Predictive models and risk of biopsy progression in active surveillance patients

Abstract

ObjectiveTo analyze the performance of different radical prostatectomy–based prognostic tools in predicting the biopsy progression in our active surveillance cohort. Materials and methodsWe analyzed 326 patients with biopsy Gleason grade≤6,≤2 positive biopsy cores,≤20% tumor present in any core, prostate-specific antigen<15ng/dl, and clinical stages T1–T2a all of whom had at least single surveillance biopsy. Probabilities of pathologically relatively aggressive disease were estimated using Partin and Dinh risk tables and Kattan, Truong, and Kulkarni nomograms for each individual patient. Using these predictions, performance of these tools was quantified regarding discrimination, stratification at different cut-points, calibration, and the clinical net benefit. ResultsPredictions of Partin and Dinh tables were not associated with the biopsy progression. The predictive value of Kattan and Truong nomograms was higher when compared with the other tools, although it was significant only on the first and second surveillance biopsies. Both nomograms were able to identify low- and high-risk subgroups within the cohort. Kattan nomogram demonstrated better correlation with the observed rate of progression over the first 3 biopsies and higher clinical net benefit. ConclusionKattan and Truong nomograms demonstrated the best performance in predicting biopsy progression, although their value was largely limited to the first 2 surveillance biopsies. Both tools were able to stratify patients into subgroups with different risks of progression. These nomograms have important differences, which suggest that a more effective predictive model combining the strong sides of both tools and possibly some other variables could be developed.