BackgroundMany studies have reported the associations of polymorphic CAG repeats in androgen receptor (AR) gene with PCOS risk, but with inconsistent results. So, the aim of present meta-analysis was to clarify such inconsistence, so as to provide more conclusive results. MethodsPubMed was searched for the eligible reports published until February 2012 without language limitation. The studies reporting the relationship between CAG repeat length and PCOS were selected for the meta-analysis according to the inclusion criteria. Two reviewers independently extracted the data and evaluated the study quality. Principal findingsAs for the relationship between CAG repeat length and PCOS risk, the pooled results showed that the biallelic mean was not significantly different between PCOS and controls (SMD −0.03, 95% CI −0.16–0.10, P=0.603), and that the ORs of PCOS were not demonstrated for the individuals with the biallelic mean less than median (OR 0.96, 95% CI 0.68–1.35, P=0.794), with the short CAG allele (OR 0.94, 95% CI 0.80–1.10, P=0.424), or with the X-weighted biallelic mean (OR 0.81, 95% CI 0.46–1.41, P=0.447). Further, as for the relationship between CAG repeat length and T levels in PCOS patients, the biallelic mean was not significantly different between PCOS patients with high T and those with low T (SMD 0.79, 95% CI −0.12–1.70, P=0.088), while the summary correlation r indicated that the CAG biallelic mean appeared to be positively associated with T levels in PCOS (r 0.20, 95% CI 0.11–0.30, p=0.000). ConclusionsThis meta-analysis demonstrates no evident association between the CAG length variations in AR gene and PCOS risk, while the CAG length appears to be positively associated with T levels in PCOS patients.