The main findings in this study showed that the majority of patients with duodenal fistulae can be successfully managed with surgical intervention consisting of duodenal repair or resection, especially when performed by a surgeon with experience in emergency surgery. In selected cases, nonoperative management with or without percutaneous drainage was successful even though some patients required surgery later.

Duodenal fistulas are rare and occur in 2–7% after repair of perforated peptic ulcer, in about 4% in associated with severe acute pancreatitis, 3% after gastrectomy for gastric cancer and 1% after endoscopic retrograde cholangio-pancreatography (ERCP) [20,21,22,23]. The main controversy regarding the management of postoperative duodenal fistulas is between initial surgical or nonoperative management, and as shown in Table 1, the successful fistula closure rates in series with more than 10 patients vary from 62 to 100%.

In a series of 24 patients with postoperative duodenal fistulas, management consisted of aggressive nutritional support, localization and drainage of intra-abdominal sepsis, and definitive surgical closure for those fistulas which did not close spontaneously. Spontaneous closure occurred in 22 (92%) cases; however, 14 patients had a total of 19 operations for drainage of localized wound or intra-abdominal abscess. The remaining two patients subsequently underwent definitive surgical closure at five and six weeks. The mortality rate was 8% [2].

Parc et al. reported 49 cases with postoperative peritonitis originating from a duodenal leak. The surgical management consisted of insertion of a spiral drain into the duodenum through the leaking site with its intraluminal end directed distally, external drainage and feeding jejunostomy. Infusion of 2000 ml/24 h with normal saline (containing thrombin, tranexamic acid and rifampicin) was started at the end of the operation and was continued for a mean of 21 days after which the spiral drain was removed and replaced with a 12-French silicone drain which was progressively removed to permit the closure of the fistula. In 32 (65%) patients, the duodenal fistula closed spontaneously at a median time of 39 (range 19–120) days. The overall mortality rate was 22% [6].

In the present series, the majority of patients (38/50, 76%) underwent initial surgical management and in all but one there was an attempt to close the duodenal defect by suture (22 patients), resection and anastomosis (6) or rectus muscle patch (one), resulting in successful fistula closure in 28/37 (76%) cases (Table 5). Specifically, in all 6 patients undergoing resection and anastomosis, the fistula remained closed.

One key element in successful repair seems to be the protection of the suture or anastomotic line with adequate duodenal decompression, either via antegrade naso-gastro-duodenal or retrograde duodenal tube, and sometimes augmented with cholecystectomy and insertion of a T-tube to achieve biliary diversion (Table 4). Duodenal decompression can also be achieved with a tube duodenostomy. It was used in a series of 31 patients with potentially insecure duodenal stump closure (12 patients) or postoperative duodenal leakage (12) through the open end of duodenum and augmented with a T-tube for biliary diversion in 19 patients. Only one patient (3%) had a subsequent duodenal stump leak which healed spontaneously [24].

Protection of the duodenal suture line by pyloric exclusion in the management of duodenal fistulae was reported already in 1907 and has subsequently been used in the management of traumatic duodenal perforations, although its benefit in traumatic setting has not been established [25,26,27]. In extreme situations, a duodenal diverticulization procedure (gastric antrectomy, tube duodenostomy, gastrojejunostomy, external periduodenal drainage and insertion of a T-tube for biliary drainage) has been used for extensive duodenal injuries [28] and was successfully used in one case in the present series.

Our current practice includes double-decompression; a nasogastric tube to the stomach and a naso-gastro-duodenal tube with extra side holes to the duodenum. In selected cases, we add a T-tube to facilitate the decompressive effect. Inserting a periduodenal external drain is mandatory and can sometimes control a secondary leak and help to avoid a reoperation. Enteral nutrition via feeding jejunostomy or in cases of a Roux-en-Y reconstruction via nasojejunal tube should be started as soon as possible. Although we used tube duodenostomy for decompression in past, we currently avoid making extra holes to the duodenum and replace tube duodenostomy with intraluminal decompression methods.

In selected cases, nonoperative management often combined with percutaneous drainage can be successful. In a systematic review of duodenal stump fistulae after gastrectomy for gastric cancer, conservative approach was performed in 79 stable patients with complete resolution achieved in 92% with a healing time ranging from 17 to 71 days [29]. Among 29 patients with external duodenal fistulae following closure of duodenal perforation and surviving 48 h, 14 patients (48%) were initially managed nonoperatively, out of which six required later surgery. In 15 patients, the indications for early surgery were peritonitis or failure to establish enteral feeding. The fistula closed spontaneously in 8/14 patients managed conservatively within 9–58 (mean 28) days, 2/6 patients with delayed surgery died. The mortality rate after initial operated patients was 9/15 (60%), but the successful closure rate was not reported [30].

Of the 12 patients undergoing initial nonoperative management in the present series, 5 were managed successfully with one or more percutaneous or endoscopic interventions, and one patient died of MOF and persistent fistula. Of the 6 patients with failed nonoperative management requiring subsequent surgery, the fistula remained closed in 4 patients, out of which one died later of sepsis caused by therapy-resistant intra-abdominal abscesses while the duodenum remained intact. In the remaining two cases, the fistula could not be closed despite surgery, and both patients died of MOF. It seems that in selected cases of stable patients and with no generalized peritonitis, nonoperative management including percutaneous and endoscopic drainage procedures can be attempted, but if failing to control the fistula, prompt operative intervention might rescue some of the patients.

In a literature review from 1865 to 1937, Bartlett and Holwell reported 130 cases of postoperative duodenal fistulae and added 12 cases of their own. It included two reports with 61 and 44 cases, where the mortality rates were 51% and 18%, respectively [31]. In more recent series, the mortality rate has varied between 8 and 42% [2, 4, 6, 22, 24, 30].

The overall 40% mortality rate in this series reflects the severity and the challenges facing surgeons treating these patients. Although some of the mortality can partly be associated with the underlying disease such as severe acute pancreatitis, obviously the duodenal leak and associated sepsis can be considered a significant contributing factor in all cases. Due to the small number of patients, no single independent risk factor for mortality could be identified.

The limitations of this study are related to its retrospective nature, small and heterogenous study population, and wide range of the management strategy.