Real-world molecular profile and treatment patterns in advanced gastroesophageal adenocarcinoma: A Brazilian multicenter cohort.

313 Background: Advanced (A) gastroesophageal adenocarcinoma (GA) prognosis remains poor, with limited data on molecular profiles and treatment in resource-diverse populations. Methods: We retrospectively analyzed clinical and molecular data from GA patients who underwent biomarker testing in centralized laboratories between January 2021 and June 2024. Patients were categorized by PD-L1 combined positive score (CPS: <1, 1–4, ≥5), HER2, and MMR/MSI status. Treatment patterns, including systemic therapy and immunotherapy (IO), were assessed among advanced cases with electronic medical record access (AGA-EMR). Time to treatment discontinuation (TTD) was compared between 1L chemotherapy alone and chemo+IO. Results: A total of 334 GAs were tested for PD-L1 CPS. Distribution was: CPS <1, 41.0%; CPS 1–4, 13.5%; CPS ≥5, 43.4%; and inconclusive, 2.1%. Among 112 AGA-EMR patients (59.8% male; median age 66.6 years; 76.8% gastric; 46.4% diffuse and 25.0% intestinal histology; 54.5% M1 at diagnosis), CPS distribution was similar: 44.6% <1, 12.5% 1–4, 42.0% ≥5, and 0.9% inconclusive. MSI-H/dMMR prevalence was 8.0%, and HER2 positivity was 9.8%. A biomarker-negative subgroup (PD-L1 CPS <5, pMMR/MSS, HER2–) represented 47.3%. Systemic therapy was administered to 83.9%, primarily fluoropyrimidine–platinum chemotherapy (87.2%). Immunotherapy (nivolumab or pembrolizumab) was incorporated in 48.9% of 1L regimens. Across all lines, 43.8% received IO. Trastuzumab was used in 72.7% of HER2+ patients. Treatment intensity decreased with successive lines: 22.3% received 2L, 8.9% 3L, and 1.8% 4L. Among pMMR/HER2– patients, IO use strongly correlated with PD-L1 status (84.6% of CPS ≥5 vs 17.1% of CPS <5 received IO, p<0.0001). In the biomarker-restricted TTD cohort (pMMR/HER2-; chemotherapy backbone; n = 73), 41% received chemo+IO and 59% chemo alone. Median TTD was 6.9 months with chemo+IO vs 5.0 months with chemotherapy alone (log-rank p=0.04; HR 0.55, 95% CI 0.30–0.98). Conclusions: In this real-world multicenter cohort, treatment decisions closely followed molecular profiling, with PD-L1 status guiding IO incorporation and trastuzumab used in most HER2+ patients. Nearly half of treated patients received 1L immunotherapy, and in the biomarker-restricted pMMR/HER2– population, chemo+IO was associated with a significantly longer TTD compared with chemotherapy alone. A substantial biomarker-negative subgroup remains without targeted options, underscoring the need for broader molecular testing and novel strategies.