Abstract 2231: Utility of plasma tumor fraction (TF) to inform sensitivity of FoundationOne Liquid CDx (F1LCDx)

Abstract

Abstract Background: F1LCDx is an FDA-approved liquid biopsy test offering pan-tumor profiling of >300 genes with CDx indications for multiple cancer types. One key limitation is that sensitivity for detection of genomic alterations in ctDNA is impaired when the plasma tumor fraction (TF) is low. TF is known to vary in advanced cancer and has been correlated with overall survival (Stover, JCO, 2018). While some ctDNA assays use variant allele fraction (VAF) to quantify tumor content, such methods may be prone to interference from germline variants, copy number alterations and a limited number of mutations. We studied SNP aneuploidy to develop a variant-independent measure of TF which we hypothesized could inform assay sensitivity. Methods: For each liquid sample, coverages from heterozygous SNPs, across the genome, were used to measure allelic deviations from normal via a root-mean-square (RMS) metric. The relationship between RMS and TF was established based on a training data set of clinical samples with known TF that were diluted in silico. In this way, the RMS metric for any sample could be linked to a distribution of TF values. TF is reported when the lower bounds of the predicted TF distribution is >0%. Positive percentage agreement (PPA) of liquid NGS versus tissue genotype was studied in those with a TF reported vs TF not quantified. Results: For a pilot concordance analysis, we studied a real-world cohort of NSCLC patients where NGS was performed on both tissue and on a 70-gene ctDNA assay (FoundationOne®Liquid (F1L), median 314 days between results). Among 251 cases with tissue NGS showing an EGFR L858R or 19del mutation and a F1L result available, the EGFR mutation was detected 90% of the time (57/63) when TF was reported (range 13-74%) but was detected 51% of the time (96/188) when TF was not quantified. Next, we analyzed a cohort of 3 clinical trials in advanced cancers where tissue genotype was known and with F1LCDx results available, including studies of activating PIK3CA mutations in breast cancer, MET exon 14 mutations in NSCLC and BRCA1/2 mutations in prostate cancer. For the subset of patients with TF reported (range 9.6-72%), PPA was 94.7% (breast, n=38), 100% (NSCLC, n=19), and 90.1% (prostate, n=46), versus 66.7% (breast, n=33), 66.7% (NSCLC, n=45) and 89.6% (prostate, n=48) for TF not quantified. Conclusions: We describe a novel method to measure TF in plasma ctDNA based on aneuploidy, which may overcome some limitations of TF measurement based on VAF. When TF confirms tumor content, sensitivity of F1LCDx for short variant detection is excellent (90%-100%), suggesting reduced value from a reflex to tissue genotyping in the absence of a targetable genomic alteration. When TF is not quantified, sensitivity can be lower and the value of reflex to tissue genotyping may be heightened. Citation Format: Meijuan Li, Bernard Fendler, Lei Yang, Russell Madison, Brennan Decker, Ole Gjoerup, Kimberly McGregor, Aparna Aiyer, Jason Hughes, Priti Hegde, Christine Vietz, Geoffrey R. Oxnard. Utility of plasma tumor fraction (TF) to inform sensitivity of FoundationOne Liquid CDx (F1LCDx) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 2231.