This meta-analysis evaluated the effectiveness of metformin in subfertile women with polycystic ovary syndrome (PCOS). Only randomized trials investigating the effectiveness of metformin and PCOS definition consistent with the Rotterdam consensus criteria, were eligible. Primary outcome was live birth rate. A literature search identified 27 trials. In therapy naïve women, we found no evidence of a difference in live birth rate when comparing metformin with clomifene citrate (CC) [relative risks (RR) 0.73; 95% confidence interval (CI) 0.51-1.1] or comparing metformin plus CC with CC (RR 1.0; 95% CI 0.82-1.3). In CC-resistant women, metformin plus CC led to higher live birth rates than CC alone (RR 6.4; 95% CI 1.2-35); metformin also led to higher live birth rates than laparoscopic ovarian drilling (LOD) (RR 1.6; 95% CI 1.1-2.5). We found no evidence for a positive effect of metformin on live birth when added to LOD (RR 1.3; 95% CI 0.39-4.0) or FSH (RR 1.6; 95% CI 0.95-2.9), or when co-administered in IVF (RR 1.5; 95% CI 0.92-2.5). In IVF, metformin led to fewer cases of ovarian hyperstimulation syndrome (OHSS) (RR 0.33; 95% CI 0.13-0.80). This meta-analysis demonstrates that CC is still first choice therapy for women with therapy naïve PCOS. In CC-resistant women, the combination of CC plus metformin is the preferred treatment option before starting with LOD or FSH. At present, there is no evidence of an improvement in live birth when adding metformin to LOD or FSH. In IVF, metformin leads to a reduced risk of OHSS.