Clinical decision tree for optimizing endoscopic assessment of signet ring cell carcinoma in hereditary diffuse gastric cancer surveillance

Abstract

Prophylactic total gastrectomy is the definitive treatment for hereditary diffuse gastric cancer syndrome (HDGC). Endoscopic surveillance informs the requirement for and optimal timing of surgery. However, endoscopic recognition of early signet ring cell carcinoma (SRCC) remains challenging. We developed an endoscopic framework to optimize SRCC assessment during HDGC surveillance.We retrospectively analyzed data from 147 individuals with HDGC undergoing endoscopic surveillance to evaluate the diagnostic accuracy of the endoscopic Cambridge criteria. We used machine learning to develop a clinical decision tree (cDT) to guide the application of the Cambridge criteria. We then prospectively validated the cDT in 66 CDH1 pathogenic-variant carriers. The interobserver agreement and diagnostic accuracy of the Cambridge criteria and cDT were assessed through a multi-reader multi-case study.Retrospective analysis of 537 endoscopies showed that the Cambridge criteria achieved 82.8% (48/58) sensitivity and 78.2% (140/179) specificity for SRCC diagnosis. The presence and number of neoplastic pale areas were independent predictors of higher cancer burden in HDGC individuals. In the prospective study, cDT had 77.8% (21/27) sensitivity and 90.7% (49/54) specificity, and improved performance of both experts and non-experts.We developed and validated a practical endoscopic framework for enhancing SRCC assessment during HDGC endoscopic surveillance.