Background
Detecting anastomotic leakage in an early stage after colorectal surgery is essential to minimize morbidity. Computer Tomography (CT) has a relatively good specificity in detecting anastomotic leakage but sensitivity is low. This study investigates which radiological criteria are most predicitive for anastomotic leakage and if a set of criteria can improve sensitivity of CT- imaging.

Materials and Methods
All patients that underwent CT-imaging within 16 days after colorectal surgery between jan 2006 and december 2012 because of clinical suspicion of anastomotic leakage were included for analysis (108 patients). Two independent radiologists, blinded for the clinical outcome, reviewed all the CT-scans based on a set of potential predictive criteria for AL.

Results
Anastomotic leakage was determined at relaparotomy in 34 (30.6%) of the 108 patients with a clinical suspicion of AL. Review of CT-scans based on the set of criteria resulted in sensitivity of .68 (95% CI .52-.83) for radiologist 1 and of .79 (95% CI 0.66-0.93) for radiologist 2. Specificity was .87 (95%CI .79-.94) for both radiologists. Univariate analysis revealed fluid, air near the anastomosis, air intra-abdominally and contrast leakage to be associated with. Multivariate analysis showed that only contrast leakage was a significant predictor for both radiologists.

Conclusion
Contrast leakage is the only independant predictor for AL. Sensitivity is not improved to a satisfactory level using a selected set of criteria. Cautiousness in the interpretation of CT-images is required.