Tailoring 4Kscore thresholds to multiparametric MRI results.

31 Background: 4Kscore was developed and validated agnostic to prostate multiparametric MRI (mpMRI) findings, with current clinical paradigms utilizing a value of 7.5% to delineate the potential for high-grade disease. A growing body of evidence suggests improved diagnostic utility when used in conjunction with mpMRI results. Incorporation of mpMRI PIRADS scores may have potential to enhance diagnostic accuracy of 4Kscore, especially for non-definitive lesions. This study aims to examine the optimal 4Kscore threshold in PIRADS 3/4 lesions to maximize test utility and help guide clinical decision-making. Methods: A single-institution review of all patients with recorded 4Kscore, prostate MRI with dominant PIRADS 3/4 lesions, and final biopsy pathology from 2016-2020 was conducted from an IRB-approved database. Clinically significant prostate cancer (csPCa) was defined as Gleason score ≥3+4 on final biopsy. Receiver operating characteristic curves were generated, and the primary data point was chosen to maximize sensitivity and specificity. Results: 88 patients with dominant PIRADS 3 (n = 40) or PIRADS 4 (n = 48) lesions were identified. For patients with PIRADS 3 lesions, area under the curve (AUC) was 0.8083 (p < 0.0039) with optimal 4Kscore threshold to detect csPCa as 18.5% (sensitivity = 70.00%, specificity = 80.00%). Negative predictive value (NPV) at 4Kscore of 18.5% in PIRADS 3 lesions was 0.93. In patients with dominant PIRADS 4 lesions, AUC was 0.7735 (p < 0.002), and optimal threshold to detect csPCa was 21.5% (sensitivity = 70.59%, specificity = 76.67%). NPV at 4Kscore of 21.5% in PIRADS 4 lesions was 0.82. Conclusions: Stratification of mpMRI PIRADS 3/4 lesions suggests that utilization of more permissive 4Kscore thresholds can improve prediction of csPCa without sacrificing NPV, especially for PIRADS 3 lesions. These findings may enhance risk-adapted prostate cancer screening, reduce unnecessary prostate biopsies, and optimize clinical management.