Treatment of poor prognosis metastatic castration-resistant prostate cancer (mCRPC) includes taxane chemotherapy and androgen receptor pathway inhibitors (ARPI). We sought to determine optimal treatment in this setting.

This multicentre, randomised, open-label, phase II trial recruited patients with ARPI-naive mCRPC and poor prognosis features (presence of liver metastases, progression to mCRPC after <12 months of androgen deprivation therapy, or ≥4 of 6 clinical criteria). Patients were randomly assigned 1 : 1 to receive cabazitaxel plus prednisone (group A) or physician's choice of enzalutamide or abiraterone plus prednisone (group B) at standard doses. Patients could cross over at progression. The primary endpoint was clinical benefit rate for first-line treatment (defined as prostate-specific antigen response ≥50%, radiographic response, or stable disease ≥12 weeks).

Ninety-five patients were accrued (median follow-up 21.9 months). First-line clinical benefit rate was greater in group A versus group B (80% versus 62%, P = 0.039). Overall survival was not different between groups A and B (median 37.0 versus 15.5 months, hazard ratio (HR) = 0.58, P = 0.073) nor was time to progression (median 5.3 versus 2.8 months, HR = 0.87, P = 0.52). The most common first-line treatment-related grade ≥3 adverse events were neutropenia (cabazitaxel 32% versus ARPI 0%), diarrhoea (9% versus 0%), infection (9% versus 0%), and fatigue (7% versus 5%). Baseline circulating tumour DNA (ctDNA) fraction above the cohort median and on-treatment ctDNA increase were associated with shorter time to progression (HR = 2.38, P < 0.001; HR = 4.03, P < 0.001). Patients with >30% ctDNA fraction at baseline had markedly shorter overall survival than those with undetectable ctDNA (HR = 38.22, P < 0.001).

Cabazitaxel was associated with a higher clinical benefit rate in patients with ARPI-naive poor prognosis mCRPC. ctDNA abundance was prognostic independent of clinical features, and holds promise as a stratification biomarker.