The etiology of bleeding is unknown. Following Roux-en-Y gastrojejunostomy, lack of inactivation of pepsin by bile acids, ischemia [13], or reverse peristalsis have all been proposed as an underlying mechanism. For this reason, prophylactic therapy with proton pump inhibitors and even bilateral truncal vagotomy has been advocated [14–16]. In a systematic review comparing open versus laparoscopic Roux-en-Y gastric bypass, the risk of bleeding was found to be significantly higher following the laparoscopic procedure, suggesting that ischemic injury may be an important etiological factor [3].

Following surgery, patients may bleed into the abdominal cavity or into the lumen from the anastomosis [17]. Some have proposed that the primary role of placing drains at the site of anastomosis is to allow the early diagnosis of postoperative hemorrhage and to distinguish this from an anastomotic or staple line leak [18]. However, drains are not always a reliable indicator, particularly if the bleeding is intraluminal. For this reason, monitoring of clinical signs is very important in the postoperative patient. A large quantity of bloody fluid from the abdominal drains, hemoperitoneum, tachycardia, a drop in the hemoglobin level, bright red blood per rectum, hematemesis, and melena have all been found to indicate postoperative hemorrhage [19].

Patients with evidence of extra-luminal hemorrhage usually require abdominal re-exploration using either a laparoscopic or open approach. The operative goals are to evacuate the majority of the clots, attempt to identify and control the site of hemorrhage, or, if there is no obvious bleeding point, to oversew all staple lines.

In an interesting study [20], endoscopy was used intraoperatively to assess the anastomosis at the time of primary surgery. A total of 118 patients underwent resectional surgery and 5 were found to have bleeding at the site of the anastomosis at the time of surgery. Unfortunately, in the follow-up phase, a further 8 cases of anastomotic bleeding were reported. These cases had not bled at the time of the surgery; it is therefore possible that intraoperative endoscopy identifies early, self-limiting bleeding points but is unable to prevent cases of delayed bleeding perhaps developing as a consequence of ischemic injury. A smaller study, using a similar intraoperative endoscopic assessment of the surgical anastomosis, reported some benefit during bariatric procedures [21]. However, it is still uncertain if this approach should become part of the standard intrasurgical assessment of the anastomosis.

There is evidence that bleeding from anastomotic ulcers are difficult to treat endoscopically. A retrospective analysis of 427 patients with bleeding from an anastomotic ulcer reported an odds ratio of 3.39 (95 % CI 1.37–7.29) for therapeutic failure [22].

In a smaller, retrospective review of 393 patients who had undergone laparoscopic-assisted gastrectomy, all cases of bleeding could be managed endoscopically [23]. I am only aware of a single report of fatal bleeding from an anastomotic ulcer [24].

Emergency gastroscopy is able to identify the site of bleeding in approximately 95 % of cases of peptic ulcer bleeding [25, 26]. However, there is no corresponding data available on the success rate for bleeding anastomoses. Nevertheless, it seems reasonable to initially attempt endoscopy with the aim to (1) identify the source of hemorrhage, (2) stop the bleeding, (3) assess the risk of rebreeding, and (4) outline a strategy of how the case should be managed if bleeding cannot be stopped or recurs.

Naturally, not every patient who develops bleeding following surgery suffers bleeding from the surgical anastomosis. Other causes to be considered include ischemia, “stress” ulcers, reflux esophagitis, coagulopathy, and aortoenteric fistula. Patients with an intraluminal upper GI source of bleeding will usually present with hematemesis and/or melena. Intraluminal bleeding in the lower GI tract would normally present with rectal bleeding.

The management of a patient with early postoperative gastrointestinal bleeding does not differ significantly from the management of common GI hemorrhage. After the bleeding site has been identified, it should be risk stratified depending on the stigmata of recent hemorrhage (SRH). Major SRH includes active arterial bleeding or a non-bleeding visible vessel [28]. Intermediate SRH includes an adherent clot on an ulcer or an ulcer that is oozing slightly without concomitant major SRH. Minor SRH includes ulcers with a pigmented flat spot or a clean ulcer base. Ulcers with major SRH have the greatest risk of continued bleeding or rebleeding without endoscopic therapy, whereas minor SRH have the lowest risk of rebleeding even without endoscopic therapy [29–32]. Ulcers with intermediate SRH have an intermediate risk of bleeding without endoscopic therapy [33]. Although this stratification was developed for bleeding peptic ulcers, it would be reasonable to assume that the same is true for bleeding ulcers at the anastomotic site.

Bleeding sites should first be treated with an initial injection of dilute adrenaline to provide tamponade. In bleeding peptic ulcers, adrenaline provides initial hemostasis in around 80 % of bleeding peptic ulcers [34, 35]. Unfortunately, bleeding may recur after 20 min as the adrenaline is absorbed [26].

The use of a sclerosing agent such as alcohol or ethanolamine does cause tissue injury at the site of the anastomosis and should be avoided. Furthermore, there is no evidence that they provide any advantage. In a study of 170 cases of bleeding peptic ulceration, injecting a sclerosant did not provide any better hemostasis than adrenaline [36]. The use of histoacryl tissue glue has been shown to be of benefit in bleeding gastric varices but is associated with risks of embolization [37] and is unproven in the treatment of bleeding from surgical anastomoses.

Fibrin, when injected into the bleeding vessel, may aid the formation of a clot. Unfortunately, clinical studies have not demonstrated any clear advantage over adrenaline in bleeding peptic ulcers. In a European multicenter trial [38] of more than 800 patients, bleeding ulcers were initially treated with adrenaline. The ulcers were then either injected with sclerosant (group A), fibrin glue (group B), or an intensive program of multiple fibrin glue injections over several days (group C). Initial hemostatic success was similar in the groups (92.5 % vs 95.1 % vs 96.4 %). However, patients in group C were less likely to rebleed than in group A (10.0 % vs 18.1 %). These results were corroborated in a subsequent, smaller study encompassing 135 patients. In this study, bleeding peptic ulcers treated with a single injection of adrenalin were just as likely to rebleed as ulcers treated with single injection fibrin glue (24 % vs 21 %) [39]. There has been one case report of the successful use of tissue glue to treat a bleeding varix at the site of a surgical anastomosis [40].

Hemoclip s are short endoscopic clips only available in countries which do not solely rely on disposable equipment (such as the UK). Initial results were very promising with an Italian study reporting a rebleeding rate of 21 % following heater probe therapy versus only 1.8 % after the placement of hemoclips [41]. Unfortunately, a subsequent study [42] did not replicate the initial, promising results. Injection therapy achieved initial hemostasis in 85 %, compared with 63 % for hemoclips, or 75 % when a combination of clips and injection therapy was used. At the site of a small bowel anastomosis, clips have the theoretical advantage that relatively little tissue injury is incurred by their use. Furthermore, the main shortcoming of hemoclips is that the main simply scrape the surface of a firm and sclerotic peptic ulcer. At a fresh, bleeding ulcer at a small bowel anastomosis, less fibrosis may be expected.

Proton pump inhibitor (PPI) therapy is recommended for patients with bleeding peptic ulceration as the neutralization of gastric acid stabilizes clots [43–45], reduces the need for endoscopic therapy, and reduces the risk of rebleeding but has no proven effect on mortality [46]. However, when the bleeding is from an anastomotic ulcer, the effect of PPI therapy is unproven and cannot be recommended.

The largest series reporting on the outcomes of treating a bleeding anastomosis was published by Lee et al. [47] which included a series of 50 patients with anastomotic ulcer bleeding following Billroth-II procedures. The bleeding ulcers were usually found at the level of the anastomosis (64 %) or at the anastomotic bifurcation (20 %). The ulcers were classified as Forrest Ia (32 %), Ib (38 %), IIa (20 %), or IIb (10 %). A total of 20 patients were treated with hemoclips and 30 with a combination of adrenaline and heater probe coagulation. In all cases, initial hemostasis was achieved. However, the risk of rebleeding was significantly greater (33 % vs 5 %) following adrenaline + heater probe coagulation than after endoclip application. The likely reason for this is the combination of tissue ischemia, a thin small bowel wall and tissue damage from the heater probe. Endoscopic therapy for non-anastomotic bleeding is associated with a 0.5–1.0 % risk of gastrointestinal perforation [48, 49] and the risks when treating a bleeding ulcer at a surgical anastomosis are probably greater.