Management of acute upper GI bleeding

BJA Education, 17 (4): 117 123 (2017) Matrix reference 1A01, 2A05, 3C00 doi: 10.1093/bjaed/mkw054 Advance Access Publication Date: 10 December 2016 Management of acute upper GI bleeding Ingi Adel Salah Elsayed MRCP FFICM 1, Pavan Kumar Battu FRCA 2, and Sarah Irving MRCP FRCA FFICM 3, * 1 ST7 Trainee, Intensive Care and Renal Medicine, Sheffield Teaching Hospitals NHS Foundation Trust, Sheffield S5 7AU, UK, 2 ST7 Trainee, Anaesthesia, Sheffield Teaching Hospitals NHS Foundation Trust, Sheffield S5 7AU, UK, and 3 Consultant in Critical Care Medicine and Anaesthesia, Sheffield Teaching Hospitals NHS Foundation Trust, Northern General Hospital, Herries Road, Sheffield S5 7AU, UK *To whom correspondence should be addressed. Tel: 0114 2434343; Fax: 0114 2269342; E-mail: Sarah.Irving@sth.nhs.uk. Key points Acute upper gastrointestinal bleeding has a high mortality rate. Initial priorities are resuscitation, risk assessment, and early endoscopic treatment as described in National Institute for Health and Care Excellence Guidance 141. Non-variceal re-bleeds require interventional radiological embolization, or surgery. Repeat oesophagogastroduodenoscopy may allow decisions to be made at this point. Variceal bleeds not responding to endoscopic therapy need management in a specialist centre, benefit from balloon tamponade if unstable and should be considered for transjugular intrahepatic portosystemic shunt. Upper gastrointestinal bleeding due to stress ulceration in the ICU will occur in up to 15% of patients not on prophylactic treatment. Current National Institute for Health and Care Excellence guidance recommends either H2-receptor antagonists or proton pump inhibitors for primary prophylaxis. The reported incidence of acute upper gastrointestinal bleeding (UGIB) in the United Kingdom varies over the range 84 172/ 100 000 year 1. 1 Mortality due to upper GI bleeding was found to be 7% among new admissions, rising to 30% in those who bled as inpatients. 2 UGIB refers to bleeding from any point proximal to the duodenojejunal flexure. It presents clinically with haematemesis and/or melaena. Fresh bleeding per rectum is usually indicative of lower GI bleeding, however massive UGIB can present with passing of red blood clots rectally. In 90% of cases, melaena results from any upper GI bleeding. Bleeding from parts of small bowel distal to the ligament of Treitz and the right hemicolon may also lead to melaena. Causes of UGIB can be classified in two ways. The first is based on the underlying pathophysiological mechanism (see Table 1). The second is based on the type of bleed i.e. variceal or non-variceal. In developing countries with endemic viral hepatitis, variceal bleeding is the commonest category. In Western communities, peptic ulcer disease (PUD) is the commonest cause. IntheUK,thecommonestcausesofUGIBarePUD,oesophagogastric varices with or without portal hypertensive gastropathy and oesophagitis. Management of acute UGIB The National Institute for Health and Care Excellence (NICE) Clinical Guideline 141 (CG 141), published in June 2012 in the UK, outlines the management of acute upper GI bleeding. NICE Quality Standards Document 38 was then published in July 2013. 3 A subsequent Evidence Update 63 was published in VC The Author 2016. Published by Oxford University Press on behalf of the British Journal of Anaesthesia. All rights reserved. For Permissions, please email: journals.permissions@oup.com 117

Table 1 Classification of UGIB according to underlying pathophysiological mechanisms Ulcerative Portal hypertension Tumours Gastric Gastric varices Benign Idiopathic Oesophageal varices Polyps Drug induced e.g. NSAIDs Hypertensive gastropathy Lieomyoma Infective: H. pylori Vascular malformations e.g. Malignant Stress ulcers Dieulafoy s lesion Adenocarcinoma Zollinger Ellison Hereditary hemorragic telangiectasia Carcinoid (Osler Weber Rendu syndrome) Oesophageal Traumatic/surgical e.g. Lymphoma Infective e.g. candidal, CMV Mallory Weiss syndrome Metastatic carcinoma Drug induced e.g. alendronate, potassium chloride Aortoenteric fistula Post anastomosis or polypectomy August 2014. 4 Much of this review is based on the evidence presented in these documents. Initial assessment Mortality associated with UGIB is high, so assessment of severity, general supportive measures and identification of the cause, should occur in parallel with direct, specific therapies. Taking an adequate history is an essential part of managing UGIB. As discussed, UGIB usually presents with haematemesis and/or melaena. Associated symptoms at presentation such as postural dizziness or collapse may point towards severity of bleeding. It is imperative to clarify if the patient has previously presented with UGIB. Sixty per cent of patients who re-bleed, do so from the same source. A history of recent GI surgery might suggest a very different origin for bleeding and necessitate early surgical and potentially radiological input. Drug history is equally important. For example, nonsteroidal anti-inflammatory drugs (NSAIDS) may be identified as a precipitant for bleeding. Anticoagulant therapies such as warfarin and new oral anticoagulants (NOACs; direct inhibitors of thrombin or activated factor Xa) increase the likelihood of severe bleeding and require specific treatment. Co-morbid illness may point to a cause for bleeding, such as chronic liver disease and variceal bleeding. It also identifies patients at higher risk of developing complications, so requiring tailored treatment. Examples are those with ischaemic heart disease who require a lower threshold for blood transfusion, or patients with heart or renal failure who may need invasive monitoring of resuscitative fluid therapy. Careful physical examination can further establish a potential cause for bleeding and determine severity, so aiding appropriate triage of patients. Signs of chronic liver disease such as jaundice and spider naevi point to variceal bleeding. Significant upper abdominal tenderness and rigidity may suggest a complication such as perforation. Severity of bleeding determines the changes in vital signs, with some caveats. Initial signs of mild blood loss would include resting tachycardia. Fifteen per cent loss of blood volume is associated with postural hypotension indicated by a drop in systolic blood pressure greater than 20 mm Hg or with an increase in heart rate greater than 20 beats min 1 on changing posture from recumbent to standing. A loss of 40% blood volume leads to recumbent hypotension. However, betablockade masks the development of reflex tachycardia expected with blood loss. Potent vasodilating antihypertensive drugs will exaggerate hypotension. Initial laboratory tests should include full blood count, liver enzymes, bilirubin, serum albumin, urea and electrolytes, coagulation screen, and samples for cross matching of blood. Other specific tests may be required as appropriate, such as screening for viral hepatitis. General supportive measures Patients may require supplemental oxygen to increase oxygen delivery, particularly with co-morbid lung or heart disease. Early elective tracheal intubation should be considered in those with ongoing haematemesis and altered mental or respiratory status to minimize the risk of aspiration, whilst aiding endoscopy. Significant co-morbidities lead to a much higher risk of complications. For those patients and others with signs of refractory shock (e.g. lactate > 4 mmol litre 1, ph< 7.3 after resuscitation), admission to a critical care area should be considered early. Invasive arterial monitoring allows closer monitoring of haemodynamic status and facilitates frequent blood gas and laboratory samples to be taken. Central venous line insertion, provided coagulation is not deranged, gives secure i.v. access and allows central venous pressure measurements to be taken. Flow guided methods of monitoring such as LidCO or oesophageal pressure monitoring give a more accurate guide to volume status. In all patients, two wide-bore cannulas (16G at least) should be established. Patients should not be given anything by mouth. Fluid resuscitation starts with rapid infusion of isotonic crystalloid solution, such as Hartmann s. In patients non-responsive to crystalloids, resuscitation is continued using blood and fresh frozen plasma (FFPs). In many centres, a major haemorrhage protocol can be activated. In extreme circumstances, O negative blood may be required initially. The authors do not advocate the use of colloids in resuscitation, given concerns over kidney injury, worsened coagulopathy, risk of anaphylaxis, and a lack of benefit over crystalloids. NICE CG 141 recommends individualizing decisions for blood transfusion according to the full clinical picture, whilst recognizing the risks of overtransfusion. Blood transfusion should usually be targeted at patients with haemoglobin (Hb) level below 70 g litre 3. 4 Evidence comes from a randomized controlled trial which compared mortality rates at 45 days between patients randomised to either a restrictive transfusion strategy (if Hb <70 g litre 1 ) or liberal transfusion strategy (if Hb <90 g litre 1 ). The trial reported significantly less mortality in the restrictive group (P ¼ 0.02) and fewer adverse events. The TRIGGER (Transfusion in Gastrointestinal Bleeding) trial 5 randomly compared clinical outcomes including 28-day mortality 118 BJA Education Volume 17, Number 4, 2017

between restrictive transfusion (Hb <80 g litre 1 ) and liberal transfusion (Hb <100 g litre 1 ). Neither study included patients with massive ongoing bleeding. To summarize, restrictive strategies are generally preferred with the exception of patients with ischaemic heart disease and possibly those with ongoing haemorrhage. The NICE CG 141 also recommends giving FFP to bleeding patients with a prothrombin time (or INR) or activated partial thromboplastin time greater than 1.5 times normal, which are generally accepted times in other scenarios with major haemorrhage. Cryoprecipitate should be given for fibrinogen levels below 1.5 g litre 1 despite FFP administration. Platelet transfusion should be given to patients with platelet count <50 10 9 litre 1 and considered in those actively bleeding while established on anti-platelet therapy such as clopidogrel. Patients actively bleeding while taking warfarin therapy should receive Prothrombin complex concentrate (PCC). In bleeding, the activation of a major haemorrhage protocol and early involvement of senior clinicians, including haematologists is vital. This is particularly important for those taking NOAC drugs, the effects of which are not easily reversed. The European Heart Rhythm Association produced practical guidance on their use. 6 Recommendations include the use of PCC alongside general resuscitation in life threatening bleeding. Dabigatran is the only NOAC drug which may be fully reversed using an antidote. Idarucizumab has been shown in a pilot trial to achieve haemostasis in serious bleeding from dabigatran. 7 This monoclonal antibody drug is currently pending approval for use in the UK. intermittent injections. It reduces portal blood flow, hence variceal bleeding. It is the only vasoactive drug proven to reduce mortality. 9 It should be given to patients with suspected variceal bleeding during resuscitation measures and prior to endoscopic confirmation. Terlipressin can be stopped once definitive haemostasis is achieved and in any case after 5 days. Other agents have been studied. In a recent RCT, somatostatin and its metabolite octreotide have proven to be as effective as terlipressin. 10 In 2015 the British Society of Gastroenterology (BSG) published guidance on managing variceal haemorrhage. 11 This suggests using terlipressin or somatostatin and if unavailable using octreotide off license. I.V. erythromycin as a prokinetic administered prior to endoscopy vs no erythromycin was evaluated in a systematic review of seven RCTs. 12 Erythromycin was associated with a greater chance of adequate visualization of the gastric mucosa, less need for second endoscopy, less blood transfusion and shorter hospital stay. Due to limitations in the studies (e.g. gastric visualization definitions varied across the studies), an update to NICE CG 141, did not recommend erythromycin use as standard. Tranexamic acid is not currently recommended by NICE for acute UGIB. The BSG document recommends further study to look for benefit in variceal bleeding. Antibiotics with gram negative cover, such as third generation cephalosporins, improve mortality rates and reduce bacterial infections. Cochrane meta-analysis 13 also shows fewer early re-bleeds. Antibiotic choice is dictated by local resistance patterns and availability. Pharmacotherapy Proton pump inhibitors (PPIs) increase gastric ph to above 6.0. This optimizes platelet function, inhibits fibrinolysis and reduces peptic activity, so increasing clot stabilization over bleeding ulcers. PPI use prior to endoscopy in UGIB was evaluated in a Cochrane systematic review of RCTs comparing PPI vs control [placebo, or H2-receptor antagonists (H2RAs) or no treatment]. 8 It found that PPI significantly reduced the need for endoscopic therapy and reduced the stigmata of recent haemorrhage during index endoscopy, but there was no evidence of improved clinically important outcomes, namely mortality, re-bleeding or need for surgery. NICE therefore advises against the use of PPI prior to endoscopy. 3 PPIs should be prescribed postendoscopy for those with non-variceal UGIB and stigmata of recent haemorrhage. Terlipressin is a synthetic analogue of Vasopressin, with a slow, sustained release allowing administration via Risk stratification The Rockall and Blatchford scores are well-validated risk stratification systems used in UGIB. The full Rockall score (RS) incorporates preendoscopic and postendoscopic data 14 to predict rebleeding and mortality. The Glasgow Blatchford score 1 (GBS) uses only clinical and laboratory data (see Tables 2 and 3). Used appropriately, these scores help identify patients likely to require Critical Care support, ensure timely interventions and help identify those suitable for early discharge from hospital. NICE advises the use of GBS at first assessment and full RS postendoscopy. 3 Specific therapies in non-variceal UGIB Oesophagogastroduodenoscopy (OGD) is the cornerstone in managing acute UGIB. NICE CG 141 stipulates that haemodynamically unstable patients with UGIB should be offered OGD Table 2 Rockall score Score 0 1 2 3 Age (yr) <60 60 79 80 N/A Pulse <100 >100 >100 N/A Systolic blood pressure >100 >100 <100 N/A Comorbidities None Nil major Ischaemic heart disease, cardiac failure, other major co-morbidity Renal or liver failure, disseminated malignancy Diagnosis Mallory Weiss or no lesion/pathology All other diagnosis Malignant lesion N/A Stigmata of haemorrhage on endoscopy None, or dark spot only None, or dark spot only Blood, adherent clot, visible/spurting vessel N/A Score of 2 good prognosis. Score >7 10 35% mortality, 25 50% mortality with re-bleed. BJA Education Volume 17, Number 4, 2017 119

Table 3 Glasgow Blatchford score Admission risk marker Score component value Blood urea (mmol litre 1 ) 6.5 8.0 2 8.0 10.0 3 10.0 25 4 >25 6 Haemoglobin (g litre 1 ) for men 120 130 1 100 120 3 <100 6 Haemoglobin (g litre 1 ) for women 100 120 1 <100 6 Systolic blood pressure (mm Hg) 100 109 1 90 99 2 <90 3 Other markers Pulse 100 (min 1 ) 1 Presentation with melaena 1 Presentation with syncope 2 Hepatic disease 2 Cardiac failure 2 Score 0 low risk, consider outpatient management. Score >5 high risk of needing intervention immediately after resuscitation and within 24 h in all other patients. 3,4 OGD is both a diagnostic and therapeutic tool. Endoscopic therapies These are warranted in patients with high risk lesions, meaning active bleeding, visible non-bleeding vessels and adherent clots. They include injection therapies, mechanical therapies, and thermal interventions. Adrenaline (1:10 000 or 1:20 000) is injected in increments of 0.5 1.5 ml aliquots to the four quadrants around high-risk stigmata. It is easy to administer and can produce a clear field around a bleeding vessel. The tissue adhesives, thrombin and fibrin, create a tissue seal at the site of bleeding and also a tamponade effect. Due to the risk of tissue necrosis, a volume limit of 1.0 ml is used. The endoscopic clip is the most commonly used mechanical device. Lesions on the gastric lesser curvature or posterior wall may not be amenable to mechanical therapy. Thermal therapies include electrocautery probes and the argon plasma coagulator (APC). Haemostasis is achieved by direct tamponade of visible vessel and tissue coagulation using delivered heat or electric current. APC does not use direct tissue contact and is used mainly in superficial lesions. NICE CG 141 advocates the use of either a mechanical method with or without adrenaline, thermal coagulation with adrenaline, or fibrin or thrombin with adrenaline. It advises against the use of adrenaline as monotherapy. After endoscopy Acid suppression using a PPI should be used after endoscopic haemostasis. An updated large Cochrane meta-analysis showed that compared with H2 blocker or placebo, PPI reduced re-bleeding, surgical intervention and need for repeated endoscopic treatment. Use of PPI reduced mortality in those with active bleeding or non-bleeding vessels. 13 The optimal dose and route of administration in terms of clinical outcomes, remains controversial. However, high dose IV PPI seems the most costeffective. 19 Current international consensus guidelines recommend high-dose i.v. PPI therapy e.g. Pantoprazole 80 mg bolus followed by 8 mg h 1 for 3 days. 15 Helicobacter pylori eradication should be offered to all who test positive for the infection. Eradication therapy was shown to be superior to PPI alone in preventing re-bleeding. Those with PUD, but negative for H. pylori atthetimeofacuteugib,shouldberetested. 3 NICE suggest offering a repeat endoscopy in the event of rebleeding with a view to further endoscopic treatment or emergency surgery. However, unstable patients who re-bleed require interventional radiology. If there is a delay in such therapy, urgent referral for surgery should be made. Specific therapy in variceal UGIB The BSG guidelines on the management of variceal bleeding complement NICE guidance and add more detail, some of which have been included below. 11 Broad spectrum antibiotic therapy should be given, as dictated by local protocol. Terlipressin should be commenced during resuscitation. Procedures for control of bleeding Endoscopic therapy for oesophageal varices: Definitive therapy should take place at the time of diagnostic endoscopy. The techniques involved are either endoscopic variceal ligation (EVL) or endoscopic sclerotherapy (ES). Initial success rates of 90% are reported. EVL was found to be as effective as ES in achieving hemostasis, but in meta-analysis EVL was found superior in rates of re-bleeding, strictures and mortality. 16 NICE CG 141, recommends the use of EVL for controlling bleeding oesophageal varices. Gastric varices are more difficult to control. Endoscopic injection of N-butyl-2-cyanoacrylate into the lumen of bleeding gastric varices (where it polymerises into firm clot in aqueous medium) is the currently established initial treatment and is recommended by NICE. It appears more effective than EVL and better in preventing re-bleed than beta blockers. Balloon tamponade is effective at achieving haemostasis in acute variceal bleeding, but has high rates of re-bleeding and complications. Balloons should be used temporarily when endoscopic therapy has failed and until either further endoscopic treatment, transjugular intrahepatic portosystemic shunt (TIPSS) or surgery is performed. Definitive treatment will depend on local resources and expertise. The BSG suggest the use of the Sengstaken Blakemore tube but alternatives include a Minnesota tube. Balloon insertion is preferably preceded by tracheal intubation to avoid aspiration. Endoscopic insertion over a guide wire has been suggested to reduce the risk of oesophageal rupture. The gastric balloon is inflated up to a volume of 500 ml with air, with the oesophageal balloon inflated up to 45 mm Hg for ongoing bleeding. Balloons are usually left in situ for 24 48 h, with deflation regularly to assess for re-bleeding. Endoscopically placed oesophageal stents, which remain in place for up to 2 weeks have been used as an alternative in oesophageal bleeding, but no data have yet been published to compare outcomes with balloon tamponade. TIPSS should be considered when variceal bleeding is not controlled via endoscopic therapy alone. This involves radiologically establishing a porto-caval shunt, using the transjugular intrahepatic route. Its success rate at stopping bleeding is quoted as 90 100%, especially in oesophageal varices. With gastric varices, development of spontaneous splenorenal collaterals may cause re-bleeding. With refractory bleeding and no other therapeutic options, salvage TIPSS may be considered but 120 BJA Education Volume 17, Number 4, 2017

Resuscitation Risk Stratification Haemodynamically Unstable Endoscopy within 2 h of optimal resuscitation Diagnosis Haemodynamically Stable Endoscopy within 24 h Non-Variceal Bleeding Variceal Bleeding -Antibiotic treatment -Terlipressin -Combination or mechanical endoscopic method -PPI if active bleed or stigmata of recent bleed Oesophageal Varices Endoscopic Variceal ligation Gastric Varices Endoscopic injection with N-butyl-2- cyanoacrylate Continued Bleed/re-bleed -Consider repeat endoscopy -Interventional radiology or surgery Consider balloon tamponade N O Controlled Bleeding? Y E S Secondary Prophylaxis EVL+NSBB -Referral for TIPSS if portal vein patent -Surgical /radiological options Fig 1 Acute UGIB management summary. may be associated with high risk of hepatic encephalopathy. NICE recommends consideration for TIPSS in any uncontrolled variceal bleed without contraindications such as portal vein thrombosis, severe systemic infection, or severe pulmonary hypertension. Some studies suggest a reduction in mortality when TIPSS is used early rather than salvage therapy. The Child-Pugh classification is used to assess the prognosis of chronic liver disease, mainly cirrhosis and to predict mortality during surgery. BSG guidance is that TIPSS should be considered within 72 h of control of bleeding in Child-Pugh class B cirrhosis or Child-Pugh class C cirrhosis with score below 14. More research is required in this area. Surgical techniques include decompression of the portal circulation (emergency surgical shunt operations) and oesophageal transection. Both can achieve long term haemostasis, but require specialist surgical expertise and have associated mortality rates of 50% and higher in Child-Pugh class C cirrhosis. The BSG document recommends that these are considered in bleeding patients with Child s B cirrhosis, when TIPSS is unavailable, or not possible due to portal vein thrombosis. TIPSS has a well-established role in variceal bleeding, as discussed. Other interventional radiological procedures have been used. Balloon-occluded retrograde transvenous obliteration techniques and percutaneous transhepatic variceal embolization with BJA Education Volume 17, Number 4, 2017 121

cyanoacrylate have been described in gastric varices, but are not common practice. 11 Prevention of further variceal bleeding Recommendations for secondary prophylaxis are long term beta blockade combined with regular EVL at 2 4 weekly intervals until eradication, evaluation for liver transplantation and consideration of elective shunt operations. TIPSS was found to decrease re-bleeding rates, but with no survival advantage and higher morbidity rates. Hence this is currently indicated for secondary prophylaxis only in the context of re-bleeding or when combination therapy is not possible. In Child s A and B cirrhosis, shunt surgery is an alternative option. Resumption of antiplatelet therapy after UGIB Patients who need secondary prevention of vascular events, should restart low-dose aspirin in combination with a PPI, once haemostasis has been secured. All NSAIDs, including COX-2 inhibitors, should be withheld during the acute phase of UGIB and should only be restarted in required cases, with PPI cover and after risk assessment. 3 Evidence for when to start clopidogrel or NOAC drugs is lacking. A multidisciplinary discussion between specialists should weigh the risks and benefits of restarting such medication. Prophylaxis of UGIB due to stress ulceration in the critically ill The incidence of clinically important GI bleeding due to stress ulcers in ICU patients is between 1.5 and 8.5% overall and up to 15% without ulcer prophylaxis. 17 It is associated with increased mortality. Major risk factors are mechanical ventilation and coagulopathy. Other risk factors include severe burns, traumatic brain or spinal cord injury and GI ulceration. NICE CG 141 recommends either H2RAs or PPIs for primary prevention of UGIB. Oral preparations are preferred. A meta-analysis of 20 RCTs showed thatprophylaxisusingeitherh2rasorppisreducedgibleeding, without effect on mortality or rates of ventilator-associated pneumonia (VAP). 18 A further systematic review 19 comparing PPIs with H2RAs, found less UGIB, without effect on mortality, pneumonia or ICU length of stay with a PPI. A study in the US found that PPIs were more cost-effective than H2RAs among ICU patients, in preventing stress ulceration and reducing serious UGIB without increasing VAP. 20 However, NICE Guidance Evidence Update 63 did not change the recommendations in CG 141. Summary Mortality due to acute UGIB remains high. Successful management relies on simultaneous resuscitation and identification of the cause to allow timely, definitive treatment. Involvement of specialist services early is important. Figure 1 provides a quick summary of the management. Primary prophylaxis against stress ulceration on the ICU reduces rates of UGIB, but without overall effect on mortality. Declaration of interest None declared. MCQs The associated MCQs (to support CME/CPD activity) can be accessed at https://access.oxfordjournals.org by subscribers to BJA Education. References 1. Blatchford O, Murray WR, Blatchford M. A risk score to predict need for treatment for upper-gastrointestinal haemorrhage. Lancet 2000; 356: 1318 21 2. Hearnshaw SA, Logan RF, Lowe D, Travis SP, Murphy MF, Palmer KR. Acute upper gastrointestinal bleeding in the UK: patient characteristics, diagnoses and outcomes in the 2007 UK audit. Gut 2011; 60: 1327 35 3. National Institute for Health and Care Excellence. Acute upper gastrointestinal bleeding: management NICE guidelines [CG141], 2012. Available from https://www.nice.org.uk/ guidance/cg141 (accessed 4 May 2015) 4. National Institute for Health and Care Excellence. Acute upper gastrointestinal bleeding, 2014. Available from https:// arms.evidence.nhs.uk/resources/hub/1037665/attachment (accessed 4 May 2015) 5. Jairath V, Kahan BC, Gray A et al. Restrictive versus liberal transfusion for acute upper gastrointestinal bleeding (TRIGGER): a pragmatic, cluster randomised feasibility trial. Lancet 2015; 386: 137 44 6. Heidbuchel H, Verhamme P, Alings M et al. Updated European Heart Rhythm Association Practical Guide on the use of nonvitamin K antagonist anticoagulants in patients with nonvalvular atrial fibrillation. Europace 2015; 17: 1467 507 7. Pollack CV Jr, Reilly PA, Eikelboom J et al. Idarucizumab for dabigatran reversal. N Engl J Med 2015; 373: 511 20 8. Sreedharan A, Martin J, Leontiadis GI et al. Proton pump inhibitor treatment initiated prior to endoscopic diagnosis in upper gastrointestinal bleeding. Cochrane Database Syst Rev 2010; 7: CD005415 9. Ioannou G, Doust J, Rockey DC. Terlipressin for acute esophageal variceal hemorrhage. Cochrane Database Syst Rev 2003; 1: CD002147 10. Seo YS, Park SY, Kim MY et al. Lack of difference among terlipressin, somatostatin, and octreotide in the control of acute gastroesophageal variceal haemorrhage. Hepatology 2014; 60: 954 63 11. Tripathi D, Stanley AJ, Hayes PC et al. UK guidelines on the management of variceal haemorrhage in cirrhotic patients. Gut 2000; 46: III1 15 12. Theivanayagam S, Lim RG, Cobell WJ et al. Administration of erythromycin before endoscopy in upper gastrointestinal bleeding: a meta-analysis of randomised controlled trials. Saudi J Gastroenterol 2013; 19: 205 10 13. Chavez-Tapia NC, Barrientos-Gutierrez T, Tellez-Avila F et al. Meta-analysis: antibiotic prophylaxis for cirrhotic patients with upper gastrointestinal bleeding an updated Cochrane review. Aliment Pharmacol Ther 2011; 34: 509 18 14. Rockall TA, Logan RF, Devlin HB, Northfield TC. Risk assessment after acute upper gastrointestinal haemorrhage. Gut 1996; 38: 316 21 15. Barkun AN, Bardou M, Kuipers EJ et al. International consensus recommendations on the management of patients with nonvariceal upper gastrointestinal bleeding. Ann Intern Med 2010; 152: 101 13 16. Laine L, Cook D. Endoscopic ligation compared with sclerotherapy for treatment of esophageal variceal bleeding. A meta-analysis. Ann Intern Med 1995; 123: 280 17. Cook DJ, Griffith LE, Walter SD et al. Canadian Critical Care Trials Group. The attributable mortality and length of 122 BJA Education Volume 17, Number 4, 2017

intensive care unit stay of clinically important gastrointestinal bleeding in critically ill patients. Crit Care 2001; 5: 368 18. Krag M, Perner A, Wetterslev J et al. Stress ulcer prophylaxis versus placebo in critically ill patients: a systematic review of randomised clinical trials with metaanalysis and trial sequential analysis. Intensive Care Med 2014; 40: 11 22 19. Alhazzani W, Alenezi F, Jaeschke RZ et al. Proton pump inhibitors versus histamine 2 receptor antagonists for stress ulcer prophylaxis in critically ill patients: a systematic review and meta-analysis. Crit Care Med 2013; 41: 693 705 20. Barkun AN, Adam V, Martel M et al. Cost-effectiveness analysis: stress ulcer bleeding prophylaxis with proton pump inhibitors, H2 receptor antagonists. Value Health 2013; 16: 14 22 BJA Education Volume 17, Number 4, 2017 123