Physical Activity in Different Periods of Life, Sedentary Behavior, and the Risk of Ovarian Cancer in the NIH-AARP Diet and Health Study

Ovarian cancer is the eighth most common cancer, and the most lethal gynecologic cancer among U.S. women (1). Its etiology remains poorly understood, and only a few factors have been consistently linked to ovarian cancer risk. The most important risk factor is a family history of breast and/or ovarian cancer, whereas parity and the use of oral contraceptive have protective effects (2).

It has been hypothesized that physical activity may protect against ovarian cancer through its influence on hormonal, inflammatory and immune pathways, or suppression of ovulation (3–5). However, epidemiologic studies have reported conflicting results. While most studies found a null or weak inverse association (6–13), two prospective studies reported an elevated risk of ovarian cancer with a high level of vigorous activity (14, 15). Several studies have reported that physical activities at mid-teens or early adulthood (<40 years) and are positively associated with ovarian cancer risk (9–11, 16), suggesting that the age period at which women engage in physical activity may be an important factor. Moreover, some studies suggested that the association may also differ by BMI, hormone use, and histologic subtypes of the disease (17, 18). In summary, much remains to be understood about the relationship between physical activity and ovarian cancer and how it is influenced by various factors.

Recent literature suggested that sedentary behavior may play an important role in cancer etiology, independent of moderate to vigorous activity (19). A few studies have examined sedentary behavior in relation to ovarian cancer. One prospective study (11) and 2 case–control studies (20, 21) have suggested that prolonged sitting may be associated with increased risk of ovarian cancer.

An early study using data from a large prospective cohort, the NIH-AARP Health and Diet Study, found no association between the frequency of moderate or vigorous physical activity participation and ovarian cancer risk (13). Comparing to women reporting no moderate or vigorous activity, women who engaged in both activities had a relative risk (RR) of 1.10 [95% confidence interval (CI), 0.82–1.48] for ovarian cancer. However, the early study only included 309 cases, which limited the study power. Now with 3 more years in follow-up, the number of cases has more than doubled. Therefore, we reevaluated the relationship between physical activity and ovarian cancer risk in this cohort, extended our analysis to include physical activity in different life periods and sedentary behavior, and examined whether these relationships differ by other risk factors and cancer subtypes.

The NIH-AARP Diet and Health Study was established in 1995–1996. Participants were recruited from AARP members who were 50 to 71 years old and resided in 6 U.S. states (CA, FL, LA, NJ, NC, and PA) and 2 metropolitan areas (Atlanta, GA, and Detroit, MI). Details of the NIH-AARP study were reported previously (22). The baseline questionnaire captured information on physical activity, dietary intake, demographic factors, medical and reproductive history, and family history of cancer. Of the 617,119 questionnaires returned, 566,398 were nonduplicate and satisfactorily completed. In 1996–1997, a second questionnaire, the risk factor questionnaire, was mailed to the baseline respondents who did not have self-reported prevalent cancer at baseline or within 6 months from the time of the baseline questionnaire. It collected further information on physical activity, sedentary behavior, and medical history. Of the initial cohort, 337,074 men and women completed the risk factor questionnaire. The study was approved by the National Cancer Institute Special Studies Institutional Review Board.

In the baseline questionnaire, participants were asked to select a response that best described their daily activity in the past 12 months, beside exercising or playing sports: sit without walking very much; sit but walk fair amount; stand or walk a lot without carrying or lifting things; lift or carry light loads or climb stairs/hills often; or do heavy work or carry heavy loads. Participants were also asked about how often (never, rarely, 1–3 times/month, 1–2 times/week, 3–4 times/week, or ≥5 times/week) they engaged in vigorous household, occupational and recreational physical activities during a typical month in the past year. Vigorous activity was defined as physical activities that lasted at least 20 minutes and caused increases in breathing or heart rate or sweating. We combined the never and rarely response categories in analysis.

In the risk factor questionnaire, participants were given two separate lists of examples of “moderate and vigorous” recreational and household physical activities (MVPA). They were asked to indicate how often (never, rarely, weekly but <1, 1–3, 4–7, and >7 hours/week) they participated these activities during various ages and time periods: 15–18, 19–29, and 35–39 years and in the past 10 years. The never and rarely response categories were combined for analysis. The risk factor questionnaire also collected information on sedentary behaviors. Participants reported the amount of time they spent in a typical day sitting overall (<3, 3–4, 5–6, 7–8, and ≥9 hours) and watching television or videos (none, <1, 1–2, 3–4, 5–6, 7–8, and ≥9 hours), during the past year. For analysis, we collapsed the lowest categories for both measures, and they were each coded into 4 categories (<3, 3–4, 5–6, and ≥7 hours/day).

Cohort members were followed annually through the US Postal Service's National Change of Address database for address changes and through linkage to the US Social Security Administration Death Master File and the National Death Index Plus for vital status. Incident ovarian cancer cases were identified through linkage to eight original and two additional (AZ and TX) state cancer registry databases. A previous validation study found that approximately 90% of cancers were identified through registry linkage (23). Cancer registry data included cancer diagnosed, diagnosis date, morphology code, grade, and stage information.

Using the International Classification of Diseases for Oncology, Third Edition (ICD-O-3) (24), we defined cases as first primary epithelial ovarian cancer (ICD-O-3 code: C56.9). Borderline and nonepithelial ovarian cancer cases were excluded in this study. The cases were further classified into the following subtypes: serous (code: 8260, 8441, 8450, 8460, 8461, 8462), endometrioid (code: 8380, 8381, 8560, 8570), mucinous (code: 8470, 8471, 8472, 8480, 8481, 8490), clear cell (code: 8310, 8313), and other epithelial (code: 8010, 8020, 8021, 8050, 8070, 8120, 8140, 8240, 8246, 8255, 8260, 8323, 8440, 8450, 8490, 8562).

Of those who satisfactorily completed the baseline questionnaire, we excluded individuals who were male (n = 339,666), whose questionnaire was completed by proxies of the intended respondent (n = 1,265), or who died, requested to be withdrawn or had moved out of the study area before baseline (n = 91), leaving a total of 225,376 potentially eligible women.

For analysis of baseline physical activity, we further excluded those who had prevalent cancer except nonmelanoma skin cancer (n = 23,876), had a history of oophorectomy or unknown oophorectomy status (n = 48,201), and developed nonepithelial or borderline ovarian cancer during follow-up (n = 63). In addition, we excluded 113 women who were identified through National Death Index with ovarian cancer as the cause of death but had no information on timing of diagnosis date or cancer histology. The final baseline analytic sample included 153,123 women of whom 753 developed invasive epithelial ovarian cancer. We additionally removed women who had missing information on daily routine (n = 1,760, including 16 cases) and vigorous activity (n = 4,234, including 8 cases) in the respective analysis.

For analysis of physical activity and sedentary behavior reported in the risk factor questionnaire, we restricted to women who responded to this survey. Of the baseline analytic sample, 99,372 women completed the risk factor questionnaire. We excluded women whose questionnaire was completed by proxies (n = 835, including 4 cases), who died or moved out of the study area before their risk factor questionnaire was received and scanned (n = 14), and who reported to have prevalent cancer in the risk factor questionnaire (n = 687, including 5 cases), leaving 96,247 women with 467 cases. We further excluded women with missing information for any of the recreational/household activity and sedentary behavior questions in the respective analysis (overall sitting: n = 479, 4 cases; TV watching: n = 405, 4 cases; MVPA: n = 1,495, 7 cases in past 10 years; n = 2,031, 9 cases at age 35–39; n = 2,222, 11 cases at age 19–29; n = 2,150, 9 cases at age 15–18).

RRs and 2-sided 95% CIs were estimated with the Cox proportional hazards model, using the SAS PROC PHREG procedure (SAS 9.2; SAS Institute). Person-years of follow-up time were calculated from the baseline (for the analysis of baseline physical activity) or from the scan date of the risk factor questionnaire (for the analysis of the MVPA and sedentary behavior in the risk factor questionnaire) until the diagnosis of invasive epithelial ovarian cancer, moved out of the study area, died from any cause, or December 31, 2006, whichever came sooner. We evaluated and confirmed the proportional hazards assumption for the main exposures by including interaction terms with time and using the Wald χ² procedure to test if coefficients equaled zero.

We present RRs and 95% CIs from both age adjusted and multivariate models adjusted for risk factors of ovarian cancer that were identified as a priori potential confounders, including age at entry (continuous), race/ethnicity (non-Hispanic White, non-Hispanic Black, and other), education (<12 years, high school, some college, college, and graduate school), number of live birth (0, 1, 2, 3-4, 5–9, and 10+), age at menarche (≤10, 11–12, 13–14, and 15+ years), age at menopause (<40, 40–44, 45–49, 50–54, 55+ years and still menstruating), oral contraceptive use (never, 1–4, 5–9, 10+ years), any menopause hormone therapy (MHT) use (never, current, and former), and smoking (never, current smoker, and former smoker). Body mass index (BMI) is a potential mediator of the effect of physical activity or sedentary behavior on ovarian cancer risk. Therefore, separate multivariate models with or without BMI were used to assess the associations. In additional models, we further adjusted for sedentary behavior and light activities in the analysis of current MVPA and current MVPA in the analysis of historical MVPA. We found minimal changes in results after including BMI in the model and therefore excluded BMI in our analysis. In the analysis of MVPA at age 15–18, 19–29, and 35–39 years, we additionally adjusted for MVPA in the past 10 years in a separate model to examine the independent effect of MVPA at younger ages. We conducted stratified analysis according to BMI and MHT status. Test for interaction were done using the likelihood-ratio test, comparing models with models without the product term between BMI and the covariate of interest.

At baseline, most women described their daily routine physical activity (excluding recreational activity) as walking and standing or walking and sitting. About 22% of the women reported never or rarely engaging in vigorous activity in the past 12 months, whereas 15%, 21%, 25%, and 17% engaged in vigorous activity 1–3 times/month, 1–2 times/week, 3–4 times/week, and ≥5 times/week. Compared to women who never or rarely engaged in vigorous physical activities, those who engaged in higher levels of vigorous activities had lower average BMI, were more likely to be college educated, to report excellent health status, and to have ever used MHT, and were less likely to be current smokers (Table 1). We also found an inverse correlation between overall sitting and MVPA in the previous 10 years reported in the risk factor questionnaire (Spearman correlation coefficient: −0.19).

No association was observed between either daily activity or vigorous physical activity level and ovarian cancer risk as reported on the baseline questionnaire (Table 2). Compared with the reference group (never/rarely engaging in vigorous physical activity), women who reported engaging in vigorous physical activities more than 5 times/week had a RR of 1.04 (95% CI, 0.83–1.30) for developing invasive epithelial ovarian cancer. Women who reported doing heavy work or carrying heavy loads as daily routine had a nonsignificant 50% lower risk of ovarian cancer than did women who reported sitting without much walking. However, the number of cases in the former category was extremely small (n = 6). No reduction in risk was observed among women who climbed stairs or carried light loads routinely. Including BMI in the models had minimal impact on the results. In subgroup analysis stratified by BMI (BMI < 25 and BMI ≥ 25), the associations did not differ by strata (data not shown).

We also examined ovarian cancer risk in relation to sedentary behavior and moderate and vigorous physical activity among women who responded to the risk factor questionnaire. The amount of time spent sitting overall was not associated with ovarian cancer risk (RR _{7+ h vs. <3 h}, 1.06; 95% CI, 0.81–1.39; Table 3). Similarly, no excess risk was observed among women who spent prolonged hours watching TV (RR _{7+ h vs <3 h}, 1.02; 95% CI, 0.67–1.55). Further adjusting for recent MVPA or stratifying by BMI or MHT use did not modify these results (data not shown). We also did not detect an interaction between overall sitting and MVPA (P_interaction = 0.66).

MVPA in the 10 years before the completion of the risk factor questionnaire and at age 15–18 was not associated with ovarian cancer risk (Table 4). Although there was a suggestion of an increased risk for women who engaged in an intermediate amount of MVPA (<1, 1–3, or 4–7 hours/week) at early adult life period, the RR estimates were not statistically significant, and no association was found with higher amount of MVPA (7+ hours/week) at these life periods. The results remained largely similar when we included overall sitting and MVPA in the 10 years (for analysis of historical MVPA) the model (data not shown).We also found no significant interactions between MVPA at different life periods and BMI or MHT status (data not shown).

We also ran analysis specific to serous versus nonserous type of ovarian cancer, but due to small numbers, we were unable to further examine specific subtypes of nonserous ovarian cancers. Neither physical activity at baseline nor sedentary behavior was related to serous or nonserous ovarian cancer (data not shown). MVPA in the past 10 years was also not associated with either subtype. We observed that among women engaged in intermediate amount of MVPA at earlier age periods, there appeared to be an elevated risk of nonserous ovarian cancer, but not serous subtype, although the increase in risk was only at borderline statistical significance (Table 5). For all the aforementioned analysis, we conducted sensitivity analysis by excluding cases diagnosed in the first 2 years of follow-up and it did not alter the results (data not shown).

In this population of middle-aged and older U.S. women, we found that ovarian cancer risk was not associated with daily physical activity pattern, time spent on sedentary behaviors, and the amount of MVPA after the age of 40 years. However, we cannot rule out the possibility of an increase in risk associated with intermediate levels of MVPA in early adulthood.

The lack of association between physical activity during middle-to-old ages and ovarian cancer risk is largely consistent with results from previous studies, especially prospective studies. A meta-analysis published in 2007 summarized findings from 12 case–control and cohort studies and reported a 20% risk reduction associated with the highest versus lowest levels of recreational physical activity (25). However, the summary association did not reach statistical significance when the analysis was restricted to the 6 cohort studies (RR, 0.81; 95% CI, 0.57–1.17), and the 3 large prospective studies published after the meta-analysis all reported a null association (12, 13, 26).

Three studies to date have looked into the effect of sedentary behavior as a possible risk factor for ovarian cancer. A case–control study conducted in China found that long hours spent sitting at work was associated with increased risk (21). In another case–control study in Turkey, the authors reported no association between sitting time at work and risk of ovarian cancer, but the number of cases was quite small (20). In the only other prospective study of which we are aware, Patel and colleagues reported that among U.S. women ages 50–74 years, prolonged sitting in leisure time (6+ vs. < 3 hours/day) was associated with a significant 55% increase in ovarian cancer risk after adjusting for physical activity (11). In contrast, we did not observe an elevated risk with prolonged sitting overall or television watching, both of which have been associated with all-cause mortality (27) and endometrial cancer (28) in our population.

When we examined the effect of MVPA at earlier ages, we found a statistically nonsignificant increase in risk in women who engaged in intermediate levels of MVPA at ages 19–29 and 35–39 years. The timing at which physical activity might have an effect is intriguing. The etiology and critical period of ovarian cancer development are not well-understood, although the protective effects of oral contraceptive use and pregnancy suggest that childbearing ages may be particularly vulnerable to carcinogenic factors. A number of previous studies also evaluated the effects of physical activity at early ages, and the results were mixed (9–11, 15–18, 25, 29–31). Most of these studies reported largely null associations, but there were a few exceptions. Three articles reported reduced risk of ovarian cancer in relation to historical activity levels (18, 30, 31). In an Italian case–control study, occupational physical activity at ages 30–39 years were inversely associated with ovarian cancer, but no relationship was found for leisure time activity (30). Pan and colleagues also found decreased risk with increased level of occupational activities, but the results were only significant for the age period of early 50s (18). A third case–control study reported inverse association with leisure time activity at teens and early adulthood (31). In contrast, 2 studies reported suggestive positive associations between ovarian cancer and physical activity at early life or adulthood (9, 16). Carnide and colleagues showed 30% to 70% increase in risk with physical activity of <2 times/week, when compared with none/slight activity, in mid-teens and early 30s (16). In the Netherlands Cohort Study, the authors found that a history of sports at 1–2 hours/week was associated with doubling the risk of ovarian cancer, although no age period was specified in the article (9). However, the increase in risk in all the studies only achieved borderline significance. Overall, these findings do not support a clear relationship between physical activity at different life periods and ovarian cancer risk.

The association between physical activity and ovarian cancer may be further complicated by the different risk factors associated with different subtypes of ovarian cancer. As elucidated by earlier studies, epithelial ovarian cancer is a group of heterogeneous diseases and each subtype may have a different set of risk factors (26, 32, 33). Although our findings suggested that the modest increase in risk with MVPA in early adulthood was stronger in non-serous types, unfortunately, we were unable to examine the relationship with individual subtypes due to small numbers. Future studies with adequate sample size of non-serous ovarian cancers will be particularly valuable to evaluate the subtype-specific effect of physical activity.

It has been postulated that there may be multiple mechanisms through which physical activity may influence ovarian cancer risk. However, the activation of different pathways may lead to different or even opposite effects on ovarian cancer risk. Physical activity may suppress chronic inflammation, enhance immune function, improve DNA repair mechanism, and decrease circulating levels of endogenous estrogens, all of which may lead to decreased risk of ovarian cancer (3–5). In addition, it is well-established that regular strenuous activity disrupts menstrual function and increases the frequency of anovulatory cycles in young women (34, 35), which, according to the “incessant ovulation” hypothesis, may also lead to reduced risk (36). On the other hand, physical activity may plausibly increase ovarian cancer risk. For example, physical activity may stimulate the production of pituitary gonadotropins (37, 38), a high level of which has been proposed as a risk factor for ovarian cancer (2, 39). In addition, circulating androgen levels tend to increase after exercise (40), and androgenic stimulation of ovarian epithelial cells may lead to tumorigenesis (3). Moreover, exercise may also lower progesterone (41, 42), a potential protective factor against ovarian cancer (3, 39). Given the complexity of the possible physiologic and pathologic changes involved, the overall effect of physical activity on ovarian cancer risk may be determined by a combination of factors, and as suggested by our study and the works by others, such factors include the timing, dose, and type of activity, woman's exposure to other risk factors, and the subtypes of the cancer.

There are several limitations of this study. The main limitation is the potential measurement error and misclassification of the self-reported levels of physical activity and sedentary behavior. Our physical activity questions were not validated, although using the same exposure measures, previous studies in this cohort have linked physical activity and sedentary behavior to the risk of postmenopausal breast (43, 44), colorectal (45), and endometrial cancer risk (28), the major cancer sites for which an etiologic connection with physical activity have been consistently established. In addition, we did not have enough cases of most of the non-serous cancers, making it impossible to examine the association specific to each ovarian cancer subtypes. Finally, although we tried to control for multiple ovarian cancer risk factors, and adjusting which did not make a big impact on the results, we could not rule out the possibility of residual confounding.

Our study has several strengths. Its prospective nature decreased the likelihood of differential recall bias, a serious limitation of many case–control studies. The large cohort with a fairly long follow-up time has allowed us to not only to test the overall effect of physical activity and ovarian cancer but also to assess the associations among people with different BMI, MHT use and with serous and non-serous types of ovarian cancer. We have excluded women with a history of any cancer diagnosis, and this reduces the influences of existing malignancies on physical activity. We also conducted a sensitivity analysis, excluding cases diagnosed in the first 2 years, further minimizing the possibility of reverse causation due to undiagnosed ovarian cancer.

In summary, we found no evidence supporting a protective effect of physical activity or an adverse effect of prolonged time in sedentary behavior at middle to old ages on ovarian cancer risk. Given the many health benefits of physical activity, including reduced risk for early mortality, heart disease, and certain cancers, the current findings should not dissuade women of any age from engaging in physical activity. To better understand the potential positive association between physical activity in early adulthood and ovarian cancer, future studies should be designed to have a better measure of different domains of physical activity and activity at different ages, have adequate power to study specific subtypes of the cancer and to evaluate the influence of other risk factors, and address specific pathways of the complex mechanisms.

A.R. Hollenbeck is a consultant/advisory board member of Love/Avon Army of Women and SPIM Board of Directors. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Q. Xiao, N. Wentzensen, A.R. Hollenbeck, C.E. Matthews

Development of methodology: Q. Xiao

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.R. Hollenbeck

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Q. Xiao, H.P. Yang, N. Wentzensen, C.E. Matthews

Writing, review, and/or revision of the manuscript: Q. Xiao, H.P. Yang, N. Wentzensen, A.R. Hollenbeck, C.E. Matthews

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.R. Hollenbeck, C.E. Matthews

Study supervision: C.E. Matthews

This research was supported by the Intramural Research Program of the NIH, National Cancer Institute, Department of Health and Human Services.

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.