Journal: Scientific reports
This prospective multicenter study assessed predictive models that integrate CT and ultrasound to evaluate the response to neoadjuvant therapy in locally advanced gastric cancer.
Five RECIST 1.1-derived submodels and an ordinal regression-based nomogram were developed using pathological tumor regression grade (TRG) as the reference.
Key findings include:
- V-RECIST model: Demonstrated the highest diagnostic accuracy for distinguishing TRG categories, with AUCs of 0.951 and 0.868 for different TRG groupings.
- U-RECIST model (ultrasound-only): Showed good accuracy and offers a radiation-free, cost-effective option for preliminary assessment.
- Other models (RECIST_Expansion, UC_RECIST, CT_RECIST): Provided satisfactory accuracy.
- Ordinal regression model: Slightly outperformed the RECIST-based models with better AUC and cross-validation accuracy.
These results support the clinical utility of revised RECIST 1.1 models, particularly V-RECIST for highest performance and U-RECIST for safe, accessible evaluation.