All men who underwent a transrectal prostate biopsy in a European tertiary care centre between 2004 and 2012 were retrospectively identified. The probability of detecting prostate cancer and significant cancer (Gleason score ≥7) was calculated for each man using the novel versions of the ERSPC-RC (DRE-based version 3/4) and the PCPT-RC (version 2.0) and compared with biopsy results. Calibration and discrimination were assessed using the calibration slope method and the area under the receiver operating characteristic curve (AUC), respectively. Additionally, decision curve analyses were performed.
ResultsOf 1 996 men, 483 (24%) were diagnosed with prostate cancer and 226 (11%) with significant prostate cancer.Calibration of the two RCs was comparable, although the PCPT-RC was slightly superior in the higher risk prediction range for any and significant prostate cancer. Discrimination of the ERSPC-and PCPT-RC was comparable for any prostate cancer (AUCs 0.65 vs 0.66), while the ERSPC-RC was somewhat better for significant prostate cancer (AUCs 0.73 vs 0.70). Decision curve analyses revealed a comparable net benefit for any prostate cancer and a slightly greater net benefit for significant prostate cancer using the ERSPC-RC.
ConclusionsIn our independent external validation, both updated RCs showed less optimistic performance compared with their original reports, particularly for the prediction of any prostate cancer. Risk prediction of significant prostate cancer, which is important to avoid unnecessary biopsies and reduce overdiagnosis and overtreatment, was better for both RCs and slightly superior using the ERSPC-RC.