Abstract
PURPOSE: To externally validate recently published prostate cancer risk calculators (PCa-RCs) incorporating multiparametric magnetic resonance imaging (mpMRI) for the prediction of clinically significant prostate cancer (csPCa) and compare their performance to mpMRI-naïve PCa-RCs.
MATERIAL AND METHODS: Men without previous PCa diagnosis undergoing transperineal template saturation prostate biopsy with fusion-guided targeted biopsy between 11/2014 and 03/2018 in our academic tertiary referral center were identified. Any Gleason pattern ≥4 was defined to be csPCa. Predictors (age, PSA, DRE, prostate volume, family history, previous prostate biopsy and highest region of interest according to PIRADS) were retrospectively collected. Four mpMRI-PCa-RCs and two mpMRI-naïve PCa-RCs were evaluated for their discrimination, calibration and clinical net benefit using a ROC analysis, calibration plots and a decision curve analysis, respectively.
RESULTS: Out of 468 men, 193 (41%) were diagnosed with csPCa. Three mpMRI-PCa-RCs showed similar discrimination with area-underneath-the-receiver-operating-characteristic-curves (AUC) from 0.83 to 0.85, which was significantly higher than the other PCa-RCs (AUCs: 0.69-0.74). Calibration-in-the-large showed minimal deviation from the true amount of csPCa by 2% for two mpMRI-PCa-RCs, while the other PCa-RCs showed worse calibration (11-27%). A clinical net benefit could only be observed for three mpMRI-PCa-RCs at biopsy thresholds ≥15%, while none of the six investigated PCa-RCs demonstrated clinical utility against a biopsy all strategy at thresholds <15%.
CONCLUSIONS: Performance of the mpMRI-PCa-RCs varies, but they generally outperform mpMRI-naïve PCa-RCs in regard to discrimination, calibration and clinical usefulness. External validation in other biopsy settings is highly encouraged.
Saba, Karim