Review article: the management of lower gastrointestinal bleeding

Transcription

1 Aliment Pharmacol Ther 2005; 21: doi: /j x Review article: the management of lower gastrointestinal bleeding J. J. FARRELL* & L. S. FRIEDMAN à *Division of Digestive Diseases, UCLA School of Medicine, Los Angeles, CA; Harvard Medical School, Boston, MA, USA; Departments of Medicine, ànewton Wellesley Hospital, Boston, MA, USA, and Massachusetts General Hospital, Boston, MA, USA Accepted for publication 30 January 2005 SUMMARY Several recent advances have been made in the evaluation and management of acute lower gastrointestinal bleeding. This review focuses on the management of lower gastrointestinal bleeding, especially acute severe bleeding. The aim of the study was to critically review the published literature on important management issues in lower gastrointestinal bleeding, including haemodynamic resuscitation, diagnostic evaluation, and endoscopic, radiologic, and surgical therapy, and to develop an algorithm for the management of lower gastrointestinal bleeding, based on this literature review. Publications pertaining to lower gastrointestinal bleeding were identified by searches of the MEDLINE database for the years 1966 to December Clinical trials and review articles were specifically identified, and their reference citation lists were searched for additional publications not identified in the database searches. Clinical trials and current clinical recommendations were assessed by using commonly applied criteria. Specific recommendations are made based on the evidence reviewed. Approximately, 200 original and review articles were reviewed and graded. There is a paucity of high-quality evidence to guide the management of lower gastrointestinal bleeding, and current endoscopic, radiologic, and surgical practices appear to reflect local expertise and availability of services. Endoscopic literature supports the role of urgent colonoscopy and therapy where possible. Radiology literature supports the role of angiography, especially after a positive bleeding scan has been obtained. Limited surgical data support the role of segmental resection in the management of persistent lower gastrointestinal bleeding after localization by either colonoscopy or angiography. There is limited high-quality research in the area of lower gastrointestinal bleeding. Recent advances have improved the endoscopic, radiologic and surgical management of this problem. However, treatment decisions are still often based on local expertise and preference. With increased access to urgent therapeutic endoscopy for the management of acute upper gastrointestinal bleeding, diagnostic and therapeutic colonoscopy can be expected to play an increasing role in the management of acute lower gastrointestinal bleeding. INTRODUCTION Lower gastrointestinal bleeding is one-fifth to one-third as common as upper gastrointestinal bleeding and generally has a less severe course. The annual incidence rate of lower gastrointestinal bleeding in the US ranges Correspondence to: Dr J. J. Farrell, Division of Digestive Diseases, UCLA School of Medicine, UCLA Center for the Health Science, Los Angeles, CA 90095, USA. jfarrell@mednet.ucla.edu from 20.5 to 27 cases per adult population at risk (0.03%). In contrast, the annual incidence rate for upper gastrointestinal bleeding is reported to range from 100 to 200 cases per As in upper gastrointestinal bleeding, lower gastrointestinal bleeding stops spontaneously in most cases (80 85%). 1 The mean age of patients with lower gastrointestinal bleeding ranges from 63 to 77 years, with a reported mortality rate of 2 4% (Table 1). 2 7 The incidence rate of lower gastrointestinal bleeding increases with Ó 2005 Blackwell Publishing Ltd 1281

2 1282 J. J. FARRELL & L. S. FRIEDMAN Table 1. Age, mortality rates, and causes of acute lower gastrointestinal bleeding Study Mean age (years) Mortality (%) Diverticulosis (%) Angiodysplasia (%) Cancer/ polyp (%) Colitis/ ulcer* (%) Anorectal (%) Otherà (%) Boley et al. (1978) 2 > NA NA Jensen and Machicado (1988) NA Leitman et al. (1988) NA NA Richter et al. (1995) Jensen and Machiado (1997) 4 77 NA Longstreath (1997) Strate and Syngal (2003) NA, not available. * Includes inflammatory bowel disease, infectious colitis, radiation colitis, vasculitis and inflammation of unknown origin. Includes haemorrhoids, anal fissure and idiopathic rectal ulcer. à Includes postpolypectomy, aortocolonic fistula, trauma from faecal impaction, and anastomotic bleeding. Study included only critically ill patients undergoing angiography. age, with a >200-fold increase from the age of years. 8, 9 This rise in incidence with age is most likely explained by the increasing prevalence of colonic diverticulosis and colonic angiodysplasia with age. Haematochezia signifies bright red blood per rectum and needs to be differentiated clinically from melaena, the passage of darkened, digested blood per rectum, which suggests an upper gastrointestinal source. Although severe lower gastrointestinal bleeding is defined as blood loss originating from a source distal to the ligament of Treitz and resulting in haemodynamic instability or symptomatic anaemia, approximately 10 15% of patients presenting with acute severe haematochezia have an upper gastrointestinal source of bleeding identified on upper endoscopy. Small bowel sources account for % of cases of severe haematochezia. The most common causes of lower gastrointestinal bleeding are diverticulosis, angiodysplasia, haemorrhoids, and ischaemic colitis (see 3, Table 1). This article focuses on the investigation and management of acute lower gastrointestinal bleeding. A practical approach to the management of lower gastrointestinal bleeding, especially severe bleeding, is proposed. Finally, predictors of outcome in patients with lower gastrointestinal bleeding are discussed. METHODS This review aims to evaluate existing published data regarding the management of lower gastrointestinal bleeding. Publications pertaining to lower gastrointestinal bleeding were identified by searches of the MEDLINE database for the years 1966 to December 2004, by using the terms gastrointestinal haemorrhage, gastrointestinal bleeding, and colonoscopy. Clinical trials and review articles were specifically identified, and their reference citation lists were searched for additional publications not identified in the database searches. Clinical trials and current clinical recommendations were assessed by using commonly applied criteria. Grade A evidence ranges from single or cumulatively large, well-designed randomized controlled trials (RCTs) with adequate statistical power to extremely positive all vs. none outcomes among cohorts receiving alternative therapies and then evidence from smaller or single RCTs or meta-analyses. Grade B evidence ranges from high-quality studies of non-randomized cohorts with and without the index treatment, to high-quality case control studies, and finally to one or more high-quality case series. Grade C evidence is that of expert opinion based on first principles, known physiology, or bench research but without access to clinical studies of the higher grades. Specific recommendations are made based on the evidence reviewed. MANAGEMENT APPROACH TO PATIENTS WITH ACUTE LOWER GASTROINTESTINAL BLEEDING The triage and evaluation of patients with lower gastrointestinal haemorrhage remains variable and largely institution-specific (Figure 1). The dynamic nature of severe bleeding leads to limitations with all treatment strategies. To best manage bleeding, it is useful to stratify patients based on the severity of haemorrhage. For the purposes of management, patients with lower gastrointestinal haemorrhage can

3 REVIEW: THE MANAGEMENT OF LOWER GI BLEEDING 1283 Treat Acute severe hematochezia History, physical examination and resuscitation Upper gastrointestinal source? No Yes + Colonoscopy EGD Treat Source identified Treat + Source not identified, adequate exam Active bleeding? No Enteroscopy Video capsule Endoscopy Source not identified, inadequate exam Yes Angiography +/ treatment +/ surgery Figure 1. Algorithm for management of lower gastrointestinal bleeding (EGD, oesophagogastroduodenoscopy). be broadly grouped into four overlapping categories. The first category includes 75 90% of patients and is characterized by minor bleeding that resolves with conservative therapy. The second category is comprised of patients with chronic intermittent bleeding. The aetiology of bleeding in this group is often elusive and probably best evaluated with colonoscopy. Angiography is especially limited in this group of patients because of the slow, sporadic nature of bleeding. The third group has episodes of severe, life-threatening bleeding with haemodynamic stability in between episodes. Because of the inconsistent nature of bleeding in this group, technetium (Tc)-99m red blood cell scans are useful prior to angiography. Alternatively, urgent colonoscopy (with or without colonic purge) may play a diagnostic and therapeutic role in this group. The fourth category comprises patients with continual active bleeding. These patients may be hypotensive and are best served by urgent angiography or even surgery. Clinical evaluation and resuscitation The patient who presents with acute lower gastrointestinal haemorrhage may complain of passing bright red blood per rectum, dark blood with clots, or, less commonly, melaena. Pallor, fatigue, chest pain, palpitations, dyspnoea, tachypnoea, tachycardia, postural changes, or syncope are suggestive of haemodynamic compromise. Resuscitation should take place concurrently with the initial evaluation of the patient. Associated symptoms may provide clues to the source of the bleeding. Although lower gastrointestinal bleeding is usually painless, a history of abdominal pain, weight loss, fever, diarrhoea, vomiting, or partial small intestinal or colonic obstruction are important findings in the differential diagnosis of inflammatory, infectious, or malignant lesions. Similarly, the patient s age, medical history and medication history [e.g. nonsteroidal anti-inflammatory drug (NSAID) use] may prove critical in elucidating the cause of bleeding. For example, colonic diverticula or angiodysplasia are more likely to be a cause of lower gastrointestinal bleeding in 2, 13, 14 a person over 70 years of age (grade B evidence). Similarly, a history of pelvic radiation therapy (for prostatic or gynecologic malignancy) may point to radiation proctitis as a cause of rectal bleeding. This may occur anywhere from 9 months to 4 years after radiation therapy Assessment should include careful cardiac, pulmonary, abdominal and rectal examinations. A digital rectal examination is helpful in excluding anorectal pathology as well as confirming the patient s description of the appearance of the stool. In addition, approximately 40% of rectal carcinomas are palpable during a digital rectal examination. 18 The presence of coagulopathy [international normalized ratio (INR) >1.5] or thrombocytopenia (<50 000/ ll) should prompt correction with transfusion of fresh frozen plasma or platelets respectively (grade C evidence). Anticoagulant use does not preclude endoscopic intervention. For most patients on warfarin who are hospitalized for gastrointestinal bleeding, anticoagulation should be reversed with fresh frozen plasma and vitamin K. Blood transfusion requirement is determined by the patient s age and the rate of bleeding and is also influenced by the presence of co-morbid conditions such as coronary artery disease, cirrhosis, or chronic obstructive pulmonary disease. Orthostatic hypotension,

4 1284 J. J. FARRELL & L. S. FRIEDMAN a decrease in the haematocrit value of at least 6%, a transfusion requirement of more than two units of packed red blood cells, or continuous active bleeding merit admission to an intensive care unit for close observation. During a bleeding episode, persistent haemodynamic instability despite aggressive resuscitation efforts warrants intervention. Nonsurgical therapy may be performed during either angiography or colonoscopy (see below). Other than the removal of possible aetiologic causes (e.g. NSAIDs) and correction of coagulopathy, there are few specific medical therapies aimed directly at the management of lower gastrointestinal bleeding. Empiric hormonal therapy (oestrogen) to control gastrointestinal bleeding of obscure origin thought to be caused by colonic angiodysplasia is controversial and may be ineffective (grade A and B evidence) There is no evidence to support the initial use of octreotide or other systemically administered drugs in the management of lower gastrointestinal bleeding. In radiation proctitis, vascular telangiectasia and nonhealing mucosal ulceration, perhaps caused by an underlying obliterative arteritis, may lead to severe recurrent haemorrhage. The nonendoscopic management of bleeding secondary to radiation proctitis includes the use of sucralphate or formalin enemas. Sucralphate is a highly sulphated polyanionic dissacharide. In this setting, its postulated mechanisms of action include stimulation of epithelial healing and formation of a protective barrier. The strongest evidence for the use of sucralphate enemas comes from a prospective randomized, double-blind, controlled trial of 37 patients in which anti-inflammatories were compared with rectal sucralphate (grade A evidence) 24 and a single prospective study of rectal sucralphate (grade B evidence). 25, 26 Formalin may sclerose and seal fragile neovasculature in radiation-damaged tissues, thereby preventing further bleeding. Application of formalin directly to the mucosa produces local chemical cauterization and can stop bleeding by sealing the neovascularized telangiectatic spots and ulcers. The success of bleeding control is related to the accurate localization of and application of formalin to all the bleeding sites. Either a 3.6% formalin solution or 4% formalin solution is used for irrigation. An alternative method is the direct application of gauze soaked in formalin (4% or 10%). Three prospective case series (grade B evidence) and nine retrospective reports (grade B evidence), support the use of formalin in the management of lower gastrointestinal bleeding secondary to radiation proctitis. There have been no prospective randomized-controlled studies of the use of formalin in the management of radiation proctitis. Endoscopy Upper endoscopy, push enteroscopy and colonoscopy are generally considered to be safe procedures, even in 39, 40 elderly patients with gastrointestinal bleeding. Elderly persons are at greater risk of complications of gastrointestinal endoscopy ( %) than younger patients ( %). 41 The principal complications that occur in the elderly are haemorrhage, aspiration pneumonia, myocardial infarction and bowel perforation. Cardiopulmonary events may account for more than 50% of complications associated with endoscopy; the majority of complications are aspiration, oversedation, hypoventilation, vasovagal episodes and airway obstruction. Therapeutic procedures, especially when performed in an emergency setting, generally result in a higher rate of complications than do diagnostic procedures. 41, 42 The risks of endoscopy performed in an emergency setting may be minimized through adequate resuscitation of the patient before the procedure and appropriate sedation and monitoring during the endoscopy (grade C evidence). Standard monitoring of sedated patients undergoing gastrointestinal endoscopic procedures includes recording of the heart rate, blood pressure, respiratory rate, and oxygen saturation before, during and after sedation. Continuous electrocardiographic (ECG) monitoring is reasonable in high-risk patients, although improved outcomes with such monitoring have not been shown conclusively in controlled trials. Patients who may benefit from ECG monitoring include those who have a history of a serious dysrhythmia or cardiac dysfunction, elderly persons, and those in whom an extensive therapeutic endoscopic procedure is anticipated. Supplemental oxygen administration has been shown to reduce the magnitude of oxygen desaturation during endoscopic procedures performed under sedation and should be considered mandatory, especially in patients with impaired pulmonary function or significant pre-sedation oxygen desaturation and patients in whom a prolonged or complex procedure is anticipated (grade B evidence) Care must be taken to avoid suppression of the hypoxic ventilatory drive, which can 13, 44 lead to profound hypercapnoea.

5 REVIEW: THE MANAGEMENT OF LOWER GI BLEEDING 1285 Upper endoscopy. Diagnostic endoscopic studies should be undertaken only after the patient has been haemodynamically resuscitated. For patients with severe haematochezia and hypovolaemia, an upper gastrointestinal source should be considered, as an upper source will be found in 10 15% of such patients (grade B evidence). 3 Endoscopy or push enteroscopy ought to be considered early in patients with a history of or risk factors for peptic ulcer, portal hypertension, or angiodysplasia. Nasogastric (NG) lavage before upper endoscopy is warranted if there is medium to low suspicion of an upper gastrointestinal source but may be misleading if only clear fluid without bile (or blood) returns (grade C evidence). 46 If blood, clots, or coffee grounds-appearing material is present in the NG aspirate, upper endoscopy must be performed to exclude an upper gastrointestinal source as the cause of haematochezia. In the presence of large-volume upper gastrointestinal bleeding, it is prudent to provide airway protection by intubating the patient prior to upper endoscopy (grade C evidence). Diagnostic colonoscopy. A lower endoscopic study is well established as the diagnostic procedure of choice in the setting of acute lower gastrointestinal haemorrhage. If sigmoidoscopy is selected as the initial endoscopic approach, it should be undertaken with the caveat that the procedure should only be considered diagnostic if an actively bleeding lesion is visualized. However, many endoscopists prefer to perform a total colonoscopy as the initial evaluation. Although the historical view has been that colonoscopy in patients with severe haematochezia is impractical because of inadequate visualization, colonoscopy is, in fact, feasible and useful after rapid cleansing (grade B evidence). 47 The diagnostic accuracy of colonoscopy ranges from 72 to 86% in patients with 3, lower gastrointestinal bleeding (grade B evidence). In one study in which upper endoscopy was performed before colonoscopy in 80 patients with ongoing haematochezia, upper gastrointestinal lesions were diagnosed in 11% of patients, presumed small-bowel lesions in 9%, and no lesions to account for bleeding in only 6% (grade B evidence). 3 These results are superior to those achieved with arteriography. For example, >75% of diverticula are found in the left colon. When colonoscopy is used to diagnose diverticular bleeding, the bleeding is observed to be from the left colon in 60% of cases. 6 When angiography is used for diagnosis, the source is more likely to be from the right colon This inconsistency may reflect differences in diagnostic sensitivity between the two tests, with angiography being less sensitive than colonoscopy for the diagnosis of less-severe diverticular bleeding. 103 At colonoscopy, angiodysplasia are recognized by their characteristic appearance as red, fern-like flat lesions consisting of ectatic blood vessels that appear to radiate from a central feeding vessel; they may have a diameter of 2 10 mm. A pale mucosal halo may be seen around the lesion. When the colon is examined completely, the sensitivity of colonoscopy for detecting angiodysplasia exceeds 80% (grade B evidence) Colonic angiodysplasia are most common in the caecum and proximal ascending colon (54%), followed by the sigmoid colon (18%) and rectum (14%) (grade B evidence). 57 Although angiodysplasia can be found throughout the small intestine, bleeding from angiodysplasia in the small bowel usually presents as irondeficiency anaemia with faecal occult blood and rarely as severe haematochezia. However, a poor bowel preparation may lead to incomplete evaluation of the colonic mucosa. Additionally, the use of a narcotic medication for sedation and analgesia may decrease the sensitivity of colonoscopy for detecting angiodysplasia by transiently decreasing mucosal blood flow. Administration of intravenous naloxone has been demonstrated to enhance the appearance of angiodysplasia during colonoscopy in patients who have received meperidine for sedation (grade B evidence). 58 Unfortunately, the use of naloxone may result in discomfort for the patient, particularly when the procedure is prolonged by a therapeutic intervention. Other specific diagnostic features at colonoscopy include NSAID-related disease, ischaemic colitis and radiation colitis. Colonic ulcers caused by NSAIDs are often sharply demarcated, with a predilection for the terminal ileum and proximal colon, where pills may reside for a longer period of time than in other segments of the bowel. The development of diaphragm-like strictures is pathognomonic of NSAID injury These strictures are typically multiple in number, with normal intervening mucosa. Non-occlusive colonic ischaemia most commonly involves the so-called watershed areas: splenic flexure, right colon, or rectosigmoid junction. In patients with ischaemic colitis, sigmoidoscopy reveals ulceration of the colonic mucosa, with the exception of the rectum in most cases. Histology reveals necrosis, non-acute and chronic inflammatory changes as seen in inflammatory bowel disease. Radiation

6 1286 J. J. FARRELL & L. S. FRIEDMAN proctitis typically demonstrates characteristic telangiectasias at colonoscopy. Several recent studies have shown that colonoscopy performed in an emergency (within 12 h of admission) is safe and effective (grade B evidence). 4, Early intervention, particularly for massive diverticular haemorrhage, may improve the diagnostic and therapeutic outcome and prevent the need for surgical intervention (grade B evidence). 4, 65, 66 In addition, early colonoscopic evaluation may reduce the duration of hospitalization and lower overall costs per patient (grade B evidence). 4, 65, In fact, time to colonoscopy has been shown to be an independent predictor of the length of hospital stay. In a widespread spectrum of patients with lower gastrointestinal bleeding, the reduction in the length of hospital stay was shown to relate primarily to improved diagnostic yield rather than therapeutic intervention. 69 There is no consensus regarding the need for a colonic purge prior to colonoscopy in a patient with active lower gastrointestinal bleeding (grade B and C evidence). 3, 48, 64, 71, 72 If it is prescribed, extra effort of the nursing staff and patient is required to ensure a successful purge. A sulphate or polyethylene glycol (PEG)-based purge (e.g. GoLytely; Braintree Laboratories, Braintree, MA, USA) is administered orally. [For patients who are not able to drink a litre of purge solution every min until the effluent clears (usually 5 8 L total), administration via an NG tube is recommended.] Approximately 30 min before the purge is started, 10 mg metoclopramide can be administered intravenously for its prokinetic and antiemetic properties. The dose can be repeated every 4 6 h if nausea results or if further purge is necessary. Occasionally, patients with chronic kidney disease may require dialysis after purging, and those with severe congestive heart failure may require diuresis. However, complication rates are low with PEG-based purges. Only an experienced endoscopist should perform colonoscopy in an actively bleeding patient with an unprepared colon, because the risk of colonic perforation may be increased in this setting (grade C evidence). Therapeutic colonoscopy. Suitable large-channel endoscopes and endoscopic accessories are necessary for effective diagnostic and therapeutic colonoscopy in the management of acute lower gastrointestinal bleeding. Applications of colonoscopic haemostasis techniques is based on identification of the same stigmata of haemorrhage that have been identified in the upper gastrointestinal tract as predictors of recurrent upper gastrointestinal haemorrhage from ulcers. Insufficient numbers of patients with these stigmata in the setting of lower gastrointestinal bleeding have been reported to determine their usefulness in predicting outcome. For most actively bleeding lesions or those with adherent clots in the colon, except in association with haemangiomas and internal haemorrhoids, a combination of adrenaline injection and thermal coagulation (with a bipolar or heater probe) is recommended (grade C evidence). This recommendation is supported in part by demonstration of the safety of colonic endotherapy in experimental animals, even with the use of energy and pressure parameters far higher than those used in a clinical setting (grade C evidence). 73 Use of either small or large probes, low-power settings (10 15 W), placement of the probe directly on the bleeding point, moderate tamponade pressure, and application of the probe until adequate whitening of the site (i.e. coagulation) is observed are recommended (grade C evidence). 74 These guidelines are in contrast to those for gastroduodenal ulcers, for which large probes, firm tamponade, and long coagulation pulses are recommended (Table 2). Colonic diverticular bleeding is amenable to haemostasis with adrenaline injection therapy, bipolar coagulation, or both. 65, One study reported that patients with demonstrable diverticular bleeding who underwent endoscopic therapy with adrenaline injection, bipolar coagulation, or both, had no recurrence of bleeding during the 30-month follow-up period, when compared with a 53% rebleeding rate in patients who received conservative medical therapy alone (grade A evidence). 65 Less frequently employed methods for controlling diverticular bleeding include endoscopic band ligation and placement of haemoclips (grade B and C evidence) Conventional endoscopic treatment of colonic angiodysplasia is performed with contact thermal probes (grade C evidence) To prevent brisk bleeding from angiodysplasia when contact electrocautery is performed, large angiodysplasia should be cauterized from the outer margin towards the centre to obliterate feeder vessels. Injection therapy with sclerosing agents, such as ethanolamine, has also been described for control of bleeding from colonic angiodysplasia but is not widely employed. 89 Argon plasma coagulation, a noncontact method, has been used increasingly for the treatment of bleeding colonic angiodysplasia (grade B evidence)

7 REVIEW: THE MANAGEMENT OF LOWER GI BLEEDING 1287 Table 2. UCLA Center for Ulcer Research and Education (CURE) Hemostasis Research Group colonoscopic technical parameters for heater probe and bipolar electrocoagulation of bleeding colonic lesions 4 Angiodysplasia, radiation colitis Polyp stalk bleeding Diverticulosis, delayed polypectomy bleeding, or ulcer bleed Cause Active bleeding Nonbleeding visible vessel Active bleeding* Active bleeding* Nonbleeding visible vessel Adherent clot* Bipolar coagulation Large Large or Large Large or Large or small Large or small probe size small small Power setting (W) Pulse duration (s) Heater probe size Large Large or small Large Large Large Large Power setting (J) Pressure Light Light Moderate Moderate Moderate Moderate Endpoint Bleeding stops White coagulum Bleeding stops Bleeding stops Flatten visible vessel and coagulum Clot eliminated, underlyingstigma treated * Consider injection with 1 : adrenaline prior to endoscopic coagulation with bipolar or heater probe. Extra care must be taken when treating lesions in the caecum to avoid perforation. 90 No comparative prospective studies have compared contact and noncontact endoscopic treatment of bleeding colonic angiodysplasia. Postpolypectomy bleeding is the most frequent complication of colonoscopy performed for polypectomy and accounts for approximately 2 8% of cases of acute lower 3, 7, 48 gastrointestinal bleeding (grade B evidence). Massive bleeding that occurs at the time of polypectomy (early postpolypectomy bleeding) is typically arterial in nature and results from inadequate haemostasis of the blood vessel in the polyp stalk. Reduction in the risk of early postpolypectomy bleeding can be achieved by the use of blended, rather than pure cutting electrocautery currents in the polypectomy snare (grade C evidence). In the event of early postpolypectomy bleeding, haemostasis can generally be controlled by resnaring the stalk of the polyp and applying pressure (grade B and C 93, 94 evidence). Delayed bleeding may occur up to 15 days after polypectomy, likely as a result of the sloughing of the eschar at the polypectomy site. 93, 94 Delayed bleeding is usually self-limited and resolves with supportive care in more than 70% of cases. In contrast to early postpolypectomy bleeding, the frequency of delayed postpolypectomy bleeding as a cause of acute severe haematochezia is increasing, possibly because of the increasing use of a blended (rather than pure coagulation) current in the polypectomy snare (grade C evidence). For persistent or severe bleeding at a polypectomy site, a variety of endoscopic techniques have proven safe and effective. These include loop ligation of the remaining polyp stalk, endoscopic band ligation, injection of adrenaline followed by thermal therapy, and application of endovascular clipping 80, devices (grade B and C evidence). The endoscopic management of lower gastrointestinal bleeding from radiation proctitis represents a specific management problem. 24, 25, 32, 34, 35, Endoscopic coagulation with a variety of devices has been reported to be effective for the control of radiationinduced bleeding. The technique is generally used to coagulate focal bleeding telangiectasias rather than the entire friable mucosa. Several treatment sessions are often required. Scarring and re-epithelization with more normal tissue tend to occur over time. Sixteen relevant studies have looked at the role of local endoscopic control of bleeding from radiation proctitis. Apart from one randomized controlled trial and one prospective series, the remaining 14 have been retrospective series or case reports. Although these case series all refer to the use of ablative therapy in lateradiation proctitis, different types of coagulation probes or lasers have been used: neodynium-yag laser in eight, argon lasers in four, bipolar electrocautery in three and heater probes in two reports. The only randomized prospective study in this area compared endoscopic bipolar electrocoagulation and heater probe therapy in the treatment of chronic rectal bleeding from rectal telangiectasia (grade A evidence). 114 Twenty-one patients were randomized to treatment with either a

8 1288 J. J. FARRELL & L. S. FRIEDMAN heater probe or a bipolar electrocoagulation probe, and treatment sessions were repeated with the same probe until the bleeding resolved. In both groups, there was a statistically significant decrease in severe bleeding, but a reduction in the number of units of blood transfused per cases was only seen in the heater-probe group. During follow-up endoscopy, there was resolution of the telangiectasias, scarring, or epithelial replacement in all cases in both groups. Patient interviews 6 months after treatment revealed that rectal bleeding, tenesmus, and general health had improved. Other nonrandomized studies that reported success with either heater probe or bipolar electrocautery used either higher or lower power settings than those used in the randomized study (grade B evidence). 115, 116 The benefit of YAG laser in the management of bleeding related to radiation proctitis has also been reported in several 17, 27, 112, retrospective series (grade B evidence). Typically a power setting of W has been reported. More recently, argon plasma coagulation has gained popularity for the management of lower gastrointestinal bleeding related to radiation proctitis. Argon plasma coagulation is a noncontact technique of electrocoagulation in which electrosurgical current is delivered to the tissue through argon gas; the depth of coagulation is limited (2 3 mm). The technique can be used to treat large surface areas of mucosa to achieve haemostasis easily, safely, and on an out-patient basis. For large bleeding surfaces, as in radiation-induced proctitis, several treatment sessions usually are required to achieve control of bleeding (grade B evidence) , 113, Short-term complications include anorectal pain, tenesmus and abdominal distention; long-term complications include anorectal pain, chronic rectal ulcer and rectal stricture. Some of the more recent developments in endoscopic haemostasis (e.g. endoscopic band ligation, endoscopic clipping and argon plasma coagulation) have been applied to the management of other sources of lower gastrointestinal bleeding (e.g. bleeding colonic varices, bleeding rectal Dieulafoy s lesion), and are typically reported as single cases without any formal prospective or randomized evaluation (grade B and C evidence) For example, endoscopic band ligation has gained popularity as an alternative to surgical treatment for persistent or recurrent haemorrhoidal bleeding despite conservative therapy (grade B evidence). 130 Yet, there have been no formal prospective randomized trials that compare surgical and endoscopic management of rectal bleeding from haemorrhoids. Radiography The role of double-contrast barium enema (DCBE) in the evaluation of lower gastrointestinal bleeding is decreasing. In addition to the suboptimal quality of DCBE, patients often prefer colonoscopy. However, plain abdominal radiography should be performed prior to colonoscopy if bowel perforation or obstruction is suspected. Radiographic evidence of thumbprinting is indicative of transmural injury to the colon as a result of ischaemic or infectious colitis. There may be an evolving role for multidetector computed tomography (MDCT) for localizing acute lower gastrointestinal bleeding as well as predicting the treatment potential of arteriography and embolization. MDCT is highly sensitive and specific for the diagnosis of colonic angiodysplasia (grade A evidence). 131 Bleeding rates <0.4 ml/min are detectable in swine, provided that peak aortic enhancement reaches 100 Hounsfield units (Grade C Evidence). 132 Several small retrospective reports have reported an accuracy rate of 54 79% for localizing large bowel 133, 134 bleeding (grade B evidence). Radionuclide imaging Evaluation with either radionuclide imaging or angiography (see later) may be appropriate in patients with massive haemorrhage that precludes colonoscopy or in whom a bleeding source is not identified on colonoscopy. Radionuclide imaging detects active bleeding at rates of ml/min and is more sensitive than angiography but less specific than endoscopic or angiographic study (grade B evidence). 135 Either technetium sulphur colloid or [99Tcm] pertechnetatelabeled red blood cells can be used. The relative disadvantage of using [99Tcm] pertechnetate-labelled red blood cells is the persistence of background activity in blood vessels and the blood pool throughout the study, thereby theoretically increasing the threshold for the amount of bleeding needed for detection. In contrast, technetium sulphur colloid, which completely clears the blood pool by min after injection, is easier to detect because background activity is absent. Although imaging with technetium sulphur colloid can detect a bleeding rate as low as 0.1 ml/min, the short

9 REVIEW: THE MANAGEMENT OF LOWER GI BLEEDING 1289 half-life of the colloid within the vascular system requires active bleeding at the time the radionuclide is present in the intravascular space. Therefore, for evaluation of lower gastrointestinal bleeding, which may be episodic, imaging following injection of [99Tcm] pertechnetate-labelled red blood cells is preferred and can be performed at 30-min intervals for up to 24 h, if necessary, thereby allowing detection of intermittent bleeding. Radionuclide imaging is well tolerated by patients but is limited by highly variable accuracy rates for localizing bleeding, ranging from 24 to 91% (grade B 50, evidence). A bleeding scan study may be done while the patient has ongoing haematochezia, and imaging can be repeated over the next 24 h, because the labelled red blood cells stay in the vascular space for at least 24 h. However, the patient must have active bleeding when the image is taken in order to demonstrate extravasation. Whereas early scans (<4 h after baseline) may be helpful in localizing the bleeding site, delayed scans are less efficient in localizing the bleeding site (grade B evidence). 137 Therefore, early-bleeding scans (i.e. at baseline and up to 1 4 h later, before or after the patient starts the oral purge prior to colonoscopy) are recommended in patients who are hospitalized for severe, ongoing haematochezia. Even if the bleeding scan is positive, a confirmatory test such as colonoscopy, angiography, or push enteroscopy is recommended before emergency surgery is considered. Radionuclide imaging is generally performed before angiography. In addition to its role in determining which patients are bleeding sufficiently to warrant an angiographic study, localization of the bleeding source by radionuclide imaging may allow a more selective angiographic study, thereby decreasing the contrast dye load. 138, 140 Radionuclide screening appears to increase the diagnostic yield of arteriography by a factor of 2.4 by screening out-patients who are not actively bleeding at the time of the examination, thus sparing them the risks and costs of a nondiagnostic arteriographic study (grade B evidence). 141 Haemodynamically stable patients with severe but intermittent bleeding should be evaluated with [99Tcm] red blood cell scanning. A positive red blood cell diagnosis should necessitate an urgent angiography, which should be performed within 1 h of positive scintigraphy, day or night (grade C evidence). 142 Patients who are haemodynamically unstable with severe unremitting bleeding should forego nuclear scintigraphy and instead undergo resuscitation and angiography as soon as possible (grade C evidence). Angiography Angiography is performed only if there is a gastrointestinal bleeding rate of at least 1 ml/min for accurate detection of extravasation of contrast into the bowel lumen. 143 Unfortunately, bleeding is frequently intermittent and may occur at a lower rate, thereby limiting the detection of the causative lesion. Indirect evidence of a bleeding lesion (such as an early-filling vein of angiodysplasia or neovascularity of a neoplasm) suggests, but does not confirm, a potential bleeding site. The examination is not definitive unless extravasation of contrast into the lumen is observed. The overall yield of angiography for the detection of a gastrointestinal bleeding source ranges from 40 to 78% (grade B evidence). 5, Diverticular disease and angiodysplasia are the most common findings when angiography is positive and account for 50 80% of sources in bowel supplied by the superior mesenteric artery (grade B evidence). 148 Angiography remains the gold standard for the diagnosis of angiodysplasia. Following injection of contrast, angiodysplasia are recognized by ectatic slow-emptying veins, vascular tufts, or small veins that fill early. Angiography has a specificity of 100% but a sensitivity of only 30 47% (grade B evidence). 149 Advantages of angiography include the lack of requirement for bowel preparation, ability to localize the bleeding source (when one is identified), and possibility of therapeutic intervention in some cases. The study can be performed without a colonic purge or while a purge is being administered. Haemostasis can be achieved by intraarterial infusion of vasopressin or arterial embolization via the angiographic catheter. 147 Intra-arterial infusion of vasopressin is successful in controlling gastrointestinal haemorrhage in up to 91% of patients with lower gastrointestinal bleeding caused by either diverticular 145, 150 disease or angiodysplasia (grade B evidence). Unfortunately, bleeding recurs in up to 50% of patients after cessation of the vasopressin infusion. 145 Vasopressin infusion is labour-intensive, requiring admission to the intensive care unit for most patients. It also causes important side-effects including abdominal pain and is contraindicated in patients with clinically significant coronary artery disease. 151 A longer acting synthetic vasopressin analogue (terlipressin) has been used

10 1290 J. J. FARRELL & L. S. FRIEDMAN successfully as a single bolus intra-arterial injection to stop lower gastrointestinal bleeding. 152 Transcatheter embolization is a more definitive means of controlling haemorrhage than is intra-arterial infusion of vasopressin. When first introduced, embolization proximal to the mesenteric border of the colon via larger catheters led to a rate of bowel infarction that ranged from 13 to 33% (grade B evidence) These initial complications deterred enthusiasm for the technique, and embolotherapy was eventually eschewed in favour of local vasoconstrictive therapy (i.e. vasopressin infusion), which has remained the procedure of choice until recently, despite its serious shortcomings. However, the disadvantages of vasopressin coupled with the availability of microcatheters led to the development of microcatheter embolization using microcoils, gelfoam, and polyvinyl alcohol particles. To date, there have been over 150 reported cases of superselective lower gastrointestinal embolization, with a rate of clinical success (cessation of bleeding) between 44 and 91% and without major ischaemic complications (grade B evidence). 142, Embolotherapy has a number of distinct advantages over local vasoconstrictive therapy, including immediate cessation of bleeding without the need for prolonged infusions or management of an indwelling arterial catheter. The systemic side-effects of vasopressin are also avoided. Despite a dearth of comparative studies, many interventional radiologists have switched from vasoconstrictive therapy to superselective embolization (grade C evidence). The location and aetiology of bleeding have important therapeutic implications for angiotherapy. In terms of location, bleeding from the right colon and caecum may be less amenable to embolotherapy than is bleeding in the left colon. Angiodysplasia is more difficult to treat with embolization than is diverticular bleeding and has a greater propensity for rebleeding (grade C evidence), 142, 157 probably because of the predilection of angiodysplasia for the right colon. As a result, 7 40% of the patients with angiodysplasia treated by embolotherapy required emergency surgery for either failure to 142, 157 control bleeding or rebleeding (grade B evidence). Angiography should be reserved for the patient who has massive bleeding that precludes colonoscopy, has persistent or recurrent bleeding, or has undergone a colonoscopy that has failed to identify the bleeding source. The endoscopic and angiographic examinations are complementary, and the order in which the investigations are undertaken often depends on local availability and expertise (grade C evidence). Small bowel evaluation An evaluation of the small bowel is indicated in those patients with gastrointestinal bleeding in whom upper gastrointestinal endoscopy and colonoscopy are negative. Small bowel evaluation in patients who are haemodynamically stable may be performed with push enteroscopy, which allows endoscopic evaluation of the proximal 60 cm of the jejunum. If examination of the remaining jejunum and ileum is desired, video capsule endoscopy provides imaging of the entire small bowel and is well tolerated by patients. Video capsule endoscopy is reported to identify the bleeding source in 55 65% of the examined patients with gastrointestinal bleeding (grade B evidence). 169, 170 A radionuclide scan for a Meckel s diverticulum may be appropriate in young patients presenting with otherwise unexplained lower gastrointestinal bleeding. 171 Surgery Most patients with severe lower gastrointestinal bleeding, and even those with prolonged bleeding, will not require surgery. Most have intermittent bleeding or can be controlled with nonsurgical therapies, including endoscopic, angiographic, or (for haemorrhoidal bleeding) anoscopic techniques. Surgical intervention is required when haemodynamic instability persists despite aggressive resuscitation, the blood transfusion requirement is greater than 6 U, or severe bleeding 12, recurs (Table 3) (grade B and C evidence) Surgical intervention for lower gastrointestinal bleeding is necessary in 18 25% of patients who require blood 12, 176 transfusion (grade B evidence). In some cases of non-occlusive colonic ischaemia, particularly in patients with renal failure or severe atherosclerosis, the presentation is fulminant, with colonic infarction, and there is a need for urgent surgery because of an otherwise high mortality rate. 36 Except in fulminant cases, treatment of non-occlusive colonic ischaemia is supportive, and most cases resolve spontaneously within several days to weeks. However, the absence of colonic infarction does not ensure a favourable outcome, and some patients who are felt to be candidates for non-operative therapy have a substantial mortality rate. 177

11 REVIEW: THE MANAGEMENT OF LOWER GI BLEEDING 1291 Table 3. Indications for emergency surgery for severe lower 12, 145, , 184 gastrointestinal bleeding Hypotension and shock despite resuscitation Continued bleeding (>6 U of packed red blood cells transfused) and lack of diagnosis (bleeding source) despite emergency colonoscopy, push enteroscopy, radionuclide imaging, and angiography Active bleeding from a segmental gastrointestinal lesion that is amenable to cure or permanent hemostasis by surgery The patient is an emergency surgical candidate without a con traindicating comorbidity and with a reasonable life expectancy The overall operative mortality rate for emergency surgery for lower gastrointestinal bleeding is 10%, despite improved methods to localize the bleeding site that permit segmental rather than subtotal colectomy (grade B evidence). 145, 176, 178 Age, probably by association with increase comorbidity, is an important risk factor for postoperative mortality. The post-operative mortality rate in patients who undergo surgery for colorectal cancer increases with age (3.7% in patients aged years, 9.8% in patients aged 80 to 89 years, and 12.9% in those over 89 years) (grade B evidence). 179 Surgery should be considered in patients in whom a bleeding source has clearly been identified and in whom more conservative therapies have failed. Accurate preoperative localization of the bleeding site is essential if segmental resection of the colon is to be successful. Blind segmental resection of the colon, segmental resection based solely on tagged red blood cell scan localization, and emergency subtotal colectomy are associated with substantial rates of rebleeding (as high as 33%) and mortality (33 57%) (grade B evidence). 50, 145, 175, Of 49 patients who underwent total abdominal colectomy in one study, the overall mortality rate was 27% (grade B evidence). 187 In another report, only 25% of patients who underwent a total abdominal colectomy for massive lower intestinal bleeding survived without complications, and the mortality rate was a formidable 33% (grade B evidence). 188 However, mortality rates as low as 5 10% have been reported for total abdominal colectomy (grade B evidence). 175, 183, 189 Mortality can also be more in patients who undergo a blind limited resection, with rates as high as 30 57% (grade B evidence). 184, 189 Comparison of blind limited resection with total abdominal colectomy has shown mortality rates that are similar if not higher for limited resection, although the number of patients studied was small. 183, 189 The rebleeding rate following blind limited resection has been as high as 33%. 184 These data support the importance of pursuing an aggressive approach to preoperative localization of the bleeding source. When angiography is successful in localizing the bleeding site, limited intestinal resection has resulted in significantly lower morbidity rates than did surgery in historic controls without angiographic localization (8.6% vs. 37%) (grade B evidence) 145, 184, 186, 190 In one study, the rebleeding rate over a 1-year follow-up period was 14% after segmental colectomy directed by angiography and 42% after blind segmental colectomy with a negative angiographic result (grade B evidence). 189 PREDICTING OUTCOME AND RECURRENCE OF LOWER GASTROINTESTINAL BLEEDING Mortality rates for lower gastrointestinal bleeding are less than 5% (Grade B Evidence). 6, 8, 49, 191 Although the highest mortality rates are reported for patients in whom bleeding occurs during a hospitalization for other reasons, in most series lower gastrointestinal bleeding per se has uncommonly been the cause of death. 6 A population-based study of patients enrolled in a health maintenance organization in the USA reported an in-hospital mortality rate of 3.6% for patients hospitalized with lower gastrointestinal haemorrhage and 23.1% for those in whom lower gastrointestinal bleeding developed following hospital admission (grade B evidence). 6 When managed conservatively, diverticular bleeding resolves spontaneously in over 75% of patients (grade B evidence). 12 The majority of patients require <4 U of transfused blood. Bleeding recurs in 14 38% of cases following the primary episode and in up to 50% following a second episode of bleeding (grade B evidence). 12, 53 For patients with a discharge diagnosis of diverticular bleeding who did not require definitive therapy, the rate of recurrent bleeding is 9% at 1 year, 10% at 2 years, 19% at 3 years, and 25% at 4 years (grade B evidence). 6 A number of studies have proposed clinical prognostic criteria to distinguish patients with a high and a low risk of recurrent haemorrhage. 7, 191, 192 A reliable predictive model that can accurately forecast the outcome of an episode of acute lower gastrointestinal bleeding in terms of risk of recurrent bleeding, need for therapeutic intervention, and, importantly, mortality