Utility of ctDNA tumor fraction to inform negative liquid biopsy (LBx) results and need for tissue reflex in advanced non-small cell lung cancer (aNSCLC)

9076 Background: Negative LBx results can be challenging to interpret. Absence of a targetable alteration may accurately reflect the tumor genotype (true negative) or may represent a false negative due to insufficient ctDNA shed, concealing a targetable oncogenic driver. Reflex to tissue biopsy (TBx) profiling is thus advised, per FDA label, for negative LBx results. Here we investigate whether algorithmic quantification of ctDNA in a LBx sample can increase confidence in negative LBx results, reducing need for confirmatory TBx testing. Methods: This study used the nationwide (US-based) de-identified Flatiron Health-Foundation Medicine NSCLC clinico-genomic database (CGDB) from approximately 280 US cancer clinics (~800 sites of care), limited to patients (pts) with aNSCLC receiving any multi-gene LBx followed by TBx CGP. In parallel, we queried an institutional database of pts with both LBx (FoundationOneLiquid CDx, F1LCDx) and TBx (FoundationOneCDx, F1CDx) to calculate positive predictive agreement (PPA) and negative predictive value (NPV) of LBx compared to TBx for detection of 7 targetable NSCLC driver alterations: KRAS G12X, EGFR, MET exon 14, BRAF V600E mutations or ALK, ROS1, and RET rearrangements. Foundation Medicine’s ctDNA tumor fraction (TF) on F1LCDx is a composite algorithm prioritizing aneuploidy at higher levels to avoid germline signal and prioritizing variant allele frequency of canonical alterations at lower levels to maximize dynamic range. Results: Of 1,734 LBx sent prior to 1 st -line therapy in pts with aNSCLC, 59% (1,024) were negative for an NCCN biomarker. For 465 pts with TBx after negative LBx (median 2.4 weeks from LBx to TBx result), 179 (38%) had a driver detected on reflex to TBx. In 45 receiving matched targeted therapy, median real-world PFS was 25 months. For pts starting therapy after reflex to TBx (n = 233), median time from LBx result to start of therapy was 4.7 weeks. To understand whether reflex to tissue may be avoided in some cases, LBx (F1LCDx) sensitivity was studied across 2,138 pairs in the institutional database. PPA for the 7 driver alterations was 59% overall (679/1,157) and improved to 96% (433/453) when limited to cases with TF ≥1%. Similarly, NPV was modest overall (67%) but improved when limited to LBx specimens with TF ≥1% (96%). Studying TF in the CGDB cohort, 33 of 85 pts (39%) with reflex TBx after negative LBx had a LBx TF ≥1% and, given high NPV for driver alterations, might have avoided reflex to confirmatory TBx. Conclusions: Elevated TF (≥1%) on F1LCDx informs assay sensitivity and identifies a subset of LBx results with high NPV where reflex to TBx, and the associated treatment delay, may be less valuable. Conversely, for pts with a negative LBx with low TF ( < 1%) on F1LCDx, reflex to TBx is an important follow-on step to identify potentially missed driver alterations with potential for durable benefit on targeted therapy.