P2748 - Risk Prediction for Early Postoperative Bleeding in Endoscopic Sleeve Gastroplasty: Development and Internal Validation of a Logistic Regression-Based Scoring System

Faisal Inayat, MBBS¹, Muhammad Sajeel Anwar, MD², Gul Nawaz, MD³, Ahtshamullah Chaudhry, MD⁴, Faisal Ibrahim, MBBS⁵, Hassam Ali, MD^6
1Allama Iqbal Medical College, Lahore, Punjab, Pakistan; ²UHS Wilson Medical Center, Johnson City, NY; ³Marshfield Medical Center, Marshfield, WI; ⁴St. Dominic's Hospital, Jackson, MS; ⁵Wexham Park Hospital, Slough, England, United Kingdom; ⁶East Carolina University Brody School of Medicine, Greenville, NC

Introduction: Gastrointestinal (GI) bleeding is a rare but important complication of endoscopic sleeve gastroplasty (ESG). Patients frequently express concerns about the likelihood of experiencing postprocedure GI bleeding. Despite the pressing need, no scoring systems are specifically designed to predict low-risk post-ESG patients. Supervised machine learning may offer a promising avenue for accurately categorizing the procedural risk profiles. We aim to develop and internally validate a scoring system determining the individualized risk of GI bleeding within the first 30 days after ESG.

Methods: Using the 2016-2021 Metabolic and Bariatric Surgery Accreditation Quality Improvement Program (MBSAQIP) database, 1,624 patients who underwent ESG were identified. Stepwise multivariate logistic regression was utilized to select the most promising predictors of GI bleeding following ESG within 30 days. The regression coefficients of predictive variables selected by stepwise regression were used to develop an algorithm. The algorithm's performance was assessed using receiver operator curves by 5-fold cross-validation.

Results: Stepwise multivariate logistic regression selected five of the 26 initial variables assessed with a P value < 0.05. These were female gender with an odds ratio (OR) of 0.22 (P = 0.008), preoperative weight (lbs) (OR 1.02, P < 0.001), hypertension (OR 0.14, P = 0.003), hyperlipidemia (OR 5.60, P = 0.005), and previous foregut surgery (OR 5.67, P = 0.009). In the logistic regression model, these variables were weighted based on their regression coefficients. For the risk model, an algorithm was subsequently developed. The cutoff score was established by the Youden index at 4.12; scores below this value were classified as having a low risk of bleeding following ESG. The area under the curve (AUC) for the risk model was determined to be 0.88 (95%CI: 0.72-0.93) (Figure 1). This resulted in sensitivity and specificity of 88.2% and 70.8%, respectively.

Discussion: This is the first population-based study that successfully developed and validated a scoring system predicting the risk of GI bleeding within 30 days post-ESG. The excellent discriminatory power of our risk assessment tool demonstrated its accuracy in predicting person-specific risk of gastrointestinal bleeding within the first 30 days post-ESG. This scoring system could prove to be a helpful tool for clinicians in determining which patients would benefit from targeted monitoring and therapy.



Disclosures:


Faisal Inayat, MBBS¹, Muhammad Sajeel Anwar, MD², Gul Nawaz, MD³, Ahtshamullah Chaudhry, MD⁴, Faisal Ibrahim, MBBS⁵, Hassam Ali, MD⁶. P2748 - Risk Prediction for Early Postoperative Bleeding in Endoscopic Sleeve Gastroplasty: Development and Internal Validation of a Logistic Regression-Based Scoring System, ACG 2024 Annual Scientific Meeting Abstracts. Philadelphia, PA: American College of Gastroenterology.