal polyps between 1986 and 1990. The average of the 1980, 1984, 1986, and 1990 FFQs was used to predict adenomas from 1990 to 1994. Finally, the average of the 1980, 1984, 1986, 1990, and 1994 FFQs was used to predict adenomas from 1994 to 1998. Women were grouped according to categories of cumulatively updated consumption of fruit and vegetables up to the questionnaire cycle before the sigmoidoscopy/colonoscopy. Total caloric intake was included in the covariate-adjusted model as is standard for a model, including foods to control for confounding by total energy intake and to minimize extraneous variation due to physical activity, body size, or general underreporting or overreporting of food items on the FFQ (23). Including total caloric intake as a separate term in a food model, including fruits and vegetables, has been found not to distort estimates (24). In an isocaloric model, the risk of adenomas is related to substituting fruits and vegetables for other energy-contributing foods.

Analyses were adjusted for age (5-year categories), family history of colorectal cancer, height (continuous), body mass index (BMI = weight/height²; continuous), regular vigorous exercise (<1.7, 1.7-4.5, 4.6-10.5, 10.6-22.1, >22.1 metabolic equivalent task score per week), regular aspirin use (<2 tablets per week, ≥2 tablets per week), pack-years of smoking (never, 0.1-10, 10.1-20, 20.1-40, >40), current multivitamin supplement use, alcohol consumption (none, 0.1-4.9 g, ≥5 g/d), total caloric intake (continuous), red meat consumption (≤1/wk, 2-4/wk, 5-6/wk, ≥1/d), calcium intake (quintiles), menopausal status, and postmenopausal hormone use (never, current, and past). All covariates were repeatedly assessed, and for each covariate, values were used from the questionnaire before the sigmoidoscopy/colonoscopy.

Because adenomas are typically asymptomatic and may have been present for many years, we also calculated the risk of incident adenomas by restricting our analysis to women who had undergone earlier endoscopy >2 years before their most recent endoscopy and had been determined to be adenoma-free. We conducted additional analysis restricted to women undergoing sigmoidoscopy or colonoscopy for routine screening and stratified by smoking status.

All Ps are two sided.

Among the 34,467 women in our cohort who had undergone a sigmoidoscopy or colonoscopy, 1,720 were found to have one or more adenomas of the distal colon or rectum.

Characteristics of our study population in the year before their first sigmoidoscopy, by their fruit and vegetable intake, are presented in Table 1. Women who reported high consumption of fruits were leaner, more likely to exercise regularly and use multivitamin supplements, and less likely to smoke and drink alcohol than women with lower fruit consumption. Women who reported frequent consumption of vegetables were more likely to exercise regularly but more likely to consume alcoholic beverages than women with infrequent consumption of vegetables. Folate intake and total caloric intake increased with higher consumption of fruits and vegetables.

Frequent consumption of fruits and vegetables was inversely related to the prevalence of distal colorectal adenomas in age-adjusted analyses (Table 2). Women who consumed seven or more servings of fruits and vegetables per day had an OR of 0.82 [95% confidence interval (95% CI), 0.71-0.95] for adenomas compared with women who consumed two or fewer servings of fruits and vegetables per day. After adjustment for the potential influence of smoking, physical activity, regular aspirin use, multivitamin supplement use, height, body mass index, family history of colorectal cancer, menopausal status, use of postmenopausal hormones, alcohol consumption, calcium intake, red meat consumption, and total caloric intake, the association remained essentially unchanged (OR, 0.81; 95% CI, 0.68-0.96). When we examined the influence of fruit and vegetables separately, fruit consumption was inversely related to the prevalence of colorectal adenomas, whereas the association with vegetable consumption was not statistically significant (Table 2). Women who reported consuming five or more servings of fruit a day had a covariate-adjusted OR of 0.60 (95% CI, 0.44-0.81) for colorectal adenomas compared with women who consumed one or fewer servings of fruit per day (P for trend = 0.003). The respective OR for vegetable consumption was 0.82 (95% CI, 0.65-1.05; P_trend = 0.1). In an analysis adjusting fruit consumption for vegetable consumption and vice versa, the covariate-adjusted OR for colorectal adenomas associated with the highest category of fruit consumption was 0.62 (95% CI, 0.46-0.85) and for the highest category of vegetable consumption was 0.89 (95% CI, 0.70-1.14). Among specific subtypes of fruits and vegetables, frequent consumption of citrus fruit (which included oranges, orange juice, grapefruit, and grapefruit juice) was inversely associated with polyp prevalence (covariate-adjusted OR, 0.81; 95% CI, 0.65-1.00; P_trend = 0.02). Regular consumption of legumes was marginally inversely related to adenomas (for each additional serving per day OR, 0.96; 95% CI, 0.93-1.00; P_trend = 0.04), whereas potato consumption showed a marginal positive trend (for each additional serving per day OR, 1.06; 95% CI, 1.01-1.10; P = 0.04; Table 2). When individual fruits were considered, frequent consumption of apples and bananas were inversely related to polyp prevalence. Women in the highest quintiles of apple consumption had a covariate-adjusted OR of 0.83 (95% CI, 0.70-0.98) for colorectal adenomas compared with women in the lowest quintile of intake (P_trend = 0.05); the respective value for frequent banana consumption was OR of 0.79 (95% CI, 0.66-0.93; P_trend = 0.001).

When we restricted our analysis to incident cases of colorectal adenomas, the results were strengthened for citrus fruit consumption and for fruits and vegetables rich in vitamin C (Table 3). Women who consumed two or more servings of citrus fruit daily had a covariate-adjusted OR of 0.60 (95% CI, 0.42-0.86; P_trend ≤ 0.001). Frequent consumption of legumes was also associated with a reduced incidence of polyps, but the association was restricted to women in the highest category of intake, consuming four or more servings per week (OR, 0.67; 95% CI, 0.51-0.90; P_trend = 0.005). Similarly, women who consumed green leafy vegetables more than once per day had a reduced incidence of colorectal adenomas compared with women who reported consuming one serving per week or less (OR, 0.78; 95% CI, 0.61-0.98; P_trend = 0.02).

Associations were generally somewhat stronger among nonsmokers than among smokers (Table 4). When we restricted analysis to women who underwent sigmoidoscopy or colonoscopy as routine screening, associations were similar (covariate-adjusted OR for adenoma for each additional serving of fruit and vegetables consumed per day: 0.95; 95% CI, 0.91-0.99; for fruit: OR, 0.92; 95% CI, 0.85-0.99). Results did not differ according to size of the adenomas (data not shown). Results were also similar but marginally weaker if we only used dietary information obtained from the 1980 FFQ (data not shown).

In this large cohort of women who had undergone colonoscopy or sigmoidoscopy, consumption of fruit was inversely related to the risk of distal adenomas of the colon or rectum after controlling for potential confounding factors. This association was most pronounced among women with incident adenomas. No significant association was found between vegetable consumption and adenomas; however, regular consumption of legumes was associated with a decreased incidence of adenomas.

The relation of fruit and vegetable consumption with risk of colorectal adenomas has been studied previously. Of a total of 11 studies conducted, nine were case-control (3–11), one was a prospective cohort (15), and one was a randomized controlled trial (12–14). All case-control studies addressed prevalent adenomas, and the HPFS reported on prevalent and incident adenomas (15). One case-control study considered both prevalent and recurrent adenomas but combined them in the analysis (11). Although some case-control studies called adenomas “first incident” or “newly diagnosed,” they truly reported prevalence data, because adenomas were diagnosed at the first endoscopy; thus, it was unknown how long the adenomas had been present. To report incident adenomatous polyps, at least two endoscopies are necessary with the first documenting absence of adenomas. The Polyp Prevention Trial considered only recurrent adenomas.

Six of the nine case-control studies found significant inverse associations between fruit and vegetable consumption and colorectal adenomas (3–8). Total vegetable consumption was inversely associated with colorectal adenoma risk in four studies (3–5, 8). A nonsignificant inverse association with cruciferous vegetables was reported by two early studies (8, 9). Fruit or fruit juice has been found to be particularly important in the development of adenomas in five studies (one prospective cohort and four case-control; refs. 5–7, 10, 15). However, case-control studies typically suffer from the potential for selection bias and are limited by recall and reporting bias if dietary intake is assessed after the diagnosis of disease. Diet was assessed after sigmoidoscopy or colonoscopy in the majority of case-control studies on fruit and vegetable consumption and the risk of adenomas.

Two analyses from the HPFS reported significant inverse associations of fruit and vegetable fiber on the risk of colorectal adenomatous polyps in men (15, 25). In the updated analysis in 1997, frequent consumption of fruit but not vegetables was associated with a reduced prevalence of adenomas (15). In the NHS and the HPFS, frequent consumption of fruits and vegetables was not associated with risk of colorectal carcinoma (21).

In the Polyp Prevention Trial, a multicenter randomized controlled trial, no association was found between a diet low in fat and high in fiber, fruits, and vegetables and the risk of recurrent colorectal adenomas or rectal mucosal cell proliferation rates (13, 14). This trial, however, was limited to a follow-up of 4 years, and the difference in fruit and vegetable intake among the intervention group (6.4 servings per day) and the comparison group (4.3 servings per day) was modest. Serum carotenoid levels did not substantially differ between the intervention and the control group after 4 years of follow-up (13). This trial could not address whether fruit or vegetable consumption might exert a protective effect only before formation of first adenoma or in later neoplastic stages.

The major strengths of our study include its prospective design, large sample size, and a response rate exceeding 90%. The large array of potential confounders assessed in this cohort enabled us to control for the influence of numerous indicators of a healthy lifestyle. Our study is based on 18 years of prospective follow-up and includes repeated assessment of dietary data and analyses of both “incident” and prevalent adenomas.

Although food consumption is assessed with error, the prospective assessment eliminated the possibility of recall bias and repeated assessment of diet reduced random within-person variation. Because polyps are asymptomatic, their existence is unlikely to lead to dietary changes in the years immediately preceding diagnosis. Women in this study reported a wide variation in intake of fruits and vegetables; thus, we were able to examine a gradient in intake and relate it to the risk of adenomas. Detection bias was reduced by including only those women who had the opportunity for detection of polyps through sigmoidoscopy or colonoscopy during the study period. Furthermore, detection bias is likely limited because the majority of those who underwent endoscopy did so for routine screening and not gastrointestinal symptoms that could be related to diet. Finally, our follow-up period of 18 years is substantially longer than the period of observation in most studies and trials, which ranged between 4 and 8 years. However, if diet during early periods of life is more important for polyp formation, we may not be able to capture these effects with our study (26). Our observation of an inverse association between frequent fruit consumption, in particular citrus fruit, and colorectal adenomas may be due to the high folate content of these foods. Folate intake has been associated with a reduced risk of adenoma (27, 28). Although fruits are also high in fruit fiber, the overall evidence does not support an important role for fiber in adenoma prevention (29).

Conversely, our findings could be the result of residual or unmeasured confounding by a healthy lifestyle. We found that women with a higher fruit consumption were more likely to exercise regularly, consumed less or no alcohol, and had a lower BMI and were less likely to smoke or consume red meat. Thus, fruit consumption may be more indicative of a healthy lifestyle and healthy diet than vegetable consumption. Fruits are consumed mostly uncooked as dessert or a snack or as juice, whereas vegetables are usually part of the main course and often consumed with meat dishes. Potatoes are also often consumed with meat; thus, residual confounding by red meat consumption is a possible explanation for the increased risk observed for frequent potato consumption. Furthermore, because French fries and potato chips were included in the potato group, preparation and fat content may be more important for the risk of polyp formation than potatoes themselves. The difference in associations observed for fruits and vegetables may also be due to the more accurate assessment of fruit consumption while vegetable consumption seems less accurately reported, possibly because vegetables are often consumed as part of mixed dishes. Limitations of our study include lack of data on food preparation and storage, which can alter the nutrient content of fruits and vegetables. Furthermore, the relevant lifetime period when dietary factors exert most of their influence on the development of polyps is unknown as is the latency period for adenoma formation, and this may have contributed to the varying relations between fruit and vegetable consumption and adenomas across different studies. The results from the present study may not be generalizable to the proximal colon, because we considered in our analysis cases with adenomas of the distal colon and rectum only, and the proximal colon and distal colon have a different spectrum of genetic alterations in tumors.

Although we found fruit consumption inversely related to the incidence and prevalence of adenomas of the distal colon and rectum in both NHS and HPFS, we did not find the consumption of fruit or vegetables associated with the incidence of colorectal cancer in either of these two large prospective cohorts of women and men (20). It is possible that the latent period between diet and carcinoma spans several decades, and only an association with an intermediate, such as adenoma, is captured in studies spanning a shorter time period. Cigarette smoking is also a strong risk factor for colorectal adenomas, but colorectal cancer is affected much less and only after several decades of latency (18). It might be sensible in future prospective studies on fruit and vegetable consumption and colorectal cancer to consider large time lags between the dietary exposure and cancer occurrence as an effect might only emerge after several decades.

Only about 15% of adenomas progress to carcinomas (1). Possibly, fruit consumption may selectively affect a subgroup of adenomas with less propensity to progress. However, we did not find a difference in the association depending on size of the adenoma.

In conclusion, the results from the NHS suggest that frequent fruit consumption is particularly important in reducing the risk of adenoma occurrence, whereas no consistent benefit from vegetable consumption was evident except for regular consumption of legumes. Further investigations of constituents of fruits that may contribute to the occurrence of polyps are warranted. Results from this study add support to current dietary recommendations that include several servings of fruit per day.

Grant support: National Cancer Institute, NIH grant CA87969 (NHS); National Institute of Diabetes and Digestive and Kidney Diseases, NIH grant DK54900 (K.B. Michels); and American Gastroenterological Association/Foundation for Digestive Health and Nutrition Research Scholar Award (A.T. Chan).

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.