External Validation of the Rotterdam Prostate Cancer Risk Calculator and Comparison with Stockholm3 for Prostate Cancer Diagnosis in a Swedish Population-based Screening Cohort

Abstract

BackgroundThe Rotterdam Prostate Cancer Risk Calculator (RPCRC) and Stockholm3 can be used to aid urologists in their decision to refer men to magnetic resonance imaging (MRI) or biopsy for early detection of prostate cancer. ObjectiveTo assess the external validity of the RPCRC and compare it with using PSA and Stockholm3 to detect clinically significant prostate cancer. Design, setting, and participantsUsing data from the prospective, population-based, randomised STHLM3-MRI screening trial, we included participants with prostate-specific antigen (PSA) ≥3 ng/ml or Stockholm3 risk threshold ≥11% in the standard group who underwent systematic prostate biopsies. Outcome measurements and statistical analysisProbabilities for clinically significant prostate cancer (csPC, International Society of Urological Pathology grade ≥2) were calculated for each participant using the RPCRC and Stockholm3 with and without prostate volume. Performance of the risk calculators was assessed by discrimination, calibration, and clinical benefits. Results and limitationsIn total, 666 men with a median age of 67 yr (interquartile range [IQR]: 61–71) and PSA of 3.4 ng/ml (2.5–5.0) were included, of whom 154 (23%) had csPC. Risk distribution of the RPCRC was narrow: median risks of 2% (IQR 1–4%) compared with 14% (IQR: 9.5–23%) for Stockholm3. Using RPCRC’s recommended risk threshold of ≥4% for finding csPC, 54% of all csPC cases would be detected versus 94% using Stockholm3 with a threshold of ≥11%. Calibration of Stockholm3 was adequate while RPCRC underestimated the risk of csPC. The Stockholm3 test showed positive net benefits at clinically relevant thresholds, while the RPCRC showed negative net benefits. Compared with PSA, the RPCRC was associated with lower detection of csPC (84 vs 103; 0.82 [0.71–0.93]), while Stockholm3 was associated with higher detection of csPC (143 vs 103; 1.40 [1.23–1.57]). The main limitation was that Stockholm3 was evaluated in a similar population to where it was developed. ConclusionsThe performance of the RPCRC in a Swedish population-based cohort is suboptimal with a considerable underestimation of prostate cancer risk, while the Stockholm3 test showed superior performance and a positive clinical benefit. Patient summaryThe use of the Rotterdam Prostate Cancer Risk Calculator available online to predict the risk of prostate cancer in a Swedish cohort was found to be clinically harmful as it underpredicted the risk of clinically significant prostate cancer, while the Stockholm3 test performed well showing clinical benefits.