PURPOSE: To evaluate whether free prostate specific antigen (PSA), PSA velocity or urokinase can predict the result of biopsy for prostate cancer – and thus reduce unnecessary biopsies – by using decision curve analysis, a novel decision analytic technique. METHODS: Men with elevated PSA are typically referred for biopsy. The positive predictive value of an elevated PSA is low (~25%) so that approximately 750,000 unnecessary biopsies are conducted yearly in the US. Several additional markers have been proposed to improve detection of prostate cancer. We analyzed three data sets from studies of prostate cancer markers: free PSA, PSA velocity, and urokinase. We used two different approaches for evaluating the value of these markers: the change in predictive accuracy (as area under the ROC curve [AUC]) associated with adding the new markers to a model including PSA alonewe also used decision curve analysis. This is a decision analytic method based on the principle that the threshold probability of disease at which a patient would opt for action (e.g. biopsy) reflects his valuation of the relative harms associated with a false-positive and a false-negative prediction. This theoretical relationship is then used to derive the net clinical benefit of the model across different threshold probabilities. Plotting clinical benefit against a range of threshold probabilities yields the decision curve. We assumed that few men would opt for biopsy if their risk of cancer was <10% and all men would opt for biopsy if their risk was >40%. RESULTS: In all cases, the additional marker was an independent statistically significant predictor of a cancer diagnosis and, moreover, added importantly to the accuracy of the predictive model. The decision curve for free PSA was superior to the strategy of biopsying all or no men across the full range of threshold probabilities, urokinase was equivalent to the strategy of biopsying all men, and PSA velocity was superior only for thresholds between 15% - 35%. Hence, free PSA is of clear benefit, urokinase not clinically useful, and PSA velocity would help some, but not all men, depending on their threshold probability. CONCLUSIONS: Decision curve analysis, but not evaluation of predictive accuracy, was able to distinguish useful markers for prostate cancer from those less useful.