Impact of Moderate–Vigorous Physical Activity Trajectories on Colon Cancer Risk over the Adult Life Course

This article is featured in Highlights of This Issue, p. 1

Colorectal cancer is a major public health and clinical concern, as it is the third leading cause of cancer in men and women in the United States (1). In 2021, there are expected to be over 100,000 individuals diagnosed with colon cancer and over 52,000 deaths (1). Development of colon cancer is multifaceted, influenced by genetic, lifestyle, and environmental factors (2). However, the strongest known risk factors for developing colon cancer are related to lifestyle, including insufficient physical activity levels, sedentary lifestyles, obesity, unhealthy diet, high alcohol consumption, and smoking (3–7). The 2018 United States Physical Activity Guidelines found strong evidence of an association that engaging in moderate-vigorous physical activity (MVPA) at recommended amounts (i.e., 150–300 minute/week) reduces colon cancer risk (8–10). In a pooled analysis of 1.4 million adults, those who engaged in high levels of leisure-time physical activity had a 16% reduction in colon cancer risk (11). Numerous types and intensities of physical activity are linked with a reduced risk of colon cancer, however the dosage and the timing of physical activity throughout the adult life course associated with low risk are not well established (12–18).

Major health organizations and government bodies have highlighted the need to better understand the relationship between physical activity over the life course and colon cancer risk (8). The current literature is inconclusive on whether physical activity levels earlier, throughout, or later in the life course are most protective (14, 19–24). Many of these studies measure levels of MVPA at only one time point (i.e., only adolescence or midlife) or measure MVPA upon cohort entry and follow prospectively for a fixed set of time for colon cancer outcomes (14, 19, 23). These approaches are not able to address when in the life course physical activity may be most protective. Few studies have examined the impact of MVPA across various points in the life course and how increasing or decreasing levels of MVPA over the life course impact colon cancer risk (20, 21). There is a need to understand the stage in the life course where MVPA is critical in minimizing colon cancer risk and developing interventions and disseminating cancer prevention messages.

Therefore, the purpose of our study was to evaluate whether maintenance of and changes in MVPA between late adolescence and later adulthood, measured as life course trajectories, are associated with colon cancer risk. We hypothesize that sustained maintenance of sufficient MVPA (i.e., at and above Physical Activity Guidelines) through the adult life course will be the most protective factor against colon cancer risk.

Study participants are from the NIH American Association for Retired Persons (AARP) Diet and Health Study (25). In 1995 to 1996, 566,398 AARP members (ages 50–71 years) from six states (CA, FL, LA, NJ, NC, PA) and two metropolitan areas (Atlanta and Detroit) completed a baseline questionnaire assessing medical conditions and demographics. In 1996 to 1997, those who were free of colon, breast, or prostate cancer at the time of the baseline questionnaire were sent a risk factor questionnaire (RFQ) that included more detailed questions about physical activity and other lifestyle factors (25). The NIH-AARP study was approved by the Special Studies Institutional Review Board of the US NCI. All participants provided written informed consent by completing and returning the baseline questionnaire and the study was conducted in accordance with ethical guidelines and the United States Common Rule.

Our study population includes adult men and women who completed both the baseline questionnaire and RFQ (response rate 67%), resulting in 334,905 men and women. The completion of the baseline and RFQ are collectively grouped as “baseline.” For the present analysis, individuals were then excluded if the RFQ was completed by a proxy (N = 10,383), previous history of cancer (N = 18,330), missing physical activity levels (N = 8,901), or poor health (N = 4,093). We selected the study sample with the goal of minimizing confounding and reverse causation due to health status and preserving the accuracy of the responses for physical activity. These exclusions resulted in an analytical sample size at baseline of 293,198 individuals. Follow-up time was calculated from completion of the RFQ until (any) colon cancer diagnosis, death, or the end of study follow-up (December 31, 2011).

Our primary exposure was MVPA throughout the adult life course. On the RFQ, the participants self-reported their physical activity levels during adulthood. Participants were provided examples of 16 common moderate and vigorous physical activities and asked “How often did you participate in moderate-vigorous physical activities” at ages:15 to 18, 19 to 29, 30 to 35, and in the past 10 years (i.e., midlife or roughly 41–60 years old).¹³ Reported MVPA duration in each age category included: 0 (rarely or none), 0.5 (<1 hour/week), 2.0 (1–3 hours/week), 5.5 (4–7 hours/week), and 7.0 (>7 hours/week). We used these reports of physical activity levels in each age period to derive trajectories of physical activity over the adult life course, as described below.

All observed covariates were self-reported at baseline using the baseline questionnaire and the RFQ. Covariate selection was guided by our assessment of previous literature. We included age, body mass index (BMI) at age 18 years, sex, race, ethnicity, education, smoking status, red meat intake, health history, hormone therapy use, diabetes, and colon cancer screening history as potential covariates. Height and weight were reported at age 18, and BMI was recorded in kg/m². We selected BMI at age 18 as a covariate because it was the body size value that roughly corresponded to our earliest age-period for the physical activity assessment, or the beginning of our exposure assessment period. We elected not to select BMI at other life-periods because physical activity participation is associated with prevention of weight gain and development of obesity (26). Education was categorized as: (i) less than 8 years, (ii) 8 to 11 years, (iii) 12 years or completed high school, (iv) post-high school or some college, (v) college and postgraduate, or (vi) unknown. Smoking status was grouped as follows: (i) never smoker, (ii) former smoker, (iii) current smoker, and (iv) unknown. The self-reported dietary factors were obtained through a 124-item food frequency questionnaire that was previously validated in the NIH-AARP cohort (25, 27). We included the dietary factors of alcohol intake, red meat intake, and processed meat intake (quantified as grams per day). Alcohol intake included consumption from all dietary sources (including cooking alcohol); red meat intake included total red meat intake; and processed meat intake accounted for the total consumption of processed meat, ham, bacon, sausage, hot dogs, and cold cuts. Hormone therapy use was categorized as yes/no, to assess whether the participant was currently taking replacement hormones. We selected hormone therapy use as a potential covariate as it has previously shown to be associated with colon cancer risk, as well as modify the effect of physical activity on colon cancer in women (28, 29). Diabetes was categorized as yes/no, as indicated by self-reported diabetes diagnosis by a medical professional. Colon cancer screening history was categorized as yes/no, assessing whether the participant had undergone any procedure to examine the colon or rectum in the past 3 years.

Information on the date of cancer diagnosis was obtained from SEER cancer registries. Colon cancer (excluding the rectum) was classified using the International Classification of Diseases of Oncology (tenth edition) morphology codes C18.0, C18.2–C18.9 (3). A comparison of cancer registry case ascertainment with SEER estimates and self-reporting determined that more than 90% of incidence cancers across state registries were identified (30).

We used latent class trajectory models to identify trajectories of physical activity at four time points across adulthood (SAS Proc Traj; refs. 31, 32). This method has been previously applied to physical activity and outcomes such as mortality and liver cancer in the NIH-AARP cohort (33, 34). The number of trajectories selected was an iterative process that balanced the number of trajectories with the number of colon cancer outcomes as well as Bayesian information criteria (BIC) related to the fit of the trajectories modeled. BIC is based on the likelihood function, where the lowest value is the best. Linear, quadratic, and cubic trajectories were considered. In making the final determination, we also examined the probability of correct classification and maintained >5% of the population in each group to have reasonable precision in our analysis (31, 32). Furthermore, the probability of assignment in this study was high, exceeding 0.9 probability for all groups, indicating an overall good match between participants’ MVPA values and assigned MVPA long-term patterns.

Cox proportional hazards models were used to estimate HR and 95% confidence intervals (CI) using SAS Version 9.4 (SAS Proc Phreg), where time was calculated from age at RFQ to age at cancer diagnosis or the end of follow-up, whichever came first. We modeled the interaction terms of the trajectory groups to assess the proportional hazard assumption. No deviations in proportional hazard assumptions were observed (all P-values >0.1). Adults who reported consistently low levels of MVPA over their adult life course were set as the reference group for all analyses. We tested a series of models that included covariates based on a previous physical activity colon cancer analysis in this cohort. Model 1 was adjusted for age and gender. Model 2 was further adjusted for education, smoking, alcohol, red meat, processed meat intake, and hormone therapy use. Model 3, the final model, was further adjusted for BMI at age 18. We used a missing indicator method to account for data on missing covariates.

We assessed effect measure modification by comparing stratified analyses by age (<65 and ≥ 65 years at baseline) and sex (male and female) by life course MVPA trajectory on colon cancer risk. In addition, to minimize the risk of reverse causation and potential latent disease, we performed a sensitivity analysis, excluding individuals who developed cancer within the first 2 years of follow-up. All statistical tests were two-sided, with P‐values <0.05 considered statistically significant.

Informed consent was obtained from the participants through the return of the original baseline questionnaire and the subsequent risk factor questionnaire.

Data were maintained by the NCI, Division of Cancer Epidemiology and Genetics, and are available upon submission of a proposal to be approved by the NIH-AARP Steering Committee. For more information, see https://www.nihaarpstars.com/.

We observed seven unique MVPA trajectories based on the long-term (e.g., 15–18 years old through midlife) patterns of MVPA, which we grouped into three basic patterns: maintainers, increasers, and decreasers (Fig. 1). Maintainers were those who reported consistent activity patterns at three activity volumes: adults who maintained consistently low (<1 hour/week) levels of MVPA over time—maintain low (n = 55,605), adults who maintained consistently high (∼7 hours/week) of MVPA over time—maintain high (n = 91,417), and adults who maintained moderate (∼2.5–6 hours/week) levels of MVPA over time meeting Physical Activity Guidelines—maintain moderate (n = 17,198). Increasers were those who reported increasing levels of MVPA over time, including adults who increased MVPA between adolescence and adulthood, early increasers (n = 28,630), and adults who increased MVPA between adulthood and midlife—late increasers (n = 16,172). Decreasers reported activity declines over time, including adults who decreased MVPA between adolescence and adulthood (early decreasers; n = 47,487), and adults who decreased MVPA between adulthood and midlife (late decreasers; n = 36,689).

Comparing demographic characteristics across these seven trajectories, a greater percentage of men consistently maintained physical activity levels throughout the adult life course that met physical activity guidelines than women, and a greater percentage of men reported decreasing activity levels over the adult life course (Table 1). Women reported an increase in physical activity levels over the adult life course, with low activity in the teenage years followed by higher physical activity levels through midlife.

From the baseline questionnaire in 1995 to 1996 to 2011, there were 5,072 incident cases of colon cancer (NIH-AARP), with an average follow-up of 13.1 years. In the adjusted analyses shown in Table 2, compared with those with consistently low MVPA over the adult life course (referent group), adults who maintained high MVPA over time had a 15% lower risk of colon cancer (HR = 0.85; 95% CI, 0.78–0.93), and those who maintained moderate levels of MVPA had a 13% lower risk of colon cancer (HR, 0.87; 95% CI, 0.76–1.00), after adjusting for covariates (Model 3). Compared with those who maintained low MVPA, the increaser groups had a slightly lower risk of colon cancer, though not significant. Specifically, those who increased MVPA earlier in life had a 10% reduced risk (0.90; 95% CI, 0.80–1.02) and those who increased MVPA later in life had an 8% reduced risk (0.92; 95% CI, 0.80–1.06). Compared with maintained low activity, those who decreased MVPA over time showed higher risks of colon cancer, with a 12% elevated cancer risk for those who decreased MVPA levels early in life to below the physical activity guidelines (1.12; 95% CI, 1.02–1.23).

The associations between the life course MVPA trajectory and colon cancer were consistent after excluding the first 2 years of follow-up (Table 2). However, we found evidence of effect modification of the association between life course MVPA trajectory and colon cancer for both age (P_interaction < 0.001) and sex (P_interaction < 0.001; Table 3 and 4, respectively).

In multivariate adjusted models, we found that, compared with those who maintained low MVPA levels over the adult life course, those who maintained moderate to high levels of MVPA over the life course had a reduced risk of colon cancer. We also found that those who were active early in adulthood but not in midlife did not have a lower risk for colon cancer. These findings offer evidence that maintaining some MVPA throughout adulthood is associated with a lower risk for colon cancer. Future research should continue to explore how changes in MVPA participation throughout adulthood can impact cancer risk.

Our results were consistent with previous studies and are broadly consistent with public health recommendations (3, 5, 7, 11, 12, 18–21, 35). In our study, those who maintained moderate levels of MVPA (2.5–6 hours/week) and those who maintained high levels of MVPA (≥7 hours/week) through midlife had a 13% lower risk of colon cancer. Our findings align with those of a recent meta-analysis by Hidayat and colleagues (2020), which explored the impact of physical activity at young ages and lifetime physical activity across various populations on the risk of cancer (22). Hidayat and colleagues (2020) concluded that high levels of physical activity throughout the lifetime were protective against colon cancer risk, with a 25% reduced risk (22). In studies measuring physical activity across the adult life course, maintaining high levels of physical activity had an inverse association with colon cancer risk (20, 21). Most of the previous studies have only assessed physical activity changes over relatively short periods in mid-life (e.g., baseline cohort entry to 5–10 year change). Our study extends these findings by using trajectory models to examine changes in MVPA from late adolescence into middle-older age (approximately 40-year span), providing a realistic behavioral pattern of MVPA that reflects MVPA shifts through adulthood. Our study adds to the collective evidence that staining moderate and high amounts of physical activity supports the recommendations of the U.S. Physical Activity Guidelines and other national organizations for general health benefits and colon cancer risk reduction (8–10, 36).

When examining the presence of effect modification, we found stronger associations between MVPA and colon cancer risk among those ages <65 years at baseline and among men. To examine if the validity of the questionnaire was driving the different associations by age, we examined the reliability of the RFQ. We found reasonable reliability with ICCs between 0.5 and 0.6 when the questionnaire was administered 6 months apart and importantly found that ICCs were similar across age groups and gender subgroups (33). This reliability is in line with what has been found in other studies using historical recall of physical activity (37–41). Previous studies that have looked at the validity of physical activity questionnaires that captured historical PA have found validity of 0.4 when compared with objective measures (42). Given the similar design and consistency of the RFQ, our findings suggest that engaging in MVPA throughout early and middle ages within the adult life course can help lower the risk of colon cancer; however, the benefits were attenuated with older age. Our results did not show any residual benefit among those who were active earlier in life and decreased in adulthood or midlife. Therefore, being active and staying active (e.g., meeting physical activity guidelines, ≥2.5 hours/week of MVPA) throughout the adult life course suggests the greatest protection from colon cancer risk.

We also found effect modification by gender. Males that met or exceeded physical activity guidelines had a stronger colon cancer risk reduction (15–19%). This is consistent withs previous research that has found a stronger association of physical activity among men (35, 43, 44). Although incidence rates of colon cancer are generally higher in men than in women, these findings suggest that there may be variation in the biological mechanisms that link physical activity and colon cancer risk by sex (35, 44). Understanding the amount and types of physical activity that can reduce risk among males and females, as well as when in the life course, physical activity may be most protective, may help counteract inherent and non-modifiable risk factors of colon cancer, including age and sex linked to physical activity and decreased colon cancer rates. The protective elements of physical activity may be linked with control of body weight, reduction in body fat, regulation of metabolic health, improved insulin sensitivity, increased glucose transport, and decreased oxidative stress and inflammation (2, 9, 14, 45). In our study, those who decreased MVPA over the adult life course were more likely to be smokers, have diabetes, consume higher levels of red meat and processed meat, report their current health as fair or poor, and have a higher BMI at age 18 compared with maintainers and increasers (Table 1). Although our study did not report changes in BMI or weight status over the life course, previous studies support that maintenance of high levels of physical activity and increasing activity levels through adulthood attenuates weight gain and increases in BMI levels (46–49). Our study further supports these biological mechanisms and established risk factors, providing additional insight into the long-term lifestyle and health behaviors that can increase the risk of colon cancer.

Our findings on the association between MVPA levels and colon cancer risk are also consistent with those examining lifetime trajectories in other cancer populations. In a similar trajectory-based analysis, Saint-Maurice and colleagues (2021) found that those who maintained high levels of activity or increased activity levels had a significantly reduced risk of endometrial cancer compared to those who were consistently inactive. In addition, women who maintained and increased physical activity over the adult life course were at a lower risk of obesity in midlife (26). In a similar study, Arem and colleagues (2018) examined the association of lifetime physical activity on liver cancer incidence, finding that those who maintained activity levels over time had a 26% to 36% lower risk of liver cancer compared with those who were consistently inactive, whereas those who decreased activity over time had a higher risk of liver cancer, although this was nonsignificant (34). Our findings parallel these studies, reaffirming the current evidence of the cumulative impact of physical activity throughout the adult life course on the reduced risk of various cancers and comorbidities (9, 10).

The strengths of our study include the large prospective sample and reporting of physical activity levels at four distinct ages, from late adolescence to midlife. In addition, our data were collected from individuals who were healthy at baseline, and the participants were followed up for disease outcomes over time. Our study had some limitations. First, information about the life course was self-reported retrospectively through the baseline and RFQ questionnaires, which is prone to bias. Self-reported MVPA levels may have introduced exposure misclassification, as activity levels may have changed in the interval between the time of the RFQ and future diagnosis. This risk was minimized by utilizing trajectories and reporting of MVPA across the adult life course to provide a better picture of the effects of patterns in lifetime activity recalled by participants at specified time intervals. However, there is a chance that physical activity levels may be misreported or differentially reported according to BMI status. Further, there is a possibility of less precise recall of MVPA levels earlier in the adult life course and given the challenge of rigorously validating historical MVPA estimates we cannot rule out some bias from this source. We also cannot rule out the potential measurement error in the historical recall of physical activities that may explain the observation of effect modification by age and sex. In addition, covariates were measured through the baseline questionnaires, which may have introduced some misclassification of covariates over time or residual confounding from unknown or unmeasured variables. Our study may also be subject to selection bias, as only participants who were alive and in good health were included in the study. Finally, the generalizability of our results is limited to adults in the NIH-AARP Diet and Health Study. There was a lack of diversity in race and ethnicity in our current sample, as the majority of our sample was identified as non-Hispanic white (>90% in each MVPA trajectory).

In conclusion, our study suggests that the consistent participation in MVPA throughout the adult life course may lower colon cancer risk, whereas increasing activity levels from low activity levels throughout the adult life course has the potential to lower colon cancer risk. We observed effect modification by age and sex, with stronger associations in adults aged less than 65 years and in men. Our findings provide suggestive evidence that adults starting an exercise program later in adulthood may obtain some of the health benefits associated with physical activity and lower their risk of colon cancer. This is particularly noteworthy, considering that many adults in the US are inactive. Our findings are aligned with the current physical activity recommendations, continue to support the benefits of sustained physical activity and add to the evidence of promoting physical activity at all stages of life in population-based cancer prevention (8–10, 36, 50), Given the national and global burden of colon cancer, it is important for future research to continue exploring whether there is a critical period of life in which the effects of an inactive and sedentary lifestyle cannot be reversed. Future studies should obtain more detailed information on the dosage of physical activity levels at multiple points over the adult life course and among diverse populations to confirm our study's findings.

No author disclosures were reported.

This research was supported [in part] by the Intramural Research Program of the NIH, NIH. Cancer incidence data from the Atlanta metropolitan area were collected by the Georgia Center for Cancer Statistics, Department of Epidemiology, Rollins School of Public Health, Emory University, Atlanta, GA. Cancer incidence data from California were collected from the California Cancer Registry, California Department of Public Health Cancer Surveillance and Research Branch, Sacramento, CA. Cancer incidence data from the Detroit metropolitan area were collected by the Michigan Cancer Surveillance Program, Community Health Administration, Lansing, MI. The Florida cancer incidence data used in this report were collected by the Florida Cancer Data System, Miami, FL, in contract with the Florida Department of Health, Tallahassee, FL. The views expressed herein are solely those of the authors, and do not necessarily reflect those of the FCDC or FDOH. Cancer incidence data from Louisiana were collected from the Louisiana Tumor Registry, Louisiana State University Health Sciences Center, School of Public Health, New Orleans, LA. Cancer incidence data from New Jersey were collected from the New Jersey State Cancer Registry, Rutgers Cancer Institute of New Jersey, New Brunswick, NJ. Cancer incidence data from North Carolina were collected from the North Carolina Central Cancer Registry, Raleigh, NC. Cancer incidence data from Pennsylvania were supplied by the Division of Health Statistics and Research, Pennsylvania Department of Health, Harrisburg, PA. The Pennsylvania Department of Health specifically disclaims responsibility for any analysis, interpretation, or conclusion. Cancer incidence data for Arizona were collected from the Arizona Cancer Registry, Division of Public Health Services, Arizona Department of Health Services, Phoenix, AZ. Cancer incidence data from Texas were collected by the Texas Cancer Registry, Cancer Epidemiology and Surveillance Branch, Texas Department of State Health Services, Austin, TX. Cancer incidence data from Nevada were collected from the Nevada Central Cancer Registry, Division of Public and Behavioral Health, State of the Nevada Department of Health and Human Services, Carson City, NV. We are indebted to the participants of the NIH-AARP Diet and Health Study for their outstanding cooperation. We also thank Sigurd Hermansen and Kerry Grace Morrissey from Westat for ascertainment and management of study outcomes, and Leslie Carroll at Information Management Services for data support and analysis. The NIH-AARP Diet and Health Study cohort is in part funded by the Intramural Division of Cancer Epidemiology and Genetics at the U.S. National Cancer Institute.