Comprehensive genomic profiling of matched ctDNA and tissue from patients with four common cancers enrolled to the NCI-MATCH trial.

3045 Background: In the NCI-MATCH trial (NCT02465060), tumor tissue from 5,954 patients with advanced cancers underwent next-generation sequencing using the Oncomine Comprehensive Assay v2 (OCAv2) to determine eligibility. Most tumors lacked a qualifying mutation of interest (MOI) for assignment. Plasma from 1,301 patients with common cancers – colorectal (COADREAD), breast (BREAST), non–small cell lung (NSCLC), and prostate (PRAD) (OncoTree codes are shown in parentheses) was analyzed to characterize circulating tumor DNA (ctDNA) and assess its utility for detecting clinically relevant MOIs. Methods: Cell-free DNA was extracted from plasma collected in Streck tubes at enrollment. ctDNA profiling was performed using the NCI ctDNA Research v2 assay (523 genes) on the Illumina NovaSeq 6000. Matched tumor tissue was analyzed using OCAv2 (143 genes). Positive percent agreement (PPA) was calculated using tissue as the reference. Blood-based microsatellite instability (bMSI) was assessed across ~2,400 loci, with bMSI-high (bMSI-H) defined as sum Jensen-Shannon Distance (sumJSD) ≥0.2. Blood-based tumor mutation burden (bTMB) was defined as total SNVs and indels per Mb. Results: Of 1,301 patients, 1,148 (88%) yielded evaluable ctDNA results (COADREAD, n=487; BREAST, n=367; NSCLC, n=220; PRAD, n=74). Overall PPA with matched tissue was 90.3%. Discordant samples had significantly lower median tumor fraction by maximum somatic allele frequency (MSAF; 0.39%) than concordant samples (12.04%). Median MSAF by histology was 14% (COADREAD), 7% (BREAST), 8% (NSCLC), and 5% (PRAD). Clinically relevant fusions detected exclusively in ctDNA included RET (1% NSCLC), EML4::ALK (2% NSCLC), FGFR2 (1% COADREAD; 2% BREAST), and NTRK1 (<1% COADREAD and BREAST), corresponding to actionable NCI-MATCH arms FGFR2/3 - arm K (erdafitinib), ALK - arm F (crizotinib), NTRK - arm Z1E (larotrectinib). Actionable ctDNA-only mutations in PRAD included ATM and MLH1. Twenty-eight cases were bMSI-H (MSAF ≥0.02; sumJSD ≥0.2), of which 13 were mismatch repair-deficient by tissue testing (MLH1/MSH2 nuclear stain-negative). bMSI-H was most frequent in COADREAD (7%). Ten cases (seven COADREAD, two NSCLC, one PRAD) were MMR-proficient by tissue but bMSI-H by ctDNA. Twenty-four cases had high bTMB (≥20 mut/Mb; MSAF ≥0.02), all of which were also bMSI-H. Conclusions: ctDNA - tissue concordance NCI-MATCH in these four cancer histologies was high (90.3%), supporting liquid biopsy as a practical alternative when tissue is unavailable. Detection of ctDNA-only alterations highlights tumor heterogeneity in advanced cancers and identifies additional therapeutic opportunities.