Physical Activity in Relation to Colon Cancer in Middle-aged Men and Women

Transcription

1 American Journal of Epidemiology Copyright O 1996 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved Vol. 144, No. 1 Printed In L/.&A Physical Activity in Relation to Colon Cancer in Middle-aged Men and Women Emily White, Eric J. Jacobs, and Janet R. Dating A population-based case-control study was conducted to assess the relation between physical activity and colon cancer among men and women aged 3-62 years. were 251 men and 193 women diagnosed with colon cancer in in three counties in the Seattle metropolitan area who were identified from the Seattle-Puget Sound Surveillance, Epidemiology, and End Results registry. were 233 men and 194 women identified by random digit telephone dialing who were selected by stratified random sampling to approximate the age, sex, and county distribution of cases. Physical activity was assessed by questions on frequency and duration of types of recreational and occupational activities during the 1-year period ending 2 years before diagnosis. Each activity was classified as low intensity (<4.5 METs) or moderate to high intensity (^4.5 METs). For men and women combined, moderate or high intensity recreational activity was associated with a decreased risk of colon cancer (relative risk () for two or more times per week vs. none =.7, 95% confidence interval (Cl).49-). This relation was stronger for men than women. Occupational activity was not associated with colon cancer, except among men younger than 55 ( for a 14.5 hours per week of moderate activity vs. none =.29, ). Among men and women combined, total moderate or high intensity activity (occupational plus recreational) was marginally related to colon cancer ( for ^5 hours per week vs. none =.78, ). These results were adjusted for age (and sex in the combined sex analyses) and were not confounded by body mass Index, dietary factors, or other measured health behaviors. The results of this study provide modest support to the growing number of studies showing that recreational and/or occupational physical activity is associated with a reduced risk of colon cancer. Am J Epidemiol 1996;144:42-5. colonic neoplasms; exercise; occupations In the mid-198s, three studies reported an increased risk of colon cancer in men with sedentary occupations (1-3). This association was intriguing because it was a possible explanation for the different time trends between men and women in colon cancer incidence over the middle of this century: Men experienced a 53 percent increase in colon cancer incidence over the 4-year period , whereas among women, the increase was only 17 percent over the same period (4). This corresponded to a decline in job activity in the United States during this century as our economy shifted from agriculture to manufacturing to service jobs (5), a decline that has affected men primarily. The percentage of men in active jobs (manual and farm) declined from 7 percent in 193 to 52 percent in 197, and the percentage of women in such jobs remained low (5). Received for publication December 24, 1994, and accepted for publication October 25, From trie Division of Public Health Sciences, Fred Hutchinson Cancer Research Center, Seattle, WA, and the Department of Epidemiology, University of Washington, Seattle, WA. Since the early studies on job activity, a large number of additional studies have investigated the risk of colon cancer associated with occupational, recreational, or total activity among men and women, with fairly consistent findings that increased activity decreases risk (6-23). However, few of these studies collected detailed information on both recreational and occupational activity, none reported on household activity, and few had complete information on the potentially confounding effects of diet and other factors associated with physical activity. We conducted a population-based, case-control study of the relation between physical activity and colon cancer among men and women in western Washington State. To ensure a large percentage of physically active study subjects, the study was limited to persons who were age 62 years, the median age of diagnosis of colon cancer, or younger. Detailed information was collected on frequency and intensity of occupational, recreational, and household activity so that physical activity could be quantified in terms of estimated metabolic equivalents (METs) in kilocalo- Downloaded from at Pennsylvania State University on February 27,

2 Physical Activity and Colon Cancer 43 ries expended per kilogram of body weight. An additional advantage of this study is that many Washington State residents were early participants in the trend toward increased recreational physical activity (24). METHODS Selection of cases and controls All incident cases of invasive adenocarcinoma of the colon diagnosed between July 1985 and September 1989 among 3- to 62-year-old residents of King, Pierce, or Snohomish County were identified through the Seattle-Puget Sound Surveillance, Epidemiology, and End Results registry. A total of 659 cases were identified. Of these, 12 died before being approached for the study. A population control group was selected by random digit dial calls made to 2,51 numbers, of which 9,384 were determined to be households. For 4.5 percent of the numbers called, it was not possible to determine whether the telephone was residential (busy signal or no answer for 2 calls). Because this percentage is consistent with the percentage of numbers that are nonworking or assigned to telephone booths, these losses were not included in the response rates below. Household screening to determine the age and sex of adults in the household was successful for 9,51 (97 percent) of the 9,384 households identified. Of those screened, 549 subjects were selected by stratified random sampling to approximate the age, sex, and county distribution of the cases. Inclusion criteria were similar for cases and controls, in particular, absence of antecedent colon or rectal cancer, polyposis, or inflammatory bowel disease; current residence in a private household in one of the three selected counties; white race; and ability to communicate adequately in English. The study was limited to whites because of the small percentage of minorities in the area. Among the cases, 55 were found to be ineligible (11 had prior colorectal cancer, 1 had polyposis or inflammatory bowel disease, 1 had no phone or lived in an institution, 14 had moved from the Seattle-Puget Sound area, six had participated in a prior study, and four were not able to communicate adequately). Permission to contact 39 cases was not granted by the treating physician, 16 cases refused to participate, and three were not reachable. Thus, 88 percent (444 of 52) of the alive, potentially eligible cases completed the interview for an overall case response proportion of 75 percent (.88 X.85, the proportion of cases who were alive). Among the controls, 26 were found to be ineligible (19 were nonwhite, three had colorectal cancer, three had polyposis or inflammatory bowel disease, and one could not communicate adequately). Ninety-four potentially eligible controls refused at screening or refused or were not reachable at the interview, and two died before the interview. Thus, 82 percent (427 of 523) of potentially eligible controls completed the interview for an overall estimated control response proportion of 79 percent (.82 X.97, the proportion of residential phone numbers that were successfully screened). Ascertainment of physical activity and covariates A structured telephone interview was conducted that covered demographic characteristics, physical activity, bowel habits, height and weight, vitamin/mineral supplement use, and alcohol consumption. To minimize any influence of symptoms of colon cancer on the exposures of interest, a reference date was set as 2 years before diagnosis for the cases or a corresponding date for the controls. In addition, a semiquantitative food frequency questionnaire was mailed to all subjects and was completed by 96 percent of those interviewed (424 cases, 414 controls). This instrument has been described previously (25). Briefly, information on consumption of 71 foods and nine beverages was taken into account in calculating energy and nutrient intakes. Subjects were asked about lifetime occupational physical activity and about recreational and household physical activity during a 1-year reference period ending at the reference date. Recreational physical activity was ascertained using a questionnaire similar to that developed by Taylor et al. (26). Subjects were asked to recall each activity they performed on a regular basis, i.e., at least twice per month in any year over the 1-year reference period. The subjects were asked for the age they started and ended each activity, the months per year the activity was performed, and for those months, the frequency (times per week or month) and duration (hours and/or minutes) they performed each activity. Each type of recreational activity was assigned an intensity code, in units of METs, using Ainsworth et al.'s Compendium of Physical Activities (27). To create the MET activity variable, we calculated the total hours in each activity during the 1-year period and multiplied this by the intensity code for that activity, summed over all activities and divided by 52 (weeks in 1 years), to yield METs per week during the 1-year period. Because the MET intensity code is an approximate measure of kilocalories per hour per kilogram of body weight, the total MET activity variable is a measure of energy expended in recreational activity that is independent of body weight (total kilocalo- Downloaded from at Pennsylvania State University on February 27, 214

3 44 White et al. ries per kilogram). Each activity was also classified as low intensity (<4.5 METs) (e.g., walking), moderate intensity ( METs) (e.g., dancing), or high intensity (^6. METs) (e.g., running) as suggested by Taylor et al. (26). Occupational activity was assessed by a modification of the method developed for the Framingham Study (28). For each job held more than 1 year, subjects were asked to report the job title and industry, the year they began and ended each job, and the average days per week and the average hours per day worked. For each job, subjects were then asked to divide a typical work day into hours spent in each of the following activity levels: 1) sitting or standing; 2) walking; 3) "moderate" physical work, such as frequently rifting or carrying up to 25 pounds; and 4) "heavy" physical work, such as frequently lifting or carrying 25 pounds or more. In calculating the job physical activity variables, only categories 2-4 were included since sitting has a MET of 1. and standing, a MET of 1.2 (27). Intensity codes were assigned to each level of job activity: category 2, 3.5 METs; category 3, 4. METs; and category 4, 5.5 METs, based on similar work activities reported by Ainsworth et al. (27). From this information, we calculated hours per week in nonsedentary work (categories 2-4), hours per week in moderate intensity work (category 4), and METs per week of work. Averages per week were averages over the subject's entire working life from age at first job to reference age or over the entire 1-year reference period, even if the subject had not been employed the entire time. Mean years of work before the reference date were similar for male cases (34.5 years) and controls (34. years) and for female cases (3.6 years) and controls (3.4 years). For household activity, the subjects were asked the number of hours per week they spent in household chores or active child care that required walking or physical work during the 1-year reference period. All housework was assumed to be low intensity (3.5 METs), based on the MET code for cleaning (3.5) in Ainsworth et al. (27). Data analyses Unconditional logistic regression was used to estimate the relative risk (estimated by the odds ratio) for colon cancer and its 95 percent confidence interval associated with each level of each factor of interest after adjustment for other covariates. Variables were converted from continuous variables to ordered categorical variables based on meaningful cut points or on tertiles or quartiles of the distribution among the controls. When the distribution of a variable differed substantially for men and women, the categorization was done separately for men and women. In addition, a test for trend was based on the significance of a single trend variable coded as the category of exposure. All analyses were adjusted for age by inclusion of five age categories (ages, in years, 3-44, 45-49, 5-54, 55-59, and 6-62) in the model. Analyses of men and women combined were also adjusted for sex. No other measured factors confounded the associations presented (see Results). The relation of physical activity to colon cancer within certain subgroups of the population (e.g., age groups) was modeled by separate analyses within each subgroup of interest. The presence of effect modification was tested on the entire sample by use of an interaction term between the physical activity trend variable and the group variable. RESULTS In table 1, the distribution of cases and controls by selected factors related to colon cancer in this study and/or factors known to be related to physical activity (29) are presented. Among both men and women, cases reported greater alcohol intake; and among men only, cases reported greater energy intake and had higher body mass index. Constipation was also more common among both male and female cases. Frequency of any recreational activity during the 1-year reference period was associated with decreased risk of colon cancer for men and women combined (table 2). However, the lowest risk was for exercise two to less than four times per week with less reduction in risk for exercise more often or less often. When recreational physical activity was limited to moderate or high intensity activities, there was a decreasing risk with increasing frequency for men and women combined. intensity activity (e.g., running) at least once per week was protective for men and women combined. For each of these relations, the pattern of risk was statistically significant (or of borderline significance) for men, whereas for women, the patterns were weaker and nonsignificant (table 2). Other methods of classifying recreational activity (by hours or METs) showed similar but weaker associations (table 2). Because men reported many more hours of occupational activity than women, and women reported many more hours of housework, only these associations are presented separately for men and women (table 3). For both sexes, there were no clear associations between colon cancer risk and reported hours of nonsedentary job activity, moderate intensity job activity, or jobrelated energy expenditure (METs) over the 1-year reference period. However, when the analyses in table Downloaded from at Pennsylvania State University on February 27, 214

4 Physical Activity and Colon Cancer 45 TABLE 1. Comparison of colon cancer cases to population controls on factors related to colon cancer or physical activity, western Washington State, Age at diagnosis (years) Alcohol intake (g/day)* Energy intaket Lowest quartile Quartile 2 Quartile 3 est quartile Body mass indext Lowest quartile Quartile 2 Quartile 3 est quartile Constipation* (times/year) (n = 251) Men (n = 233) Women (n=193) (77=194) * During the 1-year reference period ending 2 years before diagnosis. t Based on 238 case men, 224 control men, 186 case women, and 19 control women who completed a food frequency questionnaire on estimated food intake during the year midpoint in the 1-year reference period. Quartiles (kcal/day) for men:,567.83, 1, ,13.3, 2,13.4-2,866.69, 2, Quartiles for women:,368.29, 1, ,698.51, 1, ,155.74, 2, \ At reference date (2 years before diagnosis for cases and similar date assigned to controls). Quartiles (kg/m 2 ) for men: <23.48, , , Quartiles for women: <21.35, , , were stratified by age (data not shown), job activity of moderate intensity was associated with a large decrease in colon cancer risk among men less than 55 years of age (relative risk for 14.5 hours per week vs. none =.29, 95 percent confidence interval = , =.5), whereas among older men, job activity was not associated with reduced risk. The difference between older and younger men was highly significant (p for interaction =.3). Lifetime occupational activity (data not shown) and household activity (table 3) were not related to colon cancer risk. Two measures of total physical activity during the 1-year reference period were considered (table 4). The first was hours per week of moderate or high intensity recreational or job activity. Although no associations were statistically significant, the pattern for men and women combined was suggestive of a reduced risk of colon cancer associated with 5 or more hours of moderate or high intensity physical activity per week. There were no clear associations when we evaluated total energy expended in METs, which we based only on recreational and occupational activities because of the large number of hours of housework reported by women. None of the other factors in table 1, other correlates of physical activity including education and smoking (29), or other predictors of colon cancer in this study including vitamin supplement use, fiber intake, and calcium intake (25) confounded the associations shown. After adjustment for each of these factors separately and combined, the relative risks changed by less than 1 percent. Body mass index and frequency of constipation are measures of potential mechanisms of any physical activity-colon cancer relation as well as potential confounders. Inclusion of either of these factors in the statistical model had no effect on the associations presented. There were no important differences in the association of job or recreational physical activity to colon cancer by anatomic subsite (data not shown). The results were similar for cancer of the proximal colon (cecum to and including splenic flexure) and distal colon (descending and sigmoid colon). Similarly, there were no important effect modifiers of the relation between physical activity and colon cancer, except for the effect of age on the relation between occupational activity and colon cancer among men, noted above. DISCUSSION Our findings of a modest reduction in colon cancer risk associated with recreational physical activity could be due to self-selection of controls with healthier lifestyles. However, adjustment for health behaviors other than recreational activity, e.g., use of vitamin supplements, fiber intake, smoking, and alcohol use, did not modify our results. It is unlikely that our results are spuriously due to our failure to interview all cases, unless those cases who died (our primary reason for loss of cases) were more physically active than those who survived. Another concern is that differential and nondifferential measurement error in our assessment of physical activity may have biased our results. The Minnesota Leisure Time Physical Activity Questionnaire (3), the instrument on which our questionnaire is based, Downloaded from at Pennsylvania State University on February 27, 214

5 46 White et al. TABLE 2. Relative risk () of colon cancer associated with recreational physical activity,* western Washington State, Actlvtty Total (episodes/week) 1-<2 2-<4 4 Moderate-high intensity (episodes/week) 1-<2 2 intensity (episodes/week) 1 Total (hours/week) 1-< <4 4 Moderate-high intensity (hours/week) 1-<2.5 ^2.5 METS/week < (n = 25i) (n = 233) Men %Clt (/J=193) (n = 194) Women Agead)usted Ageand sexadjusted During the 1-year reference period ending 2 yeare before diagnosis for cases and a similar assigned period for controls, t Cl, confidence interval Both Downloaded from at Pennsylvania State University on February 27, 214 has significant measurement error. An additional source of error is that our reference period may not be the appropriate etiologic time period. Differential measurement error could have led to spurious results if cases underreported their activity and/or controls overreported their true activity. However, to our knowledge, the association between physical activity and colon cancer is not well known by the general public. Our results are in substantial agreement with past studies of recreational physical activity and colon cancer. We found recreational activity to be significantly associated with reduced risk in men and in men and women combined, with the magnitude of the reduction in risk similar to the findings of other studies. Four previous studies, all prospective, have reported moderate or high levels of leisure physical activity to be significantly associated with reduced risk of colon or colorectal cancer among men, with relative risk estimates between.5 and.7 (6, 7, 11, 19). Of two cohort studies among women, one in Sweden reported a rel-

6 Physical Activity and Colon Cancer 47 TABLE 3. Relative risk () of colon cancer by occupational and household physical activity,* western WashingtonState, ActMty Occupational Total nonsadentary wortcf: (hours/week) Low est Moderate intensity work (hours/week) None METS/weekll Low est Household Active housework^ (hours/week) Low est n = 251) (n = 233) C*>) Men %Clt (no 193) (n=194) Women * During the 1-year reference period ending 2 years before diagnosis for cases and a similar assigned period for controls. t Cl, confidence interval. t For men (hours/week): tow,.; medium, 1.-<2.; high, 2-<35.; highest, 35.. For women (houra/week): low, ; medium, 3.5; high For men (hours/week): medium, 4.5; high, For women (hours/week): medium, <4.5; high, 4.5. I For men (METS/week): low, <35.; medium, 35.-<7.1; high, 7.1-c147.6; highest, For women (METS/week): tow, ; medium, <47.8; high, H For men (hours/week): low, 6; medium, >6-; high, >-2; highest, >2. For women (houra/week): low, 2; medium, >2-3; high, >3-48; highest, >48. ative risk of.5 for colon cancer associated with regular exercise (7), and one in a US retirement community reported little association with colorectal cancer (6). A case-control study in Utah in which few women reported any occupational activity reported an odds ratio of.5 for the highest tertile of combined occupational and leisure activity (8, 13). In the present study, the association with recreational activity was weaker and nonsignificant among women, perhaps due to the high background level of low intensity household activity among women. In men, physically active occupations have consistently been associated with relative risks for colon or colorectal cancer of between.5 and.8 compared with more sedentary occupations (1-3, 7, 9-11, 13, 14, 17, 18, 2-23). In women, occupational physical activity has been associated with significantly reduced risk in some studies (7, 9, 22) but not in others (13, 21). We observed little reduction of risk associated with measures of occupational activity, except among men younger than 55. Among the younger men, the association was strong (relative risk for > 14.5 hours per week of moderate intensity work vs. none =.29), which may be due to higher intensity of occupational activity among younger men compared with women and older men. This is the first study to evaluate risk estimates for household activity. We did not find household activity to be associated with risk of colon cancer. The amount of physically active household work is difficult to recall, and many women reported 4 hours or more of household activity per week. Thus, the lack of association between colon cancer and household activity may be due in part to measurement error, and better methods are needed to measure this major source of activity among women. Downloaded from at Pennsylvania State University on February 27, 214

7 48 White et al. TABLE 4. Relative risk () of colon cancer by total occupational and leisure physical activity,* western Washington State, Activity Moderate-high intensity activity (hours/week) < <5 5 Total METS/week^ Low est (n=251) (n = 233) Men S5%Clt (n = 193) (no 194) Women Ageand saxadjusted Both * During the 1-year reference period ending 2 years before diagnosis for cases and a similar assigned period for controls, t Cl, confidence interval. i For men: low, <44. METS/week; medium, 44. e83.8; high, ; highest, For women: low, 2.3 METS/week; medium, 12.3-<36.8; high, 36.8-<84.1; highest Several previous studies have examined the association between colon or colorectal cancer and total physical activity measured using a wide range of methods (8, 11, 12, 15-17). These studies have generally reported reduced risk of colon cancer associated with total physical activity. Our measures of total activity were based on summing hours or METs for each type of recreational activity and for reported occupational activity on each job over a 1-year reference period. Although we observed no significant relation, there was a suggestion that 5 hours or more of moderate or high intensity activity per week (job and leisure combined) was associated with a small reduction in risk. Our approach to quantifying total energy expenditure (beyond sedentary activities) in terms of METs may not have been successful in predicting colon cancer because it may not be a measure of the appropriate characteristic of activity. Periodic vigorous activity (as measured, for example, by frequency of moderate or intense activity) may be more important in the etiology of colon cancer than total energy expenditure. Several prior studies suggest that the protective effect of physical activity may be greater for cancer of the distal colon (1, 9, 12, 21). We found no important differences by colon subsite. The association of physical activity with reduced risk of colon cancer is also supported by animal studies. Experiments have shown that chemically induced colon cancer is reduced by physical activity including both forced strenuous activity, which led to weight loss (31), and moderate voluntary activity, which did not lead to weight loss (32). The primary hypothesized mechanism of the physical activity-colon cancer relation is that exercise reduces intestinal transit time, which would limit the time of contact between the bowel mucosa and the bowel contents (which may contain mutagens or cytotoxic agents such as secondary bile acids). The relation between physical activity and transit time is supported by experimental studies in humans, one of which showed a 3 percent reduction in transit time with moderate exercise (33). However, there is only limited evidence that transit time is associated with colon cancer. Animal studies (34, 35) and ecologic studies (36, 37) on this topic are inconsistent. The present study and some earlier epidemiologic studies (38,39) found constipation, an indicator of long transit time, to be a risk factor for colon cancer. However, adjustment for frequency of constipation did not reduce the estimated relative risks for physical activity in this study. This indicates that any effect of physical activity on colon cancer was not even partially explained by the association of constipation with colon cancer risk in this study. Other postulated mechanisms are that the protective effect of physical activity is mediated by a reduction in body iron stores (4-42) or by changes in prostaglandin levels during physical activity (43, 44). In summary, this study provides additional support to the evidence that physical activity reduces the risk of colon cancer. Studies that investigate the mecha- Downloaded from at Pennsylvania State University on February 27, 214

8 Physical Activity and Colon Cancer 49 nism by which physical activity modifies colon cancer risk may further our understanding of this major disease. ACKNOWLEDGMENTS Support for this study was provided by Public Health Service grant R29 CA4479 from the National Cancer Institute. Case ascertainment was supported by contract N1-CN-523 from the National Cancer Institute. The authors thank the senior staff on this project: Janet Reusser Munger, Ann Shattuck, Jude Ballard, and Judy Stilson. REFERENCES 1. Garabrant DH, Peters JM, Mack TM, et al. Job activity and colon cancer risk. Am J Epidemiol 1984;119: Vena JE, Graham S, Zielezny M, et al. Lifetime occupational exercise and colon cancer. Am J Epidemiol 1985;122: Gerhardsson M, Norell SE, Kiviranta H, et al. Sedentary jobs and colon cancer. Am J Epidemiol 1986;123: Schottenfeld D, Fraumeni J. Cancer epidemiology and prevention. Philadelphia, PA: W. B. Saunders, US Bureau of Census. Historical statistics of the US, colonial times to 197. Bicentennial edition part 2. Washington, DC: GPO, Wu AH, Paganini-Hill A, Ross RK, et al. Alcohol, physical activity, and other risk factors for colorectal cancer: a prospective study. Br J Cancer 1987;55: Gerhardsson de Verdier M, Broderus B, Norell SE. Physical activity and colon cancer risk. Int J Epidemiol 1988; 17: Slattery ML, Schumacher MC, Smith KR, et al. Physical activity, diet, and risk of colon cancer in Utah. Am J Epidemiol 1988;128: Fredriksson M, Bengtsson NO, Hardell L, et al. Colon cancer, physical activity, and occupational exposures. Cancer 1989; 63: Peters RK, Garabrant DH, Yu MC, et al. A case-control study of occupational and dietary factors in colorectal cancer in young men by subsite. Cancer Res 1989;49: Severson RK, Nomura AMY, Grove JS, et al. A prospective analysis of physical activity and cancer. Am J Epidemiol 1989; 13: Gerhardsson de Verdier M, Hagman U, Steineck G, et al. Diet, body mass and colorectal cancer: a case-referent study. Int J Cancer 199;46: Slattery ML, Abd-Elghany N, Kerber R, et al. Physical activity and colon cancer: a comparison of various indicators of physical activity to evaluate the association. Epidemiology 199; 1: Kato I, Tominaga S, Dcari A. A case-control study of male colorectal cancer in Aichi Prefecture, Japan: with special reference to occupational activity level, drinking habits and family history. Jpn J Cancer Res 199;81:l Ballard-Barbash R, Schatzkin A, Albanes D, et al. Physical activity and risk of large bowel cancer in the Framingham Study. Cancer Res 199;5: Kune GA, Kune S, Field B, et al. Body weight and physical activity as predictors of colorectal cancer risk. Nutr Cancer 199:13: Whittemore AS, Wy-Williams AH, Lee M, et al. Diet, physical activity, and colorectal cancer among Chinese in North America and China. J Natl Cancer Inst 199;82: Brownson RC, Chang JC, Davis JR, et al. Physical activity on the job and cancer in Missouri. Am J Public Health 1991;81: Lee IM, Paffenbarger RS, Hsieh CC. Physical activity and risk of developing colorectal cancer among college alumni. J Natl Cancer Inst 1991:83; Markowitz S, Morabia A, Garibaldi K, et al. Effect of occupational and recreational activity on the risk of colorectal cancer among males: a case-control study. Int J Epidemiol 1992;21: Arbman G, Axelson O, Fredreiksson M, et al. Do occupational factors influence the risk of colon and rectal cancer in different ways? Cancer 1993;72: Chow WH, Dosemeci M, Zheng W, et al. Physical activity and occupational risk of colon cancer in Shanghai, China. Int J Epidemiol 1993;22: Fraser G, Pearce N. Occupational physical activity and risk of cancer of the colon and rectum in New Zealand males. Cancer Causes Control 1993;4: Choi JW. Exercise and participation in sports among persons 2 years of age and oven United States, Advancedata 1978;19:l Meyer F, White E. Alcohol and nutrients in relation to colon cancer in middle-aged adults. Am J Epidemiol 1993;138; Taylor HL, Jacobs DR Jr, Schucker B, et al. A questionnaire for the assessment of leisure time physical activities. J Chronic Dis 1978;31: Ainsworth BE, Jacobs DR Jr, Leon AS, et al. Compendium of physical activities: classification of energy costs of human physical activities. Med Sci Sports Exerc 1993;25: Kannel WB, Sorlie P. Some health benefits of physical activity: the Framingham Study. Arch Intern Med 1979;139: Dishman RK, Sallis JF, Orenstein DR. The determinants of physical activity and exercise. Public Health Rep 1985;1: Richardson MT, Leon AS, Jacobs DR, et al. Comprehensive evaluation of the Minnesota Leisure Time Physical Activity Questionnaire. J Clin Epidemiol 1994;47: Klurfeld DM, Weber MM, Kritchevsky D. Inhibition of chemically induced mammary and colon tumor promotion by caloric restriction in rats fed increased dietary fat. Cancer Res 1987;47: Reddy BS, Sugie S, Lowenfels A. Effect of voluntary exercise on azoxymethane-induced colon carcinogenesis in male F344 rats. Cancer Res 1988;48: Oettle GJ. Effect of moderate exercise on bowel habit. Gut 1991;32: Baker AR, Ward JM, Simon RM, et al. Intestinal transit time as an important factor in the pathogenesis of dimethyl-hydrazine-induced rat colon cancer. Surg Forum 1978;29: Cleveland JC, Cole JW. Relationship of experimental induced intestinal tumors to laxative ingestion. Cancer 1969;23: Glober GA, Kamiyama S, Nomura A, et al. Bowel transit-time and stool weight in populations with different colon-cancer risks. Lancet 1977;2: Cummings JH, Bingham SA, Heaton KW. Fecal weight, colon cancer risk and dietary intake of fiber. Gastroenterology 1992; 13: Wynder EL, Shigematsu T. Environmental factors of cancer of the colon and rectum. Cancer 1967;2: Kune GA, Kune S, Field B, et al. The role of chronic constipation, diarrhea, and laxative use in the etiology of largebowel cancer data from the Melbourne Colorectal Cancer Study. Dis Colon Rectum 1988;31: Lakka TA, Nyyssc-nen K, Salonen JT. er levels of conditioning leisure time physical activity are associated with reduced levels of stored iron in Finnish men. Am J Epidemiol 1994; 14: Downloaded from at Pennsylvania State University on February 27, 214

9 5 White et al. 41. Stevens RG, Jones DY, Micozzi MS, et al. Body iron stores thelial cell proliferation and xenograft growth. Br J Cancer and the risk of cancer. N Engl J Med 1988;319:lO ;41: Nelson RL, Davis FG, Sutter E, et al. Body iron stores and risk 44. Vapaatalo H, Laustiola K, Seppala E, et al. Exercise, ethanol of colonic neoplasia. J Natl Cancer Inst 1994;86: and arachidonic acid metabolism in healthy men. Biomed 43. Tutton PJM, Barkla DH. Influence of prostaglandins on epi- Biochim Acta 43:S Downloaded from at Pennsylvania State University on February 27, 214