Analysis of clinically actionable alterations in baseline tumor versus plasma samples in participants of the TOPAZ-1 study of durvalumab plus gemcitabine and cisplatin in advanced biliary tract cancer.

625 Background: In the TOPAZ-1 study (NCT03875235), durvalumab + gemcitabine and cisplatin (D+GC) significantly improved overall survival (OS) versus placebo + GC (P+GC) in participants (pts) with advanced biliary tract cancer. Updated results have shown a clinically meaningful long-term OS benefit for D+GC versus P+GC at 3 years. This exploratory analysis tested circulating tumor DNA (ctDNA) in plasma samples for blood-based detection of clinically actionable alterations (CAAs) and investigated the potential of this method to guide treatment decisions. Methods: Baseline genomic alterations were retrospectively assessed in evaluable tumor (FMI biomarker evaluable population [BEP], n=441) and plasma samples (GH BEP, n=643) using FoundationOne (Foundation Medicine Inc., Cambridge, MA) and Guardant INFINITY (Guardant Health, Redwood City, CA) assays, respectively. Mutation prevalence and association with outcomes were compared in the FMI BEP and GH BEP. Positive and negative percent agreement of CAAs detected in tumor versus ctDNA were assessed in 419 pts with both tumor and plasma samples (FMI-GH BEP). Results: The FMI BEP and GH BEP represented 64% and 94% of the TOPAZ-1 final analysis set (685 pts), respectively. The relative prevalence and overall mutational landscape detected in plasma ctDNA was consistent with that observed by tumor profiling, with the notable exception that genes harboring complex alterations (e.g. gene amplification, rearrangements, or homozygous deletions) were less frequently detected in ctDNA (e.g. ERBB2 , FGFR2 , and CKDN2A/2B/MTAP ). The most common alterations (>15% prevalence in both BEPs) observed were in TP53 (49%/52%), KRAS (24%/17%), and ARID1A (21%/16%) (in the FMI/GH BEPs, respectively). The relative prevalence of alterations within geographic and anatomic subgroups was similar in tumor versus plasma for most CAAs. The overall percent agreement in CAAs was ≥93%, and negative percent agreement was ≥97%. However, positive percent agreement was notably low for ERBB2 amplification (52%) and FGFR2 fusions (47%). OS hazard ratios for D+GC versus P+GC in the GH BEP were <1 for both CAAs and wild-type, except for ERBB2 amplification, as previously reported in the FMI BEP. Conclusions: The overall concordance and relative prevalence of simple mutations (e.g. single nucleotide variants) were similar using the FoundationOne tumor assay and Guardant INFINITY ctDNA assay, suggesting that plasma ctDNA testing has potential utility in clinical practice. However, negative status by ctDNA for the complex alterations found in FGFR2 and ERBB2 would require further testing of tumors, based on their low detectability in plasma. Clinical trial information: NCT03875235 .