INTRODUCTION — Upper gastrointestinal (GI) bleeding commonly presents with hematemesis (vomiting of blood or coffee-ground like material) and/or melena (black, tarry stools) (table 1). The diagnostic and initial therapeutic approach to the patient with upper GI bleeding will be reviewed here. This approach is consistent with a multidisciplinary international consensus statement updated in 2010 (table 2) [1,2].
The causes of upper GI bleeding are discussed separately. (See "Major causes of upper gastrointestinal bleeding in adults" and "Uncommon causes of upper gastrointestinal bleeding".)
APPROACH TO THE PATIENT — The initial evaluation of the patient with UGI bleeding involves an assessment of hemodynamic stability and resuscitation if necessary. Diagnostic studies (usually endoscopy) follow, with the goal of both diagnosis, and in some circumstances, treatment of the specific disorder.
History — Important elements of the history include use of aspirin and other NSAIDs, alcohol, history of liver disease or variceal bleeding, history of ulcers, weight loss, dysphagia, heartburn, an abdominal aortic aneurysm (AAA), or an abdominal aortic vascular graft. A history of an AAA or graft increases concern regarding an aortoenteric fistula and may require a CT scan for evaluation. (See "Uncommon causes of upper gastrointestinal bleeding".)
Resuscitation — All patients with hemodynamic instability (shock, orthostatic hypotension, decrease in hematocrit of at least 6 percent, or transfusion requirement over two units of packed red blood cells) or active bleeding (manifested by hematemesis, bright red blood per nasogastric tube, or hematochezia) should be admitted to an intensive care unit for resuscitation and close observation with automated blood pressure monitoring, ECG monitoring, and pulse oximetry (table 3).
Adequate resuscitation and stabilization is essential prior to endoscopy to minimize treatment-associated complications . Two large caliber (16 gauge or larger) peripheral catheters or a central venous line should be inserted for intravenous access.
Gastroenterological consultation should be obtained. The decision to obtain a surgical consultation prior to endoscopy should be based on the likelihood for persistent or recurrent bleeding, or risks/complications stemming from endoscopic therapy (perforation, precipitation of massive bleeding). As a general rule, we obtain surgical consultation if endoscopic therapy is unlikely to be successful, or if the patient is deemed to be at high risk for rebleeding or complications associated with endoscopy. In addition, a surgeon should be promptly notified of all patients with severe acute UGI bleeding .
High-risk patients (eg, older adults or those who have severe comorbid illnesses such as coronary disease or cirrhosis) should receive packed red blood cell transfusions to maintain the hemoglobin above 10 gm/dL. Older adults can do as well as younger patients with acute UGI bleeding . However, older adults with cardiovascular disease are at increased risk for a myocardial infarction and should thus be monitored appropriately; consideration should be given to ruling out a myocardial infarction. Young and otherwise healthy patients should be transfused to maintain their hemoglobin above 7 gm/dL. Patients with active bleeding and a coagulopathy (prolonged prothrombin time with INR >1.5) or low platelet count (<50,000/microL) should be transfused with fresh frozen plasma and platelets, respectively.
Elective endotracheal intubation in patients with ongoing hematemesis or altered respiratory or mental status may facilitate endoscopy and decrease the risk of aspiration.
Somatostatin, or its analog octreotide, which have been best studied in the treatment of variceal bleeding, may also reduce the risk of bleeding due to nonvariceal causes . It can be used as adjunctive therapy before endoscopy, or when endoscopy is unsuccessful, contraindicated, or unavailable . However, it is not routinely recommended for patients with acute nonvariceal UGI bleeding. (See "Treatment of bleeding peptic ulcers".)
Nasogastric tube — Nasogastric or orogastric tube lavage should be performed to remove particulate matter, fresh blood, and clots to facilitate endoscopy and confirm an upper source of bleeding. A nasogastric tube lavage that yields blood or coffee-ground like material confirms the diagnosis and predicts whether bleeding is caused by a high-risk lesion . However, lavage may not be positive if bleeding has ceased or arises beyond a closed pylorus. The presence of bilious fluid suggests that the pylorus is open and, if lavage is negative, that there is no active upper GI bleeding distal to the pylorus . In comparison, hematochezia (bright red or maroon colored blood or fresh clots per rectum) is usually a sign of a lower GI source (defined as distal to the ligament of Treitz). (See "Approach to the adult patient with lower gastrointestinal bleeding".)
Significance of stool color — Stool color is not a reliable indicator of the location of bleeding since melena can be seen with proximal lower GI bleeding, and hematochezia can be seen with massive upper GI bleeding. In one series of 80 patients with severe hematochezia, for example, 74 percent had colonic lesions (primarily angiodysplasia, diverticula, and polyps or cancer), 11 percent had an upper GI lesion, 9 percent a presumed small bowel source, and no lesion site was identified in 6 percent . Patients who have hematochezia from upper GI bleeding usually have massive bleeding and are orthostatic.
Diagnostic studies — Upper endoscopy is the diagnostic modality of choice for acute UGI bleeding [10,11]. Endoscopy is highly sensitive and specific for locating and identifying bleeding lesions in the upper gastrointestinal tract. In addition, once a bleeding lesion has been identified, therapeutic endoscopy can achieve acute hemostasis and prevent recurrent bleeding in most patients. (See "Treatment of bleeding peptic ulcers".) Early endoscopy (within 24 hours) is recommended for most patients with acute UGI bleeding.
It is frequently helpful to irrigate the stomach prior to endoscopy to help remove residual blood and other gastric contents. However, despite irrigation, the stomach can be obscured with blood, potentially making it difficult to establish a clear diagnosis and/or perform therapeutic maneuvers. In patients in whom bleeding stopped spontaneously, a second-look endoscopy may be required to establish a diagnosis, but routine second look endoscopy is not recommended.
Erythromycin — Pharmacologic means to address the problem of poor visualization have been the focus of at least two randomized controlled trials (one involving 105 patients and the other involving 41 patients), which have suggested that a single dose of intravenous erythromycin given 20 to 120 minutes before endoscopy can significantly improve visibility, shorten endoscopy time, and reduce the need for a second-look endoscopy [12,13]. Erythromycin promotes gastric emptying based upon its ability to be an agonist of motilin receptors and may help with gastric visualization. Treatment appeared to be safe in both studies. Thus, this approach can be considered in patients who are likely to have a stomach full of blood, such as those with severe bleeding. A reasonable is 3 mg/kg intravenously over 20 to 30 minutes, 30 to 90 minutes prior to endoscopy.
Risks of endoscopy — Risks of upper endoscopy include aspiration, adverse reactions to conscious sedation, perforation, and increasing bleeding while attempting therapeutic intervention.
The risks versus benefits of upper endoscopy should be considered in high-risk patients, such as those who have had a recent myocardial infarction. In one study, for example, 200 patients who underwent endoscopy within 30 days after myocardial infarction (AMI) were compared to 200 controls matched for age, sex, and endoscopic indication . Complications (including fatal ventricular tachycardia, near respiratory arrest and mild hypotension) occurred more often in patients who had a recent AMI (7.5 versus 1.5 percent). Complications occurred far more often (21 versus 2 percent) in patients who were very ill (Apache II score >16 or hypotension prior to endoscopy). Notably, such patients are at increased risk for complications even without endoscopy (and may be particularly vulnerable to complications from continued bleeding without endoscopy). (See "Predictive scoring systems in the intensive care unit".)
Other diagnostic tests — Other diagnostic tests for acute UGI bleeding include angiography and a tagged red blood cell scan, which can detect active bleeding [15,16]. UGI barium studies are contraindicated in the setting of acute UGI bleeding because they will interfere with subsequent endoscopy, angiography, or surgery . (See "Angiographic control of gastrointestinal bleeding" and "Evaluation of obscure gastrointestinal bleeding".)
Risk stratification — Endoscopic, clinical, and laboratory features may be useful for risk stratification of patients who present with UGI bleeding (table 4 and picture 1) [17-25]. Combining this information, several investigators have developed decision rules and predictive models that permit identification of individuals who are at low risk for recurrent or life-threatening hemorrhage . Such individuals may be suitable for early hospital discharge or even outpatient care.
Several studies have evaluated the effectiveness of such rules in a variety of clinical settings [17-23,26-34]. Most have suggested that application of these rules leads to comparable patient safety and reduced utilization. Thus, it is generally accepted that patients at low risk for recurrent bleeding can be identified. One widely cited scoring system, the Rockall score, for example, is based upon age, the presence of shock, comorbidity, diagnosis, and endoscopic stigmata of recent hemorrhage (calculator 1) . In one validation study, only 32 of 2531 (4.3 percent) who scored 2 or less (out of a maximum of 11) rebled and only one died. On the other hand, in a later study of 247 patients who underwent endoscopic therapy for bleeding peptic ulcers, the model (known as the Rockall Scoring System) performed poorly on predicting recurrent bleeding, underscoring the need for validation of these models .
Another commonly cited score is the Blatchford score, which unlike the Rockall score does not take endoscopic data into account (calculator 2) . The score is based upon blood urea nitrogen, hemoglobin, systolic blood pressure, pulse, presence of melena, presence of syncope, hepatic disease, and cardiac failure. The score ranges from zero to 14 or greater and the risk of requiring endoscopic intervention increases with increasing score. Patients with a score of zero are considered low risk and the risk of intervention increases with increasing score.
What is less clear is whether the application of these decision rules in real-life settings can reduce utilization, a point emphasized in a controlled trial involving 93 outpatients with acute upper GI bleeding . Patients were randomly assigned to urgent endoscopy (before hospitalization) or elective endoscopy after admission. Results of the urgent endoscopy and a recommendation regarding patient disposition were provided to the attending physician who made final decisions regarding patient disposition.
The timing of endoscopy did not affect resource utilization or patient outcomes. Length of stay was similar (3.98 versus 5.11 days in the urgent and delayed groups, respectively) as was the mean number of days in the intensive care unit (1.2). Outpatient care was recommended for 19 patients (40 percent) in the urgent groups, but only four were discharged.
These findings contrast with several other controlled trials that demonstrated reduced resource utilization from early endoscopy. The main difference lies in the degree to which the risk stratification actually translated into changes in patient management. Studies showing reduced utilization have incorporated processes by which patient disposition was linked directly to the risk stratification system.
Considered together, these data suggest that risk stratification is feasible, permitting identification of patients who can be managed safely without hospitalization. However, for these systems to be successful, care processes must include mechanisms by which the risk stratification system is tied directly to decisions regarding patient discharge. None of the published risk scores has yet been adopted widely.
We do not routinely use any of the scoring systems of risk stratification, though their use is recommended by the International Concensus on Nonvariceal Upper Gastrointestinal Bleeding . As a general rule, we discharge patients who are young and have no comorbidities, have negative NGT aspiration, stable vital signs, and a normal HCT provided that an upper endoscopy has identified a likely source of bleeding and the cause of bleeding is not associated with a high rebleeding risk (eg, variceal bleeding, active bleeding, bleeding from a Dieulafoy's lesion or ulcer bleeding with certain stigmata) (table 4). However, this suggestion depends upon individual-patient factors such as reliability for follow-up and confidence in the diagnosis; in some cases we admit patients for observation. We admit patients to a monitored setting or intensive care unit (depending upon the severity of bleeding, comorbidities, and stability of vital signs) if they do not meet these criteria. Most patients who have received endoscopic treatment for high-risk stigmata should be hospitalized for 72 hours to monitor for rebleeding.
Acid suppression — Several studies have examined the role of acid suppression given before or after endoscopy (with or without therapeutic intervention) . In the setting of active UGI bleeding, acid suppressive therapy with H2 receptor antagonists has not been shown to significantly lower the rate of ulcer rebleeding [37-39]. By contrast, high dose antisecretory therapy with an intravenous infusion of proton pump inhibitor (IV PPI with pantoprazole or omeprazole) significantly reduced the rate of rebleeding as compared to standard treatment in patients with bleeding ulcers. Oral and intravenous PPI therapy also decrease length of hospital stay, rebleeding rate, and the need for blood transfusion in high-risk ulcer bleeders treated with endoscopic therapy. (See "Treatment of bleeding peptic ulcers".)
An intravenous PPI given before endoscopic therapy in patients with UGI bleeding can reduce signs of bleeding and the need for endoscopic therapy. One of the most methodologically rigorous controlled trials evaluating this approach included 638 patients with UGI bleeding who were randomly assigned to intravenous omeprazole or placebo before endoscopy . Patients randomized to omeprazole had significantly shorter length of stay, fewer actively bleeding ulcers (6 versus 15 percent) and more ulcers with a clean base. There was no significant difference in the proportion needing surgery or with recurrent bleeding.
We suggest that patients with upper GI bleeding be started on an intravenous PPI. It can be started at presentation and continued until confirmation of the cause of bleeding, after which the need for specific therapy can be determined. Administering an IV PPI before endoscopy was slightly more costly and effective than administering it after endoscopy in a cost-effectiveness analysis . The benefit of PPIs has been best proven in bleeding from peptic ulcers. (See "Treatment of bleeding peptic ulcers".)
Although omeprazole has been the most extensively studied, other intravenous formulations of proton pump inhibitors given in doses that are known to inhibit gastric acid secretion are probably acceptable alternatives. Pantoprazole, lansoprazole and esomeprazole are the only intravenous formulations available in the United States. The suggested IV pantoprazole dose is 80 mg bolus followed by 8 mg/hr infusion. If there is no rebleeding within 24 hours, the patient may be switched to oral pantoprazole 40 mg/day or omeprazole 20 mg/day. Twice daily dosing of an oral proton pump inhibitor may be a reasonable alternative if intravenous formulations are not available. Patients should also be tested for H. pylori infection and treated if infected, with confirmation of eradication following treatment. Negative testing obtained in the acute setting should be repeated to confirm h. pylori negativity. (See "Indications and diagnostic tests for Helicobacter pylori infection" and "Treatment regimens for Helicobacter pylori".)
Antibiotics for patients with cirrhosis — Bacterial infections are present in up to 20 percent of patients with cirrhosis who are hospitalized with gastrointestinal bleeding; up to an additional 50 percent develop an infection while hospitalized. Such patients have increased mortality. Multiple trials evaluating the effectiveness of prophylactic antibiotics in cirrhotic patients hospitalized for bleeding suggest an overall reduction in infectious complications and possibly decreased mortality. Antibiotics may also reduce the risk of recurrent bleeding in hospitalized patients who bled from esophageal varices. A reasonable conclusion from these data is that patients with cirrhosis who present with upper GI bleeding (from varices or other causes) should be given prophylactic antibiotics, preferably before endoscopy (although effectiveness has also been demonstrated when given after endoscopy). (See "General principles of the management of variceal hemorrhage".)
TREATMENT — The specific treatment of patients with upper gastrointestinal bleeding due to various causes is discussed separately. (See "Treatment of bleeding peptic ulcers" and "Contact thermal devices, endoscopic hemoclips, and epinephrine injection for ulcer hemostasis".)
INFORMATION FOR PATIENTS — Educational materials on this topic are available for patients. (See "Patient information: Upper endoscopy" and "Patient information: Peptic ulcer disease".) We encourage you to print or e-mail these topic reviews, or to refer patients to our public web site, www.uptodate.com/patients, which includes these and other topics.
SUMMARY AND RECOMMENDATIONS
UpToDate performs a continuous review of over 440 journals and other resources. Updates are added as important new information is published. The literature review for version 18.3 is current through September 2010; this topic was last changed on October 8, 2010. The next version of UpToDate (19.1) will be released in March 2011.