Abstract A020: Utilization of liquid biopsy in metastatic castrate resistant prostate cancer (mCRPC): patient characteristics, treatment selection, and mutational status

Abstract Background: Targeted therapies for metastatic castrate resistant prostate cancer (mCRPC) increasingly rely on genomic profiling, including liquid biopsy of circulating tumor DNA (ctDNA), to enable precision treatment. Integrating multi-patient ctDNA-derived molecular insights with deeply abstracted real-world clinical data offers the opportunity to improve biomarker identification, patient selection, and reverse translation. Methods: Data were obtained from the ConcertAI Translational360™ dataset. Specifically, mCRPC patients with overlapping liquid biopsy ctDNA testing results from Guardant Health were identified. Patient characteristics, timing of ctDNA testing, prior treatments, and the prevalence of specific mutations were described. Using all mCRPC patients in the cohort, treatment-agnostic univariate Cox proportional hazards models were trained on the 3 most frequent mutations and on maximum allele frequency. Results: Among 1,167 mCRPC patients from primarily community oncology practices (78.3%) with at least one ctDNA test, median age at mCRPC diagnosis was 68 (IQR: 60, 80) years; 72% of patients were white, 17% were Black, and 2% were Asian. The first ctDNA test occurred a median 2.2 years (IQR: 0.9, 5.0) after patients became castrate-resistant. For 200 patients with longitudinal testing, a median of 1.2 years (IQR: 0.6, 1.9) elapsed between the first and last ctDNA test. Treatment patterns shifted after ctDNA testing. Specifically, PARP inhibitor therapy increased from 5% of patients before ctDNA testing to 20% after ctDNA testing. Longitudinal changes in mutations were associated with treatment. Among patients who initiated PARP inhibitor therapy between their first and last ctDNA tests (n=22; median 5 prior lines of therapy (IQR: 3.2, 7.8)), the increase in mutation count was greater than in other patients: median increase of 8.5 unique mutations (IQR 5.0, 15.5) vs. 3 (IQR 0, 7.2), Wilcoxon rank-sum p