Incremental Value of Transrectal Ultrasound Target Visibility Beyond Clinical and MRI Variables for Predicting...

To determine whether three-level transrectal ultrasound (TRUS) target visibility adds predictive information beyond clinical and magnetic resonance imaging (MRI) variables for patient-level biopsy-detected clinically significant prostate cancer (csPCa).

This retrospective two-center study included 2,700 men undergoing cognitive transperineal biopsy. Center A provided development (n = 1,642) and temporal validation (n = 565) cohorts, and Center B provided external validation (n = 493). Gradient-boosting models were evaluated. TRUS visibility was retrospectively classified from contemporaneous biopsy-planning materials by readers blinded to pathology. Performance was assessed using AUC, average precision (AP), Brier score, quantitative calibration, exploratory decision-curve analysis, and interobserver reproducibility.

Adding visibility to the Clinical-MRI model produced modest but consistent improvement. In temporal validation, ΔAUC was 0.039 (95 % CI, 0.021-0.058), ΔAP 0.074 (0.034-0.116), and ΔBrier score - 0.016 (-0.023 to - 0.009); corresponding external differences were 0.023 (0.009-0.037), 0.045 (0.014-0.075), and - 0.016 (-0.023 to - 0.008). Calibration intercept/slope for the visibility model were - 0.074/1.160 temporally and 0.316/1.304 externally. Visibility-specific csPCa gradients were preserved in both validation cohorts, with no evidence of center-by-visibility heterogeneity. Interobserver agreement was high (linear-weighted Cohen's κ = 0.82).

TRUS target visibility provides modest incremental post-TRUS risk information beyond clinical and MRI variables. Because csPCa was defined at the patient level from targeted or systematic cores, the observed association cannot distinguish lesion biology from improved localization or sampling. Its potential role is risk refinement and sampling assistance, not biopsy avoidance or evidence of causal benefit.