Grant Number:	1R01CA295896-01A1 Interpret this number
Primary Investigator:	Rubenstein, Joel
Organization:	University Of Michigan At Ann Arbor
Project Title:	Advancing the Use of the Kettles Esophageal and Cardia Adenocarcinoma Prediction Tool
Fiscal Year:	2026

Esophageal and gastric cardia adenocarcinomas (EAC/GCA) are clinically similar cancers that have been rising in incidence while incidence for most other cancers has been decreasing. EAC/GCA rarely present at an early stage, and overall 5-year survival is less than 20%. And yet, NIH funding for EAC/GCA research is disproportionally low compared to its lethality. The individuals at highest risk include White men with symptoms of gastroesophageal reflux disease who are at least 60 years of age, but they represent less than 1/3 of all cases of EAC/GCA. Screening by esophagogastroduodenoscopy (EGD) for the precursor, dysplastic Barrett’s esophagus (BE), is associated with reduced mortality, and has been endorsed by specialty societies for higher risk individuals. But fewer than 20% of those with EAC/GCA had an EGD prior to their cancer diagnosis. EAC/GCA are not common enough to warrant population-wide screening, so a personalized approach to screening is necessary. Validated prediction tools exist but are not commonly used. A tool that can be easily implemented is critical, such as by incorporation in the electronic health record (EHR). We used machine learning to adapt the tools for use within the Veterans Health Administration (VHA) EHR among 11,395 cancer cases and 10 million controls. By imputing missing data and leveraging novel associations with routine laboratory values and diagnostic codes, we developed and internally validated the Kettles Esophageal and Cardia Adenocarcinoma predictioN tool (K-ECAN). Despite these advances, there are key unanswered questions regarding K-ECAN, including the accuracy of K-ECAN in other healthcare systems, the optimal threshold risk above which screening should be offered, and barriers and facilitators to implementing K-ECAN across diverse care delivery settings. Therefore, we aim to: 1. Improve K-ECAN and externally validate its accuracy for identifying patients with dysplastic BE and predicting EAC/GCA. 2. Estimate the optimal cost-effective threshold K-ECAN score above which to offer screening. 3. Identify barriers and facilitators to implementing K-ECAN, and implementation strategies to support its uptake and adoption across diverse care delivery settings. EAC/GCA are screening-preventable cancers, but they are too uncommon to justify mass screening. Instead, a personalized approach to screening is necessary. K-ECAN can be rapidly scaled up for use by providers to select patients for screening. But key knowledge gaps need to be addressed first by this proposal. Completion of this award will lead to improvements in, and adaptation and validation of K-ECAN for EHRs beyond VHA and to planning trials of using K-ECAN to select patients for screening. Implementation of such an intervention would bridge the gap between population-health screening and precision medicine.





None