Acute non-variceal upper gastrointestinal bleeding
Colin J. McKay and James Lau
A major audit of upper gastrointestinal (UGI) bleeding in the UK carried out over 15 years ago reported an incidence of 103 cases per 100 000 adults per year.1 A much higher incidence was reported in an audit from the West of Scotland, which found 172 cases per 100 000.2 The difference was thought to be due to regional variation in deprivation and possible Helicobacter pylori prevalence. Whatever the actual incidence at regional level, UGI bleeding remains a significant health problem and in Scotland alone accounts for approximately 7000 hospital admissions each year. Overall mortality from an episode of UGI bleeding was 14% in the UK study1 and 8% in the study from the West of Scotland,2 and both demonstrated a marked increase in mortality if bleeding occurred following hospitalisation with another complaint (33% and 44% in the respective studies). A more recent audit3 covering patients admitted to UK hospitals during a 2-month period in 2007 demonstrated a marked reduction in the proportion of patients undergoing surgery, as well as an increase in the proportion of patients admitted with variceal bleeding. Overall mortality had fallen slightly at 10%, compared with 14% in the earlier UK audit, and the mortality of 26% for inpatients was also lower than in the pevious study. However, mortality in patients admitted with peptic ulcer bleeding had fallen from 8.8% to 5.8% and the overall reduction in mortality from 14 to 10% occurred despite the increase in incidence of variceal bleeding. Mortality is now rare in the absence of comorbidity. The typical patient with severe peptic ulcer bleeding is now elderly, often suffering medical comorbidity and taking antiplatelet therapy. Such patients are at greater risk of death despite skilled intervention and are less able to withstand surgery should this be necessary. Although the need for surgical intervention is now much reduced with the widespread availability of skilled endoscopic haemostasis, mortality following surgery remains high. In the recent UK audit, postoperative mortality was 30% and, while this is likely to reflect the fact that surgery has become a last resort in patients who are often elderly and suffering from other conditions, there are inevitable concerns about the availability of appropriate expertise in the management of what has become an uncommon surgical emergency. As a result alternatives to surgical intervention are increasingly being employed and will be discussed in this chapter.
In the Rockall study1 only 4% of patients had UGI bleeding due to varices, with the majority (35%) being attributed to peptic ulcer disease. Of some concern was the 25% of patients in this study where no cause for bleeding was identified at all, particularly as this group had a mortality of 20%. Very similar figures were reported in the Scottish study.2 Of those patients with a diagnosis, approximately 25% were due to duodenal ulcer and 25% to peptic ulceration in the stomach, oesophagus or a combination of sites. The remainder were due to a number of other conditions, including oesophagitis, gastritis and duodenitis, with 10% of cases explained by malignancy or Mallory–Weiss syndrome (Table 7.1). In the most recent UK audit, the proportion of patients with variceal bleeding had increased to 11%, whereas the overall proportion with peptic ulcer remained unchanged.
Endoscopic diagnoses to patients who presented with acute upper gastrointestinal bleeding in the 2007 UK Audit.
Total number of patients (5004)
Others including vascular ectasia
No abnormality seen
Modified from Hearnshaw S et al. Gut 2011; 60: 1327-1335
Initial assessment and triage
Patients with acute UGI bleeding present with haematemesis, melaena or a combination of the two. Haematemesis may be defined as the vomiting of blood from the upper gastrointestinal tract and is indicative of significant bleeding from a site in the oesophagus, stomach or duodenum. Melaena is the passage of black, tarry stools and usually indicates a bleeding site in the upper gastrointestinal tract, although bleeding from the small bowel or even the right colon may present in a similar way, depending on speed of passage. Coffee-ground vomiting (the vomiting of black, particulate material that tests positive for blood) is not indicative of active bleeding. Presentation with haematemesis is associated with an increased risk of mortality compared with melaena or coffee-ground vomiting.4 Patients with upper gastrointestinal blood loss may occasionally present with frank rectal bleeding (haematochezia), but this is indicative of major blood loss and, not surprisingly, is associated with an increased need for transfusion, surgery and mortality.5 A good example of this would be a patient who has previously undergone aortic aneurysm surgery who subsequently develops an aortoduodenal fistula.
Many patients with UGI haemorrhage have trivial bleeding, require no intervention and can safely be discharged early. Others, however, have catastrophic bleeding and many lie between these two extremes. On initial assessment, differentiation between patients can sometimes be difficult. Two separate studies have assessed the initial, pretreatment risk factors in an attempt to enable triage of patients to appropriate early endoscopy, observation or even non-admission. In the English audit discussed above, Rockall and colleagues identified risk factors for mortality.5 Patients with an initial score of zero (i.e. age < 60, no tachycardia, no hypotension and no comorbidity) had very low mortality, although in a subsequent study 18% of patients with an initial Rockall score of zero required further intervention. Assessment of the re-bleeding risk therefore requires the full Rockall score, which necessitates endoscopic assessment. This is discussed below.
In contrast, a scoring system was developed in the West of Scotland, the aim being to allow early assessment of the need for intervention rather than risk of death.6 The Glasgow Blatchford Score (GBS) performed significantly better than the Rockall score at predicting the need for intervention and has now been validated in prospective studies from the UK7 and Hong Kong.6 In particular, the GBS is the better scoring system for predicting low-risk patients who can be safely discharged. The proportion of patients with UGI haemorrhage who fulfil these low-risk criteria (i.e. a GBS of 0) is, however, low (4.6% in the Hong Kong study). The proportion of patients who require endoscopic therapy increases with higher scores. There remains a small but significant proportion of patients with low to moderately high scores who require endoscopic therapy and it is difficult to define a cut-off score beyond which urgent endoscopy becomes mandatory.
Full assessment of mortality risk demands endoscopic determination of the cause of bleeding and assessment of the appearance of any ulcer present. These findings are incorporated into the full Rockall score. This composite Rockall system has been validated in prospective studies,8–10 but is more accurate in predicting mortality than the risk of re-bleeding8,9 (Table 7.2).
Rockall score for predicting risk of re-bleeding or death In non-variceal UGI haemorrhage
IHD, ischaemic heart disease; SBP, systolic blood pressure; SRH, stigmata of recent haemorrhage.
Reproduced from Rockall TA, Logan RFA, Devlin HB et al. Variation in outcome after acute upper gastrointestinal haemorrhage. Lancet 1995; 346(8971):346–50. With permission from Elsevier.
A GBS of 0 on presentation identifies a low risk group of patients who may be safely discharged without the need for urgent UGI endoscopy.
A full Rockall score should be calculated for all patients with UGI bleeding following initial endoscopic assessment.
While it may be appropriate to consider discharge for young patients with no haemodynamic compromise and a GBS of zero, particularly where there has been no witnessed frank haematemesis, most patients with a history of UGI bleeding will be admitted for observation and endoscopy. This will entail insertion of a large-bore cannula, administration of prewarmed crystalloid or colloid as required and immediate blood samples taken for crossmatch, biochemistry, full blood count, coagulation screen and arterial blood gases. There is some evidence that admission to a dedicated UGI bleeding unit is associated with a reduction in mortality.11 Such a unit requires a 24-hour on-call service for immediate endoscopy if required and early (within 24 h) consultant-led endoscopy for all patients. Indications for urgent (out-of-hours) endoscopy vary, but the main factor determining the degree of urgency is the necessity for endostasis. Therefore, patients with evidence of continuing haemorrhage, declared either by continuing haematemesis or haemodynamic instability despite initial fluid resuscitation, require emergency endoscopy with a view to endostasis. Repeated ‘coffee-ground’ vomiting or melaena in a haemodynamically stable patient are of less urgent concern. Most stable patients will undergo UGI endoscopy within 24 h (usually on the morning after admission). In such patients the purpose of endoscopy is twofold: firstly, patients with minor bleeds undergo full diagnostic assessment and if considered at low risk of re-bleeding can be discharged home; secondly, to identify the group of patients who have significant lesions and who require endoscopic therapy to reduce the risk of re-bleeding. This will be discussed in the next section.
In those patients who have evidence of haemodynamic compromise, initial resuscitation should follow the appropriate guidelines.11 A Cochrane review of 55 trials found no evidence of any benefit of administering colloid rather than crystalloid solutions during resuscitation in critically ill patients.12 As colloid solutions are more expensive, initial resuscitation with appropriate crystalloids is recommended.
Use of blood and blood products
Red cell replacement is likely to be required when 30% or more of the blood volume is lost. This can be difficult to assess, particularly in young patients, and clinical assessment of blood loss, coupled with the response to initial volume replacement, must guide the decision on the necessity of transfusion. Haemoglobin levels are a poor indicator of the degree of acute blood loss. It should very rarely be necessary to administer unmatched blood to such patients, but group-specific blood may be required in life-threatening haemorrhage. In the majority of patients there is usually sufficient time to allow full compatibility testing before transfusion. In patients with evidence of continuing haemorrhage, however, arrangements for emergency endoscopic intervention must be made in parallel with resuscitation.
Administration of platelets should aim to maintain a platelet count of more than 50 × 109/L, but in practice a platelet transfusion should be triggered by a count of less than 75 × 109/L in a patient with ongoing haemorrhage, and may be anticipated in a patient who has required more than two blood volumes of fluid/blood replacement. Platelet therapy may be required in the presence of a normal platelet count where a patient has been receiving antiplatelet therapy.
Coagulation factors are likely to be required when more than one blood volume has been lost. These are most commonly given in the form of fresh frozen plasma (FFP). The use of platelets, FFP and other agents such as recombinant factor VIIa should be guided by local protocols and early involvement of a haematologist.
Early Pharmacological Treatment
Upper gastrointestinal endoscopy is the mainstay of investigation and management of UGI bleeding, but there may also be a role for early treatment with acid suppression therapy. In vitro studies have shown that, at pH < 6, platelet aggregation and plasma coagulation are markedly reduced,13 a situation exacerbated by the presence of pepsin. It is therefore reasonable to expect acid suppression therapy to promote clot formation and stabilisation. Six randomised trials comparing pre-endoscopy proton-pump inhibitor (PPI) therapy with histamine-2 receptor antagonist or placebo were analysed in a Cochrane review.14 No significant impact of PPI therapy was demonstrated on mortality, surgery or re-bleeding rates. There was, however, a reduction in the proportion of patients with stigmata of recent haemorrhage at the time of endoscopy, and a reduction in the requirement for endoscopic therapy at the index endoscopy. While these findings may suggest that pre-endoscopy PPI therapy is warranted, there was no evidence of a reduction in any of the clinically significant parameters. In a large, single-centre, randomised study,15 pre-endoscopy intravenous omeprazole infusion was associated with a reduction in the incidence of active bleeding and significant reduction in the necessity for endoscopic intervention. However, no difference was observed for re-bleeding, surgery or mortality rates, which were observed in 3%, 2% and 2% of patients, respectively. An accompanying cost-effective analysis concluded that the strategy of pre-emptive use of PPI infusion was cost-saving, probably because of reduced endoscopic therapy and hospitalisation.16 It therefore seems reasonable to propose pre-endoscopic PPI therapy in patients admitted with UGI bleeding, but this should not delay or act as a substitute for early endoscopic intervention. There is no evidence to support the use of other agents such as somatostatin, octreotide or vasopressin in the pre-endoscopy setting, except where variceal bleeding is suspected.
Pre-endoscopy treatment with PPIs is recommended as it reduces the number of actively bleeding ulcers and increases the number of clean-based ulcers seen at the time of endoscopy.14 Early use of PPI reduces the need for endoscopic intervention and hospitalisation.
Upper gastrointestinal endoscopy is required for any patient with significant UGI bleeding. Patients with haemodynamic instability or evidence of continuing haemorrhage require emergency endoscopy, whereas the majority of patients will undergo endoscopy within 24 hours of admission. In a study from Hong Kong,17 70 patients aged less than 60 years with a clean ulcer base and stable vital signs were safely discharged on the same day as endoscopy with appropriate anti-ulcer medication. A systematic review of the literature supports a policy of early endoscopy, as this allows the safe discharge of patients with low-risk haemorrhage and improves outcome for patients with high-risk lesions.18
Early endoscopy is recommended for all patients with UGI haemorrhage.18
Endoscopy for UGI bleeding requires the support of a dedicated endoscopic unit with trained nursing staff, availability of additional endoscopes and equipment, ready access to anaesthetic staff and operating theatre, and, increasingly, access to interventional radiology services. These procedures are not ideal for the unsupervised trainee and should be performed or supervised by experienced consultant staff.
For the majority of stable patients, procedures can be safely carried out using standard diagnostic endoscopes. In the unstable patient or where continuing haemorrhage is suspected, the twin-channel or large (3.7-mm) single-channel endoscope is preferable and allows better aspiration of gastric contents as well as more flexibility with regard to the use of heater probes and other instruments. In unstable or obtunded patients, anaesthetic support is mandatory as an endotracheal tube should be passed before endoscopy to guard against aspiration. In rare cases, the use of a pharyngeal over-tube and gastric lavage tube is necessary to remove blood and clot from the stomach before adequate visualisation of the bleeding site can be achieved. Water is poured down the lavage tube via a funnel, which is then placed in a bucket at floor level, allowing siphoning of gastric contents. This, however, can be a time-consuming, unpleasant and messy experience, and in the presence of continuing bleeding should not be allowed to delay endostasis. The use of a tilting trolley allows repositioning of the patient, which can facilitate visualisation of the proximal stomach when obscured by blood and clot. Initially, placing the patient in an upright position may suffice, and if necessary rolling the patient into a right lateral and upright position may be needed for complete visualisation of the gastric fundus. In general, lavage is more successful in achieving visualisation than endoscopic aspiration, as endoscope working channels rapidly block with clot. Lavage can be achieved using repeated flushes of saline down the endoscope working channel or with the use of the powered endoscopic lavage catheters such as that provided with the heater probe. With experience, it should rarely be necessary to proceed to surgery or angiography because of inability to visualise the bleeding site due to blood and clot in the stomach or duodenum.
Bleeding gastric ulcers are most likely within the antrum or at the incisura (77%), or less commonly higher on the lesser curve (15%), with ulcers at other sites within the stomach being uncommon. Ulcers at the incisura and proximal lesser curvature can be readily overlooked unless the endoscope is retroflexed within the stomach. The most common site for a bleeding duodenal ulcer is the posterior wall, sometimes with involvement of the inferior and superior walls of the first part of duodenum. Superficial anterior duodenal wall ulcers can ooze, but usually these ulcers perforate. Ulcers elsewhere in the duodenum are seen in less than 10% of patients.19 The presence of active bleeding at the time of endoscopy and the size of the ulcer, rather than its anatomical site, are the main endoscopic determinants of the risk of therapeutic failure.
Management of bleeding due to causes other than peptic ulceration
Bleeding due to gastritis or duodenitis may be associated with non-steroidal anti-inflammatory drug (NSAID) therapy or ingestion of alcohol. It may also be due to Helicobacter pylori and can be severe enough to cause superficial erosions. Such bleeding, however, is almost always self-limiting in the absence of bleeding disorders and therapeutic intervention is not required at the time of endoscopy. Treatment with appropriate acid suppression therapy and early discharge is usually appropriate in the absence of other comorbid illness.
Mallory–Weiss syndrome was first described in 1929 and refers to haematemesis following repeated or violent vomiting or retching. It is caused by a linear tear of the mucosa close to the oesophagogastric junction. It accounts for approximately 5% of patients with UGI haemorrhage and most will settle without the need for therapeutic intervention. However, if bleeding is seen at the time of endoscopy, several approaches have been described. The simplest and most readily available technique is the injection of 1:10 000 adrenaline, as for bleeding peptic ulcers, which is sufficient in the great majority of patients.20 Mechanical methods of endostasis such as endoscopic band ligation or clip application have not been shown to be superior to adrenaline injection alone but are appropriate alternatives, particularly when major bleeding or shock has occurred or where adrenaline injection fails to achieve endostasis.20,21
Gastro-oesophageal reflux disease is responsible for approximately 10% of cases of UGI haemorrhage and is rarely severe. Treatment is with oral PPI therapy.
Major bleeding is occasionally associated with oesophageal, gastric or duodenal tumours. Gastrointestinal stromal tumours (GISTs) may present with bleeding, which can be severe in very occasional cases. Malignancies of the UGI tract commonly cause occult, chronic bleeding but major bleeding can occur and may be difficult to control endoscopically. Management will be dependent on the specific circumstances and may include endoscopic techniques such as argon plasma coagulation, angiographic embolisation or, as a last resort, surgical resection. Where possible, however, if a malignancy is suspected, non-operative methods of achieving haemostasis should be employed, allowing full staging investigations to be organised to guide appropriate management.
This rare cause of UGI bleeding is due to spontaneous rupture of a submucosal artery, usually in the stomach and often within 6 cm of the cardia. The characteristic endoscopic appearance is of a protruding vessel with no evidence of surrounding ulceration. They are commonly missed due to their small size and relatively inaccessible position. Endoscopic clip application or band ligation offers durable and definitive treatment when the lesion is identified. In a small randomised trial,22 haemoclip application was associated with a lower rate of re-bleeding than adrenaline injection, although both achieved similar rates of initial haemostasis.
Endoscopic management of bleeding peptic ulcers
Endoscopy has a central role in the management of non-variceal UGI bleeding. It enables an early diagnosis and allows for risk stratification. Endoscopic signs or stigmata of bleeding are of prognostic value and, in patients with actively bleeding ulcers or stigmata associated with a high risk of recurrent bleeding, endoscopic therapy stops ongoing bleeding and reduces re-bleeding.23 When compared to placebo in pooled analyses, endoscopic therapy has been shown not only to reduce recurrent bleeding, but also the need for surgery and mortality.24,25
Endoscopic Stigmata Of Bleeding
Forrest et al.26 categorised endoscopic findings of bleeding peptic ulcers into those with active bleeding, stigmata of bleeding and a clean base. A modified nomenclature has been in common use in the endoscopy literature. Laine and Peterson27 summarised published endoscopic series of ulcer appearances in which endoscopic therapy was not used and provided crude figures in both the prevalence and rate of recurrent bleeding associated with these stigmata of bleeding. In ulcers that are actively bleeding (Fig. 7.1) or exhibit a non-bleeding visible vessel (NBVV; Fig. 7.2), endoscopic treatment should be offered. There has, however, been observer variation in the interpretation of endoscopic signs28 and the National Institutes of Health Consensus Conference29 defined an NBVV as ‘protuberant discoloration’ (Fig. 7.2). The endoscopist should search the ulcer base diligently in patients judged to have bled significantly or when there is circumstantial evidence of ongoing or recent bleeding, e.g. presence of fresh blood or coffee-ground materials in the gastroduodenal tract. There has been until recently a controversy whether to wash away adherent clot overlying an ulcer (Fig. 7.3). Endoscopists vary in their vigour in clot irrigation before declaring a clot adherent. There have been several randomised studies and a pooled analysis has demonstrated that recurrent bleeding is reduced following clot elevation and treatment to the underlying vessel when compared to medical therapy alone.30 Techniques for clot elevation include target irrigation using a heat probe and cheese-wiring using a snare with or without pre-injection with diluted adrenaline. One should, however, be cautious in elevation of clots overlying large deep bulbar and lesser curve ulcers as some of these may have eroded into larger arteries. A recourse to angiographic embolisation without clot elevation and possible provocation of bleeding may be a better option in sucvh cases. Ulcers with flat pigmentation and clean base (Fig. 7.4) are associated with minimal risk of recurrent bleeding. Stable patients with such ulcers can be discharged home early on medical treatment (Table 7.3).
Prevalence and outcomes of ulcers based on endoscopic appearance
Data from Laine L, Peterson WL. Bleeding peptic ulcer. N Engl J Med 1994; 331 (11):717–27.
FIGURE 7.1 Bleeding vessel in base of ulcer.
FIGURE 7.2 Visible vessel.
FIGURE 7.3 Adherent clot.
FIGURE 7.4 Ulcer with clean base.
Endoscopic therapy should be applied where there is active bleeding or a non-bleeding visible vessel in the ulcer base.23 Adherent clot should be removed and endoscopic therapy applied to the underlying vessel.30
Modalities of endoscopic treatment can be broadly categorised into: injection, thermocoagulation, haemoclipping and, recently, the use of haemostatic nano-powder.
Injection therapy has been widely used because of its simplicity. Injection therapy works principally by volume tamponade. Aliquots (0.5–1 mL) are injected near the bleeding point at four quadrants using a 21- or 23-gauge injection needle. Adrenaline 1:10 000 has an added local vasoconstrictive effect. There is no role for the use of sclerosants as there have been fatal case reports of gastric necrosis following its injection and the added injection of a sclerosant such as polidocanol or sodium tetradecylsulphate after pre-injection of diluted adrenaline does not confer any advantage over injection of diluted adrenaline alone. Injection of fibrin or thrombin has been shown to be effective in some studies but repeated injections are required. These products are costly.31,32
In a canine mesenteric artery model, contact thermocoagulation is superior to injection therapy and non-contact coagulation such as laser photocoagulation in securing haemostasis. Contact thermocoagulation using a 3.2-mm probe consistently seals arteries up to 2 mm in diameter in ex vivo models. Johnston and colleagues emphasised the need for firm mechanical tamponade before sealing of the artery with thermal energy, introducing the term ‘coaptive thermocoagulation’. Mechanical compression alone stops bleeding, reduces heat-sink effect and dissipation of thermal energy. The footprint after treatment provides a clear end-point to therapy. Non-contact thermocoagulation in the form of laser photocoagulation is no longer used as a laser unit is bulky and difficult to be transported. At least in animal experiments, non-contact coagulation in the form of laser photocoagulation is not as effective as contact thermocoagulation. There has also been interest in the use of argon plasma thermocoagulation, with two randomised trials comparing this to injection sclerotherapy or heat-probe treatment, respectively.33 Neither demonstrated any significant difference in treatment outcome.
Haemoclips are commonly used. Their application may be difficult in awkwardly placed ulcers such as those on the lesser curvature of the stomach and the posterior bulbar duodenum. In a meta-analysis of 15 studies with 390 patients34 that compared haemoclipping versus injection and thermocoagulation, successful application of haemoclips (81.5%) was superior to injection alone (75.4%) but comparable to thermocoagulation (81.2%) in producing definitive haemostasis. In this pooled analysis, haemoclipping also led to a reduced need for surgery when compared to injection alone.
Recently the endoscopic use of a haemostatic nano-powder was reported in a small series of 20 patients with actively bleeding peptic ulcers.35 The powder was approved in the United States for external use in traumatic injuries. During endoscopy, the tip of a catheter is placed 1–2 cm from the ulcer. With the push of a button, the powder is then sprayed onto the bleeding ulcer with a pressurised canister with carbon dioxide. It was successful in the control of bleeding in 19 of 20 patients. Comparative studies are required to determine the efficacy of this haemostatic powder. The simplicity of its application certainly appeals to endoscopists.
Single versus combined methods
Soehendra introduced the concept of combination treatment that involved pre-injection with adrenaline allowing a clear view of the vessel, which then allowed targeted therapy using a second modality to induce thrombosis. In a meta-analysis of 16 studies and 1673 patients,36 adding a second modality after adrenaline injection further reduces bleeding from 18.4% to 10.6% (odds ratio (OR) 0.53, 95% CI 0.4–0.69), emergency surgery from 11.3% to 7.6% (OR 0.64, 95% CI 0.46–0.90) and mortality from 5.1% to 2.6% (OR 0.51, 95% CI 0.31–0.84). In an independent meta-analysis of 22 studies (2472 patients),37 dual therapy was shown to be superior to injection alone. However, treatment outcomes following combination treatments were not better than either mechanical or thermal therapy alone. Based on the above pooled analyses, adrenaline alone should no longer be considered an adequate treatment for bleeding peptic ulcers. The current evidence suggests that after initial adrenaline injection to stop bleeding, the vessel should either be clipped or thermocoagulated. In ulcers with a clear view to the vessel, direct clipping or thermocoagulation should yield similar results.
Endoscopic therapy for bleeding peptic ulcers should use dual therapy or mechanical therapy rather than adrenaline injection alone in order to reduce the risk of re-bleeding.36,37
Limit Of Endoscopic Therapy
As mentioned previously, the size of the bleeding artery is a critical determinant in the success of endoscopic treatment. In an ex vivo model, a vessel size of 2 mm could be consistently sealed by a 3.2-mm contact thermal device. In clinical studies that examined factors that might predict failure of endoscopic treatment, ulcer size greater than 2 cm, ulcers on the lesser curvature and ulcers on the superior or posterior wall of bulbar duodenum were consistently identified as major risk factors for recurrent bleeding.23 These ulcers erode into large artery complexes such as the left gastric and the gastroduodenal artery, which are usually sizeable. Consideration should therefore be given to prophylactic measures against recurrent bleeding in these ulcers judged endoscopically to be at significant risk of re-bleeding.
Many endoscopists re-scope their patients the next morning and re-treat ulcers with remaining stigmata of bleeding. In a pooled analysis on the role of second-look endoscopy,38 the authors found a modest 6.2% reduction in the absolute risk of re-bleeding (number needed to treat (NNT) to reduce one episode of recurrent bleeding). The NNT for reduction of surgery and surgery was 58 and 97, respectively. A subsequent meta-analysis and a third meta-analysis carried out for an international consensus conference confirmed that routine second-look endoscopy did reduce the incidence of re-bleeding. The findings were strongest in studies including a high proportion of high-risk ulcers. However, in many of these trials, adrenaline injection alone was used. The role of second-look endoscopy following dual therapy or mechanical therapy remains unclear. In addition, adjuvant treatment with PPI therapy following endoscopic haemostasis can be expected to reduce the benefit of second-look endoscopy even further. With aggressive first endoscopic treatment, the risk of complications, especially perforation, with second treatment is substantial. There may, however, be a role for second-look endoscopy in selected high-risk patients, although this would require further studies and the most recent international consensus guidelines did not recommend this approach on the basis of the available evidence.23
Second-look endoscopy is not indicated as a routine if primary optimum endoscopic haemostasis has been performed.23
Pharmacological management of bleeding peptic ulcers
It has been shown in an in vitro study that platelet aggregation is dependent on plasma pH. It is thought that a pH of 6 is critical for clot stability and an intragastric pH above 4 inactivates stomach pepsin, preventing the digestion of clots.13 To raise intragastric pH consistently above 6, a high-dose PPI given intravenously is required. The antisecretory effect of histamine receptor antagonists, due to tolerance, is less reliable than PPIs. In a study from the Hong Kong group,39 a 3-day course of high-dose omeprazole infusion given after endoscopic therapy to bleeding ulcers reduced the rate of recurrent bleeding from 22.5% to 6.7% at day 30. The majority of recurrent bleeding occurred within the first 3 days of endoscopic treatment. This trial demonstrated the importance of early endoscopic triage, selecting only the high-risk ulcers for aggressive endoscopic treatment followed by profound acid suppression. In a Cochrane systematic review of 24 controlled trials and 4373 patients,40 PPI treatment was shown to reduce re-bleeding (pooled rate of 10.6% vs. 17.3%, OR 0.49, 95% CI 0.37–0.65) as well as surgery (pooled rate of 6.1% vs. 9.3%, OR 0.61, 95% CI 0.48–0.78) when compared with placebo or histamine-2 receptor antagonist. There was no evidence of an effect on all-cause mortality, although when the analysis was confined to patients with high-risk stigmata (active bleeding or visible vessels) there was an associated reduction in mortality with PPI therapy. A multicentre study randomised 767 patients from 91 hospitals in 16 countries to intravenous esomeprazole or placebo following successful endoscopic haemostasis.41 Esomeprazole was associated with significant reductions in re-bleeding and endoscopic re-intervention rates and non-significant reductions in mortality and the need for surgery.
High-dose intravenous PPI therapy (80 mg omeprazole followed by 8 mg/h for 72 h) is recommended for patients with active bleeding or visible vessels at the time of endoscopy.40
Surgical management of bleeding peptic ulcers
The first UK audit revealed a mortality of 24% in those patients (251 of 2071, 12%) who required surgery for bleeding peptic ulcers.1 However, in 78% of these patients, no previous attempt at endoscopic haemostasis had been made. In the most recent UK audit,3 surgery was required in only 1.9% of patients but mortality remained high in this group (30%). The high mortality is probably related to an aged population, with the mean age at 68, and the high incidence of comorbidity. In some patients with severe comorbid illnesses and bleeding peptic ulcer, gastrointestinal haemorrhage is an agonal event. Bleeding ulcers that fail endostasis are typically ‘difficult’ ulcers – larger chronic ulcers that erode into major arterial complexes. The decline in elective ulcer surgery also means the atrophy of surgical techniques in dealing with these ulcers. Ideally, a specialist team with an experienced upper gastrointestinal surgeon should be involved in managing these patients.
Although emergency ulcer surgery has diminished significantly, it has an important gatekeeping role in the management algorithm. The clear indication for surgery is loosely defined as ‘failure of endoscopic treatment’. In a patient with massive bleeding that cannot be controlled by endoscopy, immediate surgery should obviously follow. Similarly, in a patient with bleeding controlled at endoscopy, most clinicians would adopt a non-operative approach. However, the difficulty lies in deciding the exact role of surgery in ulcers judged to have a high risk of recurrent bleeding (e.g. > 2 cm and at difficult locations), in whom endoscopic haemostasis has been initially successful. Increasingly, angiographic embolisation is replacing emergency surgery in these circumstances.
Choice Of Surgical Procedure For Bleeding Peptic Ulcers
The choice of surgical procedure for bleeding peptic ulcers, when required, has not been adequately examined in the era following routine eradication of Helicobacter pylori and high-dose PPI therapy. Many surgeons maintain that under-running of ulcers alone combined with acid suppression using high-dose PPI therapy is safer than definitive surgery by either gastric resection or vagotomy. Two randomised studies looking at the different surgical procedures used to control bleeding peptic ulcers have been reported,42,43 but both predate the PPI and routine H. pylori eradication era and therefore their results must be interpreted with considerable caution. One of these was a multicentre study comparing minimal surgery (under-running the vessel or ulcer excision alone plus intravenous histamine receptor antagonist) versus definitive ulcer surgery (vagotomy and pyloroplasty or partial gastrectomy) in patients with gastric and duodenal ulcers.42 The trial was terminated, however, because of the high rate of fatal re-bleeding in the minimal surgery group (6 of 62 vs. 0 of 67, P = 0.02).
The other trial was carried out by the French Association of Surgical Research and included only bleeding duodenal ulcers.43 The patients in this trial were randomly assigned to either under-running plus vagotomy and drainage (58 patients) or partial gastrectomy (60 patients). Recurrent bleeding occurred in 10 of 58 patients (17%) after under-running and vagotomy. In the group assigned to partial gastrectomy, only two patients (3%) re-bled and both recovered without the need for further surgery. The rate of duodenal stump leak in the gastrectomy group was 8 in 60 (13%). When the results were analysed on an intention-to-treat basis, and those with duodenal leaks after re-operations for re-bleeding in the under-running and vagotomy group were included, duodenal leak rate was similar in both groups (7/58 vs. 8/60). The mortality in both groups was similar (22% after vagotomy and 23% after gastrectomy). In the era of PPI therapy, the role of vagotomy has disappeared. A proper ligation of the gastroduodenal artery complex including the right gastroepiploic and the transverse pancreatic branches is the key to avoid recurrent bleeding.
In a survey of UK surgeons reported in 2003, more than 80% of respondents rarely or never perform vagotomy for bleeding peptic ulcer.44 Despite the absence of recent randomised evidence, surgeons have clearly adopted a more conservative approach based on the efficiency of PPI treatment and H. pylori eradication in the healing of peptic ulceration. With improvements in endoscopic therapy and the increasing age and comorbidity of patients, the risks of definitive ulcer surgery may outweigh any potential benefit from reduction in re-bleeding. For duodenal ulcer haemorrhage, longitudinal duodenotomy is carried out and control of bleeding achieved by digital pressure or by grasping the posterior duodenal wall in tissue forceps. If possible, preservation of the pylorus is preferred, but extension of the duodenotomy to include the pylorus may be required if access is difficult. Control of bleeding may be aided by mobilisation of the duodenum (Kocher's manoeuvre), allowing pressure to be applied posteriorly. In the majority of patients, simple under-running of the bleeding vessel can be achieved using 0 or 1/0 absorbable sutures above and below the bleeding point, ensuring deep enough tissue penetration to completely occlude the vessel. Due to the variation in anatomy of the gastroduodenal artery (Fig. 7.5), four or five sutures should be placed to ensure enduring haemostasis. The duodenotomy can then be closed longitudinally or converted into a formal pyloroplasty if the pylorus has been divided.
FIGURE 7.5 The anatomy of the gastroduodenal (GD) artery complex with confluence of several branches into the artery itself. RGE, right gastroepiploic artery; SPD, superior pancreatico-duodenal artery; TP, transverse pancreatic.Reproduced from Berne CJ, Rosoff L. Peptic ulcer perforation of the gastroduodenal artery complex: clinical features and operative control. Ann Surg 1969; 169:141–4. With permission from Lippincott, Williams & Wilkins.
In cases of a massive duodenal ulcer, it may be necessary to exclude the ulcer, perform a distal gastrectomy and close the distal duodenum. This can be a challenging procedure in an elderly, unstable patient, particularly where duodenal thickening and scarring prevent safe stump closure. In this situation it may be better to anticipate a controlled duodenal fistula by closing the duodenal stump around a Foley catheter rather than attempting more complex closures, such as with the Billroth I reconstruction.
In the case of surgery for a bleeding gastric ulcer, the common scenario is for the ulcer to be located high on the lesser curve of stomach. Anterior gastrotomy, identification of the bleeding site and simple under-running of the ulcer (with biopsy of the ulcer edge) is the procedure of choice, and is also suitable for rare cases of Mallory–Weiss tear or a Dieulafoy lesion that does not respond to endoscopic management. In the rare case of a distal gastric ulcer that does not respond to endoscopic therapy, there may occasionally be a case for ulcer excision or even distal gastrectomy, but it is difficult to justify such a course of action in the hands of a non-specialist surgeon, and simple under-running should be the aim in the majority of patients.
The choice of operation in patients with bleeding peptic ulcers who have failed endoscopic treatment should involve, where possible, simple under-running of the bleeding ulcer, without either vagotomy or gastric resection. Biopsies should be taken from the edge of a gastric ulcer.
Management of recurrent bleeding
The decision on management of patients who re-bleed after initial endoscopic control can be difficult. In a randomised study that compared endoscopic re-treatment to surgery in such patients,45 endoscopic re-treatment secured bleeding again in 75% of patients. With intention-to-treat analysis, complications following endoscopic re-treatment were significantly less when compared to those who received surgery. The gastrectomy rate in the surgery group was 50%. In a subgroup analysis, those re-bleeding with hypotensive shock from ulcers greater than 2 cm were less likely to respond to a repeat endoscopic treatment. It is therefore suggested that a selective approach can be used in re-bleeding patients. Patients with smaller ulcers and subtle signs of re-bleeding should be re-endoscoped and therapy repeated, often with successful outcome. If not successful, surgery should obviously follow. It remains probable that patients with large chronic ulcers and in shock are better treated by expeditious surgery without recourse to endoscopic re-treatment. Some of these patients may benefit from early ‘elective/pre-emptive’ surgery or (increasingly) angiographic embolisation.
Management of re-bleeding following successful endostasis will depend on the specific circumstances. Further endoscopic haemostasis may be appropriate for many patients,46 but high-risk ulcers, particularly those where good endostasis was difficult to achieve at the first procedure, may be better considered for surgery or even transarterial angiographic embolisation.
The Role Of Selective Mesenteric Embolisation
Transarterial angiographic embolisation (TAE) is an alternative rescue procedure for bleeding duodenal ulcers and the technique has been in use for over two decades. In the 1980s, there were reports of visceral infarcts46,47 following TAE, and its use was restricted to a small group of patients with refractory bleeding considered unfit for surgical intervention. With advances in embolisation techniques and specifically the use of superselective coiling (Fig. 7.6), the success rate in the control of bleeding has been reported to be between 64% and 91%, and mortality between 5% and 25%. There have since been two retrospective comparative studies comparing angio-embolisation with surgery,48,49 involving a total of 179 patients of whom 72 had TAE and 107 underwent surgery. In the Hong Kong series,48 re-bleeding was higher after TAE compared with surgery but complications were higher in the surgical group and overall mortality was similar at 25%for TAE and 30% for surgery. The Swedish study49 had a lower mortality in both groups (3% and 14% for TAE and surgery, respectively), perhaps reflecting different selection criteria, but again there were no significant differences between the TAE and surgical groups. In the latter comparative study, the lack of outcome difference and the more advanced age in the TAE group suggest that TAE may be at least as good an option as surgery in the management of refractory ulcer bleeding. TAE is certainly now the procedure of choice in the small group of patients who re-bleed after surgery, but a randomised controlled study comparing surgery and TAE in this group of patients would be of great interest. Similarly, the role of semi-elective TAE following successful endostasis in patients considered at high risk of re-bleeding or death has not been studied, but is another area where further research may be interesting.
FIGURE 7.6 Active bleeding from the gastroduodenal artery complex during transfemoral angiography. Coils were used to embolise the artery leading to cessation of bleeding.
Transarterial embolisation should be considered in patients who re-bleed following surgery for bleeding peptic ulcers and as an alternative to surgery when endoscopic haemostasis has failed, provided appropriate facilities and expertise are available. This may be particularly useful in elderly patients with medical comorbidity. TAE should also be considered as a possible pre-emptive treatment option in high-risk surgical patients who are at high risk of re-bleeding after endostasis.
Helicobacter pylori eradication
A Cochrane review50 concluded that H. pylori eradication was associated with a significant reduction in the risk of re-bleeding compared with no H. pylori eradication, from 20% to 2.9%. If antisecretory therapy was continued, the risk was 5.6%, still significantly higher than achieved with H. pylori eradication. The overall risk of re-bleeding following H. pylori eradication was less than 1% per year. It is therefore of concern that in a UK review of consultant behaviour, fewer than 60% routinely tested patients for H. pylori following treatment for complicated peptic ulcers.44
Following treatment for bleeding duodenal ulcer, patients should be tested for H. pylori and receive eradication therapy where appropriate. Patients should have further testing to ensure successful eradication.50
Use of NSAIDs
In patients who continue to require NSAIDs, co-therapy with PPI reduces recurrence in peptic ulcers and bleeding. In these patients, H. pylori should first be tested and treated if confirmed. A randomized controlled trial compared the use of a traditional NSAID plus a PPI to COX-2 inhibitors and found that the rate of further ulcer complications is between 4-6% in 6 months.51 A subsequent randomized trial combined the use of COX-2 inhibitor to PPI and compared them to the use of a COX-2 inhibitor alone.52 At 1 year, the use of COX-2 inhibitor alone was associated with a rate of 8.9% in recurrent bleeding. The risk of recurrent bleeding was completely abolished in those who the combined treatment. COX-2 inhibitor plus PPI appears to offer the best protection to these high risk patients.
The challenge posed by peptic ulcer bleeding has altered with the increasing age of the population at risk and the increasing availability of skilled therapeutic endoscopy. Failure of endoscopic haemostasis is increasingly uncommon but the surgical challenge presented by the elderly patient with refractory bleeding from a large ulcer is considerable. Successful management of UGI bleeding will involve the close cooperation of a multidisciplinary team, which will increasingly include interventional radiologists, aided by local protocols based on evidence-based best practice (Fig. 7.7).
FIGURE 7.7 Management algorithm. SBP, systolic blood pressure.