Journal Logo

BASIC SCIENCES: Epedimiology

Physical Activity and Mammographic Density in a Cohort of Midlife Women

OESTREICHER, NINA1; CAPRA, ANGELA1; BROMBERGER, JOYCE2; BUTLER, LESLEY M.3; CRANDALL, CAROLYN J.4; GOLD, ELLEN B.3; GREENDALE, GAIL A.4; MODUGNO, FRANCESMARY2; STERNFELD, BARBARA1; HABEL, LAUREL A.1

Author Information
Medicine & Science in Sports & Exercise: March 2008 - Volume 40 - Issue 3 - p 451-456
doi: 10.1249/MSS.0b013e31815f5b47
  • Free

Abstract

Physical activity is one of the few risk factors for breast cancer that is modifiable. An International Agency for Research on Cancer working group systematically reviewed the evidence on more than 150 studies of physical activity and cancer prevention and found strong evidence that higher levels of physical activity reduce the risk of developing breast cancer by 30-40% (16). Evidence for the protective effect of physical activity is strongest for postmenopausal women and has been more equivocal (10) or conflicting among premenopausal women (17). Results of recent investigations have also suggested that among the domains of physical activity (e.g., recreational, household, and work), recreational activity may be the most important contributor to the physical activity-breast cancer association (17,18,22,23), because that domain is most likely to involve moderate to vigorous intensity activity that is of a structured and continuous nature (3). Physical activity may decrease breast cancer risk by lowering levels of endogenous estrogen and insulin-like growth factors or by decreasing obesity (15), all of which may also be associated with mammographic density (5,8,12).

High mammographic density is one of the strongest established risk factors for breast cancer, with four- to sixfold increased risks reported among those women with the highest compared with those with the lowest categories of mammographic density (4). Mammographic density refers to the extent of radiographically dense tissue in the breast. The earliest approach to evaluating mammographic density was semiquantitative (34). Later approaches have quantified breast density in both relative (percentage of the breast occupied by radiographically dense tissue) and absolute (area of radiographically dense tissue, measured in centimeters squared) terms, and these quantitative approaches have generally proven to be more predictive of breast cancer risk than semiquantitative breast density-assessment methods (6); they also demonstrate high reliability (4). Although not used as frequently as percent density, recent studies suggest that area of density may be equally predictive of breast cancer risk (21), may equally reflect racial/ethnic differences in risk (21), and may possibly better reflect relevant changes from dietary interventions (4). Mammographic density is strongly associated with breast cancer risk and is modifiable. Therefore, it may be useful to examine the association of physical activity and mammographic density as a preliminary means of assessing the value of regular physical activity in breast cancer prevention.

There have been limited observational studies of the association between physical activity and mammographic density in healthy women (11,19,20,26,27,30,32). Investigations have been conducted primarily among postmenopausal women, and there are few known studies in multiethnic populations (27). Several previous studies found null (19,26,27,30,32) or weak associations (11,20) of physical activity with mammographic density. Two studies have found strong inverse associations of physical activity and mammographic density among postmenopausal breast cancer survivors (17,18). Only some of these studies have assessed specific activity domains: recreational activity alone (17,18) and in addition to work activity (11) and household activity (30). Few of the studies account for frequency or duration of activity (27,30).

The aim of our study was to comprehensively examine the association of physical activity and mammographic density in a multiethnic cohort of pre- and early perimenopausal women. Multiple domains of physical activity were examined using indices that accounted for frequency, intensity, and duration of exposure.

MATERIALS AND METHODS

Study Setting and Population

We included participants from the UC Davis-Kaiser, UCLA, and University of Pittsburgh field centers for the Study of Women's Health Across the Nation (SWAN), a multisite, multiethnic, community-based longitudinal study of midlife women (28). Briefly, at each site, a sample of women who self-identified as Caucasians, and a prespecified sample of women who self-identified as belonging to another specific racial/ethnic group, were enrolled. At entry, eligible women had to be premenopausal or early perimenopausal (i.e., had a menstrual period in the prior 3 months and without or with changes in regularity of cycles, respectively) and could not be taking contraceptive/menopausal hormones during the previous 3 months. Only information on ever- versus never use of contraceptive or menopausal hormones prior to that interval was collected. Information on a comprehensive range of exposures, including anthropometric, reproductive, and lifestyle factors, was collected at baseline and at annual follow-up visits. For our ancillary study, we obtained screening mammograms up to 2 yr prior to the baseline SWAN visit through 2 yr after annual follow-up visit 6, and we linked the density assessments for each screening mammogram to SWAN data from the study visit closest to that mammogram.

For the current analyses, we selected the earliest mammogram (index mammogram) among eligible mammograms, and we restricted analyses to mammograms done when women were pre- or early perimenopausal. We made these restrictions in order to evaluate physical activity associations at a time closest to that of peak breast density, as density generally declines with age and through the menopausal transition (33). A total of 1005 women were potentially eligible for the current analyses, which constitutes 86, 84, and 71% of the SWAN core study participants at the UC Davis-Kaiser, UCLA, and University of Pittsburgh sites, respectively. Women were excluded from these analyses if they were missing mammographic density measures (N = 8), were taking hormones after the baseline SWAN visit (N = 29), or were not pre- or early perimenopausal at the time of their index mammogram (N = 196). This left 772 women available for these analyses. We obtained informed written consent to participate in the study from all women in the study. Our study was approved by institutional review boards at all participating study sites.

Data Collection

Mammographic density.

Films from the craniocaudal view for the right breast were retrieved and assessed for breast area and area of radiographically dense tissue (area of density, measured in centimeters squared). A single mammographic density assessor who was blinded to participant characteristics, including physical activity levels, performed these assessments. Martine Salane, the mammographic density assessor for our study, is an established expert in the techniques of measuring mammographic density. She has been considered the standard expert against which computer-based methods have been evaluated (31). Ms. Salane's measurements are therefore highly correlated with computer-assisted density measurements (P = 0.90) (14). Most importantly for this study, her evaluation of mammographic density has been validated in a number of research studies (i.e., shown to be strongly associated with breast cancer risk) (7,24,25,35). Films from the left breast were used for women with a prior surgery on the right breast (N = 14). The area of the breast and area of density were traced onto acetate overlays. A compensating polar planimeter was used to measure the total breast area (cm2) and the total area of density (cm2). Percent mammographic density (percent density) was calculated as area of density divided by breast area, expressed as a percentage. Unidentified to the reader, 10% of all films were sent for re-review. Multiple readings were compared for concordance of these assessments. The within-person Spearman correlation coefficient of percent density was 0.96, and the mean difference in percent density assessments was 2.2%, (CI 1.7%, 2.7%). There was strong agreement for parenchymal pattern assessments (kappa = 0.7).

Physical activity.

Physical activity was assessed with the Kaiser Physical Activity Survey, an adaptation by Sternfeld et al. (29) of the Baecke Short Physical Activity Questionnaire (2). The Kaiser Physical Activity Survey was designed specifically to assess activity in women, including questions about household and caregiving activities. It has been validated in Caucasians (1) and members of racial/ethnic minorities (29). This survey is a comprehensive assessment of physical activity in the past year, ascertaining activity frequency in multiple domains by means of a self-administered questionnaire with Likert-type questions about frequency of activity in each domain. Four mutually exclusive physical activity indices were derived from questionnaire responses: the active living index, sports index, work index, and household/caregiving index. The active living index measures nonsports leisure activities using responses to two questions about the frequency with which the participant watched television and walked or bicycled to and from work, school, or errands. The household/caregiving questions ask about frequency of various activities only and assign values for duration. An example of a question that contributes to the household/caregiving index is, "During the past year, how much time did you spend preparing meals or cleaning up from meals?" The sports index and work index are based on questions that ask about frequency and duration of activity. The sports index also assigns intensity values to each activity and the work index assigns values for intensity based on job title. The individual activity indices range in value from 1 to 5, with 5 indicating the highest level of activity. Because the activity indices are developed from questions with categorical responses, quantifiable measures of energy expenditure (i.e., kilocalories per week) cannot be derived from these indices. Rather, the indices represent relative rankings of activity levels that enable comparisons between indices. One-month test-retest reliability was high for all indices (r = 0.79-0.91) (1). Correlations between activities from the Kaiser Physical Activity Survey and 7-d physical activity diaries ranged from r = 0.03 to 0.64. Daily, habitual activities from the Kaiser Physical Activity Survey were the most highly correlated with the daily physical activity diaries (r > 0.28), whereas correlations between infrequent activities from the Kaiser Physical Activity Survey and daily physical activity diaries were very low (r < 0.05). We used tertiles of the sports index and work index and categorized values of the household/child care index and active living index at levels below, at, and above the median for our analyses. Because of the limited range of values for each index, it was not possible in all cases to divide the study population evenly into their respective categories.

Statistical Analyses

Two measures of mammographic density, area of density and percent density, were analyzed in separate multivariable linear regression models. Although the distribution of area of density was slightly right-skewed, linear regression inference uses the t-distribution, which is robust against minor deviations from normality, and, therefore, values were not log transformed. A priori, we planned to adjust for race/ethnicity, age at mammogram, body mass index (BMI) (continuous, measured in kilograms per meter squared), parity, menopausal status, past use of contraceptive hormones, and past use of menopausal hormones, because these factors were associated with mammographic density in previous analyses of this cohort (13). Additionally, we considered the remaining exposures in Table 1 as potential confounding factors. Those factors that changed the beta estimate for any of the physical activity variables by 10% or more, after adjusting for the prespecified adjustment variables, were considered to be confounders and were included as covariates in the adjusted models.

TABLE 1
TABLE 1:
Characteristics of 772 pre- and early perimenopausal study participants.

We performed several secondary analyses. There was wide variation in BMI across race/ethnicity groups, making it difficult to disentangle the associations of BMI and race/ethnicity with mammographic density. We therefore also performed analyses restricted to women in the second and third BMI quartiles based on the entire study population, because there were some women from each racial/ethnic group in both of these strata. Stratified analyses according to World Health Organization obesity criteria were also performed (36). To minimize potential misclassification of physical activity levels at the time of the mammogram, we restricted some analyses to women whose dates of assessment of physical activity and mammogram date were fewer than 6 months apart. We also examined whether associations varied by race/ethnicity and by menopausal status.

RESULTS

The average age of the 772 participants was 47 yr, and the majority were parous (82%), premenopausal (54%), never-smokers (69%), and well educated (55% had a college education or higher degree, Table 1). Mean percent density was 45% (median 46%), and mean area of density was 48 cm2 (median 42 cm2). The largest difference in mean percent density between the highest and lowest levels of an individual physical activity index was observed for the active living index (44 vs 47%, respectively), and the smallest difference was observed for the household/child care activity index (46 vs 47%, respectively) (data not shown). The mean time interval between physical activity assessment and mammogram date was approximately 7.8 months (median, 5.3 months). Several of the physical activity indices had weak inverse associations with percent density and/or area of density, after adjusting for age, menopausal status (premenopausal and early perimenopausal), race/ethnicity, parity, BMI, waist circumference, education, and past use of hormones (contraceptive and menopausal hormones) (Table 2). For the active living index, women who had values above the median had lower percent and area of density (beta = −2.62 95% confidence interval (CI) (−5.84, 0.60) and beta = −4.75 (CI −10.40, 0.88)), but these associations did not reach statistical significance. We ran the models for percent density and area of density and the active living index, with an additional adjustment for sports index, given that these activities may be correlated. Adjustment for sports index did not materially change results for the active living index (beta for above median level of active living index for percent density = −1.92 vs −2.62 in model not adjusted for sports index) and = −4.21 (vs −4.75 in model not adjusted for sports index) for area of density. Although the associations of sports and household/child care activity and density (percent density and area of density) and work activity (percent density only) also did not approach statistical significance, their inverse trends were in the expected direction. We did not find associations between work activity and area of density.

TABLE 2
TABLE 2:
Association between physical activity and breast and density measures.a

We performed several secondary analyses, adjusting for the same covariates as for the primary analyses. Our multivariable analyses restricted to women whose interval between physical activity assessment and mammogram date was 6 months or less did not yield substantively different results (data not shown). Overall results were also similar among African American, Caucasian, Chinese, and Japanese women.

When multivariable models were stratified by BMI according to World Health Organization criteria (< 25 kg·m−2, ≥ 25 to < 30 kg·m−2, and ≥ 30 kg·m−2), results did not differ across BMI strata for the work, active living, and household/child care indices for either percent density or area of density (data not shown). However, for the sports index, there were negative associations for women whose BMI was < 25, whereas there were positive associations for those whose BMI was > 30, for both percent density and area of density. Analyses restricted to women in the second and third BMI quartiles produced similar results to analyses of the full cohort (data not shown).

It is interesting to note that BMI was the factor most strongly associated (inversely) with mammographic density in our cohort; results unadjusted for BMI in the full cohort were in the reverse direction of adjusted results. In analyses stratified by race/ethnicity, the confounding effect of BMI on results for percent density was especially strong for African American and Caucasian women, and less strong for Chinese and Japanese women. In contrast to percent density, not adjusting for BMI did not affect the direction of associations of most physical activity variables with area of density in the full cohort, and this was also true when analyses were stratified by race/ethnicity.

DISCUSSION

To our knowledge, this is the second study to examine physical activity and mammographic density in a multiethnic cohort (27) and the first to examine a comprehensive group of specific activity domains. We observed inverse associations for the highest versus the lowest category of active living and both measures of mammographic density. We also observed modest inverse associations with the highest level of each of the other domains of physical activity and percent mammographic density, and for all but work index for area of density. All estimates for the associations of physical activity and mammographic density were imprecise and findings may be attributable to chance.

While our study was one of the largest investigations of its kind, there was limited statistical power because of small group sizes, particularly at the highest levels of physical activity. Our study had other limitations as well. We had information on recent physical activity only. However, the etiologically relevant time period during which physical activity affects breast cancer risk (or mammographic density) remains to be established (9). Inclusion of physical activity outside the relevant time window would attenuate associations. Also, the units for the physical activity indices are ordinal rankings. They cannot be compared with variables that are expressed in terms of MET-hours per week or other similar units of energy expenditure. Although this is a limitation of the study, as an example, the sports index has been shown to be a valid measure of activity against activity records, accelerometers, and aerobic capacity; plus, it shows the expected relations with other outcomes such as BMI, bone density, and weight change (1).

Another limitation was the dissimilarity of BMI across race/ethnicity, limiting our ability to fully adjust for BMI in multivariable models. However, restricting analyses to women in the second and third BMI quartiles did not substantively change results. While residual confounding by BMI is possible, it would result in an attenuation of the strength of our inverse associations.

Our results of an inverse association between multiple domains of physical activity and mammographic density are in agreement with some studies (11,20), but not the majority (19,26,27,30,32). Most studies have been conducted among primarily postmenopausal Caucasian women, and it possible that the relationship between physical activity and mammographic density differs by race/ethnicity or menopausal status. However, we did not observe differences across racial/ethnic subgroups. The only other known study of physical activity and mammographic density in a multiethnic cohort (27) did not stratify results by race/ethnicity.

Two recent studies that stratified results by menopausal status did not find physical activity-mammographic density associations in either premenopausal or postmenopausal women (27,30). However, a study by Irwin et al. (17) of sports and recreational activity and mammographic density in premenopausal breast cancer survivors observed a positive association in lean women and an inverse association in heavy women. An inverse association was observed in both lean and heavy postmenopausal breast cancer survivors. Our study of pre/perimenopausal women found an inverse sports activity-mammographic density association in lean women and a positive association in heavy women-the opposite of the findings of Irwin et al. It is possible that associations between mammographic density and physical activity may differ between healthy women and breast cancer survivors.

In addition, studies that used less-quantitative density measurements might have had attenuated estimates, because the more-quantitative measurements are generally more strongly associated with breast cancer risk. Mean percent density among our multiethnic study population of premenopausal and perimenopausal women was higher than what has been reported in other physical activity studies conducted in primarily postmenopausal Caucasian women: 45% in our study versus 34% (30,32) and 17% (20). We would expect lower mammographic density in studies of older and postmenopausal women, and previous studies have found that breast composition differs by race/ethnicity (21).

In conclusion, we found that a lower percent density was inversely associated with the highest versus the lowest category of each domain of physical activity, although modestly so. Area of density was inversely associated with all domains but work index. However, associations were generally nonlinear. Contrary to expectations, we did not find the strongest association between sports activity and mammographic density. Boyd and others have estimated that for every 1% increase in density, there is a 2% increase in breast cancer risk (4,7). In our study, mean percent density at the highest versus the lowest levels of the active living index (adjusted for confounding factors) was 44% and 47%, respectively. Although causality cannot be established from these observational data, if high levels of physical activity are truly associated with a 30-40% reduction in breast cancer risk, our more modest inverse associations between physical activity and density reinforce the strategy of breast cancer risk reduction accomplished through regular physical activity.

This analysis was supported by PHS grant CA89552. In addition, the Study of Women's health Across the Nation (SWAN) has grant support from the National Institutes of Health, DHHS, National Institute on Aging, the National Institute of Nursing Research, and the NIH Office of Research on Women's Health (NR004061, AG012505, AG012531, AG012539, AG012546, AG012553, AG012554, and AG012495). Dr. Crandall was supported by National Institutes on Aging grant 5K12AG01004. We thank Ninah Achacoso for her programming assistance.

REFERENCES

1. Ainsworth BE, Sternfeld B, Richardson MT, Jackson K. Evaluation of the Kaiser physical activity survey in women. Med Sci Sports Exerc. 2000;32(7):1327-38.
2. Baecke JA, Burema J, Frijters JE. A short questionnaire for the measurement of habitual physical activity in epidemiological studies. Am J Clin Nutr. 1982;36:936-42.
3. Bernstein L, Patel AV, Ursin G, et al. Lifetime recreational exercise activity and breast cancer risk among black women and white women. J Natl Cancer Inst. 2005;97:1671-9.
4. Boyd NF, Byng JW, Jong RA, et al. Quantitative classification of mammographic densities and breast cancer risk: results from the Canadian National Breast Screening Study. J Natl Cancer Inst. 1995;87:670-5.
5. Boyd NF, Lockwood GA, Byng JW, Little LE, Yaffe MJ, Tritchler DL. The relationship of anthropometric measures to radiological features of the breast in premenopausal women. Br J Cancer. 1998;78:1233-8.
6. Boyd NF, Lockwood GA, Byng JW, Tritchler DL, Yaffe MJ. Mammographic densities and breast cancer risk. Cancer Epidemiol Biomarkers Prev. 1998;7:1133-44.
7. Byrne C, Schairer C, Wolfe J, et al. Mammographic features and breast cancer risk: effects with time, age, and menopause status. J Natl Cancer Inst. 1995;87:1622.
8. Diorio C, Pollak M, Byrne C, et al. Insulin-like growth factor-I, IGF-binding protein-3, and mammographic breast density. Cancer Epidemiol. Biomarkers Prev. 2005;14:1065-73.
9. Dorn J, Vena J, Brasure J, Freudenheim J, Graham S. Lifetime physical activity and breast cancer risk in pre- and postmenopausal women. Med Sci Sports Exerc. 2003;35(2):278-85.
10. Friedenreich CM. Physical activity and breast cancer risk: the effect of menopausal status. Exerc Sport Sci Rev. 2004;32:180-4.
11. Gram IT, Funkhouser E, Tabar L. Moderate physical activity in relation to mammographic patterns. Cancer Epidemiol Biomarkers Prev. 1999;8:117-22.
12. Greendale G, Palla S, Ursin G, et al. The association of endogenous sex steroids and sex steroid binding proteins with mammographic density: results from the Postmenopausal Estrogen/Progestin Interventions Mammographic Density Study. Am J Epidemiol. 2005;162:826-34.
13. Habel L, Capra A, Oestreicher N, et al. Mammographic density in a multiethnic cohort. Menopause. 2007;14:891-9.
14. Haiman CA, Bernstein L, Berg D, Ingles SA, Salane M, Ursin G. Genetic determinants of mammographic density. Breast Cancer Res. 2002;4:R5.
15. Hoffman-Goetz L, Apter D, Demark-Wahnefried W, Goran MI, McTiernan A, Reichman ME. Possible mechanisms mediating an association between physical activity and breast cancer. Cancer. 1998;83:621-8.
16. IARC vol. 6: weight control and physical activity. In: IARC Handbook on Cancer Prevention. Lyon (France): IARC; 2002. p.144-54.
17. Irwin ML, Aiello EJ, McTiernan A, et al. Pre-diagnosis physical activity and mammographic density in breast cancer survivors. Breast Cancer Res Treat. 2006;95:171-8.
18. Irwin ML, Aiello EJ, McTiernan A, et al. Physical activity, body mass index, and mammographic density in postmenopausal breast cancer survivors. J Clin Oncol. 2007;25(9):1061-6.
19. Jeffreys M, Warren R, Gunnell D, McCarron P, Smith GD. Life course breast cancer risk factors and adult breast density (United Kingdom). Cancer Causes Control. 2004;15:947-55.
20. Lopez P, Van Horn L, Colangelo LA, Wolfman JA, Hendrick RE, Gapstur SM. Physical inactivity and percent breast density among Hispanic women. Int J Cancer. 2003;107:1012-6.
21. Maskarinec G, Pagano I, Lurie G, Wilkens LR, Kolonel LN. Mammographic density and breast cancer risk: the multiethnic cohort study. Am J Epidemiol. 2005;162:743-52.
22. McTiernan A, Kooperberg C, White E, et al. Recreational physical activity and the risk of breast cancer in postmenopausal women: the Women's Health Initiative Cohort Study. JAMA. 2003;290:1331-6.
23. Patel AV, Callel EE, Bernstein L, Wu AH, Thun MJ. Recreational physical activity and risk of postmenopausal breast cancer in a large cohort of US women. Cancer Causes Control. 2003;14:519-29.
24. Saftlas AF, Hoover RN, Brinton LA, et al. Mammographic densities and risk of breast cancer. Cancer. 1991;67(11):2833-8.
25. Saftlas AF, Wolfe JN, Hoover RN, et al. Mammographic parenchymal patterns as indicators of breast cancer risk. Am J Epidemiol. 1989;129:518.
26. Sala E, Warren R, McCann J, Duffy S, Luben R, Day N. High-risk mammographic parenchymal patterns, hormone replacement therapy and other risk factors: a case-control study. Int J Epidemiol. 2000;29:629-36.
27. Siozon CC, Ma H, Hilsen M, Bernstein L, Ursin G. The association between recreational physical activity and mammographic density. Int J Cancer. 2006;119:1695-1701.
28. Sowers M, Crawford S, Sternfeld B. SWAN: a multicenter, multiethnic community-based cohort study of women and the menopausal transition. In: Menopause: Biology and Pathobiology. New York (NY): Academic Press; 2000. p. 175-88.
29. Sternfeld B, Ainsworth BE, Quesenberry CP. Physical activity patterns in a diverse population of women. Prev Med. 1999;28:313-23.
30. Suijkerbuijk KP, Van Duijnhoven FJ, Van Gils CH, et al. Physical activity in relation to mammographic density in the dutch prospect-European prospective investigation into cancer and nutrition cohort. Cancer Epidemiol Biomarkers Prev. 2006;15:456-60.
31. Ursin G, Astrahan MA, Salane M, et al. The detection of changes in mammographic densities. Cancer Epidemiol Biomarkers Prev. 1998;7:43-7.
32. Vachon CM, Kushi LH, Cerhan JR, Kuni CC, Sellers TA. Association of diet and mammographic breast density in the Minnesota breast cancer family cohort. Cancer Epidemiol Biomarkers Prev. 2000;9:151-60.
33. Wolfe JN. Breast parenchymal patterns and their changes with age. Radiology. 1976;121:545-52.
34. Wolfe JN. Risk for breast cancer development determined by mammographic parenchymal pattern. Cancer. 1976;37:2486-92.
35. Wolfe JN, Saftlas AF, Salane M. Mammographic parenchymal patterns and quantitative evaluation of mammographic densities: a case-control study. AJR Am J Roentgenol. 1987;148:1087.
36. World Health Organization. Global strategy on diet, physical activity and health: obesity and overweight. Available at: http://www.who.int/dietphysicalactivity/publications/facts/obesity/en. Accessed February 4, 2007.
Keywords:

BREAST CANCER; ETIOLOGY; PARENCHYMAL PATTERNS; EXERCISE

©2008The American College of Sports Medicine