Dynamic CEA kinetics as a prognostic biomarker for HER2-negative advanced gastric cancer treated with chemoimmunotherapy.

[BACKGROUND] Programmed death-1 (PD-1) inhibitors combined with chemotherapy have become a standard treatment for human epidermal growth factor receptor 2 (HER2)-negative advanced gastric cancer (GC). However, responses vary, and effective prognostic biomarkers are needed. This study aimed to investigate the prognostic value of dynamic changes in carcinoembryonic antigen (CEA) levels in this patient population.

[METHODS] This retrospective cohort study included 96 patients with HER2-negative metastatic or locally advanced GC treated at our institution between June 2021 and June 2024. All patients received PD-1 inhibitors plus chemotherapy. Serial CEA measurements were used to construct a linear model (Y = ax + b) for each patient, where 'a' represents the slope of CEA change. Using a receiver operating characteristic (ROC) curve analysis for predicting 6-month progression-free survival (PFS), an optimal cutoff slope of a = 0.85 was determined. Patients were stratified into a Slow-riser group (a ≤ 0.85, n = 70) and a Rapid-riser group (a > 0.85, n = 26). We compared treatment efficacy, survival outcomes, and adverse events between the groups. Prognostic factors for PFS were identified using univariate and multivariate Cox proportional hazards models, in line with TRIPOD principles.

[RESULTS] During treatment, serial CEA levels exhibited a fluctuating pattern with an overall upward trend. Compared to the Rapid-riser group, the Slow-riser group had a significantly higher disease control rate (DCR) (71.4% vs. 46.2%, P = 0.028) and a lower incidence of progressive disease (PD) (24.3% vs. 53.8%, P = 0.012). The Slow-riser group also experienced a lower incidence of grade ≥ 3 adverse events (22.9% vs. 46.2%, P = 0.040). The median PFS (mPFS) was significantly longer in the Slow-riser group than in the Rapid-riser group (3.9 vs. 2.4 months, P = 0.022), although median overall survival (mOS) was not statistically different (16.3 vs. 12.7 months, P = 0.245). Multivariate Cox regression analysis identified CEA trend (Rapid-riser vs. Slow-riser: HR 7.15, 95% CI 2.85-12.11, P < 0.001), number of metastatic sites, peritoneal metastasis, liver metastasis, ECOG performance status, and differentiation grade as independent predictors of poor PFS.

[CONCLUSION] The dynamic trajectory of CEA levels during treatment is a significant and independent prognostic factor for patients with HER2-negative advanced GC receiving PD-1 inhibitors plus chemotherapy. Slower CEA elevation is associated with better disease control, improved safety, and longer PFS. Monitoring dynamic CEA changes could be a valuable, accessible tool for clinical risk assessment and patient stratification.