A meta-analysis was undertaken to quantitatively determine if men with low selenium levels were at increased risk of prostate cancer. PubMed, EMBASE and current contents were searched to identify relevant studies. The effect size was calculated by pooling the mean difference for serum, plasma and toenail selenium levels (95% confidence intervals) separately and combined using a random effects model. Meta-regression analysis explored possible sources of heterogeneity. Twenty epidemiologic studies were selected. Mean differences were: -5.55 microg/l (-9.82; -1.27; p=0.01), -0.01 microg/g (-0.03; 0.006; p=0.19), -0.52 microg/l (-4.63; 3.58; p=0.80) for serum, toenail and plasma studies, respectively. Overall, the pooled standardized mean difference between cases and controls was; -0.23 (-0.40; -0.05; p=0.01) indicating a possible inverse association between selenium levels and risk of prostate cancer. Differences in selenium levels between populations, a possible threshold effect and the relationship between selenium and the different stages of prostate cancer require further investigation.