Physical Activity, Obesity, and Risk for Colon Cancer and Adenoma in Men

Methods

Study Population

The Health Professionals Follow-up Study [36] was started in 1986; in that year, 51 529 male dentists, optometrists, osteopaths, podiatrists, pharmacists, and veterinarians in the United States between 40 and 75 years of age responded to a mailed questionnaire. They reported on their leisure-time physical activity (described below); current body weight; weight at age 21 years; height; history of cancer and other medical conditions; parental history of various diseases, including colorectal cancer; and use of aspirin and other nonsteroidal anti-inflammatory medications. They also reported dietary and alcohol intake using a validated [37, 38], semi-quantitative food-frequency questionnaire. We mailed an optional questionnaire in 1987 to assess waist and hip circumferences. In 1988, 1990, and 1992, we updated exposure information and ascertained newly diagnosed medical conditions and history of colonoscopy or sigmoidoscopy, including the indications for endoscopy: bleeding in stool, positive results from tests for occult fecal blood, abdominal pain, diarrhea or constipation, family history of colorectal cancer, routine screening without symptoms, or follow-up [39].

Most of the deaths in the cohort were reported by family members or by the postal system in response to the follow-up questionnaires. We also used the National Death Index to identify deaths among nonrespondents [40].

Assessment of Physical Activity

The 1986 questionnaire included a section about mainly recreational or leisure-time physical activity. The reliability and validity of questionnaires designed to assess physical activity have been investigated [41-43]. A questionnaire such as the one used in our cohort was evaluated in a cohort of U.S. nurses and was found to provide useful information [44], and similar results were found during a similar validation study done within the Health Professionals cohort (Chasan-Taber S. Personal communication). Participants reported the average time per week spent doing each of eight moderate and vigorous activities, choosing from among 10 possible responses that ranged from 0 minutes to 11 or more hours per week. The specific activities listed were walking or hiking outdoors (including walking during golf); jogging (slower than 10 minutes/mile); running (10 minutes/mile or faster); bicycling (including that done on a stationary machine); lap swimming; tennis, squash, or racquetball; and calisthenics or rowing. In addition, each respondent reported the number of flights of stairs he climbed daily and his usual walking pace.

The reported time spent at each activity per week was multiplied by its typical energy expenditure requirements expressed in metabolic equivalents (METs) [45] to yield a MET-hour score. One MET, which is the energy expended by sitting quietly, is equivalent to 3.5 mL of oxygen uptake per kilogram of body weight per minute for a 70-kg adult. For example, 1 hour per week of running contributed 10.2 MET-hours, 1 hour of tennis contributed 6 MET-hours, and 1 hour of walking at a moderate pace contributed 3.2 MET-hours. Body weight was excluded from the derivation of energy expenditure from physical activity to avoid confounding the expenditure variable by body weight. If more than one published intensity level was available for a given activity, the moderate or general MET value was chosen. An average MET value was assigned to the categories that listed more than one activity, such as rowing or calisthenics, and squash or racquetball.

Assessment of Anthropometric Variables

Each man reported his current weight and height and his weight at age 21 years on the 1986 questionnaire. In 1987, we mailed an optional questionnaire once to obtain additional exposure information, including body circumference measurements. We instructed each participant to measure (to the nearest quarter inch) his waist at the umbilicus and his hips at the largest circumference between the waist and thighs while standing and without measuring over bulky clothing [46]. We provided a tape measure and an illustration to help standardize the measurements. Sixty-five percent of the cohort responded.

We used the Quetelet index (kilograms/height in meters²) as a measure of total adiposity, waist-to-hip ratio to measure relative distribution of fat, and waist circumference to estimate total abdominal fat. Although the waist-to-hip ratio has been used more widely, waist circumference provides an estimate of absolute abdominal adiposity, the component most closely related to important metabolic abnormalities, including hyperinsulinemia, hypertension, and high triglyceride levels. To remove extraneous variation in the waist circumference because of height (taller men tend to have larger waist circumferences due to their larger body size rather than to obesity), we adjusted waist for height using residual analysis [47]. We first regressed waist on height using multiple linear regression and then added the residual to the average waist size (37.4 inches) for a man of average height (70 inches) to convert this measure back to the initial scale. This conversion created for each man a standardized waist circumference unconfounded by height.

We evaluated the precision of self-reported anthropometric measures among 123 cohort members who were part of a dietary validation study [46]. Trained technicians paid the substudy participants two visits, approximately 6 months apart, to measure current weight and waist and hip circumferences. The Pearson correlation between self-report and the average of the technicians' two measurements was 0.97 for weight, 0.95 for waist circumference, 0.88 for hip circumference, and 0.69 for waist-to-hip ratio. The men's self-measurements of their waist circumferences were 0.36 inches greater, their self-measurements of hip circumferences were 0.78 inches smaller, and their self-measurements of weight were 2.3 pounds less than the technician's measurements.

Identification of Patients with Colorectal Cancers

In 1988, 1990, and 1992, we asked each participant whether cancer had been diagnosed during the previous 2 years. The response rate to the follow-up questionnaires was 94% through 31 January 1992. When a participant (or a decedent's next-of-kin) reported a diagnosis of cancer of the colon or rectum, we sought permission to obtain hospital records and pathology reports. A study physician, blinded to exposure information, reviewed all medical records received and extracted data about histologic type, anatomic location, and stage of the cancer. Proximal colon cancers were defined as those from the cecum to and including the splenic flexure, and distal colon cancers were defined as those in the descending and sigmoid colon. We confirmed 249 new cases of colorectal adenocarcinoma (excluding carcinoma in situ), 90% by medical records and the remainder with corroborating information about diagnosis and treatment from the cohort member. Two hundred three cancers were in the colon and 46 were in the rectum.

Identification of Patients with Colorectal Adenomas and Controls

Because more than 90% of the adenomas were diagnosed during endoscopic procedures for screening or for unrelated gastrointestinal conditions, we restricted the adenoma analysis to men who had had a colonoscopy or sigmoidoscopy. This was done to reduce the potential for detection bias. Most procedures were sigmoidoscopies; thus, we analyzed only adenomas of the distal colorectum. Although we did not examine proximal colon adenomas, this should not have biased inferences for the distal colorectum. However, different causes for proximal adenomas may exist. A total of 12 879 men who did not meet any of the exclusion criteria (see Data Analysis) reported having had an endoscopy between 1986 and 1992. In 1993, we sent a mailing to a random sample of 200 controls (men who reported negative results from an endoscopy) to confirm that they did not have adenomas. After one mailing, 140 (70%) controls granted us permission to review the medical records of their endoscopic procedure; none had an unreported, histologically confirmed adenoma.

We were able to recontact 96% of the men who reported a diagnosis of polyp, and we received medical records in response to more than 96% of the requests sent to medical record departments, physicians' offices, or pathology departments. We considered for analysis only cases of adenoma confirmed by histopathologic report (including carcinoma in situ). After exclusion of all hyperplastic polyps and all adenomatous polyps proximal to the descending colon, 568 adenomas in the distal colorectum and 455 in the distal colon remained.

Data Analysis

We excluded from these analyses men who had completed none of the questions about physical activity; men who had reported previous cancer (other than nonmelanoma skin cancer), ulcerative colitis, or a familial polyposis syndrome at baseline; men with implausibly high or low scores for total energy intake (outside the range of 800 to 4200 kilocalories per day); or men who left 70 or more items blank. The dietary criteria for exclusion were based on a review of the records by the study investigators at baseline, which indicated that most of the men with extreme intake scores had misinterpreted the instructions. After exclusions, 47 723 men remained eligible for analysis.

Follow-up time began at the month during which the initial questionnaire was returned and ended at the month during which colorectal cancer was diagnosed, the month during which death occurred from other causes, or the end of the study period, which was 31 January 1992. We computed the incidence rate within a specified category by dividing the number of new cases by the number of person-years in the category; we calculated the relative risk by dividing the incidence rate by the rate in the specified reference group. We used the Mantel-Haenszel estimator and logistic regression to adjust for age (across 5-year categories) and potential confounding variables [48]. All relative risks were adjusted for age.

We used the same exclusion criteria for the adenoma analyses but also excluded men with a previous diagnosis of colorectal polyp. We analyzed the adenomas by size because large adenomas (≥ 1 cm) are more likely to reflect the influence of a tumor promoter. Men diagnosed with both adenomas and cancers were included only in the cancer analyses.

We did analyses on the subcohort (31 055 of 47 723 men) that responded to the 1987 questionnaire to evaluate the potential influence of adipose distribution on risk for colon cancer and adenoma. We first examined whether any observed associations with physical activity and body mass index were similar to those seen in the full cohort; if they were, we examined whether body fat distribution provided additional information over body mass.

For the analyses, physical activity and waist-to-hip ratio were divided into quintiles, and body mass index, height, and waist circumferences were divided into informative increments on the basis of an examination of the distribution of values. All decisions about categorization were made before any analyses were done.

Results

We examined physical activity and body mass index in relation to potential confounders of risk for colon cancer (Table 1). Men who were more physically active tended to be leaner, to smoke less, and to use multivitamins more often. Physical activity was also inversely related to total fat intake, primarily that of animal and not vegetable fat. The men who were more physically active tended to consume more dietary fiber and more nutrients related to multivitamin use. Except for smoking prevalence, which was highest among the leanest men, the covariate profile across levels of leanness was similar to that across levels of physical activity.

Physical Activity, Body Fat, Weight Change, Height, and Risk for Colon Cancer

Physical activity (in total MET-hours) was inversely associated with risk for colon cancer. Men in the highest quintile of physical activity had approximately half the incidence of colon cancer seen in men in the lowest quintile (Table 2). After we controlled for various factors, including body mass index, the inverse association was somewhat attenuated and remained statistically significant. The reduced risk for cancer was noted for the distal colon (for an increment of physical activity from lowest to highest quintile: age-adjusted relative risk, 0.50 [CI, 0.25 to 1.00]) but was weaker and nonsignificant for the proximal colon (relative risk, 0.75 [CI, 0.36 to 1.55]), although the CIs overlapped substantially. A nonsignificant direct association was seen between physical activity and rectal cancer (relative risk, 1.83 [CI, 0.83 to 3.84]), but this was based on only 46 cases.

We did an analysis, excluding the first 2-year follow-up period, to examine whether latent colon cancer could have biased our results by inducing a reduction in activity before diagnosis. The association became stronger (relative risk between high and low quintiles, 0.41 [CI, 0.22 to 0.74]), and the association in the second highest quintile became statistically significant (relative risk, 0.56 [CI, 0.32 to 0.96]). A similar activity questionnaire administered in 1988 yielded a similar association with risk for colon cancer in 1988-1992 (lowest compared with highest quintile: relative risk, 0.55 [CI, 0.31 to 0.98]).

Men with a higher body mass index had an elevated risk for colon cancer (Table 2). Although this risk was attenuated when we controlled for physical activity and other factors, the test for trend remained statistically significant. Body mass index at age 21 years was associated with a slightly elevated risk for colon cancer (high compared with low quintile: age-adjusted relative risk, 1.36 [CI, 0.90 to 2.07]), but this association did not persist (relative risk, 1.00 [CI, 0.60 to 1.67]) when we controlled for body mass index in 1986. Men categorized with both a high body mass and a low level of physical activity (defined by extreme tertiles) had an age-adjusted relative risk of 4.90 (CI, 2.59 to 9.27) relative to those in the low body mass and high activity level tertiles.

Men who had gained at least 40 pounds since age 21 years had an elevated risk for colon cancer (age-adjusted relative risk, 1.91 [CI, 1.15 to 3.16]). This relative risk was moderately attenuated when we controlled for physical activity (relative risk, 1.67 [CI, 0.99 to 2.82]). Weight gain of less than 40 pounds was only weakly associated with risk for colon cancer.

Attained adult height was related to elevated risk for colon cancer Table 2, although this association was slightly attenuated in the multivariate model. None of the covariates contributed substantially to this modest confounding; the confounding seemed to be attributable to the additive influence of several of the covariates.

Physical Activity, Obesity, and Risk for Colon Adenoma

Weak inverse associations with physical activity existed for adenomas of the distal colon (relative risk adjusted for age, previous endoscopy, and parental history of colorectal cancer, 0.79 [CI, 0.57 to 1.09], P for trend = 0.12) and for rectal adenomas (adjusted relative risk, 0.92 [CI, 0.56 to 1.52], P for trend = 0.55). The inverse association between physical activity and distal colon adenomas was primarily due to adenomas with a diameter of 1 cm or more (relative risk, 0.63 [CI, 0.36 to 1.10], P for trend = 0.08) but not to small adenomas. Body mass index was nonsignificantly associated with risk for both large (relative risk, 1.43 [CI, 0.78 to 2.62], P for trend = 0.38) and small adenomas (relative risk, 1.46 [CI, 0.90 to 2.38], P for trend = 0.13).

Body Mass Distribution and Risk for Colon Cancer and Adenoma

Associations between colon cancer and adenoma and body mass index and physical activity were similar in the subcohort that provided circumference measurements. Thus, we prospectively studied the relation between colon neoplasia and the waist and hip circumferences assessed in 1987. Waist adjusted for height, a measure of absolute abdominal obesity, and waist-to-hip ratio, a measure of relative distribution of adiposity, were strongly associated with risk for colon cancer (Table 3). The association with waist circumference was slightly stronger for distal colon cancer. Despite a correlation of 0.79 between body mass index and waist circumference, the positive association with waist circumference remained statistically significant even after adjustment for body mass index and physical activity (for waist > 39 inches compared with waist < 37 inches: relative risk, 1.94 [CI, 0.87 to 4.36], P for trend = 0.03). After adjusting for body mass, the association with waist-to-hip ratio persisted (relative risk, 3.05 [CI, 1.17 to 7.96]), although the test for trend was not statistically significant (P for trend = 0.09). After adjustment for waist-to-hip ratio, the positive trend between body mass index and colon cancer remained statistically significant (P for trend < 0.01).

Waist circumference and waist-to-hip ratio were also strong risk factors for large distal colon adenomas but were unrelated to small adenomas (Table 3). Waist-to-hip ratio remained a significant risk factor for large adenoma even after we controlled for physical activity and body mass (multivariate relative risk, 2.86 [CI, 1.28 to 6.41], P for trend = 0.04). A suggestive inverse association between large adenoma and physical activity (P for trend = 0.09) and a positive association with body mass index (P for trend = 0.09) became weaker after we controlled for waist-to-hip ratio (P = 0.21 for physical activity; P = 0.27 for body mass index). Absolute waist circumference was no longer statistically significant after adjustment for body mass, mostly because of widening CIs due to the high correlation between waist circumference and body mass.

Discussion

A moderate level of physical activity was related to a substantially lower risk for colon cancer in this cohort of middle-aged to elderly men. Obesity, particularly abdominal adiposity, increased the risk for colon cancer independently of physical activity level. Similar patterns were seen for large but not for small adenomas of the distal colon. These findings suggest that physical inactivity and obesity influence the promotion or growth of adenomas and are consistent with Hill's premise [49] that factors influencing the formation and growth of adenomas may differ. Although we had limited statistical power, we found—as did most previous investigators—that physical activity did not protect against rectal neoplasia. Our finding of a particularly high risk among men who both were physically inactive and had a high body mass index was consistent with findings from a recent report by Lee and colleagues [30] from the Harvard Alumni Study.

Our results are particularly informative at the low range of physical activity because the cohort consisted of men in professions generally requiring low levels of physical activity or who were retired. (The few possible exceptions were some veterinarians involved with the care of large animals, but they would account for only a small percentage of the cohort). The median activity level of 11.3 MET-hours/week for the cohort was low, approximately equivalent to 1 hour of running, 2 hours of tennis, or 3 hours of walking at a moderate pace. The median level is an underestimation because we assessed only eight common activities, but our results suggest that relatively modest levels of physical activity may substantially reduce the risk for colon cancer. However, up to at least 46.8 MET-hours/week (the 90th percentile), the higher the activity level, the lower the risk for colon cancer.

The deleterious effect of a sedentary lifestyle may at least partly account for the higher rates of colon cancer in industrialized countries and urban areas [50]. Our results confirm previous analytic epidemiologic studies showing an inverse association between physical activity level and risk for colon cancer [4-23], an association seen in several countries (the United States, China, Sweden, and Japan), among both men and women, and with both occupational and recreational activity. The association appears stronger with, but is not limited to, the distal colon. Obesity has also been shown to be directly associated with risk for colon cancer [13, 24-33] and adenoma [51, 52].

An unmeasured confounder is unlikely to account for all of these observations. Diet is perhaps the most important potential confounder to exclude because dietary factors probably influence risk for colon cancer and are difficult to measure and, thus, control. Several case–control [5, 17, 21, 22] and cohort [16, 19] studies have found that dietary factors did not confound the association between physical activity and risk for colon cancer. However, because the assessments of diet in the cohort studies were done using a limited dietary questionnaire [16] and 24-hour recall [19], it is unclear whether the investigators adequately controlled for long-term diet. We found significant dietary associations for colon cancer [53] and adenoma [54], indicating that our questionnaire was precise enough to adequately control for diet. We also controlled for nondietary factors, including smoking, aspirin use, family history of colorectal cancer, and screening behavior.

The influence of error on the results from the physical activity and anthropometric measures depends on the precision of the particular measure and on how closely these variables track over many years. Body mass and circumference measures are assessed precisely [46], but the measure of physical activity, although informative, may have appreciable error [41-44]. These variables appear to track over time; the correlation coefficient was 0.91 between reported body mass in 1986 and 1990 and 0.54 between MET-hours in 1986 and 1988. The time interval over which these variables act during carcinogenesis is also crucial. For example, recent evidence suggests that a positive association between cigarette smoking and colorectal cancer may have been missed because smoking may initiate cancers that are diagnosed only several decades later [55, 56]. The association of physical inactivity and obesity with large but not with small adenomas suggests that these variables may have an influence at relatively late stages of the disease, thus lessening (but not eliminating) the effect of any imprecision in assessing exposure over a limited period before the diagnosis. Most importantly, because our data were collected before the diagnosis, any imprecision was likely to cause underestimations rather than overestimations of associations.

Because abdominal obesity [57-60] and physical inactivity [61, 62] are strong independent determinants of insulin resistance and hyperinsulinemia, and because insulin is an important growth factor for colonic mucosal cells and colonic carcinoma cells in vitro [63-65], we hypothesize that hyperinsulinemia may mediate the effect of a sedentary lifestyle on risk for colon cancer. Consistent with this hypothesis, a diet high in refined sugars and low in dietary fiber, often linked to colon cancer [3], also causes hyperinsulinemia [66]. The male tendency toward abdominal distribution of fat and higher insulin levels may account for the stronger association between high body mass and risk for colon cancer in men than in women [13, 24-33]. Potential mechanisms whereby hyperinsulinemia may promote colon cancer have recently been reviewed in detail [67].

Other mechanisms may contribute to the beneficial influence of physical activity on risk for colon cancer. At the very low range of activity, physical activity probably stimulates colonic motility [68, 69], although the relation between activity and colonic transit time above a very low threshold level of activity is unclear [70-72]. Moreover, the influence of colonic motility on risk for colon cancer is not firmly established [3].

We found that increased stature, independent of obesity, was associated with an elevated risk for colon cancer. The few studies that have considered height as a risk factor have also shown this, with odds ratios of 2.1 for men and 1.6, 1.6, and 1.2 for women [33-35]. The mechanism underlying this association is unknown. Height may be a proxy for net energy intake during childhood, and some for the energy-balance effects may be mediated through insulin, insulin-like-growth factors, or their receptors and binding proteins. Inadequate nutritional intake in early life will stunt overall growth and organ cellularity in particular [73]. Adult stature correlates closely with the total length of the human colon [74], and the greater number of stem cells at risk for transformation in tall persons could be enough to put them at higher risk for colon cancer [75]. Alternatively, the rate of growth and, therefore, the rate of mitosis may determine the likelihood of the occurrence and the propagation of a genetic mistake.

Our findings support efforts to increase the level of physical activity and to reduce the prevalence of obesity in the population to lower the incidence of colon cancer. These findings require confirmation in other populations, especially women, among whom weak associations between total body mass and risk for colon cancer may have obscured the strong influence of abdominal obesity. Because of the accumulating evidence that a sedentary lifestyle is associated with risk for colon cancer, more intensive study of potential underlying mechanisms is warranted.