Upper GI (UGI) bleeding is any GI bleeding originating proximal to the ligament of Treitz. The overall annual incidence of UGI bleeding ranges from 39 to 172 per 100,000 in Western countries.^1,2,3 Difference in prevalence between countries is attributed to variations in Helicobacter pylori rates, socioeconomic conditions, and prescription patterns of ulcer-healing and ulcer-promoting medications.² Increasing age, coexistent organ system disease, and recurrent hemorrhage are factors associated with increased morbidity and mortality.³

PEPTIC ULCER DISEASE

Despite a downward trend in prevalence over the past 20 years, peptic ulcer disease, which includes gastric, duodenal, esophageal, and stomal ulcers, is still considered the most common cause of UGI bleeding.^2,4 However, the Analysis of Clinical Outcomes Research Initiative found gastric and duodenal ulcers in only 20.6% of 7822 endoscopies performed for suspected UGI bleeding.⁴ This number is much lower than previous estimates of up to 50%.^5,6 Awareness that aspirin, nonsteroidal anti-inflammatory drugs (NSAIDs), and smoking cause bleeding and increased recognition and treatment of H. pylori infection may be responsible for decreased incidence.^7,8,9,10

EROSIVE GASTRITIS AND ESOPHAGITIS

Erosive gastritis, esophagitis, and duodenitis are also common causes of GI hemorrhage.¹¹ Common predisposing factors include alcohol, salicylates, and NSAIDs. Infection, toxic ingestion, radiation, and stress from severe illness may also cause erosive gastritis. Stress-related mucosal disease occurs in patients with overwhelming sepsis, trauma, or respiratory failure requiring mechanical ventilation. Candida, herpes simplex virus, cytomegalovirus, and human immunodeficiency virus are potential sources of esophageal bleeding from infection.

ESOPHAGEAL AND GASTRIC VARICES

Esophageal and gastric varices result from portal hypertension and, in the United States, are most often a result of alcoholic liver disease.¹² Although varices account for a small percentage of all cases of UGI hemorrhage, they can rebleed and carry a high mortality rate. However, many patients with end-stage cirrhosis never develop varices; many patients with documented varices never bleed; and many patients with a documented history of varices presenting with UGI bleeding will actually bleed from nonvariceal sites. Variceal bleeding is the cause of UGI bleeding in cirrhotics 59% of the time, followed by peptic ulcer disease in 16% of cases.¹³ In-hospital mortality rates for any type of GI bleed in cirrhotics are essentially double those of noncirrhotic patients.¹⁴

MALLORY-WEISS SYNDROME

Mallory-Weiss syndrome is bleeding secondary to a longitudinal mucosal tear at the gastroesophageal junction. The classic history is repeated vomiting followed by bright red hematemesis. The syndrome can be associated with alcoholic binge drinking, diabetic ketoacidosis, or chemotherapy administration. The Valsalva maneuver, such as from coughing or seizures, is also a reported cause.

DIEULAFOY LESIONS

Dieulafoy lesions are arteries of the GI tract that protrude through the submucosa. They are most commonly found in the lesser curvature of the stomach but may be found anywhere in the GI tract; 80% to 95% are found within 6 cm of the gastroesophageal junction.¹⁵ These lesions are characterized by intermittent massive GI bleeding, without the standard predisposing factors of liver disease or NSAID use. Dieulafoy lesions are difficult to diagnose endoscopically, and sometimes patients report multiple previous diagnostic maneuvers with negative results.

OTHER CAUSES

Arteriovenous malformation and malignancy are other causes of UGI hemorrhage. Significant bleeding from ear, nose, and throat sources can also masquerade as GI hemorrhage. An aortoenteric fistula secondary to a preexisting aortic graft is an unusual but important cause of bleeding to keep in mind. Classically, this presents as a self-limited “herald” bleed with hematemesis or hematochezia, which precedes massive hemorrhage and exsanguination.

HISTORY

Ask about hematemesis, coffee-ground emesis, or melena. Classically, hematemesis and coffee-ground emesis suggest a UGI source. The presence of melena and age <50 years old more likely indicate an upper GI bleed versus a lower GI bleed, even in patients without hematemesis.¹⁶ Vomiting and retching, followed by hematemesis, suggest a Mallory-Weiss tear. Be sure to ask about prior episodes of GI bleeding and any interventions performed. A history of an aortic graft should suggest bleeding from an aortoenteric fistula. Review the patient’s medication list carefully. Salicylates, glucocorticoids, NSAIDs, and anticoagulants all place the patient at high risk for GI bleed. Alcohol abuse is strongly associated with a number of causes of bleeding, including peptic ulcer disease, erosive gastritis, and esophageal varices. Ingestion of iron or bismuth can simulate melena. Liquid medications with red dye, as well as certain foods, such as beets, can simulate hematochezia. In such cases, stool guaiac testing will be negative. Inquire about past history of GI bleeding, even though recurrent bleeding episodes may originate from different sources.

Although the medical history may suggest the source of bleeding, history can also be misleading. For instance, what initially appears to be lower GI bleeding may actually be a UGI bleed in disguise. Bright red or maroon rectal bleeding unexpectedly originates from UGI sources about 14% of the time.¹⁶ Although patients volunteer complaints of hematemesis or melena, if there is no vomiting or the patient has not noted tarry stools, signs may be subtle. Patients with hypotension, tachycardia, angina, syncope, weakness, confusion, or cardiac arrest may have underlying GI hemorrhage.

PHYSICAL EXAMINATION

Visual inspection of the vomitus for a bloody, maroon, or coffee-ground appearance is the most reliable way to diagnose UGI bleeding in the ED. Consider keeping a sample of the vomitus or nasogastric (NG) aspirate at bedside for the gastroenterologist to view.

Vital signs may reveal obvious hypotension and tachycardia or more subtle findings such as decreased pulse pressure or tachypnea. Younger patients and those without comorbidities can tolerate substantial volume loss with minimal or no changes in vital signs. Paradoxical bradycardia may occur even in the face of profound hypovolemia. Remember that comorbid conditions and medications may mask the body’s physiologic response to volume loss. β-Blockers, for example, will prevent tachycardia. Patients with baseline hypertension may have relatively normal blood pressure in the setting of hypovolemia.

Cool, clammy skin is an obvious sign of shock. Spider angiomas, palmar erythema, jaundice, and gynecomastia suggest liver disease. Petechiae and purpura suggest an underlying coagulopathy. Facial lesions, cutaneous macules, or telangiectasias may be suggestive of the Peutz-Jeghers, Rendu-Osler-Weber, or Gardner’s syndromes. A careful ear, nose, and throat examination can reveal an occult bleeding source that has resulted in swallowed blood and subsequent coffee-ground emesis. Abdominal examination may disclose tenderness, masses, ascites, or organomegaly.

Perform rectal examination to detect the presence of blood and its appearance, whether bright red, maroon, or melanotic.

LABORATORY TESTING

In patients with significant bleeding, the single most important laboratory test is to obtain blood for type and cross-match in case transfusion is needed. A CBC is also important, although the initial hematocrit level may not reflect the actual amount of acute blood loss. In addition, consider BUN, creatinine, electrolyte, glucose, coagulation, and liver function studies. UGI hemorrhage will elevate BUN levels through digestion and absorption of hemoglobin. A BUN:creatinine ratio ≥30 suggests a UGI source of bleeding.¹⁷ Coagulation studies, including INR, partial thromboplastin time, and platelet count, are useful in patients taking anticoagulants and those with underlying hepatic disease. Obtain an ECG in patients with underlying coronary artery disease. Silent cardiac or mesenteric ischemia can develop if bleeding decreases cardiac or mesenteric perfusion. A single elevated lactate level is a sentinel sign of severe illness. The success or failure of resuscitation efforts can be assessed by following dynamic lactate levels, because a rising lactate level in the hospital setting is a clear predictor of in-hospital mortality.¹⁸

Routine abdominal and chest radiographs are of limited value and are not needed in the absence of specific clinical indications. Barium contrast studies are contraindicated because barium may hinder subsequent endoscopy or angiography.

In cases where traditional endoscopy is unavailable or endoscopic visualization is unable to find the source, consider tagged red-cell scintigraphy or visceral angiography. Both of these tests will demonstrate the source only in cases of active bleeding. Scintigraphy and angiography help localize the source of bleeding to determine whether medical or surgical management is optimal.

NASOGASTRIC LAVAGE

NG intubation and aspiration are diagnostic and therapeutic.¹⁹ In patients without a history of hematemesis, a positive aspirate provides strong evidence for a UGI source of bleeding. High-risk lesions are more likely in patients with bloody aspirates. Visual inspection of the aspirate to identify bloody, maroon, or coffee-ground material verifies UGI bleeding. Early NG lavage is associated with decreased time to endoscopy.²⁰ NG tube placement and lavage can confirm the diagnosis of UGI bleeding and stratify risk.

A negative NG aspirate does not conclusively exclude a UGI source. Intermittent bleeding, pyloric spasm, or edema preventing reflux of duodenal blood can cause false-negative results. Ultimately, NG aspiration yields a positive result in only 23% of patients without hematemesis who have occult UGI bleeding.²¹