Dietary Risk Factors for the Incidence and Recurrence of Colorectal Adenomatous Polyps: A Case-Control Study

Methods

Our study sample included patients having colonoscopy at three colonoscopy practices in New York City between April 1986 and March 1988. In total, 2988 patients were evaluated. Of these, 2443 (81.8%) were eligible for our study (patients had to be between 35 and 84 years of age; reside in New York, New Jersey, or Connecticut; speak English or Spanish; and have colonoscopy to at least the splenic flexure). The colonoscopists completed data sheets indicating the reason for colonoscopy and the clinical findings at the time of colonoscopy. The study pathologist reviewed slides of all suspected neoplastic lesions.

All eligible participants received a letter signed by their colonoscopist introducing the study. A trained interviewer then contacted and interviewed participants by telephone. Alternatively, the questionnaire was mailed for self-completion and was followed by a telephone interview to resolve any remaining questions. An earlier study indicated that the results obtained for dietary factors were similar for both interview methods [13].

The interview itself consisted of a general questionnaire that focused on demographic characteristics, medical history, lifestyle, family history, and other topics. The dietary interview consisted of the Block food frequency questionnaire and specified food intake for a period 3 to 5 years before the colonoscopy [14].

Ultimately, 1956 dietary questionnaires were completed (80.1% of eligible patients). Of these, 71% were conducted by telephone, and 29% were returned by mail.

An incident case of adenomatous polyps was defined as an eligible participant with no history of colon carcinoma, adenomatous polyps, or inflammatory bowel disease who was found to have one or more pathologically defined polyps on the index colonoscopy. The incident control group consisted of persons who were found to be free of colorectal neoplasia on index colonoscopy and who were without a history of adenomatous polyps, colon cancer, or inflammatory bowel disease.

A case of recurrent polyps was defined as an eligible participant with a self-reported history of one or more polyps who had a pathologically confirmed polyp on the index colonoscopy. The recurrent control was defined as a participant whose index colonoscopy showed no colorectal neoplasia but who had a history of one or more polyps. Cases and controls with a history of colorectal cancer or inflammatory bowel disease were excluded. Although we did not have pathologic confirmation of all initial polyps, we did obtain pathology reports on a random sample of 100 recurrent cases and controls and found 97 to be adenomatous.

By these criteria, 286 incident cases (162 men and 124 women) and 480 incident controls (210 men and 270 women) were identified, whereas 186 recurrence cases (130 men and 56 women) and 330 recurrence controls (187 men and 143 women) were found.

Food item and nutrient data were generated by software programs provided by Block and coworkers [14] at the National Cancer Institute. The main analyses were done using logistic regression modelling and maximum likelihood ratios [15] in the BMDP-LR program. Analyses were conducted separately for men and women. Age, Quetelet index, and caloric intake were entered as covariates for most analyses. A previous study by our group had shown obesity, as measured by Quetelet index, to be a risk factor for polyps among women; the trend for men was not significant [16]. Analyses in which nutrients were standardized per 1000 kilocalories were also done for comparison [17]. For each nutrient or food group, quartiles were defined by review of the control group data; the lowest quartile was given a reference value of 1.0, and odds ratios were calculated for each of the other quartiles, with 95% confidence intervals (CIs) for the highest-to-lowest quartile comparison. The probability of a linear trend was calculated by entering the four quartiles as ordered categories.

Results

The case and control groups for the incidence and recurrence studies were generally similar in age distribution and education. Table 1 shows a comparison of the characteristics of the case and control groups for both the incident and recurrent groups. Most polyps were 5 mm or larger in size and had at least some degree of atypia. The site distribution of the incident polyps showed a preponderance to the left. Most incident case and control participants had colonoscopy because of overt or occult rectal bleeding. A larger proportion of the recurrent polyps were right-sided (P = 0.005). The time interval from initial polypectomy to index colonoscopy was 4.3 years for cases and 3.7 years for controls (P > 0.2).

Tables 2 and 3 show the odds ratios by quartile, using the lowest quartile as the referent group, for some of the 15 nutrients and food groups evaluated. The results for vegetables, red meat, beef, cheese, protein, vitamin C, and carotene are not shown; however, no consistent differences were found.

Men

The only dietary risk factor statistically associated with the risk for colorectal adenomatous polyps in men was caloric intake; however, this association was in a direction opposite to that ordinarily expected [18].

Women

In contrast, various dietary factors were observed to be associated with the risk for colorectal adenomatous polyps in women (Tables 2 and 3). Increased saturated fat, decreased fish and chicken, increased meat-to-fish and -chicken ratio, and decreased vitamin A intake increased the risk for incident polyps (Table 2). Increased caloric intake, increased total fat, increased saturated fat, and decreased fiber intake all raised the risk for recurrent polyps in women, whereas vitamin A and carbohydrate intake had borderline protective effects (Table 3).

Analyses were also done using nutrient density (nutrient compared with caloric intake) instead of entering calories as a covariate. The results are not shown, but the same risk factors were statistically significant for incident polyps in women, although the estimated odds ratios were larger. In addition, fiber was protective for incident polyps in women (odds ratio, 0.6; CI, 0.3 to 1.1; P = 0.06). The same dietary factors were also associated with recurrent polyps in women, although the odds ratio estimates were again larger. Both vitamin A (odds ratio, 0.5; CI, 0.2 to 1.1; P = 0.06) and carbohydrate intake (odds ratio, 0.4; CI, 0.2 to 1.0; P = 0.001) were more clearly protective.

Subgroup Analyses

For each of the dietary factors associated with colorectal polyps in women, further subgroup analyses were done for right-sided polyps only, for left-sided polyps only, and for polyps 5 mm or larger in size. Generally, no major variations were observed for the various subgroups, although some reduction was seen in statistical power because of the smaller number of cases.

To determine the independent effect of each of the variables found to be associated with polyps in women, we conducted further multiple logistic regression analyses using various dietary factors as covariates. The elevated risk associated with increased consumption of saturated fats remained after adjustment for fiber or vitamin A.

Discussion

Many studies have suggested that diet plays a role in the etiology of colorectal cancer [19-35]. Evidence suggests that increased consumption of saturated fat is a causal factor and that increased consumption of fiber, (particularly fruit and vegetable fiber) is protective [26]. Similarly, an increased risk has been associated with greater consumption of red meat compared with chicken or fish [19], and a protective effect has been linked to consumption of vegetables [35]. A protective effect of such micronutrients as vitamin A, carotene [32-34], and calcium [27-30] has also been suggested, although the evidence is less compelling.

Because adenomatous polyps are known precursors for colorectal cancer, three case–control studies [9-11] have explored their association with diet. Despite their limitations, each study has suggested a protective effect for fiber. A recent cohort study [12] of male health professionals found saturated fat and decreased fiber, as well as increased red meat-to-fish and meat-to-chicken ratio, to be risk factors for left-sided incident polyps. A small study by our group also showed that supplemental vitamins had no influence on the development of incident polyps in women [36].

No previous observational study has explored dietary risk factors for the recurrence of polyps; such recurrence has been used as an end point in chemoprevention trials to determine the efficacy of specific nutrients. One study failed to find a protective effect for vitamins C and E [37], and another [38] found no overall effect for a low-fat, high-fiber diet.

Our study found that increased caloric intake and saturated fat, as well as an elevated meat-to-fish and meat-to-chicken ratio, were associated with an increased risk for developing colorectal adenomatous polyps in women, whereas vitamin A was found to be protective. Interestingly, increased caloric intake and saturated fat were also associated with the recurrence of polyps in women, whereas vitamin A intake had a marginally protective effect. In addition, total fiber intake was associated with decreased recurrence of polyps. Vegetable intake did not play a role in any group.

Except for a protective effect of increased caloric intake, no consistent associations were observed for men. This difference between our observations for men and women is puzzling, because most studies have observed similar associations between dietary factors and risk for colorectal cancer in both men and women. Nonetheless, a randomized trial of polyp recurrence similarly found a protective effect for a low-fat, high-fiber diet in women but not in men [38]. One possible explanation is that the observed dietary risk factors operate through a reproductive hormonal mechanism or interact with some important hormonal factor. In particular, bile acids have been strongly implicated in the etiology of colorectal neoplasia [39-42] and appear to be highly influenced by various reproductive hormonal relations [43, 44]. Another possible explanation for the gender-related differences is that women have a better recall of dietary consumption than do men. Thus, more random misclassification errors would be introduced into the measurement of the diets of male participants, resulting in attenuation of any significant association. A third possibility is that the observed differences represent a quirk of our particular patient sample.

The greatest strength of our study was the selection of the control group from patients having colonoscopy done by the same colonoscopists. Because polyps are usually incidental findings in a given colonoscopy, their recognition is made more likely by selection factors that increase the likelihood of a given person having colonoscopy. These factors include the nature of the symptoms as well as the socioeconomic status and the general level of anxiety of the patient. By using a different control group, a study might simply determine risk factors for colonoscopy and not for incident polyps.

The use of a patient sample that had colonoscopy also allows the study of polyps proximal to the splenic flexure and insures minimal misclassification of controls. Given an estimated overall prevalence rate of 15% to 30% for polyps in an asymptomatic population of middle or advanced age [45], this degree of misclassification is not trivial.

The most obvious study limitation was the retrospective assessment of dietary intake. This is common to all case–control studies of diet and chronic disease. Another potential problem was a participation rate of just over 80%, although a similar compliance rate was achieved in all the case and control groups.

The most serious potential limitation of our incident polyp study is that, despite the advantages of using a “clean” colonoscopy group as our controls, these persons were not totally healthy and asymptomatic, as indicated by the need for colonoscopy. Thus, the observed associations may, in fact, reflect some systematic bias in the control group rather than a significant association with the case group.

Our findings indicate that dietary factors (specifically, increased saturated fat and decreased fiber intake) increase the risk for both the incidence and recurrence of adenomatous polyps in women. The lack of a similar association in men is also consistent in both the incident and recurrent studies; however, the reason is unclear. The findings with the recurrence of polyps represent the first observational study of this phenomenon and support the use of a low-fat, high-fiber diet as a chemopreventive regimen in intervention studies using polyp recurrence as an end point. Dietary modification may be warranted in patients, especially women, diagnosed with polyps. Further studies are needed to confirm and explain the observed gender differences.