Prospective evaluation on the effect of interobserver variability of digital rectal examination on the performance of the Rotterdam Prostate Cancer Risk Calculator

Objectives To assess the level of agreement between digital rectal examination findings of two urologists and its effect on risk prediction using the digital rectal examination-based Rotterdam Prostate Cancer Risk Calculator. Methods The study sample consisted of a prospective cohort of asymptomatic unscreened men with prostate-specific antigen ≤50.0 ng/mL and transrectal ultrasound volume ≤110 mL who underwent transrectal ultrasound-guided prostate biopsy. Both urologists’ digital rectal examination findings were graded normal or abnormal (nodularity and/or induration), and volume classified as 25, 40 or 60 mL, according to the risk calculator algorithm. Interrater agreement analysis using Cohen's kappa (κ) statistic was carried out to determine consistency of digital rectal examination outcome and volume assessment. Receiver operating characteristic curve analysis and calibration plots were constructed to determine the effect of interrater differences. Decision curve analysis was applied to evaluate the clinical usefulness of the model. Results Of the 241 men included in the study, 41% (n = 98) had prostate cancer (81 were clinically significant, i.e. Gleason ≥3 + 4). There was substantial agreement in the digital rectal examination (abnormal/normal; κ = 0.78; P < 0.001) and volume estimation (κ = 0.79; P < 0.001). Receiver operating characteristic analyses showed good discrimination (0.75–0.78) and were comparable for both urologists. In the high-risk cohort, at a probability threshold of 25%, the risk calculator reduced the prostate biopsy rate by 9%, without missing cancers. Conclusions Slight differences in digital rectal examination findings seem to have very limited impact on the performance of the Rotterdam Prostate Cancer Risk Calculator. Therefore, this can be considered a useful prostate biopsy outcome prediction tool.