Management of nonvariceal upper gastrointestinal bleedingE Wee
Gastroenterology, Department of General Medicine, Khoo Teck Puat Hospital, Singapore
Nonvariceal upper gastrointestinal bleeding is unique from variceal bleeding in terms of patient characteristics, management, rebleeding rates, and prognosis, and should be managed differently. The majority of nonvariceal upper gastrointestinal bleeds will not rebleed once treated successfully. The incidence is 80 to 90% of all upper gastrointestinal bleeds and the mortality is between 5 to 10%. The causes include nonacid-related ulceration from tumors, infections, inflammatory disease, Mallory-Weiss tears, erosions, esophagitis, dieulafoy lesions, angiodysplasias, gastric antral vascular ectasia, and portal hypertensive gastropathy. Rarer causes include hemobilia, hemosuccus pancreaticus, and aortoenteric fistulas. Hematemesis and melena are the key features of bleeding from the upper gastrointestinal tract, but fresh per rectal bleeding may be present in a rapidly bleeding lesion. Resuscitation and stabilization before endoscopy leads to improved outcomes. Fluid resuscitation is essential to avoid hypotension. Though widely practiced, there is currently insufficient evidence to show that routine red cell transfusion is beneficial. Coagulopathy requires correction, but the optimal international normalized ratio has not been determined yet. Risk stratification scores such as the Rockall and Glasgow-Blatchford scores are useful to predict rebleeding, mortality, and to determine the urgency of endoscopy. Evidence suggests that high-dose proton pump inhibitors (PPI) should be given as an infusion before endoscopy. If patients are intolerant of PPIs, histamine-2 receptor antagonists can be given, although their acid suppression is inferior. Endoscopic therapy includes thermal methods such as coaptive coagulation, argon plasma coagulation, and hemostatic clips. Four quadrant epinephrine injections combined with either thermal therapy or clipping reduces mortality. In hypoxic patients, endoscopy masks allow high-flow oxygen during upper gastrointestinal endoscopy. The risk of rebleeding reduces after 72 hours. In rebleeding, repeat endoscopy is useful and persistent failure of endoscopic therapy mandates either embolization or surgery. In this review, we analyze the management of nonvariceal upper gastrointestinal bleeding with evidence from the currently published clinical trials.
Keywords: Endoscopy, GI bleeding, management, proton pump inhibitor, review, upper gastrointestinal bleeding
Upper gastrointestinal bleeding can be divided into variceal bleeding and nonvariceal bleeding. This is because the patient characteristics, endoscopic management, medical management, rebleeding rates, and prognosis are different between both pathologies. Published studies from multiple countries suggest that 80 to 90% of all upper gastrointestinal bleeds are due to nonvariceal lesions. The majority are due to peptic ulcer disease. ,, Other causes of nonvariceal bleeding are due to nonacid-related ulceration (e.g., tumors, viral infections, inflammatory disease), Mallory-Weiss tear More Detailss, erosions, esophagitis, dieulafoy lesions, angiodysplasias, gastric antral vascular ectasia, and portal hypertensive gastropathy. ,,, Even rarer causes are hemobilia, hemosuccus pancreaticus, and aortoenteric fistulas. ,,,
Nonvariceal upper gastrointestinal bleeding carries a mortality of between 5 to 10%. , The majority of patients with an episode of bleeding will not rebleed once they have been treated successfully. Endoscopy is the best modality to evaluate and manage an episode of nonvariceal upper gastrointestinal bleeding as it is both effective and safe. However, as many as 7 to 16% of patients will have persistent bleeding or recurrent episodes of rebleeding during their admission. ,
Upper gastrointestinal bleeding is defined as bleeding proximal to the ligament of Treitz. This may present either as hematemesis or melena.  Occasionally, hemoptysis may be mistaken by the patient as hematemesis and this should be suspected by the astute physician when the upper gastrointestinal endoscopy is normal without any blood seen in the upper gastrointestinal tract. In addition, bleeding can also occur from the oral pharynx.
Patients with an upper gastrointestinal bleeding may present with melena. However, melena is also a feature of small bowel bleeding and can also present as bleeding from the right-sided colon.  In addition, although fresh per rectal bleeding is frequently associated with lower gastrointestinal bleeding, upper gastrointestinal bleeding can also present as fresh per rectal bleeding. This may occur when the rate of bleeding is rapid, such as that seen by an arterial bleed from a duodenal ulcer.
The initial management of any patient who has an episode of upper gastrointestinal bleeding is resuscitation. This includes stabilizing the airway, breathing, and circulation, ensuring stable hemodynamics. Patients with massive hematemesis may aspirate blood leading to respiratory compromise or may be drowsy from hypovolemic shock. Those who are hypoxemic should be administered supplemental oxygen. Patients who have high oxygen flow requirements which cannot be adequately provided by intranasal prongs may require prophylactic intubation before endoscopy. Alternatively, those who require a face mask to provide oxygen may use specially designed endoscopy masks which have one-way valves that fit an endoscope without leaking air out (e.g., Endoscopy mask, VBM Medizintechnik GmbH, Germany). These masks can provide high-flow oxygen with positive pressure during endoscopy.  However, suctioning of the oral cavity is not possible with the mask in place. Therefore, they are only suited for short durations of use.
Hypotensive patients require resuscitation with volume expanders. A recent Cochrane systematic review on the role of red cell transfusion in gastrointestinal bleeding failed to show any benefit of transfusion. However, the review was hampered by the small numbers of studies with incomplete data. Well-designed clinical trials on this topic are currently lacking.  Nonetheless, fluids such as normal saline or blood products should be given as required. One should realize that correction of hypotension is not only to maintain end-organ perfusion, but also to facilitate the administration of drugs used in conscious sedation (e.g., midazolam) during endoscopy. Benzodiazepines may aggravate hypotension and should be used with caution in a patient with a borderline normotensive blood pressure.
Coagulopathy should be corrected with blood products in a bleeding patient. Patients on anticoagulation (e.g., warfarin) or antithrombotic therapy (e.g., aspirin, dipyridamole, clopidogrel) should have their medications stopped temporarily, as the risks of continued bleeding outweigh the risks of cerebrovascular or cardiovascular thrombosis. In patients who are taking warfarin for a mechanical cardiac valve, high doses of vitamin K should not be given. Rather, fresh frozen plasma with or without low-dose vitamin K (1-2 mg) is preferred.  One should note that the optimal target international normalized ratio (INR) for a bleeding patient has yet to be determined. INR levels below 1.5 do not require further correction. For patients with supra-therapeutic INR values, endoscopy should be postponed until the coagulopathy is corrected, if possible. Patients with a therapeutic INR of 2 and below do not usually need to wait for the coagulopathy to be corrected before endoscopy.  For patients who are on antiplatelet therapy, their medications should be stopped. Platelets can be transfused if needed to correct the functional platelet defect. Those with recently inserted cardiac stents (within one year), especially drug-eluting stents, should have their case discussed with the cardiologist. 
In determining the urgency of treatment during an episode of upper gastrointestinal bleeding and the prognosis of the patient, a risk stratification score may be used. There are several scores, among which the Rockall and the Glasgow-Blatchford scores are frequently used. ,, Risk stratification scores predict the need for intervention, risk of rebleeding, and risk of mortality.
The Rockall score is a prognostic score which was developed to determine the risk of rebleeding and death from an episode of upper gastrointestinal bleed [Table 1].  This score is based on the patient's age, presence of shock, coexisting illness, diagnosis, and stigmata of hemorrhage. Therefore, before the Rockall score can be calculated, the endoscopy must first be performed. Patients with a score of 2 or below are at low risk of rebleeding and death. Although one validation study found the prediction of risk of rebleeding to be unsatisfactory with the Rockall score, another large retrospective study found the score useful in influencing the management of patients based on their severity of illness. ,
The Glasgow-Blatchford score [Table 2] is another frequently utilized score for upper gastrointestinal bleeding. It can be calculated by tallying up the points for each of the following criteria: systolic blood pressure, blood urea nitrogen, hemoglobin, and the presence of tachycardia, melena, syncope, liver, or cardiac diseases.  High scores carry a graver prognosis and require intervention, as compared with lower scores. One advantage of this score is that it can be calculated when the patient presents to the physician. Unlike the Rockall score, both endoscopy and the diagnosis are not required for the Glasgow-Blatchford score. The Glasgow-Blatchford score has been used successfully to stratify patients into different risk groups. Those with low scores can be managed outpatient, whereas those with higher scores are admitted and treated in hospital. It has also been used to determine the need for intervention and the risk of mortality. 
Intravenous proton pump inhibitors (PPI) should be initiated in any patient suspected to have bleeding from the upper gastrointestinal tract. Intravenous PPIs are available as various preparations (e.g., omeprazole, pantoprazole, esomeprazole) and there is no current consensus which favors the use of one drug over the other.
PPI therapy is given as an intravenous loading dose, followed by an infusion. PPI should be started even before endoscopy is performed, but should not delay the endoscopy.  High-dose infusion PPI is believed to promote clot stability, facilitating hemostasis by raising the intragastric pH. ,, Starting a PPI infusion before endoscopy has been shown in clinical trials and a recent Cochrane systematic review to downstage the high-risk bleeding stigmata seen in endoscopy. PPI before endoscopy leads to a faster resolution of bleeding. , Therefore, even though no mortality benefit is seen in early PPI initiation, it is beneficial to the patient because less therapy is required. ,
During upper gastrointestinal endoscopy, bleeding lesions are graded based on the Forest classification [Table 3].  Lesions which are Forest II b or greater (i.e., Forest IIa, Ib, Ia) should continue with PPI infusion, whereas lesions which are Forest II c and below can be treated with oral PPI therapy. The commonly used PPI doses are below:
In the event that the patient has an allergy to PPI, histamine-2 receptor antagonists such as intravenous ranitidine can be given. However, the acid suppression of histamine-2 receptor antagonists is inferior to PPI infusion. , In a multicenter randomized controlled trial, intravenous PPI was superior to intravenous histamine-2 receptor antagonist in the treatment of arterial spurters.  Therefore, with the availability of PPI, histamine-2 receptor antagonists cannot be recommended unless PPI are contraindicated.  Histamine-2 receptor antagonist such as ranitidine can be given as intravenous ranitidine 50 mg 6 hourly or intravenous ranitidine 50 mg as a loading dose, followed by 6.25 mg per hour infusion (a total of 150 mg per day).
A prokinetic is a useful adjunct to endoscopy. It is helpful in patients who have large amounts of blood clots in their stomach (as seen endoscopically or suspected when persistent hematemesis occurs). It is also useful in patients who have recently consumed food and require urgent endoscopy.  Prokinetic therapy should be used only when indicated and not routinely. Drugs such as intravenous metoclopramide or erythromycin may be administered. Intravenous metoclopramide is prescribed as a bolus dose of 10 mg. Intravenous erythromycin is given at a dose of 3 mg per kg body weight. Usually, 250 mg of intravenous erythromycin is diluted in 100 ml of saline and infused over 15 minutes or longer. Erythromycin should be avoided, if possible, in patients with a prolonged corrected QT on electrocardiography, due to the risks of triggering life-threatening ventricular arrhythmias.
Timing of endoscopy
After resuscitation and medical therapy, endoscopic therapy should be performed. Proper preparation before endoscopy can lower the risk of complications of emergency endoscopy. , Once ready, endoscopy should be done earlier, rather than later. Early endoscopy (within 24 hours) for nonvariceal upper gastrointestinal bleeding has been shown in a systemic analysis to reduce the length of hospitalization, lower costs, lower transfusion requirements, and lead to better patient outcomes.  However, factors such as the availability of expertise, staff, and equipment, especially after duty hours, vary from institution to institution and may mandate a delay in endoscopy. Where possible, endoscopy should not be delayed longer than 24 hours.  The type of endoscopic management depends on the stigmata of hemorrhage seen. This can be graded according to the Forest classification [Table 3].  Endoscopic therapy is advocated in lesions which present with active bleeding (spurting or oozing), nonbleeding visible vessels, or adherent clots. These lesions carry a higher risk of rebleeding as compared with an ulcer with a hematin spot. They are graded as Forest II b and above. A randomized control trial showed that treating nonbleeding lesions such as a visible vessel or adherent clot endoscopically combined with PPI therapy was superior to PPI therapy alone in preventing rebleeding.  Evidence for the removal of adherent clots is weak, although it has been advocated in a recent consensus meeting on upper gastrointestinal hemorrhage.  When an adherent clot is seen, it is advisable to inject epinephrine at 4 quadrants around the clot before removing it. Clot removal can be performed by flushing it vigorously, displacing it with an instrument, sucking at it with the endoscope, or snaring the clot away.  Snaring the clot is an atraumatic method if done correctly. Guillotine cold snaring is performed by gently opening and closing the snare over the clot. This fragments the clot and may reveal an underlying vessel. One should be extremely cautious not to snare a protuberant vessel.
Clean base ulcers and ulcers with a flat red spot or necrotic base do not require endoscopic therapy because they are at very low risk of rebleeding. These lesions (graded Forest II c and below) can be treated with medical therapy alone. If a bleeding source is identified, hemostasis is applied through one of several methods.
Epinephrine injection is a useful method to induce hemostasis for both arterial and venous bleeding. However, the vasoconstrictive and tamponading effects of epinephrine are temporary. Meta-analysis of randomized controlled trials has shown clearly that when epinephrine injection therapy is combined with a second hemostatic modality, the rebleeding, emergency surgery, and mortality rates are reduced. This benefit is irrespective of the type of second modality used. Hence, it is recommended to combine epinephrine injection therapy together another hemostatic method. , Epinephrine injection is performed by injecting a solution of epinephrine (1 : 10 000 concentration) in aliquots of 1.0 ml around the bleeding source. Larger volumes of epinephrine and repeated injections may be required if the bleeding persists. The assistant should verbalise the amount of epinephrine injected so that the endoscopist is aware of the amount delivered. This is performed in 4 quadrants before assessing for a response. A response is indicated by temporary cessation of bleeding and accompanied with blanching of the surrounding mucosa into a pale whitish-red color. This indicates that vasospasm has been successfully induced by epinephrine. The endoscopist should not inject epinephrine directly into a vessel or bleeding point, as this may cumulate in sudden tachycardia, hypertension, and a restless patient.
The aim of thermal therapy is to coagulate the vessel. This can be performed with bipolar/multipolar electrocoagulation, heater probe, argon plasma coagulation, or laser. Thermal therapy with the heater probe and electrocoagulation works through the mechanism of coaptive coagulation of the vessel. Therefore, it is important to apply firm pressure against the vessel when using both these thermal therapies. The heater probe is used with low-power settings of 15 to 30 J. Sometimes, several pulses of the heater probe are required, before the vessel is ablated. Power settings in the duodenum should be lower than settings in the stomach where the gastric walls are thicker. After coagulation, a shallow depression remains where therapy was applied, and the vessel will be absent. Further therapy is required if bleeding persists, but will lead to deeper depressions. Hence, thermal therapy is unsuitable when deep cavitating ulcers are present, especially in the duodenum or small bowel, as there is a genuine risk of causing perforation.
Bipolar/multipolar electrocoagulation probes are safer than the older monopolar probes. This is because the depth of thermal therapy is shallower and more predictable. Electrocoagulation generates heat when electricity conducts through tissues with electrical resistance. Multipolar probes are usually large 10F probes and used at 30 to 40 W for a duration of up to 10 seconds.
Thermal therapy is particularly useful in treating lesions with oozing edges without a visible vessel. It is also useful in constricted areas of the gastrointestinal tract where there is limited distance between the endoscope and the lesion, preventing deployment of a clip. However, heater probes are not useful in targeting vessels when the plane of the lesion is almost parallel with the probe, since such an orientation would not allow enough pressure to tamponade the vessel during coaptive coagulation.
Argon plasma coagulation
Unlike other thermal therapies, the argon plasma coagulation does not require tissue contact. Argon plasma coagulation causes tissue coagulation when electricity is conducted across argon gas emitted by the catheter. , As such, the tissue damage is up to a depth of 2 to 3 mm only.  Because contact is not needed, argon plasma coagulation is useful for tangential lesions which are oozing.
Argon plasma coagulation is not useful when bleeding is occurring from a spurting vessel. In such cases, coaptive thermal coagulation or clipping is required. Argon plasma coagulation is useful for lesions which ooze, such as oozing from a raw surface of a gastric tumor, oozing from edges of ulcers or vascular lesions such as angiodysplasia or gastric antral vascular ectasia. Argon plasma coagulation is also useful for hemostasis in areas where the gastrointestinal wall layer is thin, such as in the small bowel and colon, since it burns superficially. As argon plasma coagulation involves unipolar cautery, one should be cautious in patients with an implantable cardio-defibrillator or pacemaker. 
Endoscopic clipping is a useful and safe method of hemostasis. This requires the assistance of a nurse who should be familiar with the loading and deployment of the clip. Clips are available as preloaded sets too, which are more convenient to use in the emergent setting of active bleeding. Frequently, multiple clips are required due to poor positioning or inadequate hemostasis. Therefore, it is crucial to obtain good positioning before deployment of the clip. The targeted vessel should always be visible. Application of the clip blindly, with blood obscuring the vessel, is usually a futile process. To obtain good positioning, several centimeters of distance between the endoscope and the vessel are required for the clip to be advanced and opened fully. Clips should be firmly opposed against the mucosa with the vessel between the prongs, before it is deployed. If bleeding persists, further clips can be placed or thermal therapy applied.
Endoscopic clipping has been shown to be superior to epinephrine injection alone in treating nonvariceal upper gastrointestinal bleeding. However, endoscopic clipping is neither superior nor inferior to thermal therapy in terms of rebleeding, surgical rates, or mortality.  Endoscopic clipping may be difficult in locations such as the posterior wall of the gastric body, lesser curve, and the posterior wall of the duodenal bulb. Hence, the choice of applying endoscopic clipping over thermal therapy should be based on the ease of application of therapy (which may be determined by the site of the bleeding), familiarity with the modality of therapy, potential contraindication to further thermal injury (e.g., deep ulceration in the duodenum with concerns of inducing perforation), type of stigmata of hemorrhage (oozing edges are more appropriately treated with thermal therapy), and coagulation status (thermal therapy may aggravate bleeding in severe coagulopathy).
Failure of endoscopic therapy
In patients who rebleed after the initial endoscopy, there is a role to repeat endoscopy again.  Otherwise, there is usually no necessity for a re-look endoscopy. Endoscopic retreatment is comparable with surgical intervention in patients who rebleed in terms of hospitalization duration, blood transfusion requirements, and mortality. Surgery is advocated when nonvariceal upper gastrointestinal bleeding cannot be controlled endoscopically or if the patient is persistently hemodynamically unstable.  Patients with persistent hypotension or large ulcers greater than 2 cm are likely to fail repeat endoscopy and may benefit from surgery.  Surgery is also indicated if a complication such as a perforation is present.
In patients who are deemed unfit for surgery, an alternative is percutaneous angiogram and selective embolization. Embolization is performed with the use of coils, alcohol, cyanoacrylic glue, gelatin sponges, or polyvinyl.  Complications of embolization include bowel ischemia, infarction of the stomach, liver, and spleen, and subsequent duodenal stenosis. ,
The management of upper gastrointestinal bleeding is divided into nonvariceal and variceal bleeding. Prognostic markers, such as the Glasgow-Blatchford score, can aid in deciding on the appropriate timing of endoscopy. , Adequate resuscitation before endoscopy leads to better endoscopic outcomes. , PPI therapy started early downstages the high-risk bleeding stigmata.  In nonvariceal bleeding lesions, epinephrine should be injected, followed by either thermal therapy or clipping. , When rebleeding occurs, endoscopy is useful,  failing which surgery or embolization should be considered. ,,
[Table 1], [Table 2], [Table 3]