Decision analysis for treatment of early stage prostate cancer

We performed a decision analysis to evaluate the usefulness of pretreatment prediction of clinically significant or insignificant tumor in patients with prostate-specific antigen (PSA)-detected stage T1c prostate cancer nonpalpable on rectal examination. Analysis was done for otherwise healthy subjects with 20 years of life expectancy. The prevalence of insignificant tumor among those with T1c prostate cancer was initially assumed to be 0.2. Quality-adjusted life expectancy was calculated and compared between 2 strategies; one with prediction-based selection of either radical prostatectomy or watchful waiting and the other with unselective assignment of one of the treatments. The selection strategy was superior when the sensitivity and specificity for detecting clinically significant tumor were 0.92 and 0.73, respectively, as reported by Epstein et al. (1994) using criteria of PSA density and Gleason score in a needle biopsy specimen. Sensitivity analysis revealed that the prediction-based selection strategy is preferred, with sensitivity and specificity constant, when the prevalence of insignificant tumor exceeds 0.16. On the other hand, when the prevalence of insignificant tumor is kept constant at 0.2, sensitivity should be 0.85 or higher for the prediction strategy to be preferred. As the prevalence of insignificant tumor among those with T1c prostate cancer increased, the prediction-based selection strategy is preferred with lower values of sensitivity and specificity for detecting significant tumor. These results suggest that a selective treatment strategy of either radical or conservative treatment based on pretreatment prediction for significant tumor is a beneficial alternative to radical prostatectomy unselectively assigned to all patients at the T1c stage, if a reasonable accuracy in prediction is attained.