Abstract
BACKGROUND: Structured assessment of abdominal computed tomography (CT)-scans is increasingly used to identify signs of internal herniation after Roux-en-Y gastric bypass (RYGB), aiding in the decision-making process to perform a diagnostic laparoscopy (DLS). This study aimed to develop a prediction score based on structured assessment of CT-signs for internal herniation. METHODS: Patients presenting with abdominal pain after RYGB, who underwent a CT-scan for suspicion of internal herniation and subsequently DLS, were included. CT-scans were reassessed for presence of ten CT-signs for internal herniation by two radiologists and two registrars. Diagnostic accuracy for detection of internal herniation for each sign and an overall suspicion score were calculated and compared with the original CT-reports. Interobserver agreement was measured using Fleiss' kappa. A prediction score was developed based on variables identified by multivariable logistic regression. RESULTS: With DLS 44 internal herniations (114 CT-scans, 92 patients) were identified. Structured assessment improved diagnostic accuracy compared to the original CT-report (AUC of 0.69 to 0.79, p = 0.03), and the positive (67% to 81%) and negative predictive value (75% to 82%). The three-sign prediction score (venous congestion, swirl sign, right-sided anastomosis) resulted in improved diagnostic accuracy compared to the original CT-report (AUC of 0.69 to 0.79, p = 0.038). Interobserver agreement of these signs was adequate between all readers (K = 0.56-0.75). CONCLUSIONS: Structured assessment of CT-scans improves diagnostic accuracy for internal herniation after RYGB. Our three-sign prediction-model offers a simplified, reproducible alternative to extensive assessment, without compromising the improved diagnostic effectiveness.