Risk factors (RF) for poor outcomes in esophageal (EC) and gastric (GC) adenocarcinomas treated with immune-checkpoint inhibitors (ICI).

e16032 Background: There are no established risk factors or toxicity biomarkers for ECs and GCs treated with ICI, even though they are commonly used in current clinical practice. Additionally, the predictive value of programmed death-ligand 1 (PDL1) expression level is questionable. This is a retrospective review of patients with EC or GC treated with ICI at an academic institute to identify RF for poor outcomes. Methods: Patients with EC or GC who received at least one dose of ICI for Stage IV or recurrent disease between 1/1/2015 and 7/31/2021 at the Ohio State University were included in the study. The patients' baseline (BL) characteristics (at the first dose of ICI) were extracted with immune-related adverse events (irAE) details and survival outcomes. Descriptive statistics, Fisher exact test for categorical variables, log-rank test for survival outcomes, and logistic regression modeling for categorical outcomes were conducted using JMP Pro 16 (SAS Institute Inc., Cary, NC). Results: Our cohort included 67 patients (42 EC and 25 GC). The median age was 65 years (range 33-90) with 84% male, 88% Caucasian (4% African American & 8% others), and 93% with distant metastasis (7% recurrent disease). 75% received the first ICI dose in ≥ 3 lines with 4% first line, 41% had a history of radiotherapy (XRT) (22% palliative & 19% neoadjuvant [NA]), and only 9% had chemotherapy combination (CC). Most were microsatellite stable (90%), and PDL1 expression was not available in 39%, negative in 7%, and positive in 53%. Median overall survival (OS) and progression-free survival (PFS) were 5 months (m) (95% confidence interval [CI], 2-8) and 3m (95% CI, 2-4), respectively. Patients with ICI in ≥ 3 lines did worse (vs ≤ 2 lines, 3m vs 12m, respectively, p = 0.01) in OS. Patients with a history of XRT had lower PFS (yes vs no, 2m vs 3m, p = 0.02), but OS was not significantly different. BL white cell count (WBC) (p = 0.002), alkaline phosphatase (ALP) (p = 0.006), total bilirubin (TB) (p = 0.004), thyroid stimulating hormone (TSH) (p = 0.04), and free T4 (p = 0.01) significantly impacted OS, while WBC (p = 0.01), red cell count (RBC) (p = 0.02), platelet count (PLC) (p = 0.04), and absolute lymphocyte count (ALC) (p = 0.01) impacted PFS. RFs positively related to OS and PFS were the line of ICI and CC, while high ALP and prior XRT were negatively related to OS and PFS. Low albumin (p = 0.003) was negatively related to OS-only while primary tumor location (EC, p = 0.04) and prior NA XRT (p = 0.01) were negatively related to PFS-only. Other BL characteristics, including PDL-1 expression (among patients with available values), did not impact outcomes. Only 7% (n = 5) had irAE, and two had ≥ grade 3. Conclusions: Patients with identified RFs (EC primary, prior XRT, NA XRT, and BL labs such as WBC, RBC, PLC, ALP, ALC, albumin, TB, TSH, and free T4) for clinical outcomes should be monitored closely. These associations should be confirmed in larger studies.