Chapter 82. Gastrointestinal Hemorrhage

• Aggressive intravenous resuscitation with fluids and blood, and airway protection are crucial in the management of the acutely bleeding patient.
• Endoscopy should be performed with therapeutic intent for both upper and lower gastrointestinal bleeding.
• Pharmacologic therapy should be used as an adjunct to endoscopic therapy.
• An early team approach, involving medical, radiologic, and surgical personnel, should be implemented.
• In the setting of severe bleeding or bleeding refractory to endoscopic therapy, angiographic and surgical therapies should be instituted promptly.

Gastrointestinal (GI) hemorrhage continues to be a frequent indication for intensive care management, with over 300,000 hospitalizations reported annually in the United States.1 Upper GI (UGI) bleeding has continued to predominate, with lower GI (LGI) bleeding constituting approximately 25% of all GI bleeding.2 Despite improved diagnostic and therapeutic modalities in the last two decades, the mortality rates for upper and lower GI hemorrhage have demonstrated different trends. Mortality from UGI bleeding has remained stable at 10%,3 and this could be explained by an aging population with a significantly higher GI bleeding mortality due to comorbid conditions.4 In contrast, the mortality from LGI bleeding has decreased dramatically despite an aging population, and this is probably due to more aggressive diagnostic and therapeutic endoscopic intervention.

The management of GI hemorrhage in the ICU is multidisciplinary, involving the intensivist, gastroenterologist, radiologist, and surgeon. A successful outcome relies on effective fluid resuscitation, maintenance of adequate perfusion pressure, prompt hemostasis, monitoring of end-organ function, and prevention of multiple-organ failure.

Prognostic Factors

Multiple studies focusing primarily on nonvariceal UGI bleeding have been designed to define prognostic factors for GI bleeding and to identify high-risk patients.5–7 A common and pivotal feature of these studies is the combined use of clinical variables and endoscopic findings to guide risk stratification, thereby stressing the importance of integrating clinical and endoscopic information for optimal decision making. Table 82-1 outlines the clinical and endoscopic indicators associated with an increased risk of rebleeding and higher mortality. A recent study has identified similar prognostic indicators for LGI bleeding.8

Initial Presentation

GI bleeding is divided into UGI and LGI bleeding based on its location proximal or distal to the ligament of Treitz at the junction of the duodenum and jejunum. UGI bleeding commonly presents with hematemesis and/or melena, and a nasogastric (NG) lavage that yields blood or coffee-ground material supports the diagnosis. However, it is important to note that a negative or bile-stained NG aspirate (indicating an open pylorus) does not exclude a UGI source because the bleeding may be intermittent.9 In comparison, hematochezia is usually the presenting sign of an LGI source. These distinctions based on presenting signs are not absolute because melena can be seen with proximal LGI bleeding and hematochezia can be present with massive, brisk UGI bleeding.

Initial Evaluation and Management

Table 82-2 outlines the initial evaluation and management for GI hemorrhage.

Hemodynamic

Regardless of the etiology and site of GI hemorrhage, the initial management should be directed at maintaining hemodynamic stability by restoring intravascular volume. Intravenous access with two large-bore IV catheters should be maintained at all times. In the presence of hypotension or hypovolemic shock, prompt fluid resuscitation with crystalloids and packed red blood cells is essential. Monitoring end-organ perfusion and preventing ischemic organ injury improves survival. In particular, coronary and renal perfusion should be assessed. An electrocardiogram should be obtained in patients at risk for myocardial ischemia, and renal laboratory parameters and urine output should be followed to assess for possible prerenal azotemia, acute renal failure, and (in cirrhotics) hepatorenal syndrome. In the subset of patients with suspected variceal hemorrhage, central venous pressure (CVP) monitoring may be useful to prevent sustained portal hypertension and recurrent bleeding following fluid replacement; a CVP of less than 10 mm Hg may prevent recurrent variceal bleeding. When left-sided heart failure coexists, monitoring of pulmonary artery wedge pressure (Ppw) may facilitate aggressive fluid resuscitation while reducing the risk of cardiogenic pulmonary edema.

Gastrointestinal/Endoscopic

Prompt identification and hemostasis of the source of GI hemorrhage are essential in improving patient outcome. In the event that initial fluid resuscitation establishes hemodynamic stability, endoscopy may be performed under stable conditions within the first 24 hours of the bleed. However, emergent endoscopy with therapeutic hemostatic intent should be performed for patients with UGI and LGI hemorrhage who cannot be stabilized hemodynamically with intravascular volume resuscitation and continue to bleed. Multiple studies in UGI bleeding10,11 and more recent studies in LGI bleeding12,13 have demonstrated that therapeutic esophagogastroduodenoscopy (EGD) and colonoscopy using hemostatic interventions such as epinephrine injection and electrocautery facilitate early control of bleeding, reduce rebleeding rates, and improve short-term morbidity and mortality. Based on these observations, the 2001 American Society for Gastrointestinal Endoscopy (ASGE) recommendations emphasize the importance of emergent endoscopy with therapeutic intent in the setting of persistent hemodynamic instability.14,15 Absolute contraindications to endoscopy include suspected GI perforation, acute uncontrolled unstable angina, severe coagulopathy, untreated respiratory decompensation, and severe patient agitation. Apart from perforation, other conditions contraindicating endoscopy can be corrected, following which endoscopy should be performed. In the face of massive exsanguination, emergent surgical intervention (perhaps facilitated by intraoperative endoscopy) should be considered instead of endoscopy.

A potential caveat to emergent endoscopy is the marginally increased risk of complications compared to elective endoscopy, although the development of smaller and more flexible scopes is likely to reduce this risk. A specific risk is perforation, and with regard to upper endoscopy, the anatomically susceptible regions are the cricopharyngeal region at the site of esophageal intubation and the pylorus-duodenum junction during a blind duodenal intubation. The risk of perforation during emergency EGD can be lowered with pre-endoscopic NG lavage, which facilitates visualization. Recent studies have suggested that the prokinetic agent erythromycin, given as a single intravenous dose of 250 mg prior to EGD, improves visualization and diagnosis.16

In the setting of nonvariceal UGI bleeding, acid suppression with histamine receptor antagonists and proton pump inhibitors has shown short-term efficacy only after hemostasis has been achieved using some form of therapeutic endoscopic intervention.17,18 Therefore, acid suppression may be beneficial in the short-term management of UGI bleeding following endoscopic intervention. In the setting of variceal bleeding, pharmacologic therapy with a splanchnic vasoconstrictor such as octreotide and empiric antibiotic therapy should be initiated.

Hematologic

In order to ensure adequate oxygen-carrying capacity in the circulation and to prevent end-organ ischemia, the hematocrit should be maintained above 30%. It should be noted that the initial hematocrit after an acute GI bleed can be misleading because acute hemorrhage produces loss of whole blood and the hematocrit does not change initially because the initial loss of plasma and erythrocytes is equivalent. Within 24 to 72 hours of the initial bleed, plasma is redistributed from the extravascular to the intravascular space, thus resulting in a dilution of the red cell mass and a fall in the measured hematocrit. Intravenous hydration with crystalloids compounds this dilutional anemia so that red cells should be replaced promptly. In most instances, there is sufficient time to allow typing and cross-matching of red cells; however, in the setting of massive exsanguination, the transfusion of non–cross-matched type-specific blood may be necessary. In the presence of thrombocytopenia, platelets must be transfused to maintain the count above 60,000/μL. Any existing coagulopathy should be corrected with fresh frozen plasma to lower the prothrombin time to within 2 seconds of the normal range. Recent studies have demonstrated the ability of recombinant activated factor VIIa (rFVIIa) to rapidly correct severe coagulopathy in hepatic failure,19,20 and this agent may assume an important role in the treatment of coagulopathy in the future. During the process of aggressive intravascular volume resuscitation, fluids and blood products should be warmed to prevent the development of a cold coagulopathy, and the core body temperature should be maintained above 35°C (see Chaps. 68 and 69).

Pulmonary

In the presence of active hematemesis, an NG tube should be placed to decrease the risk of aspiration. Endotracheal intubation should be performed for airway protection and to decrease the risk of aspiration in the following situations: (1) in the presence of active hematemesis and decreased mental status, (2) prior to an emergent EGD for active hematemesis, and (3) prior to insertion of an esophageal tamponade tube. In the setting of shock, intubation and full mechanical ventilatory support are indicated to decrease the oxygen consumption of the respiratory apparatus. During the performance of an EGD, significant hypoxemia may occur, especially in elderly patients and in those with moderate to severe obstructive pulmonary disease (defined as a FEV1/FVC ratio of less than 0.6).21 Probable causes include hypoventilation due to sedative agents, the partial airway obstruction produced by the endoscope, and aspiration of gastric contents, resulting in bronchospasm and ventilation-perfusion mismatch.

Consultation

Radiology and surgery should be consulted early in the course of management. In the setting of nondiagnostic upper or lower endoscopy, radiologic modalities offer diagnostic and therapeutic options to achieve hemostasis. Emergent surgical intervention is indicated in the exsanguinating patient who may not be stable enough for endoscopic or radiologic evaluation.

When endoscopic evaluation is unable to identify the cause of bleeding, the two diagnostic radiologic modalities available are angiography and radionuclide scans. As outlined later, angiography with embolotherapy or selective infusion of vasoconstrictors offers therapeutic options as well (see Chap. 101).

The angiographic diagnosis of acute arterial hemorrhage is based on visualization of extravasated contrast material in the gastrointestinal lumen. Therefore, it only identifies the bleeding site if active bleeding is occurring when the study is performed. In addition, the rate of bleeding must be brisk, in the range of 0.5 to 1 mL/min. In the case of brisk UGI bleeding, angiography may demonstrate a bleeding site in 75% of patients, with most bleeding episodes originating from a branch of the left gastric artery. In the setting of LGI bleeding, the average diagnostic yield is decreased to about 60%, with diverticular disease and vascular ectasia being the most common findings.

Radionuclide studies occasionally aid in the detection of the bleeding site. The current radionuclide scan of choice is the 99mTc-pertechnetate-labeled red blood cell scan. The radionuclide scan offers the ability to detect rates of bleeding of less than 0.5mL/min, and the 48-hour stability of the tagged red blood cells allows repeated nuclear imaging for 1 to 2 days following administration of the radionuclide in the setting of intermittent bleeding. However, a positive radionuclide study localizes the bleeding only to an area of the abdomen and cannot define the mucosal location of the bleeding site precisely. Therefore, a positive result should prompt a repeat endoscopy or angiography to localize the bleeding site precisely.

Rebleeding

In most instances, the presentation of rebleeding is similar to that of the initial episode, and the source is identical to the site of the initial bleed. After hemostasis of the initial bleed following spontaneous cessation or therapeutic intervention, the patient should be monitored closely for rebleeding, especially in the presence of clinical and endoscopic indicators associated with an increased risk of rebleeding (see Table 82-1). Apart from the obvious signs of gastrointestinal blood loss characterized by melena, hematemesis, or hematochezia, more subtle signs of rebleeding may include tachycardia and hypotension owing to a decreasing intravascular volume. Therefore, continuous hemodynamic monitoring should be performed following initial hemostasis, and invasive monitoring with a central venous catheter or an arterial line may be considered for the patient at high risk for rebleeding. A falling serum hemoglobin concentration on serial measurements may suggest a recurrent bleed. The management of rebleeding should include immediate repeat endoscopic intervention targeted at the initial lesion, followed by radiologic or surgical intervention, if necessary. Specific pharmacologic and endoscopic interventions for long-term secondary prophylaxis against rebleeding will be discussed in the sections that follow.

With regard to prognosis and treatment, UGI hemorrhage can be divided into (1) variceal hemorrhage and (2) nonvariceal hemorrhage.

Variceal hemorrhage presents as a symptom of decompensated cirrhosis in as many as 50% of patients and accounts for about one-third of all deaths related to cirrhosis. Mortality is related to hepatic disease severity, as defined by the Childs-Pugh classification (Table 82-3), with an overall mortality estimated at 50%. There are two distinct phases in the course of variceal hemorrhage. In the first phase, defined by the initial episode of active hemorrhage, only 50% of patients stop bleeding spontaneously (in contrast to nonvariceal hemorrhage, in which 90% cease spontaneously). The initial bleed is followed by a second phase of an approximately 6-week duration, defined by a high risk of recurrent hemorrhage, with the greatest risk of rebleeding being within the first 48 to 72 hours.

Management

The management of variceal bleeding is outlined in Fig. 82-1. In addition to multiorgan ischemic injury from hypoperfusion, variceal hemorrhage in the setting of cirrhosis predisposes the patient to specific derangements, including hepatic encephalopathy, type 1 hepatorenal syndrome, and systemic infection. These processes contribute to the high mortality associated with variceal bleeding, and therefore, the management should address these issues in addition to achieving hemostasis and hemodynamic stability.

Cardiopulmonary

Fluid resuscitation should be aimed at achieving an isovolemic status because this approach prevents persistent portal hypertension and recurrent variceal bleeding.22 To this end, invasive hemodynamic monitoring with a central venous catheter can be used to guide fluid therapy. In the setting of hypotension that is refractory to fluid resuscitation, a peripheral vasoconstrictor such as norepinephrine should be the vasoactive agent of choice. Agents that have β2-agonist activity, such as dopamine, should be avoided because they potentially could cause splanchnic vasodilation and therefore worsen the variceal bleed. Splanchnic vasoconstrictors such as octreotide and terlipressin (discussed later) can have a beneficial effect on systemic blood pressure by diverting blood away from the splanchnic circulation. Endotracheal intubation for airway protection is critical, especially in the setting of encephalopathy, active hematemesis, or emergency endoscopy.

Infection

Bacterial infections are common after acute variceal hemorrhage, with the common sites being respiratory, ascitic, and urinary. The presence of infection has been associated with failure to control the initial bleed and an increase in the recurrence of rebleeding, likely owing to the induction of a hyperdynamic circulation and increased portal pressure.23,24 A meta-analysis of studies regarding the use of prophylactic antibiotics in variceal bleeding suggests that this intervention significantly increases the number of variceal bleeders who do not acquire an infection and improves short-term mortality.25 Although most of the pertinent studies include a quinolone, the optimal choice and duration of antibiotic therapy have not been defined, and therefore, the choice of empirical antibiotic therapy should be institution-specific. An important consideration is the potential for quinolone resistance in cirrhotic patients who have been on norfloxacin for the prevention of spontaneous bacterial peritonitis (SBP). Prior to initiating antibiotic therapy, blood, urine, and (if indicated) ascitic fluid cultures should be obtained.

Hematologic

In addition to fresh frozen plasma, recombinant activated factor VIIa (rFVIIa) is emerging as a procoagulant that can rapidly correct severe coagulopathy associated with decompensated liver disease and promote hemostasis in variceal hemorrhage.19,20 In a small uncontrolled study of 10 patients,20 a single dose of 80 μg/kg normalized the prothrombin time in all patients within 30 minutes, and immediate hemostasis was obtained in all patients. Although larger studies are needed to confirm these observations, these initial studies suggest that rFVIIa may assume an important role in the management of acute variceal hemorrhage.

Neurologic

In the setting of decompensated liver disease, variceal bleeding could induce or exacerbate hepatic encephalopathy (HE). Therefore, in the presence of decreased mental status during variceal bleeding, HE should be considered as a potential etiology in addition to cerebral hypoperfusion, and empirical lactulose therapy via an NG tube or as an enema should be considered (see Chap. 84). Elevated serum ammonia levels may be useful in supporting the diagnosis of HE.

Hemostatic Therapy

Specific therapies aimed at arresting active bleeding include pharmacotherapy and endoscopic therapy. Multiple studies have proven the increased hemostatic efficacy of combined pharmacotherapy and endoscopic therapy over either treatment alone.

Pharmacotherapy

In the setting of variceal bleeding, pharmacologic agents are aimed at causing splanchnic vasoconstriction and reducing portal hypertension. Empiric pharmacotherapy should be initiated in suspected variceal hemorrhage prior to endoscopic diagnosis and intervention. Selective splanchnic vasoconstriction has the added advantage of diverting blood flow from the splanchnic circulation to the systemic circulation, thereby improving systemic blood pressure. In particular, improved renal perfusion could prevent or ameliorate the hepatorenal syndrome.

The current agent of choice in the United States is the somatostatin analog octreotide. Somatostatin and its analogs inhibit the release of vasodilator hormones such as glucagon, thereby indirectly causing splanchnic vasoconstriction and decreased portal inflow. Although octreotide has a longer half-life than somatostatin, its therapeutic efficacy is obtained only with a continuous infusion; the recommended dose is a 50-μg IV bolus, followed by an infusion of 50 μg/h for 5 days. A recent meta-analysis of trials of octreotide has demonstrated improved control of bleeding compared with other therapies, including the previous agent of choice, vasopressin.26 Furthermore, the adverse extrasplanchnic vasoconstrictive effects observed with vasopressin, such as myocardial and cerebral ischemia, are not observed with octreotide.

More recently, the long-acting vasopressin analog terlipressin has received a favorable recommendation based on European studies, which report fewer side effects than vasopressin and an efficacy similar to that of octreotide and endoscopic sclerotherapy.27 In addition, a systematic review comparing trials of terlipressin with other pharmacotherapies identified terlipressin as the only pharmacologic agent that reduced mortality, therefore suggesting that terlipressin may be the future vasoactive agent of choice in acute variceal bleeding.28 Larger prospective trials are awaited to confirm these encouraging preliminary studies. Terlipressin is currently not available in the United States.

Endoscopic Therapy

Following the administration of pharmacotherapy, emergent endoscopy with therapeutic hemostatic intent is imperative. As outlined in Fig. 82-1, endoscopic evaluation can localize the source of the variceal bleed to an esophageal or gastric varix. This is an important distinction because, while esophageal variceal bleeding is amenable to endoscopic therapy, gastric variceal bleeding may require more aggressive salvage measures, as outlined below.

Endoscopic therapy is based on the interruption of blood flow through the venous collateral system lining the distal esophagus and gastric cardia using either stimulation of thrombosis (e.g., sclerotherapy) or immediate occlusion (e.g., band ligation). The two established forms of endoscopic therapy are endoscopic sclerotherapy (EST) and endoscopic variceal band ligation (EVL). While sclerotherapy involves the intravariceal or paravariceal injection of a sclerosant (e.g., sodium morrhuate), band ligation involves the placement of small bands around varices in the distal 5 cm of the esophagus (Fig. 82-2). A meta-analysis has shown that EVL is superior to EST in initial hemostasis, obliteration of varices, rates of recurrent bleeding, complications, and mortality,29 and EVL is now the endoscopic treatment of choice for acute variceal hemorrhage. However, a technical challenge during use of the band ligator apparatus is the decreased endoscopic field of view in a setting already complicated by active hemorrhage. Therefore, EST may be indicated in the setting of poor initial visualization, followed later by definitive EVL treatment.

Gastric Varices

While the preceding endoscopic interventions are effective in esophageal variceal bleeding, gastric variceal bleeding presents a technical challenge. Gastric varices are located deeper in the submucosa, where EVL and EST are not successful in obtaining sustained hemostasis. Initial studies reporting successful hemostasis with intravariceal injection of cyanoacrylate tissue glue30 and thrombin 31 suggest novel approaches to endoscopic intervention in gastric variceal bleeding. However, these therapies need further validation. In the setting of gastric variceal bleeding, EVL may be attempted to obtain initial hemostasis. However, given the limited success of endoscopic hemostasis, the emergent application of nonendoscopic interventions should be anticipated, which may include balloon tamponade, transjugular intrahepatic portosystemic shunt (TIPS), and surgery.

Complications of Endoscopy

Most complications have been associated with EST, and the advent of variceal band ligation (EVL) has decreased the incidence of complications significantly after therapeutic endoscopy. Following endoscopic intervention, local complications include ulceration, dysmotility, and stricture formation, and regional complications include esophageal perforation and mediastinitis. In addition, both EST and EVL increase the risk of developing portal hypertensive gastropathy (PHG) and its bleeding sequelae because blood is shunted away from the venous system of the gastroesophageal junction to that of the gastric mucosa. Bleeding from PHG is characterized by a diffuse, slow bleed from the gastric mucosa that typically is not amenable to localized endoscopic therapy.

Salvage Therapy for Endoscopically Uncontrolled Bleeding

In the event that hemostasis cannot be achieved by initial endoscopic therapy or there is evidence of rebleeding after initial hemostasis, a repeat trial of endoscopic therapy may be attempted. However, following a second failed endoscopic trial, nonendoscopic interventions need to be implemented emergently. These interventions may include balloon tamponade, TIPS, and surgical therapy.

Balloon Tamponade

Esophageal tamponade tubes (such as the Sengstaken-Blakemore tube) effectively achieve short-term hemostasis by compressing the gastric and distal esophageal mucosa. In most cases, tamponade is effective after inflating only the gastric balloon. Endotracheal intubation is imperative prior to insertion of tamponade tubes to decrease the risk of aspiration. Once the airway is secured, the tube can be passed either nasally or orally to the stomach. The gastric balloon should be inflated only partially (with approximately 30 mL of air) pending radiographic confirmation of correct placement because inadvertent inflation of the balloon in the esophagus can lead to esophageal perforation. Once adequate positioning is ensured, the gastric balloon can be inflated fully according to the manufacturer's recommendations (gastric balloons are inflated to predetermined volumes, whereas the esophageal balloon component is inflated according to pressure). If hemostasis is not achieved by isolated inflation of the gastric balloon, the esophageal balloon can be inflated to 35 mm Hg, a pressure exceeding the intravariceal pressure. Following inflation, traction should be applied to the apparatus at the insertion site to maintain proper positioning. A maximum duration of 48 hours is recommended for variceal compression because prolonged tamponade can lead to esophageal wall ischemia. An NG tube inserted above the esophageal balloon is mandatory to prevent aspiration of oropharyngeal secretions that collect above the inflated apparatus, unless the tube has its own lumen for esophageal suction. A major limitation of tamponade therapy is the high risk of rebleeding following deflation of the balloon. Furthermore, given the serious complications of pulmonary aspiration and esophageal ulceration and perforation, this mode of therapy should be performed by skilled personnel and generally as a temporizing step while planning definitive treatment.

Transjugular Intrahepatic Portosystemic Shunt (TIPS)

Following temporary stabilization with balloon tamponade, further definitive treatment to achieve hemostasis involves the creation of an artificial vascular shunt between the systemic and portal circulation in order to decompress the variceal vasculature. This can be accomplished surgically (discussed below) or via TIPS, which offers a less invasive method for obtaining a vascular shunt. The TIPS consists of an expandable metallic stent placed intrahepatically between portal and hepatic veins using radiologically guided access (see Chap. 101). Traditionally, TIPS has been recommended as secondary prophylactic therapy for variceal bleeding in the setting of mild to moderate liver disease, and advanced cirrhosis has been regarded as a contraindication to TIPS therapy due to increased mortality after TIPS in this setting. However, recent studies with favorable hemostasis and mortality data have supported the use of emergent TIPS as salvage therapy for refractory variceal hemorrhage, even in advanced cirrhosis.32 Therefore, when variceal hemorrhage is refractory to pharmacologic and endoscopic therapy, urgent TIPS therapy should be considered irrespective of the severity of hepatic disease. The hemodynamic benefits of TIPS therapy can be attributed not only to portal decompression and variceal hemostasis but also to increased venous return due to intravascular mobilization of any existing ascites. Additional beneficial effects of TIPS therapy include treatment of refractory ascites and the hepatorenal syndrome.

Surgical Therapy

Since the advent of TIPS as salvage therapy for refractory variceal hemorrhage, there has been a reduction in the need for surgical intervention. However, shunt surgery for portal decompression is indicated for variceal hemostasis in patients with preserved hepatic synthetic function (i.e., Childs-Pugh A disease).33 Esophageal transection with or without devascularization may be another surgical option in massive exsanguination refractory to other interventions.

Shunt operations can be divided into (1) nonselective shunts (e.g., portacaval shunts) that decompress the entire portal system and divert all blood flow away from the portal vein and (2) selective shunts (e.g., distal splenorenal shunt) that compartmentalize the portal tree into a decompressed variceal system and a hypertensive superior mesenteric vein that maintains sinusoidal perfusion. A selective shunt is the preferred operation because portacaval shunts significantly alter vascular anatomy and therefore complicate future liver transplant surgery. In addition, emergency portacaval shunts are associated with a higher rate of thrombosis and shunt failure.

Distal esophageal transection in the setting of massive variceal exsanguination may control bleeding, but mortality remains above 80%. Since the transection does not address the underlying portal hypertension, varices recur after a variable period, and rebleeding should be anticipated. Esophageal transection with devascularization of the gastroesophageal junction (Sugiura procedure) may be considered in patients who have an absolute contraindication to shunt surgery, such as extensive thrombosis in the portal venous circulation involving the splenic, superior mesenteric, and portal veins.

Variceal hemorrhage may occur in noncirrhotic patients. Patients with extrahepatic portal hypertension are better operative shunt candidates than cirrhotics. When large gastric varices accompanied by small or absent esophageal varices are identified, splenic or portal vein thrombosis should be considered as a possible etiology rather than cirrhosis. Splenic and portal vein thrombosis may occur in the setting of acute pancreatitis, pancreatic cancer, abdominal trauma, and hypercoaguable states. It is essential to identify this subset of patients with variceal hemorrhage because splenectomy rather than a portosystemic shunt may be curative. Celiac angiography is diagnostic.

Secondary Prophylaxis

If successful hemostasis is achieved by any of the interventions just discussed, it is essential that secondary prophylaxis to prevent variceal rebleeding is initiated in the ICU. Once the hemodynamic status is stabilized, nonspecific β-blocker therapy (nadolol 20 to 40 mg/d) should be initiated to decrease portal hypertension. This pharmacologic intervention should be coupled with an endoscopic variceal band ligation schedule as defined by the gastroenterologist because the combination of these therapeutic modalities decreases the risk of variceal recurrence and rebleeding.34 In addition, the patient should be evaluated for vascular shunt procedures, including TIPS and surgical shunts. Most important, early referral to a transplant center should be initiated to evaluate the patient for liver transplantation.

In contrast to variceal hemorrhage, nonvariceal hemorrhage presents a favorable prognosis. The improved outcome can be attributed to a greater than 90% spontaneous cessation rate of bleeding and to the absence of a predisposition to multiorgan dysfunction that exists in decompensated cirrhosis. As outlined in Table 82-1, multiple studies pertaining to nonvariceal hemorrhage have identified clinical and endoscopic indicators that increase the risk for continued or recurrent bleeding, and therefore predict an adverse prognosis. The remaining mortality rate of 10% is largely due to rebleeding in patients with these factors. The important role of early therapeutic upper endoscopy in achieving hemostasis, reducing rebleeding, and improving short-term morbidity and mortality has been established10,11 and reiterated in a recent consensus statement addressing the management of nonvariceal bleeding.11a Furthermore, with regard to peptic ulcer disease, endoscopic characterization of high-risk ulcer lesions has led to the development of specific endoscopic therapies that have improved hemostatic efficacy and outcome compared with prior medical therapy.35

The evaluation and management of nonvariceal hemorrhage are outlined in Fig. 82-3. Following initial hemodynamic, pulmonary, and hematologic management, early upper endoscopic evaluation with therapeutic intent should be conducted. Emergent endoscopy should be pursued in the setting of hemodynamic instability that is refractory to fluid resuscitation. Acid-suppression therapy has demonstrated short-term efficacy only after endoscopic hemostasis has been achieved17,18 and therefore should be used as an adjunct to endoscopy. Since peptic ulcer disease accounts for the majority of nonvariceal bleeds and has been the focus of therapeutic developments, the different treatment modalities will be discussed in this context. The management of some nonulcer lesions, including stress-related mucosal damage (SRMD), will follow.

Peptic Ulcer Disease (PUD)

Endoscopic Therapy

Progress in endoscopic diagnosis and therapy has been due largely to major improvements in endoscopic techniques and equipment. A number of hemostatic endoscopic methods have been developed,36 but the two used most commonly in the United States are contact thermal devices and injection therapy with epinephrine.

Contact Thermal Devices

Thermal therapy is designed to produce coagulation and dehydration in the ulcer base surrounding the bleeding vessel, and this results in constriction and destruction of the submucosal feeding vessels supplying the surface artery. The two types of thermal therapy that have gained popularity are the bipolar probe and the heater probe. Bipolar probes heat contacted tissue by passing electricity via tissue water between positive and negative electrodes located at the tip of the probe. Once the contact tissue is fully desiccated, electrical conduction ceases, and deeper tissue coagulation is restricted. On the other hand, the heater probe uses a thermocouple at the end of the probe to generate heat, and this process does not depend on tissue water. Therefore, deeper tissue coagulation is achievable despite desiccation, although this increases the risk of perforation.

Injection Therapy

Injection therapy is aimed at causing vasoconstriction and necrosis of the bleeding vessel and surrounding tissue. Injection with epinephrine (1:10,000 dilution in saline) or absolute alcohol has been shown to be effective in achieving acute hemostasis.37 While epinephrine exerts a vasoconstrictive effect on the vessel, pure ethanol causes dehydration, contraction, and necrosis of the vessel and surrounding tissue. It should be noted that the rebleeding rate is high if epinephrine injection therapy is performed in isolation, and therefore, it should be combined with thermal coagulation therapy. The technique used in injection therapy involves the injection of the agent via an injector catheter in four quadrants within 2 to 3 mm of the active bleeding point in the ulcer base. While the volume of total epinephrine administered can range from 5 to 10 mL, the total volume of ethanol should not exceed 1 mL because extensive ulceration may occur.

Endoscopic Evaluation and Treatment in Pud

As outlined in Fig. 82-3, the endoscopic management of PUD is guided by the characteristics of the ulcer, which defines the lesion as exhibiting “major” or “minor” stigmata of ulcer hemorrhage. The major stigmata consist of (1) active bleeding, (2) a visible, nonbleeding vessel, and (3) an adherent clot, and these three lesions are associated with a high risk of rebleeding and therefore increased short-term mortality. The stigmata associated with a low risk of rebleeding include (1) a clean base and (2) a flat/pigmented spot, and these lesions have a favorable prognosis. Specific endoscopic treatment guidelines have been developed by the ASGE for each of these lesions14 based on multiple studies addressing the response of these lesions to thermal and injection therapy. These guidelines and subsequent studies have led to the following endoscopic treatment recommendations: (1) combination therapy with injection and thermal coagulation is superior to monotherapy either in active bleeding38,39 (Fig. 82-4) or for an adherent clot,40 (2) monotherapy with thermal coagulation is indicated for a nonbleeding visible vessel, and (3) no endoscopic therapy is required for low-risk lesions, which include a clean base and a flat/pigmented spot.

In anticipation of initiating secondary prophylaxis for PUD, the visualization of a duodenal or gastric ulcer should prompt the endoscopist to obtain a gastric antral mucosal biopsy to test for Helicobacter pylori. The gastric biopsy may be difficult to perform during the acute phase of the bleeding, and the presence of blood or antiulcer medications may interfere with a biopsy urease test. In the absence of a gastric biopsy, a venous blood sample should be tested for H. pylori serology, or stool or breath testing should be performed. Furthermore, a negative gastric antral mucosal biopsy should be confirmed with a serologic test, especially if antiulcer therapy has been initiated prior to the biopsy.

Pharmacotherapy

While intravenous histamine receptor antagonists (H2RAs) are used commonly in the initial management of PUD bleeding, there is no evidence that this intervention improves short-term outcomes, such as rebleeding rates or transfusion requirements.41 The concept that more profound acid suppression with proton pump inhibitors (PPIs) will improve short-term outcomes has been investigated, and most studies have shown short-term efficacy of PPIs when administered after successful endoscopic hemostasis.42,43 Therefore, acid-suppressive therapy in the form of PPIs should be administered as an adjunct to endoscopic therapy and should not be used in lieu of an endoscopic evaluation. Studies comparing intravenous and oral PPI therapy are awaited; however, intravenous PPI therapy may ensure improved bioavailability in the setting of a bleeding gastrointestinal tract. The improved hemostatic efficacy of PPI over H2RA therapy may be due to the superior ability of PPI therapy to maintain a gastric pH above 6.0 and therefore protect an ulcer clot from fibrinolysis.44

Several studies have evaluated the efficacy of the splanchnic vasoconstrictors somatostatin and octreotide in the treatment of acute nonvariceal hemorrhage. A meta-analysis of these studies suggests that somatostatin and octreotide are associated with a reduced risk of continued bleeding and rebleeding in PUD.45 Therefore, in addition to acid-suppressive therapy, these agents also may be considered as pharmacologic adjuncts to endoscopic therapy.

Angiography

In most cases of nonvariceal hemorrhage, endoscopic evaluation is able to visualize the bleeding lesion and deliver effective hemostatic therapy. However, in a minority of patients, the bleeding source is not visualized by endoscopy, thereby necessitating angiographic localization. Also, angiography offers the option of hemostatic therapy using arterial vasoconstrictors or embolization; however, this generally is reserved for patients who are poor surgical candidates or for the control of bleeding in an unstable patient awaiting surgery.

Localization

As mentioned previously, angiography can successfully localize brisk UGI bleeding (rate >0.5 mL/min) in 75% of cases,46 with most bleeding episodes (85%) originating from a branch of the left gastric artery. The right gastric and short gastric arteries account for the remainder of the sources.

Hemostatic Therapy

UGI arterial bleeding can be controlled by the selective arterial infusion of vasoconstrictors such as vasopressin or embolization of particulate matter. Most studies indicate that selective intraarterial vasopressin is more effective than a peripheral intravenous infusion in achieving hemostasis.47 Since the vasoconstrictive action of vasopressin is more pronounced on terminal blood vessels such as arterioles, venules, and capillaries, this therapy is more effective in controlling bleeding from such vessels, as in diffuse hemorrhagic gastritis, rather than a duodenal ulcer bleed originating from a large gastroduodenal artery. In the setting of gastric bleeding, therapy appears to be equally effective when administered via the left gastric or celiac artery.48A usual therapeutic vasopressin dose is 0.2 unit/min, with a recommended maximum dose of 0.4 unit/min. If hemostasis is achieved, the infusion is continued in the ICU for 24 to 36 hours and then tapered over 24 hours. Rebleeding after cessation of the infusion is a concern48 and can be treated with repeat vasopressin treatment or embolization.

Embolization therapy uses various substances, most frequently a gelatin sponge (Gelfoam), to selectively embolize the bleeding vessel. Since this technique carries the risk of causing bowel wall ischemia and infarction due to nonspecific embolization, a target vessel must be accessible for selective catheterization of the bleeding site. Necrosis of the stomach, duodenum, gallbladder, liver, and spleen has been documented following nonspecific embolotherapy. Since the duodenum has a dual blood supply from the celiac artery and the superior mesenteric artery, spontaneous infarction of the duodenum is rare. The patient with advanced atherosclerotic vascular disease or with prior gastric surgery involving ligation of collateral vessels is at greater risk of infarction due to a compromised collateral circulation. In view of the higher morbidity associated with embolization therapy, it should be used only after unsuccessful intraarterial vasopressin therapy. Furthermore, embolization therapy should not be followed immediately by vasopressin therapy because this may compromise the collateral circulation, resulting in tissue infarction.

Surgical Therapy

Surgical intervention for bleeding peptic ulcers should be considered in two situations. First, surgery is used to control life-threatening hemorrhage that is refractory to medical and endoscopic intervention. In fact, in patients who have a low operative risk and who have been stabilized effectively to allow surgery, angiographic therapy should not be attempted. Second, surgery should be considered in the patient in whom medical management has failed to heal or prevent recurrence of peptic ulceration, particularly if there have been previous complications attributable to PUD, such as bleeding. The patient who has recurrent hemorrhage owing to noncompliance with maintenance ulcer therapy should be considered for elective surgical therapy once bleeding has stopped. It should be emphasized that surgical morbidity and mortality are greatly reduced when the surgeon operates electively in the nonbleeding patient. Therefore, successful initial hemostasis using endoscopic and pharmacologic therapy is preferable prior to surgical intervention. In the setting of hemorrhage refractory to nonsurgical intervention, stabilization of cardiopulmonary status and optimization of hematologic parameters prepares the patient for emergent surgery and improves postoperative outcome.

The choice of surgical procedure depends on the location of the ulcer and on the stability of the patient. In the patient with an actively bleeding duodenal ulcer undergoing an emergent operation, the bleeding point of the ulcer will be oversewn and truncal vagotomy and pyloroplasty performed. Vagotomy and antrectomy may be considered if the patient has been stabilized adequately. In the setting of exsanguinating hemorrhage, gastric resection may be considered, but this procedure carries a high mortality of approximately 50%.49 Selective vagotomy with either pyloroplasty or antrectomy, an option in the elective situation, is not advisable in an unstable patient.

Gastric ulcer bleeding is treated with the same approach as a bleeding duodenal ulcer, except that resection is recommended if the situation permits. Partial gastrectomy carries a slightly lower mortality in the setting of gastric ulcer bleeding than when performed for a bleeding duodenal ulcer.50 Resection for an actively bleeding gastric carcinoma is recommended only when performed electively because it is a prolonged procedure that may be unsuitable for an unstable patient.

Following surgical intervention for bleeding peptic ulcers, mortality approaches 30%, with postoperative wound infection being the major complication.

Secondary Prophylaxis

Once successful hemostasis is achieved, secondary prophylaxis is initiated to prevent recurrent ulcer bleeding, especially following nonsurgical hemostatic therapy. If histologic or nonhistologic evaluation for H. pylori is positive, appropriate treatment should be initiated because this reduces the long-term (1-year) rate of rebleeding from gastric or duodenal ulcers.51 In addition, long-term acid suppressive therapy with oral H2RAs or PPIs is indicated, and nonsteroidal anti-inflammatory drugs (NSAIDs) should be avoided.

Mallory-Weiss Tear

Mallory-Weiss tear generally is a self-limited cause of nonvariceal bleeding, rarely requiring more than supportive intervention. However, patients with portal hypertension are at increased risk of massive bleeding from Mallory-Weiss tears compared with those with normal portal pressures.52 In the rare instance of continued bleeding from a Mallory-Weiss tear in a patient without portal hypertension, endoscopic therapy with either thermal coagulation or injection therapy should be attempted prior to surgical oversewing of the lesion. In the presence of portal hypertension, thermal coagulation may worsen the bleeding; therefore, band ligation or sclerotherapy should be performed. Following hemostasis, acid-suppression therapy with H2RAs or PPIs may be given as adjunctive therapy to accelerate healing.

Dieulafoy's Lesion

A Dieulafoy's lesion is a dilated aberrant submucosal vessel of unclear etiology that erodes the overlying epithelium in the absence of a primary ulcer. It is usually located along the high lesser curvature of the stomach near the gastroesophageal junction, although it has been found in all areas of the GI tract, including the esophagus and duodenum. Massive bleeding can occur when the eroding submucosal vessel is an artery. The endoscopic treatment of choice is a combination of epinephrine injection therapy and thermal coagulation.53 Also, endoscopic band ligation has been used successfully to achieve hemostasis in bleeding Dieulafoy's lesions.54 The risk of rebleeding after endoscopic therapy remains high (up to 40% in some reports) owing to the usually large size of the underlying artery. In the event of rebleeding, repeat endoscopic intervention may be attempted, following which surgical wedge resection of the lesion should be performed to achieve permanent hemostasis.

Stress-Related Mucosal Damage (SRMD)

SRMD, also referred to as stress ulcers or stress-related erosive syndrome (SRES), is the result of multiple-organ-system failure in the critically ill patient. The incidence of hemorrhage due to SRMD appears to be decreasing, probably as a result of significant advances in the intensive care management of the critically ill patient, including optimization of hemodynamic status and tissue oxygenation,55 and the early initiation of stress ulcer prophylaxis. However, in the event of SRMD-induced hemorrhage, the mortality rate is greater than 30% owing to the difficulty in controlling such bleeding and the poor prognosis of the underlying disease. With regard to etiology, gastric mucosal ischemia secondary to systemic (and splanchnic) hypoperfusion is considered to be the major inciting factor, with acid and pepsin assuming minor roles. Of note, acid and pepsin secretion are normal to low in most critically ill patients, and increased gastric acidity is observed only in patients exhibiting Cushing's ulcers related to central nervous system (CNS) trauma or infection.

Bleeding from SRMD may be overt and significant, resulting in hemorrhage and hemodynamic compromise, or occult and minimal, detectable only by Gastroccult testing of the gastric contents. Although occult bleeding due to SRMD may occur frequently in critically ill patients, it is of little clinical significance because few of these patients progress to overt bleeding. Multiple studies have attempted to assess the relative importance of the underlying disease processes and biochemical abnormalities in inducing SRMD.56,57 Two major risk factors identified are coagulopathy and mechanical ventilation for greater than 48 hours.56 Other suggested risk factors include sepsis, hypotensive shock, acidosis, peritonitis, extensive burns, hepatic failure, and renal failure, with multiple risk factors having an additive effect on the probability of SRMD.

Endoscopically, SRMD may appear as multiple shallow erosions or submucosal hemorrhage during the early stages. After the first several days of the ICU course, SRMD lesions are characterized by multiple, deeper, acute ulcerations, predominantly in the gastric lesser curvature or fundus, and these lesions can erode into the submucosa, causing massive hemorrhage. Bleeding usually manifests as oozing of blood from the margins of these lesions. However, submucosal penetration can cause hemorrhage from a major artery, with the typical endoscopic appearance of an ulcer with a visible vessel.

Therapy

The mainstay of therapy for SRMD is supportive, with an attempt to reverse the underlying precipitating factors. Acid suppression in the form of intravenous H2RAs or PPIs may be used as adjunctive therapy to endoscopic or angiographic intervention. The role of endoscopic therapy in SRMD may be limited because the lesions usually are diffuse and not amenable to directed therapy. However, in the setting of a single dominant lesion or a few bleeding lesions, endoscopic therapy may achieve successful hemostasis in 90% of such cases.58 Therapeutic angiography is recommended for bleeding that is refractory to endoscopic therapy. Both intraarterial vasopressin therapy59 and embolotherapy60 are equally successful at controlling hemorrhage without major ischemic complications owing to the rich collateral blood supply of the gastric mucosa. Since the left gastric artery is the source in most cases of SRMD-induced bleeding, this vessel is a convenient target for embolization therapy.

Surgery usually should be avoided because a near-total gastrectomy is required in most instances, and mortality exceeds 50%. Gastrectomy should be undertaken only when massive hemorrhage persists despite nonsurgical therapy in a viable patient with treatable medical problems.

Prophylaxis

Since hemorrhage from SRMD presents a therapeutic challenge and carries a high mortality, much attention has been given to prophylactic therapy. Despite the existence of SRMD in the setting of low or normal acid secretion, prophylaxis has been directed toward acid suppression or neutralization. The superior efficacy of intravenous H2RAs compared with sucralfate in preventing SRMD has been demonstrated,61and therefore, H2RAs are preferred. Furthermore, prior concerns regarding the increased incidence of nosocomial pneumonia with acid-suppressive therapy has not been observed in subsequent studies. More recently, oral and intravenous PPIs have been added to the regimen of acid-suppressive therapy in the ICU and should assume increasing importance in the prophylaxis of SRMD.

In addition to pharmacologic therapy, adequate nutritional support and, in particular, enteral nutrition have been shown to decrease the incidence of SRMD.62,63 The prophylactic effect of enteral nutrition is not mediated by an increase in gastric pH and instead may involve an increase in gastric epithelial energy stores, which, in turn, maintain epithelial integrity and prevent necrosis and ulceration. Therefore, the initiation of adequate nutrition support in the critically ill patient, preferably via the enteral route, may play a prophylactic role against SRMD. In the absence of a functional gastrointestinal tract, total parenteral nutrition (TPN), which has demonstrated a protective effect against SRMD, can be considered, although the net effect may be harmful62 (see Chap. 11).

LGI bleeding is defined as bleeding originating from a source distal to the ligament of Treitz. Hematochezia is the common presenting sign of LGI bleeding, and the two frequent LGI sources are diverticulosis and angiodysplasia. However, in the patient with hematochezia and hemodynamic compromise, a briskly bleeding UGI source should be included in the differential diagnosis. Studies have indicated that as many as 10% of patients suspected initially to have LGI bleeding ultimately are found to have a UGI source.64

In the past decade, technologic advances in endoscopy have greatly improved the diagnostic and therapeutic utility of colonoscopy in LGI bleeding. Traditionally, colonoscopic evaluation in the setting of LGI bleeding was delayed owing to the need for bowel cleansing, the lack of proven diagnostic and therapeutic efficacy, and the fear of increased procedural risks. However, in recent years, several studies have suggested that emergency therapeutic colonoscopy for acute LGI hemorrhage facilitates early control of bleeding, reduces rebleeding and surgical intervention rates, and improves short-term morbidity and mortality.12,13 Based on these studies, the practice guidelines of the ASGE have emphasized the diagnostic and therapeutic utility of colonoscopy in the early management of LGI bleeding.15

The evaluation and management of LGI hemorrhage is outlined in Fig. 82-5. Most patients who experience severe LGI bleeding are elderly, with an average age of 65, and have comorbidities, including cardiac and respiratory disease. Therefore, prompt resuscitative measures aimed at multiple-organ support should be initiated. As mentioned previously, in the setting of hematochezia and hypotension, a UGI source should be considered in the differential diagnosis. The presence of a positive NG aspirate or historical risk factors for UGI bleeding should prompt an emergent EGD. A negative NG aspirate and a clinical suspicion for an LGI source should lead to a diagnostic and potentially therapeutic colonoscopy following a rapid oral purge with polyethene glycol (PEG). In the presence of massive LGI bleeding, immediate angiographic or surgical intervention without endoscopic evaluation is indicated.

The different diagnostic and treatment modalities used in the management of LGI hemorrhage (see Fig. 82-5) are outlined below. In contrast to UGI hemorrhage, a role for adjunctive pharmacotherapy has not been established in LGI bleeding.

Colonoscopy

Following the exclusion of a UGI source, an emergent colonoscopy after a rapid oral purge is the initial examination of choice for diagnosis and treatment. Studies have indicated that following colonic cleansing, colonoscopy offers a higher diagnostic yield and a lower complication rate than the traditional angiographic approach.64,65 An accepted bowel cleansing protocol is 4 L polyethylene glycol (golytely) given orally or via an NG tube over 2 hours.64Metoclopramide (10 mg) is administered at the start of the purge to facilitate intestinal transit and to minimize the risk of emesis. This bowel preparation regimen has not been associated with an increase in the rate of bleeding or rebleeding.64

The overall diagnostic yield of emergent colonoscopy in LGI bleeding is 69% to 80%.64,66 Diverticular bleeding and angiodysplasia are the most common findings in the majority of large series, with colitis (ischemic, inflammatory, or radiation-induced), neoplasia, and anorectal disease being some of the minor etiologies. If an adequate colonoscopic evaluation does not reveal a bleeding source, and an upper endoscopy is negative, a small intestinal etiology should be considered. Intubation of the terminal ileum at the time of colonoscopy may be useful because fresh blood emanating from the ileum may be indicative of small intestinal bleeding. Small bowel evaluation with push enteroscopy, capsule endoscopy, or enteroclysis can be initiated once hemostasis is achieved spontaneously or with nonendoscopic methods. A further discussion regarding the evaluation of small intestinal bleeding will follow.

Endoscopic Therapy

The hemostatic techniques used in therapeutic colonoscopy are similar to those used in therapeutic upper endoscopy. Both thermal coagulation and injection therapy with epinephrine have been used successfully to obtain hemostasis in acute LGI bleeding. With further expertise and advances in endoscopic technology, this relatively new field of therapeutic colonoscopy is expected to evolve.

Diverticular Bleeding

The endoscopic visualization of a visible vessel or pigmented protuberance within a diverticular segment identifies patients who are at high risk for persistent or recurrent diverticular bleeding.67 In this setting, thermal coagulation or injection therapy with epinephrine may be used individually or together to achieve successful hemostasis (Fig. 82-6). Studies have suggested that endoscopic hemostasis may prevent recurrent diverticular bleeding and the need for hemicolectomy.12,15 Therefore, endoscopic therapy offers a viable long-term alternative to surgical therapy for diverticular bleeding in the elderly patient, who may not be an ideal candidate for surgical intervention. However, massive diverticular bleeding may not be amenable to endoscopic therapy because of poor endoscopic visualization or failed endoscopic therapy, thereby necessitating angiographic or surgical therapy.

Angiodysplasia

Bleeding from angiodysplastic lesions frequently is responsive to endoscopic therapy. These lesions often are located in the cecum and right colon and represent acquired arteriovenous malformations. Successful hemostasis can be achieved with both injection therapy and thermal coagulation.68 The periphery of the lesion should be treated before the center in order to obliterate the feeder vessels. With respect to thermal coagulation, the recommended power settings are lower than those used for a bleeding peptic ulcer owing to the increased risk of perforation in the right colon.

Additional lesions responsive to thermal and injection therapy include postpolypectomy sites, radiation colitis, and anorectal sources. Noncontact modalities such as the argon-plasma coagulator (APC) have been used effectively in the management of radiation colitis and postpolypectomy bleeding. In addition, a band ligation technique similar to esophageal variceal band ligation can be used to treat bleeding internal hemorrhoids.69

Radionuclide Studies

In the event of a negative colonoscopy, a radionuclide scan using 99mTc-pertechnetate-labeled red blood cells is used frequently to localize the LGI bleeding site prior to subsequent angiographic evaluation. The nuclear scan offers the ability to detect rates of bleeding as low as 0.1 mL/min, and the 48-hour stability of the tagged red blood cells allows repeated imaging during this time period. However, the high sensitivity of radionuclide imaging is offset by its low specificity compared with a positive endoscopic or angiographic examination.70 A positive radionuclide study localizes the bleeding only to an area of the abdomen and cannot define precisely the mucosal location of the bleeding site. Therefore, a positive scan should be used to direct attention to specific sites of the GI tract that can be examined subsequently by angiography or a repeat endoscopy. Furthermore, surgical intervention should not be based on the results of a radionuclide scan alone but should be guided by accurate angiographic or endoscopic evaluation.

Angiographic Therapy

Angiography with therapeutic intent is the appropriate treatment modality in the setting of massive bleeding that precludes colonoscopy or after a nondiagnostic colonoscopy. The overall diagnostic yield of angiography ranges from 40% to 78%, with diverticular disease and angiodysplasia being the most common findings.71 In a stable patient, a radionuclide scan is performed often for initial localization prior to angiographic evaluation because the nuclear scan is a more sensitive examination and detects slower rates of bleeding. In the absence of localization via nuclear imaging, the superior mesenteric artery (SMA) is examined first because most diverticular and angiodysplastic bleeds occur in bowel supplied by this artery.72 If the evaluation of the SMA is negative, the inferior mesenteric and celiac vessels are studied.

The intermittent nature of LGI bleeding in many cases presents a problem for angiographic evaluation and treatment because active bleeding at the time of dye injection is required for a positive examination. Initial angiographic hemostasis using intraarterial vasopressin or embolization ranges from 60% to 100%, although recurrent bleeding may be as high as 50%, especially following vasopressin therapy.73 Intraarterial vasoconstrictive therapy with vasopressin is used predominantly in diverticular bleeding and angiodysplasia and is associated with a major complication rate of 10% to 20%, including arrhythmias, ischemia, and pulmonary edema. Transcatheter embolization therapy with various agents, including gelatin sponge and microcoils, may be a more definitive means of controlling hemorrhage but is associated with a risk of intestinal infarction as high as 20%. Attempts to reduce this risk have included the application of a highly selective catheterization technique and the use of a relatively distal site for embolization with temporary occluding agents.74

In the event that therapeutic angiography does not achieve permanent hemostasis, emergent surgical therapy is indicated. However, initial angiographic therapy in the actively bleeding patient may achieve temporary hemostasis and hemodynamic stability and thereby allow surgical intervention in a controlled setting with an improved operative mortality.73

Surgical Therapy

In the patient presenting with exsanguinating LGI hemorrhage, colonoscopy and angiography should be deferred, and an emergent subtotal colectomy should be performed. In all other cases, surgical therapy should be reserved for hemorrhage that is refractory to nonsurgical interventions. Furthermore, preoperative localization of bleeding is crucial to avoid extensive surgical resection and to ensure that the bleeding is truly arising from the LGI tract.

Preoperative Localization

The role of angiography and colonoscopy in identifying the bleeding site was outlined previously. In addition to angiography and colonoscopy, exploratory laparotomy with intraoperative endoscopy can be used to localize the bleeding source, especially in the small intestine. Intraoperative endoscopy can be performed with oral, rectal, or enterotomy introduction of the endoscope. Following incision of the abdominal wall and exposure of small bowel, the endoscope can be introduced orally. The endoscope, generally a pediatric colonoscope, can be advanced easily to the ligament of Treitz. Subsequently, the entire small bowel is examined by pleating the small bowel over the colonoscope. The endoscopist must limit the amount of air insufflation because excessive distention of the bowel will result in prolonged postoperative ileus. Following inspection of the small bowel to the ileocecal valve, a second inspection is performed as the colonoscope is withdrawn slowly. The surgeon assists the examination by carefully inspecting the serosal side of the bowel for abnormalities, such as a transilluminated angiodysplastic lesion. Alternatively, a sterilized colonoscope may be placed through an enterotomy site in the small bowel and then can be passed proximally and distally into the small bowel to facilitate examination. This approach carries a risk of contamination of the exposed peritoneum. A collaborative effort between the endoscopist and the surgeon is essential for the efficient and safe performance of intraoperative endoscopy.

Once the source of bleeding is localized, a segmental colectomy involving the bleeding lesion can be performed. In a patient with extensive diverticular disease and a localized diverticular bleeding site, a segmental resection eradicating the bleeding site is adequate without the need to resect segments involving nonbleeding diverticula.73 With respect to angiodysplasia, the presence of cecal angiodysplasia should alert the surgeon to the possibility of angiodysplasia in the distal terminal ileum. It should be noted that angiography that demonstrates cecal angiodysplasia may fail to identify a similar small bowel lesion, and therefore, intraoperative small bowel endoscopy should be used. In addition, when a right hemicolectomy for suspected angiodysplasia is undertaken, resection of the distal 30 to 60 cm of the terminal ileum should be considered.

A subtotal colectomy is indicated for exsanguinating hemorrhage or persistent hemorrhage without an identifiable site of bleeding and involves colonic resection from the cecum to proximal rectum with an ileoproctostomy. This procedure is associated with a high morbidity and mortality, but rebleeding rates are extremely low.75 Blind segmental resection is contraindicated because this procedure is associated with excessive rates of rebleeding, morbidity, and mortality.75

Occasionally, a patient presenting with acute gastrointestinal bleeding undergoes a nondiagnostic evaluation with endoscopy, radionuclide imaging, and angiography but clinically stabilizes owing to intermittent or permanent cessation of bleeding. In such a situation, emergent surgical exploration and intraoperative endoscopy may not be indicated. Instead, further diagnostic evaluation should be directed specifically at the small bowel, which is the likely source of obscure bleeding in many cases, with angiodysplasia being the most common lesion. Recent advances in endoscopic technology have enabled the gastroenterologist to visualize more distal portions of the small bowel compared with traditional upper endoscopy. These novel endoscopic modalities will be reviewed and will be followed by a discussion of specific lesions associated with obscure bleeding.

Novel Endoscopic Modalities

Push Enteroscopy

An orally inserted adult/pediatric colonoscope or special enteroscope is passed with or without an overtube as far as possible beyond the ligament of Treitz and allows visualization of the proximal 60 cm of jejunum. The diagnostic yield of this procedure is as high as 50%, with angiodysplasia being the most common lesion.76 In addition, this technique allows biopsy or therapy of visualized lesions.

Sonde Enteroscopy

This technique should be considered in the setting of a negative push enteroscopy and allows visualization of small bowel distal to the limits of push enteroscopy. During this procedure, a long flexible fiberoptic enteroscope without controls is propelled passively by intestinal peristalsis, allowing a view of the distal small bowel beyond the visualization by push enteroscopy. The endoscopic examination is performed during withdrawal of the instrument. In contrast to push enteroscopy, this technique does not allow biopsy or therapy of visualized lesions, and additional limitations include high equipment costs and a long procedure time.

Wireless Capsule Endoscopy

This technique has been introduced into clinical practice recently and provides a noninvasive method of examining the entire small bowel via peristaltic propulsion of the endoscopic capsule. The patient swallows a capsule that produces approximately 50,000 images while it traverses the small bowel over 12 to 15 hours. Recent versions of this technique are able to approximate the location of the bleeding lesion within the small bowel. A limitation similar to that for the sonde enteroscopy is the fact that tissue sampling or therapeutic intervention cannot be performed. Experience with this technique is growing rapidly, but extensive studies are yet to be performed.

In addition to the preceding endoscopic modalities, enteroclysis radiography can be considered for the evaluation of potential small bowel bleeding sources, although the yield of this test is only about 10%.77 This study is a double-contrast study performed by passing a tube into the proximal small bowel and injecting barium, methylcellulose, and air. Despite the low sensitivity of this study, it is considered superior to standard imaging using small bowel followthrough.

Lesions Associated with Obscure Bleeding

Small intestinal angiodysplasia accounts for the majority of small bowel lesions associated with obscure bleeding. An increased incidence of small bowel angiodyplasia has been reported in patients with end-stage renal disease (ESRD), Von Willebrand disease, and Osler-Weber-Rendu (OWR) syndrome. In the setting of proximal lesions located approximately in the first 60 cm of jejunum, therapeutic push enteroscopy may achieve permanent endoscopic hemostasis. However, more distal lesions are not amenable to endoscopic therapy. In select patient populations, including those with ESRD, Von Willebrand disease, and OWR syndrome, hormonal therapy with estrogen, with or without progesterone, may be useful in controlling bleeding from angiodysplasia.78 In patients refractory to endoscopic and medical therapy, chronic iron supplementation and periodic transfusions are indicated.

Hemobilia is bleeding from the liver, bile ducts, or pancreas and is characterized by blood emanating from the ampulla of Vater. Hepatic hemobilia usually results from blunt or sharp trauma to the liver. A hepatic artery aneurysm that erodes into the right hepatic or common bile duct produces melena and occasionally presents with right upper quadrant abdominal pain and jaundice. Pancreatic hemobilia is even less common and usually results from pancreatitis-induced pseudoaneurysm formation in the splenic artery. Acquired splenic artery aneurysms may erode into the pancreas, resulting in hemobilia without pancreatitis.

Hemobilia should be suspected when melena occurs in conjunction with jaundice or after blunt trauma or acute pancreatitis. Following a negative forward-viewing upper endoscopy, the endoscopist should examine the duodenal papilla using a side-viewing duodenoscope. Active bleeding or a clot emanating from the papilla may be seen. Alternatively, angiography may reveal active bleeding or associated aneurysms in the hepatic or splenic artery. Angiographic therapy may provide temporary control of hematobilia,79 but generally definitive surgery is required.

Aortoenteric fistula is a rare development following abdominal vascular surgery involving placement of a synthetic graft. The fistula arises commonly from the proximal anastomosis of the graft and communicates with the fourth portion of the duodenum. Graft infection generally is present and likely plays a role in the pathogenesis of the fistula. Bleeding from aortoenteric fistulas is typically intermittent and profuse. The evaluation of a suspected aortoduodenal fistula should begin with endoscopy. The endoscopist must examine the fourth portion of the duodenum, where bleeding or the graft itself may be seen. If endoscopy fails to identify a fistula, angiography should be performed if clinical suspicion is high. Surgical correction of the fistula, including removal of the graft, is necessary to prevent potential exsanguination.

Meckel's diverticulum should be considered in younger patients presenting with massive bleeding. The diagnosis often is made by a 99mTc-pertechnetate scan, which has a sensitivity of 75%.80 Surgical resection of an identified Meckel's diverticulum provides definitive therapy.