Pretreatment multiphasic contrast-enhanced CT predicts tumor regression grade and survival in locally advanced gastric cancer

Reliable pretreatment prediction of pathological response remains a major challenge in locally advanced gastric cancer (LAGC). This study evaluated whether quantitative parameters from multiphasic contrast-enhanced CT (CECT) can predict tumor regression grade (TRG) after neoadjuvant chemotherapy (NAC) and provide prognostic information.

We retrospectively analyzed 116 patients with cT2-4NxM0 gastric adenocarcinoma treated with NAC treatment followed by gastrectomy between January 2019 and December 2024. Quantitative imaging variables included phase-specific tumor attenuation ([Formula: see text], normalized enhancement difference ([Formula: see text]), normalized enhancement rate ([Formula: see text]), tumor area, and enhancement pattern. TRG was assessed on surgical specimens using the Mandard system; responders were defined as TRG 1-2 and non-responders as TRG 3-5. Candidate variables were screened by univariate analysis and Spearman correlation, then enrolled into weighted multivariable logistic regression with 5-fold cross-validation. Associations with overall survival were evaluated using univariate Cox and Kaplan-Meier analyses.

Forty-one patients were responders and 75 were non-responders. CA199, CEA, and age were independent clinical predictors. Among imaging features, venous-phase tumor attenuation ([Formula: see text]; OR = 8.809, 95% CI: 2.134-36.358, P = 0.003) and arterial-phase normalized enhancement rate ([Formula: see text]; OR = 3.200, 95% CI: 1.365-7.504, P = 0.007) independently predicted TRG response. The clinical, imaging, and combined models achieved mean AUCs of 0.746, 0.807, and 0.869, respectively; the combined model showed the best overall discrimination (accuracy 0.759, sensitivity 0.703, specificity 0.787) with good calibration. Higher [Formula: see text] (HR = 1.01, P = 0.004) and [Formula: see text] (HR = 1.01, P = 0.043) were associated with worse overall survival.

Pretreatment CECT-derived quantitative features can predict TRG response in LAGC, and a combined clinical-imaging model improves predictive performance. Higher normalized enhancement rates in the venous and delayed phases are associated with poorer overall survival, supporting pre-treatment risk stratification in clinical practice.