ST-DoxPCa H&E Biomarker Identifies mHSPC Patients Most Likely to Benefit From Docetaxel

ST-DoxPCa is an AI computational pathology biomarker that infers spatial gene-expression patterns from routine H&E to predict docetaxel benefit in metastatic hormone-sensitive prostate cancer, using a Vision Transformer trained on paired H&E/spatial transcriptomics and distilled to a 26-gene signature including AR, PTEN, FASN, FOXA1, and ACPP. In CHAARTED (ECOG-ACRIN E3805; NCT00309985), the OS model was trained and validated in the ADT-alone arm (n=154; 77/77 split), locked, and applied to the held-out ADT+docetaxel arm (n=129). ST-DoxPCa-positive patients derived OS benefit from adding docetaxel to ADT (HR 0.53, 95% CI 0.31–0.90; p=0.018), while ST-DoxPCa-negative patients did not (HR 1.32, 95% CI 0.74–2.34; p=0.34), refining CHAARTED’s broader overall docetaxel benefit where low-volume disease historically showed little/no benefit.