Gastrointestinal Bleeding
Margaret Shuhart, M.D. , Kris Kowdley, M.D., and Bill Neighbor
, M.D.



Gastrointestinal (GI) bleeding is among the most common gastrointestinal disorders. In this chapter we review the many causes of both upper and lower GI bleeding, and the diagnostic and therapeutic approaches in patients with GI bleeding. General guidelines for the initial management and appropriate referral to a gastroenterologist for patients presenting with GI bleeding are discussed in this chapter.

MeSH search strings used in this review included GASTROINTESTINAL BLEEDING, UPPER GASTROINTESTINAL BLEEDING, LOWER GASTROINTESTINAL BLEEDING, OCCULT GASTROINTESTINAL BLEEDING, AND OBSCURE GASTROINTESTINAL BLEEDING. MeSH terms also included specific causes of GI bleeding as outlined in Sections 3.1 and 4.1. These terms were used in a search of Medline for the years 1990 to 2002. Certain articles published prior to 1990 were also included if they were landmark studies or illustrative case reports.


2.1. Initial assessment, resuscitation and triage

Patients with GI bleeding may have different clinical presentations ranging from
hematemesis or hematochezia with hemodynamic instability, to melena or rectal bleeding without hemodynamic compromise. Finally, patients may have chronic GI bleeding with asymptomatic iron-deficiency anemia, or hemoccult-positive stool on screening for colorectal cancer.

The severity of illness at presentation dictates the initial patient assessment and resuscitative efforts. Patients presenting to the emergency room with hemodynamic instability require rapid clinical assessment, intravenous access with at least two large-bore lines, nasogastric tube placement, determination of hematocrit and coagulation studies, and type and cross for blood products. Patients with altered mental status should undergo endotracheal intubation for airway protection. Emergent evaluation by a gastroenterologist should be requested. The patient should be stabilized before proceeding to endoscopy, unless there is massive bleeding and no response to resuscitation.

During the initial assessment, it is important to determine whether the bleeding is from an upper or lower GI source (proximal or distal to the ligament of Treitz, respectively) as this will guide the diagnostic approach. Distinguishing between an upper and lower source of bleeding is usually relatively straightforward. However, approximately 5%-10% of patients who present with hematochezia are bleeding from an upper GI source, and some may not have blood in a gastric aspirate. If there is uncertainty about the presence of an upper GI bleeding source, such as when the gastric aspirate is not bile-stained, patients with hematochezia and hemodynamic compromise should undergo upper endoscopy before evaluation of the lower GI tract.

Admission to the hospital is required for most patients presenting with GI bleeding. Those who present with frank hypotension or who have evidence for ongoing bleeding require monitoring in an intensive care unit and urgent endoscopic evaluation. Those who present with mild or no orthostasis, have no evidence for continued bleeding, but have had a significant drop in hematocrit are generally hospitalized on a medical/surgical floor. In general, young patients with self-limited GI bleeding who present without orthostasis or hemodynamic instability and who have no significant comorbid conditions may be managed as outpatients. Early endoscopy may also assist in the identification of patients with a low rebleeding risk who can be managed as outpatients [Rockall, 1990].

2.2. Diagnosis

2.2.1. History and physical

During the initial stabilization and evaluation, a complete history and physical should be performed. First, a detailed history of current and past GI bleeding should be elicited. Patients with upper GI bleeding should be questioned about known ulcer disease, liver disease, malignancy, abdominal surgery or bleeding disorder, weight loss, and alcohol, aspirin or non-steroidal antiinflammatory drug (NSAID) use. A history of antecedent retching suggests a Mallory-Weiss tear. Patients with suspected lower GI bleeding should also be asked about hemorrhoids, associated diarrhea, change in bowel habits, personal or family history of inflammatory bowel disease, and a history of radiation therapy. A family history of GI disorders, malignancy or bleeding disorders should also be obtained. Physical examination should include orthostatic blood pressure and pulse even if the patient appears stable and has no symptoms of orthostasis. Although the physical examination seldom leads to a specific diagnosis, certain findings may provide helpful clues. Such findings include cutaneous stigmata of liver disease, splenomegaly or ascites, abdominal tenderness, an abdominal mass or lymphadenopathy, and cutaneous or mucocutaneous manifestations of systemic disorders associated with GI bleeding (Table 1).

2.2.2. Diagnostic studies Upper GI endoscopy

Gastrointestinal endoscopy is the preferred diagnostic modality in patients with upper GI bleeding. Barium studies generally have limited sensitivity and may interfere with subsequent endoscopy or angiography. Furthermore, many abnormalities seen on barium study will require endoscopic evaluation. In addition to its superior sensitivity, other advantages of endoscopy include the ability to obtain biopsies for an accurate histologic diagnosis, to determine the risk of rebleeding (thus influencing decisions such as level of hospital care, when to resume feeding, and the duration of hospitalization), and to provide endoscopic therapy. Endoscopic therapy of bleeding esophageal varices and peptic ulcers with a high risk of rebleeding can reduce the incidence of further bleeding and the need for surgery, and improve survival [Westaby,1985; Williams,1993; Thomopoulos,1997].

Early, judicious use of endoscopy can be of benefit from the standpoint of diagnosis, therapy, and prognosis. Patients with severe upper GI bleeding should have an upper endoscopy, or
esophagogastroduodenoscopy (EGD), performed as soon as they are stable. If resuscitation is unsuccessful, an attempt at EGD is still warranted as it might provide useful information. Patients in whom endoscopy cannot be performed due to torrential bleeding should be considered for laparotomy, with or without prior mesenteric angiography. Such decisions must be made in consultation with surgery, gastroenterology and interventional radiology. Early endoscopy is also helpful in patients with milder degrees of upper GI bleeding. The endoscopic findings in such cases can help predict the risk of rebleeding. This information can be helpful in decision-making regarding continued hospitalization versus early discharge. Variables predictive of mortality in patients with upper GI bleeding are listed in (Table 2) [Sanders,2002]. Patients who are young and who present with trivial bleeding may not require endoscopic evaluation, particularly if there is a history of antecedent vomiting, suggesting the presence of a Mallory-Weiss tear. Guidelines for the evaluation of patients with upper GI bleeding are provided in Algorithm 1. Sigmoidoscopy and colonoscopy

Patients with bright red hematochezia and minimal blood loss can undergo initial evaluation with anoscopy and flexible
sigmoidoscopy, unless the patient is age 50 or older. In the latter setting, a full colonoscopy is generally recommended to rule out a colonic neoplasm. Those with dark hematochezia or bright red blood per rectum and evidence for significant blood loss should undergo full colonoscopy after a complete bowel preparation. In such patients bowel preparation with an oral lavage solution is preferred unless there is evidence for bowel obstruction. The solution can be administered rapidly via NG tube in order to expedite the procedure. Active, brisk bleeding and continued hemodynamic instability despite ongoing resuscitation is an indication for emergency angiography rather than colonoscopy. Since lower GI bleeding can originate anywhere in the small bowel or colon, angiography is also preferable to laparotomy in the setting of such bleeding. Surgery is generally reserved for patients whose bleeding site is identified by angiography but who are inappropriate for, or fail, angiographic therapy. Guidelines for the evaluation of patients with lower GI bleeding are provided in Algorithm 1. Angiography

Angiography is helpful in the patient with massive GI bleeding from either an upper or lower source. Selective mesenteric angiography reveals a bleeding site in up to 75% of patients with massive upper GI bleeding [Laine,1997]. Among lower GI bleeding sources, diverticular bleeding and angiodysplastic lesions are the most common lesions diagnosed by angiography [Fioritto,1989]. Animal studies have demonstrated that a bleeding rate of 0.5 mL/min is necessary in order for angiography to be positive [Nusbaum,1963]. Contrast extravasation during angiography localizes the site of bleeding, and in some instances can also define the lesion even when extravasation is not noted (e.g., arteriovenous malformation, tumor).

Finally, selective angiography allows for therapy with embolization or vasopressin infusion. Radionuclide scanning

Radionuclide scanning is sometimes helpful in localizing the source of lower GI bleeding. Red blood cells obtained by venipuncture are labeled with technetium 99m (99mTc) and reinjected into the patient. The abdomen is then scanned with a gamma counter. Since prolonged or repeated scanning is possible, bleeding can be detected even if it is intermittent or too slow to be detected on angiography. However, 99mTc-labeled red blood cell scans have their limitations. In earlier studies, the red blood cell scan was negative in up to 70% of patients with a lower GI source of bleeding [Bentley,1991; Voeller,1991] and in up to 50% of positive scans the predicted bleeding location was in correct [Bentley,1991; Dusold,1991]. More recently, digital equipment has allowed for more frequent imaging and cinematic display. In one study using this technology, the site of bleeding was accurately localized in 88% of positive scans [O'Neill,2000]. Despite their limitations, 99mTc-labeled red blood cell scans may be helpful in some patients with recurrent lower GI bleeding in whom all other diagnostic studies are negative.


Since red blood cell scans detect lower rates of bleeding (0.1cc/min) than angiography, these scans are often used as a screen to determine which patients should undergo angiography. If the red blood cell scan is negative, the angiogram is very unlikely to demonstrate active bleeding.

Radionuclide scanning is particularly helpful in the setting of bleeding from a Meckel’s diverticulum. 99mTc pertechnetate is given intravenously, and is taken up by the ectopic gastric mucosa in the Meckel’s diverticulum.

Using surgery as the gold standard, the sensitivity of a Meckel's scan ranges from 75% to 91%, while the specificity is 97% to 100% [Fries,1984; Kong,1993]. This study should be considered early in the evaluation of a young patient with lower GI bleeding. Radiographic and endoscopic studies of the small bowel

As previously discussed, barium studies are not immediately helpful in the setting of overt GI bleeding. However, a small bowel follow-through or enteroclysis is generally recommended to evaluate the small bowel when endoscopic studies are unrevealing and bleeding has ceased. Enteroscopy also may be helpful in this setting, particularly if small bowel angiodysplasia is suspected.

Wireless capsule endoscopy is currently being studied undergoing study in the evaluation of GI bleeding from a possible small bowel source. The Given Diagnostic Imaging System (Given Imaging Ltd, Norcross, GA) was recently approved by the Food and Drug Administration. The capsule endoscope, which is slightly larger than a large vitamin capsule, is swallowed with water and is propelled through the GI tract by peristalsis. Prior to capsule administration, patients either fast or complete a limited purge to rid the distal small bowel of debris. Images are transmitted to an antenna array worn on the abdomen and stored on a portable data recorder. The capsule battery allows for 8 hours of imaging, after which time the data are downloaded to a personal computer for viewing. In one recent study, video capsule endoscopy was superior to small bowel radiographs in the evaluation of 13 patients with obscure GI bleeding; 12 of 13 capsule endoscopy studies yielded a source of bleeding, while only 1 of 13 small bowel radiographs was positive [Costamanga,2002]. Angiodysplasia was the most common finding. It should be noted that capsule endoscopy has no role in the patient with acute GI bleeding because the test takes 8 hours and the results are not immediately available. Furthermore, capsule endoscopy will not replace push enteroscopy in patients with obscure GI bleeding, since the latter has the ability to provide therapy. It also should not be performed in lieu of a small bowel radiograph. A small bowel follow-through or enteroclysis is generally recommended before capsule endoscopy because a stricture is a contraindication to capsule administration. Finally, the interpretation of nonspecific findings remains problematic, and to date no study has determined its sensitivity and specificity.

2.3. Empiric treatment

Treatment of specific causes of GI bleeding is discussed below. To date, no study has demonstrated any benefit from empiric treatment of acute upper GI bleeding with intravenous H2 receptor antagonists [Zuckerman,1984; Collins,1985]. Furthermore, a large, randomized controlled trial of omeprazole in patients presenting with acute upper GI bleeding found no significant difference between the treatment and placebo groups with respect to transfusion requirements, rates of rebleeding, surgery, and mortality [Daneshmend,1992].

Patients who have known portal hypertension and esophageal or gastric varices may be treated empirically with agents that reduce portal pressure. These agents are discussed in more detail below.


3.1. Epidemiology

Upper GI bleeding is defined as bleeding from a gastrointestinal source that is proximal to the ligament of Treitz. It is more common than lower GI bleeding. Over 50% of upper GI bleeds are due to erosive or ulcerative disease of the stomach or duodenum [Gilbert,1990]. Etiologies of upper GI bleeding are listed in (Table 3).

Epidemiologic data from Europe indicated an upper GI bleeding annual incidence of 48 to 145 per 100,000 population in the 1960s and 1970s [Gilbert,1990]. In 1978 it was estimated that the total number of hospital admissions for upper GI bleeding in the U.S. was 150 per 100,000 population [Cutler,1990]. A more recent single HMO population-based study found an annual incidence of 102 hospitalizations for upper GI bleeding per 100,000 [Longstreth,1995]. Finally, 1992-1999 data from the National Hospital Discharge Sruvey found an annual hospitalization rate for UGI bleeding ranging from 149-172/100,000 [Lewis,2002].

Despite recent advances in endoscopic therapy, the mortality associated with GI bleeding remains significant at 5% to 11% [Lonstreth,1995; Laine,1993; Rockall,1995a; Rockall,1995b; Lewis,2002]. Factors associated with mortality due to upper GI bleeding have been identified in prospective studies [Rockall,1996; Gilbert,1990]. These included underlying renal, liver, neoplastic, central nervous system or lung disease, and physical findings consistent with cardiorespiratory or hemodynamic compromise, or liver failure. Patients with active bleeding at the time of endoscopy, blood transfusion requirement greater than 5 units, and requirement for surgery also had increased mortality. In addition, patients requiring emergency surgery had increased mortality compared with those undergoing more elective surgery. Other patients with increased mortality include those with recurrent bleeding after hospitalization [Fleischer,1983] and those who develop GI bleeding after hospitalization for other reasons [Longstreth,1995].

3.2. Pathogenesis, diagnosis and therapy of specific lesions

3.2.1. Gastric and duodenal erosive or ulcerative disease

Peptic ulcer disease is the most common cause of upper GI bleeding; over 50% of episodes are due to gastric or duodenal erosions or ulcers [Longstreth,1995; Silverstein,1981].

GI bleeding due to ulcer disease may be brisk, subacute or chronic. In erosive disease, the mucosal defect by definition is not deep enough to result in brisk bleeding. Patients with gastric/duodenal erosions typically have modest blood loss and may present with coffee-ground emesis or melena. If gastric and/or duodenal erosive disease is diagnosed on EGD, therapy should be initiated with either proton-pump inhibitors or H2-receptor antagonists.

In patients with acute GI bleeding from a gastric or duodenal ulcer, endoscopic findings can be used to guide medical therapy. While treatment with H2-receptor antagonists, when compared to placebo, has not been shown to reduce the risk of rebleeding in such patients [Laine,1994], several studies have shown that omeprazole is beneficial in selected patients [Khuroo,1997; Schaffalitzdy,1997; Hasselgren,1997; Lau,2000; Javid,2001]. Specifically, in those who have undergone endoscopic therapy for acute UGI bleeding from a peptic ulcer that is at high risk of rebleeding, treatment with omeprazole therapy is associated with a significant reduction in rebleeding rates [Khuroo,1997; Schaffalitzdy,1997; Lau,2000; Javid,2001], emergency surgery [Khuroo,1997; Schaffalitzdy,1997; Javid,2001], transfusion requirement [Schaffalitzdy,1997; Javid,2001], and length of hospital stay [Javid,2001]. Studies have differed with respect to route of administration (p.o or I.V.) and duration of treatment (3 or 5 days). Since patients with a high risk of rebleeding should be kept NPO, the I.V. route is generally preferred. Both of the positive studies with I.V. omeprazole used an 80 mg bolus followed by an 8 mg/hour continuous infusion for 72 hours, after which time patients were transitioned to omeprazole 20 mg p.o. daily for up to 8 weeks [Schaffalitzdy,1997; Lau,2000]. Finally, two studies have compared omeprazole and H2 antagonists in patients with acute GI bleeding from high-risk ulcers [Lin,1998; Lanas,1995]. One of the studies found a significant reduction in rebleeding rates in the omeprazole group [Lin,1998], while the second study found trends toward decreased rates of rebleeding and surgery in those receiving omeprazole [Lanas,1995]. IV omeprezole is currently unavailable in the United States. The only intravenous PPI available in the U.S. is pantoprazole, and we await studies of this agent in acute ulcer bleeds. Given that the beneficial effect of IV omeprezole is likely related to the ability of an IV PPI to maintain continuous elevation of the intragastric pH, it appears reasonable to extrapolate from the IV omeprezole studies and use IV pantoprazole at a dose of 80 mg bolus, 8 mg/hr for ulcers at high risk of rebleeding.

There is no evidence that high-dose I.V. PPI therapy is beneficial in patients with GI bleeding from an ulcer that has a low risk of rebleeding (i.e., a clean-based ulcer on endoscopy). In these patients, oral PPI therapy can be instituted after the patient has recovered from the endoscopy. PPIs are generally recommended over H2 blockers because healing of uncomplicated gastric and duodenal ulcers is more rapid with proton-pump inhibitors (PPIs) than with H2-antagonists [Lauritsen,1985; McFarland,1990; Marks,1991; Hotz,1992; Delle,1992]. Helicobacter pylori

The discovery of
Helicobacter pylori (H. pylori) infection has revolutionized our diagnostic and therapeutic approaches to ulcer disease. Over 90% of duodenal ulcers and 80% of gastric ulcers can be attributed to H. pylori, while the vast majority of remaining ulcers are due to NSAID use [Graham,1996]. The pathogenesis of H. pylori-related ulcer disease, and H. pylori testing and treatment are discussed in Chapter 2.

Patients who are found to have a
gastric ulcer, duodenitis or duodenal ulcer on endoscopy should have gastric mucosal biopsies to assess for H. pylori infection. Infection with H. pylori can be diagnosed by rapid urease testing of a single antral biopsy or by submitting biopsies for histologic examination. Non-invasive tests such as the urea breath test or stool antigen test are as sensitive and specific as endoscopic biopsy and may be more appropriate in patients with brisk bleeding (performed once bleeding has stopped), when it may be unwise to prolong the endoscopy in order to take biopsies. Patients with duodenitis or duodenal/gastric ulcer who test positive for H. pylori should be treated with an acid-reducing agent (PPI BID) and at least two antibiotics (clarithromycin 500 mg BID, amoxicillin 1 gm BID) for at least 1 week. More information about specific H. pylori treatment regimens can be found in Chapter 2. Patients with complicated ulcer disease (bleeding or perforation) should have confirmation of H. pylori cure by either urea breath test, stool antigen test or endoscopic biopsy. Repeat testing should be performed a minimum of four weeks after completion of therapy [Rollan,1997]. Proton pump inhibitor therapy should also be discontinued for 1 week prior to repeat testing for H. pylori because of the suppressive effect of these drugs on H. pylori growth, which may lead to false negative urea breath tests and stool antigen tests [Laine,1998]. Because serology for H. pylori may remain positive for up to two years following successful eradication, it cannot be relied upon for confirmation of successful cure patients. NSAID erosions/ulcer

It is well recognized that NSAID use is associated with an increased risk of gastric or duodenal ulcer. Antral erosions are present within 1 to 2 days in virtually all individuals taking NSAIDs [O'Laughlin,1981; Larkai,1987]. However, these erosions are usually asymptomatic, often disappear with continued NSAID use, and do not predict ulcer development. Symptomatic ulcers and ulcer complications develop in only 2% to 4% of patients taking NSAIDs for a year [Paulus,1985]. The risk of hospitalization or death from such complications is 1.3% to 1.6% per year in patients with rheumatoid arthritis [Fries,1991]. Although the risk appears small, the millions of U.S. patients taking NSAIDS for arthritis or aspirin for cardiovascular prophylaxis translates into a large number of patients at risk. Even low-dose aspirin, with or without enteric coating, is associated with an increased risk of UGI bleeding (standardized incidence ratio 2.6 in both groups) [Sorenson,2000].

The pathogenesis of NSAID-induced ulcer disease relates to both local and systemic effects [Bjorkman,1995]. Aspirin and NSAIDs are lipid soluble at acid pH and thus diffuse across gastric epithelial cell membranes. Furthermore, these agents are cytotoxic, leading to a breakdown of the gastric mucosal barrier. These local effects rarely lead to acute ulceration and symptomatic disease. It is the systemic inhibition of prostaglandins that leads to symptomatic chronic ulceration. Such inhibition further compromises the gastric mucosal barrier by decreasing mucus and bicarbonate secretion and mucosal blood flow. NSAIDs also inhibit platelet function, and their use has been associated with GI bleeding from throughout the GI tract [Lanas,1996].

Nabumetone and etodolac, NSAIDs that have partial selectivity for prostaglandin synthesis inhibition in inflamed tissues, appear to have lower rates of GI ulceration and related complications. In three large postmarketing survey studies of nabumetone, GI ulcer and complication rates ranged from .02% to .7% [Jenner,1990; Stroelmann,1990; Willkens,1990]. Two large open-label studies of etodolac reported adverse reactions in 0.1% and .005%, respectively [Benhamou,1990]. Although these newer agents appear promising with respect to GI complications, complication rates attributed to newer versus older NSAIDs cannot be directly compared due to differences in dose, patient characteristics and study design. Large, randomized trials comparing these newer agents with older NSAIDs are lacking. Small, short-term endoscopic studies have reported fewer GI erosions and/or ulcers in patients taking nabumetone or etodolac compared with other NSAIDs [Roth,1987; Bianchi,1991; Laine,1995]. Newer NSAIDs (celecoxib, rofecoxib) that are even more selective for prostaglandin synthesis in inflamed tissues are now available. In prospective endoscopic studies, these drugs caused fewer ulcers when compared with other NSAIDs [Simon,1999; Laine,1999; Hawkey,2000; Goldstein,2001], and caused no more ulcers than placebo [Simon,1999; Laine,1999; Hawkey,2000]. Two large randomized trials also found significantly reduced rates of symptomatic ulcer and ulcer complications in patients assigned to rofecoxib [Bombardier,2000] or celecoxib [Silverstein,2000] when compared to nonselective NSAIDs (naproxen and ibuprofen/diclofenac, respectively). In the latter study, subjects were allowed to take low-dose aspirin for cardiovascular prophylaxis. Interestingly, aspirin use appeared to negate the GI benefits of celecoxib. A recently published study randomized over 8000 patients with rheumatoid arthritis to rofecoxib or naproxen and followed them for a median of 9 months [Laine,2002]. Overall, upper GI events (symptomatic or complicated ulcer) were significantly decreased in the rofecoxib group, and the number needed to treat to prevent an event was 31. However, when taking into account the highest risk patients (age = 75, prior upper GI event, or severe rheumatoid arthritis), the number needed to treat was only 10-12. Risk factors for NSAID ulcer complications

Risk factors for NSAID-induced ulcer complications include age greater than 65-75 years, history of peptic ulcer or gastrointestinal bleeding, history of heart disease, prednisone or warfarin use, and severe rheumatoid arthritis [Fries,1991; Silverstein,1995; Wolfe,1999; Laine,2002]. NSAIDs should be avoided, if possible, in these high-risk patients. If an NSAID is necessary, rofecoxib or celecoxib should be considered, although the cost-effectiveness of this approach has not been determined. Other options includeprophylaxis with misoprostol or omeprazole. Prevention of NSAID ulcers and ulcer complications

In a randomized, placebo-controlled study of three
misoprostol dosing regimens in the prophylaxis of NSAID-related ulcer disease, doses of 200 mcg bid, tid and qid all significantly decreased the rate of gastric and duodenal ulcer when compared to placebo [Raskin,1995]. A dose-reponse effect was noted in the prevention of gastric, but not duodenal, ulcer. Due to side effects such as abdominal pain, nausea, flatulence, and diarrhea, twice daily dosing is generally better tolerated than more frequent dosing [Silverstein,1995; Raskin,1995]. Misoprostol also has been show to reduce the risk of NSAID-related ulcer complications [Silverstein,1995].

Prophylaxis with
H2-blockers is as effective as misoprostol in reducing the rate of NSAID-induced duodenal ulcer, but is significantly less effective in the prevention of gastric ulcers [Raskin,1996]. Three randomized, placebo controlled studies have demonstrated that omeprazole, 20 mg daily, is effective in the primary prophylaxis against gastroduodenal ulcers in patients taking NSAIDs [Ekstrom,1996; Cullen1998; Bianchi,1998]. In a recently published randomized, placebo-controlled trial of omeprazole versus misoprostol for the prophylaxis of NSAID-related ulcer, omeprazole 20 mg daily was more effective than misoprostol 200 mcg twice daily in patients with prior NSAID-related erosive or ulcerative disease, and either drug was more effective than placebo [Hawkey,1998]. Finally, a recent placebo-controlled trial demonstrated that lansoprazole 30 mg daily significantly reduced the risk of recurrent aspirin-related ulcer complications in patients who continued on low dose aspirin [Lai,2002].

In a two-week placebo-controlled study of sucralfate gel prophylaxis in NSAID users, the incidence of gastroduodenal erosions and/or ulcers on 2-week endoscopy were significantly less frequent in the sucralfate group [Miglioli,1996]. However, no study has demonstrated that long-term prophylaxis with sucralfate prevents gastroduodenal ulcer or ulcer complications. A randomized trial comparing misoprostol to sucralfate found that the former was significantly more effective in preventing gastric ulcers in patients on chronic NSAID therapy [Agrawai,1991]. Thus sucralfate and related agents have no role in the prophylaxis of NSAID-related ulceration. Treatment

In patients with GI bleeding from an NSAID-induced gastric or duodenal ulcer, it is best to discontinue the NSAID and treat with a PPI for four weeks. If it is necessary to continue the NSAID, the ulcer is likely to heal if treated, but PPIs rather than H2 blockers should be used in this setting [Yeomans,1998]. In a recently published randomized trial of NSAID ulcer treatment, omeprazole 20 mg or 40 mg daily and misoprostol 200 mcg qid all were equally effective in healing NSAID-related erosive or ulcerative disease in patients who continued to take NSAIDs, with healing rates of 71% to 76% at 8 weeks [Hawkey,1998]. However, patients with "clinically important" GI bleeding were excluded from this study. Certain factors should be considered before deciding if NSAID use should be continued in the setting of GI bleeding. Such variables include the significance of the GI bleed, patient comorbidity (that is, the ability to tolerate another bleed), and whether or not the patient had warning symptoms such as dyspepsia before presenting with GI bleeding. If long-term NSAID therapy is deemed necessary, ulcer prophylaxis should be continued indefinitely. Whether to use misoprostol or a proton pump inhibitor in this setting is uncertain. Hawkey and colleagues recently reported that duodenal ulcer recurrence was significantly lower with omeprazole maintenance at 20 mg daily (39%) than with misoprostol maintenance at 200 mcg bid (52%), but that both agents were superior to placebo (73%) [Hawkey,1998]. Role of H. pylori infection in NSAID ulcer

The role of H. pylori infection in NSAID ulcer is not certain. Furthermore, it may be difficult to definitively prove that an ulcer is due to NSAIDs rather than H. pylori, since an ulcer may have been present before NSAID therapy was initiated. In general, all patients who develop an ulcer while on NSAIDs should be tested for H. pylori infection and treated if positive. However, although there are some data to indicate that endoscopic ulcers may be less frequent if patients are treated for H.pylori before beginning NSAIDs, we await further clinical outcome studies of this question before recommending routine testing and treatment of H. pylori prior to initiating NSAIDs. Ulcers not associated with H. pylori or NSAID use

An increasing proportion of gastric and duodenal ulcers cannot be attributed to H. pylori infection or NSAID use [Laine,1998]. In the case of gastric ulcers, it is important to rule out malignancy. If malignancy (in the case of gastric ulcer) and Zollinger-Ellison syndrome have been excluded, it is recommended that the patient with a complicated ulcer continue on life-long acid-suppression therapy once the ulcer has healed. In the majority of cases a single dose of H2-antagonist at bedtime will reduce the risk of ulcer recurrence to approximately 15% to 25%, although some patients will require twice this dose [Strum,1986; Texter,1986] or a proton pump inhibitor for effective prophylaxis . In general the appropriate dose for prophylaxis can be individualized according to the patient’s symptoms. Patients who develop recurrent ulcer disease despite medical prophylaxis should be considered for anti-ulcer surgery. Indications for follow-up endoscopy

Since approximately 1% of gastric ulcers are malignant, gastric ulcers should be biopsied at the time of diagnosis. If they are not, follow-up endoscopy to document ulcer healing is recommended regardless of endoscopic appearance or evidence for H. pylori infection or NSAID use. The follow-up endoscopy is performed approximately 8 weeks after the initial diagnosis. If the ulcer is still present and has not shown significant healing, several biopsies are taken from its edge. Further management is individualized and obviously depends on whether the non-healing ulcer is benign or malignant. Stress gastritis

Stress gastritis refers to hemorrhagic gastric erosions in patients who are critically ill [Ben-Menachem,1994; Maier,1994]. These lesions can be diffuse and typically involve the gastric body and fundus. With extensive involvement bleeding can be massive. Physiologic stress associated with trauma, burns, major surgery or severe medical illness requiring ICU hospitalization have been associated with the development of hemorrhagic gastritis [Chamberlain,1993]. The pathogenesis is due to an imbalance between aggressive and protective mucosal factors. While infection with H. pylori may coexist, it is thought to play no role in the pathogenesis of stress gastritis. In a recent study, the independent risk factors for development of stress ulcer bleeding were respiratory failure and coagulopathy [Cook,1994]. These patients should receive prophylaxis with H2 antagonists [Cook,2000]. The diagnosis of stress gastritis is readily made on endoscopy and treatment is with aggressive reduction of gastric pH using H2 antagonists or proton pump inhibitors [Collier,1989; Hadi,1989].

3.2.2. Bleeding associated with portal hypertension Esophageal and gastric varices

At the time of diagnosis of cirrhosis, esophageal and/or gastric varices are present in approximately 60% with decompensated cirrhosis and in approximately 30% with compensated cirrhosis [D'Amico,1995].

A discussion of other causes of portal hypertension is beyond the scope of this chapter. The incidence of bleeding in patients with known esophageal varices is approximately 10% to 15% per year, and in patients with large varices is 20% to 30% [D'Amico,1995]. Gastrointestinal bleeding due to varices is typically brisk and associated with hemodynamic compromise, although sometimes patients with variceal bleeding present with melena and are hemodynamically stable. The mortality associated with the first variceal bleed ranges from 30% to 50% [D'Amico,1995].

Endoscopic sclerotherapy and band ligation of varices are the mainstay of emergent therapy in patients presenting with esophageal variceal bleeding. Patients in whom early rebleeding is successfully prevented with endoscopic therapy should undergo repeated sessions until varices are eradicated. With such therapy the risk of rebleeding is significantly reduced but remains as high as 50% with sclerotherapy and 35% with band ligation [Steigmann,1992; Laine,1993; Gimson,1993]. Data from recent trials indicate that band ligation, when compared to sclerotherapy, is associated with fewer treatment-related complications [Stiegmann,1992; Laine,1995; Lo,1995] and requires fewer sessions to achieve variceal eradication [Gimson,1993; Laine,1995; Lo,1995]. Lower rebleeding rates [Gimson,1993; Lo,1995] and improved survival [Stiegmann,1992; Lo,1995] have also been reported in patients undergoing variceal ligation. Band ligation of esophageal varices has largely supplanted sclerotherapy.

Although medical therapy for variceal hemorrhage has been extensively studied, the benefit of such therapy remains uncertain. Earlier studies demonstrated that vasopressin or vasopressin analogue reduced early rebleeding rates, but often at the expense of significant toxicity from systemic vasoconstriction [Mallory,1980; Fogel,1982]. More recently somatostatin and the somatostatin analogue, octreotide, have replaced vasopressin in the acute treatment of variceal hemorrhage (only octreotide is available in the U.S.). These agents reduce portal pressure by decreasing splanchnic blood flow [Bosch,1981]. While many studies have demonstrated a benefit of using somatostatin or octreotide in acute variceal hemorrhage, overall results have been mixed [Valenzuela,1989; Burroughsk,1990; Saari,1990; Hwang,1992; Shields,1992; Planas,1994; Sung,1995; Goetschze,1995]. In some studies, somatostatin and octreotide were as effective as endoscopic therapy in controlling acute variceal hemorrhage [Shields,1992; Planas,1994]. In addition, patients with acute variceal hemorrhage treated with both octreotide and endoscopic therapy had significantly lower rebleeding rates than those treated with endoscopic therapy alone [Besson,1995; Sung,1995; Avgerinos,1997; Zuberi,2000; Cales,2001]. In all studies, somatostatin or one of its derivatives was continued for a duration of 5 days. A meta-analysis was recently published that included these trials and three additional studies published only in abstract form [Banares,2002]. This analysis found that somatostatin/somatostatin analogues, when combined with endoscopic treatment, improved initial and 5-day hemostasis compared to endoscopic therapy alone. Given the demonstrated benefit of these agents and the significant mortality associated with variceal hemorrhage, it is reasonable to use octreotide in patients with documented bleeding from esophageal varices. A loading dose of 50 mcg IV is given, followed by an infusion of 25 mcg to 50 mcg per hour for a duration of 5 days. Two of the studies initiated somatostatin or its analogue before endoscopy [Avgerinos,1997; Cales,2001]. In clinical practice, octreotide is often initiated before the endoscopy if variceal hemorrhage is suspected. This is a reasonable approach, provided that the endoscopy can be expedited and the octreotide can be discontinued if an alternate source of bleeding is discovered.

The Senkstaken-Blakemore or Minnesota tube remains an important therapeutic tool in patients with brisk esophageal or gastric variceal hemorrhage that cannot be controlled on initial endoscopy.

It is generally felt that the gastric balloon on these tubes should not remain inflated for more than 48 hours at one time. The esophageal balloon should be inflated only if bleeding continues despite inflation of the gastric balloon, and for no more than 8 hours at a time due to the high risk of tissue necrosis. Patients who continue to bleed or who have more than one rebleeding episode despite endoscopic and medical therapy should be considered for portal decompression. Due to the morbidity and mortality associated with surgical portal decompression in patients with Childs B or C cirrhosis, transjugular intrahepatic shunting (TIPS) is generally preferred in these patients, particularly if portal decompression is serving as a bridge to liver transplantation.

Patients with Childs A cirrhosis should be considered for surgical decompression, as they may not require liver transplantation for several years and TIPS is associated with occlusion due to neointimal hyperplasia in the majority of patients within two years [Somberg,1997].

Once acute esophageal variceal bleeding has been controlled, patients are generally treated with repeat outpatient endoscopic therapy until varices are eradicated. Oral beta-blocker therapy is used in combination with endoscopic therapy as it has been shown to decrease the rebleeding rate over endosocopic therapy alone [Vinel,1992]. Non-selective beta-blockers are preferred because their effect on portal pressure likely results from both splanchnic vasodilation and decreased cardiac output. It is recommended that the dose is titrated to reduce the resting pulse by 25%. Following the eradication of esophageal varices, endoscopy is repeated every 3 to 6 months in order to monitor for recurrent varices and provide further endoscopic therapy if indicated.

Some investigators have compared medical therapy (propranolo/nadolol plus isosorbide mononitrate) to endoscopic therapy for the prevention of variceal rebleeding [Villanueva,1996; Patch,2002; Lo,2002]. In one study, the combination of nadolol and isosorbide mononitrate was superior to endoscopic sclerotherapy in the prevention of rebleeding [Villanueva,1996]. Two other studies compared medical therapy to band ligation [Patch,2002; Lo,2002]. One found them to be equivalent in preventing rebleeding [Patch,2002], while the second found that band ligation was more effective than nadolol and isosorbide mononitrate [Lo,2002]. Because medical therapy has not yet been compared to the current standard of care (band ligation and beta-blocker therapy in combination), these findings are unlikely to change clinical practice.

Finally, several studies have compared TIPS to sclerotherapy [Cabrera,1996; Cello,1997; Sanyal,1997; Garcia-Vilareal,1999; Sauer,1997; Merli,1998] or band ligation [Jalan,1997; Pomier_layrargues,2001] in the prevention of esophageal variceal rebleeding. All but one [Sanyal,1997] found that rebleeding rates were significantly reduced in those treated with TIPS. These studies were included in a recent meta-analysis that confirmed reduced rebleeding rates in those undergoing TIPS [Papatheodoridis,1999]. However, encephalopathy was increased in TIPS patients, and there was no difference in overall survival [Papatheodoridis,1999]. Thus TIPS is not routinely recommended for the prevention of rebleeding from esophageal varices, but is reserved for the treatment of bleeding that is refractory to endoscopic therapy.

The management of bleeding from gastric varices is more difficult. Although medical and/or endoscopic therapy may be successful in controlling acute bleeding, endoscopic therapy does not lead to effective variceal obliteration. These patients should be referred for TIPS [Barange,1999] or surgical decompression, preferably after control of the initial bleeding episode. Portal hypertensive gastropathy

Patients with portal hypertension also may develop
portal hypertensive gastropathy.

Although the pathogenesis is not well understood, increased mucosal blood flow and passive congestion of the submucosa appear to play a role [Panes,1992; Ohta,1994]. Gastropathy accounts for 8% to 20% of acute bleeding in patients with portal hypertension and can also lead to chronic blood loss [Sarin,1992; Gostout,1993]. Many patients presenting with bleeding from portal hypertensive gastropathy have recurrent bleeding [Gostout,1993]. Propranolol may be effective in preventing recurrent hemorrhage from portal hypertensive gastropathy [Hosking,1987; P'Erez,1991]. Occasionally transjugular intrahepatic stent (TIPS) placement or surgical portal decompression is required for severe, intractable bleeding [Babb,1988; Orloff,1995]. Other

varices may cause significant upper GI bleeding. If varices are within the reach of the endoscope, sclerotherapy or band ligation may successfully control the acute hemorrhage [Tazawa,1995]. However, the long-term management of duodenal varices generally requires portal decompression.

3.2.3. Mallory-Weiss tear

Mallory-Weiss tear, a mucosal laceration of the gastric cardia or gastroesophageal junction, was first described in 1929 by Kenneth Mallory and Soma Weiss [Mallory,1929]. These tears account for approximately 5% to 15% of all cases of upper GI bleeding and are relatively common in alcoholics [Harris,1993]. The classic presentation is that of repeated retching, vomiting or coughing followed by hematemesis. However, up to 50% of patients with Mallory-Weiss tears do not give a history of antecedent retching or vomiting [Harris,1993]. Blood loss is usually modest and only 10% present with hemodynamic compromise. The typical endoscopic appearance is a mucosal tear in the cardia either just below the gastroesophageal (GE) junction or bridging the GE junction.

Bleeding is self-limited in 90% of cases and patients in whom bleeding has ceased at endoscopy require no therapeutic intervention. Active bleeding at the time of endoscopy is treated with injection of saline or 1:10,000 epinephrine and/or electrocautery. Angiography with vasopressin infusion or embolization has been reported to be successful in small, uncontrolled series [Carsen,1978; Fisher,1980]. Rarely patients with severe or recurrent bleeding require surgery and oversewing of the tear [Harris,1993].

3.2.4. Esophagitis

Heartburn and regurgitation due to gastroesophageal reflux are common GI symptoms . Since most patients with persistent symptoms are treated with acid-reducing agents, bleeding from

or esophageal ulceration is not very common. One exception is the alcoholic patient, who may not present until overt signs of bleeding develop. Prolonged nasogastric tube placement in hospitalized patients may also lead to bleeding from erosive esophagitis.

3.2.5. Alcohol and upper GI bleeding

It is a common misconception that "alcoholic gastritis" is a significant cause of upper GI bleeding in patients with excess alcohol use. Although both animal and human studies have shown that
alcohol is directly injurious to gastric mucosa [Guth,1984; Tarnawski,1987], alcohol use is most commonly associated with a reactive gastropathy. Only rarely does this entity lead to GI bleeding. Patients who are actively drinking more commonly bleed from ulcer disease, disorders related to portal hypertension, Mallory-Weiss tears, or esophagitis [Wilcox,1996]. A recent multinational case-control study found that heavy alcohol consumption was a risk factor for major gastric and duodenal bleeding [Kelly,1995].

3.2.6. Neoplasms

Adenocarcinoma of the stomach is the most common malignant
neoplasm of the upper GI tract and often presents with frank or occult GI bleeding. With the exception of vascular tumors and stromal tumors [Senewiratne,1987], benign neoplasms of the upper GI tract seldom bleed. Stromal tumors, formerly called leiomyomas, are prone to ulceration and thus may bleed. They can be found throughout the GI tract.

Hemangiomas are the most common vascular tumor of the GI tract, but are a rare cause of GI bleeding. Almost all bleeding hemangiomas are cavernous hemangiomas and they most commonly occur in the small intestine. Endoscopically they appear as single or multiple red, purple or blue nodular lesions. Cavernous hemangiomas of the skin, GI tract and other viscera are seen in patients with the blue rubber bleb nevus syndrome. Hemangiosarcomas, malignant vascular tumors, are rare in the GI tract. The treatment of hemangiomas and hemangiosarcomas is surgical.

3.2.7. Vascular anomalies

Vascular anomalies are an unusual cause of upper GI bleeding, accounting for only 5% of cases. True arteriovenous malformations are rare.
Angiodysplasia, whether sporadic or secondary, is the most common vascular anomaly seen in the GI tract. Angiodysplasia

Gastrointestinal angiodysplastic lesions are dilated, tortuous vessels in the mucosa and submucosa. These lesions may be sporadic, usually developing in the elderly, or may be found in association with a number of disorders including renal failure, cirrhosis, the CREST syndrome, radiation injury, von Willebrand’s disease, and aortic stenosis. Angiodysplastic lesions may occur anywhere in the GI tract, but are more commonly found in the colon, followed by the small intestine and the stomach. These lesions usually lead to occult blood loss, but can also cause overt GI bleeding. They are usually apparent at endoscopy, at which time therapy with laser or thermal probes may be applied. Bleeding that is refractory to endoscopic or medical therapy is an indication for surgical resection. In the case of small bowel angiodysplastic lesions where endoscopic therapy is not possible due to the location or multiplicity of lesions, low-dose combination estrogen/progesterone therapy may be beneficial [Van Custem,1990].

Hereditary hemorrhagic telangiectasia (Osler-Weber-Rendu syndrome) encompasses a group of autosomal dominant disorders associated with vascular ectasia and arteriovenous malformations of the nose, skin, lungs, brain and GI tract. Epistaxis is the most common manifestation and occurs in most patients before the third decade. Recurrent bleeding from the stomach, small intestine and colon usually begins in the fifth or sixth decade and occurs in only a minority of patients. Endoscopic therapy with laser or thermal probes is usually helpful in controlling actively bleeding ectasias, but bleeding frequently recurs due to the presence of multiple lesions. As previously mentioned, combined estrogen/progesterone therapy may be useful [Van Custem,1990]. Because multiple lesions are present throughout the GI tract, surgical therapy is not an option. Arteriovenous malformations

arteriovenous malformations (AVMs) are rare in the GI tract. They are present in younger patients, suggesting that they are congenital, but their true etiology remains uncertain. On endoscopy AVMs appear as nodular lesions.

Unlike angiodysplasia, they may involve any layer of the gut wall. Surgical resection is the treatment of choice, although poor surgical candidates may be treated with arterial embolization. Dieulafoy lesion

This lesion, first described by Gallard in 1884 and later by the French surgeon Dieulafoy in 1896, is characterized by an aberrantly large and tortuous submucosal artery that may erode through a small mucosal defect, resulting in massive hemorrhage [Juler,1984]. Approximately 75% to 95% occur in the proximal stomach, usually on the lesser curvature and within 6 cm of the gastroesophageal junction, although they have been reported to occur thoughout the GI tract [Scheider,1994; Dy,1995; Goins,1995; Farrell,1992]. The initial endoscopy in patients bleeding from a
Dieulafoy lesion may reveal no obvious bleeding because of the lesion’s small size, rapid healing of the mucosal defect, or the presence of large amounts of retained blood obscuring the lesion. Not infrequently a second endoscopy is required to identify the lesion; in some patients the diagnosis is made only on angiography or during surgery [Reilly,1991]. The typical endoscopic appearance is that of a bright or dark red "nipple" protruding from the mucosa without an associated ulcer base, distinguishing it from a visible vessel in an ulcer.

These lesions often respond to endoscopic therapy with injection of saline or 1:10,000 epinephrine and/or endoscopic application of thermal probes [Pointner,1988; Baettig,1993]. Patients who do not respond to endoscopic therapy require surgical intervention. Although the literature in this area is limited, it is generally felt that a wide wedge resection of the artery and bleeding site is preferable to oversewing the artery in the area of the mucosal defect [Veldhuyzen-van-Zanten,1986]. Patients who are poor surgical candidates may respond to angiographic embolization [Sherman,1979]. Gastric antral vascular ectasia

Gastric antral vascular ectasia (GAVE), or watermelon stomach, is an unusual cause of upper GI bleeding. Patients typically present with melena or occult blood loss with iron deficiency anemia [Katz,1993]. GAVE is usually idiopathic, but has been reported in association with several disorders including scleroderma, cirrhosis and after bone marrow transplantation [Watson,1996; Payen,1995; Tobin,1996]. Endoscopy is often diagnostic, revealing a watermelon-like striped pattern of mucosal erythema radiating proximally from the pylorus and usually limited to the antrum.

Small mucosal vascular malformations may also be seen. Mucosal biopsy typically reveals foveolar hyperplasia, fibromuscular hyperplasia of the lamina propria and dilated mucosal capillaries with intraluminal thrombi [Suit,1987]. Although the pathogenesis is unknown, the histologic appearance suggests that prolapse of antral folds through the pylorus may play a role. Endoscopic laser or thermal probe therapy of the lesions is usually successful [Bourke,1996; Ng,1996]. Rare refractory cases may require surgical resection of the involved area.

3.2.8. Miscellaneous causes

Aortoduodenal fistulas, representing 80% of all aortoenteric fistulas, are a very unusual but catastrophic cause of bleeding from the upper GI tract. It is estimated that 0.4% to 4% of patients with aortic grafts develop an aortoenteric fistula [Champion,1982]. Primary aortoduodenal fistulas associated with atherosclerotic aneurysms or trauma are much less common [Voorhoeve,1996]. The classic presentation of an aortoduodenal fistula is that of a "herald bleed" (brief, with spontaneous resolution), followed anywhere from hours to weeks later by a massive upper GI bleed. A high index of suspicion and rapid diagnosis is imperative as exsanguinating hemorrhage may occur. Intermittent or occult bleeding can occur in patients who have an aortoenteric erosion rather than a true fistula [Bessen,1985]. An aortoduodenal fistula or erosion rarely is seen endoscopically [139,140]. More commonly fresh blood or clots may be found in the third portion of the duodenum. Angiography is seldom helpful unless a pseudoaneurysm is seen [Bessen,1985]. Although CT visualization of an aneurysm with associated extraluminal gas is virtually diagnostic, this finding is also uncommon. The diagnosis is usually made or confirmed on laparotomy. Surgical repair is the only therapeutic option.

Hemobilia, another rare cause of UGI bleeding, is usually associated with intraductal neoplasm, trauma, or iatrogenic injury such as percutaneous liver biopsy [Bloechle,1994]. Other reported causes include cystic artery pseudoaneurysm [Nakajima,1996]. The diagnosis is suggested by jaundice, right upper quadrant pain and gastrointestinal bleeding. Hemobilia may be confirmed at endoscopy by finding blood limited to the second portion of the duodenum or blood emanating from the ampulla, but often requires angiography. Angiographic therapy is the treatment of choice, although occasionally surgical therapy is necessary [Katz,1993].

Bleeding from the ampulla may also be due to hemosuccus pancreaticus. This is most commonly due to a pseudoaneurysm of the splenic artery in patients with a pancreatic pseudocyst or chronic pancreatitis but rarely may occur in patients with pancreatic duct malignancy [Risti,1995]. Diagnosis and therapy are dictated by the underlying pathology.

Finally, patients who present with symptoms of upper GI bleeding may occasionally be bleeding from a non-GI source, as in patients with epistaxis and swallowed blood. Factitious bleeding from ingestion of animal blood or human blood after auto-phlebotomy has also been reported [Rose,1996].


4.1. Epidemiology

Bleeding from a site distal to the ligament of Treitz is less common than upper GI bleeding. More commonly bleeding is from a colonic rather than a small bowel source. Causes of lower GI bleeding are listed in (Table 4).

A recent U.S. population-based study estimated an annual incidence rate of lower GI bleeding at 20.5/100,000 [Longstreth,1997]. The most common diagnoses were diverticulosis,
colorectal cancer and ischemic colitis. Bleeding rate was increased in males and in older patients. The mortality rate was 2.4% in patients admitted with lower GI bleeding compared with 23.1% in patients who developed lower GI bleeding after hospitalization for another reason. Among all patients presenting with lower GI bleeding, hemorrhoidal bleeding is probably the most common cause, followed by diverticula, vascular anomalies, colorectal cancer, and infectious or inflammatory colitis.

4.2. Pathogenesis, diagnosis and therapy of specific lesions

4.2.1. Hemorrhoids

Hemorrhoidal bleeding is the most common cause of lower GI bleeding. Bleeding is typically small volume and intermittent, with bright red blood on the surface of the stool, on the toilet paper and/or in the toilet bowl. Patients often describe blood dripping into the toilet bowl following a bowel movement. Occasionally bleeding is severe. The diagnosis can be confirmed on anoscopy and/or flexible sigmoidoscopy.

Patients with small volume, bright red hematochezia should undergo flexible sigmoidoscopy if anoscopy does not reveal bleeding hemorrhoids. Fiber supplementation is the mainstay of therapy. Severe or recurrent bleeding are indications for hemorrhoidal band ligation or hemorrhoidectomy.
Anal fissures may also bleed, but bleeding is usually minimal (blood on the toilet paper or scant blood on the stool surface) and is associated with anal discomfort. Fiber supplementation is advised. Sitz baths may also be helpful in relieving anal discomfort.

4.2.2. Colonic diverticula

After hemorrhoids, colonic diverticular bleeding is the most common cause of lower GI bleeding [Longstreth,1997]. Colonic diverticula are felt to develop as a result of increased intraluminal pressure and occur at sites of natural weakness in the colonic wall where transmural arteries are located. Local erosion into one of these arteries leads to brisk but usually self-limited bleeding. Rarely bleeding is massive on presentation, requiring emergent diagnostic angiography followed by intra-arterial infusion of vasopressin or segmental resection. In patients with self-limited bleeding the diagnosis is often made by finding diverticula on colonoscopy in the absence of another identifiable cause of lower GI bleeding. Rarely a clot is seen in the diverticulum that has bled [Bokhari,1996]. Patients with a single self-limited episode of diverticular hemorrhage can be managed conservatively. Stool bulking agents are generally recommended. Recurrent bleeding, which develops in 9% and 10% of patients at 1 and 2 years, respectively [Longstreth,1997], is an indication for surgical resection if the bleeding location can be identified.

4.2.3. Vascular anomalies

Sporadic and secondary angiodysplasia are a common cause of bleeding from the small bowel and colon. These are discussed in detail in section As previously mentioned, Dieulafoy lesions rarely lead to bleeding from the small bowel and colon.

4.2.4. Colorectal neoplasm

Although colorectal cancer is most commonly associated with occult blood loss rather than overt bleeding, patients with rectosigmoid lesions may present with hematochezia. The diagnosis is readily made on endoscopy.

Although endoscopic laser therapy can slow bleeding, surgical resection is usually recommended to control the bleeding and to treat the cancer.

4.2.5. Infectious, inflammatory or ischemic colitis

A variety of infectious, inflammatory and ischemic colitides may present with bloody diarrhea. The diagnosis of
infectious colitis is usually confirmed by stool culture or assay for Clostridium difficile toxin, but occasionally stool studies are negative. Sigmoidoscopy is indicated in the setting of negative stool studies and persistent bloody diarrhea. Endoscopy may reveal pseudomembranes characteristic of C. difficile, and mucosal biopsies may help to distinguish among infection, inflammatory bowel disease, or ischemia. However, certain infectious agents (C. difficile and E. Coli O157:H7) may mimic ischemia on histology. Endoscopy is always indicated in the setting of possible inflammatory or ischemic colitis, unless there is clinical evidence for perforation. Again, mucosal biopsies are usually diagnostic.

4.2.6. Solitary rectal ulcer syndrome

A solitary rectal ulcer may develop in the setting of rectal prolapse and is due to mucosal ischemia. These patients may present with hematochezia. The absence of obvious rectal prolapse on history or exam does not reliably exclude this diagnosis. Endoscopic biopsies are usually diagnostic.

4.2.7. Other causes

Jejunal and rarely mesenteric varices may cause significant lower GI bleeding [Hansen,1990; Yuki,1992]. Accurate diagnosis requires angiography and therapy is portal decompression. Rectal varices may also be a cause of significant hematochezia and are readily detected on sigmoidoscopy.

Small bowel diverticula are uncommon and rarely lead to lower GI bleeding [Akhrass,1997]. Duodenal diverticula are more common than jejunoileal diverticula. Approximately 5% of both duodenal and jejunoileal diverticula may bleed [Akhrass,1997]. Meckel’s diverticulum is the most common congenital anomaly of the small intestine, occurring in up to 4% of the general population [Arnold,1997], and is the most frequently recognized cause of bleeding small bowel diverticula. It represents a persistence of the Vitelline duct and often has heterotopic gastric mucosa resulting in ulceration and bleeding. The roles of H. pylori infection and NSAID ingestion in the pathogenesis of bleeding in patients with Meckel’s diverticulum remain uncertain [Mathur,1992; Bemelman,1993; Cserni,1996]. The diagnosis is made on a "Meckel’s scan," a technetium pertechnetate scan where ectopic gastric mucosa is identified by taking up radiolabeled technetium. Definitive treatment with diverticulectomy or segmental intestinal resection is necessary [Cullen,1996].

Small intestinal ulcers are found on autopsy in 8% of patients taking NSAIDs at the time of their death [Allison,1992]. NSAID induced small bowel and colonic ulcers should be considered in every patient with lower GI bleeding who is taking these drugs.

Small bowel neoplasms are a particularly uncommon cause of GI bleeding; in one study of 258 patients with bleeding of obscure etiology (normal upper endoscopy, colonoscopy and small bowel series), small bowel tumors were diagnosed by enteroscopy in only 5% [Lewis,1991].


Occult GI bleeding manifests as either iron deficiency anemia or a positive test for fecal occult blood, and should be evaluated with endoscopy. As previously discussed, barium studies have limited sensitivity for many sources of GI bleeding and should not be utilized as the primary diagnostic modality in patients with occult GI bleeding. Guidelines for the evaluation of patients with occult GI bleeding are provided in Algorithm 2.

5.1. Iron-deficiency anemia

The initial evaluation of iron deficiency anemia should be guided by the patient’s symptoms. If the patient is asymptomatic, the evaluation should begin with a colonoscopy, particularly if the patient is age 50 years or older. If the initial endoscopy (whether EGD or colonoscopy) is normal, the alternate exam should be performed. In the patient whose initial endoscopy reveals a source of occult GI bleeding, it is difficult to make definitive statements regarding the yield of further endoscopic evaluation. While some authors report a low yield of further endoscopy in this setting [Rockey,1993], others have found lesions in both the upper and lower GI tracts in the setting of iron-deficiency anemia [Zuckerman,1992; McIntyre,1993]. Clearly, if the lesion seen on initial endoscopy is not deemed "significant" enough to be the source of iron deficiency anemia, further endoscopic evaluation should be performed. If no lesions are identified on endoscopy, a duodenal biopsy should be considered at the time of EGD in order to rule out iron malabsorption due to celiac sprue and other small bowel disorders. Finally, if all tests are negative, a small bowel series or enteroclysis is usually recommended, although the yield of these studies is quite low [Rockey,1993; McIntyre,1993].

5.2. Positive fecal occult blood testing

When a positive fecal occult blood test is obtained in the context of evaluating upper abdominal symptoms, an EGD should be the first diagnostic study. If the EGD is negative, patients who are age 40 or older should undergo colonoscopy. Patients with positive fecal occult blood testing obtained in the context of screening for colorectal cancer should first undergo evaluation with colonoscopy. If the colonoscopy is normal, it is reasonable to conclude the evaluation unless the patient has previously unrecognized upper GI symptoms, is anemic, or fecal occult blood remains positive on subsequent testing. This approach can be recommended based on the high false positive rate of a single fecal occult blood test and its limited ability to detect bleeding from an upper GI source. On the other hand, some studies have reported a "significant" upper GI source of occult bleeding in up to 30% of patients with hemoccult-positive stool and normal colonoscopy [Zuckerman,1992; Hsia,1992]. More recently, Rockey and colleagues reported that 60% of patients with hemoccult-positive stool on rectal exam or routine screening had an upper GI tract lesion [Rockey,1998]. The higher prevalence of upper GI lesions in this study may in part be due the inclusion of positive hemoccults on rectal examination, since the latter is ostensibly being done as part of a physical examination in an individual with GI symptoms. However, although the presence of upper GI symptoms was predictive of a positive finding on EGD, more lesions were found in the upper GI tract than the lower GI tract in both patients with upper GI symptoms and those without symptoms. Until these findings are confirmed, the approach recommended above is likely to remain the standard.

NSAID use in the context of hemoccult-positive stool warrants some mention. Although NSAID use is frequently associated with GI mucosal injury, a positive test for fecal occult blood should not be attributed to NSAID use. Furthermore, even if NSAID-related GI injury is found on endoscopy, such lesions are common enough that they cannot be presumed to be the main source of GI blood loss. Such patients may have clinically important lower GI lesions and should undergo evaluation of the lower GI tract with colonoscopy [Bahrt,1984; Pye,1987].


Obscure GI bleeding is defined as recurrent acute or chronic GI bleeding for which no source has been found despite evaluation with EGD and colonoscopy with or without routine small bowel follow-through. The optimal diagnostic study in such patients is not certain, but additional studies should be considered in patients who require blood transfusions or repeated hospitalizations for obscure GI bleeding. In general, EGD and colonoscopy are repeated once in such patients in order to exclude lesions that may have been missed on the prior endoscopic exams. Further diagnostic options include enteroscopy (endoscopy to approximately the midjejunum using an enteroscope or pediatric colonoscope), capsule endoscopy, small bowel enteroclysis, or 99mTc-labeled red blood cell scintigraphy. Enteroscopy can also be performed intraoperatively, where the entire small bowel is "sleeved" over the enteroscope and the mucosa is visualized upon withdrawal. This approach is most useful in patients with significant bleeding who are candidates for bowel resection. Guidelines for the evaluation of patients with obscure GI bleeding are provided in Obscure Bleeding Algorithm.

Patients who present with recurrent acute bleeding should undergo a 99mTc-labeled red blood cell scan while actively bleeding. Although this study is not always accurate with respect to the site of bleeding, it is noninvasive and may give valuable information either when used alone or in combination with other studies. Patients with a positive red blood cell scan can then undergo mesenteric angiography for more definitive diagnosis as well as possible therapy. If the bleeding site identified on red blood cell scan is within the reach of an endoscope (via EGD, enteroscopy or colonoscopy), confirmation can be sought endoscopically rather than angiographically and endosocopic therapy can be applied at the same time. Finally, patients who are being considered for exploratory laparotomy with or without enteroscopy also should undergo mesenteric arteriography. Lesions such as vascular anomalies or vascular tumors may be visualized, even in the absence of active bleeding.

Patients with slow, chronic blood loss should initially undergo enteroclysis as it is simple and noninvasive. In two recent case series, 20% of patients with obscure GI bleeding and negative EGD and colonoscopy had a positive finding on enteroclysis, and lesions were surgically proven in the majority of cases [Rex,1989; Moch,1994].

Similarly, case series of patients undergoing enteroscopy for obscure GI bleeding have reported positive findings in 38% to 47% of patients [Foutch,1990; Schmit,1996]. Not uncommonly, the site of bleeding is found in the stomach or proximal small bowel, within reach of the standard upper endoscope [Foutch,1990; Schmit,1996; Berner,1994]. Finally, capsule endoscopy may be useful in patients with obscure GI bleeding (see Section for a discussion of this technique). In one recent study, video capsule endoscopy was superior to small bowel radiographs in the evaluation of 13 patients with obscure GI bleeding; 12 of 13 capsule endoscopy studies yielded a source of bleeding, while only 1 of 13 small bowel radiographs was positive [Costamanga,2002]. Angiodysplasia was the most common finding. As previously mentioned, a small bowel stricture should be excluded with a radiographic study before proceeding to capsule endoscopy.


In general, most patients presenting with GI bleeding will need referral to a gastroenterologist for a definitive diagnosis of the bleeding source. However, young patients with minor bleeding, particularly from an obvious anorectal source, may be managed without the assistance of a gastroenterologist. General guidelines for referral are provided in (Table 5).