Longitudinal tumor-informed cfDNA whole genome sequencing to capture residual disease during neoadjuvant immune checkpoint inhibition in resectable gastroesophageal cancer.

2550 Background: Although circulating tumor DNA (ctDNA) detection represents a promising approach to capture minimal residual disease (MRD), the clinical performance of ctDNA MRD during neoadjuvant immune checkpoint inhibition (ICI) remains understudied. Here we employ a tumor-informed whole genome sequencing (WGS) approach to capture residual disease and link ctDNA dynamics with pathologic response and clinical outcomes. Methods: WGS was performed on tumor (n = 28), matched WBC (n = 28), and longitudinal plasma samples (baseline, post-ICI cycle 1, post-ICI cycle 2 and pre-op; n = 97) from 28 patients with resectable gastroesophageal cancer treated with neoadjuvant ICI and chemoradiation prior to surgical resection (NCT03044613). Tumor-specific single nucleotide variants were identified from tumor and WBC datasets, from which a high confidence candidate variant set was used to determine the presence of ctDNA through a random forest machine learning model. ctDNA status and tumor fraction (TF) were determined based on the level of signal compared to a reference population of noncancerous donor plasma samples (n = 80). Serial ctDNA TF dynamics were correlated with overall (OS) and recurrence-free survival (RFS) in comparison to a tumor-naïve targeted NGS gene panel liquid biopsy approach. Results: Twenty-four of the 28 patients (86%) with evaluable specimens had ctDNA detected in a least one timepoint: 22 of 25 (88%) evaluable patients had ctDNA detected at baseline, 20 of 25 (80%) evaluable patients had ctDNA detected post-ICI cycle 1, 18 of 26 (69%) evaluable patients had ctDNA detected post-ICI cycle 2, and 5 of 21 (24%) evaluable patients had ctDNA detected at the pre-op timepoint. In contrast, the tumor-naïve targeted NGS approach detected 13 of 30 (43%), 12 of 30 (40%), 11 of 30 (37%) and 5 of 25 (20%) patients at baseline, post-ICI cycle 1, post-ICI cycle 2 and pre-op respectively. A ctDNA TF peak was detected at either the post-ICI cycle 1 or cycle 2 timepoint for 50% of the patients. A 50% reduction in ctDNA TF at the post-ICI cycle 2 timepoint showed a sensitivity of 80% and specificity of 69% for prediction of complete pathologic response, which was improved compared to the tumor-naïve liquid biopsy approach and importantly showed a significantly higher evaluable rate (86% vs 62% for tumor-informed and tumor-naïve respectively). Similar trends were observed between major pathologic response and ctDNA TF. A dramatic reduction of ctDNA TF (≥65%) at the pre-op timepoint predicted longer OS and RFS (log-rank p = 0.0035 and p = 0.0032 respectively). Conclusions: Tumor-informed cfDNA whole genome sequencing analyses showed reliable and sensitive detection and quantification of ctDNA during neoadjuvant ICI and adds to the body of evidence supporting the clinical utility of ctDNA residual disease in interpreting clinical outcomes.