MONDAY, Aug. 28, 2023 (HealthDay News) -- The Kettles Esophageal and Cardia Adenocarcinoma predictioN (K-ECAN) Tool can predict incident esophageal adenocarcinoma (EAC) and gastric cardia adenocarcinoma (GCA) using electronic health record data, according to a study published online Aug. 17 in Gastroenterology.
Joel H. Rubenstein, M.D., from the LTC Charles S. Kettles Veterans Affairs Medical Center in Ann Arbor, Michigan, and colleagues accessed the Veterans Health Administration (VHA) Corporate Data Warehouse to identify veterans with one or more encounters between 2005 and 2018. Cases diagnosed with EAC or GCA (8,430 and 2,965, respectively) were compared to 10,256,887 controls. The K-ECAN Tool was developed and validated internally using a machine learning method. Training, preliminary validation, and final testing was performed in 50, 25, and 25 percent of the data, respectively.
The researchers found that compared with previous models such as HUNT and Kunzmann or published guidelines, K-ECAN was well calibrated and had better discrimination (area under the receiver operating characteristic curve [AuROC], 0.77 versus 0.68 and 0.64, respectively). The accuracy was slightly attenuated using only data between three and five years before index (AuROC, 0.75). The AuROCs of HUNT and Kunzmann improved on undersampling men to simulate a non-VHA population, but K-ECAN was still the most accurate (AuROC, 0.85). Gastroesophageal reflux disease was strongly linked to EAC, but contributed a small proportion of gain in information for prediction.
"Our devoted team was able to use sophisticated machine learning tools to develop this unique tool, and we are very excited that this could potentially lead to increased screening and a decrease in preventable deaths," Rubenstein said in a statement.
Two authors disclosed ties to the pharmaceutical industry.
Abstract/Full Text (subscription or payment may be required)
