Ultra-sensitive ctDNA dynamics to capture therapy response in pembrolizumab-treated gastroesophageal cancer.

4025 Background: Metastatic esophagogastric cancer (mEGC) poses a significant challenge due to its poor long-term survival. The KeyLargo trial (NCT03342937) was a single-arm phase II study of pembrolizumab combined with oxaliplatin and capecitabine in first line HER2 negative mEGC patients (pts). Although the study revealed encouraging response rates, not all pts responded, underscoring the need for improved biomarkers. To address this challenge, we utilized an ultra-sensitive tumor-informed circulating tumor DNA (ctDNA) strategy to monitor patient response and identify predictive biomarkers for immune checkpoint blockade (ICB) therapy in mEGC. Methods: Between January 2018 and January 2020, a total of 36 pts were treated. 33 pts were evaluable with an overall response rate of 76%. This included 6 pts with a complete response and 19 pts with a partial response. Our analysis focused on 25 pts with FFPE tumor samples of sufficient quality for whole-genome sequencing (WGS), complemented by >200 corresponding on-treatment plasma samples, collected up to cycle 35. NeXT Personal, a tumor-informed ctDNA assay, was employed to construct bespoke liquid biopsy panels for each patient utilizing single nucleotide variants (SNV) identified through tumor WGS. Each personalized assay comprises up to ~1,800 SNVs, offering enhanced sensitivity for the detection of molecular residual disease (MRD). Results: In the cohort of 25 pts, ctDNA levels were detected across a broad range from 406,067 down to 5.5 parts per million (PPM). All 25 pts (100%) tested positive for MRD at baseline prior to treatment in the study, with a median limit of detection of 2.03 PPM. Molecular progression based on ctDNA increases, consistently preceded imaging progression, with a median lead time of 55 days. Patients exhibiting an early decrease or clearance in ctDNA levels showed significant improvement in overall survival (OS) (p-value=0.002 and 0.005, respectively) and progression-free survival (PFS) (p-value=0.005 and 0.012, respectively). Univariate Cox regression analysis indicated greater risk and a higher hazard ratio (HR) for pts when ctDNA did not significantly decrease (HR=15 for PFS and HR=6.6 for OS) and a protective effect when ctDNA was cleared (HR=0.52 for PFS and HR=0.15 for OS). ctDNA dynamics were highly correlated with changes in the sum of the diameters across all target lesions (rho=0.54, p-value = 7.2e -8 ). Conclusions: ctDNA was detected in all pts at baseline. A decrease in ctDNA levels prior to the first restaging or ctDNA clearance up to cycle 7 were strong predictors of improved patient outcomes. ctDNA-based molecular progression predated progression by imaging by nearly two months, and ctDNA dynamics correlated with changes in tumor diameter. These results are suggestive that an ultra-sensitive liquid biopsy platform could improve treatment decision-making for pts receiving ICB.