Incidence and management of postoperative bile leaks: A prospective cohort analysis of 467 liver resections

Abstract

Background: Bile leaks from the parenchymal transection margin are a major cause of morbidity following major liver resections. The aim of this study was to benchmark the incidence and identify the risk factors for postoperative bile leakage after hepatic resection.
Patients and methods: A prospective database of 467 consecutive liver resections performed by the University of Cape Town HPB surgical unit between January 1990 and January 2016 was analysed. The relationship of demographic, clinical and perioperative factors to the development of bile leakage was determined. Bile leak and postoperative complications severity were graded using the International Study Group of Liver Surgery and Accordion classifications.
Results: Overall morbidity was 24% (n = 112), with bile leaks occurring in 25  (5.4%) patients. Significantly more bile leaks occurred in patients who had major resections (≥ 3 segments) and longer total operative times (p < 0.05). There were 5 Grade A bile leaks which stopped spontaneously. Seventeen Grade B leaks required a combination of percutaneous drainage (n = 15), endoscopic biliary stenting (n = 8) and percutaneous transhepatic biliary drainage (n = 3). All 3 Grade C leaks required laparotomy for definitive drainage. Median hospital stay in the 442 patients without a bile leak was 8 days (IQR 1–98) compared with 12 days (IQR 6–30) for the 25 with bile leaks (p < 0.05) with no mortality. Major resections (≥ 3 segments) and total  operative time (> 180mins) were significantly associated with bile leaks.
Conclusion: The incidence of bile leakage was 5.4% and occurred after major liver resections with longer operative times and resulted in significantly extended  hospitalisation. Most were effectively treated nonoperatively by percutaneous drainage of the collection and/or endoscopic or percutaneous biliary drainage without mortality.

Key words: operation, liver resection, bile leak.