Physical Activity and Colorectal Cancer Risk by Sex, Race/Ethnicity, and Subsite: The Multiethnic Cohort Study

Past studies have provided consistent evidence that physical activity is associated with a lower risk of colorectal cancer (1–3). A meta-analysis of 12 prospective studies, conducted by the World Cancer Research Fund/American Institute for Cancer Research Continuous Update Project (CUP) in 2017, showed a 20% [95% confidence interval (CI), 0.72–0.88] decreased risk of colon cancer comparing the highest and lowest levels of total activity levels when including activities of various intensity levels (1).

However, data is lacking on whether the association is consistent between sexes and across different races/ethnicities and anatomic subsites of tumors. Given that incidence rates of colorectal cancer vary widely by sex and race/ethnicity, and that biological mechanisms may differ by anatomic subsite, the beneficial effects of physical activity may be greater for certain populations and subsites. To address this gap, we analyzed data from the Multiethnic Cohort Study (MEC), which mostly consists of African Americans, Native Hawaiians, Japanese Americans, Latinos, and whites living in Hawaii and California.

The design and characteristics of the MEC have been described in detail elsewhere (4). In brief, the MEC is a prospective cohort study designed to examine the relation of lifestyle and genetic factors with cancer among representative population groups of five races/ethnicities: African American, Native Hawaiian, Japanese American, Latino, and white in Hawaii and Los Angeles (4). Between 1993 and 1996, more than 215,000 men and women ages 45–75 years entered the cohort by completing a comprehensive questionnaire that included sections on physical activity, body weight and height, eating habits, and medication history. The study was approved by the review boards of the University of Hawaii (Honolulu, HI) and University of Southern California (Los Angeles, CA). For this analyses, we excluded participants who were not in one of the targeted racial/ethnic groups (n = 13,987), who had previous colorectal cancer reported on the baseline questionnaire (n = 2,251) or identified from the tumor registries (n = 301), or who reported implausible diet based on macronutrients (n = 8,116; ref. 5). We also excluded participants with missing information on physical activity variables (n = 18,483). These were more likely to be women than men and African American or Latino, compared with the other three racial/ethnic groups. As a result, the analyses included 172,502 participants.

The physical activity questions were designed to reflect a participant's behavior over an average 24-hour period during the previous year and inquired about time spent sleeping, in various sitting activities, and in strenuous sports, vigorous work, and moderate activities. Sleep duration, including naps, was asked using six categories: ≤5, 6, 7, 8, 9, and ≥10 hours per day. Five categories of sitting activities were asked, including sitting in car or bus, sitting at work, watching TV, sitting at meals, and other sitting activities (such as reading, playing cards, or sewing), each in seven categories: never, <1, 1–2, 3–4, 5–6, 7–10, and ≥11 hours per day. Total daily sitting hours per day were calculated as the sum of the midpoints of the sitting categories, using 0, 0.5, 1.5, 3.5, 5.5, 8.5, and 11, respectively. Strenuous sports (such as jogging, bicycling on hills, tennis, racquetball, swimming laps, or aerobics), vigorous work (such as moving heavy furniture, loading or unloading trucks, shoveling, weight lifting, or equivalent manual labor), and moderate activity (such as housework, brisk walking, golfing, bowling, bicycling on level ground, or gardening) were asked in eight categories: never, 0.5–1, 2–3, 4–6, 7–10, 11–20, 21–30, and ≥31 hours per week. Hours in each activity level per day was calculated using the midpoint of the categories: 0, 0.75/7, 2.5/7, 5/7, 8.5/7, 15/7, 25.5/7, and 31/7. Light physical activity was calculated by subtracting the total time spent in all activities (sleeping, sitting, or in moderate and vigorous activity) from 24 hours.

The metabolic equivalents (MET) for a 24-hour day were created using the following formula: (hours in sleep × 0.91 + hours in sitting × 1.0 + hours in light activity × 2.4 + hours in moderate activities × 4.0 + hours in vigorous activity × 7.2)/24. MET-hours of moderate and vigorous activity per day were calculated as the sum of hours in moderate activities × 4.0 and hours in vigorous activity × 7.2. The physical activity questionnaire used in the MEC has been validated against the objective measure of total energy expenditure based on doubly labeled water in 230 adults (6). The correlation between the objective and self-reported values was reasonable with r = 0.29 for METs.

The same questions on physical activity were asked in a 10-year follow-up survey (2003–2007) except for moderate activity, which separated recreational (such as brisk walking, golfing, bicycling on level ground, gardening, dancing, or softball) and work activities (such as housework, yard work, restaurant work, sales work, or equivalent moderate physical activity). Total recreational activity was calculated as the sum of hours in strenuous sports and moderate recreational activities. Total work-related activity was calculated as the sum of hours in vigorous work and moderate work.

Incident cases of colorectal cancer were identified by linking the cohort to the tumor registries in Hawaii and California through December 31, 2013. Cases in this study were limited to participants diagnosed with invasive adenocarcinoma of the large bowel with International Classification of Disease (ICD)-O2 codes of C18.0–C18.9, C19.9, and C20.9. For anatomic subsite–specific analyses, cases were categorized using ICD-O2 codes: C18.0–C18.5 for right colon, C18.6-C18.7 for left colon, and C19.9 and C20.9 for rectum, excluding multisite cases. Deaths were identified by linkage to death files in Hawaii and California and the National Death Index through December 31, 2013.

Cox proportional hazards models of colorectal cancer with age as the time metric were used to calculate HRs and 95% CIs for men and women separately. Colorectal cancer cases other than adenocarcinoma were censored at the date of diagnosis. Physical activity variables were parameterized as quintiles or quartiles, based on the overall distributions of the variables in men and women combined. Tests of proportional hazards assumption were based on the Schoenfeld residual method and found to be met (7). Trend tests were conducted by modeling continuous variable-assigned sex/ethnic-specific median values within the quantile categories. All models were adjusted for race/ethnicity as a strata variable and age at cohort entry as a covariate. Multivariate models were further adjusted for family history of colorectal cancer (yes, no), history of intestinal polyps (yes, no), body mass index (BMI; <25, 25–<30, ≥30 kg/m²), pack-years of cigarette smoking, multivitamin use (yes, no), NSAID use (yes, no), daily intake of total energy (log transformed kcal), alcohol (g), red meat (g/1,000 kcal), dietary fiber (g/1,000 kcal), calcium (mg), folate (dietary folate equivalents), and vitamin D (IUs), and for women, menopausal hormone therapy (MHT) use (never, former, or current use of estrogen). These covariates were chosen because they are established risk factors or were found to be associated with risk of colorectal cancer in the MEC. In the multivariate models, participants with missing data on covariates (n = 11,145) were excluded, resulting in 161,357 participants. For subgroup analyses, we only presented the variable of MET-hours of moderate/vigorous activity because this variable showed the strongest association. The models were run separately by race/ethnicity and anatomic subsite of tumors (right colon, left colon, and rectum). We ran the models for combined association of physical activity and BMI with a group of the lowest level of moderate/vigorous activity and the highest BMI category (≥30 kg/m²) as the reference. We also ran the models for combined association of physical activity and sitting time with a group of the lowest level of physical activity and the longest sitting hour (≥10 hours/day) as a reference. Among postmenopausal women, the models were run for MHT never and ever users, separately. Heterogeneity was tested by Wald statistics for cross-product terms of trend variables and subgroup membership. Heterogeneity by anatomic subsite was tested by Wald statistics using competing risk methodology (7).

For analyses of recreational and work-related physical activity, we repeated the models in the 76,010 men and women who participated in 10-year follow-up survey and met the inclusion criteria at the time of follow-up. Physical activity variables and covariates were from the 10-year follow-up survey and outcome was incident colorectal cancer occurring after the 10-year follow-up. For all analyses, we used SAS version 9.4 (SAS Institute). All P values were two-sided.

Table 1 presents baseline characteristics by physical activity level in men and women. Men and women who had higher MET-hours of moderate or vigorous activity were more likely to be younger, Native Hawaiian or white, to have family history of colorectal cancer, to have higher education, to use multivitamin supplements, to have higher intakes of energy, calcium, folate, vitamin D, and alcohol, while they were less likely to be obese and NSAID users. Women with higher physical activity tended to be current users of MHT at baseline.

A total of 4,430 incident colorectal cancer cases (2,341 men and 2,089 women) were identified in the study population during an average follow-up of 16.8 years. Table 2 shows the association of sleep, sitting hours, and various types of physical activity with colorectal cancer risk in men and women. Sleep and total sitting hours were not related to colorectal cancer risk in either men or women. Sitting time watching TV was associated with an increased risk especially in women; however, this association was no longer significant after adjustment for covariates. In men, moderate activity (HR, 0.83; 95% CI, 0.72–0.95 for the highest vs. lowest quintile, P_trend = 0.01) was associated with a decreased risk in multivariate-adjusted models. MET-hours of moderate and vigorous activity showed a stronger association (HR, 0.76; 95% CI, 0.66–0.87 for the highest vs. lowest quintile, P_trend < 0.001), compared with each activity type, total hours spent in three types of physical activity, and total METs (HR, 0.88; 95% CI, 0.77–1.00; P_trend = 0.05). When considering Bonferroni correction for multiple comparisons (P < 0.005), the associations for moderate and vigorous activity in men both as hours and MET-hours were still statistically significant. In women, however, no association was found. There was a suggestion for heterogeneity by sex for strenuous sports (P = 0.07) and MET-hours of moderate/vigorous activity (P = 0.07), but not for the other types (P ≥ 0.13). In the sensitivity analysis excluding the cases diagnosed within the first 2 years of follow-up, the findings remained similar (in men: HR, 0.76; 95% CI, 0.66–0.87 for the highest vs. lowest quintile of MET-hours of moderate/vigorous activity, P_trend < 0.001).

In racial/ethnic-specific analyses (Table 3), all HRs in the upper quintiles were below 1 across the five groups in men (P_{heterogeneity} = 0.36), with an inverse trend in African Americans (P_trend = 0.003) and Japanese Americans (P_trend = 0.02). Among women, only whites showed a decrease in risk (HR, 0.64; 95% CI, 0.45–0.90 for the highest vs. lowest quintile; P_trend = 0.08), but no indication was seen for racial/ethnic differences (P_{heterogeneity} = 0.36).

The inverse association with moderate/vigorous activity was more apparent for right colon and rectal cancer than for left colon in men (P_{heterogeneity} = 0.55; Table 4). In women, decreases in risk for right colon cancer were shown (P_{heterogeneity} = 0.36).

In joint analyses of physical activity and BMI (Table 5), no difference in the association between moderate/vigorous activity on colorectal cancer risk was found across the BMI groups in either men or women (P_{heterogeneity} = 0.62 and 0.19, respectively), although among women, an inverse trend with physical activity was suggestive in the BMI ≥ 30 group. The BMI < 30 groups were at lower risk regardless of physical activity levels in both men and women, compared with the reference group (BMI ≥ 30 and least physically active). There was further decrease in risk with physical activity in the BMI < 25 group among men but not among women.

In combined analyses of physical activity and sitting time (Table 6), the inverse association with moderate/vigorous activity was seen in men with total sitting time longer than 6 hours per day, but not in those with 6 or less hours (P_{heterogeneity} = 0.06). When examining sitting time watching TV, the inverse association was significant in men watching TV for 3 hours or longer, but not in those with less than 3 hours (P_{heterogeneity} = 0.52). Among women, no association was found compared with those with the lowest physical activity level and the longest sitting time. In stratified analyses by MHT use at baseline among postmenopausal women (Table 7), MHT ever users who were at lower risk of colorectal cancer than never users showed further decrease in risk with physical activity, while no association was found in never users (P_{heterogeneity} = 0.03).

A total of 34,089 men (mean age: 69.0 ± 8.3 years) and 41,921 women (mean age: 68.6 ± 8.4 years) were included in the analysis conducted among participants who returned the 10-year follow-up questionnaire, (mean follow-up: 8.0 ± 2.1 years). In men, both recreational and work-related activities in the 10-year follow-up survey were associated with a lower risk of subsequent colorectal cancer (Table 8). The decrease in risk was seen for all four upper quintiles without further decrease in the highest quintile. No significant association was found in women. Similarly, moderate/vigorous activity measured in MET-hours and calculated from the 10-year follow-up was inversely associated with colorectal cancer risk in men but not in women.

In this large multiethnic population, physical activity was associated with a lower incidence of colorectal cancer; this association was more apparent in men than in women. Sleep and sitting durations were not associated with colorectal cancer risk either in men or women. The association did not vary by race/ethnicity, anatomic subsite, and BMI group. The inverse trend was seen among men with longer sitting time (>6 hours per day) but not among those with short sitting time, while no association was found across all sitting time categories among women.

In the CUP meta-analysis based on the literature published up to April 2015, total physical activity in a comparison with the highest versus lowest levels was consistently associated with a 20% lower risk of colon cancer (95% CI, 0.72–0.88) but was not associated with rectal cancer [relative risk (RR), 1.04; 95% CI, 0.92–1.18]. Similarly, recreational physical activity was related to a 16% lower risk of colon cancer (95% CI, 0.78–0.91) but not to rectal cancer (RR, 0.95; 95% CI, 0.85–1.07). Prospective studies published since the CUP have generally supported the conclusions of the CUP, with some differences. A report from the Women's Health Initiative Observational Study (WHI-OS) found an inverse association between leisure time physical activity and colorectal cancer risk, particularly for rectal cancer in postmenopausal women (8), which was not consistent with the CUP results. The Singapore Chinese Healthy Study found that strenuous–vigorous physical activity (sports and work) was associated with a lower risk of colorectal cancer in men and women combined, which was similar for colon and rectum tumors (9). In the AARP Diet and Health Study, physical activity (mostly recreational) was related to a lower risk of colon cancer among nondiabetic, but not diabetic, participants (10). In this study, we found a lower risk of colon cancer with physical activity in both groups of men with and without a history of diabetes at baseline (P_{heterogeneity} = 0.93 for MET-hours of moderate/vigorous activity). A U.K. cohort study also reported that total physical activity was associated with a lower risk of colon cancer (11).

Although recent prospective studies and a meta-analysis (12) have supported similar strengths of association between men and women, the inverse association between physical activity and colorectal cancer risk in the MEC appears to be stronger in men than in women. The difference in benefits from physical activity between the sexes may reflect hormonal differences (13). However, when we examined the associations among MHT ever users and never users in postmenopausal women, the inverse association was only found among MHT ever users. As a group, MHT users were at lower risk of colorectal cancer than MHT never users or men. In addition to the possibility of biologically distinct responses to exercise between men and women (13), the potential for gender differences of physical activity levels and types even within the same category (e.g., moderate activity) should be also considered as a potential explanation for our findings (14, 15).

In another recent meta-analysis of 17 cohort and 21 case–control studies (16), occupational, recreational, and transportation-related physical activity were related to a 26% (95% CI, 0.67–0.82), 20% (95% CI, 0.71–0.89), and 34% (95% CI, 0.45–0.98) decrease in risk of colon cancer, and a 12% (95% CI, 0.79–0.98), 13% (95% CI, 0.75–1.01), and 12% (95% CI, 0.70–1.12) decrease in risk of rectal cancer, respectively, comparing the highest versus lowest levels. In the MEC, we found a similar risk reduction for colon (19%) and rectal cancer (21%) in men. We found similar HRs for right and left colon tumors, which was also reported by two meta-analyses (17, 18). Although we were not able to examine domain-specific physical activity at baseline, a 10-year follow-up survey in the MEC asked separate questions for recreational and work-related activities. We found a decreased colorectal cancer risk with both recreational and work-related activity among men. This is consistent with the World Health Organization recommendation that moderate or vigorous activity can be accumulated in any domain for health benefits (2, 19).

In this study, the inverse association of physical activity and colorectal cancer risk appeared to be stronger in African American men, compared with the other racial/ethnic groups, although there was no indication for heterogeneity overall across races. Racial/ethnic differences have been reported regarding types of exercise and sports participation, which might not be captured by the MEC questionnaire but be related to colorectal cancer risk (20). However, in addition to the relatively small number of African Americans in the MEC, they tended to be less physically active than the other racial/ethnic groups and thus the group with the highest level of physical activity was small. Therefore, caution needs to be exercised when interpreting these findings.

A meta-analysis of 15 case–control and cohort studies found a stronger relative risk for physical activity and colorectal cancer risk in the higher BMI group (21). Although in our study women with higher BMI showed a suggestive inverse association, we found no evidence for an interaction between BMI and physical activity in relation to colorectal cancer risk. One of the potential mechanisms by which physical activity may lower colorectal cancer risk is through a reduction in insulin resistance and inflammation, which have been related to colorectal tumor development (1). However, it is not clear whether physical activity without weight loss has a significant impact on these pathways (1).

An interaction between physical activity and sitting time has previously been reported in relation to colorectal cancer risk. In the Singapore Chinese Health Study, inverse associations between physical activity and colorectal cancer risk were the clearest among those with longer sitting time (9). No such interaction was found in the WHI-OS (8). In the MEC, we found a greater decrease in risk among men with longer sitting hours. Thus, our findings suggest that moderate/vigorous physical activity may be particularly beneficial for colorectal cancer prevention among people with longer sitting times.

Strengths of our study include the prospective design, the large study population with various racial/ethnic backgrounds, and the comprehensive information collected. However, several limitations need to be taken into account when interpreting our findings. In the validation study of the physical activity questionnaire used in the MEC, the correlation with the doubly labeled water standard was modest (6). Although we considered a wide range of confounding factors, there might still be uncontrolled factors related to colorectal cancer risk. For example, information on colorectal cancer screening was not available at baseline. However, using data from a 5-year follow-up survey that were available for 80% of the participants, we found that further adjustment for colorectal cancer screening did not change the associations with subsequent colorectal cancer. For this analysis, we used physical activity measured once either at baseline or in the 10-year follow-up survey. When analyzing data with updated physical activity information from the 10-year follow-up survey, we found a similar inverse association of physical activity with subsequent colorectal cancer. In addition, we were only able to distinguish recreational activity from work-related activity on the basis of the 10-year follow-up, but not the baseline questionnaire. Also, we were not able to distinguish type of activity within an intensity level.

In summary, our findings confirm the inverse association between physical activity and colorectal cancer; this association appears to be stronger in men, especially those with longer sitting time, and suggest possible differences in the strength of the association by race/ethnicity and anatomic subsite of the tumor.

No potential conflicts of interest were disclosed.

Conception and design: S.-Y. Park, L.R. Wilkens, L.L. Marchand

Development of methodology: S.-Y. Park, L.L. Marchand

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L.R. Wilkens, C.A. Haiman, L.L. Marchand

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.-Y. Park, L.R. Wilkens, L.L. Marchand

Writing, review, and/or revision of the manuscript: S.-Y. Park, L.R. Wilkens, L.L. Marchand

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L.L. Marchand

Study supervision: L.L. Marchand

This work was supported by the NCI at the NIH (grant no. U01CA164973).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.