The role of body weight in the relationship between physical activity and endometrial cancer: Results from a large cohort of US women
Factors influencing circulating estrogen levels, insulin-mediated pathways or energy balance through obesity-related mechanisms, such as physical activity, have been proposed as potential risk factors for endometrial cancer. We examined measures of physical activity in relation to endometrial cancer risk in the American Cancer Society Cancer Prevention Study II Nutrition Cohort, a prospective study of cancer incidence and mortality, using information obtained at baseline in 1992. From 1992 to 2003, 466 incident endometrial cancers were identified among 42,672 postmenopausal women with intact uteri who were cancer-free at enrollment. Cox proportional hazards modeling was used to compute hazard rate ratios (RR) while adjusting for potential confounders. To assess the role of body mass index (BMI) in this relationship, we computed multivariate RR with and without adjustment for BMI and stratifying by BMI. All measures of physical activity and the avoidance of sedentary behavior were associated with lower endometrial cancer risk. Baseline recreational physical activity was associated with 33% lower risk (RR = 0.67, 95% CI 0.44–1.03 for 31.5+ vs. <7 MET-hr/week, trend p = 0.007) in the multivariate model without BMI. However, the trend was attenuated after further adjustment for BMI (trend p = 0.18). BMI significantly modified the association between physical activity and endometrial cancer risk (heterogeneity of trends p = 0.01). The inverse relationship was seen only among overweight or obese women (trend p = 0.003) and not in normal weight women (trend p = 0.51). In summary, light and moderate physical activity including daily life activities were associated with lower endometrial cancer risk in our study, especially among women who are overweight or obese. © 2008 Wiley-Liss, Inc.
Endometrial cancer is the fourth most common incident cancer among US women.1 Important risk factors for endometrial cancer include obesity, postmenopausal unopposed estrogen use and nulliparity.2 Type II diabetes has also been associated with increased risk.3 Few other risk factors for endometrial cancer have been well-established. Physical activity has been proposed to protect against endometrial cancer. Physical activity influences circulating estrogen levels and insulin-mediated pathways both through its effects on energy balance and adiposity and directly through independent pathways.4–6
To date, 17 observational studies have examined the relationship between recreational (leisure-time) physical activity and endometrial cancer risk (reviewed in Refs. 7–10). Although only half of these studies reached statistical significance in their findings,9–16 the majority suggest a benefit with regular physical activity in lowering endometrial cancer risk. A recent meta-analysis8 provided summary risk estimates of a 27% decreased risk of endometrial cancer from case-control studies (95% CI, 0.62–0.86) and a 23% decreased risk from cohort studies (95% CI, 0.70–0.85) when comparing the most active women with the least active women.
Researchers have also examined the association between endometrial cancer and nonrecreational activities in daily life, such as household chores, shopping and gardening. These activities are usually less intense than the recreational activities generally recommended for chronic disease prevention, but are as or more commonly done. Whether these activities have any potential health benefits is unknown. Five previous studies (reviewed in Ref. 7) have examined the role of these household activities; 4 of these reported a significant inverse relationship between the highest levels of household activity and risk of endometrial cancer.9, 11, 17, 18 Two other studies reported an increased risk of endometrial cancer in sedentary women.11, 19
Another important unresolved question is whether body weight confounds, modifies or is an intermediary in the relationship between physical activity and endometrial cancer risk. Most previous studies have adjusted for measures of body mass in multivariate models (reviewed in Ref. 8). Body mass index (BMI) attenuates the relationship between physical activity and risk when added to multivariate models. Eleven previous studies also examined whether body weight is an effect modifier of the relationship between physical activity and endometrial cancer.9, 11, 12, 15, 17–23 Most of these studies found no statistical interaction between BMI and physical activity on the multiplicative scale; however, 3 studies reported significantly lower relative risk estimates associated with regular physical activity among overweight or obese women than normal weight women.17, 18, 21
To further clarify the relationship between physical activity and risk of endometrial cancer, we examined whether recreational physical activity, nonrecreational household activities or sedentary behavior was associated with endometrial cancer risk, and whether these associations differed by body weight among postmenopausal women in the American Cancer Society Cancer Prevention Study II (CPS-II) Nutrition Cohort, a large prospective study in the US.
Material and methods
Women in this analysis were drawn from the 97,786 female participants in the CPS-II Nutrition Cohort, a prospective study of cancer incidence and mortality established by the American Cancer Society in 1992 as a subgroup of the larger 1982 CPS-II baseline mortality cohort.24 Most participants were aged 50–74 years at enrollment in 1992. At baseline, they completed a 10-page self-administered questionnaire that included questions on demographic, reproductive, medical, behavioral, environmental and dietary factors. Beginning in 1997, follow-up questionnaires were sent to cohort members every 2 years to update exposure information and to ascertain newly diagnosed cancers. All follow-up questionnaire response rates (after multiple mailings) among living cohort members are at least 88%. End of follow-up for the present analysis was June 30, 2003.
We excluded from this analysis 3,190 women who were lost to follow-up (i.e., alive at the first follow-up questionnaire in 1997 but did not return the 1997 or any subsequent follow-up questionnaire), who reported prevalent cancer (except nonmelanoma skin cancer) at baseline (N = 12,053), who reported not being postmenopausal (N = 4,291) or who had a hysterectomy or unknown hysterectomy status at baseline (N = 30,724). We also excluded women with missing information on recreational physical activity at baseline (N = 640) or BMI at baseline (N = 756). To eliminate the strong effect of estrogen-only hormone replacement therapy (ERT) on endometrial cancer and the possibility that reports of ERT only use were in error (because this regimen is contraindicated in women with an intact uterus), we excluded women who reported current ERT use in 1992 (N = 1,583) and those with current or past postmenopausal hormone use of unknown type in 1992 (n = 1,812). Finally, we excluded reported cases of endometrial cancer that could not be verified through medical or cancer registry records (N = 33) or cases with missing or potentially unrelated histologies such as endometrial stromal sarcoma (histology code 8930), adenosarcoma (code 8933), Mullerian mixed tumor (code 8950), endometrial adenofibroma (code 8381) and carcinosarcoma (code 8980) (n = 32). The etiology of these less common tumors is thought to be different from endometrial carcinomas. Women who did not return a 1999, 2001 or 2003 questionnaire were censored at the return of their last questionnaire. Women who underwent a hysterectomy were censored when first reported on the 1997, 1999 or 2001 questionnaire. After all exclusions, the final analytic cohort consisted of 42,672 women with a mean age of 62.8 (± 6.0 SD) years when enrolled in the study.
This analysis included 466 verified incident cases of endometrial cancer diagnosed between the date of enrollment and June 30, 2003. Of these, 433 cases were identified initially by self-report on a follow-up questionnaire and subsequently verified from medical records (n = 326) or linkage with state cancer registries (n = 107). A previous study linking cohort participants with state cancer registries has shown that the Nutrition Cohort participants are highly accurate (93% sensitivity) in reporting any past cancer diagnoses.25 An additional 9 cases were reported by participants as another type of cancer, but were found to be endometrial cancer upon examination of registry records. Lastly, 24 incident cases were initially identified as interval deaths (deaths that occurred between baseline in 1992 and the end of follow-up in 2003) through automated linkage of the entire cohort with the National Death Index,26 and subsequently verified through linkage with state cancer registries.
Measures of physical activity and sedentary behavior
Baseline information on recreational physical activity was collected using the question “During the past year, what was the average time per week you spent at the following kinds of activities: walking, jogging/running, lap swimming, tennis or racquetball, bicycling or stationary biking, aerobics/calisthenics and dancing?” Response to each activity included “none,” “1–3 hr per week,” “4–6 hr per week” or “7+ hr per week.” Summary MET-hr/week were calculated for each participant. A MET, or metabolic equivalent, is the ratio of metabolic rate during a specific activity to resting metabolic rate.27 Because of the older age of this population, the summary MET score for each participant was calculated by multiplying the lowest number of hours within each category times the moderate intensity MET score for each activity according to the Compendium of Physical Activities27 to provide conservatively estimated summary measures. The MET scores for various activities were27: 3.5 for walking, 7.0 for jogging/running, 7.0 for lap swimming, 6.0 for tennis or racquetball, 4.0 for bicycling/stationary biking, 4.5 for aerobics/calisthenics and 3.5 for dancing. Recreational physical activity at baseline was categorized in MET-hr/week as none, >0–<7, 7–<17.5, 17.5–<31.5 or ≥31.5. For reference, 31.5 MET-hr/week corresponds to approximately 1 hr of moderate-paced walking (3.0 mph) per day.
In addition to recreational leisure activity at baseline, nonrecreational leisure activity was also examined based on information collected from the question “During the past year, what was the average time per week you spent at the following kinds of activities: gardening/mowing/planting, heavy housework/vacuuming, heavy home repair/painting and shopping?” The above algorithm was used to calculate MET-hr/week using the following values for each activity27: 3.0 for gardening/mowing/planting, 2.5 for heavy housework/vacuuming, 3.0 for heavy home repair/painting and 2.5 for shopping. Baseline nonrecreational leisure activity was categorized in quartiles of MET-hr/week as none, >0–5.0, >5.0–<10.0, 10.0–<18.5 or ≥18.5.
For both recreational and nonrecreational physical activity, women who reported being inactive were not used as the referent group because of the possibility that their complete inactivity may be due to underlying conditions related in some way to endometrial cancer risk. If inactive women suffer from other health conditions that are hormone-related and impair their ability to engage in physical activity (such as severe osteoporosis), the association between inactivity and endometrial cancer risk may be confounded.
We assessed sedentary behavior based on the question “During the past year, on an average day, (not counting time spent at your job) how many hours per day did you spend sitting (watching TV, reading, etc.)?” Responses included “none, less than 3, 3–5, 6–8, more than 8 hr per day.” Sedentary behavior at baseline was categorized as 0–<3, 3–5, ≥6 or missing hr/day.
The baseline questionnaire also asked participants to recall physical activity at age 40 using the question, “At age 40, what was the average time per week you spent at the following kinds of activities: walking, jogging/running, lap swimming, tennis or racquetball, bicycling or stationary biking, aerobics/calisthenics and dancing?” A summary MET score at age 40 was created using the same method for baseline recreational activity described above. Recreational physical activity at age 40 was categorized in MET-hr/week as none, >0–<7, 7–<17.5, 17.5–<31.5, 31.5–<42.0 or ≥42.0. Another measure of past physical activity was obtained from a questionnaire completed in 1982 when participants in the CPS-II Nutrition Cohort were enrolled in the larger CPS-II mortality study. The 1982 questionnaire asked “How much exercise do you get (work or play)?” with possible responses: “none, slight, moderate or heavy.” Although crude, this measure of physical activity has been shown to correlate with all-cause mortality rates.28 The self-reported activity level in 1982 was combined with the more detailed information on 1992 recreational physical activity to examine whether risk of endometrial cancer was reduced among women who consistently reported being physically active in both 1982 and 1992. Women who reported being “none or slight” in 1982 and <17.5 MET-hr/week in 1992 were categorized as “consistently low,” those reporting “moderate or heavy” in 1982 and 17.5+ MET-hr/week in 1992 were categorized as “consistently high,” those reporting “none or slight” in 1982 and 17.5+ MET-hr/week in 1992 were categorized as “increasing low to high” and those reporting “moderate or heavy” in 1982 and <17.5 MET-hr/week in 1992 were classified as “decreasing high to low.”
We used Cox proportional hazards modeling29 to calculate hazards rate ratios (RR) and corresponding 95% confidence intervals (CI) to examine the relationship between measures of physical activity (recreational and nonrecreational), sedentary behavior and endometrial cancer risk. Statistical Analysis Software (SAS), v 9.1 was used for all analyses. For each exposure variable, we assessed risk in 3 models, one adjusted only for age, the second adjusted for age and other potential confounding factors except BMI and the third adjusting for all potential confounding factors including BMI. All Cox models were stratified on exact year of age at enrollment, and follow-up time in days was used as the time-axis. We tested the Cox proportional hazards assumption for each exposure measure and found no violations. Potential confounders included in the multivariate models were BMI [weight (kg)/height (m2)] (<25.0, 25.0–<27.5, 27.5–<30.0, ≥ 30.0), oral contraceptive use (never, <5 years, 5+ years, ever use with unknown duration, missing), parity (nulliparous, 1–2, 3+, missing), age at menopause (<45, 45–49, 50–54, 55+, unknown), age at menarche (<12, 12+, missing), postmenopausal hormone therapy use (HT) (never, current, former, other, missing/unknown), personal history of diabetes (yes, no), smoking status (never, current, former, ever unknown status, missing) and total energy intake (in quartiles). HT use and history of diabetes were modeled as time-varying covariates using information obtained in 1992, 1997, 1999 and 2001.
Trend tests for baseline recreational and nonrecreational activity, physical activity at age 40 and duration of sedentary behavior were calculated by assigning the median value within each category to that category. Trend tests for physical activity in 1982 were obtained by using an ordinal variable corresponding to each level of physical activity. To test whether physical activity across multiple time points was associated with endometrial cancer risk, we combined baseline recreational physical activity with physical activity in 1982 as an index of consistency in the 10-years prior tobaseline. To test whether any of the potential confounders described above modified the association between the main effect measures and endometrial cancer risk, we examined each factor in a separate model by constructing multiplicative interaction terms with each risk factor and comparing the interaction model to the base model without the interaction terms. Because of small numbers in some strata, categories of potential effect modifiers were sometimes combined. Statistical interaction was assessed in multivariate models using the likelihood ratio test and a p-value <0.05 was considered statistically significant.30
Approximately 9% (n = 3,854) of women reported no recreational physical activity at baseline (Table I). Among physically active women (defined as those reporting any recreational physical activity at baseline), the median MET expenditure was 8.0 MET-hr/week, corresponding to approximately 2 hr of moderately paced walking per week. Physically active women, regardless of level of energy expenditure, engaged primarily in activities judged to be of low to moderate intensity (walking, biking, aerobics or dancing) rather than higher intensity (jogging/running, swimming, tennis/racquetball). Physically active women were more likely to be lean and have ever used oral contraceptives and postmenopausal HT. Physically active women were also more likely to have been physically active in the past (both as measured in 1982 and recalled from age 40) and to engage in various household (nonrecreational) activities (Table I).
Table I. Selected Study Participant Characteristics1 in Relation to Recreational Physical Activity at Baseline Among 42,672 Women in the CPS-II Nutrition Cohort, 1992–2001
|Median recreational activity MET-hr/week||0||3.5||11.5||24.0||39.5|
|Median nonrecreational MET-hr/week||8.0||8.0||8.0||12.5||13.0|
|Median recreational MET-hr/week, age 40||3.5||3.5||9.5||18.0||28.5|
|Moderate or high exercise in 1982 (%)||56.9||65.8||73.6||81.5||88.4|
|Median hr/day spent sedentary||4.0||4.0||4.0||4.0||1.5|
|BMI1 (mean ± SE)||26.9 ± 0.08||25.9 ± 0.04||25.2 ± 0.04||24.7 ± 0.05||24.1 ± 0.09|
|Age at menopause1 (mean ± SE)||50.0 ± 0.09||50.4 ± 0.05||50.5 ± 0.04||50.5 ± 0.06||50.6 ± 0.10|
|Age at menarche1 (mean ± SE)||12.8 ± 0.02||12.8 ± 0.01||12.8 ± 0.01||12.8 ± 0.02||12.8 ± 0.03|
|Estimated kcal/day1 (SE)||1,374 (7.9)||1,358 (4.2)||1,353 (4.1)||1,370 (5.4)||1,411 (9.0)|
|Race1 (% White)||97.2||97.8||97.8||97.7||97.7|
|Parity1 (%)|| || || || || |
|Oral contraceptive use1 (%)|| || || || || |
| Never use||65.6||64.2||62.2||64.0||61.4|
| <5 years||17.0||17.7||18.7||17.8||19.0|
| 5+ years||14.0||15.6||16.7||15.8||17.5|
| Ever use/years unknown||1.7||1.3||1.2||1.2||1.2|
|Ever use of HT1 (%)||36.3||41.3||43.2||42.2||43.4|
|History of diabetes1 (%)||7.6||5.9||5.3||5.1||4.6|
Every measure of baseline physical activity and the avoidance of sedentary behavior were associated with lower endometrial cancer risk (Table II). Recreational physical activity at baseline was associated with a 33% lower risk (RR = 0.67, 95% CI 0.44–1.03 for 31.5+ vs. <7 MET-hr/week, trend p = 0.007) in the multivariate model without adjustment for BMI (Table II). However, the association was attenuated when BMI was added to the model (RR = 0.79, 95% CI 0.52–1.22 for 31.5+ vs. <7 MET-hr/week, trend p = 0.18). Similarly, baseline household activity was marginally associated with lower endometrial cancer risk (RR = 0.79, 95% CI, 0.61–1.03 for >18.5 MET-hr/week vs. >0–<5 MET-hr/week; trend p = 0.07); further adjustment for BMI only slightly attenuated this association (Table II). Finally, adjustment for BMI greatly influenced the relationship between endometrial cancer and sedentary behavior at baseline (6+ vs. <3 hr/day sitting RR = 1.40, 95% CI 1.03–1.89 without adjustment for BMI versus RR = 1.18, 95% CI 0.87–1.59 with adjustment for BMI).
Table II. Hazard Rate Ratios (RR) and 95% Confidence Intervals (CI) for Measures of Physical Activity at Various Points in Time and Endometrial Cancer, CPS-II Nutrition Cohort, 1992–2001
|Baseline recreational activity MET-hr/week|| || || || || |
| None||43||34,622||0.92 (0.66–1.28)||0.93 (0.67–1.30)||0.84 (0.60–1.17)|
| 0<–<7||170||124,302||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)|
| 7–<17.5||157||133,553||0.86 (0.69–1.07)||0.86 (0.70–1.07)||0.92 (0.74–1.14)|
| 17.5–<31.5||72||75,456||0.70 (0.53–0.92)||0.70 (0.53–0.93)||0.80 (0.60–1.05)|
| 31.5+||24||27,264||0.65 (0.42–1.00)||0.67 (0.44–1.03)||0.79 (0.52–1.22)|
| || || || ||p-trend = 0.007||p-trend = 0.18|
|Baseline nonrecreational activity MET-hr/week4|| || || || || |
| None||11||5,473||1.45 (0.78–2.69)||1.44 (0.78–2.66)||1.31 (0.71–2.44)|
| >0–5.0||132||96,620||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)|
| >5.0–<10.0||110||93,290||0.87 (0.67–1.11)||0.87 (0.67–1.12)||0.91 (0.70–1.17)|
| 10.0–<18.5||103||96,592||0.77 (0.60–1.00)||0.77 (0.60–1.00)||0.79 (0.61–1.02)|
| 18.5+||106||99,250||0.78 (0.60–1.01)||0.79 (0.61–1.03)||0.83 (0.64–1.07)|
| || || || ||p-trend = 0.07||p-trend = 0.13|
|Baseline sitting4 (hr/day)|| || || || || |
| <3||195||184,173||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)|
| 3–5||203||165,561||1.13 (0.93–1.38)||1.13 (0.92–1.38)||1.02 (0.83–1.25)|
| 6+||56||36,584||1.40 (1.04–1.89)||1.40 (1.03–1.89)||1.18 (0.87–1.59)|
| || || || ||p-trend = 0.05||p-trend = 0.41|
|Exercise in 19824|| || || || || |
| None/Slight||156||105,260||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)|
| Moderate||285||264,391||0.71 (0.58–0.86)||0.74 (0.60–0.90)||0.83 (0.68–1.02)|
| Heavy||20||20,608||0.64 (0.40–1.02)||0.67 (0.42–1.07)||0.80 (0.50–1.28)|
| || || || ||p-trend = 0.003||p-trend = 0.08|
|Recreational activity MET-hr/week at age 404|| || || || || |
| None||64||57,878||0.91 (0.67–1.23)||0.93 (0.68–1.26)||0.92 (0.68–1.25)|
| 0<–<7||119||97,393||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)|
| 7–<17.5||146||115,712||1.02 (0.80–1.30)||1.02 (0.80–1.30)||1.03 (0.81–1.32)|
| 17.5–<31.5||78||74,574||0.84 (0.63–1.12)||0.83 (0.63–1.11)||0.81 (0.61–1.08)|
| 31.5–<42||28||20,991||1.09 (0.73–1.65)||1.11 (0.73–1.68)||1.11 (0.73–1.68)|
| 42+||25||22,236||0.92 (0.60–1.42)||0.94 (0.61–1.45)||0.96 (0.63–1.49)|
| || || || ||p-trend = 0.74||p-trend = 0.72|
|Long-term exercise4,5|| || || || || |
| None/consistently low||142||90,185||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)|
| Low 1982, High 1992||14||15,075||0.58 (0.34–1.01)||0.58 (0.34–1.01)||0.65 (0.37–1.13)|
| High 1982, Low 1992||229||203,277||0.69 (0.56–0.86)||0.72 (0.58–0.89)||0.81 (0.65–1.00)|
| Consistently high||76||81,722||0.58 (0.44–0.76)||0.61 (0.46–0.80)||0.75 (0.56–0.99)|
With regard to past measures of physical activity, prospectively reported exercise in 1982 showed a similar reduction in risk of endometrial cancer as baseline recreational physical activity, before adjusting for BMI (RR = 0.67, 95% CI 0.42–1.07 heavy vs. none/slight). Recreational physical activity at age 40 (as recalled in 1992) was not associated with endometrial cancer risk (Table II). There were no appreciable changes in any risk estimates when simultaneously adjusting for recreational physical activity, nonrecreational activity or sedentary behavior in multivariate models (data not shown). Consistently high levels of physical activity over the 10-years prior to baseline (1982 and 1992) were associated with lower risk of endometrial cancer even after adjustment for BMI (RR = 0.75, 95% CI, 0.56–0.99 for consistently high vs. consistently low activity levels) (Table II).
When we tested for potential effect modification by other endometrial cancer risk factors, we found evidence of effect modification by BMI (heterogeneity of trends p = 0.01). In normal weight women (BMI < 25.0), we observed no association between recreational physical activity and endometrial cancer risk (RR = 1.01, 95% CI, 0.69–1.48 for >17.5 MET-hr/week vs. <7 MET-hr/week; trend p = 0.51), whereas risk of endometrial cancer was significantly lower in active women who were overweight or obese (BMI ≥ 25.0) than in inactive women (RR = 0.59, 95% CI, 0.42–0.83; trend p = 0.003) (Table III). We examined the associations while adjusting more finely for BMI and did not observe any appreciable differences in results (data not shown). Thus, in an effort to provide the most stable risk estimates, without compromising quality, we collapsed BMI categories. Results for overweight and obese women also did not differ appreciably and were therefore combined to provide more stable risk estimates. We also examined effect modification by BMI with both nonrecreational activity and sedentary behavior; however, did not observe any statistically significant interactions (data not shown). This was likely due to our limited power to examine associations in the highest levels of these exposures when stratifying by BMI. We found no suggestion of interactions between measures of physical activity or sedentary behavior and any of the other potential risk factors included in this analysis (data not shown).
Table III. Hazard Rate Ratios (RR) and 95% Confidence Intervals (CI) for Baseline Recreational Physical Activity Among Normal and Overweight/Obese Women in Relation to Endometrial Cancer, CPS-II Nutrition Cohort, 1992–2001
|BMI < 25.0|| || || || |
| RR1 (95% CI)||0.52 (0.22–1.21)||1.00 (ref.)||0.94 (0.64–1.36)||1.01 (0.69–1.48)|
| ||trend p = 0.51|
|BMI 25+|| || || || |
| RR1 (95% CI)||0.99 (0.68–1.43)||1.00 (ref.)||0.88 (0.67–1.15)||0.59 (0.42–0.83)|
| ||trend p = 0.003|
| ||Heterogeneity of trends p = 0.01|
Results from this prospective study provide further support for a role of recent light and moderate physical activity (recreational and household) in lowering risk of endometrial cancer. Previous studies have reported a decrease in risk among the most physically active study subjects ranging from 20 to 90%, with a recent meta-analysis showing a pooled average of approximately 25% reduction in risk.8 Our results are consistent with this pooled estimate for recreational physical activity. Additionally, our results agree with 4 of 5 previous studies that have shown associations between endometrial cancer and nonrecreational (daily life, household) activity9, 11, 17, 18 as well as sedentary behavior.11, 19
There is strong biologic rationale to support the role of physical activity in lowering endometrial cancer risk. Exposure to unopposed estrogen is the major determinant of endometrial carcinogenesis,2, 31 and physical activity has been shown to decrease postmenopausal estrogen levels directly or indirectly through reducing peripheral fat stores, the major source of postmenopausal estrogen synthesis.32–35 Hyperinsulinemia has also been implicated in endometrial carcinogenesis through several proposed mechanisms (reviewed in Ref. 2). Higher levels of insulin are associated with decreased levels of SHBG, resulting in increased levels of free estradiol.36 Insulin may also act directly on endometrial tissue as a mitogenic growth factor, and may downregulate IGFBP-1 leading to a greater bioavailability of free IGF-1.2 Independent of its effects on body mass, physical activity increases insulin sensitivity and decreases plasma insulin levels in postmenopausal women who engage in low to moderate levels of activity.37
Excess weight is a strong risk factor for endometrial cancer and is also associated with sedentary behavior. Therefore, we assessed whether the relationship between physical activity and endometrial cancer was confounded or modified by BMI. Adjustment for BMI in the multivariate analyses greatly attenuated the association between physical activity and endometrial cancer risk. However, a modest inverse association between physical activity and endometrial cancer remained even after adjustment for BMI. This suggests that physical activity has an effect on endometrial cancer that is not entirely mediated by BMI.
Furthermore, physical activity was strongly associated with lower risk of endometrial cancer only among overweight and obese women in our study. Our findings are consistent with 317, 18, 21 of 11 previous studies9, 11, 12, 15, 17–23 that reported a greater risk reduction with regular physical activity among overweight or obese women compared to normal weight women. It is unclear whether the other studies observed interactions on less than a multiplicative scale, i.e. an additive scale, or were not adequately powered to detect an interaction based on data provided. Since physical activity, even in the absence of weight loss, significantly improves insulin sensitivity and has direct effects on bioavailable estrogen,37 it is biologically plausible that overweight or obese women engaging in regular physical activity may experience a greater risk reduction compared to active, normal weight women.
Our study has several limitations. We have no information on the intensity with which individuals engage in each behavior thus increasing the likelihood of misclassifying true energy expenditure. While the physical activity questions we used have not been validated and are subject to misreporting, they are very similar to those used and validated in another prospective study.38 Wolf et al. found strong correlations between activity reported on past-week activity recalls and 7-day diaries and that reported on the questionnaire (0.79 and 0.62, respectively).38 Despite the limitations in our physical activity measures, these measures have also been associated with lower risk of breast and colon cancer in this cohort.39, 40 We had limited statistical power to examine higher intensity activities since most highly active women engaged in walking with the addition of modest amounts of the other 6 reportable activities.
Strengths of our study include the prospective design which eliminated the possibility of recall bias and our ability to control for potential confounding by known endometrial cancer risk factors. The relatively homogenous characteristics of women in our study reduced the likelihood of residual confounding by unknown factors even though it also reduced the range of the physical activity exposure variables.
In summary, our results add to the growing body of evidence that light and moderate levels of physical activity, including daily life activities like household chores, may reduce the risk of endometrial cancer, especially among overweight and obese women. Our study also suggests that in addition to its effects mediated through BMI, physical activity may have an independent effect on lowering risk of endometrial cancer possibly through directly suppressing estrogen or increasing insulin sensitivity. Future studies should further examine the association between light-intensity activities and endometrial cancer risk to strengthen public health recommendations in this regard.