A Novel Biopsy-Free Predictive Model for Clinically Significant Prostate Cancer Incorporating Stratified PSA, PI-RADS, and PSMA PET-CT SUVmax: A Dual-Center Validation Study

Background The standard diagnostic pathway for prostate cancer necessitates a biopsy, an invasive procedure associated with potential complications. Identifying patients who can safely forego biopsy before radical treatment is an important clinical objective. Objective To develop and externally validate a preoperative prediction model for clinically significant prostate cancer (csPca) using stratified levels of serum prostate-specific antigen (PSA), PI-RADS scores from multiparametric MRI, and semi-quantitative PSMA PET-CT parameters (SUVmax). The goal is to define criteria for a safe biopsy-free approach to radical prostatectomy. Methods This retrospective, dual-center study analyzed data from 312 patients with suspected prostate cancer enrolled between January 2019 and June 2024. All patients underwent PSA testing, 3.0T multiparametric MRI, and ¹⁸F-PSMA-1007 PET-CT prior to any intervention. The reference standard was pathological examination of specimens from systematic combined targeted biopsies or radical prostatectomies. The cohort was randomly split 7:3 into a training set (n = 218) and an internal validation set (n = 94). Key predictors were categorized: PSA (< 10, 10–20, > 20 ng/mL); PI-RADS v2.1 (scores 3, 4, 5); and lesion SUVmax (< 4, 4–8, > 8). Multivariable logistic regression was employed to identify independent predictors of csPca (ISUP grade group ≥ 2) in the training set, leading to the construction of a nomogram. Model performance was rigorously assessed using the area under the receiver operating characteristic curve (AUC), calibration plots, and decision curve analysis (DCA). Furthermore, the diagnostic utility of various combination strategies of the three criteria was evaluated. Results Pathological examination confirmed prostate cancer in 256 patients (82.1%), with 220 cases (70.5%) classified as csPca. Multivariable analysis revealed that each stratified variable was an independent, strong predictor of csPca, demonstrating a clear gradient effect. Compared to their respective reference categories, the adjusted odds ratios were: for PSA 10–20, 3.86 (95% CI: 2.12–7.03); for PSA > 20, 6.94 (95% CI: 3.87–12.45); for PI-RADS 4, 5.73 (95% CI: 3.21–10.23); for PI-RADS 5, 12.46 (95% CI: 7.18–21.64); for SUVmax 4–8, 4.92 (95% CI: 2.68–9.03); and for SUVmax > 8, 10.87 (95% CI: 6.23–18.96) (all P < 0.001). The nomogram derived from these factors exhibited excellent discrimination, with an AUC of 0.934 (95% CI: 0.901–0.967) in the training cohort and 0.919 (95% CI: 0.876–0.962) in the validation cohort. Calibration was excellent, and DCA confirmed the model's clinical utility across a wide range of threshold probabilities. For clinical application, a "low-threshold" strategy (PSA ≥ 10, PI-RADS ≥ 4, SUVmax > 4) identified 106 patients, yielding a csPca positive predictive value (PPV) of 97.2% (103/106). A "high-threshold" strategy (PSA > 20, PI-RADS 5, SUVmax > 8) identified 56 patients with a csPca PPV of 100% (56/56). Conclusions A novel prediction model incorporating stratified PSA, PI-RADS, and PSMA PET-CT SUVmax accurately predicts the presence of csPca. The proposed low-threshold combination (PSA ≥ 10 ng/mL, PI-RADS ≥ 4, SUVmax > 4) identifies a substantial patient subgroup with a PPV exceeding 97%, for whom a biopsy-free approach to radical prostatectomy appears safe and justifiable. The high-threshold strategy offers absolute predictive certainty for select patients. This risk-stratified approach has the potential to streamline care pathways and reduce unnecessary invasive procedures in men with suspected high-risk prostate cancer.