Objective To investigate the value of a computed tomography (CT) radiomics-based model for predicting the efficacy of neoadjuvant immunotherapy combined with chemotherapy for locally advanced gastric cancer (LAGC).

Methods Data on 114 patients with LAGC who underwent radical surgery after neoadjuvant immunotherapy combined with chemotherapy at The Second Affiliated Hospital of Xingtai Medical College between June 2019 and June 2021 were retrospectively collected. These patients’ data were divided into a training set (n=67) and a validation set (n=47) based on the time of admission. High-throughput features were extracted from baseline portal phase CT images of all patients, and the selected features were used to construct the radiomics prediction model. The predictive performance of the model was evaluated using receiver operating characteristic (ROC) and calibration curves. The prognostic ability of the model was assessed using Kaplan–Meier curves.

Results Based on the maximum relevancy min-redundancy (mRMR) algorithm and least absolute shrinkage and selection operator (LASSO) regression model, 5 out of 584 assessed features were incorporated into the radiomics (Rad) score. The respective areas under the curve for predicting pathological complete response (pCR) in the training and validation sets were 0.865 and 0.830, respectively, and good fits were obtained (Hosmer-Lemeshow test: P>0.05). The optimal cut-off value for the Rad score was determined based on the Youden index. Patients with high Rad scores had significantly higher 3-year recurrence-free survival rates (82.7% vs. 60.4% in the training set and 78.9% vs. 53.8% in the validation set) and 3-year overall survival rates (78.9% vs. 60.2% in the training set and 79.3% vs. 50.0% in the validation set) than those with low Rad scores (P<0.05).

Conclusions The CT radiomics prediction model effectively predicted the pathological response and prognosis of patients with LAGC after neoadjuvant immunotherapy combined with chemotherapy and is expected to serve as a practical clinical tool.