Models to predict positive prostate biopsies using the Tyrol screening study.

Objective: To describe two predictive models that predict for prostate cancer on biopsy derived from a large screening population. There are no published predictive models that predict prostate cancer in a screened population.

Methods: The patients from the Tyrol screening study of known age, total prostate-specific antigen (PSA) level, digital rectal examination (DRE) findings, prostate volume, and percentage of free PSA, and who underwent an initial prostate biopsy from January 1992 to June 2004 were included (n = 2271). Multivariate logistic regression models were used to develop the biopsy positivity predictive models: nomogram 1, age, DRE, and total PSA; and nomogram 2, age, DRE, total PSA, and percentage of free PSA. The predictive accuracy of the models was assessed in terms of discrimination and calibration. External validation of the nomograms was performed using a urologically referred population of patients who underwent prostate biopsy (n = 599).

Results: Both nomograms were well-calibrated internally and externally and discriminated well between patients with positive and negative biopsy findings for both the European and U.S. cohorts (model 2 better than model 1).

Conclusion: Our nomogram with age, total PSA, and DRE had good predictive ability to differentiate between screened patients with cancer on the initial prostate biopsy and those without. Adding the percentage of free PSA improves this predictive power further. These models might aid in clinical decision making regarding the need for biopsy in both European and U.S. populations.