Construction and validation of a predictive model for the efficacy of neoadjuvant chemotherapy combined with immunotherapy in locally advanced gastric cancer: a single-center retrospective study

Abstract Objective Exploring the factors influencing the efficacy of Neoadjuvant chemotherapy combined with immunotherapy (NACI) in locally advanced gastric cancer (LAGC) and constructing a nomogram model to predict treatment response. Methods Data of patients with LAGC who underwent NACI at the Department of Gastroenterology, Xijing Hospital from June 2021 to December 2024 were retrospectively collected. According to tumor regression grading (TRG), patients were divided into response group (TRG 0–1) and non-response group (TRG 2–3). The feature selection of the model was optimized by least absolute shrinkage and selection operator (LASSO) regression, the predictive model was constructed using multifactor logistic regression and plotted in nomogram, the distinction, calibration and clinical applicability of the predictive model were assessed by receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA), and internal validation was carried out using the Bootstrap method . Results A total of 144 patients were included (52 in the response group and 92 in the non-response group).LASSO regression identified six predictive variables: diabetes mellitus, tumor differentiation grade, signet ring cell carcinoma, programmed cell death-ligand 1 (PD-L1), carbohydrate antigen 72 − 4 (CA72-4), and neutrophil-lymphocyte ratio (NLR); When included in the results of multivariate logistic regression, having diabetes mellitus (OR = 1.94, 95% CI: 0.72–5.26, p = 0.190), tumors with moderate to high differentiation (OR = 3.91, 95% CI: 1.65–9.25, p = 0.002), local absence of signet ring cells (OR = 0.30, 95% CI: 0.10–0.90, p = 0.032), PD-L1 score ≥ 1 (OR = 3.03, 95% CI: 1.31–7.00, p = 0.010), CA724 ≤ 6.9 U/mL (OR = 0.38, 95% CI: 0.15–0.96, p = 0.040), and low NLR (OR = 0.68, 95% CI: 0.48–0.96, p = 0.029) were identified as independent factors influencing the benefit of neoadjuvant therapy for patients.Using the above five variables with P < 0.05 to construct a stepwise regression prediction model, the area under the ROC curve (AUC) was 0.799 (95% CI: 0.725–0.873). Calibration curve results showed that the model's predicted results fit well with the actual results, with a Hosmer-Lemeshow test P value of 0.772. DCA results showed that the model had a high net benefit. Conclusion The nomogram model constructed in this study can effectively predict the efficacy of NACI for LAGC and has high clinical practical value.