BACKGROUND: Most men with elevated levels of prostate-specific antigen (PSA) do not have prostate cancer, leading to a large number of unnecessary biopsies. A statistical model based on a panel of four kallikreins has been shown to predict the outcome of a first prostate biopsy. In this study, we apply the model to an independent data set of men with previous negative biopsy but persistently elevated PSA. METHODS: The study cohort consisted of 925 men with a previous negative prostate biopsy and elevated PSA (>= 3 ngml(-1)), with 110 prostate cancers detected (12%). A previously published statistical model was applied, with recalibration to reflect the lower positive biopsy rates on rebiopsy. RESULTS: The full-kallikrein panel had higher discriminative accuracy than PSA and DRE alone, with area under the curve (AUC) improving from 0.58 (95% confidence interval (CI): 0.52, 0.64) to 0.68 (95% CI: 0.62, 0.74), P<0.001, and high-grade cancer (Gleason >= 7) at biopsy with AUC improving from 0.76 (95% CI: 0.64, 0.89) to 0.87 (95% CI: 0.81, 0.94), P 0.003). Application of the panel to 1000 men with persistently elevated PSA after initial negative biopsy, at a 15% risk threshold would reduce the number of biopsies by 712; would miss (or delay) the diagnosis of 53 cancers, of which only 3 would be Gleason 7 and the rest Gleason 6 or less. CONCLUSIONS: Our data constitute an external validation of a previously published model. The four-kallikrein panel predicts the result of repeat prostate biopsy in men with elevated PSA while dramatically decreasing unnecessary biopsies. British Journal of Cancer (2010) 103, 708-714. doi:10.1038/sj.bjc.6605815 www.bjcancer.com Published online 27 July 2010 (C) 2010 Cancer Research UK