Risk Factors for Advanced Colonic Neoplasia and Hyperplastic Polyps in Asymptomatic Individuals FREE

Colorectal cancer is the second leading cause of cancer death in North America. Evidence exists that screening asymptomatic populations beginning at age 50 years can reduce mortality due to colorectal cancer^1- 8 and that removal of precursor adenomas may reduce the incidence of colorectal cancer.^9- 10 Identification of important risk factors for advanced colonic neoplasia could inform both risk stratification and development of risk reduction strategies.

Many potential risk factors for colorectal neoplasia have been investigated previously.^11- 12 Few studies have evaluated risk factors for advanced colorectal neoplasia in asymptomatic individuals, compared risk factors between persons with and without polyps, or included most purported risk factors in a multivariate analysis. We conducted a cross-sectional study to determine risk factors for advanced colonic neoplasia in asymptomatic persons undergoing a complete screening colonoscopic examination. The unique features of this study are (1) the inclusion of persons known to be free of colonic polyps based on colonoscopy as a comparison group vs patients with advanced neoplasia and (2) a sample size sufficient to analyze the most important risk variables in a multivariate analysis. Suspected risk factors were selected based on associations found in prior studies and biological plausibility. We studied family history of colorectal cancer, diet, smoking, alcohol use, nonsteroidal anti-inflammatory drug (NSAID) use, body mass index, prior cholecystectomy, serum cholesterol level, and physical activity as potential risk factors.

We performed a secondary analysis of patients whose worst colonic lesions were 1 or more hyperplastic polyps. Hyperplastic polyps are benign and common.^13- 14 Current guidelines suggest that patients with hyperplastic polyps be treated similarly to patients with no polyps.¹⁵ No prior studies have evaluated risk factors in patients with hyperplastic polyps to determine if they resemble those in patients with no polyps. Our study provided an opportunity to determine risk variables for patients with hyperplastic polyps compared with those in patients with no polyps and with those in patients with advanced neoplasia.

Patients aged 50 to 75 years who did not have lower gastrointestinal tract symptoms were recruited from 13 Veterans Affairs medical centers between February 1994 and January 1997.¹⁶ The study protocol was approved by a central human rights committee and by institutional review boards at each participating center, and all participants provided written informed consent. Patients completed a clinic survey to determine eligibility. Patients were excluded if they reported symptoms of lower gastrointestinal tract disease, including rectal bleeding (on ≥1 occasion in the previous 6 months), a marked change in bowel habits, or lower abdominal pain that would normally require medical evaluation. Other exclusion criteria were current participation in other studies, history of disease of the colon (such as colitis, polyps, or cancer), prior colonic surgery, a colonic examination (ie, sigmoidoscopy, colonoscopy, or barium enema) within the previous 10 years, a medical condition that could increase the risk associated with colonoscopy or that would preclude a benefit from colonoscopic screening (cancer or any terminal illness), a prosthetic heart valve, anticoagulant therapy, nonmedical problems (psychiatric disorders, lack of transportation, homelessness or lack of support at home, or excessive use of alcohol), or a need for special precautions in performing colonoscopy (such as antibiotic prophylaxis). Women with childbearing potential were excluded. A total of 17 732 patients were screened for enrollment; 2346 patients declined to complete the clinic survey. The most common reasons for exclusion were colonic examination in the previous 10 years (n = 6486), history of colon disease (n = 2221), serious medical disorder (n = 1052), and anticoagulation therapy (n = 712). Among patients who met the eligibility criteria, 1463 (31.4%) declined to participate and 3196 eligible patients were enrolled. Patients who had complete colonoscopy examinations were included in this analysis (n = 3121). We determined our sample size requirements to have sufficient power (>90%) to detect a clinically important doubling of risk for key categorical risk factors. We analyzed the 329 patients with advanced neoplasia and 1441 without polyps (N = 1770) (Table 1).

Procedures for performing colonoscopy and histologic evaluation have been described previously.¹⁶ Advanced colonic neoplasia was defined as an adenoma of 10 mm or more, a villous adenoma (at least 25% villous), adenoma with high-grade dysplasia, or invasive cancer. Patients with a pathologic interpretation of intramucosal carcinoma or carcinoma in situ were classified in the high-grade dysplasia group. Cancer was defined as invasion of malignant cells beyond the muscularis mucosa. Patients were classified based on the most histologically advanced lesion. A second cohort of patients in whom hyperplastic polyps were the worst colonic lesions was studied to determine if their risk factors differed from those of patients with no polyps or from those of patients with advanced neoplasia.

Diet. Diet was assessed with a semiquantitative food frequency questionnaire that has been validated in prior studies.^17- 18 Patients were excluded if they left 70 or more items blank or reported implausibly high or low scores for total energy intake. Based on this questionnaire, the following measurements were made in each patient: total energy intake; total fat; fat derived from beef, pork, or lamb; total dietary fiber; fiber from cereal, fruits, and vegetables; total calcium; total folate; vitamin D; and daily multivitamin use.

Physical Activity. Two methods were used to assess the impact of regular daily physical activity and the role of sporadic but vigorous exercise. The physical activity index was adopted from the Framingham Study.¹⁹ Patients were provided definitions of very light, light, moderate, and heavy activity for home, occupation, and recreation and recorded their time spent in each category to generate a 24-hour score. The second measure used a calculation of metabolic equivalents, a well-validated assessment of energy expenditure during specific forms of moderate to heavy exercise, expressed as metabolic equivalents per week.^20- 21 The 2 scoring systems reflect different profiles of physical activity. On the physical activity index, an individual who engages in many hours per day of mild activity may have a similar score to a person who performs a short period of exertional activity. In contrast, the metabolic equivalent score discriminates individuals who engage in moderate to heavy physical activity; an individual with several hours of mild activity will have a lower score than a person with short periods of heavy exertion.

Alcohol, Smoking, and NSAIDs. Past and current smoking habits were ascertained. Patients were queried regarding current use, number of cigarettes per day, and number of years of smoking. For prior smokers, the year of cessation, number of years of smoking, and number of cigarettes per day were determined. The number of pack-years was calculated for both groups. Current and past use of alcohol was measured by ascertaining the number of drinks per week, defined as wine (4 oz per glass), beer (12 oz per can or bottle), or liquor (1 oz per shot), each equivalent to approximately 11 g of alcohol per drink.

Use of aspirin or other NSAIDs was ascertained by a query that included every known NSAID. This list was revised and updated during the study as new products became available. Patients indicated specific product used, daily dosage, and duration (in years) of consumption. The data were analyzed based on current use (never vs occasional or daily) and on duration of any use.

Other Factors. Body mass index was calculated as weight in kilograms divided by the square of height in meters. History of prior cholecystectomy and date of surgery were ascertained. Serum cholesterol was measured from a fasting blood sample.

Study database management and all statistical analyses were performed with SAS software, version 6.12 (SAS Institute Inc, Cary, NC). Descriptive statistical analyses included calculation of rates and proportions for categorical data and means and SEs for continuous data. In addition, standard logistic regression methods were used for calculating relative risks as odds ratios (ORs) with 95% confidence intervals (CIs).²² In the analysis of dietary and other continuous variables, participants were categorized according to quintiles ranging from lowest to highest values, with the lowest quintile serving as the reference group. Dietary variables were adjusted first for age and then for age plus total energy intake and medical history factors (family history, body mass index, smoking, alcohol use, physical activity index, and NSAID use). Nondietary risk factors were adjusted for age.

A primary objective was to assess in a multivariate model the relative importance of the risk factors. These factors fell into general classes within which there were multiple measures (eg, physical activity; dietary fiber as total and cereal, fruit, and vegetable). The development of the multivariate model required variable selection to avoid problems of redundancy and overspecification. In cases of ambiguity, a statistical approach involving stepwise regression procedures was used within classes of factors to select the most statistically relevant variables. The multivariate model included the following list of variables, rescaled with reasonable increments: family history (≥1 first-degree relatives with colorectal cancer vs none); body mass index (continuous in 5-unit increments); physical activity index (continuous in 5-unit increments); smoking (current vs never or past); alcohol use (continuous by number of servings per week); NSAID use (any vs never); beef, pork, or lamb as main dish (continuous as a percentage of total energy intake) representing the fat class of dietary factors; cereal fiber (continuous in 1-g increments) as the fiber variable; and vitamin D (continuous in 100-IU increments). For all analyses, P<.05 was considered statistically significant.

A total of 3121 persons had complete colonoscopy to the cecum and have been previously described.¹⁶ Men accounted for 96.8% of this cohort, so these results cannot be generalized to women. Patients were classified by the worst histologic lesion found in the colon. A total of 1441 patients had no polyps, 118 (3.8%) had nonadenomatous lesions, 391 (12.5%) had hyperplastic polyps, and 842 (27.0%) had 1 or more tubular adenomas less than 1 cm in diameter. Three hundred twenty-nine patients (10.5%) had at least 1 advanced neoplastic lesion in the colon and are the subject of this analysis. These patients were compared with 1441 patients who had no polyps or growths in the colon. Complete medical history information was obtained for all patients. Complete dietary information was obtained for 312 patients with advanced neoplasia (94.8%) and for 1359 patients with no neoplasia (94.3%). Patients with advanced neoplasia were older than patients with no neoplasia (65.1 years vs 62.7 years; P<.001). All ORs are adjusted for age.

The results of the age-adjusted analysis of medical history risk factors are summarized in Table 1. Patients with at least 1 first-degree relative with colorectal cancer had an increased risk of advanced neoplasia (OR, 1.68; 95% CI, 1.21-2.31), which was higher if patients had 2 first-degree relatives with colorectal cancer. Sixteen (28%) of 57 participants with young index relatives (<60 years) had advanced neoplasia compared with 41 (25%) of 164 participants with older index relatives (≥60 years), a nonsignificant difference (OR, 0.77; 95% CI, 0.39-1.51).

Other selected patient characteristics are presented in Table 1. There were no significant differences in risk of advanced neoplasia when whites were compared with blacks and Hispanics. There was no association between body mass index and risk of advanced neoplasia. A physical activity index level of greater than 36 was associated with reduced relative risk of advanced neoplasia (OR, 0.67; 95% CI, 0.49-0.93). Metabolic equivalents per week did not demonstrate an association with risk. No differences were found in serum cholesterol levels or rates of cholecystectomy among patients with vs without advanced neoplasia.

Current smoking (Table 1) was strongly associated with increased risk (OR, 2.12; 95% CI, 1.49-3.01). Risk was increased for current users who had more than 25 pack-years of smoking and for past smokers of more than 49 pack-years before cessation of smoking. Current alcohol consumption was associated with increased risk if participants were consuming an average of 1 or more alcoholic drinks per day (OR, 2.12; 95% CI, 1.44-3.14).

Occasional or daily use of NSAIDs was associated with a decreased risk of advanced neoplasia. When the duration of use was analyzed as a continuous variable, the age-adjusted OR for advanced neoplasia was strongly associated with increased duration of use (OR, 0.83 for 10 years of use; 95% CI, 0.73-0.95).

Dietary variables were adjusted for age, total energy consumption, and medical history risk factors (Table 2). Total animal fat consumption had a weak, nonsignificant association with advanced neoplasia. A stronger association was observed for the highest quintile of fat intake derived from red meat. Participants who consumed beef, pork, or lamb as a main dish 5 or more times per week had an increased relative risk of advanced neoplasia compared with those who did not consume these products (OR, 2.73; 95% CI, 1.05-3.63). Among the components of fiber intake, only cereal fiber was associated with a reduced relative risk of advanced neoplasia. Statistically significant reductions in relative risk of advanced neoplasia were found with increased intake of total calcium, total folate, and vitamin D in the highest quintile compared with the lowest quintile. Daily consumption of multivitamins was associated with a significant reduction in risk (OR, 0.75; 95% CI, 0.56-0.99).

Patients with hyperplastic polyps were compared with the cohort with no polyps (Table 3). There were no risk variables that distinguished the 2 groups except for past and current smoking. Past (OR, 1.65; 95% CI, 1.23-2.23) and current (OR, 2.71; 95% CI, 1.93-3.81) smokers had a significantly increased risk of hyperplastic polyps compared with polyp-free controls. Daily use of NSAIDs was associated with a decreased OR for hyperplastic polyps (OR, 0.75; 95% CI, 0.56-0.99). When we compared risk variables in patients with hyperplastic polyps with patients with advanced neoplasia, results were generally similar to those in patients with no polyps. Increased age was associated with increased risk of advanced neoplasia, and higher intake of vitamin D and multivitamins was associated with a lower risk. Other associations were not statistically significant.

A multivariate model including each medical history factor and 1 factor from each of the dietary groups was developed (Table 4). Dietary variables were selected statistically by within-category stepwise regression. This analysis identified cereal fiber, vitamin D, and beef/pork/lamb as main dish (as percentage of total energy intake) to be the best candidates for the model. In the multivariate analysis, factors associated with a significantly increased risk of advanced neoplasia were family history, current smoking, and moderate to heavy current alcohol use. Risk factors associated with a significantly decreased risk of advanced neoplasia were use of NSAIDs, cereal fiber intake, and vitamin D. Vitamin D dominated total calcium and folate intake in the multivariate analysis. Higher levels of physical activity were associated with a marginal non–statistically significant reduction in risk.

The current study is the first cross-sectional study to assess a wide range of risk factors for advanced colonic neoplasia in asymptomatic individuals who have had a complete colon examination. The choice of advanced colonic neoplasia as the study end point was designed to identify risk factors in the patients most likely to develop malignancy, with the goal of developing preventive strategies. This end point has some limitations since many patients with advanced neoplasia will not develop colon cancer in their lifetime.

Several risk factors were not informative. Prior studies have found weak associations between colorectal neoplasia risk and body mass index,^23- 24 serum cholesterol level,^25- 26 and prior cholecystectomy.^27- 29 These variables were not significantly associated with advanced colonic neoplasia in our study, but modest associations cannot be excluded.

There was discrepancy between the results of 2 indexes used to assess physical activity. Patients who scored in the higher range of the physical activity index, with either longer periods of mild exercise or some heavy activity, had a lower age-adjusted risk of advanced neoplasia. There was no difference in risk across quintiles of the metabolic equivalent score. These findings suggest that individuals may be just as likely to benefit from long periods of mild physical activity as from shorter periods of heavy activity. In the multivariate analysis, physical activity had a marginal, non–statistically significant benefit.

The relationship of dietary fat and fiber and colorectal neoplasia is uncertain. Our analysis found no significant increase in risk of advanced neoplasia associated with total fat and a marginal, nonsignificant increase associated with fat derived from red meat sources. Although we did not find a strong relationship between fat and advanced adenomas, we cannot exclude the possibility that dietary fat may promote malignant change in precursor adenomas, thus leading to higher rates of cancer. The current data show that among the fiber variables, the highest quintiles of cereal fiber intake were associated with lower risk of advanced neoplasia in the univariate model, and the relationship remained after adjustments in the multivariate model. Prior interventional studies with fiber have failed to demonstrate a protective effect over a 3- to 4-year study period.^30- 31 However, these studies used any adenoma (large or small) as an end point and evaluated a short period of intervention. Our results do not necessarily conflict with these studies. Our data suggest that higher intake of cereal fiber over a long period may protect against development of advanced neoplasia.

Vitamin D emerged as an important protective factor. Several prior studies have suggested that vitamin D and, in particular, its active form (1,25-hydroxyvitamin D) exerts an antiproliferative effect on colonic mucosa in vitro,³² and several clinical studies have suggested a beneficial effect.^33- 35 Some prior studies have found a marginal benefit from calcium supplementation in the prevention of adenomas.^36- 37 In our multivariate analysis, dietary vitamin D dominated calcium as a beneficial risk factor. We did not measure sunlight exposure in our patients, which can influence 25-hydroxyvitamin D levels. These data suggest that vitamin D may be an important protective factor, and further study is warranted to determine the relative contributions of diet vs sunlight. The impact of total folate and daily multivitamin intake are difficult to separate in the analysis; both factors were associated with a protective effect in the univariate analysis.

We found that a history of 1 or more first-degree relatives with colorectal cancer is strongly associated with advanced neoplasia, confirming results from many other studies.^38- 41 When we compared the risk of advanced neoplasia in participants with family members diagnosed as having cancer either before or after age 60 years, we did not find a significant difference. Therefore, our data reinforce recommendations to treat all patients who have a first-degree relative with colorectal cancer as a high-risk group and offer screening with colonoscopy.¹⁵

Current cigarette smoking and consumption of more than 7 drinks of alcohol per week were strongly associated with increased risk, consistent with prior studies.^42- 46 Although the physiologic mechanisms are unclear, these data provide further reason for health care practitioners to advise smoking cessation and moderation of alcohol consumption.

Participants taking NSAIDs on a regular basis had a lower risk of advanced neoplasia compared with those who did not use these drugs, consistent with many prior studies.^47- 51 There was a strong relationship between duration of drug exposure and reduction in risk. This study precedes the availability of the cyclooxygenase-2 inhibitor drugs, which have been shown to reduce the development of adenomas in patients with familial polyposis.⁵² More than 80% of our participants already were taking aspirin as prophylaxis for cardiovascular disease or other NSAIDs for arthritis and pain. Should adults take NSAIDs solely for prevention of colonic neoplasia? Some intervention studies have not shown a benefit.^53- 54 Moreover, there are potential harms. The risk of peptic ulcer disease and gastrointestinal bleeding⁵⁵ and other health effects must be balanced against the potential benefits. The decision to begin therapy with aspirin or other NSAIDs solely for reduction of risk of colonic neoplasia requires further study.

In a secondary analysis, we determined risk factors in patients whose most serious colonic lesions were 1 or more hyperplastic polyps. Most guidelines regard hyperplastic polyps as benign,¹⁵ and several large screening studies have found that distal hyperplastic polyps are not risk factors for proximal neoplasia.^16,56 We hypothesized that if hyperplastic polyps are incidental, nonneoplastic growths, patients with hyperplastic polyps should closely resemble patients with no polyps. No prior study has determined if there are any distinguishing risk variables for hyperplastic polyps. Our results suggest that patients with only hyperplastic polyps are similar to polyp-free controls. Only past or current smoking emerged as a significant risk factor. This is of interest because smoking is the only risk factor associated with both hyperplastic polyps and advanced neoplasia. Smoking may promote DNA hypermethylation, and development of advanced adenomas may depend on methylenetetrahydrofolate reductase genotype.⁵⁷ When we compared the advanced neoplasia cohort to patients with hyperplastic polyps, we found that higher intake of vitamin D intake and multivitamins was associated with a lower risk of advanced neoplasia. Other differences were not statistically significant. Since this was a secondary analysis, our study was not powered to detect such differences. In addition, patients with advanced neoplasia as their worst lesions may also have hyperplastic polyps, creating some overlap between groups.

There are several important limitations of our study. First, this is a cross-sectional study. Some of the purported risk factors may not be associated with prevalent advanced neoplasia but could be risk factors for subsequent incident lesions. Follow-up of this population is planned. This analysis focuses on risk factors in a high-risk population with advanced neoplasia. Cancer progression is not inevitable in this population, and other factors may promote progression from advanced neoplasia to cancer. Second, some results may be subject to recall bias. The food frequency questionnaire has been validated in many studies^17- 18 but does not measure whether dietary habits have changed over time. A total of 38.7% of our participants reported a major dietary change over the past 10 years. Similar rates of diet change were found in patients with and without advanced neoplasia. Therefore, we cannot exclude the possibility that prior dietary behaviors could have had an effect on current risk of advanced neoplasia. Third, our results can only be generalized to a male population; further study of risk factors in women is needed. Also, the veteran population may differ from other male populations in the United States. Finally, we recognize the complexity of analyzing confounding variables. Therefore, an important feature of this study was the inclusion of all major purported historical and dietary risk factors.

An important strength of this study is that asymptomatic patients had complete structural examinations of the colon to define the at-risk population, reducing the likelihood of misclassification. The study was sufficiently powered to examine a broad range of potentially confounding risk variables.

What new information have we learned from this study? This is one of the first studies to identify vitamin D as a protective factor. Because vitamin D and calcium are confounders, only a study of sufficient sample size can determine the relative benefits of either vitamin D or calcium. We were able to determine the relative risk and benefit of obesity and exercise. We conclude that obesity alone is not a significant risk factor and that regular exercise can be protective. There have been many prior publications citing the benefits and risks of high-fiber or high-fat diets. Our data suggest that diets high in cereal fiber are protective and that diets that include large amounts of fat derived from meat may increase the risk of advanced neoplasia. The risk associated with smoking and alcohol is now confirmed in a large multivariate analysis. Many prior studies have analyzed these variables without considering other risk factors. Finally, we found that patients with hyperplastic polyps as the most serious finding in the colon closely resemble polyp-free controls. Past or current smoking was the only significant risk factor for hyperplastic polyps. This new information is consistent with recommendations to treat hyperplastic polyps as insignificant colonic lesions.

In conclusion, these results show that several risk factors are positively associated with prevalence of advanced colonic neoplasia, including family history of colorectal cancer, current smoking, and current moderate to heavy alcohol consumption. Factors with an inverse association include use of NSAIDs and higher levels of consumption of cereal fiber and vitamin D. Marginal, nonsignificant protective factors were moderate physical activity and total calcium, total folate, and multivitamin intake.

Our study was not designed to demonstrate that modification of risk factors can modulate risk of colonic neoplasia. Nevertheless, it is prudent to recommend that patients stop smoking, reduce alcohol intake, and exercise regularly as part of general preventive health measures. Consuming vitamin D plus a calcium supplement or regular dairy products represents a low-risk strategy that may benefit patients. The benefit of a daily multivitamin is uncertain and requires further study but is associated with very low risk to patients. The strong association with family history of colorectal cancer reinforces existing recommendations to offer screening with colonoscopy. The potential protective effect of NSAIDs must be carefully balanced against the risks. Further study is needed to determine if risk factors at baseline colonoscopy are predictive of future incidence or recurrence of advanced neoplasia.