Most previous studies addressing the association of body size, weight change and body fat distribution with the risk of breast cancer were conducted in Western societies with a high proportion of overweight people. It remains unclear whether the dose-response relation observed in earlier studies can be extended to women with "normal" weight based on prevailing Western standards. To address this issue, we analyzed data from a population-based case-control study of breast cancer recently completed among Chinese women in urban Shanghai. In-person interviews and anthropometric measurements were completed for 1,459 women newly diagnosed with breast cancer from 25 to 64 years of age and 1,556 controls frequency-matched to cases on age. Unconditional logistic regression was employed to estimate adjusted odds ratios (ORs) and 95% confidence intervals (CI) related to anthropometric variables and self-reported body weight. Currently measured weight, body mass index [BMI: weight (kg)/height(m) 2 ] or height was each found to be positively related to risk of postmenopausal breast cancer in a doseresponse manner, with ORs (95% CI) being 2.0 (1.4 -3.0), 2.0 (1.2-3.2) or 1.7 (1.2-2.5), respectively, for the highest category of weight, BMI or height compared to the lowest category of these variables. These variables were unrelated to premenopausal breast cancer risk. Reported weight at ages >40 years and weight gain after age 20 were more predictive for postmenopausal breast cancer than weight at an earlier age. After adjustment for BMI, waist-to-hip ratio was related to an increased risk of premenopausal [OR ؍ 1.7 (1.3-2.3) for the highest category compared to the lowest category] but not postmenopausal breast cancer. This study suggests that, even in a relatively thin Chinese population, weight gain and height are related to an increased risk of postmenopausal breast cancer, while central fat distribution was associated with premenopausal breast cancer. General weight control may be an effective measurement for breast cancer prevention.