Physical Activity and Outcomes in Patients with Stage III Colon Cancer: A Correlative Analysis of Phase III Trial NCCTG N0147 (Alliance)

Each year in the United States, more than 50,000 deaths are attributed to colorectal cancer (1). Many more such deaths are likely prevented by virtue of early detection and curative therapy. To further reduce the mortality burden of colorectal cancer, it is important to identify factors associated with prognosis among colon or rectal cancer patients, particularly factors that may be intervened upon or modified. Although stage at diagnosis remains the strongest known predictor of colorectal cancer survival outcomes (1, 2), and certain tumor attributes [e.g., DNA mismatch repair (MMR) status, BRAF mutation status] have also been consistently associated with prognosis (2–7), increasing evidence suggests that modifiable lifestyle factors may also play a role in survival (8–13). One such factor that may plausibly be associated with colorectal cancer survival is physical activity (12–14).

A sizable literature supports the association between physical activity and reduced risk of developing colon cancer (15, 16). Among numerous plausible mechanisms underlying this association with risk, physical activity contributes to reductions in circulating levels of insulin and insulin-like growth factor-1 (IGF-1), both of which inhibit apoptosis and promote cell proliferation in colon tumor cells (17–23). Such mechanisms could also be expected to contribute to a relationship between physical activity and colon cancer survival. In a recent meta-analysis, Wu and colleagues found significantly more favorable disease-specific and overall survival (OS) after colorectal cancer diagnosis associated with any versus no exercise and with high versus low physical activity levels, regardless of whether physical activity was ascertained with regard to the time period before or after cancer diagnosis (14). However, questions remain as to whether the association between physical activity and colon cancer survival outcomes varies across case groups defined by patient and/or tumor characteristics.

Using data from a large, multicenter clinical trial of adjuvant chemotherapy for stage III colon cancer, we evaluated the association of physical activity patterns with disease-free survival (DFS), OS, and time to recurrence (TTR), with consideration for possible heterogeneity according to tumor and patient attributes.

N0147, led by the North Central Cancer Treatment Group (NCCTG; now a part of the Alliance for Clinical Trials in Oncology), is a multicenter phase III randomized trial of adjuvant 5-fluorouracil, oxaliplatin, and leucovorin (FOLFOX) with or without cetuximab in patients with resected stage III colon cancer (24). A total of 3,397 patients were enrolled, including 2,686 who were randomized to the primary treatment comparison arms (FOLFOX vs. FOLFOX plus cetuximab). The current study is limited to N0147 participants enrolled in the primary treatment comparison arms who had completed a patient questionnaire at the baseline study visit, and, specifically, answered physical activity questions (N = 1,992). All participants signed institutional review board–approved, protocol-specific informed consent forms in accordance with federal and institutional guidelines.

A patient questionnaire eliciting information on a variety of risk factors, including smoking history, alcohol consumption patterns, and usual physical activity patterns was administered to N0147 participants prior to treatment. With respect to physical activity, participants were asked to report on their usual level of physical activity “during a routine day” (almost none/mild activity/moderate/heavy activity). They were also asked to report on how often during their “free time” they took part in “any” physical activity (never/about once a month/several times a month/several times a week/daily). In addition to the question of any free-time physical activity, participants were asked separate questions about the frequency of their participation in “moderate” physical activity (e.g., golf, gardening, long walks, bowling) and “vigorous” physical activity (e.g., jogging, racket sports, swimming, aerobics) during their free time. A specific reference time point was not provided in the physical activity questionnaire items; however, these questions were administered at a time shortly following surgical resection for stage III colon cancer and prior to receipt of chemotherapy.

Tumor tissue blocks were collected from the original surgical resection for study participants. Testing for DNA MMR status, as well as KRAS- and BRAF mutation status, was conducted centrally at the Mayo Clinic using previously described methods. Briefly, MMR status was determined by IHC assessment of hMLH-1, hMSH-2, and hMSH-6 (25); patients with tumors exhibiting a loss of expression for any of these proteins were classified as having tumors with defective MMR (dMMR), whereas patients with no loss of expression were classified as having tumors with proficient MMR (pMMR). DNA isolated from tumor specimens was used to test for 7 mutations in KRAS exon 2 hotspot codons 12 and 13, and for the BRAF V600E mutation (26).

We evaluated associations with respect to three overlapping clinical outcomes: TTR, OS, and DFS. TTR was defined as the time from randomization to first documented disease recurrence; participants who died before a recurrence were censored at their last disease assessment date. OS was defined as the time from randomization to death from any cause. DFS was defined as the time from randomization to the first documented cancer recurrence or death from any cause, whichever came first. Patients were censored at 5 years postrandomization in analyses of DFS and TTR and censored at 8 years postrandomization in analyses of OS. The current analysis was based on follow-up through December 3, 2014.

We evaluated associations between the following self-reported physical activity attributes and colon cancer outcomes: usual daily physical activity, frequency of any free-time physical activity, frequency of moderate intensity physical activity, frequency of vigorous physical activity. For the latter three frequency variables, based on observed similarities in point estimates and small numbers, we grouped together exposure categories for those who reported never participating in physical activity and those who participated only about once a month; similarly, we grouped together exposure categories for those who reported participating in free-time physical activity several times a month to daily.

The distribution of patient and tumor attributes was tabulated and compared across grouped categories of physical activity frequency using Wilcoxon (continuous; ref. 27) and χ² (categories; ref. 28) tests. We used Cox proportional hazards regression models (29, 30) to assess associations between physical activity variables and study outcomes, with adjustment for the following attributes: age at diagnosis, sex, T-stage, N-stage, Eastern Cooperative Group (ECOG) performance score, histologic grade, tumor location, KRAS mutation status, BRAF mutation status, MMR status, body mass index (BMI), smoking frequency, alcohol use, and treatment assignment. Confounders were selected a priori and were included in the analytic model as parameterized in Table 1. Participants with missing data on model covariates were excluded from multivariable analyses (n = 211). We also conducted analyses stratified by each of the aforementioned adjustment variables and evaluated potential interaction by these variables in the association between physical activity and patient outcomes. Proportional hazards assumptions were verified by testing for a nonzero slope of the scaled Schoenfeld residuals on ranked failure times (31). Two-sided P values <0.05 were considered statistically significant.

All statistical analyses were performed by the Alliance Statistics and Data Center. Analyses were conducted using SAS version 9.3 (SAS Institute, Cary, NC).

Approximately 75% of study participants reported participating in any physical activity during their free time at least several times a month; more specifically, 68% reported participating in moderate-intensity physical activity more than once a month and 16% reported participating in vigorous intensity physical activity more than once a month. Compared with those who reported no participation in free-time physical activity or participation only once a month, participants who took part in any physical activity more than once a month were slightly younger (P = 0.04), more likely to be male (P < 0.001), had lower ECOG performance scores (P = 0.003), were more likely to consume alcohol (P < 0.001), and had a lower BMI (P < 0.001; Table 1).

Over the course of study follow-up, 28% (n = 505) of study participants died and 30% (n = 541) experienced a disease recurrence (89% of whom subsequently died during follow-up). The median follow-up time was 6.9 years (IQR, 6.1–7.6 years). After multivariable adjustment, there was no association between usual daily activity level and any of the evaluated outcomes (Table 2). However, reported participation in free-time physical activity more than once a month versus once a month or less was significantly associated with OS [HR = 0.76; 95% confidence interval (CI), 0.63–0.93; P = 0.007] and DFS (HR = 0.82; 95% CI, 0.69–0.99; P = 0.04); these associations persisted when we restricted the comparison group to those who reported no participation in any free-time physical activity (HR = 0.73 and HR = 0.77, respectively). Reported participation in moderate-intensity physical activity more than once a month was similarly associated with outcomes, particularly with respect to OS (HR = 0.80; 95% CI, 0.66–0.96; P = 0.02). Observed associations between vigorous-intensity physical activity and patient outcomes were more modest and not statistically significant, as were associations between any or moderate-intensity physical activity and TTR.

In stratified analyses, the association between reported participation in any free-time physical activity more than once a month (vs. once a month or less) and more favorable OS appeared to be restricted to those with smaller tumors (stage T3 vs. T4, P_interaction = 0.03). This association was also suggestively but not significantly stronger among those who were normal weight or obese (but not overweight, P_interaction = 0.07), among those with tumors not exhibiting mutations in BRAF or KRAS (vs. with a somatic mutation in either gene, P_interaction = 0.20), and among those who received FOLFOX without cetuximab (vs. with cetuximab, P_interaction = 0.16; Fig. 1). Similar patterns of difference were evident in analyses of DFS (Fig. 2).

In this cohort of clinical trial participants with stage III colon cancer, we observed a modest association between physical activity and colon cancer outcomes, suggesting more favorable DFS and OS in individuals who reported participating in recreational physical activity more than once a month around the time of colon cancer diagnosis. These associations were consistent across most participant subgroups defined by demographic, tumor, and lifestyle factors.

Similar to our overall findings, most previous studies into the association between prediagnostic physical activity and subsequent colorectal cancer outcomes have suggested modestly more favorable survival in those who were physically active than in those who were not active or minimally active during this time (12, 13, 32–38). Most recently, Walter and colleagues reported a significant inverse association between average leisure time activity in the years immediately preceding colorectal cancer diagnosis and subsequent OS, and a more modest, borderline significant inverse association with DFS (38). Although physical activity was parameterized in quartiles of MET-hours/week in that study, point estimates comparing outcomes for those in upper quartiles of physical activity with those in the lowest quartile (HR_OS = 0.64 to 0.81, HR_DFS = 0.78 to 0.92) are within the range of point estimates reported here. Previous meta-analyses have similarly reported modestly more favorable OS and disease-specific survival associated with participation in any versus no physical activity among individuals with colorectal cancer, with summary point estimates of 0.74 for OS and 0.75 for disease-specific survival (32–34). However, this pattern of association has not been uniformly observed across all studies (39).

To our knowledge, only one prior study has investigated the association between physical activity and colon cancer outcomes with consideration for the tumor attributes evaluated here (13). In contrast to our findings, Hardikar and colleagues reported suggestive evidence that the association between prediagnostic physical activity and subsequent colorectal cancer survival may be slightly stronger among those with tumors exhibiting high levels of microsatellite instability (mimicking the dMMR profile evaluated in this study); their results also indicated no difference in the magnitude of association by BRAF/KRAS somatic mutation status (13). Other analyses of postdiagnostic physical activity in relation to colorectal cancer outcomes have suggested that the benefits of physical activity may be greater among individuals with colorectal tumors that were negative for nuclear cadherin-associated protein β1 (CTNNB1; ref. 40), positive for PTGS2 expression (41), and negative for expression of insulin receptor substrate 1 (IRS1; ref. 42). Although CTNNB1, PTGS2, and IRS1 expression levels were not assayed in the N0147 study population, these prior studies highlight potential heterogeneity in the association between physical activity and cancer outcomes. Thus, although the broad-range benefits of physical activity are well established, it is plausible that the magnitude of that benefit may vary across the population of colon cancer patients.

The precise mechanisms underlying the association between physical activity and colon cancer outcomes are unknown; however, several factors are likely to contribute. In particular, physical activity has been suggested to lower insulin and IGF levels, and to elevate levels of IGF-binding proteins, thereby reducing the potential impact of insulin and IGF on tumor growth (43, 44). Other studies have suggested that moderate physical activity is associated with greater T-cell proliferation relative to in sedentary individuals (45–48), which may enable an enhanced immune response in colon cancer. Similarly, physical activity has been shown to reduce levels of chronic inflammation and serum prostaglandin levels (44, 49, 50); chronic overproduction of prostaglandins has previously been linked with cancer progression and metastasis (51, 52). Several additional mechanisms may plausibly underlie the observed associations with physical activity, both with respect to OS and disease-specific outcomes (44).

Results from the current analysis should be interpreted in the context of study limitations. In particular, information as to the timing of reported physical activity was nonspecific: Participants were asked about current physical activity patterns at a time shortly following colon cancer diagnosis, with no specification as to whether or not those patterns had recently changed in response to cancer symptoms, treatment, or life circumstances. Furthermore, because this physical activity information was not updated in follow-up interviews, we were not able to evaluate associations with continued patterns of physical activity after colon cancer diagnosis and treatment. We also lacked information on participation in specific activities (e.g., swimming, running) and the typical duration of activity sessions. As such, we were unable to estimate metabolic equivalent hours or other metrics of actual energy expenditure from physical activity, as previous studies have done (12–14, 35, 53–55). Given that the current study was conducted within a randomized clinical trial population, the generalizability of study findings to the broader population of all colon cancer patients is not fully known. Patients who enroll in randomized trials represent a more selected population and, thus, their personal characteristics, such as physical activity patterns, may differ from the general population of colon cancer patients. Residual confounding is also a potential concern; for example, we lacked detailed information on income and socioeconomic status, which are plausibly associated with both physical activity (56) and colon cancer outcomes (57). We also lacked information on other comorbidities (e.g., diabetes, cardiovascular disease) that could also have an impact on OS and also related to physical activity (58, 59). Also, although we were able to investigate heterogeneity by a number of participant and tumor attributes, it is plausible that the association between physical activity and colon cancer outcomes could vary according to attributes not captured by this study (e.g., NRAS mutation status). Finally, although the overall included sample size was large, our numbers were limited within some strata defined by patient or tumor characteristics, which reduced our ability to identify heterogeneity in the association between physical activity and colon cancer survival.

Our study also has several important strengths, including the availability of information regarding multiple tumor markers, potential confounders, and standardized treatment data. Incorporating this information allowed us to better disentangle the association between physical activity and outcomes after colon cancer diagnosis. The associations we observed in this context are consistent with the existing literature (13, 36, 37, 53, 60–62), indicating a modest favorable association between physical activity and colon cancer outcomes.

Current guidelines suggest that cancer survivors, including colon cancer survivors, engage in 150 minutes of moderate-intensity or 75 minutes of vigorous-intensity physical activity per week (63, 64). Although we did not collect information on duration of activity, and although physical activity information from our study cannot be clearly classified as reflecting strictly pre- or postdiagnostic patterns, our findings do support these guidelines, suggesting that higher physical activity around the time of colon cancer diagnosis is associated with more favorable colon cancer outcomes.

P.J. Limburg has ownership interest in Exact Sciences. No potential conflicts of interest were disclosed by the other authors.

The content of this article does not necessarily reflect the views or policies of the NCI.

Conception and design: A.I. Phipps, Q. Shi, R.M. Goldberg, D.J. Sargent, S.R. Alberts

Development of methodology: Q. Shi, R.M. Goldberg, D.J. Sargent, S.R. Alberts

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Gill, R.M. Goldberg, B. Jahagirdar, A. Shields, F.A. Sinicrope, D.J. Sargent, S.R. Alberts

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.I. Phipps, Q. Shi, T.J. Zemla, R.M. Goldberg, S. Hardikar, P.J. Limburg, P.A. Newcomb, F.A. Sinicrope, D.J. Sargent, S.R. Alberts

Writing, review, and/or revision of the manuscript: A.I. Phipps, Q. Shi, T.J. Zemla, E. Dotan, S. Gill, R.M. Goldberg, S. Hardikar, P.J. Limburg, P.A. Newcomb, A. Shields, F.A. Sinicrope, S.R. Alberts

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R.M. Goldberg, D.J. Sargent

Study supervision: R.M. Goldberg

Research reported in this publication was supported by the NCI of the NIH under Award Numbers K07CA172298; U10CA180821, U10CA180882 and UG1CA189823 (to the Alliance for Clinical Trials in Oncology); U10CA180790; U10CA180850; UG1CA189863; U10CA180820 (ECOG-ACRIN); U10CA180863 and CCSRI #021039 and 704970 (CCTG); U10CA180835 and U10CA180888 (SWOG).

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.