Discriminative Performance of Routine Inflammation- and Nutrition-Based Indices for Histopathologically Confirmed...

Background/Objectives: This study aimed to evaluate the discriminative performance of routine systemic inflammation- and nutrition-based indices for distinguishing histopathologically confirmed gastric malignancy from non-malignant endoscopic findings. Methods: This retrospective case-control study included 211 patients who underwent upper gastrointestinal endoscopy with histopathological evaluation: 139 non-malignant endoscopic controls (129 benign/non-neoplastic and 10 premalignant findings) and 72 gastric malignancy cases. Premalignant lesions were not considered malignant outcomes. Discrimination was evaluated using univariable ROC analyses and multivariable logistic models. The fixed predictor set was assessed using 10-fold out-of-fold validation and 500-resample bootstrap optimism correction; a separate fully nested selection-inclusive audit was used to quantify uncertainty arising from variable selection. Results: NLR (AUC = 0.792), INR (AUC = 0.791), and SIRI (AUC = 0.783) showed the highest univariable discrimination. Among the variables retained in Model 1, higher ESR and RDW, lower MPV and HDL, higher INR values, and male sex showed statistically significant multivariable associations. Model 1 had an apparent AUC of 0.928 (95% CI, 0.890-0.959), a fixed-set 10-fold out-of-fold AUC of 0.890, and a bootstrap optimism-corrected AUC of 0.901. A fully nested selection-inclusive audit yielded an out-of-fold AUC of 0.873 and a bootstrap-corrected AUC of 0.869. Conclusions: Routine inflammation- and nutrition-related laboratory measures discriminated gastric malignancy from non-malignant endoscopic findings in this endoscopy-selected cohort. The model remains exploratory and requires prospective external validation. These data do not establish screening performance or diagnostic accuracy for early or asymptomatic gastric cancer.