Bardia A, Hartmann LC, Vachon CM, Vierkant RA, Wang AH, Olson JE, Sellers TA, Cerhan JR. Recreational Physical Activity and Risk of Postmenopausal Breast Cancer Based on Hormone Receptor Status. Arch Intern Med. 2006;166(22):2478-2483. doi:10.1001/archinte.166.22.2478
Copyright 2006 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2006
Physical activity is a potentially modifiable breast cancer risk factor. There is considerable recent evidence to suggest that risk factors for breast cancer differ based on its subtype, particularly estrogen receptor (ER)/progesterone receptor (PR) status, but this has been less well studied for physical activity. The objective of this study was to examine the association of physical activity with breast cancer incidence based on ER/PR status of the tumor.
The Iowa Women's Health Study is a prospective cohort study of 41 836 postmenopausal women. Recreational physical activity was self-reported on the baseline questionnaire, and 3 levels (high, medium, and low) were defined. Breast cancer incidence and ER/PR status, through 18 years of follow-up, were ascertained by linkage with the Iowa Surveillance, Epidemiology, and End Results Cancer Registry. Cox proportional hazards models were used to estimate multivariate relative risks (RRs) and 95% confidence intervals (CIs) of breast cancer, adjusting for other breast cancer risk factors.
During 554 819 person-years of follow-up, 2548 incident cases of breast cancer were observed. Compared with low physical activity, high physical activity levels were inversely associated with risk of breast cancer (RR, 0.86; 95% CI, 0.78-0.96), and there were inverse associations for ER-positive (ER+)/PR-positive (RR, 0.87; 95% CI, 0.75-1.00), ER+/PR-negative (PR−) (RR, 0.67; 95% CI, 0.47-0.96), and ER-negative/PR− (RR, 0.80; 95% CI, 0.56-1.14) tumors. Further adjustment for body mass index attenuated the overall association with breast cancer (RR, 0.91; 95% CI, 0.82-1.01) and for ER+/PR-positive tumors (RR, 0.94; 95% CI, 0.81-1.08), while there was no change for ER+/PR− tumors (RR, 0.66; 95% CI, 0.46-0.94).
Higher recreational physical activity might reduce the risk of postmenopausal breast cancer overall. Risk reduction varies by ER/PR status of the tumor, being most marked for ER+/PR− tumors, which, in general, have been associated with a clinically more aggressive tumor phenotype. If confirmed in additional studies, these results would suggest that additional mechanisms, besides an effect on body mass, may account for observed protective effects of physical activity in reducing breast cancer.
Breast cancer is the most common noncutaneous cancer and is the second leading cause of cancer-related death among women in the United States.1,2 The incidence of breast cancer is increasing worldwide, and this seems to be partially related to an increase in some lifestyle risk factors for breast cancer, including obesity.3- 9
Physical activity is a potentially modifiable breast cancer risk factor. Increased physical activity has been associated with a decreased risk of breast cancer among premenopausal and postmenopausal women.10- 25 However, a few studies have reported no benefit26- 31 or an increased risk.32 It is suggested that breast cancer may be biologically heterogeneous and, therefore, risk factor associations may differ based on the tumor characteristics, including estrogen receptor (ER) and progesterone receptor (PR) status.5,33- 48 Recent studies have found that various well-established risk factors for breast cancer vary by the ER/PR profile of the tumor, including age,34 menopausal status,34,37 parity,33,38- 40 age at menarche,38,41 age at first pregnancy,34,37,39,40 hormonal use,33,34,42- 44 family history,42,45 body mass index (BMI),33,34,38,42,46 waist-hip ratio,47 alcohol consumption,35,42,48 dietary fat intake,49 and folate level.5,48 However, only 2 recent case-control studies46,50 and 1 cohort study16 have evaluated the association of physical activity with postmenopausal breast cancer incidence defined by joint ER/PR status, and the single cohort study was small (411 cases) and only reported results for ER-positive (ER+)/PR-positive (PR+) tumors.16
The Iowa Women's Health Study is a large prospective cohort of postmenopausal women. A nonsignificant inverse association of physical activity with breast cancer was previously reported after 10 years of follow-up (adjusted relative risk [RR], 0.97; 95% confidence interval [CI], 0.87-1.08).31 We present the updated results of this previous study after an additional 8 years of follow-up, along with results stratified by ER/PR status of the tumor.
The Iowa Women's Health Study is a prospective cohort study of 41 836 women aged 55 to 69 years initiated in 1986, and details about characteristics of the cohort have been described previously.51,52 Briefly, a 16-page questionnaire was mailed to 99 826 randomly selected women and returned by 41 836 women (41.9% response rate), with follow-up questionnaires mailed in 1987, 1989, 1992, 1997, and 2002.
Physical activity during “free time” was ascertained using 2 questions that assessed frequency (“rarely or never,” “a few times a year,” “a few times a month,” “about once a week,” “two to four times a week,” or “more than four times a week”) and type of physical activity in the form of moderate activity (such as bowling, golf, light sports, gardening, or taking long walks) and of vigorous activity (such as jogging, racket sports, swimming, aerobics, or strenuous sports). This level of moderate physical activity generally requires fewer than 6 metabolic equivalent (resting metabolic rate or metabolic equivalent of energy expenditure) hours per week, and vigorous physical activity generally requires 6 or more metabolic equivalent hours per week.53
Based on these questions, 3 levels of physical activity were derived. High physical activity was defined as participation in vigorous activity 2 or more times per week or moderate activity more than 4 times per week. Medium physical activity was defined as participation in vigorous activity once per week or moderate activity 1 to 4 times per week. Low physical activity composed the rest of the cohort. This physical activity index has been indirectly validated in the cohort by the observation that BMI and mean energy intake vary by level of physical activity, and higher physical activity is inversely associated with overall and cardiovascular disease–related mortality.54,55
Incident cases of breast cancer, including ER/PR status, were identified through 2003 using the Iowa Cancer Registry, a member of the National Cancer Institute's Surveillance, Epidemiology, and End Results program.2 Each year, registry cases and cohort members were matched against registry files on name, maiden name, ZIP code, birth date, and social security number. Deaths were identified through annual linkage to Iowa death certificates, supplemented by linkage to the National Death Index.
Women who at baseline were premenopausal (n = 569), had cancer other than skin cancer (n = 3830), had undergone a total or partial mastectomy (n = 1884), or did not have data about physical activity (n = 742) were excluded from the analysis, leaving 36 363 participants for this analysis (exclusions are not mutually exclusive).
Follow-up for incident events was calculated as the time from completion of the baseline questionnaire until the date of breast cancer diagnosis, date of move from Iowa, or date of death. Cox proportional hazards regression analysis was used to estimate RRs and 95% CIs of the association of physical activity with breast cancer risk based on ER/PR tumor subtype, controlling for potential confounding factors as outlined in Table 1. Because one of the mechanisms by which physical activity is postulated to decrease breast cancer risk is through decreasing adiposity,56- 58 we modeled the association with and without controlling for obesity (BMI and BMI at the age of 18 years). Incidence was modeled as a function of age, because age is a better predictor of breast cancer risk in this cohort than length of follow-up.59 Tests for trend were calculated using an ordinal variable for level of physical activity, and including it in the Cox proportional hazards model as a linear variable. In the receptor-specific analyses, events not of that specific cancer type were considered censored observations. Population-attributable risk estimates were calculated based on the coefficients generated by the Cox proportional hazards models and the distribution of physical activity in the cohort.60 Confidence intervals were generated using the bootstrap resampling method.61 All statistical tests were 2-sided, and all analyses were performed using SAS statistical software (SAS Institute Inc, Cary, NC) and Splus software (Insightful, Inc, Seattle, Wash).
There were 36 363 postmenopausal women aged 55 to 69 years at baseline in the analytic data set, and 99.2% were white. Compared with women with a low physical activity index, women with a high physical activity index had a lower BMI. They were also slightly more likely to be higher educated, to have a later age at first live birth, to consume alcohol, and to not smoke (Table 1). All other risk factors showed negligible differences with level of physical activity.
During 554 819 person-years of follow-up, 2548 incident cases of breast cancer were observed in the postmenopausal cohort. The mean age at diagnosis of breast cancer was 71.4 years. Estrogen receptor/PR status was available for 73.0% of the cases and of those with receptor status available, most were ER+/PR+ (71.1%), followed by ER+/PR negative (PR−) (13.5%), ER negative (ER−)/PR− (13.1%), and ER−/PR+ (2.3%). The availability of ER or PR was related to stage (highest for local and regional cases and lowest for in situ), but, of those cases with ER or PR, the percentage positive did not vary strongly with stage (eg, 73.1% of in situ, 85.9% of local, 82.7% of regional, and 75.0% of distant cases were ER+).
Compared with women with low physical activity levels, women with high physical activity levels had a 14% decreased risk of breast cancer after adjusting for major breast cancer risk factors, with the exception of the BMI variables (Table 2). This inverse association was strongest for ER+ and PR− tumors, while there were weaker inverse associations for ER− and PR+ tumors. Cross classification by receptor status showed an inverse relationship of physical activity with each breast cancer subtype except ER−/PR+, which showed a statistically nonsignificant positive association (P=.56); the fewer cases in this subtype limits interpretation of this result. The strongest association was seen for ER+/PR− tumors.
Next, we further adjusted the RRs for BMI at study baseline (1986) and for BMI at the age of 18 years (Table 2). The inverse association of physical activity attenuated for breast cancer overall and for ER+, PR+, and ER+/PR+ tumors. In contrast, the results changed little for ER−, PR−, and ER+/PR− tumors. Further adjustment for waist-hip ratio did not alter these results (data not shown).
There were no significant interactions between previously reported common effect modifiers for breast cancer48,62 (BMI [P=.10], waist-hip ratio [P=.43], family history of breast cancer [P=.87], and smoking [P=.37]) and physical activity on incidence of breast cancer, either overall or by ER/PR status (results not shown).
Population-attributable risks (given in percentages) were calculated to estimate the potential public health significance if physical activity levels were increased to the high level of the physical activity index. Multivariate-adjusted population-attributable risks (not including BMI and BMI at the age of 18 years) were 10.9 (95% CI, 3.9-18.0) for all breast cancer, 7.9 (95% CI, 0-17.0) for ER+/PR+ tumors, 28.4 (95% CI, 7.4-49.4) for ER+/PR− tumors, and 21.9 (95% CI, 0-42.7) for ER−/PR− tumors; the latter estimate included no effect because the RR estimate for ER−/PR− tumors was not statistically significant.
This is the first large prospective cohort study, to our knowledge, to comprehensively report the association of physical activity on incidence of breast cancer by ER/PR status among postmenopausal women. We found that higher recreational physical activity was associated with about 14% decreased risk of breast cancer. The inverse association seemed to vary by ER/PR status and was most marked among ER+/PR− tumors (33% lower risk). Adjustment for BMI attenuated associations for all breast cancer and for ER+/PR+ tumors, but not for ER+/PR− tumors.
The study findings of a 14% lower risk of breast cancer among the most physically active women is within the range reported by most case-control and cohort studies10- 25 involving postmenopausal women. A recent report25 from the Women's Health Initiative Cohort Study involving 74 171 women reported a 14% lower risk of breast cancer for women engaged in regular strenuous physical activity at the age of 35 years (95% CI, 0.78-0.95) and an 8% lower risk for women engaged in regular physical activity at the age of 50 years (95% CI, 0.83-1.01). Similar results were found for ER+ tumors, but results for PR and the joint classification of ER/PR were not conducted. Lee et al16 also reported an inverse association of physical activity with breast cancer among postmenopausal women (RR, 0.67; 95% CI, 0.44-1.02) in the Women's Health Study cohort, and results were similar when restricted to ER+/PR+ tumors (RR, 0.76; 95% CI, 0.43-1.34); no data were reported for other subtypes because of the small sample size.
Two case-control studies46,50 have provided results for joint ER/PR status among postmenopausal women. Similar to our results, Enger et al46 found that higher physical activity had a stronger effect in reducing ER+/PR− tumors (odds ratio, 0.43; 95% CI, 0.19-0.98) than ER+/PR+ tumors (odds ratio, 0.69; 95% CI, 0.42-1.13). Another case-control study63 found the effect of vigorous physical activity to be slightly stronger among ER+ tumors (odds ratio, 0.79; 95% CI, 0.68-0.93) than ER− tumors (odds ratio, 0.86; 95% CI, 0.70-1.05), but did not further cross classify the tumors. Two other case-control studies found little evidence for causative heterogeneity for the association of physical activity with ER/PR subtypes among premenopausal50,64 or postmenopausal64 women.
Our ER/PR subtype results for physical activity parallel those for reductions in fat intake seen in the Women's Health Initiative, a randomized trial of 48 835 postmenopausal women. Women who reduced intake of total fat to 20% of energy had a markedly lower risk of ER+/PR− tumors (RR, 0.64; 95% CI, 0.49-0.84), but not ER+/PR+ tumors (RR, 0.97; 95% CI, 0.86-1.10).36 The ER+/PR− tumors seem to represent a more clinically aggressive phenotype.65 For example, ER+/PR− tumors are often larger,65 have a higher grade,65 have more resistance to selective ER modulators,66- 68 and have a worse prognosis,65,68- 70 compared with ER+/PR+ tumors.
Physical activity decreases the endogenous production of estrogen by reducing adipose tissue, the major source of estrogens in postmenopausal women.56- 58 Lowering estrogen levels could lead to decreased ER+/PR+ tumors, the opposite of which is seen in obesity, in which increased circulating estrogens are associated with increased ER+/PR+ tumors.33,34,38,41,42,46 Consistent with this mechanism, the association of physical activity with ER+/PR+ tumors attenuated after adjustment for BMI. In contrast, there was no change in the association for ER+/PR− tumors after adjustment for BMI, suggesting that additional mechanisms are likely to be important for this subtype. Higher levels of certain growth factors, particularly epidermal growth factor receptor or human growth factor receptor 2, have been associated with ER+/PR− tumors.70 Physical activity seems to impact growth factor levels,57,71,72 and this might in part explain the ER+/PR− results.
This study had a few limitations. First, physical activity was measured as a single self-reported assessment at baseline (1986), with no update during follow-up. Also, it did not include detailed aspects of physical activity, such as lifetime activity, seasonal patterns, or occupational activity, and we were not able to calculate metabolic equivalents. However, previous reports54,55 from the Iowa Women's Health Study cohort have indicated that this physical activity index is sensitive enough to identify major disease trends. Second, information about ER/PR status of breast cancer was obtained through multiple pathological laboratories, rather than a single reference laboratory. However, the ER/PR distribution in our study was similar to that reported by other studies.34,73 Finally, most of the population in the study was white. However, physical activity has been shown to have a beneficial effect on breast cancer among other races.63
In summary, higher physical activity was associated with about a 14% decreased risk of postmenopausal breast cancer. The inverse association was most marked for ER+ tumors, particularly ER+/PR− tumors; the latter, in general, have been associated with a more aggressive phenotype. The results also suggest that additional mechanisms, besides an effect on body mass, may account for observed effects of physical activity for this subtype. Further studies are needed to confirm these novel findings, and to evaluate similar relationships among premenopausal women. If found to be causally related to breast cancer, physical activity would have a substantial public health effect on the prevention of this disease, along with its other positive health benefits.74
Correspondence: James R. Cerhan, MD, PhD, Department of Health Sciences Research, Mayo Clinic College of Medicine, 200 First St SW, Rochester, MN 55905 (firstname.lastname@example.org).
Accepted for Publication: September 1, 2006.
Author Contributions: Dr Cerhan had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Bardia, Sellers, and Cerhan. Acquisition of data: Bardia, Sellers, and Cerhan. Analysis and interpretation of data: Bardia, Hartmann, Vachon, Vierkant, Wang, Olson, and Cerhan. Drafting of the manuscript: Bardia, Vachon, Vierkant, and Cerhan. Critical revision of the manuscript for important intellectual content: Bardia, Hartmann, Vierkant, Wang, Olson, Sellers, and Cerhan. Statistical analysis: Vierkant, Wang, and Cerhan. Obtained funding: Sellers and Cerhan. Administrative, technical, and material support: Olson and Cerhan. Study supervision: Hartmann and Cerhan.
Financial Disclosure: None reported.
Funding/Support: This study was supported in part by grant R01 CA39742 from the National Cancer Institute.
Role of the Sponsor: The funding body had no role in data extraction and analyses, in the writing of the manuscript, or in the decision to submit the manuscript for publication.
Previous Presentation: This study was presented in part at the American Society of Clinical Oncology meeting; June 5, 2006; Atlanta, Ga.