Development and Validation of a Predictive Model for Internal Hernia after Roux-en-Y Gastric Bypass in a Multicentric Retrospective Cohort: The Swirl, Weight Excess Loss, Liquid Score.

Objective The aim of this study was to develop and validate a prediction score for internal hernia (IH) after Roux-en-Y gastric bypass (RYGB). Summary background data The clinical diagnosis of IH is challenging. A sensitivity of 63% to 92% was reported for computed tomography (CT). Methods Consecutive patients admitted for abdominal pain after RYGB and undergoing CT and surgical exploration were included retrospectively. Potential clinical predictors and radiological signs of IH were entered in binary logistic regression analysis to determine a predictive score of surgically confirmed IH in the Geneva training set (January 2006-December 2014), and validated in 3 centers, Geneva (January 2015-December 2017) and Neuchâtel and Strasbourg (January 2012-December 2017). Results Two hundred twenty-eight patients were included, 80 of whom (35.5%) had surgically confirmed IH, 38 (16.6%) had a negative laparoscopy, and 110 (48.2%) had an alternate diagnosis. In the training set of 61 patients, excess body weight loss >95% (odds ratio [OR] 6.73, 95% confidence interval [CI]: 1.13-39.96), swirl sign (OR 8.93, 95% CI: 2.30-34.70), and free liquid (OR 4.53, 95% CI: 1.08-19.0) were independent predictors of IH. Area under the curve (AUC) of the score was 0.799. In the validation set of 167 patients, AUC was 0.846. A score ≥2 was associated with an IH incidence of 60.7% (34/56), and 5.3% (3/56) had a negative laparoscopy. Conclusions The score could be incorporated in the clinical setting. To reduce the risk of delayed IH diagnosis, emergency explorative laparoscopy in patients with a score ≥2 should be considered.