Early detection of anastomotic leakage in upper gastrointestinal surgery

BackgroundAlthough surgical methods for upper gastrointestinal tract cancer continue to advance with the aim of reducing the incidence of anastomotic leakage (AL), it remains a prevalent and serious complication. Therefore, early identification of patients at high risk for AL is necessary for timely therapeutic measures. MethodsWe retrospectively identified patients with AL who underwent elective gastric/esophageal resection at our department between 2005 and 2017. Using propensity score matching, a comparison group without AL but with comparable baseline characteristics was developed. Several previously published risk scores (o-POSSUM, E-PASS, Steyerberg, NUn score) were calculated, and their predictive accuracy for presence of AL was compared. ResultsSteyerberg Risk Score, o-POSSUM, and E-PASS were found to be unsuitable for early detection of an anastomotic problem. However, an increased NUn score on the fourth to seventh postoperative day was independently associated with the presence of AL. The test accuracy (0.631-0.714), sensitivity (28.9%-50.5%), and specificity (72.5%-89.5%) were marginally satisfactory. When only C-reactive protein levels were considered, similar test accuracy (0.629-0.717), sensitivity (47.5%-69.9%), and specificity values (51.5%-81.6%) were observed using a cut-off of 150 mg/l. ConclusionsThe NUn score showed no superiority over CRP values in the prediction of AL. Therefore, further diagnostics should be carried out from the fourth postoperative day if the CRP is > 150 mg/l. However, large-scale registry studies and artificial intelligence may aid in more appropriate determination of patient-specific risks in the future.