Use of a tissue-free epigenomic circulating tumor DNA (ctDNA) assay for quantification of tumor fraction (TF) and association with outcomes from RADIOHEAD real-world advanced pan-cancer cohort.

2570 Background: The validity of ctDNA assays is well established for evaluating molecular response to therapy, and ctDNA levels may be monitored throughout a patient’s (pts) journey to indicate when relapse or progression is present, often sooner than current methods (RECIST). Genomic assays using variant allele fraction (VAF) have limitations such as low ctDNA levels and interference from copy number variation and clonal hematopoeisis (CH), which may be overcome by methylation-based quantification. Here we describe the performance of a methylation-based assay to quantify ctDNA levels and correlate changes with outcomes in a real-world pan-cancer cohort. Methods: RADIOHEAD is an observational study of ~1200 solid tumor patients receiving standard of care ICI regimens with blood samples collected prospectively for retrospective analysis. 552 patients with stage III/IV lung, melanoma, bladder and other cancers were randomly chosen for evaluation. Available plasma samples from baseline and on-treatment timepoints (C3D1, 6mo and 12mo) were analyzed with an analytically validated NGS ctDNA assay measuring methylation and TF quantification. Cox proportional hazards (CPH) were used for comparison of real-world progression free survival (rwPFS). Gender, age, disease stage, and tobacco use were included as co-variates. Median rwPFS and rwOS were calculated using Kaplan Meier analysis. rwPFS ordinal groups (95% reduction in TF from baseline to C3D1 or low ctDNA at both timepoints (rwPFS <3mo = 1.8%, 3-6mo = 7.8%, 6-12mo = 15.1%, ≥12mo = 75.3%; N=166 p<0.001). For pts with ctDNA not detected at C3D1, ctDNA detection at 6mo post first dose was strongly associated with shorter rwPFS vs ctDNA not detected at 6mo (mPFS 13.6mo vs. NR; HR= 7.43 p<0.001). For pts with ctDNA not detected at 6mo, ctDNA detection at 12 months post first dose was also associated with shorter rwPFS vs ctDNA not detected at 12mo (mPFS 22.8 mo vs. NR; HR= 5.2 p=0.02). Conclusions: These data demonstrate a significant association of methylation-based ctDNA detection and on-treatment changes in TF with rwPFS and rwOS. Furthermore, subsequent TF detection 6mo or 12mo on-treatment, without TF detected at previous timepoint was associated with worse outcomes. This suggests the value of serial monitoring throughout treatment for early detection of progression and the potential to inform treatment decisions.