Computational pathology to predict docetaxel benefit in patients with metastatic hormone-sensitive prostate cancer from the CHAARTED trial (ECOG-ACRIN E3805).

Abstract

1560 Background: The CHAARTED trial demonstrated the efficacy of early docetaxel (DTX) in combination with androgen deprivation therapy (ADT) for metastatic hormone-sensitive prostate cancer (mHSPC), particularly in patients (pts) with high volume (HV) disease. Volume of metastases has been shown to assist pts selection, however more precise biomarkers are needed to select pts that might benefit from early DTX, especially in low volume (LV) metastasis. We aim to validate a computational AI pathology image classifier (APIC), that quantifies the tumor-immune microenvironment from biopsy images to predict DTX benefit in pts from the CHAARTED trial. Methods: The study included a subset of pts from CHAARTED for whom high-quality biopsy images were available. Outcomes were defined as overall survival (OS) and time to castration resistance (CRPC). After segmenting nuclei and identifying lymphocytes, we derived features that captured immune-tumor spatial patterns and nuclear diversity to construct APIC and stratify patients into positive or negative groups. DTX benefit was evaluated using Cox proportional hazards with interaction terms, log-rank tests and Kaplan-Meier analyses by comparing endpoint estimates between treatment arms within APIC-stratified groups. Results: Among CHAARTED trial pts, we analyzed H&E images that met quality control and excluded prostatectomies (N = 286/790, 36.2%). Half of the ADT arm was used for training (78 pts), and 208 pts were used for validation (ADT = 77, DTX = 131). Among these, 118 pts (56%) were classified as APIC-positive and 90 pts (44%) as APIC-negative. In APIC-positive, DTX significantly improved OS (HR = 0.52 [95% CI: 0.31–0.85], p = 0.0075, interaction p < 0.05) and delayed time to CRPC (HR = 0.48 [95% CI: 0.33–0.71], p = 0.00019, interaction p < 0.05). APIC-positive pts who received DTX derived a 24.3% higher 5-year OS and remained castration-sensitive 21.8% longer than those who received ADT alone. In APIC-negative pts, no prolongation effects of OS or time to CRPC were observed from the addition of DTX. APIC was able to identify pts benefiting from DTX in the HV group for OS (HR = 0.43 [95%CI: 0.24-0.77], p = 0.0035), and in both HV (HR = 0.50 [95%CI: 0.31-0.79], p = 0.0027) and LV (HR = 0.42 [95%CI: 0.20-0.91], p = 0.023) groups for time to CRPC. While CHAARTED showed modest CRPC delay in LV pts and no clear OS benefit, APIC identified a subset of LV pts who derived substantial CRPC delay from DTX. Conclusions: We validated APIC as a predictive biomarker for DTX benefit in mHSPC pts from CHAARTED. Notably, in unselected pts with LV disease where the benefit of DTX is less, APIC identified an LV subset who derived significant delayed progression to CRPC from adding DTX to ADT. This work, validated in the context of ADT alone, warrants investigation alongside androgen receptor axis-targeted agents.