A multivariable prediction model combining 18F-PSMA PET/CT and mpMRI for clinically significant prostate cancer:...

Multiparametric MRI (mpMRI) and prostate-specific membrane antigen (PSMA) PET/CT provide complementary information for prostate cancer diagnosis, but optimal integration strategies remain unclear. We aimed to develop and validate a multivariable model combining clinical and imaging parameters to predict clinically significant prostate cancer (csPCa).

This study included 1305 consecutive patients with suspected prostate cancer who underwent both mpMRI and 18F-PSMA PET/CT with biopsy at Peking University First Hospital. The development and temporal validation cohorts were temporally divided; the development cohort was split into training (70%) and internal test (30%) sets. csPCa was defined as Gleason Grade Group ≥2. Predictors were selected using LASSO regression (1-standard error criterion) followed by forward stepwise logistic regression (AIC, p < 0.05). Model performance was assessed by AUC, calibration plots, and decision curve analysis.

The final model retained 3 predictors: PRIMARY score, PI-RADS score, and PSAD. The training AUC was 0.916. At the Youden Index cutoff (≥84%), sensitivity was 79.3% and specificity 76.6%; at the recommended screening cutoff (≥46%), sensitivity reached 96.0%. The AUC was 0.914 (95% CI: 0.882-0.941) in the internal test set and 0.837 (95% CI: 0.778-0.891) in temporal validation. The model showed good calibration (Brier score: 0.096) and superior clinical utility over individual imaging parameters.

The multivariable model integrating 18F-PSMA PET/CT and mpMRI parameters provides accurate risk stratification for csPCa and may help optimize biopsy decisions.