Comprehensive Summary
This study evaluates the clinical utility of an artificial intelligence–powered prognostic tool, PATHOMIQ_PRAD, to accurately predict the risk of biochemical recurrence (BCR) and distant metastasis (DM) using hematoxylin and eosin (HE)-stained whole-slide images from biopsy or radical prostatectomy (RP) specimens for risk stratification in patients with intermediate-risk prostate cancer (PCa). The authors retrospectively analyzed PATHOMIQ_PRAD scores from a cohort of 176 patients who underwent RP without adjuvant therapy at the Icahn School of Medicine at Mount Sinai. Patients were stratified into high- and low-risk categories for BCR and DM based on predefined score thresholds. The Kaplan-Meier method with a log-rank test was used to compare BCR-free survival and metastasis-free survival. Univariate results for the entire cohort show that PATHOMIQ_PRAD effectively identified patients at high risk of BCR (hazard ratio 4.347; p < 0.0001) and metastasis (hazard ratio 4.656; p < 0.0005). A subset of 129 patients with prognostic genomic scores was analyzed to compare PATHOMIQ_PRAD against existing risk stratification tools; DCA results for the 3-yr and 5-yr probability of BCR and the 5-yr probability of metastasis confirmed that PATHOMIQ_PRAD offers a superior net benefit to existing nomograms including CAPRA-S and genomic scores.
Outcomes and Implications
These findings suggest that PATHOMIQ_PRAD and similar AI-powered tools could augment clinical decision-making by improving risk stratification for intermediate-risk PCa patients. Within this patient population, a majority lack well-defined treatment strategies, and a significant proportion will experience disease progression after initial treatment. Given this, the ability to identify patients at higher risk of progression is critical for enabling timely intervention with adjuvant therapies following initial treatment, thereby reducing recurrence and metastasis rates and improving overall health outcomes. By incorporating AI-driven insights into clinical workflows, PATHOMIQ_PRAD supports precision medicine-based approaches by enabling more personalized treatment plans which tailor interventions to a patient's specific risk profile. If integrated into clinical practice, PATHOMIQ_PRAD could standardize risk assessment for this patient population, providing physicians with a data-driven framework for more effective treatment planning.