Laboratory Examination

Initial laboratory studies should include a complete blood count, serum urea nitrogen, creatinine, coagulation profile (prothrombin time, INR, and partial thromboplastin time), liver tests, blood type and cross-match, and electrolytes. As discussed above, positive fecal occult tests (guaiac-based tests and immunochemical fecal occult blood tests) may favor LGIB. The blood urea nitrogen–to–creatinine ratio has been used as a noninvasive test to help distinguish UGIB from colonic sources of bleeding (59–61). In the study of Chalasani et al. (59), a ratio of 33 or higher had a sensitivity of 96% for UGIB, although overlap was observed with LGIB, especially in patients with UGIB without hematemesis.

In the ICU, patients suspected of having GI blood loss secondary to findings of persistent anemia will have a fecal occult blood study as the first investigative step. While there are numerous types of fecal occult blood tests, the test used determines the location in the GI tract where blood is likely to be detected (62).

Endoscopic Approach: General

Safety of Endoscopic Procedures for Critically Ill Patients. Colonoscopy remains the procedure of choice for evaluating patients with acute LGIB, and enteroscopy is considered for proximal small bowel bleeding. However, for ICU patients with acquired bleeding and hemodynamic instability, questions remain due to the possibility of severe comorbid illness, including respiratory distress, aortic aneurysm, life-threatening dysrhythmias, AMI, history of GI tract perforation or operation, severe lower GI obstruction, and bleeding tendency, to list only a few conditions. Colonoscopy complication rates in LGIB of the general population or patients admitted to the ICU primarily for hemorrhage are low, and the bowel preparation itself appears to be safe (71–73). Zuckerman and Prakash (74) reviewed 13 studies and found an overall complication rate of 1.3%. A study of 55 ICU patients with acquired LGIB demonstrated an acceptable diagnostic rate of 67% (37 of 55 cases) without procedure-related complications, suggesting that bedside colonoscopy after preparation is safe for the critically ill (5).

Preparation before Endoscopic Procedures. In patients with massive active bleeding, colonoscopy is often frustrating, nonproductive, and sometimes dangerous. However, some clinicians still consider it the first diagnostic maneuver, even in the patient with severe ongoing bleeding (48,75). Traditionally, colonoscopy for LGIB was delayed because of the need for bowel preparation and the fear of increased procedural complications. Indeed, urgent colonoscopy in an unprepared colon can be challenging, or even dangerous. Good bowel preparation is important for an adequate and sensitive colonoscopy. Studies of urgent colonoscopies performed in an unprepared colon to evaluate for LGIB revealed completion rates as low as 35% (76,77). Jensen et al. (78) studied patients with severe diverticular hemorrhage, noting that completion rates may reach up to 100% if aggressive bowel preparation is performed before urgent colonoscopy. Because of variable comorbid conditions, such as decreased bowel motility, obstruction, and electrolyte disturbance, colon preparation in ICU patients may not be as thorough as in the general population. There are limited studies of the effectiveness of colon preparation for LGIB in ICU patients. Lin et al. (5) noted that for patients with LGIB occurring after admission to ICU, the reach rate of the cecum with a colonoscope was 58% after enemas or an oral polyethylene glycol (PEG) solution were administered.

In spite of providing a relatively feces-free colon, old methods for colon preparation with clear liquids, laxatives, and enemas or peroral gut lavage 48 to 72 hours before colonoscopy are time consuming, uncomfortable, and inconvenient for patients. For those in the ICU or with active LGIB, these methods are clearly not useful. Peroral gut lavage with saline or balanced electrolyte solutions has been proposed and was found to provide rapid, effective cleansing of the colon; however, the method is not tolerated in 11% of patients due to the high fluid volume—7 to 12 L—and might cause fluid and electrolyte disturbances (79). Recently, development of osmotically balanced solutions may provide minimal water absorption or secretion into the bowel lumen. Generally speaking, there are iso-osmotic preparations, hyperosmotic preparations, and stimulant laxatives for colon cleansing. Polyethylene glycol–electrolyte lavage solution (PEG-ELS) is an isotonic, nonabsorbable electrolyte solution that clears the bowel by washing out ingested fluid without significant fluid and electrolyte shifts. Rapid purge is best accomplished with PEG-based solutions. They can be administered by a nasogastric tube or by drinking 1 L every 30 to 45 minutes, to a median dose of 5.5 L (range 4 to 14 L); 3 to 4 hours are required to cleanse the colon (80). Two liters of PEG-ELS plus bisacodyl or magnesium citrate have also been suggested for colon preparation (81). The studies comparing a standard 4-L PEG-ELS with 2-L PEG-ELS with either a magnesium citrate or bisacodyl preparation have shown equal efficacy for colon cleansing (82–84). With PEG-ELS, 5% to 15% of patients have difficulty drinking the large amounts of fluid, or they develop symptoms such as nausea, vomiting, abdominal fullness, and cramps, leading to incomplete colon preparation (85,86). Hyperosmotic preparations, including monobasic and dibasic sodium phosphate (NaP) and magnesium citrate, draw plasma water into the bowel lumen to promote evacuation. In addition to hyperosmotic action, magnesium citrate also stimulates fluid secretion and intestinal motility through the action of cholecystokinin. Fleet Phospho-Soda (C. B. Fleet Co., Lynchburg, VA) is a poorly absorbed salt that produces intestinal fluid retention due to the osmotic load, and causes fluid evacuation from the bowel. Patients often prefer NaP to PEG-ELS because there is much less fluid to drink. Liquid and tablet forms of NaP are currently available. Stimulant laxatives include bisacodyl and senna extract. The former is a poorly absorbed diphenylmethane that stimulates colonic peristalsis, and the latter contains anthracene derivatives that may be metabolized by colonic bacteria into substances that enhance colonic motility.

The timing of colon preparation varies among institutions. Early methods required preparation 48 to 72 hours before the procedure; of course, this is not ideal for ICU patients who need urgent interventions. Some institutions administer the solution on the day of the examination, even though the manufacturer recommends taking the medication the day before the procedure (81). For the critically ill patient with LGIB, there are no data on the “proper timing” of colon preparation. Further, preparation may be incomplete due to poor motility or ileus. Chiu et al have suggested that PEG-ELS on the day of procedure may be adequate (81).

The safety of colon preparation is important for critically ill patients with comorbid illness; of particular concern are fluid overload and electrolyte disturbances. PEG-ELS causes no significant change in weight, vital signs, serum electrolytes, or complete blood count (87–89). It is a relatively safe colon preparation solution for patients with electrolyte imbalance, advanced liver disease, poorly compensated congestive heart failure, or renal failure, although in the study of Granberry et al. (90), exacerbation of congestive heart failure after PEG-ELS administration was noted. NaP, a hyperosmotic preparation, may cause alterations in serum electrolytes and extracellular fluid status (91,92). Asymptomatic hyperphosphatemia is seen in up to 40% of patients, but clinically significant hyperphosphatemia is rare and usually limited to patients with renal failure (93–95). Twenty percent of patients had abnormally low serum potassium levels after bowel preparation with NaP (91). It is suggested that NaP is contraindicated in patients with renal failure, acute myocardial infarction (MI) or unstable angina, congestive heart failure, ileus, intestinal malabsorption, and significant ascites. Gremse et al. (95) reported that the degree of asymptomatic hyperphosphatemia in children was greater than in adults, and recommended avoiding NaP in children with renal failure, congestive heart failure, ileus, and ascites.

Colonoscopy for Critically Ill Patients. Three primary diagnostic tools for LGIB are colonoscopy, radionuclide scintigraphy, and mesenteric arteriography. Advances in endoscopic technology have brought colonoscopy to the forefront of the management of LGIB. Colonoscopy as the first choice for occult or stable LGIB is not in dispute. However, for brisk LGIB or LGIB with hemodynamic compromise, whether the attending physicians consider colonoscopy as the first diagnostic maneuver for brisk LGIB or LGIB with hemodynamic compromise is controversial. Colonoscopy in patients with severe hematochezia is impractical because of inadequate visualization caused by brisk blood loss (2,74). Some are reluctant to perform colonoscopy in hemodynamically unstable patients with ongoing bleeding, suggesting that these patients are best served by urgent angiography, perhaps in conjunction with surgical consultation (80,96). However, there is a reason to favor colonoscopy for acute LGIB. In addition to affording a rapid diagnosis, colonoscopy may indicate specific therapy when (and if) the bleeders are found (78,97–99), although the rate for intervention in an ICU study was low (5). Rapid endoscopic identification of a bleeding source, regardless of whether therapy is administered, may contribute to the clinical management of recurrent bleeding, if it occurs. Finally, compared with angiography, urgent colonoscopy has a higher diagnostic yield and a lower complication rate. In a retrospective study of 107 patients with severe LGIB, colonoscopy was diagnostic in 90% of patients and angiography in 48%; the former was therapeutic in 12% versus 22% for the latter (100). In another study, the diagnostic yield of colonoscopy was 82% versus 12% for angiography (48). Interestingly, most patients undergoing radiographic evaluation for LGIB—regardless of findings and interventions—will subsequently require a colonoscopy to establish the cause of bleeding. In addition to diagnosis, occasional therapy, and management planning, earlier colonoscopy does contribute to a shorter length of hospital stay (101). Hemodynamic instability, higher comorbidity, performance of a tagged red blood cell nuclear scan, and surgery for hemostasis were significantly associated with a decreased likelihood of discharge (101).

While most episodes of LGIB will stop spontaneously, 10% to 15% of patients undergoing urgent colonoscopy received endoscopic therapy. The lesions most amenable to colonoscopic treatment of LGIB, in most studies, are angiodysplasia or diverticulosis. Once it is identified as the source of bleeding, angiodysplasia is usually coagulated by methods including the following (5,78,102,103):

• Injection therapy (epinephrine, saline, or ethanol)
  
• Heater probe
  
• Monopolar and multipolar electrocoagulation
  
• Argon plasma coagulation (APC)
  
• Hemoclips
  
• Band ligation

In a study of ICU patients with nonprimary LGIB, spontaneous cessation occurred in 53% of patients; 29% achieved hemostasis with endoscopy, but had a higher rate (19%) of recurrent bleeding (5). Other studies, not limited to ICU patients, have reported rebleeding rates of 13% to 53%, and many patients may require more than one treatment (102,103).

Timing of Colonoscopy

The use of colonoscopy is controversial for critically ill patients with LGIB. Some believe that colonoscopy is best utilized in patients whose bleeding has stopped or slowed down (80,96), and others agree with early colonoscopy for LGIB. Evidence suggests that earlier intervention leads to more diagnostic and therapeutic opportunities (78,104). It has further been noted that early colonoscopy reduces the length of hospital stay, and therefore should decrease treatment costs (91,92). Urgent colonoscopy after bowel preparation with endoscopic treatment of patients with active diverticular bleeding or stigmata of bleeding has been shown to be highly effective in decreasing the need for surgical intervention (78,80).

In conclusion, prompt intervention may decrease the need for surgical exploration, as well as the rate of recurrent bleeding and length of hospital stay. Who needs urgent colonoscopy? Early identification of high-risk patients would allow the more selective delivery of urgent therapeutic interventions to those who will benefit. Clinical high-risk predictors in the first hour of evaluation in patients with severe LGIB have been proposed, and included an initial hematocrit of no more than 35%, the presence of abnormal vital signs 1 hour after initial medical evaluation, and gross blood on initial rectal examination (6). How early should the urgent colonoscopy be performed? The definition of urgent colonoscopy varies widely in the literature—from within 8 to 24 hours of presentation (5,48,78,105–109). Most definitions consider the procedure within 12 to 24 hours; more recently, the literature defines urgent colonoscopy as within 12 hours (48,78).

Sigmoidoscopy or Colonoscopy?

In studies of the general population with LGIB, diverticular and angiodysplastic bleeding are the most frequent events. Although anatomically prone to be located in the left colon, bleeding most often occurs in the right colon; thus, most endoscopists prefer a total colonoscopy. Flexible sigmoidoscopy can be performed in the initial evaluation of patients with LGIB, but the diagnostic yield of flexible sigmoidoscopy in LGIB is low, ranging from 9% to 58% (10). Regardless of presentation, flexible sigmoidoscopy may miss serious proximal pathology (110). However, in the study of ICU patients with acquired bleeding, 78% of responsible lesions were in the left colon (5). If the critical situation precludes the use of total colonoscopy, sigmoidoscopy as the first maneuver may be acceptable for this patient group. Nonetheless, unless a definite and compatible bleeding source is identified with flexible sigmoidoscopy, the study of LGIB should proceed to a full colonoscopy in most patients.

In 2005, the ASGE offered the following guidelines regarding LGIB (111):

• Colonoscopy is effective in the diagnosis and treatment of LGIB (prospective controls).
  
• Colonoscopy is recommended in the early evaluation of severe acute LGIB (prospective controls).
  
• Thermal contact modalities, including heat probe, and bipolar/multipolar coagulation and/or epinephrine injection can be used in the treatment of bleeding diverticula, vascular ectasia, or postpolypectomy bleeding (prospective controls).

Enteroscopy and Capsule Endoscopy for Critically Ill Patients

The small bowel has traditionally been a problematic area to evaluate because of the long length, looping, free intraperitoneal location, active contractility, and limits of standard endoscopy. It is estimated that 10% to 25% of LGIB originates in the small bowel and can pose a diagnostic dilemma for clinicians (112–114). Small bowel sources account for 0.9% to 7% of cases presenting with blood per rectum (48,100,115), and comprises approximately 5% of obscure GI bleeding (116). Upper endoscopy and colonoscopy appear to have limited roles in the investigation of small bowel bleeding, and are only useful when the bleeding source is the duodenum or the most distal segment of the small intestine (terminal ileum), respectively (55). Some clinicians have utilized peroral intubation with a standard colonoscope, reaching a point 20 to 60 cm distal to the ligament of Treitz; this may increase the diagnostic yield of GI bleeding of obscure origin by 17% to 46%—not an insignificant proportion (13,117–119). Small bowel enteroscopy is currently the best endoscopic investigative modality. Indeed, it has become the cornerstone of management in patients with obscure GI bleeding. Current tools for ruling out small bowel diseases include push-type enteroscopy (PE), double-balloon enteroscopy (DBE), intraoperative enteroscopy, and capsule endoscopy.

Small bowel evaluation may be performed with PE, which allows endoscopic evaluation of the proximal 60 cm of the jejunum or 150 cm distal to the pylorus. PE is probably the most commonly performed small bowel procedure today, and is often pursued when upper GI endoscopy and colonoscopy have failed to find the source for blood loss. PE may be performed using an enteroscope or pediatric colonoscope. The obvious limitation of push enteroscopy is the inability to reach lesions distal to the middle jejunum. PE can only examine a relatively short portion of the small bowel, even under fluoroscopy, and the true depth of insertion is unreliable. Some experts consider it complementary to capsule endoscopy, and believe that it should be performed only if capsule endoscopy or other modalities are positive for a proximal small bowel lesion. The use of an overtube to prevent looping of the instrument in the stomach increases the insertion depth by 10 to 25 cm (120). Prototype, variable stiffness enteroscopes are emerging in an attempt to achieve maximal insertion depth without the use of an overtube (121,122). Although PE cannot investigate the entire small bowel, an important benefit of PE is the ability to provide diagnostic and therapeutic capabilities with one procedure if the bleeder can be found.

The diagnostic yield of PE is between 38% and 65% of patients in whom upper and lower endoscopies are negative (123). Multivariate analysis in a retrospective, two-center study by Lepère et al. (124) showed that melena and chronic renal failure increase the diagnostic yield of PE in patients with unexplained GI bleeding. The positive findings noted with PE for patients with renal failure were most often in the distal duodenum or jejunum, including ulcers (17%) and AVMs (41%). Others have shown similar results in that most small bowel lesions diagnosed by PE are vascular in nature (angiodysplasia/AVM) (125–127), followed by ulcerations and malignancies (128). A delay between the bleeding and PE (less than or more than 4 days) and a history of recurrent intestinal bleedings before PE were not associated with positive findings; thus, rapid performance of PE may not be necessary, except in patients with continuous active bleeding (124). Another benefit of PE is that it provides a “second look” for lesions that may have been missed on original endoscopy. Interestingly, 25% to 40% of lesions found on PE are within reach of a standard upper endoscopy (129,130). Complications of PE are infrequent, occurring in less than 1% of cases. Most complications, including bleeding and perforation, are related to the use of an overtube (131,132). The data of effectiveness and safety of PE for ICU patients are limited.

Even when PE reaches a maximal depth of insertion of 160 cm below the ligament of Treitz, there is still over 250 cm of small intestine remaining unexamined. It is possible to close this gap with a newly developed DBE system (133). The newest modality for imaging the small bowel was introduced in 2001, when Yamamoto et al. (134) reported their results using a double-balloon method in four patients. The DBE represents the first successful provision of both diagnostic and therapeutic intervention to the entire small bowel. The goal is to reach the ileocecal valve, but this often is not possible. The total inspection of the small bowel is usually attainable with the peroral approach and retrograde approach per rectum. On average, approximately 250 cm is achieved via the oral route and 130 cm via the anal route, with a mean examination time of 75 minutes (135). If total enteroscopy is necessary, it can be achieved in 60% to 86% of cases, depending on the experience of the endoscopist (136,137). Yamamoto et al. (136) performed 50% of the studies in an antegrade fashion, with the remainder being retrograde. The bleeding source was found in 76%, and hemostasis using electrocautery was performed successfully in 18%. Multicenter experience with DBE in the United States showed that the mean procedure time was 115 minutes, and the yield of DBE for a GI bleed ranged between 52% and 75% (135,138). Oral DBE requires no specific preparation other than a 6- to 8-hour fast before the procedure. If a retrograde (anal) approach is undertaken, standard colonic preparation is necessary; conscious sedation or general anesthesia may be utilized. Endoscopic hemostasis using injection therapy, APC, electrocautery, and hemoclipping may be used as is done in routine upper GI endoscopy and colonoscopy. Complications of DBE are noted in 1.1% to 8.5% (136,138), and include aspiration pneumonia, abdominal pain, perforation, and acute pancreatitis (135,136,138–140). The effectiveness and safety of DBE for critically ill patients are unknown.

Before the advent of capsule endoscopy and DBE, intraoperative endoscopy was the only way to detect and treat lesions beyond the reach of PE. Now the role of diagnostic intraoperative enteroscopy is likely to decrease with the introduction of the less-invasive capsule endoscopy and DBE; however, it still plays a role in specific clinical situations. Intraoperative endoscopy is performed in conjunction with a surgeon in the operating room, and with the patient under general anesthesia. Intubation may be achieved transorally, transanally, or through an operative enterotomy, depending on the clinical circumstance and physicians’ preference.

During laparotomy combined with intraoperative enteroscopy, the endoscopist carefully inspects the intestinal lumen with a push enteroscope, while the surgeon slowly guides the bowel over the endoscope using the air-trapping technique and examines the external wall with palpation and transillumination (141). Lesions can be treated endoscopically or marked with a tattoo for surgical resection. The terminal ileum is reached in more than 90% of patients (142). The yield in detecting bleeding lesions reaches 70% to 100% (55,143), making intraoperative endoscopy the most sensitive method of diagnosing small bowel disorders. However, the high sensitivity comes at the cost of extreme invasiveness, making it a procedure of last resort. The complication rate is estimated at about 3%, including mucosal tears and bleeding of the mesentery due to traction (138).

Capsule endoscopy is a safe and promising diagnostic tool for GI bleeding of unknown origin, focusing especially on the small bowel; it may obviate the need for angiography in some difficult patients. The idea of wireless imaging of the small intestine was conceived simultaneously by Paul Swain, a British gastroenterologist, and Gavriel Iddan, an Israeli scientist. They merged research efforts in 1998 and soon developed a pill-sized camera with sufficient battery life to image the entire small intestine (144). Capsule endoscopy was introduced into clinical practice in 2001 and made it possible for the first time to visualize intraluminal conditions throughout the entire small bowel. The first commercially available video capsule (Given) is composed of three main subsystems: an ingestible capsule endoscope, a data recorder, and a workstation. The capsules are equipped with a miniaturized image-capturing system, battery, light source, and transmitter. After an overnight fast, the patient swallows the capsule, which travels through the GI tract by means of the actions of normal peristalsis. The capsule device captures two images per second and has a battery life of approximately 8 to 12 hours. Captured images are transmitted by a digital radiofrequency communication channel to an external data recorder unit.

Studies comparing capsule endoscopy with other diagnostic procedures, including enteroclysis, PE, computed tomography (CT) scan, and intraoperative enteroscopy, showed that capsule endoscopy was clearly superior in the diagnosis of occult/obscure/overt small bowel bleeding (132,145–150). Capsule endoscopy has proven superior to enteroclysis (147–149) and PE (146). Van Gossum et al. (151) noted that, with obscure GI bleeding, no significant difference in diagnostic yield was found between push and wireless-capsule endoscopy. Capsule endoscopy, compared to CT scanning, is reported to be superior in detecting small bowel lesions (147,149).

The overall diagnostic yield rate of capsule endoscopy in patients with GI bleeding ranges from 45% to 66% (145,147,152–154). The timing of capsule endoscopy has been addressed in two studies and appears related to diagnostic yield. When administered to patients with ongoing overt bleeding, the diagnostic yield is higher—87% to 92%—than in those with previous overt bleeding or iron deficiency anemia—46% to 56% (152,155). Hartmann et al. (150) noted that capsule endoscopy identified lesions in 100% of patients with ongoing overt bleeding, 67% of patients with previous overt bleeding, and 67% of patients with obscure/occult bleeding. To increase the diagnostic yield of capsule endoscopy, prokinetic agents (156), oral bowel preparation (oral sodium phosphate and polyethylene glycol), simethicone, and erythromycin have been recommended. Prokinetic agents prompt the passage of the capsule and prevent the exhaustion of batteries before study completion. Bowel preparation with oral sodium phosphate has been suggested to offer better visualization than overnight fasting alone, and is associated with fewer disturbances by intraluminal turbid fluid (157). However, there are concerns that increasing bowel motility may result in missing a lesion. Fireman et al. (158) showed that erythromycin markedly reduced gastric emptying time and had a negative effect on the small bowel images. Preparation of elderly subjects with PEG or sodium phosphate also had a negative effect on small bowel transit time. PEG increased the visibility in the proximal small bowel in one study, but had no effect in the second investigation (159,160). Presently, bowel preparation is preferred by most practitioners. Erythromycin leads to faster gastric emptying at the expense of small bowel transit time and poorer visualization (161). Bowel preparation with simethicone, which can decrease intraluminal gas bubbles, resulted in significantly better visibility (162). Simethicone may be added to the routine preparation for capsule endoscopy to improve visualization of the small bowel mucosa. A delay in bowel transit time may result in an incomplete study due to capsule battery drainage.

Despite the higher diagnostic yield, capsule endoscopy limitations are evident: biopsy specimens cannot be obtained, therapeutic intervention cannot be performed, and localization of some lesions is imprecise (163). Capsule endoscopy, however, is regarded as a low-risk procedure that is well tolerated. The primary risk with capsule endoscopy is capsule entrapment within the GI tract; this occurs in 0.75% to 5% of cases. Most entrapment occurs in the small intestine, although case studies report impaction at the cricopharyngeus, tracheal aspiration, and retention in diverticula (10). Risk factors for entrapment include NSAID-induced strictures, prior abdominal radiation, Crohn enteritis, prior major abdominal surgery, and known diverticula. A trapped capsule may be retrieved endoscopically or surgically. In cases where the colon is not visualized on capsule endoscopy and the patient does not see the capsule pass, an abdominal radiograph should be obtained to document passage. Absolute contraindications to its use include GI obstruction and pseudo-obstruction—meaning ileus. Relative contraindications include a history of GI motility disorders, such as gastroparesis; history of intestinal strictures or fistulae; pregnancy; history of multiple small bowel diverticula; history of Zenker diverticulum; history of abdominal surgeries or radiation; and an active swallowing disorder or dysphagia. Although there is concern about the use of capsule endoscopy in patients with pacemakers, new evidence suggests that capsule endoscopy may be safely utilized in these patients (164).

Capsule endoscopy has been reported to change patient management in up to 75% of cases (146), although the studies focused solely on small bowel lesions. Colonic bleeding is difficult to evaluate via capsule endoscopy because of retained stool, limited battery life, and poor visual field due to the colon’s large diameter. A recently developed PillCam Colon capsule endoscopy appears promising for colonic evaluation (165), although in comparison to conventional colonoscopy, false-positive findings were recorded in 33% cases (166).

The use of capsule endoscopy in critically ill patients has been limited. These patients often cannot ingest the capsule by themselves, especially if endotracheally intubated; an endoscopic technique of capsule placement has been described for such patients (167). In addition to swallowing problems, bowel transit time may be delayed due to sepsis, electrolyte imbalance, medication use, and anatomic changes due to surgeries. Abnormal bowel transit time affects the diagnostic yield of capsule endoscopy. The utilization of capsule endoscopy for small bowel or obscure bleeding must be made on a case-by-case basis.

Nuclear Medicine

As mentioned previously, for LGIB, endoscopy, nuclear medicine, and mesentery angiography are three main diagnostic modalities. The two techniques of radionuclide scanning commonly use either ^99mTc-labeled red blood cells or technetium-99m sulfur colloid. A ^99mTc-labeled red blood cell scan is the preferred technique, with images that can be detected for up to 12 to 24 hours after injection. If the rate of bleeding is insufficient to give an immediate positive test, or if the bleeding is intermittent, the labeled red blood cells can sometimes accumulate to detect the site of bleeding—when rescanned—up to 24 hours after injection. This technique can detect bleeding at a rate as low as 0.1 to 0.5 mL/min, and is thought to be a sensitive diagnostic tool for LGIB (25). ^99mTechnetium sulfur colloid is rapidly cleared by the reticuloendothelial system after injection, with a half-life of only 2 to 3 minutes. Therefore, if there is no active bleeding when administered, the ^99mtechnetium sulfur colloid is quickly cleared, with a resultant nondiagnostic test.

Angiography

Angiography, first employed in the diagnosis of GI bleeding more than 40 years ago, provides imaging of the entire mesenteric system, localizes the sites of hemorrhage, and affords the opportunity for transcatheter interventions. It now holds an established place in dealing with difficult GI bleeding, both for diagnosis and treatment. Mesenteric angiography is more invasive than technetium-labeled red blood cell scanning, and requires a bleeding rate of at least 0.5 to 1.0 mL/min to detect bleeding (192). Unfortunately, bleeding is frequently intermittent and may occur at a much lower rate, resulting in the inability to detect the causative lesion (193). Angiography is usually undertaken when patients have clinical indicators of severe bleeding (e.g., tachycardia and/or syncope). Although colonoscopy is the diagnostic modality of first choice for LGIB, many endoscopists are reluctant to perform colonoscopy in hemodynamically unstable patients with ongoing bleeding; these patients usually undergo radiographic studies. In addition, the bleeding of colonic lesions, such as vascular abnormalities, can be too massive for colonoscopic visualization, thus precluding the procedure. In some centers, a radionuclide scan is requested before mesenteric angiography, because a negative radionuclide scan is unlikely to have a positive angiogram. Angiography has been reported to be especially useful in patients presenting with postoperative GI hemorrhage (194).

Angiography will localize the site of bleeding in 40% to 86% of patients with LGIB (113,182,195). Even if a bleeding site is identified on angiography, localizing the site intraoperatively can be difficult; angiography has a specificity of 100% but a sensitivity of only 30% to 47% (196). Diverticula and angiodysplasia are the most common findings when angiography is positive, with 50% to 80% of the bowel bleeding sites being supplied by the superior mesenteric artery (197). Diverticular hemorrhage is most likely to produce extravasation on angiography (198). Following the injection of contrast media, bleeding and nonbleeding angiodysplastic lesions are characteristically seen as ectatic slowly emptying veins, vascular tufts, or small veins, with early filling in the arterial phase (199). Angiography is more sensitive than colonoscopy for detecting angiodysplasia (200), and when angiography identifies a bleeding site, treatment with embolization therapy or directed infusion of vasopressin may be performed. The overall rate of complication for mesenteric angiography is similar to most selective angiography, and is acceptable at less than 5% (201). Complications include hematoma or bleeding at the catheter site; access site thrombosis; contrast reactions; injury to the target vessels, including dissection and distal embolization; and renal failure (202). The injured vessels usually involve the SMA, IMA, and celiac artery.

In addition to its diagnostic role, angiography offers therapeutic possibilities via pharmacologic vasoconstriction or selective embolization (transcatheter arterial embolization [TAE]), and therefore may reduce the need for surgical resection. Once the bleeder is confirmed with contrast injection, embolization of the vessel is performed, usually with one of three embolic agents: microcoils, polyvinyl alcohol sponge particles, or gelatin sponge particles, alone or in combination (203). Pharmacologic vasoconstriction is achieved with intra-arterial vasopressin infusion.

TAE may be a more definitive means of controlling bleeding, but is associated with a risk of intestinal infarction. Selective embolization initially controls bleeding in up to 100% of patients, but rebleeding rates have been reported to be 15% to 40% (198,199). The major complication rate was 10% to 20% and included dysrhythmias, pulmonary edema, hypertension, and ischemia (204,205). The bowel infarction or colonic necrosis rate from embolization ranged from 10% to 20% (206–208). Superselective TAE may decrease the incidence of ischemia and rebleeding (207,209,210). A literature review of 144 cases by Kuo et al. (203) showed a minor complication rate of 9% and 0% for major complications.

In the past, embolization has been reserved for treatment of UGIB, whereas LGIB has been controlled with vasopressin infusion. The reason is based on reports in the older literature in which infarction frequently occurred after LGI embolization. With advances in superselective embolization techniques, clinically significant bowel ischemia has become an uncommon complication (211,212). Although the efficacies of vasopressin and embolization are reasonably comparable, embolization allows more rapid completion of therapy and a decreased likelihood of systemic complications. Embolization should be considered a primary option for LGIB, although vasopressin is still preferable for diffuse lesions and cases in which superselective catheterization is not technically possible.

Pharmacologic vasoconstriction for LGIB involves an intra-arterial infusion of vasopressin, started at a rate of 0.2 units/min. If the bleeding continues, the rate of infusion can be increased up to maximal dose of 0.4 units/min. A repeat angiogram can be performed after 20 to 30 minutes to assess whether the bleeding is continuing or slowing down. If the bleeding seems to stop, infusion continues at the same rate for 12 hours, and subsequently, the dose of vasopressin is decreased by 50% provided that no bleeding recurs. After 12 hours of only saline infusion, the catheters are removed (168). Success rates for hemostasis are variable, with some reports as low as 36% and others as high as 100% (113,182,213,214). Bleeding recurrence is high, and may occur in up to 50% of patients after cessation of the infusion (182). Vasopressin should not be used in patients with significant coronary artery disease (CAD) or peripheral vascular disease; mesenteric thrombosis, intestinal infarction, and death have been reported with its use (113). During the vasopressin infusion, patients need to be in an ICU setting where they can be monitored for myocardial, bowel, and peripheral ischemia; hypertension; dysrhythmias; and hyponatremia. Nitroglycerin reverses the vasopressin-induced coronary vasoconstriction without affecting the therapeutic vasoconstriction of the mesenteric artery (215). If standard angiography is negative, provocative angiography has been suggested with anticoagulants, vasodilators, and thrombolytics; of course, their use may cause bleeding (128) and is not routine. The use of provocative angiography should be reserved for selected patients at competent centers with well-trained radiologists.

For critically ill patients, ensuring the adequacy of intravascular volume is very important before mesentery angiography. Dehydration may exacerbate the nephrotoxicity of the contrast medium.

Computed Tomography Scan and Magnetic Resonance Imaging

CT scans are not usually considered diagnostic tools for LGIB, except in the context of bleeding bowel tumors. Several recent reports suggest that helical CT scans may be useful (216–218) in that this mode of scanning has the potential to detect hemorrhage rates of 0.5 mL/min or less (219,220)—between 72% and 79% (216,217). In the evaluation of colonic vascular lesions, helical CT has a sensitivity of 70% and specificity of 100% compared to colonoscopy and conventional angiography (221).

Helical CT angiogram is a modified form of angiography. A rapid-acquisition CT scan is performed 30 seconds after contrast is injected into the abdominal aorta. The images are taken in 10-mm slices, 5 mm apart. A positive result is seen as extravasation of contrast medium into the intestinal lumen. Helical CT angiography reports 70% sensitivity in the diagnosis of colonic angiodysplasia through the demonstration of vessel accumulation in the colon wall, early filling vein, and enlarged supplying artery (221). Currently, there is no role for the use of magnetic resonance imaging (MRI) in the evaluation of LGIB.

Ultrasound

Ultrasonography is a convenient, noninvasive, nonradiation-emitting, and easily available diagnostic tool in the emergency department and ICU. Data regarding the use of US for LGIB have been limited. Yamaguchi et al. (222) noted that the colonic bleeding site was localized by US in 59 of the 90 (66%) patients compared with 81% by colonoscopy. When the bleeding site was in the rectum, the US detection rate was only 30% (10 of 33 patients), but the US detection rate was 82% to 100% when the bleeding site was elsewhere. These clinicians concluded that rectal and diverticular bleeding were difficult to diagnose by US, but for the other diseases, diagnosis by US was possible in 91% to 100% of cases. In our experience, angiodysplasia cannot be detected by US, while diverticulitis—but not diverticular bleeding—can. Other causes of bleeding—especially due to tumors, enteropathy, and colitis—can also be detected by US. In contrast to Yamaguchi’s study (222), we think rectal lesions may be visualized via US through a urinary bladder window. We have proposed the “ultrasonographic bisection approximation method” to localize and detect GI obstructive lesions. The accuracy of US in predicting obstructive levels in the gastric outlet; duodenum, the jejunum, and ileum; and the colon were 100%, 74%, and 98%, respectively (223).

Other Modalities

Small bowel follow-through (SBFT) and enteroclysis are used to detect small lesions, while barium enema is the image study for the colon. All of these are less sensitive for superficial lesions such as angiodysplasia, a common bleeder of LGIB. SBFT is of little use in evaluating obscure GI bleeding, with a diagnostic yield that may be as low as 0% (224). For patients with a high index of suspicion for the presence of small bowel diseases, such as small bowel tumor or Crohn disease, the diagnostic yield will be higher; SBFT detected 83% of small bowel tumors. In patients with suspected Crohn disease, SBFT may have a sensitivity of over 90% (225,226).

Enteroclysis is a modified form of SBFT in which a 10-French catheter is inserted into the distal duodenum or proximal jejunum under fluoroscopy, followed by the infusion under high pressure of a double-contrast solution with barium and air, water, or methylcellulose. This rapid rate of infusion allows better distention and visualization of the small bowel. Studies revealed that enteroclysis seems superior to SBFT for evaluation of the small bowel (227,228). Small bowel tumors seem to be the most common diagnosis made by enteroclysis, followed by Meckel diverticulum and Crohn disease of the terminal ileum (229,230). Angiodysplasia is not detected by enteroclysis. With the advent of capsule endoscopy, the use of SBFT or enteroclysis for GI bleeding has declined.

Barium enema cannot detect superficial lesions or confirm a definitive bleeding source of the colon. Furthermore, it may complicate subsequent colonoscopy or angiography, and is less useful for critically ill patients with LGIB.