Assessing the impact of clonal hematopoiesis in disease monitoring using targeted cell-free DNA (cfDNA) sequencing technology.

e14530 Background: Somatic variants found in plasma cell-free DNA (cfDNA) may derive from either solid tumors or clonal hematopoiesis (CH). Little is known about how this may impact plasma-based longitudinal disease monitoring using targeted sequencing of circulating tumor DNA (ctDNA). Methods: To assess the potential impact of CH in disease monitoring, we evaluated monitoring algorithms by targeted sequencing with and without matched peripheral blood mononuclear cells (PBMC). Samples were collected from a prospective observational study, where 62 late stage lung adenocarcinoma subjects were treated with first-line chemo or chemoradiation therapy. Pre-treatment plasma cfDNA and matched PBMC were analyzed with the AVENIO ctDNA Surveillance Kit (For Research Use Only, not for use in diagnostic procedures), a sequencing panel of 198 kilobases targeting cancer genes. Median input amounts of 25 ng cfDNA and 50 ng PBMC DNA were sequenced to median deduplicated depths of 4582 and 6134, respectively. Results: A median of 120 single nucleotide variants were detected per cfDNA sample, with 93.1% of these identified in matched PBMC. Most PBMC-matched cfDNA variants were germline SNPs, with allele frequency (AF) ~ 50% or 100%. A median of 1 (range 0-5) PBMC-matched cfDNA variants per sample were detected with an AF < 10%, consistent with CH. The number of these variants was positively associated with age (p-value = 0.0039) and the most frequently mutated gene was TP53. The remaining somatic variants (i.e., in cfDNA and not PBMC) had an AF range 0.03-40.9%. These PBMC-informed variants (median of 7 per sample) were used in longitudinal monitoring in the first post-treatment plasma sample to assess early response to therapy. Association between ctDNA level and progression-free survival using the same monitoring algorithm yielded nearly identical results on somatic variants derived from filtering approaches independent of matched PBMC (HR 0.32; 95% CI 0.16 - 0.65; log-rank P = 0.0009) and the PBMC-informed method (HR 0.31; 95% CI 0.14 - 0.66; log-rank P = 0.0013). Conclusions: A targeted panel focused on solid tumors by design has limited impact from CH. For disease monitoring applications in a non-MRD setting, measuring multiple variants instead of a single variant further enables robust classifiers that can moderate the impact of variants, if any, from CH.