A Model for Predicting Clinically Significant Prostate Cancer using Prostate MRI and Risk Factors

Objective To develop and validate a predictive model for clinically significant prostate cancer (csPCa) using prostate MRI and patient risk factors. Methods We identified 960 men who underwent MRI from 2015-2019 and biopsy either 6 months prior or 6 months following the MRI. We identified men diagnosed with csPCa and modeled csPCa risk using known patient factors (age, race, PSA) and prostate MRI findings (location, PI-RADS score, extraprostatic extension, dominant lesion size, PSA density [PSAD]). csPCa was defined as Gleason Sum ≥ 7. Using a derivation cohort, a multivariable logistic regression model and point-based scoring system were developed to predict csPCa. Discrimination and calibration were assessed in a separate independent validation cohort. Results 552 of 960 MRI reports (57.5%) were from men diagnosed with csPCa. Using the derivation cohort (n=632), variables that predicted csPCa were PI-RADS 4 and 5, presence of extraprostatic extension, and elevated PSAD. Evaluation using the validation cohort (n=328) resulted in AUC of 0.77, with adequate calibration (Hosmer-Lemeshow p=0.58). At a risk threshold of >2 points, the model identified csPCa with sensitivity of 98.4%, negative predictive value (NPV) of 78.6%, but only prevented 4.3% (0-2 points; 14/328) potential biopsies. At a higher threshold of >5 points, the model identified csPCA with sensitivity of 89.5%, NPV of 70.1%, and avoided 20.4% (0-5 points; 67/328) of biopsies. Discussion Our point-based model can potentially identify a vast majority of men at risk for csPCa, while sparing ∼ 1 in 5 men with an elevated PSA, a biopsy.