Abstract / Summary
Background/Objectives: We compared the prognostic performance of preoperative prognostic nutritional index (PNI) and Gustave Roussy Immune Score (GRIm-s) after curative-intent gastrectomy. Methods: This retrospective cohort included 168 patients, including five postoperative in-hospital deaths. Continuous PNI and ordinal GRIm-s were compared using censoring-adjusted time-dependent AUC and concordance, with paired bootstrap confidence intervals. Preoperative prediction models underwent bootstrap internal validation; restricted mean survival time and time-varying Cox analyses addressed non-proportional hazards. Results: There were 92 deaths and 107 disease-free survival (DFS) events. At 36 months, overall survival (OS) AUC was 0.659 for PNI and 0.527 for GRIm-s (difference 0.132, 95% CI 0.039–0.226); corresponding DFS values were 0.650 and 0.516 (difference 0.134, 95% CI 0.035–0.228). After optimism correction, preoperative PNI-model AUCs were 0.610 for OS and 0.595 for DFS. PNI associations weakened over time. The cohort-derived 49.6 threshold was used only in secondary exploratory analyses. Conclusions: PNI showed better unadjusted discrimination than GRIm-s at three years, but modest internally validated model performance limits clinical interpretation. These findings do not establish clinical utility, a validated cutoff, or a substitute for pathological assessment. External validation is required.