On-treatment plasma ctDNA fraction and treatment outcomes in metastatic castration-resistant prostate cancer.

5051 Background: Androgen receptor pathway inhibitors (ARPI) are standard of care for treatment-naive metastatic castration-resistant prostate cancer (mCRPC), but primary or rapidly acquired resistance is common. Early identification of resistant disease is critical for improving management strategies. We investigated whether on-treatment changes in plasma ctDNA fraction (ctDNA%; the proportion of cell-free DNA that is tumor-derived) associate with ARPI treatment outcomes. Methods: We collected serial plasma cell-free DNA from 81 patients with mCRPC: at baseline and following 4 weeks of treatment during two prospective multi-center observational studies (NCT02426333; NCT02471469). CtDNA% was calculated by integrating deep targeted and shallow whole-genome sequencing. Samples were dichotomized at above and below 1%, and designated as high or low ctDNA%, respectively. Outcome measurements were radiographic and/or clinical progression-free survival (rcPFS) and overall survival (OS). Non-durable responses were defined as rcPFS < 6 months, excluding patients with PSA progression alone (n = 3) or treatment cessation for toxicity (n = 4). Results: Median follow-up was 27.4 months (IQR 17.7-34.9). CtDNA% was high in 47/81 (58%) patients at baseline and 29/81 (36%) patients at 4 weeks. The median ctDNA% for patients with high ctDNA was 15.0% (IQR 4.9-43.8%) at baseline and 5.0% (IQR 2.0-20.6%) at 4 weeks. High baseline ctDNA% was prognostic for rcPFS and OS (Table). RcPFS and OS was shortest for patients that retained high ctDNA% at 4 weeks. However, patients converting from high to low ctDNA% at 4 weeks did not experience different outcomes to patients with low ctDNA at both timepoints. ctDNA% associations with rcPFS and OS were independent of established clinical prognostic factors. 23/27 (85%) patients experiencing non-durable responses had high ctDNA% at baseline and 4 weeks. Only 3/47 patients (6%) experiencing durable responses had high ctDNA% at both timepoints. Sensitivity and specificity for predicting non-durable response was 85% and 94%, respectively. Conclusions: Early changes in ctDNA% are strongly linked to duration of first-line ARPI treatment benefit in mCRPC and may have utility for informing clinical trials testing early therapy switches in patients unlikely to experience durable disease responses. Clinical trial information: NCT02426333; NCT02471469. [Table: see text]