When less is more: A simple predictive model for repeated prostate biopsy outcomes

ObjectivesTo present a new predictive model for repeated prostate biopsy outcomes. Several practical problems are described that arise when searching for a proper model among those that already exist. A new model is developed with only two explanatory variables and a simple graphical output.MethodsThis is a retrospective cohort study based on data collected from December 2006 to June 2011 at the Clinic of Urology of the University Hospital in Olomouc, Czech Republic. The cohort consists of 221 patients who underwent the first repeated biopsy after an initial biopsy with a negative outcome. All patients had prostate-specific antigen (PSA) levels between 1.5 and 16.5 ng/mL and a prostate volume not greater than 100 mL. A logistic regression model was fitted.ResultsOf the 221 patients, 29 (13%) were diagnosed with prostate cancer on the repeated biopsy. The final model includes the PSA level and the transitory zone volume as predictors. Its accuracy is 76.4%. The cut-off point of 0.0687 in the predicted positive repeated biopsy outcome assures 95% sensitivity and prevents 42% of unnecessary biopsies.ConclusionsThe accuracy of the model is comparable to that of more complex models (with more than two predictors) published in the literature. The model includes only two routinely measured variables, and hence it is accessible for a wide range of practitioners. The simple graphical outcome makes the model even more attractive.