Multicenter study evaluates multimodal model for gastric cancer chemotherapy response

Neoadjuvant chemotherapy is standard treatment for locally advanced gastric cancer, but patient responses vary. Researchers analyzed 1,568 patients: 1,150 from six centers for model development and validation, 352 in a real-world incremental cohort, and 66 from two prospective trials. NeoGEM combined clinical information with radiomic features from pretreatment CT scans of tumors and L3-level body composition to predict response. Pathological response was used in the main cohorts; the incremental cohort used radiological response.

NeoGEM achieved an AUC of 0.888 and accuracy of 0.809 in the training cohort. AUCs were 0.870–0.875 in internal validation, 0.859–0.877 in external validation, and 0.864–0.867 in the prospective cohorts. In the incremental cohort, the AUC was 0.840. The model outperformed clinical, tumor-only and body-composition-only models across the validation cohorts examined. Low- versus high-score groups also differed in 5-year overall survival (35.5% vs. 62.4%) and disease-free survival (21.5% vs. 50.8%).

These results support further evaluation of multimodal response prediction, but do not establish that using NeoGEM to guide treatment improves outcomes. The prospective validation involved only 66 patients, and the incremental cohort used a different response measure. Transcriptomic findings were exploratory.