Predicting strangulated small bowel obstruction: how the combination of CT findings with clinical and laboratory parameters improves diagnostic accuracy.

purposeComputed tomography (CT) is a promising tool for the evaluation of small bowel obstruction (SBO) and assessing the risk of strangulation. Despite its usefulness, CT alone might be insufficient to accurately predict strangulated SBO (SSBO). Therefore, the present study addressed the diagnostic performance of the combination of clinical, laboratory, and CT findings for predicting SSBO.

methodsMedical records of patients with a diagnosis of acute SBO who subsequently underwent surgery were reviewed. These patients were subsequently divided into two groups according to operative and histopathological findings: the SSBO (resection) and the non-SSBO (non-resection) groups. Eight clinical, seven laboratory, and six CT findings were then compared between the two groups to determine the most clinically useful variables and develop prediction models for accurate identification of SSBO.

results150 patients (mean age: 54.5 ± 18.6 years, 51 (34%) patients with SSBO) were included in the analysis. The univariate analysis identified the density of ascites (ROC-AUC: 0.87, cut-off point = 10 HU, sensitivity = 96%, specificity = 62%), mesenteric fluid (ROC-AUC: 0.80, sensitivity = 76%, specificity = 83%), C-reactive protein (CRP) (ROC-AUC: 0.79, sensitivity = 75%, specificity = 76%) and abdominal guarding (ROC-AUC: 0.76, sensitivity = 98%, specificity = 54%) as the variables most strongly associated with SSBO. Multivariate analysis demonstrated that a prediction model consisting of density of ascites, mesenteric fluid, CRP, and abdominal guarding demonstrated the highest predictive performance (ROC-AUC: 0.96, sensitivity = 94%, specificity = 86%), being superior to a bivariate radiologic (density of ascites + mesenteric fluid ) model (ROC-AUC: 0.88, sensitivity = 81%, specificity = 88%) and multivariate models with all clinical (ROC-AUC: 0.88, sensitivity = 76%, specificity = 90%), laboratory (ROC-AUC: 0.89, sensitivity = 86%, specificity = 78%), and radiologic (ROC-AUC: 0.92, sensitivity = 81%, specificity = 90%) parameters.

conclusionIn summary, although a bivariate radiologic model (mesenteric fluid + density of ascites) could predict SSBO with moderate sensitivity (81%) and high specificity (88%), the addition of CRP level and abdominal guarding further improves sensitivity to 94%. These findings could assist surgeons in facilitating the identification of SSBO and help avoid unnecessary surgeries, particularly when CT findings are equivocal.