Abstract
Dense breast is one of the strong risk factors for breast cancer among women. While it has been established that physical activity is associated with decreased risk for breast cancer, results have been inconsistent in terms of mammographic density. Thus, we examined physical activity in relation to mammographic density among Japanese women in Tokyo.
We used 123,026 records from 33,698 Japanese women without a history of breast cancer who got mammograms at St. Luke's International Hospital in Tokyo, Japan from 2004 to 2019. Mammographic density was classified according to the Breast Imaging Reporting and Data System (BI-RADS), and women self-reported their physical activity level over the past year. ORs were estimated using logistic generalized estimating equations after adjusting for age, body mass index, menopausal status, parity, family history of breast or ovarian cancer, hormone therapy use, smoking status, alcohol consumption, and year.
We observed inverse associations of physical activity with dense breasts. Adjusted ORs were 0.96 (95% confidence interval: 0.91–1.00) for women with physical exercise for 1–2 days per week, 0.94 (0.88–0.99) for those with physical exercise for 3–5 days per week, and 0.91 (0.84–0.99) for those with daily physical exercise when compared with those reported seldom physical exercise.
Higher levels of physical activity may be associated with decreased mammographic density levels in Japanese women.
Increasing physical activity may serve as a reasonable intervention to reduce mammographic density, and thereby, to mitigate the risk of breast cancer in Asian women.
Introduction
In the perspective of breast cancer prevention, breasts are typically classified into four groups: almost entirely fat, having scattered fibroglandular tissue, heterogeneously dense, or extremely dense. The latter two groups are regarded as dense breast, which is positively associated with breast cancer in women (1–4). Dense breasts are assessed by using mammography and approximately 31% to 41% of women are known to have dense breasts with racial variation (5). Asian women have the greatest prevalence of dense breasts compared with other ethnicity groups (6–8).
Studies have shown that mammographic density is positively associated with family history of breast cancer and use of hormone treatment, but inversely associated with age, body mass index (BMI), parity, and menopause (9–12). However, the literature has produced mixed results with respect to an association of physical activity with mammographic density (13–16).
It is established that physical activity is associated with lower risk of breast cancer or breast cancer–specific mortality independent of BMI (17–20). An umbrella review revealed that physical activity was most strongly associated with a lower risk of breast cancer among several types of cancers (21). With regard to mammographic density, however, the findings have been inconsistent. Two systematic reviews (13, 22) concluded a null association between physical activity and mammographic density. An epidemiologic study of Danish women concluded that physical activity is not a determinant of mammographic density (23), and the same conclusion was drawn from an analysis of Malaysian women (24). Although null findings prevail in the literature, a randomized intervention trial found that an increase in physical activity may reduce mammographic density (14). Still, whether physical activity reduces mammographic density remains unconclusive. Moreover, the majority of these have been among in Western countries (13, 25). Only few studies have examined the association in Asian women, who has the highest prevalence of dense breast, limiting the generalization of results to other ethnic groups.
In previous studies, physical activity was generally viewed as individual choice, and thus effect modification was assessed for known risk factors at the individual level such as menopausal status and BMI (23, 26). Meanwhile, a growing literature has documented that increasing green environment promotes physical activity (27–29). Along with other policy interventions for promoting individuals’ physical activity, an increase in greenness will benefit the general public in the form of a shared resource. Nevertheless, analyses assessing whether greenness modulates the association of physical activity with mammographic density have not yet been conducted.
Therefore, the objective of this study is to determine the association between physical activity and mammographic density among women in Tokyo, Japan. In doing so, we also evaluated whether the association can be modified by environmental factors such as greenness.
Materials and Methods
Study population
The longitudinal follow-up data were sampled from women who had screening mammograms at the Center for Preventive Medicine at St. Luke's International Hospital, Tokyo from 2004 to 2019. Before mammogram, women answered a questionnaire about their history of disease, family history of disease, and health-related lifestyle including physical activity, smoking, and alcohol consumption for the past year. For this study, women with a history of breast cancer were excluded.
The Japanese universal health insurance system provides citizens with annual health examinations for public health (30). Employers are obliged to offer health checks to their employees on an annual basis. The unemployed or self-employed are also given the annual health checkups by the local government. Every year, individuals are asked to examine their health at a medical center. Prior to the check-up day, they are given the questionnaire and answer health-related questions such as personal/family history of disease or other relevant life style questions. Individuals tend to revisit the same institute every year, which enables longitudinal observations on health outcomes and relevant variables for each individual on an approximately yearly basis.
The study was approved by the Institutional Review Board (IRB) of St. Luke's International Hospital.
Measurement of physical activity
For each year, women reported on their physical activity levels over the past year in the questionnaire a few days before the mammogram. In the question about regular physical exercise, they reported how many days per week they had engaged in moderate-intensity exercise inducing a light sweat for 20 minutes or more: daily, 3 to 5 days per week, within 1 to 2 days per week, and seldom.
Outcome ascertainment
Mammographic density was measured by mammograms obtained from full-field digital mammography. The images of both breasts were taken for cephalocaudal (CC) and mediolateral oblique (MLO) views, for a total of four images per patient (CC right, MLO right, CC left, MLO left).
Board-certified imaging radiologists and breast specialists assessed mammography images to assign mammographic density into four categories based on the American College of Radiology's Breast Imaging Reporting and Data System (BI-RADS): almost entirely fat (Category 1), scattered areas of fibroglandular density (Category 2), heterogeneously dense (Category 3), and extremely dense categories (Category 4). Two certified doctors made separate official radiology reports and when a discrepancy arises between two reports, the consensus is reached before the report is finalized. During the assessment, the radiologists were blinded by physical activity or other lifestyle reports of the women.
Other covariates
Weight and height were measured using an electric scale on the day of examination to calculate BMI. Women also reported parity (count), menopausal status (yes or no), smoking status (smoker, non-smoker, past-smoker), family history of breast cancer or ovarian cancer (yes or no), and alcohol consumption in the questionnaire.
To measure green space, we obtained the normalized difference vegetation index (NDVI) values from the Landsat 5 satellite at a resolution of 30 m. The value ranges from −1 to 1 and higher, with more vegetation. Individuals’ postal codes were mapped out using Quantum Geographic Information System (QGIS) version 3.32 (QGIS Project). We averaged NDVI values for each postal area, which was regarded as a green area if its mean NDVI is greater than 0.25.
Statistical analysis
We used logistic generalized estimating equations (GEE) to estimate the association of physical activity with mammographic density. The model accounted for the within-subject correlations that arise from repeated outcome measurements for each individual. The level of physical activity was treated as an ordered variable of daily, 3 to 5 days per week, 1 to 2 days per week, and none. We used the physical activity level that was reported at the same visit for the mammogram. The main model was controlled for age (continuous), BMI (continuous), menopausal status (yes or no), parity (count), family history of breast or ovarian cancer (yes or no), hormone therapy use (yes or no), smoking status (current, past, or never), the frequency of alcohol consumption (habitually, occasionally, or never), and the calendar year of the check-up. We assigned 1 as dense breast if the category of either breast is fell into BI-RADS Category 3 or 4 (heterogeneously dense and extremely dense, respectively). The other groups (BI-RADS Category 1 and 2, entirely fat and scattered fibroglandular tissue) were assigned to the value zero for not having dense breasts.
Menopausal status and parity are main risk factors for hormonal cancer such as breast cancer. BMI is correlated with both physical activity and risk of breast cancer. Therefore, we performed subgroup analyses according to menopausal status (premenopausal vs. postmenopausal), parity (nulliparous vs. parous), BMI (<25 kg/m2 vs. ≥25), and greenness. We concluded that difference is not statistically significant if the confidence intervals (CI) of two groups overlap.
We conducted several sensitivity analyses. In the first sensitivity analysis, we fit an ordinal logistic GEE model to use the raw ordinal scale of the four BI-RADS categories instead of the two combined categories of low and high.
In the second sensitivity analysis, we evaluated whether physical activity had time-lagged associations with mammographic density. We replaced the current year's physical activity with the one in the previous year (lag 1) or in the two previous years (lag 2).
While we used the default correlation structure according to a quasi-likelihood information criterion, we also applied the other correlation structures to examine whether the results change significantly under different correlation structures.
Analyses were conducted using the PROC GENMOD procedure with the REPEATED statement in SAS software version 9.4 (SAS Institute Inc.).
Data availability
The datasets used in this study are not publicly available due to the institutional restrictions on data security, but may be made available upon reasonable request from the corresponding author, subject to IRB approval and permission by the institute.
Results
The average number of measurements per woman was 3.7. The characteristics of records (n = 123,026) from 33,698 women included in this study are presented in Table 1. There were 63,502 records classified under the not dense breast group, and 59,524 records in the dense breast group based on the mammographic categories. Women with dense breasts were younger and have a smaller BMI. The mean age and BMI of the participants was 56.3 years and 22.4 kg/m2, respectively for the not dense group, and 48.6 years and 20.3 kg/m2, respectively for the dense group. Those with dense breasts also drink less and less physically active compared with women in the dense group. The majority of the women were non-smokers (80.4% for the dense group and 82.5% for the non-dense group) and lived in urban areas.
Description of data from 33,698 women 2004 to 2019 (n = 123,026).
Characteristic . | Dense breasta (n = 59,524) . | Non-dense breastb (n = 63,502) . |
---|---|---|
Age, years, mean ± SD (range) | 48.6 ± 9.9 (21–90) | 56.3 ± 10.7 (23–93) |
Body mass index, kg/m2, n (%) | 20.3 ± 2.5 | 22.4 ± 3.4 |
Parity, n, mean ± SD | 0.8 ± 1.0 | 1.4 ± 1.1 |
Menopaused, n (%) | 21,330 (35.8) | 42,745 (67.3) |
Family history of breast and/or ovarian cancer, n (%) | 7,404 (12.4) | 7,279 (11.5) |
Hormone therapy, n (%) | 7,883 (13.2) | 7,476 (11.8) |
Smoking, n (%) | ||
Never | 47,833 (80.4) | 52,379 (82.5) |
Former | 7,789 (13.1) | 7,931 (12.5) |
Current | 3,902 (6.6) | 3,192 (5.0) |
Alcohol consumption | ||
Never | 23,072 (45.7) | 29,776 (56.8) |
Sometimes | 8,903 (17.6) | 7,909 (15.1) |
Frequent | 18,554 (36.7) | 14,727 (28.1) |
Physical activityc | ||
Seldom | 20,861 (35.1) | 18,101 (28.5) |
Within 1–2 days/week | 22,047 (37.0) | 22,732 (35.8) |
3–5 days/week | 11,232 (18.9) | 15,081 (23.8) |
Almost everyday | 5,384 (9.1) | 7,588 (12.0) |
Green residenced | 7,062 (13.4) | 8,392 (16.0) |
Characteristic . | Dense breasta (n = 59,524) . | Non-dense breastb (n = 63,502) . |
---|---|---|
Age, years, mean ± SD (range) | 48.6 ± 9.9 (21–90) | 56.3 ± 10.7 (23–93) |
Body mass index, kg/m2, n (%) | 20.3 ± 2.5 | 22.4 ± 3.4 |
Parity, n, mean ± SD | 0.8 ± 1.0 | 1.4 ± 1.1 |
Menopaused, n (%) | 21,330 (35.8) | 42,745 (67.3) |
Family history of breast and/or ovarian cancer, n (%) | 7,404 (12.4) | 7,279 (11.5) |
Hormone therapy, n (%) | 7,883 (13.2) | 7,476 (11.8) |
Smoking, n (%) | ||
Never | 47,833 (80.4) | 52,379 (82.5) |
Former | 7,789 (13.1) | 7,931 (12.5) |
Current | 3,902 (6.6) | 3,192 (5.0) |
Alcohol consumption | ||
Never | 23,072 (45.7) | 29,776 (56.8) |
Sometimes | 8,903 (17.6) | 7,909 (15.1) |
Frequent | 18,554 (36.7) | 14,727 (28.1) |
Physical activityc | ||
Seldom | 20,861 (35.1) | 18,101 (28.5) |
Within 1–2 days/week | 22,047 (37.0) | 22,732 (35.8) |
3–5 days/week | 11,232 (18.9) | 15,081 (23.8) |
Almost everyday | 5,384 (9.1) | 7,588 (12.0) |
Green residenced | 7,062 (13.4) | 8,392 (16.0) |
aAlmost entirely fat and scattered fibroglandular.
bHeterogeneously dense and extremely dense.
cWomen self-reported how many days per week they had engaged in moderate-intensity exercise inducing a light sweat for 20 minutes or more over the past year.
dNDVI > 0.25.
More frequent physical exercise was associated with lower odds of having dense breasts in Japanese women (Table 2), with fully adjusted OR of 0.96 (95% CI: 0.91–1.00) for women with physical exercise for 1 to 2 days per week, 0.94 (95% CI: 0.88–0.99) for those with physical exercise for 3 to 5 days per week, and 0.91 (95% CI: 0.84–0.99) for those with daily physical exercise comparing those reported none of physical exercise. The inverse relationship was linear, which was statistically significant at an alpha level of 0.05.
Crude and adjusted ORs of dense breasts in relation to physical activity in Japanese women, 2004 to 2019.
. | Frequency (%) . | Crude . | Adjusteda . |
---|---|---|---|
Reference (Seldom) | 38,962 (31.7) | 1 | 1 |
1–2 days/week | 44,779 (36.4) | 0.84 (0.81–0.88)b | 0.96 (0.92–1.01) |
3–5 days/ week | 26,313 (21.4) | 0.65 (0.61–0.68)b | 0.94 (0.89–1.00) |
Everyday | 12,972 (10.5) | 0.62 (0.57–0.66)b | 0.91 (0.84–0.99)b |
. | Frequency (%) . | Crude . | Adjusteda . |
---|---|---|---|
Reference (Seldom) | 38,962 (31.7) | 1 | 1 |
1–2 days/week | 44,779 (36.4) | 0.84 (0.81–0.88)b | 0.96 (0.92–1.01) |
3–5 days/ week | 26,313 (21.4) | 0.65 (0.61–0.68)b | 0.94 (0.89–1.00) |
Everyday | 12,972 (10.5) | 0.62 (0.57–0.66)b | 0.91 (0.84–0.99)b |
aAdjusted for age, body mass index, menopausal status, parity, family history of breast or ovarian cancer, hormone therapy use, smoking status, alcohol consumption, and year.
bStatistically significant at α = 0.05.
Results indicated more apparent inverse associations among premenopausal, parous, and high BMI (≥25 kg/m2) women as compared with postmenopausal, nulliparous, and low BMI (<25 kg/m2) women, respectively (Table 3). The association was also stronger among women who were living in green neighborhood (OR: 0.78; 95% CI: 0.68–0.89) than those in less green area (OR: 0.91; 95% CI: 0.82–1.01). However, none of the differences were statistically significant.
ORs of dense breasts in relation to physical activity according to menopausal status, parity, BMI, and greenness in Japanese women, 2004 to 2019.
. | Menopausea . | Parityb . | BMIc . | Greenness . | ||||
---|---|---|---|---|---|---|---|---|
. | Pre- . | Post- . | Nulliparous . | Parous . | <25 . | ≥25 . | Greend . | Less Green . |
Reference (none) | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
1–2 days/week | 0.99 (0.94–1.05) | 0.92 (0.86–0.99)e | 0.97 (0.91–1.04) | 0.93 (0.87–0.98)e | 0.95 (0.90–0.99)e | 0.91 (0.80–1.05) | 0.87 (0.8–0.94) | 0.97 (0.91–1.03) |
3–5 days/ week | 0.95 (0.88–1.02) | 0.93 (0.85–1.01) | 0.93 (0.85–1.02) | 0.91 (0.85–0.98)e | 0.89 (0.84–0.95)e | 0.84 (0.71–1.01) | 0.82 (0.74–0.9) | 0.92 (0.85–0.99) |
Everyday | 0.87 (0.79–0.97)e | 0.94 (0.85–1.04) | 0.94 (0.83–1.07) | 0.88 (0.80–0.97)e | 0.88 (0.81–0.95)e | 0.78 (0.61–0.99)e | 0.78 (0.68–0.89) | 0.91 (0.82–1.01) |
. | Menopausea . | Parityb . | BMIc . | Greenness . | ||||
---|---|---|---|---|---|---|---|---|
. | Pre- . | Post- . | Nulliparous . | Parous . | <25 . | ≥25 . | Greend . | Less Green . |
Reference (none) | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
1–2 days/week | 0.99 (0.94–1.05) | 0.92 (0.86–0.99)e | 0.97 (0.91–1.04) | 0.93 (0.87–0.98)e | 0.95 (0.90–0.99)e | 0.91 (0.80–1.05) | 0.87 (0.8–0.94) | 0.97 (0.91–1.03) |
3–5 days/ week | 0.95 (0.88–1.02) | 0.93 (0.85–1.01) | 0.93 (0.85–1.02) | 0.91 (0.85–0.98)e | 0.89 (0.84–0.95)e | 0.84 (0.71–1.01) | 0.82 (0.74–0.9) | 0.92 (0.85–0.99) |
Everyday | 0.87 (0.79–0.97)e | 0.94 (0.85–1.04) | 0.94 (0.83–1.07) | 0.88 (0.80–0.97)e | 0.88 (0.81–0.95)e | 0.78 (0.61–0.99)e | 0.78 (0.68–0.89) | 0.91 (0.82–1.01) |
aAdjusted for age, body mass index, parity, family history of breast or ovarian cancer, hormone therapy use, smoking status, alcohol consumption, and calendar year.
bAdjusted for age, body mass index, menopausal status, family history of breast or ovarian cancer, hormone therapy use, smoking status, alcohol consumption, and calendar year.
cAdjusted for age, menopausal status, parity, family history of breast or ovarian cancer, hormone therapy use, smoking status, alcohol consumption, and calendar year.
dNDVI > 0.25.
eStatistically significant at α = 0.05.
The sensitivity analyses largely confirmed the results of the primary analyses despite a slight attenuation of the risk estimates (Table 4). Results from the analyses using the lagged variable for physical activity remained similar (Table 5). The application of different correlation structures produced similar results to the main analysis (see the Supplementary Table S1).
Adjusted ORs of high mammographic density in relation to physical activity in Japanese women in an ordinal scale, 2004 to 2019.a
. | OR . |
---|---|
Reference (none) | 1 |
1–2 days/week | 0.97 (0.93–1.02) |
3–5 days/week | 0.96 (0.91–1.02) |
Everyday | 0.92 (0.86–0.99)b |
. | OR . |
---|---|
Reference (none) | 1 |
1–2 days/week | 0.97 (0.93–1.02) |
3–5 days/week | 0.96 (0.91–1.02) |
Everyday | 0.92 (0.86–0.99)b |
aAdjusted for age, body mass index, menopausal status, parity, family history of breast or ovarian cancer, hormone therapy use, smoking status, alcohol consumption, and calendar year.
bStatistically significant at α = 0.05.
Adjusted ORs of mammographic density in relation to physical activity in Japanese women, 2004 to 2019 by lagged year.a
. | Previous yearb . | Two previous yearsc . |
---|---|---|
Reference (none) | 1 | 1 |
1–2 days/week | 0.97 (0.92–1.03) | 0.94 (0.89–1.00) |
3–5 days/week | 0.95 (0.89–1.01) | 0.93 (0.87–1.00) |
Everyday | 0.91 (0.84–1.00) | 0.90 (0.82–0.99)d |
. | Previous yearb . | Two previous yearsc . |
---|---|---|
Reference (none) | 1 | 1 |
1–2 days/week | 0.97 (0.92–1.03) | 0.94 (0.89–1.00) |
3–5 days/week | 0.95 (0.89–1.01) | 0.93 (0.87–1.00) |
Everyday | 0.91 (0.84–1.00) | 0.90 (0.82–0.99)d |
aAdjusted for age, body mass index, menopausal status, parity, family history of breast or ovarian cancer, hormone therapy use, smoking status, alcohol consumption, and calendar year.
bLagged associations with self-reported physical activity in the year before mammography.
cLagged associations with self-reported physical activity in the two years before mammography.
dStatistically significant at α = 0.05.
Discussion
In this large longitudinal study, we found a suggestive association of increasing physical activity with lower mammographic density among Japanese women after controlling for age, BMI, menopausal status, parity, family history of breast or ovarian cancer, hormone therapy use, smoking status, and alcohol consumption. The inverse association of physical activity with mammographic density was consistent across the subgroup analyses by menopausal status, BMI, and parity. We also found that the association was stronger in the green environment.
A substantial body of literature supports that increased physical activity reduces risk of breast cancer (17, 31–34). In terms of mammographic density, however, majority of studies have reported null associations with physical activity (35–37). Results from randomized controlled trials have shown mixed results, with protective effects of physical activity in a study with longer duration (14, 38). Studies with null associations tend to have relatively small sample sizes of a few hundreds, short exposure/intervention period, or cross-sectional study design (39, 40). Studies with a sizeable sample size, longer follow-up period, and/or the use of absolute measure for the amount of breast dense tissue tend to produce results with protective effects of physical activity (41, 42). Further studies with a large sample size and a long follow-up duration are warranted to better evaluate whether physical activity reduces mammographic density.
The hormone estrogen plays an essential role in the development of high mammographic density by stimulating breast epithelial or stromal proliferation (43–45). Physical activity can lower estrogen concentrations by reducing body fat, and subsequently lowering mammographic density (13). Several studies have demonstrated that high levels of physical activity were associated with lower level of urinary estrogens and estrogen metabolites in both premenopausal and postmenopausal women (46, 47). The inverse association was also observed when serum samples were used in postmenopausal women (48, 49).
The inverse relationship of physical activity with mammographic density was consistent across strata of BMI. Yet, the high BMI group showed a stronger association than the low BMI group. This result is broadly in agreement with findings from the prior study (19). Interestingly, women living in greener areas showed a more prominent association compared with those living in less green areas. Studies have shown that neighborhood greenness is associated with increased physical activity (27, 50), of which have demonstrated a positive association of greenness with physical activity regardless of the income level (28, 51). Greenness also mitigates exposure to environmental pollutants such as air pollution (52), which may in part explain this finding (53). This result implies that increasing green space is beneficial to promote breast health at the population level.
Major strength of our study lies in its large sample size, which makes our study discern from previous studies. The longitudinal data also enabled us to account for repeated measurement of mammographic density and time-varying covariates.
The main limitation is that physical activity was measured by answering a brief questionnaire, which can be rather subjective. Because physical activity is usually over-reported in questionnaires, potential misclassification of physical activity may have occurred. However, women had reported their physical activity level before they got a mammogram without knowing their mammographic density. Therefore, they were blind with respect to their mammographic density and error in their report occurs regardless of outcome status. Thus, we speculate that the associations were attenuated (54). Future studies of this population will benefit from more objective measure of metabolic equivalent task hours per week (55). Otherwise, the growing use of wearable devices such as wrist accelerometers may help to enhance quantify physical activity (56). The use of the BI-RADS for mammographic density is also posing a limitation to outcome assessment. Although the BI-RADS categorization is widely used in clinical practice, it relies on visual and subjective assessment by radiologists (1). Future studies would benefit from the use of quantitative assessments such as volume-based or area-based percent density measures (2). Small effect estimates can also be challenged. Although our analyses were adjusted for several covariates, we cannot rule out the possibility of unmeasured confounding as a driver of our results such as age at first birth, breastfeeding history, the use of oral contraceptive and other socioeconomic factors.
Asian women are known to have denser breasts than other ethnic groups (6), and yet, have a lower risk of developing breast cancer compared with White or Black women (57). One may interpret this inter-race comparison to mean that mammographic density does not play an important role in developing breast cancer among Asian women. When studies are conducted solely in Asian women populations, however, mammographic density is still a strong risk factor for breast cancer (58, 59). Therefore, decreasing mammographic density benefits Asian women as well as the Western counterpart. Our result suggests that increasing physical activity may serve as a reasonable intervention to reduce mammographic density, and thereby, to mitigate the risk of breast cancer in Asian women.
Authors' Disclosures
M. Lee reports grants from St. Luke's International Hospital during the conduct of the study. H. Yamauchi reports grants from AstraZeneca and Eiken Kagaku outside the submitted work. No disclosures were reported by the other author.
Authors' Contributions
M. Lee: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. R. Kotake: Data curation. H. Yamauchi: Resources.
Acknowledgments
This study was supported by the St. Luke's Breast Cancer Charity Fund. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the funding organization.
Note: Supplementary data for this article are available at Cancer Epidemiology, Biomarkers & Prevention Online (http://cebp.aacrjournals.org/).