Physical Activity and Endometrial Cancer Risk, a Systematic Review of Current Evidence

Endometrial cancer is the eighth most common cancer in women worldwide. An estimated 200,000 cases occur yearly, accounting for 1.8% of all new cases of cancer worldwide and 4% to 8% in Western industrialized countries (1). Known risk factors for endometrial cancer include obesity, postmenopausal estrogen replacement, ovarian dysfunction, type II diabetes, infertility, nulliparity, and tamoxifen use (2, 3). Recently, physical inactivity has received increasing attention as a potential risk factor for endometrial cancer (4-6). Physical activity together with maintenance of a healthy body weight would provide women with a powerful strategy for the prevention of endometrial cancer and possibly other cancers as well (7).

The majority of studies on the association between physical activity and cancer have focused on breast cancer. About 50 observational studies have been conducted on total or leisure time activities, in which both null results and inverse associations between physical activity and breast cancer have been observed (8). By contrast, much fewer studies examined the association between physical activity and endometrial cancer, and the results seem to point rather uniformly to risk reduction from physical activity. By systematically reviewing all the evidence from observational epidemiologic studies, we aim to estimate the magnitude of the effect of physical activity on endometrial cancer risk and to examine the level of evidence, taking into account the methodologic quality of the studies including those on total, leisure time, and occupational activities. Because obesity is a strong risk factor for endometrial cancer, we have paid special attention to a possible role for body weight as a confounder, effect modifier, or intermediary factor in the association between physical activity and endometrial cancer.

Studies were identified through a systematic review of published literature available on PubMed literature databases through December 2006. The databases were searched using the following terms: (“physical activity” OR “physical inactivity” OR “exercise” OR “sedentary lifestyle”) AND (“endometrial cancer” OR “cancer, endometrium” OR “neoplasms, endometrium” OR “cancer, corpus uteri” OR “neoplasms, corpus uteri” OR “cancer, uterus” OR “neoplasms, uterus”). From relevant publications, the bibliographic lists were hand searched for additional articles.

The criteria for inclusion of studies in the review were as follows: case-control or cohort studies investigating the association between physical activity and endometrial cancer; with incidence, prevalence, or mortality as end point; >10 cancer cases included in the analysis; published in English. We included studies assessing leisure time activity or total activity or occupational activity.

The data extraction and study quality assessment were independently done by two reviewers (D.W.V. and E.M.M.). Any disagreement was resolved by consensus or by consultation with a third reviewer (F.E.v.L.). Details of the standardized data extraction are described elsewhere (8). We documented study size, characteristics of the study population, components of physical activity assessment, and methodologic characteristics.

A quality scoring system was developed that captured both generic methodologic issues and issues specific to our subject (see Appendix A; ref. 8). The items of the scoring system were categorized according to three important sources of error in observational studies (the major headings): selection bias, misclassification bias, and confounding bias. Criteria for physical activity assessment methods were partly adopted from Powell et al. (9). The quality scoring system contained 19 items (selection, 5; misclassification, 11; and confounding, 3). The members of the Task Force Physical Activity and Cancer assessed which aspects of quality are more important and therefore should be weighted more in the overall quality score. In this research area, the potential for confounding bias is judged less important. Studies that adequately adjusted for potential confounders show that the magnitude of the risk estimates does not materially change after adjustment, whereas biases due to selection or misclassification are expected to lead to larger effects on risk estimates. Thus, the major headings were weighted 2:2:1. The maximum attainable score is 105 (see Appendix A), and the quality score of the studies is presented as percentage of the maximum attainable score.

We assessed statistical heterogeneity across studies using a formal test (10) and found statistical evidence for heterogeneity for total, leisure time, and occupational activities combined, both in cohort and case-control studies. Summary estimates were calculated using a general variance–based method using confidence intervals (11).

A qualitative summary of all studies was undertaken according to a best evidence synthesis, taking into account the methodologic quality of the studies and the consistency of the available evidence (see Appendix B). An inverse association between physical activity and endometrial cancer risk within a study was defined as a risk estimate for the highest versus the lowest level of activity of <0.8, irrespective of statistical significance. In our best evidence synthesis, the evidence for an inverse association was defined as “strong” if ≥75% of the high-quality studies reported an inverse association. Studies were categorized as high or low quality based on the median quality score of all studies combined. Graphical displays were used to investigate whether the risk estimates from studies with a relatively low methodologic quality differed systematically from those with relatively high quality. Possible publication bias was investigated by funnel plots.

We identified 20 relevant articles on physical activity and endometrial cancer risk. The main characteristics and results of the 7 cohort studies (4, 6, 12-16) and 13 case-control studies (5, 17-28) are summarized in Tables 1 and 2, respectively. All studies were published between 1993 and 2006. Two cohort studies and four case-control studies assessed total activity whereas the other studies reported only leisure time activities and/or occupational activities. The number of endometrial cancer cases ranged from 49 to 5,287 in the cohort studies and from 31 to 832 in the case-control studies. There was considerable variation among the studies with respect to age of the participants, physical activity assessment, and the length of follow-up (cohort design).

The total quality score, as a percentage of the maximum score, ranged between 61% and 87% (median, 71%) for the cohort studies and between 33% and 80% (median, 55%) for the case-control studies. The median quality score of all studies combined was 60.5%. All seven cohort studies and three case-control studies were thus categorized as high-quality studies (i.e., quality score above the median). Especially cohort studies scored very favorably on issues about selection bias (range, 83-100%; median, 100%) as compared with case-control studies (range, 26-90%; median, 71%). Case-control studies scored moderately on misclassification issues (range, 24-83%; median, 57%) and even less favorably on confounding (range, 0-100%; median, 19%). Cohort studies scored similarly, for misclassification between 43% and 78% (median, 60%) and for confounding between 19% and 76% (median, 38%). Among case-control studies, there was a tendency for larger studies (i.e., greater number of endometrial cancer cases) to have a more favorable quality score than smaller studies (Fig. 1).

Four cohort studies reported on leisure time activity, of which three found a decreased risk of endometrial cancer (based on our definition; see Methods) for women in the highest activity versus the lowest activity category (Table 1; Fig. 2; refs. 4, 12, 14). Risk reductions ranged between 20% and 90%, and two of three studies showed evidence for a dose-response effect (trend test, P < 0.01; refs. 4, 12). Terry et al. (12) found that the highest activity level (i.e., “hard physical training”) was associated with a very strong decrease in risk of endometrial cancer compared with having no leisure time activities [hazard ratio (HR), 0.1; 95% confidence interval (95% CI), 0.04-0.6]. However, this was based on only two cases in the most active group. Only one cohort study assessed total activity and found a 20% endometrial cancer risk reduction (HR, 0.8; 95% CI, 0.5-1.1; ref. 13). In two of four studies that assessed occupational activity, a decreased risk of endometrial cancer was found in women in the highest versus the lowest category of occupational activity (e.g., manual/standing work versus sedentary work; refs. 14, 16).

Of the 13 case-control studies, 4 assessed total activity and 9 assessed leisure time activity (3 studies assessed both). Ten case-control studies included some assessment of occupational activities. Of the nine studies assessing leisure time activity, five observed a decreased risk of endometrial cancer (based on our definition; see Methods) for the most active group compared with the least active group (Table 2; Fig. 2; refs. 5, 19-21, 24). Three of these nine studies also showed evidence for a dose-response effect (P values between <0.01 and 0.06 for trend test; refs. 5, 21, 24) and one study reported a significant trend test although no obvious trend could be observed (19). All four studies that assessed total activity found a markedly decreased risk of endometrial cancer (HRs between 0.1 and 0.5) for the most active women compared with the least active women (17-19, 23). None of these studies clearly showed a dose-response effect, although a trend test suggested differently in two studies (trend test, P ≤ 0.01; refs. 19, 23). Six of 10 studies reporting on occupational activity found a decreased risk of endometrial cancer (based on our definition; see Methods; refs. 17, 19, 22, 24, 26, 28). Two of these studies also showed some evidence for a dose-response effect; however, no P values were reported (26, 28). Dosemeci et al. (27) reported on the only study to find a 2-fold increased risk of endometrial cancer for women with active jobs as compared with women with sedentary jobs. The reported odds ratio (OR) was not statistically significant, based on only 31 cases, and the study had an unfavorable quality score.

Based on our best evidence synthesis, there is strong evidence for an inverse association between physical activity and endometrial cancer risk. Eight of 10 high-quality studies reported a risk reduction of >20% for the highest category of total, leisure time, or occupational activity. Seven of eight (88%) high-quality studies reporting on total or leisure time activities found an inverse association with either of these physical activity assessments. Only 3 of 7 (43%) high-quality studies on occupational activity found a reduction in risk for women with active jobs as compared with women with inactive jobs.

The effect estimates of eight case-control studies that investigated leisure time activities and reported 95% CIs were not found to be statistically heterogeneous. Pooling resulted in a summary estimate of a 27% decreased risk of endometrial cancer for the most active women compared with the least active women (summary OR, 0.73; 95% CI, 0.62-0.86). Similarly, effect estimates of eight case-control studies that reported on occupational activities and included 95% CIs were also not found to be statistically heterogeneous (summary OR, 0.80; 95% CI, 0.66-0.96). Pooling of all total, leisure time, and occupational activity effect estimates, irrespective of statistical heterogeneity, suggested an at least 20% decrease in endometrial cancer risk for the most active women compared with the least active women [cohort summary relative risk (RR), 0.77; 95% CI, 0.70-0.85; case-control summary OR, 0.71; 95% CI, 0.63-0.80; Fig. 2].

Four of seven cohort studies assessed whether body mass index (BMI) had a modifying effect on the association between physical activity and endometrial cancer risk and did not find any statistical evidence for interaction (at the multiplicative level; refs. 4, 6, 13, 14). Of all case-control studies, six assessed effect modification by BMI. Four studies found no differences in stratum-specific risk estimates (5, 18, 24, 25). Sturgeon et al. (17) found that the decrease in risk found with leisure time activity was limited to those with a BMI >25. Levi et al. (19) also observed stronger associations with physical activity in those with a high BMI; however, these analyses were not adjusted for potential confounders. Most studies adjusted for BMI as a confounding factor but found only slightly attenuated risk estimates.

Very few studies specifically assessed whether the association between physical activity and endometrial cancer risk differed by menopausal status. Colbert et al. (13) found no significant interaction between menopausal status and total activity. Furberg and Thune (14) stratified by age and found that the effect of physical activity was stronger in women aged ≥50 years (not statistically significant). Of the case-control studies, only two studies examined effect modification by menopausal status. Both Sturgeon et al. (5) and Matthews et al. (17) found no significant difference in effect between premenopausal and postmenopausal women.

The association between endometrial cancer risk and physical activity in different periods of life was only assessed in five case-control studies (5, 18, 19, 21, 24). The association with physical activity levels in recent years of life (above age 50, or years before physical activity assessment) was generally somewhat stronger than the association with physical activity in adolescence or young adulthood (18, 19, 21, 24). None of these studies adjusted for physical activity levels in other periods of life. Matthews et al. (5) specifically studied physical activity patterns in adolescence and adulthood in relation to premenopausal and postmenopausal endometrial cancer risk. They reported that being physically active only in adulthood (i.e., nonactive during adolescence) was associated with a decrease in postmenopausal but not in premenopausal endometrial cancer risk. Higher levels of physical activity in adolescence were associated with a decreased risk of premenopausal endometrial cancer only. However, being physically active in adolescence only (i.e., nonactive during adulthood) was associated with marginally decreased risk of both premenopausal and postmenopausal endometrial cancer (not statistically significant).

We next examined whether methodologic issues can explain heterogeneity in the magnitude of the estimated endometrial cancer risk. Substantial variation between studies was observed in exposure quantification and categorization. The reference category of leisure time activities varied from “none” to 2-3 h of (moderate) activity per week, whereas the highest activity category varied from 2 h of (vigorous) activities per week to >6 or even 9 h of moderate/vigorous activities per week. We did not observe a pattern or trend of studies with higher activity categories or a larger contrast finding stronger associations between physical activity and endometrial cancer risk (data not shown).

Besides these issues relating to exposure assessment, we also examined whether study quality could explain heterogeneity in magnitude of the risk estimates. In Fig. 2 studies are ordered according to total quality score, showing the relation between the total quality score and the magnitude of the estimate of the highest versus the lowest level of activity. Studies with an unfavorable total quality score tended to report more divergent risk estimates, ranging from a >2-fold decreased risk to a 2-fold increased risk. In particular, high variability in the magnitude of the risk estimates was found in case-control studies with a relatively unfavorable “selection bias score” (Fig. 3). In cohort studies, selection bias was not an issue as six of seven cohort studies attained the maximum score. The magnitude of the risk estimates in cohort and case-control studies did not correlate strongly with quality scores for “misclassification” or “confounding” (data not shown).

Figure 4 represents a funnel plot for cohort and case-control studies combined. The funnel plot shows that risk estimates of the larger studies cluster around a risk reduction of 20% to 50%, comparing the lowest and highest levels of physical activity, in both cohort and case-control studies. Studies with relatively small numbers of cases show a larger diversity in risk estimates, both strong risk reductions and some increased risks. These results suggest that no major bias has occurred by selective publication.

This is the first systematic and standardized review of epidemiologic studies of the association between physical activity and endometrial cancer. Considerable variation among studies was observed with respect to design, exposure assessment, and study quality (scores ranging between 33% and 87% of the maximum attainable score). The best evidence synthesis showed that the majority of high-quality studies found strong evidence for an inverse association between physical activity and endometrial cancer risk. The pooled estimate of seven high-quality cohort studies showed a statistically significantly decreased risk of endometrial cancer (OR, 0.77; 95% CI, 0.70-0.85) for the most active women compared with the least active women. We found some evidence that study quality was inversely associated with the estimated magnitude of the risk reduction, particularly with regard to selection bias in case-control studies. Therefore, several more high-quality cohort studies are needed for a definitive conclusion on the association between physical activity and endometrial cancer risk.

Most of the studies included in this review controlled for confounding factors to some extent. On the other hand, neither cohort nor case-control studies scored very favorably on study quality with respect to confounding. The incomplete assessment of confounders does not automatically imply, however, that the results of the included studies are biased. In those studies that adequately adjusted for potential confounders and reported both adjusted and unadjusted risk estimates, the magnitude of the risk estimates did not materially change by adjusting for confounding factors. An important issue with respect to confounding is the role of BMI. Body weight or BMI could be in the causal pathway between physical activity and endometrial cancer risk (i.e., hypothetically, the apparently protective effect from physical activity might be wholly due to a lower body weight in physically active women). If this is the case, adding BMI as a confounding factor to the regression models would strongly attenuate the effect of physical activity. Three of five cohort studies and 6 of 13 case-control studies specifically reported on the effect of adding BMI to the model adjusted for other confounding factors. Most studies found only slightly attenuated risk estimates, providing evidence that physical activity is associated with lower endometrial cancer risk independently of BMI. Furthermore, no convincing evidence was found for differences in the association between physical activity and endometrial cancer risk after stratification for BMI.

This review shows that only few studies assessed total physical activity. All four studies that assessed both total physical activity and leisure time activity found that the association with endometrial cancer risk was stronger for total than for leisure time activity. Overall, the evidence was less consistent for occupational activity than for total and leisure time activities. Most studies on occupational activity used crude methods for exposure assessment (i.e., job title) and a large number of women were not, or only shortly, engaged in paid employment. This may have resulted in errors in the measurement of physical activity and consequently risk estimation for risk of endometrial cancer. Although leisure time activity has become more important in Western countries and may reflect an individual's total physical activity level in most societies, this assumption does not hold for all populations. Therefore, future studies should aim at measuring both total and leisure time activities, and special attention should be paid to household activities.

Several issues have not received sufficient attention in the epidemiologic studies thus far. Some studies have used very rough assessments of physical activity, without specifically taking into account the frequency, duration, and intensity of physical activities, and the different periods in life during which activity patterns may have changed. In addition, the association of physical activity and premenopausal endometrial cancer risk has been insufficiently studied. Future epidemiologic studies will need to address these issues to specify the association between physical activity and endometrial cancer risk.

The mechanisms by which physical activity may protect against endometrial cancer are not fully understood. Obesity is one of the main risk factors for endometrial cancer and is estimated to account for ∼40% of endometrial cancer incidence in Europe (29). Although lack of physical activity is known to be associated with an increased risk of obesity (30, 31), the studies described in this systematic review show that physical activity is associated with a decreased risk of endometrial cancer in both normal weight and obese women. The combined effects of endogenous hormones, such as sex steroid hormones, insulin, and insulin-like growth factor-I, are also known to play an important role in the development of endometrial cancer, and may well relate to the mechanism underlying the effect of physical activity on endometrial cancer risk (32). Many of the effects of known risk factors for endometrial cancer can be explained by the unopposed estrogen hypothesis (2). This hypothesis proposes that endometrial cancer risk is increased in women who have high plasma bioavailable estrogens insufficiently counterbalanced by progesterone, which would result in stimulation of proliferation and induction of genetic damage in endometrial cells (2, 33). Both hormones also affect levels of insulin-like growth factor-I, possibly in opposite directions. Hyperinsulinemia is also associated with increased risk of endometrial cancer (34, 35). Although the effects of physical activity on insulin-like growth factor-I are unclear (36), physical activity is known to decrease the levels of serum estrogens (37) and serum insulin (38).

In conclusion, physical activity seems to be associated with a significant reduction in the risk of endometrial cancer. However, the number of high-quality prospective cohort studies is still limited. Based on 20 observational studies, evidence suggests that physical activity is associated with a 20% to 40% decreased risk of endometrial cancer. Further studies are needed to assess which aspects (i.e., frequency, intensity, duration) of physical activity are most strongly related to the risk of endometrial cancer; which amount of physical activity is necessary; and in which period of life physical activity contributes most to the risk reduction. Both observational epidemiologic studies, preferably prospective in design, and intervention studies should be designed to examine interactive effects of physical activity, diet, and body weight. Intervention studies should help elucidate the causal pathway through which these effects occur.

Methodologic quality score—cohort and case-control studies

(a) At least four studies are necessary to define the evidence as strong, moderate, or no evidence. When less than four studies are available, the evidence will be defined as inconclusive.

(b) A high-quality study was defined as a study with a total quality score above the median quality score of all studies (i.e., >60.5% of the maximal attainable score).

(c) A decreased risk was defined as a risk estimate of <0.80 for the highest versus the lowest level of activity; no association as a risk estimate between 0.8 and 1.25; and an increased risk as a risk estimate >1.25.