Prostate-specific antigen (PSA) is a protease produced in the prostate that cleaves insulin-like growth factor binding protein-3 and other proteins. Production is mediated by the androgen receptor (AR) binding to the androgen response elements (ARE) in the promoter region of the PSA gene. Studies of a single nucleotide polymorphism (PSA -158 G/A, rs266882) in ARE1 of the PSA gene have been conflicting for risk of prostate cancer and effect on plasma PSA levels. In this nested case-control analysis of 500 white cases and 676 age- and smoking-matched white controls in the Physicians' Health Study we evaluated the association of rs266882 with risk and survival of prostate cancer and prediagnostic total and free PSA plasma levels, alone or in combination with AR CAG repeats. We used conditional logistic regression, linear regression and Cox regression, and found no significant associations between rs266882 (GG allele vs AA allele) and overall prostate cancer risk (RR=1.21, 95% confidence intervals (CI): 0.88-1.67) or prostate cancer-specific survival (RR=0.94, 95%CI: 0.56-1.58). Similarly, no associations were found among high grade or advanced stage tumours, or by calendar year of diagnosis. There was no significant association between rs266882 and baseline total or free PSA levels or the AR CAG repeats, nor any interaction associated with prostate cancer risk. Meta-analysis of 12 studies of rs266882 and overall prostate cancer risk was null.