MMAI digital pathology AI biomarker predicts abiraterone benefit (CYP17 inhibition) in very high-risk M0 prostate cancer

MMAI is a locked multimodal AI model built from digital pathology plus PSA, T-stage, and age, tested post hoc across two sequential randomized STAMPEDE Phase 3 abiraterone comparisons in clinically very high-risk, non-metastatic prostate cancer (NCT00268476). Among 1,137 patients (LT-ADT n=583; LT-ADT+abiraterone n=554), the MMAI “very high-risk” subgroup (n=268) had a large metastasis-free survival (MFS) gain with abiraterone (HR 0.47; 95% CI 0.31–0.70), with 5-year MFS 62%→81%; in the MMAI “standard high-risk” subgroup (n=869), benefit was limited (HR 0.83; 95% CI 0.63–1.09), with 5-year MFS 82% vs 84% (interaction p=0.02). For context, in the overall STAMPEDE non-metastatic high-risk population, adding abiraterone to SOC improved 6-year MFS from ~69% to ~82% (HR ~0.53), but with higher rates of grade ≥3 hypertension and transaminitis than SOC—so a biomarker that concentrates benefit into a subset could materially improve the risk/benefit and cost equation.