Internal Hernia after Laparoscopic Roux-en-Y Gastric Bypass: Optimal CT Signs for Diagnosis and Clinical Decision Making.

Abstract

Purpose To evaluate the accuracy of computed tomography (CT) for diagnosis of internal hernia (IH) in patients who have undergone laparoscopic Roux-en-Y gastric bypass and to develop decision tree models to optimize diagnostic accuracy. Materials and Methods This was a retrospective, ethics-approved study of patients who had undergone laparoscopic Roux-en-Y gastric bypass with surgically confirmed IH (n = 76) and without IH (n = 78). Two radiologists independently reviewed each examination for the following previously established CT signs of IH: mesenteric swirl, small-bowel obstruction (SBO), mushroom sign, clustered loops, hurricane eye, small bowel behind the superior mesenteric artery, and right-sided anastomosis. Radiologists also evaluated images for two new signs, superior mesenteric vein (SMV) "beaking" and "criss-cross" of the mesenteric vessels. Overall impressions for diagnosis of IH were recorded. Diagnostic accuracy and interobserver agreement were calculated, and multivariate recursive partitioning was performed to evaluate various decision tree models by using the CT signs. Results Accuracy and interobserver agreement regarding the nine CT signs of IH showed considerable variation. The best signs were mesenteric swirl (sensitivity and specificity, 86%-89% and 86%-90%, respectively; κ = 0.74) and SMV beaking (sensitivity and specificity, 80%-88% and 94%-95%, respectively; κ = 0.83). Overall reader impression yielded the highest sensitivity and specificity (96%-99% and 90%-99%, respectively; κ = 0.79). The decision tree model with the highest overall accuracy and sensitivity included mesenteric swirl and SBO, with a diagnostic odds ratio of 154 (95% confidence interval [CI]: 146, 161), sensitivity of 96% (95% CI: 87%, 99%), and specificity of 87% (95% CI: 75%, 93%). The decision tree with the highest specificity included SMV beaking and SBO, with a diagnostic odds ratio of 105 (95% CI: 101, 109), sensitivity of 90% (95% CI: 79%, 95%), and specificity of 92% (95% CI: 83%, 97%). Conclusion The decision tree with the highest accuracy and sensitivity for diagnosis of IH included mesenteric swirl and SBO, the model with the highest specificity included SMV beaking and SBO, and the remaining signs showed lower accuracy and/or poor to fair interobserver agreement. Overall reader impression yielded the highest accuracy for diagnosis of IH, likely because alternate diagnoses not incorporated in the models were considered. © RSNA, 2016 Online supplemental material is available for this article.