Prognostic role of the naples prognostic score in advanced gastric cancer treated with immune checkpoint inhibitors: a retrospective multicenter study

Abstract Immune checkpoint inhibitors (ICIs), such as by programmed death-1 (PD-1) inhibitors, combined with chemotherapy have been established as a new first-line standard of care for patients with advanced gastric cancer (AGC), significantly prolonging survival in some patients. However, the overall benefit remains markedly heterogeneous, and prognostic assessment tools that can integrate multidimensional clinical and laboratory indicators to accurately quantify individual survival risk are lacking. Therefore, systematically analyzing the survival characteristics and independent prognostic factors of patients with AGC, and constructing a stable and reliable prognostic prediction model are important for achieving individualized treatment decisions and precise risk stratification. In this retrospective study, 252 patients with human epidermal growth factor receptor 2 (HER2)-negative AGC who received ICIs across three institutions were included. We investigated the associations between various clinical factors, including the Naples Prognostic Score (NPS), and overall survival (OS). On the basis of the predictors identified through multivariate analysis, we developed a prognostic model and evaluated its predictive performance using time-dependent receiver operating characteristic (ROC) analysis and calibration curves. Decision curve analysis (DCA) was subsequently performed to assess the clinical utility of the prognostic model. The cohort had a median OS of 15.0 months (95% CI: 13.7–17.0), a median progression-free survival (PFS) of 8.5 months (95% CI: 7.2–10.0), and an objective response rate (ORR) of 36.1%. Univariate Cox regression revealed sex, performance status (PS), Lauren classification, Nutritional Risk Screening 2002 (NRS2002) score, NPS, PD-L1 expression, number of metastatic organs, and ascites as significant prognostic factors. Variables with P < 0.05 and clinically relevant factors were subjected to least absolute shrinkage and selection operator (LASSO) regression, which included PS, Lauren classification, NPS, PD-L1 expression, and the number of metastatic organs. Multivariate Cox analysis revealed that these five factors remained independent predictors of OS (PS: HR = 2.02, 95% CI: 1.27–3.20, P = 0.003; Lauren classification: HR = 1.56, 95% CI: 1.03–2.38, P = 0.037; NPS: HR = 3.72, 95% CI: 1.48–9.34, P = 0.005; PD-L1 expression: HR = 0.46, 95% CI: 0.25–0.86, P = 0.016; number of metastatic organs: HR = 1.61, 95% CI: 1.10–2.35, P = 0.015). A nomogram integrating these variables was constructed. The combined model showed better discrimination than the clinical-only and pathological-only models (AUC = 0.800 vs. 0.672 and 0.653, respectively). The apparent C-index was 0.746 (95% CI: 0.703–0.789), and the optimism-corrected C-index was 0.730 (optimism = 0.016), indicating minimal overfitting. Time-dependent receiver operating characteristic (ROC) analysis yielded 1- and 2-year AUCs of 0.780 and 0.782, respectively; after bootstrap correction, these values remained stable at 0.778 (95% CI: 0.706–0.865) and 0.785 (95% CI: 0.691–0.847), respectively. The calibration curves showed good agreement between the predicted and observed 1-year OS probabilities; however, the 2-year calibration results should be interpreted with caution (calibration intercept = 0.376). The proportional hazards assumption was satisfied (global Schoenfeld test P = 0.67 ). DCA showed that the net benefit of the nomogram was greater than that of the clinical-only and pathological-only models, with the predictive model consistently outperforming the clinical-only model across the 0.05–0.70 threshold range (ΔNB: 0.025–0.202) and showing comparable benefit at low thresholds (0.05–0.30) but substantially greater benefit at high thresholds (0.40–0.70; ΔNB: 0.024–0.169) than the pathological-only model. In patients with AGC receiving ICIs plus chemotherapy, an elevated NPS is associated with poor outcomes and serves as an independent prognostic factor. Integrating the NPS with PS, Lauren classification, PD-L1 expression, and the number of metastatic organs may yield a composite indicator reflecting immune-nutritional status. The NPS has potential value as a prognostic biomarker in patients with AGC treated with ICIs and chemotherapy, and warrants further validation in prospective studies.