This study was performed to integrate the results of previous studies that investigated the relationship between body mass index (BMI) and prognosis in breast cancer. We reviewed the English literatures using the MEDLINE database from 1966 to 1999. The materials included 12 published articles with a total of 8,029 cases of breast cancer. The effect size was obtained from hazard ratio in each study. Homogeneity test was conducted before the integration of each effect size and the result demonstrated that the studies were heterogeneous. A random effect model was used to integrate the overall effect size. The integrated effect size was 1.56 (95% confidence interval, 1.22-2.00). In addition, publication bias should be accounted for because each published study was asymmetric in shape revealed by funnel plot. These results suggest that BMI have a prognostic significance in breast cancer. We believe that well-designed longitudinal studies, involving a large number of samples are required to resolve these issues.