abstract
- OBJECTIVE: The objective of our study was to evaluate the diagnostic performance of CT in the identification of anastomotic leaks. MATERIALS AND METHODS: This was a study of patients who underwent bowel surgery and a subsequent postoperative CT examination performed specifically for investigating for an anastomotic leak. The study group included patients with surgically confirmed anastomotic leaks (n = 59), and the control group included patients without anastomotic leaks (n = 48) confirmed by either repeat surgery or uneventful clinical follow-up for at least 6 months. Two radiologists and two radiology residents independently reviewed each CT examination for specific CT findings from a set of predetermined imaging predictors. The sensitivity and specificity for each imaging predictor were calculated for each reader, and the interobserver agreement was calculated using the Cohen kappa coefficient. Diagnostic performance was assessed using ROC curve analysis. RESULTS: The most sensitive imaging predictor was intraabdominal free fluid (95.3%). Leakage of intraluminal contrast agent was also a highly specific imaging predictor (96.6%). Substantial interobserver agreement was shown for intraabdominal free gas (κ = 0.76) and leakage of intraluminal contrast agent (κ = 0.76). Overall diagnostic performance in correctly identifying surgically confirmed leaks, as assessed by the area under the ROC curve, ranged from 0.76 to 0.86. Diagnostic performance was higher for all readers when intraluminal contrast agent was used and reached the anastomosis, with the exception of one reader, whose diagnostic performance remained unchanged. CONCLUSION: Diagnostic performance of CT was highest when an intraluminal contrast agent was used. Meticulous and careful use of an intraluminal contrast agent is therefore important in this patient population.