Prediction of prostate cancer: external validation of the ERSPC risk calculator in a contemporary Dutch clinical cohort

Background: The validity of prediction models needs external validation to assess their value beyond the original development setting. Objective: To report the diagnostic accuracy of the European Randomized Study of Screening for Prostate Cancer (ERSPC) risk calculator (RC)3 and RC4 in a contemporary Dutch clinical cohort. Design, setting, and participants: We retrospectively identified all men who underwent prostate biopsy (PBx) in the Jeroen Bosch Hospital, The Netherlands, between 2007 and 2016. Patients were included if they met ERSPC RC requirements of age (50–80 yr), prostate-specific antigen (PSA) (0.4–50 ng/ml), and prostate volume (10–150 ml). The probability of a positive biopsy for prostate cancer (PCa) and significant PCa (Gleason score ≥7 and/or higher than T2b) were calculated and compared with PBx pathology results. Outcome measurements and statistical analysis: Evaluation was performed by calibration, discrimination, and clinical usefulness using calibration plots, area under the receiver operating characteristic curves (AUCs), and decision curve analyses (DCAs), respectively. Results and limitations: A total of 2270 PBx sessions were eligible for final analysis. Discriminative ability of RC3 (AUC) was 0.78 and 0.90 for any PCa and significant PCa, respectively. For RC4 the calculated AUCs were 0.62 (any PCa) and 0.76 (significant PCa). The calibration plots of RC3 showed good results for both any PCa risk and significant PCa risk. In the repeat PBx group, RC4 tended to underestimate outcomes for PCa and showed moderate calibration for significant PCa. DCA showed an overall net benefit compared with PSA and digital rectal examination (DRE) alone. Limitations of this study are its retrospective single-institution design, retrospectively assessed DRE outcomes, no time restrictions between the first and repeat biopsy sessions, and no anterior sampling in the repeat PBx protocol. Conclusions: The ERSPC RCs performed well in a contemporary clinical setting. Most pronounced in the biopsy-naive group, both RCs should be favoured over a PSA plus DRE–based stratification in the decision whether or not to perform PBx. Patient summary: We looked at the ability of the existing European Randomized Study of Screening for Prostate Cancer risk calculator (RC), using different clinical data to predict the presence of prostate cancer in Dutch men. The RC performed well and should be favoured in the decision of whether or not to perform prostate biopsies over the conventional diagnostic pathway. The European Randomised Screening for Prostate Cancer risk calculators should be favoured over a prostate-specific antigen plus digital rectal examination–based stratification in the decision whether or not to perform prostate biopsy. An individualised approach for prostate cancer detection may reduce the adverse effects of our diagnostic approaches and/or treatments.