Several polymorphisms in the 5alpha-reductase type 2 (SRD5A2) gene have been implicated as risk factors for prostate cancer. We performed a meta-analysis of 9 studies (12 comparisons) with V89L genotyping (2558 prostate cancer cases and 3349 controls), 7 studies (8 comparisons) with A49T genotyping (1594 cases and 2137 controls), and 4 studies with TA repeat genotyping (1109 cases and 1378 controls). The random effects odds ratio (OR) for the L versus V allele was 1.02 [95% confidence interval (CI), 0.94-1.11]. There was no suggestion of an overall effect either in recessive or dominant modeling, and comparison of L/L versus V/V also showed no differential prostate cancer susceptibility (OR, 1.03; 95% CI, 0.83-1.28). The random effects OR for the T versus A allele was 1.56 (95% CI, 0.93-2.62). However, excluding the first published study there was no evidence for any effect (OR, 1.08; 95% CI, 0.72-1.61). Moreover, the T allele had a low prevalence (0%, 1%, and 2% in Asian, African and European controls, respectively). The random effects OR for longer versus short TA alleles was 0.88 (95% CI, 0.74-1.05). Longer TA allele homozygotes were nonsignificantly under-represented among prostate cancer cases (OR, 0.53; 95% CI, 0.26-1.06). We exclude a role for the V89L polymorphism in conferring susceptibility to prostate cancer. The A49T and TA repeat polymorphisms may have a modest effect on prostate cancer susceptibility, but bias and chance findings cannot be excluded; any genuine genetic effects would account only for a small proportion of prostate cancer in the population.