aceDevelopment of an Explainable Machine Learning Model for Predicting Reflux Esophagitis Among Candidates for Metabolic Bariatric Surgery
Abstract: Background Reflux esophagitis (RE) is a common gastroesophageal disorder among candidates for metabolic bariatric surgery (MBS). This study aimed to develop and internally validate an explainable machine learning (ML) model for predicting preoperative RE risk among patients undergoing MBS. Methods This retrospective study included 1,068 MBS candidates who were randomly divided into training and internal test cohorts (7:3). Feature selection was performed using correlation analysis, multicollinearity assessment,…
Sistema coletor de secre es by Robertolyra at Portuguese Wikipedia. CC BY-SA 3.0 · http://creativecommons.org/licenses/by-sa/3.0/
In a retrospective study published September 17, 2026, researchers developed and internally validated machine learning models to predict preoperative reflux esophagitis among 1,068 metabolic bariatric surgery candidates. After feature selection, six routine variables were used, and an XGBoost model achieved the best performance with AUCs of 0.881 in training and 0.825 in internal testing, supported by SHAP interpretation and an online risk tool.
The tool offers a practical way to stratify RE risk before surgery using routinely available clinical data, potentially informing perioperative management. As of the publication date, performance is based only on internal validation, and the authors note that external validation in independent cohorts is needed before broader clinical use.
- Retrospective study of 1,068 MBS candidates, 280 (26.2%) diagnosed with RE, split 7:3 training and internal test.
- Feature selection via correlation analysis, multicollinearity assessment, LASSO regression, and Boruta algorithm selected Sex, Hp, BMI, TG, HGB, and HC.
- Six ML algorithms evaluated; XGB showed best overall performance with SHAP analysis identifying TG, BMI, Hp, HC, HGB, and Sex as major contributors.
- Authors built an online risk assessment tool based on final XGB model for individualized prediction.
An explainable XGBoost model using six routine preoperative variables enabled individualized reflux esophagitis risk prediction for metabolic bariatric surgery candidates with AUC 0.881 training and 0.825 internal test.
The rundown
The study included 1,068 candidates for metabolic bariatric surgery, with 280 cases of reflux esophagitis identified preoperatively. Data were randomly divided into training and internal test cohorts at a 7:3 ratio for model development and evaluation.
Six predictors were retained after selection: Sex, Hp, BMI, TG, HGB, and HC. Six machine learning algorithms were compared on AUC, calibration, and classification metrics, with SHAP used for interpretation and an online tool deployed for individualized risk calculation.
Sources
- Peer-reviewedObesity Surgery2026-09-17
How should this claim be treated?
The debate