Physical activity and breast cancer risk among Asian-American women in Los Angeles

A case–control study

Authors

  • Dongyun Yang Ph.D.,

    1. Department of Preventive Medicine, Keck School of Medicine, University of Southern California, Los Angeles, California
    Search for more papers by this author
  • Leslie Bernstein Ph.D.,

    1. Department of Preventive Medicine, Keck School of Medicine, University of Southern California, Los Angeles, California
    Search for more papers by this author
  • Anna H. Wu Ph.D.

    Corresponding author
    1. Department of Preventive Medicine, Keck School of Medicine, University of Southern California, Los Angeles, California
    • Department of Preventive Medicine, Keck School of Medicine, University of Southern California, 1441 Eastlake Avenue, MC9175, Los Angeles, CA 90089-9175
    Search for more papers by this author
    • Fax: (323) 865-0139


Abstract

BACKGROUND

To the authors' knowledge, there have been few studies published to date regarding physical activity patterns and breast cancer risk in Asian and Asian-American women.

METHODS

The authors conducted a population-based case–control study of 501 Asian-American women with incident breast cancer and a control group of 594 Asian-American women in Los Angeles County to evaluate the role of lifetime physical activity on breast cancer risk. Information concerning lifetime recreational physical activity (i.e., type of activity, duration [years], and frequency [average hours per week]) and occupational physical activity was obtained using a structured questionnaire that was administered in person.

RESULTS

Increasing years and levels (average metabolic equivalent [MET] hours per week) of lifetime recreational activity were associated with a significantly reduced risk of breast cancer after adjusting for demographic factors, migration history, and menstrual and reproductive factors. Compared with women who had no lifetime recreational physical activity, ≤ 3 MET hours per week, > 3–6 MET hours per week, > 6–12 MET hours per week, and > 12 MET hours per week of activity were associated with significantly reduced risk, with odds ratios (and 95% confidence intervals) of 0.91 (0.55–1.49), 0.65 (0.39–1.10), 0.53 (0.31–0.90), and 0.47 (0.28–0.80), respectively (P value for trend < 0.001). The risk of breast cancer was associated inversely with occupational physical activity, although the result was not statistically significant.

CONCLUSIONS

The findings of the current study provide further support for the finding that physical activity has a protective role in breast cancer. Cancer 2003;10:2565–75. © 2003 American Cancer Society.

DOI 10.1002/cncr.11364

Breast cancer is the leading malignancy diagnosed among women across all major racial/ethnic groups in the U.S.1 Incidence rates are highest among white and African-American women, intermediate among Asian-American and Pacific Islander women, and lowest among American Indian/Alaska Native and Hispanic women.1, 2 Between 1992 and 1998, the largest increase in breast cancer incidence was observed among Asian-American and Pacific Islander women.2

Lifetime exposure to estrogens has been established as a major underlying determinant for breast cancer.3 However, menstrual and reproductive events (e.g., early age at menarche, nulliparity) associated with higher endogenous estrogen profiles are not modified easily. Physical activity is more amenable to change and, thus, is more promising as a primary preventive strategy for breast cancer. Physical activity may influence endogenous estrogen levels by delaying menstrual periods and by lowering the conversion of androgens to estrogens in body fat.4–6 Low-to-moderate intensity physical activity may improve immune function and lower the subsequent risk of developing breast cancer.6

Since the first epidemiologic study on physical activity and breast cancer risk was published in 1985,7 at least 43 epidemiologic studies have provided additional evidence. Greater than half of those studies, including cohort7–14 and case–control studies,15–33 reported reduced risk associated with occupational and/or recreational physical activity. However, many details regarding this association have not been settled. Findings regarding the optimal intensity, duration, frequency, and timing (i.e., period of life) of physical activity are not consistent. Relatively few studies have assessed lifetime physical activity in relation to risk of breast cancer.15, 24, 29, 31, 34

The vast majority of published studies were conducted among white women residing in western countries. Five studies conducted in Japan17, 22, 23 and China8, 33 have investigated physical activity patterns and breast cancer risk. Although these results generally are supportive of an inverse association between physical activity and breast cancer risk, only one study33 included information on lifetime physical activity patterns. In a case–control study of colorectal cancer that we conducted among Chinese men and women in North America and China in the 1980s,35 we found markedly reduced physical activity in adulthood among Chinese control participants who resided in North America compared with their counterparts in China. The extent to which physical activity patterns change with migration and the impact of physical activity on breast cancer risk among Asian-American women has not been studied. The current analysis provides new information on physical activity and breast cancer risk among Asian-American women in Los Angeles County, California.

MATERIALS AND METHODS

Study Population

This population-based, case–control study included Chinese, Japanese, and Filipino women who were between the ages of 25 years and 74 years at the time they were diagnosed with primary incident breast carcinoma on or after January 1, 1995 through 1997.36 Cases were identified through the Los Angeles County Cancer Surveillance Program, a population-based cancer registry that is part of the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Program and the statewide California Cancer Registry. Of the 871 Chinese, Japanese, and Filipino women who were identified, 523 women were interviewed (22 women had partially complete interviews and were excluded from the current analysis), 244 women declined to be interviewed (198 women refused, and 46 physicians refused to give permission to contact the patient), 11 women were deceased, and 93 women could not be located or had moved outside of Los Angeles County. Women in the control group (n = 594 women) were selected from the neighborhood in which women in the case group resided at the time of diagnosis. A well-established algorithm was used to identify neighborhood control participants that we have used in other case–control studies in Los Angeles County.37 This algorithm defined a specified sequence of houses to be visited in the neighborhood where the women lived at the time of their diagnosis. We sought to interview as the control the first eligible resident in this sequence. If the first eligible control participant refused to participate, then the second eligible control participant in the sequence was asked, and so on. Letters were left when no one was home and were followed up by mail and telephone (if a telephone number was determined). Control participants were frequency matched to cases on a specific basis of Asian ethnicity and 5-year age group). The average number of households walked before a suitable control participant was interviewed was 67.0 households. Sixty-eight percent of the control participants interviewed represented the first eligible control identified. However, 18% of the control participants interviewed had 1 refusal, and 14% of the control participants interviewed had ≥ 2 refusals.

Information Collected at Interview

All study participants completed a detailed, in-person, standardized, structured interview that was administered by bilingual interviewers. Signed informed consent was obtained from each participant, and study procedures were approved by the University of Southern California Institutional Review Board, in accordance with assurances approved by the U.S. Department of Health and Human Services. The information collected included demographic characteristic, migration history, menstrual and reproductive history, family history of cancer, diet, lifetime recreational physical activity and occupational history, body size at various ages, and lifetime smoking and alcohol use. Questions regarding migration history were adapted from those we have used in a previous study of breast cancer in Asian-Americans.38 The nondietary lifestyle factors mentioned above were similar to those used in questions for the Women's Contraceptive and Reproductive Experiences Study.39 The dietary questionnaire was modeled after the validated diet instrument used in the multiethnic cohort study conducted in Hawaii and Los Angeles.36

Assessment of Physical Activity

Participants were queried about recreational physical activities in which they had participated regularly (at least 1 hour per week or 52 hours per year) from age 10 years to the reference age (1 year before diagnosis for cases and 1 year before interview for control participants). Participation in physical education classes during school years was included. For each activity, a series of questions was asked that included the type of activity, the ages at start and stop, duration (years engaged in the activity and average months per year), and frequency of activity (average hours and minutes per week of activity). Separate episodes of an activity were recorded if women stopped the activity for ≥ 1 year and then continued. Our assessment methods of lifetime recreational physical activities were similar to those used by Bernstein and colleagues15 that received the highest rating in terms of data quality.40

Summary recreational physical activity variables were calculated for each participant from age 10 years to the reference age. They included 1) total years, 2) average hours per week, and 3) average metabolic equivalent (MET)-hours per week of physical activity. Years of activity from age 10 years to the reference age were computed by accumulating all years in which the woman reported any activity of at least 1 hour per week but excluding overlapping years of activity (i.e., if a woman participated in more than one activity during a particular year, then this was counted as only 1 year of activity). The average numbers of hours per week of activity were calculated by summing the products of duration and frequency of each activity divided by the total number of years. The average MET hours per week were computed by summing the average hours of activity (duration × frequency) multiplied by a MET value assigned to each type of physical activity and dividing by the total number of years. A MET value for a specific activity measured its intensity, i.e., the metabolic energy expenditure in that activity compared with the resting metabolic rate.41 For example, walking at moderate speed (< 3.5 miles per hour [mph]) was assigned a 3-MET value, which indicates that the energy expenditure for walking is 3 times that for sitting quietly. Average MET hours per week reflect the hours of physical activity per week weighted by a measure of intensity. Participation in recreational physical activity by intensity of activities (low MET scores of 3.0–4.0 vs. medium to high MET scores of > 4.0) and type of activities (walking as the only activity vs. other than walking) was examined separately. Finally, participation in physical activity during adolescence (ages 10–19 years) and adulthood (age 20 years to the reference age) was evaluated to assess timing of physical activity and its effects on risk.

Women were asked to identify the three jobs of longest duration that they held outside of the home for at least 1 year during their adult life. Job title and years of working were included. Each job title was coded using the 1980 Bureau of the Census three-digit occupation code and was classified into one of three activity categories: sedentary (code 0: a job that required activity < 20% of the time), moderately active (code 1: a job that required activity 20–80% of the time), and highly active (code 2: a job that required activity > 80% of the time), as rated by the U.S. Department of Labor.42 A weighted job activity code was created based on the assigned activity code and the corresponding number of years worked. Participants were then categorized into 1 of 4 job-related physical activity categories: sedentary (weighted code = 0: held a sedentary job only), mixed (weighted code > 0 and < 1: held both sedentary and moderately active jobs), active white collar jobs only (weighted code = 1: held moderately active jobs with occupational titles code < 400), and active blue collar jobs (weighted code ≥ 1: held at least 1 job rated as moderately active with occupational titles code ≥ 400 or highly active).

Statistical Analysis

Descriptive statistics were performed to characterize the study population and to examine differences between cases and controls and the three Asian groups of control participants. Conditional logistic regression with matched sets defined jointly by age group (≤ 39 years, 40–44 years, 45–49 years, 50–54 years, 55–59 years, 60–64 years, 65–69 years, and ≥ 70 years) and specific Asian ethnicity (Chinese, Japanese, and Filipino) was used to calculate odds ratios (ORs; relative risk estimates), their corresponding 95% confidence intervals (95% CIs) and two-sided P values. All regression models also included education (high school or less, some college, college graduate, and graduate), and migration history (born in the U.S., years lived in the U.S.: 21–60 years, 11–20 years, and 0–10 years) (Model 1 in Tables 2–5). In addition, a more fully adjusted model was employed (Model 2 in Tables 2–5) that also included parity (0, 1, 2, 3, and ≥ 4 live births), menopausal status (premenopausal or postmenopausal), first-degree family history of breast cancer (no or yes), and soy intake during adolescence and adult life (low in both periods, low intake during adolescence and high intake during adult life, high intake during adolescence and low during adult life, and high during both periods).36 Total caloric intake, body mass index (BMI; equal to weight [kg]/height [m2]) at age 18 years, BMI at reference age, weight gain between age 18 years and the reference age, age at the first live birth, and lifetime smoking and alcohol use were considered, but were not included, in the final model, because they did not seem to change the point estimates. Confounders were selected based on changes in point estimates of recreational and occupational physical activity variables by at least 10% in multivariate analyses compared with univariate analyses. Possible modifying effects by menopausal status and body size were suggested in some previous studies.9, 24, 28, 31, 32, 43 Potential modifying effects by Asian ethnicity, migration history, BMI at the reference age, total caloric intake, and menopausal status on an association between physical activity and breast cancer risk were examined using stratified models and were tested by comparing corresponding likelihood ratio statistics between the baseline and nested models that included the multiplicative product terms.44 Ordinal values were assigned to increasing levels of physical activity to assess trends in breast cancer risk. All analyses were performed using SAS statistical software (version 8.2; SAS, Inc., Cary, NC45) and Epilog Plus software (version 1.0; Epicenter Software, Pasadena, CA46).

Table 2. Adjusted Odds Ratios and Corresponding 95% Confidence Intervals for Breast Cancer in Association with Recreational Physical Activity of Any Type among Asian-American Women, Los Angeles County
All recreational physical activityCases/controlsModel 1ab ORModel 2ac
OR95% CI
  • OR: odds ratio; 95% CI: 95% confidence interval; MET: metabolic equivalent.

  • a

    Both Model 1 and Model 2 were conditional logistic regression models that were matched within 8 age groups (ages ≤ 39 years, 40–44 years, 45–49 years, 50–54 years, 55–59 years, 60–64 years, 65–69 years, and 70+ years) and three ethnic groups (Chinese, Japanese, and Filipino).

  • b

    Model 1 adjusted for education (≤ high school, some college, college graduate, and > college) and migration history (U.S. born and resided in the U.S. for 21–60 years, 11–20 years, and 0–10 years).

  • c

    Model 2 adjusted for the variables in Model 1 as well as parity (no live birth and 1, 2, 3, or 4 + live births), family history of breast cancer (no or yes), menopausal status (premenopausal or postmenopausal), years with active jobs (continuous) and job activity category (sedentary, mixed, active white collar only, and active blue collar), and soy intake during adolescence and adult life (low during both periods, low intake during adolescence and high intake during adult life, high intake during adolescence and low during adult life, and high during both periods). Adolescent soy intake was defined as low if it was less than weekly and high if it was weekly or more. Adult soy intake was defined as low if it included ≤ 6.24 mg of isoflavones per 1000 kcal and high if it included > 6.24 mg of isoflavones per 1000 kcal.

  • d

    From age 10 years to the reference age.

  • e

    For each year.

No. of yrsd    
 None53/441.001.00
 1–9169/1920.700.730.45–1.19
 10–19166/1850.700.760.47–1.25
 20–2946/970.360.350.20–0.63
 30 +50/720.460.480.27–0.87
 P for trend< 0.001< 0.001
Average hours per weekde    
 None53/441.001.00
 ≤ 0.75168/1630.780.840.51–1.37
 0.76–1.50114/1330.670.670.40–1.11
 1.51–3.0087/1330.540.560.33–0.94
 3.01 +62/1170.410.440.26–0.77
 P for trend< 0.001< 0.001
Average MET hours per weekde    
 None53/441.001.00
 ≤ 3163/1460.840.910.55–1.49
 > 3–699/1190.640.650.39–1.10
 > 6–1288/1350.520.530.31–0.90
 > 1281/1460.450.470.28–0.80
 P for trend< 0.001< 0.001
Table 3. Adjusted Odds Ratios and Corresponding 95% Confidence Intervals for Breast Cancer by Intensity and Type of Activities
VariableCases/controlsOR95% CIa
  • OR: odds ratio; 95% CI: 95% confidence interval; MET: metabolic equivalent.

  • a

    All ORs were adjusted for variables as in Model 2 (see footnotes a and c in Table 2).

  • b

    MET scores ≤ 4.0 indicate low-intensity activities; MET scores > 4.0 indicate medium or high-intensity activities.

  • c

    Walking includes walking at moderate speed (< 3.5 mph), walking briskly (3.5 mph), and heavy walking (e.g., walking uphill with a load).

  • d

    Women who ever participated in activities other than walking, as defined above.

Intensity of activityb
 None53/441.00
 1–9 yrs in low intensity only85/900.750.44–1.28
 10+ yrs in low intensity only63/580.820.46–1.46
 1–9 yrs in medium or higher intensity174/2210.660.41–1.07
 10 + yrs in medium or higher intensity109/1770.510.31–0.85
Activity type
Walking only (average MET hours per week)c
 None53/441.00
 ≤ 354/301.720.84–3.49
 > 3–629/380.630.31–1.27
 > 6–1226/350.600.29–1.24
 >1219/390.390.18–0.86
 P for trend0.002
Other than walking (average MET hours per week)d
 None53/441.00
 ≤ 3109/1160.720.42–1.23
 > 3–670/810.640.36–1.13
 > 6–1262/1000.440.25–0.78
 > 1262/1070.450.26–0.79
 P for trend< 0.001
Table 4. Adjusted Odds Ratios and Corresponding 95% Confidence Intervals for Breast Cancer in Association with Occupational Physical Activity
Occupational physical activityCases/controlsModel 1ab ORModel 2ac
OR95% CI
  • OR: odds ratio; 95% CI: 95% confidence interval; MET: metabolic equivalent.

  • a

    Both Model 1 and Model 2 were conditional logistic regression models that were matched within 8 age groups (ages ≤ 39 years, 40–44 years, 45–49 years, 50–54 years, 55–59 years, 60–64 years, 65–69 years, and 70 + years) and three ethnic groups (Chinese, Japanese, and Filipino).

  • b

    Model 1 adjusted for education (≤ high school, some college, college graduate, and > college) and migration history (U.S. born and resided in the U.S. for 21–60 years, 11–20 years, and 0–10 years).

  • c

    Model 2 adjusted for the variables in Model 1 as well as parity (no live birth and 1, 2, 3, or 4 + live births), family history of breast cancer, (no or yes), menopausal status (premenopausal or postmenopausal), average MET hours per week for all recreational activities (none, ≤3, >3–6, >6–12, and >12), and soy intake during adolescence and adult life (low during both periods, low intake during adolescence and high intake during adult life, high intake during adolescence and low during adult life, and high during both periods). Adolescent soy intake was defined as low if it was less than weekly and high if it was weekly or more. Adult soy intake was defined as low if it included ≤ 6.24 mg of isoflavones per 1000 kcal and high if it included > 6.24 mg of isoflavones per 1000 kcal.

  • d

    Additional adjustment for years of holding active jobs (continuous).

  • e

    Results were unchanged when we excluded women who reported no work outside the home in the reference group.

  • f

    Additional adjustment for weighted job activity category (continuous).

Job activity categoryd    
 Sedentary onlye117/1191.001.00
 Mixed148/1720.850.880.59–1.31
 Active white collar only106/1570.580.650.38–1.12
 Active blue collar113/1420.770.910.55–1.50
Years held active jobsf    
 0117/1191.001.00
 1–15172/2350.780.730.45–1.19
 16 +195/2360.710.630.34–1.18
 P for trend0.300.18
Table 5. Adjusted Odds Ratios and Corresponding 95% Confidence Intervals for Breast Cancer in Association with Recreational Physical Activity Stratified by Variables
VariableCases/controlsAll recreational activity (average MET hours per week): OR (95% CI)aP value
0–3> 3–6> 6–12> 12TrendInteraction
  • MET: metabolic equivalent; OR: odds ratio; 95% CI: 95% confidence interval; BMI: body mass index.

  • a

    All ORs were adjusted for variables as in Model 2 (see Table 2) except for the corresponding stratified variable.

Ethnicity       
 Chinese155/2261.000.60 (0.33–1.12)0.65 (0.35–1.20)0.42 (0.22–0.81)0.01
 Japanese144/1741.000.95 (0.47–1.94)0.84 (0.44–1.62)0.63 (0.33–1.21)0.16
 Filipino185/1901.000.67 (0.37–1.22)0.34 (0.17–0.64)0.47 (0.24–0.90)0.0020.60
Migration status       
 U.S. born/20 + yrs in U.S.271/3081.000.62 (0.38–1.01)0.48 (0.30–0.79)0.54 (0.33–0.86)0.004
 0–19 yrs in U.S.213/2821.000.75 (0.43–1.31)0.67 (0.39–1.17)0.41 (0.22–0.73)0.0030.52
Current BMI (kg/m2)       
 ≤ 21.5166/2011.000.86 (0.43–1.72)0.62 (0.33–1.19)0.58 (0.31–1.08)0.06
 21.6–24.5163/2001.000.72 (0.39–1.33)0.40 (0.21–0.76)0.35 (0.18–0.67)< 0.001
 ≥ 24.6151/1781.000.51 (0.26–0.99)0.64 (0.33–1.25)0.65 (0.30–1.39)0.180.60
Total caloric intake (kcal/day)       
 ≤ 1812233/2911.000.63 (0.36–1.09)0.55 (0.33–0.92)0.42 (0.25–0.73)< 0.001
 > 1812251/2991.000.72 (0.44–1.19)0.60 (0.36–1.02)0.62 (0.37–1.05)0.0430.99
Menopausal status       
 Premenopausal206/2881.000.81 (0.47–1.40)0.68 (0.40–1.16)0.44 (0.25–0.78)0.004
 Postmenopausal278/3021.000.56 (0.35–0.91)0.46 (0.28–0.76)0.55 (0.33–0.92)0.0030.35

RESULTS

This analysis included 501 cases (160 Chinese women, 146 Japanese women, and 195 Filipino women) and 594 controls (228 Chinese women, 175 Japanese women, and 191 Filipino women). On average, cases were 2 years older than controls but were similar in terms of birthplace, education, age at menarche, and BMI. Cases were significantly more likely to be nulliparous and had fewer live births. Details regarding these demographic, menstrual/reproductive, and lifestyle characteristics of participants were reported previously by Wu and colleagues.36

The distributions of select confounders according to recreational physical activity among control participants are shown in Table 1. Levels of recreational activity were associated significantly with age, Asian ethnicity, education, migration history, BMI, and parity. Compared with women who were inactive or had low levels of recreational physical activity, women with moderate or high levels of recreational physical activity were more likely to be younger, Japanese, more educated, born in the U.S., leaner, and nulliparous (P < 0.05; chi-square test). The recreational physical activity pattern was not influenced by caloric intake or occupational physical activity (Table 1), menopausal status (controlling for age), family history of breast cancer, or intake of soy (data not shown).

Table 1. Distribution (in percent) of Potential Confounding Factors by Levels of Recreational Activity among Women in an Asian-American Control Group
ConfounderRecreational physical activity (MET hours/week per year) (n = 590 women) (%)P value
0 (n = 44 women)≤ 3 (n = 146 women)> 3–6 (n = 119 women)> 6–12 (n = 135 women)> 12 (n = 146 women)
  1. MET: metabolic equivalent; BMI: body mass index.

Age (yrs)      
 ≤399.814.117.422.835.9
 40–497.023.018.823.927.2
 50–594.930.125.221.018.90.035
 60–695.431.518.526.118.5
 70+16.024.020.018.022.0
Race      
 Chinese12.421.720.821.723.5
 Japanese4.021.317.824.732.20.002
 Filipino4.731.621.622.619.5
Education      
 ≤ High school14.725.916.421.621.6
 Some college5.727.220.317.129.70.040
 College6.724.621.426.321.0
 > College3.319.621.726.129.3
Migration history  (yrs lived in U.S)      
 U.S. born3.123.517.326.529.6
 21–60 yrs4.125.326.721.921.90.004
 11–20 yrs12.022.024.719.322.0
 0–10 yrs11.428.811.423.525.0
No. of full-term births      
 Nulliparous4.021.017.025.033.0
 110.518.624.417.429.1
 26.721.921.925.324.20.015
 39.022.118.927.023.0
 4+7.741.318.316.316.3
Current BMI (kg/m2)      
 ≤ 21.59.519.916.924.928.9
 21.6–24.57.524.519.023.026.00.041
 ≥ 24.63.429.825.321.919.7
Total caloric intake  (kcal/day)      
 ≤ 18128.924.718.223.724.4
 > 18126.024.722.122.125.10.56
Job activity category      
 Sedentary5.027.713.426.926.9
 Mixed6.418.623.825.026.20.17
 Active white collar only7.025.523.621.722.3
 Active blue collar11.328.917.618.323.9
Years held active jobs      
 1–158.520.921.725.523.4
 16+7.627.122.018.225.00.19

The majority of women reported that they had participated regularly (i.e., at least 52 hours per year) in at least 1 type of recreational physical activity between age 10 years and the reference age. Compared with control participants who exercised, women in the case group exercised for fewer years (14.9 years for cases vs. 16.1 years for controls; P = 0.08), fewer hours per week (1.7 hours for cases vs. 2.3 hours for controls; P = 0.002), and less intensively (mean MET hours per week: 7.9 MET hours per week for cases vs. 11.2 MET hours per week for controls; P < 0.001). The five most frequently reported recreational physical activities were school-related physical education sports, walking at moderate speed, health club or home exercises at low intensity, moderate bicycling, and health club or home exercises at moderate intensity (listed in order beginning with the most frequent). Almost all participants had worked at least 1 year outside the home (n = 35 women who reported no work outside the home). Cases and control participants did not differ significantly with regard to the job activity category or years holding active jobs. The average years women held active jobs were 17.1 years and 16.6 years, respectively, among cases and controls who reported ever having an active job (P = 0.67). The three most frequently reported active, white-collar jobs were registered nurses, elementary school teachers, and managers and administrators. The three most frequently reported blue-collar jobs were sewers and stitchers (seamstress), nursing aides, and waitresses.

The overall associations between physical activity patterns and breast cancer risk are presented in Tables 2–4. Risk estimates that were obtained by using logistic regression models that included only demographic factors (Model 1) were very similar to risk estimates that were obtained with additional adjustment for menstrual characteristics, reproductive factors, family history of breast cancer, and dietary factors (Model 2). A significant trend of decreasing risk with increasing level of recreational physical activity was evident (all P for trend < 0.001). Risk patterns were similar for all three measures of reported activity: the number of years with participation in an activity, the average hours of activity per week, and the average MET hours per week. The risk of breast cancer decreased significantly in association with reported physical activity ≥ 20 years, a lifetime weekly average of at least 1.5 hours of activity, or a lifetime average of at least 6 MET hours per week (Table 2).

Risk patterns in relation to the intensity of recreational physical activity were examined in two ways. Analyses by intensity of activities (MET score ≤ 4.0 vs. > 4.0) suggested that the inverse associations with years of recreational activity were stronger for more intense activities (MET score > 4.0) compared with less intense activities (MET score ≤ 4.0) (Table 3), However, analysis by activity type (walking versus activities other than walking) suggested fairly comparable risk patterns among women who were engaged in at least > 3 MET hours per week (Table 3).

We assessed the joint effects of activity during adolescence and at later ages. Participants were classified into one of four categories: 1) no activity at any time, 2) no activity between age 10 years and 19 years but some activity since age 20 years, 3) some activity between age 10 years and 19 years but no activity since age 20 years, and 4) some activity in both periods. Compared with women who had no activity at all, women who reported some recreational physical activity during adolescence (ages 10–19 years) and adulthood (ages ≥ 20 years) showed a significant reduction in risk (OR, 0.57; 95% CI, 0.36–0.91; data not shown). Participation in recreational physical activity only during adolescence was associated with a statistically nonsignificant reduction in risk (OR, 0.63; 95% CI, 0.37–1.05). Based on only 66 cases and 34 controls, we found no risk reduction among women who participated in recreational physical activity during adulthood but did not participate in recreational physical activity at all during adolescence.

Occupational physical activity was related less strongly to breast cancer risk (Table 4). Modest reductions in risk (10–35%) were found among women who had active white collar or blue collar jobs compared with women who had only sedentary jobs. The risk of breast cancer also declined nonsignificantly with increasing numbers of years worked in a physically active job. Women who had active jobs for ≥ 16 years and exercised > 3 MET hours per week showed a significantly lower risk (OR, 0.44; 95% CI, 0.21–0.90) compared with women who did not have active jobs and who exercised < 3 MET hours per week (data not shown).

Stratified analyses were performed to evaluate possible effect modifications by menopausal status, specific Asian ethnicity, migration history, BMI at reference age, caloric intake, and menopausal status (Table 5). The pattern of risk reduction in premenopausal and postmenopausal women remained unchanged after further adjustment for BMI (data not shown). These factors did not modify the effects of occupational (data not shown) and recreational physical activity on breast cancer risk (all P values for interaction were > 0.05) (Table 5).

DISCUSSION

This is one of the first studies to investigate the association of physical activity and breast cancer risk among Asian-American women. In this population-based, case–control study, significant inverse associations between lifetime recreational physical activity and breast cancer risk were observed. The results for occupational activity were similar to the results for recreational physical activity. The risk of breast cancer declined with increasing duration (years), frequency (hours per week), and intensity (MET hours per week) of recreational physical activity. Although participation for ≥ 10 years in higher intensity activities (defined as activities with a MET score > 4.0) was associated with a more marked risk reduction compared with participation in lower intensity activities (MET score ≤ 4.0), there also was a benefit associated with substantial amounts of walking. The inverse association was most apparent for women who participated in recreational physical activity during both adolescence and adulthood. The associations between recreational physical activity and breast cancer risk were similar across subgroups of women defined by Asian ethnicity, migration history, BMI at reference age, caloric intake, and menopausal status.

Our study was limited, in that the response rates among cases and controls were modest. However, we have no reason to suspect that physical activity pattern was a determinant for participation among cases and controls, because our initial contact letters did not specifically mention physical activity. It is unlikely that we selectively recruited women for the control group who were at home and, thus, had more recreational physical activity, because > 98% of women in the control group had worked for at least 1 year outside of the home, and there was no correlation between recreational physical activity and job activity category (Table 1). It also should be noted that recruitment of the control group (i.e., control walking) occurred during the entire week. In addition, as part of the control selection algorithm, a letter of invitation was left at a residence if no one was at home at the time of visit, and two additional letters were mailed to the residence to identify an eligible control participant. An extensive, in-person interview was conducted that included assessment of lifetime recreational physical activity but did not include household activity patterns. The questions on recreational physical activity in this study were comparable to questions used in other studies:15, 20, 47 questions that experts40 have reported provide very high-quality information. Our finding of a 53% reduction in risk associated with at least 12 MET hours per week of physical activity is compatible with results of other case–control studies15, 24, 33 using similar measures of recreational physical activity.

A gradient reduction in breast cancer risk was observed with increasing lifetime duration (in years) of participation in recreational physical activity in the current study. Four other studies14, 24, 29, 33 have provided information on the correlations between duration of physical activity and breast cancer risk. A trend toward decreasing risk with increasing years of physical activity during adolescence and adulthood was found among Chinese women in Shanghai.33 Similar risk patterns in relation to total years of recreational physical activity were found in another study.24 However, no significant dose-response trends between years of recreational physical activity and risk of breast cancer were reported in two studies of Dutch women.14, 29 The correlation between type of recreational physical activity and breast cancer risk is of particular public health interest. Our findings suggest that some reduction in risk was found for women who participated in all types of activity, including women who reported walking for at least 3 MET hours per week. A Canadian case–control study48 found that occupational and household physical activities of low-to-moderate intensity were associated more strongly with breast cancer risk compared with vigorous activities. Physical activity categorized by intensity was not evaluated separately in other case–control studies.15, 24, 29, 33

A sizeable decline in breast cancer risk (OR, 0.57) was observed in association with recreational physical activity reported during both adolescence and adulthood. Heterogeneous results were reported in five case–control studies15, 24, 29, 33, 49 with data covering recreational physical activity during both adolescence and adulthood. In three studies,15, 33, 49 an inverse association was stronger for physical activity reported during later adult life; whereas, in other studies,24, 29 there were no apparent differences in breast cancer risk according to the timing of physical activity. In addition, 13 studies had data on physical activity patterns before age 20 years. Eight14, 19, 20, 27, 30, 47, 50, 51 of those studies found no significant association between physical activity reported during adolescent years and breast cancer risk. Of the five studies that found an inverse association with physical activity during adolescent years,18, 22, 25, 26, 52 results were statistically significant in two.18, 25

Heterogeneity in the effects of physical activity on breast cancer risk by periods of life may be due in part to the differences in the assessment and analysis of physical activity. In studies with information on lifetime physical activity, there were no standard questions relating to frequency, duration, intensity, or timing of activity. In the studies that did not ask about lifetime recreational physical activity, the specific periods of interest also differed.

Occupational physical activity was associated with some risk reduction, although this finding was not statistically significant. It should be noted that we did not obtain lifetime occupational histories but asked only about the three longest held jobs. At least eight previous studies8, 9, 11, 21, 25, 27, 31, 33 have found a significant, inverse association between the risk of breast cancer and reported occupational physical activity. Occupational physical activity was associated with a nonsignificantly reduced risk in two studies23, 29 and was unrelated to risk in three studies.14, 53, 54 Assessment of occupational physical activity also varied in those studies. In this and most previous studies, a level of occupational physical activity (i.e., sedentary, moderate, and high) was assigned based on the particiant's job title.8, 11, 14, 21, 23, 27, 29, 33, 54 A few studies asked the respondents to assess their level of occupational physical activity based on intensity of work or the number of hours engaged in the various activities at work (i.e., sitting, walking, or lifting).9, 25, 31, 53 Two studies31, 33 obtained lifetime occupational physical activity that included frequency, duration, and intensity of activities at work.

There is now a growing body of literature on physical activity patterns and the risk of breast cancer in non-white women. In this study, protective effects of physical activity on breast cancer risk were similar among Chinese, Japanese, and Filipino-American women. In five studies that were conducted among Asian women residing in Asia,8, 17, 22, 23, 33 a 30–65% risk reduction was observed comparing women with the highest levels versus the lowest levels of recreational or occupational physical activity. The risk of breast cancer also was associated inversely with physical activity among African-American women30 and Hispanic women in New Mexico32 and in the San Francisco Bay Area.55

Women in the control group who were born in the U.S. or were long-term residents in the U.S. reported higher levels of recreational physical activity compared with women who were more recent migrants (lived in the U.S. < 20 years) (Table 1). We are not aware of published data on physical activity patterns according to migration status among Asian Americans. In a study of colorectal cancer among Chinese in North America and China, older Chinese women who resided in China were more active physically compared with Chinese women who resided in North America, although but we did not evaluate whether physical activity patterns differed among North American Chinese women by birthplace or years of residence in the U.S.35 Our results show an inverse association between physical activity and breast cancer risk among both women who were born in the U.S./long-time migrants and women who were more recent migrants.

In the current study, BMI at the reference age did not modify the physical activity-breast cancer risk association, consistent with results from a Canadian study.31 Four other studies conducted among white women suggested a stronger association of physical activity with breast cancer risk among leaner women.9, 24, 27, 29 The relatively strong associations observed in the current study may have been related to the generally low BMI of Asian-American women who participated. The interrelations between body size, physical activity, and breast cancer risk are likely to be complex. Although there is accumulating evidence that breast cancer risk is increased in relation to central or upper body distribution, the relation between physical activity and body fat distribution has been seldom studied.56 Future investigations of the interrelations between physical activity (including household, occupational, and recreational sources), BMI, and body fat distribution in Asian and non-Asian populations may provide insights regarding the mechanism by which physical activity influences breast cancer risk.

The results of the current study show that physical activity is an important protective factor among Asian-American women. Increasing frequency, years, and intensity of physical activity were associated with a greater reduction in the risk of breast cancer. Physical activity promotion programs should be developed and tailored to the needs of girls as a public health recommendation, because girls are more likely than boys to become inactive during adolescent years.57

Acknowledgements

The authors thank all of the study participants for their support and contributions. They also thank the data collection team, particularly Betty DeBorja, Annie Fung, Diem Tran, Lydia Tran, and June Yashiki.

Ancillary