Background: Prediction models need validation to assess their value outside the development setting.
Objective: To assess the external validity of the European Randomised study of Screening for Prostate Cancer (ERSPC) Risk Calculator (RC) in a contemporary clinical cohort.
Methods: The RC calculates the probability of a positive sextant prostate biopsy (P(posb)) using serum prostate-specific antigen (PSA), results of digital rectal examination, transrectal ultrasound (TRUS) and ultrasound assessed prostate volume. We prospectively validated the RC in 320 biopsied men (55-75 years), with no previous prostate biopsy, included in five Dutch hospitals in 2008-2011. If the P(posb) was ≥ 20% a biopsy was recommended. The performance of the RC was tested by comparing the observed outcomes to predicted probabilities, using the area under the curve (AUC) and decision curves analyses.
Results: Compared to the screening cohort, men in the clinical cohort differed. They had higher PSA levels (median 6.8 versus 4.3 ng/ml, p<0.01), less TRUS-lesions (27% versus 34%, p = 0.01) and more prostate cancer (PCa) at biopsy (43% versus 25%, p<0.01). Mainly eight biopsy cores were taken. Despite the differences between these cohorts, the mean observed probability agreed with the mean predicted probability (43% versus 40%). The RC predicted P(posb) better than a model with PSA and digital rectal examination, AUC 0.77 (95% confidence interval (CI) 0.72-0.83) and 0.71 (95%CI 0.65-0.76, p<0.01), respectively. This was confirmed by the decision curves analysis. Under the 20% threshold, 17% (11/63) of the biopsied men were diagnosed with PCa. Two of 11 men had an important cancer (Gleason 3+4).
Conclusions: The ERSPC RC performs well in a Dutch clinical cohort in men with previous PSA tests and contemporary biopsy schemes, and outperforms a PSA and DRE-based approach in the decision to perform a biopsy.
Copyright © 2012 Elsevier Ltd. All rights reserved.