Statistical analysis

Pooled ORs and 95% CIs for the associations between BMI, reproductive factors, and breast cancer molecular subtypes were calculated via dividing the observed frequencies of categories by the unified cutoff values of eligible studies. Heterogeneity among the included articles was estimated by a chi-square-based Q-test. If significant heterogeneity was not detected (P > 0.05), the Mantel-Haenszel method was used to estimate the pooled OR and 95% CI. Otherwise, the DerSimonian and Laird method was selected. The WLS method was used to quantify the heterogeneity of associations between risk factors and breast cancer subtypes. Egger's regression test was used to determine publication bias.23 Publication bias is considered to be present if the p-value of the intercept (a) is less than 0.05. Sensitivity analysis was performed via excluding the study with the largest weight and then evaluating the impact of the change on the pooled OR. Crude ORs with 95% CIs for each study, Q-tests, and pooled ORs with 95% CI were calculated using Review Manager 5.0 (Nordic Cochrane Center, Copenhagen, Denmark). The Egger's and WLS tests were performed with SPSS 17.0. The p-values were subjected to a two-tailed test, with an alpha level of 0.05 for significance testing.