The association of breast cancer with dietary patterns such as a western diet has not been studied in Asian women. We examined this among Shanghai Breast Cancer Study participants. Cases were of ages 25 to 64 years, diagnosed 08/1996-03/1998, and identified through a rapid case ascertainment system supplemented by the Shanghai Cancer Registry. Controls, selected from the general population of urban Shanghai, were frequency matched to cases by 5-year age group. Participants provided information on diet, lifestyle, and reproductive factors. In principal component analysis among 1,556 controls, two patterns emerged: a "vegetable-soy" pattern (tofu, cauliflower, beans, bean sprouts, green leafy vegetables) and a "meat-sweet" pattern (shrimp, chicken, beef, pork, candy, desserts). In adjusted unconditional logistic regression analyses including 1,446 cases and 1,549 controls with complete covariate data, risk was not associated with the vegetable-soy pattern. It was associated with the meat-sweet pattern (4th versus 1st quartile: odds ratio, 1.3; 95% confidence interval, 1.0-1.7; P(trend) = 0.03), but only in postmenopausal women, specifically among those with estrogen receptor-positive tumors (4th versus 1st quartile: odds ratio, 1.9; 95% confidence interval, 1.1-3.3; P(trend) = 0.03). Our findings indicate that a western diet increases breast cancer risk in postmenopausal Chinese women. They also suggest the value of quantifying aggregate risk for common combinations of foods.