The epidemiologic association between physical activity and breast cancer has been corroborated in many studies. Some inconsistencies remain, possibly due to variation in life periods for exposure assessment, confounding and undetected effect modification. In our cohort study, we address some of these questions by taking into account physical activity in different periods of life and by investigating effect modification by birth cohort and body mass index (BMI). Altogether 9,539 same-sex twin women aged 42-70 years who answered a questionnaire about their work and leisure-time physical exercise from ages 25 to 50 during 1967 and 1970 were included in our cohort. During follow-up, 506 breast cancer cases occurred through 1997. We used multivariate Cox models to estimate relative risk (RR) with 95% confidence interval (CI). We found no associations between physical activity and breast cancer overall. Women born between 1901 and 1917 (aged 51-70 at baseline) who reported regular leisure-time activity had a borderline significant 40% lower risk compared with those who reported no activity (RR 0.6; 95% CI 0.4 -1.0; test for trend, p ؍ 0.07). This association appeared to be confined to women with a low BMI after the age of 50 and to women with a high BMI during the premenopausal period. We found no evidence that work activity reduces risk for breast cancer. Physical activity might be one of the few realistic approaches toward primary prevention of breast cancer through changes in lifestyle. A protective effect is biologically plausible because physical activity may influence several factors relevant to breast carcinogenesis, such as menstrual characteristics, body size, the immune system 1 and the hormonal system including insulin-like growth factor I (IGF-I) levels. 2 Nevertheless, epidemiologic data in studies on physical activity and breast cancer risk remain inconclusive. [3][4][5][6][7][8][9][10][11][12] Although some studies show a clear protection, 5,7,8,[12][13][14][15][16][17] others reveal only a weak, often nonsignificant association 18,19 or no association at all. 10,11,20,21 Many factors could explain these discrepancies. First of all, obtaining reliable information about physical exercise throughout adult life is a complicated affair. Hence, misclassification of exposure could easily entail underestimation of any true association or cancel it altogether. In addition, several other breast cancer risk factors may modify the effect of physical activity on breast cancer risk, as indicated by recent studies. For example, we observed that physical activity reduced breast cancer risk only among women with a low body mass index (BMI) who were nonusers of postmenopausal hormone therapy. 15 Moreover, such therapy may increase breast cancer risk, chiefly in women with a low BMI. 22 Hence, if women with a low BMI are both more susceptible to exogenous hormones and more protected by physical activity, the protective effect of physical activity on breast cancer risk may be obscured among lean women who use exogenous...