CT Radiomics for Predicting Outcomes in HER2-Positive Surgically Resectable Advanced Gastric Cancer: A Preliminary...

Accurate risk stratification in human epidermal growth factor receptor 2-positive surgically resectable advanced gastric cancer (HER2-p SRAGC) can strengthen monitoring for high-risk patients, allowing timely HER2-specific treatment and potentially improving prognosis. Therefore, we aimed to develop a CT radiomics model for predicting outcomes in HER2-p SRAGC and compare it with the 8th edition TNM staging system.

621 HER2-p SRAGC patients who received either radical gastrectomy or radical gastrectomy after neoadjuvant therapy or chemotherapy were retrospectively enrolled in two hospitals and assigned to a training (n=330), an internal validation (n=143) and an external validation (n=148) cohorts. A radiomics model incorporating Radscore and clinical scores was constructed. Model performance was assessed by Kaplan-Meier estimator, Log-rank test, and Harrell's C-index.

The radiomics model was correlated with the overall survival (OS) across all cohorts, with C-indexes of 0.711[95% confidence interval (CI): 0.666-0.756; p<0.001; training], 0.669 (95%CI: 0.585-0.753; p=0.009; internal validation) and 0.693 (95%CI: 0.597-0.789; p=0.015; external validation). In all study cohorts, the radiomics model successfully stratified patients into high-risk and low-risk groups, and outweighed individual scores, pathological staging (pTNM), and clinical staging (cTNM) but was inferior to post-neoadjuvant therapy staging (ypTNM). Additionally, the radiomics model had added value to the prognostic efficacy of pTNM and was unaffected by patient age and gender.

The radiomics model delivers individualized prognosis prediction of HER2-p SRAGC, surpassing clinical scores, and both pTNM and cTNM in forecasting OS. It confers incremental benefit to pTNM and exhibits certain universality across patient types.

Retrospectively registered.