The length of the polymorphic CAG repeat in the N-terminal of the androgen receptor (AR) gene is inversely correlated with the transactivation function of the AR. Some studies have indicated that short CAG repeats are related to higher risk of prostate cancer. We performed a case-control study to investigate relations between CAG repeat length and prostate cancer risk, tumour grade, tumour stage, age at diagnosis and response to endocrine therapy. The study included 190 AR alleles from prostate cancer patients and 186 AR alleles from female control subjects. All were whites from southern Sweden. The frequency distribution of CAG repeat length was strikingly similar for cases and controls, and no significant correlation between CAG repeat length and prostate cancer risk was detected. However, for men with non-hereditary prostate cancer (n = 160), shorter CAG repeats correlated with younger age at diagnosis (P = 0.03). There were also trends toward associations between short CAG repeats and high grade (P = 0.07) and high stage (P = 0.07) disease. Furthermore, we found that patients with long CAG repeats responded better to endocrine therapy, even after adjusting for pretreatment level of prostate-specific antigen and tumour grade and stage (P = 0.05). We conclude that short CAG repeats in the AR gene correlate with young age at diagnosis of prostate cancer, but not with higher risk of the disease. Selection of patients with early onset prostate cancer in case-control studies could therefore lead to an over-estimation of the risk of prostate cancer for men with short CAG repeats. An association between long CAG repeats and good response to endocrine therapy was also found, but the mechanism and clinical relevance are unclear.