116PLongitudinal changes in cell-free DNA (cfDNA) methylation levels identify early non-responders to treatment in advanced solid tumours

Abstract Background Aberrant methylation changes are present in nearly all cancers. The goal of this study was to determine if changes in cfDNA methylation patterns during the course of treatment could predict non-response prior to routine imaging. Methods We prospectively collected clinical data and blood from 52 patients with advanced malignancies (25 lung, 11 breast, 16 other). Blood was drawn prior to start of a new treatment, after first cycle (median 23 days), and/or second cycle (median 43 days). We performed whole-genome (WG) bisulfite sequencing (median depth 18X) on plasma cfDNA to determine methylation levels. By tracking how methylation levels change across 19,710 regions throughout the genome, from baseline to subsequent timepoints, we classified patients as either progressors or non-progressors. Treatment response at first follow-up imaging (FUI) was determined by RECIST 1.1 or clinical assessment. Study endpoints were agreement with first FUI and progression-free survival (PFS) by cfDNA classification. Results The cohort consisted of 58% females and the median age was 71. Main treatment regimens were immuno- (22), chemo- (20), targeted- (7) or endocrine therapy (3). PFS was significantly shorter (log-rank p = 0.003) in patients classified as progressors by cfDNA (N = 10; median: 90 days) compared to non-progressors (42, median: 263 days). 6 out of 10 patients classified as cfDNA progressors were later confirmed to progress at first follow-up evaluation (60% positive predictive value). The cfDNA assay for these 6 patients preceded imaging and clinical evaluation by a median of 40 days. For the remaining patients, 39 of 42 did not progress (93% negative predictive value). Thus, sensitivity for the assay for identifying progression was 67% and specificity was 91%. Conclusions Our results show that WG cfDNA methylation change is a novel cancer signature with potential to identify patients whose treatment regimen is ineffective prior to imaging, and allow switching to a more effective alternative at an earlier time. Moreover, integrating methylation-based changes with information about genomic alterations may increase performance of cfDNA-based response monitoring. Legal entity responsible for the study The authors. Funding Lexent Bio, Inc. Disclosure A.A. Davis: Travel / Accommodation / Expenses: Menarini Silicon Biosystems. W. Iams: Travel / Accommodation / Expenses, Clinical Trials Visit: EMD Serono. R. Srivas: Shareholder / Stockholder / Stoc