Erythromycin Intravenous Bolus Infusion in Acute Upper Gastrointestinal Bleeding: A Randomized, Controlled, Double-Blind Trial

Transcription

1 GASTROENTEROLOGY 2002;123:17 23 Erythromycin Intravenous Bolus Infusion in Acute Upper Gastrointestinal Bleeding: A Randomized, Controlled, Double-Blind Trial JEAN LOUIS FROSSARD, LAURENT SPAHR, PIERRE EDOUARD QUENEAU, EMILIANO GIOSTRA, BEAT BURCKHARDT, GAËLLE ORY, PHILIPPE DE SAUSSURE, BERJ ARMENIAN, RAYMOND DE PEYER, and ANTOINE HADENGUE Division ofgastroenterology and Hepatology, Geneva University Hospitals, Genève, Switzerland Background & Aims: Emergency endoscopy may be difficult in upper gastrointestinal bleeding when blood obscures the visibility. Erythromycin, a motilin agonist, induces gastric emptying. We investigated whether an intravenous bolus infusion of erythromycin would improve the yield of endoscopy in these patients. Methods: Patients admitted within 12 hours after hematemesis were randomly assigned to erythromycin (250 mg) or placebo, 20 minutes before endoscopy. The primary end point was endoscopic yield, as assessed by objective and subjective scoring systems and endoscopic duration. Secondary end points were the need for a second look, endoscopy-related complications, blood units transfused, and length of hospital stay. Results: Fiftyone patients received erythromycin and 54 received placebo. A clear stomach was found more often in the erythromycin group (82% vs. 33%; P < 0.001). This difference remained significant in patients with cirrhosis. Erythromycin shortened the endoscopic duration (13.7 vs minutes in the placebo group; P 0.036) and reduced the need for second-look endoscopy (6 vs. 17 cases; P 0.018). Length of hospital stay and blood units transfused did not significantly differ between the 2 groups. No complications were noted. Conclusions: Erythromycin infusion before endoscopy in patients with recent hematemesis makes endoscopy shorter and easier, thereby reducing the need for a repeat procedure. Upper gastrointestinal (GI) bleeding is common in clinical practice. Current estimations of its incidence range from 50 to 150 episodes per 100,000 population per year. 1,2 Upper GI tract endoscopy plays a key role in the investigation of upper GI hemorrhage and can identify the source of bleeding in 80% 90% of cases. 3 5 Endoscopic hemostatic therapy for upper GI bleeding reduces the need for surgery, decreases the use of blood transfusion, shortens hospital stay, and provides a clinically important reduction in morbidity and mortality in patients with acute nonvariceal upper GI hemorrhage. 6 8 In patients with gastric or duodenal ulcer, injection therapy is used as a first-choice treatment because it can be started at the time of diagnosis, has a low cost, and is effective in 75% 90% of patients to stop the bleeding. 8,9 In patients with variceal bleeding, sclerotherapy or variceal band ligation can control bleeding in approximately 80% of cases. 10 However, the endoscopic yield and success rate of sclerotherapy strongly depend on the quality of the endoscopic viewing. Blood in the stomach or duodenum may obscure the view or expose the patient to potential broncho-aspiration; it delays a definitive surgical treatment and explains why repeat endoscopy is performed in 5% 12% of cases. 11,12 Erythromycin, a macrolide antibiotic discovered in 1952, is a major motilin receptor agonist that accelerates gastric emptying by inducing antral contractions similar to phase III of the interdigestive migrating motor complex The aim of our study was to evaluate the effects of a short intravenous infusion of erythromycin on gastric emptying and on the endoscopic performance in patients presenting with hematemesis. Thus, we undertooka monocentric, prospective, double-blind, randomized trial designed to evaluate the effects of erythromycin in clearing the stomach before emergency endoscopy. The primary end points were the effects on endoscopic yield as assessed by objective and subjective scoring systems and the mean duration of the endoscopic procedure. Secondary end points were the need for second-look endoscopy, endoscopy-related complications, blood units transfused, length of hospital stay, and analysis of the subgroup of patients with cirrhosis. Abbreviation used in this paper: GI, gastrointestinal by the American Gastroenterological Association /02/$35.00 doi: /gast

2 18 FROSSARD ET AL. GASTROENTEROLOGY Vol. 123, No. 1 Patients and Methods All patients with hematemesis admitted to our institution from October 1999 to April 2001 were included if they met the following criteria: age 18 years; upper GI bleeding, defined as either fresh and bright red hematemesis or coffeeground hematemesis; and admission within 12 hours of the initial signs of bleeding. All patients who were admitted underwent upper GI endoscopy when they were hemodynamically stable, with heart rate less than 100 beats/min and systolic pressure greater than 100 mm Hg at the time of endoscopy, as defined by Savides and Jensen. 18 Patients were not eligible to enter the study protocol in the following cases: (1) known allergy to erythromycin or to other macrolides; (2) encephalopathy, making informed consent impossible; (3) concomitant terfenadine, astemizole, or cyclosporin use, because of possible interactions with erythromycin ; (4) nasogastric tube or gastric lavage; (5) prior gastrectomy; and (6) pregnancy or lactation. In the subgroup of patients with cirrhosis, the presence of portal hypertension was determined on the basis of a histological diagnosis of cirrhosis or the presence of esophageal varices seen during endoscopy. The Child Pugh score was used to assess the severity of liver disease on admission (15- point scale). Eligible patients were randomly assigned to erythromycin or placebo. Treatments were delivered in a double-blind way and were allocated by a manual randomization list by our central pharmacy, which delivered treatment packages every 2 weeks to our division. Treatment assignment was not released until final analysis. Active erythromycin (250 mg) and placebo were packed in identical syringes. The central pharmacy kept the identity of treatments in sealed envelopes. For final analysis, the identity of each syringe was released by the central pharmacy. The protocol was approved by the local Hospital Ethics Committee at Geneva University Hospitals, and a written, informed consent was obtained from each patient at the time of enrollment. After inclusion, placebo or erythromycin (250 mg) was mixed with 50 ml of isotonic saline solution before infusion and was administered intravenously over 5 minutes. Twenty minutes thereafter, emergency endoscopy was performed with the patient in the left lateral position by using lidocaine pharyngeal anesthesia. No other drug was used on a regular basis. Each procedure was timed from start to finish. The steering-down method to insert the instrument was used under television monitor vision by using air insufflation without suction. To reduce the possibility of missing any area, a systematic routine survey was used that included successive examination of the esophagus, the cardia, the fundus, the corpus, the antrum, the bulbus, and, finally, the second part of the duodenum. Definition of the Primary End Point The primary end point was to evaluate the effects of erythromycin on the endoscopic yield, by using subjective and objective scoring systems. We were inspired by the Avgerinos score, which was published in 1997 in the Lancet. 22 In the first study, which showed an improved endoscopic yield with use of pharmacological treatment in the early period after upper GI hemorrhage, Avgerinos et al. measured the ease of endoscopic procedure on a 10-cm visual analog scale that ranged from easy (0) to difficult (10). 22 Although this score is complex to use, we thought it worth using after modification, because it was successfully evaluated in a large cohort of patients with GI bleeding. Thus, we chose to modify and expand the Avgerinos score to make it simpler to use in a daily practice. Thus, the first subjective scoring system was based on the endoscopist s judgment in evaluating the ease of performing the diagnostic and therapeutic endoscopy, ranging from easy to difficult. In our study, a full stomach was defined if the endoscopy was difficult and if 50% 100% of the stomach mucosa was covered by blood residues or clots. An empty stomach was defined when the endoscopy was easy to perform and when less than 25% of the mucosa was covered by blood. In addition, an objective score, scored from 0 to 2 (0 [worst vision], 25% of the surface was visible; 1, 25% 75% visible; 2 [best vision], 75% visible), was derived from analysis of each area of the stomach (fundus, corpus, antrum, and bulbus). Therefore, the maximal total score was 8 for this second scoring system. A score of 6 was considered as a clear stomach, and a score of 5 was considered as a full stomach. The cutoffs of 5 and 6 were determined before the start of the study by the results obtained in 10 patients with GI bleeding who agreed to receive a unique intravenous bolus of erythromycin instead of receiving a nasogastric tube and gastric lavage (Division of Gastroenterology, unpublished data, February, 1999). The mean duration of the endoscopic procedure was also considered as a primary end point because endoscopic duration may reflect the ease of the procedure. Definition of Secondary End Points Secondary end points were the need for second-look endoscopy, blood units transfused within 24 hours after the initial endoscopy, endoscopy-related complications, length of hospital stay, and analysis of the subgroup of patients with cirrhosis. Second-lookendoscopy within 24 hours after the initial endoscopy was defined as follows: failure of the initial endoscopy to identify the bleeding site or recurrence of clinically significant GI bleeding, defined as a transfusion requirement of 2 U of blood within 24 hours of time 0 and systolic blood pressure of 100 mm Hg. 18,23 Endoscopy-related complications were defined as accidental perforation of the GI tract, broncho-aspiration, and death after the endoscopic procedure. Gastro-duodenal ulcers actively bleeding at the time of endoscopy were treated by injection therapy as usually recommended in daily routine, because no conclusive differences among the various treatment modalities available, including thermal devices, laser therapy, or injection therapy, can be claimed. 18,24 Esophageal varices actively bleeding were treated either by intravariceal sclerotherapy or by band ligation. Sclerotherapy was performed with polidocanol. Blood products

3 July 2002 ERYTHROMYCIN FOR UPPER GI BLEEDING 19 consisting of plasma or packed cells were given to patients with 1 of the following signs of bleeding: red hematemesis, hemodynamic instability as defined previously, and hemoglobin 10 g/dl in patients without cirrhosis and 8 g/dl in those with cirrhosis. The quantity of blood products transfused was monitored during the first 24 hours. Patients with peptic ulcer were given intravenous omeprazole for 48 hours (80 mg/day) and were then switched to 20 mg/day orally for 8 weeks Patients with esophageal variceal bleeding were given intravenous bolus doses of octreotide (100 g) after endoscopy and then received a continuous perfusion of octreotide (50 g/h) for 5 days. 10 We assessed treatment safety by monitoring adverse effects reported after the infusion. Statistical Analysis The number of patients to be included in each group was calculated to meet the primary end point of the trial (improving endoscopic yield from 65% to 90%; error, 0.05; error, 0.2; 51 patients to be included in each arm). The trial was not designed to detect an outcome difference in terms of the blood units transfused and the length of hospital stay. Thus, the lackof difference might be irrelevant. Results are expressed as mean SD. Comparisons between the 2 treatment groups were performed with the Pearson 2 test for categorical variables (or the Fisher exact test, if appropriate), 1-way analysis of variance, or 2-sample t tests with unequal variance for continuous variables. Analyses were performed on an intention-to-treat basis and included all patients who underwent randomization. The Mann Whitney test was used to compare transfusion requirements in the 2 groups. By using multiple logistic regression analysis, we examined which factors including erythromycin, delay between bleeding and endoscopy, hemoglobin level and systolic blood pressure at admission, bleeding source, sex, age, and presence of cirrhosis were associated with gastric emptying, our main outcome in this study. P values of 0.05 were considered statistically significant. Analyses were performed with SPSS software (version 9.0; SPSS Inc., Chicago, IL). Results From October 1999 to April 2001, 198 patients presented with hematemesis at our center. One hundred five patients who presented with 107 bleeding episodes met the criteria for inclusion and were randomized. There were 3 protocol violations (gastric lavage in 2 cases and 1 case of prior partial gastrectomy). Because only 1 patient in each group had a repeated episode, the violation-of-independence assumption could be ignored, and we chose to consider only the first episode from each patient, to keep the analysis clean. Thirty-two patients had cirrhosis (Table 1). The groups were well matched in terms of sex, cause of bleeding, delay between onset of bleeding and endoscopy, baseline hemoglobin, systolic blood pressure, and heart rate (Table 1). However, there was a small but significant difference between the 2 groups with regard to age ( years in the erythromycin group vs years in the placebo group; P 0.043). The mean time between the onset of Table 1. Characteristics ofthe Study Population and ofthe Subgroup ofpatients With Cirrhosis at Randomization Variable Erythromycin Placebo P value All patients (n) No. ofbleeding episodes Sex (M/F) 39/12 45/ Age (yr) Delayed bleeding/endoscopy (h) Hemoglobin (g/dl) a Systolic blood pressure (mm Hg) Source ofbleeding Gastric ulcer Duodenal ulcer Esophageal varix Peptic esophagitis Mallory Weiss syndrome Tumor Patients with cirrhosis (n) Sex (M/F) 11/2 14/ Age (yr) Delayed bleeding/endoscopy (h) Hemoglobin (g/dl) Systolic blood pressure (mm Hg) Child Pugh score HVPG (mm Hg) b HVPG, hepatic venous pressure gradient. a Data missing in 4 patients at the time ofrandomization. b Data available for 18 patients.

4 20 FROSSARD ET AL. GASTROENTEROLOGY Vol. 123, No. 1 Table 2. Effects of Erythromycin and Placebo in the Study Population and in the Subgroup of Patients With Cirrhosis Variable Erythromycin Placebo P value All episodes (n) Mean endoscopic score Empty stomach, n (%) 42 (82%) 18 (33%) Mean duration ofendoscopy (min) Therapeutic intervention, n (%) 22 (43%) 28 (52%) 0.34 Ulcers (sclerosed) Varices (sclerosed or banded) Quantity ofsclerosant (ml) Need for second-look endoscopy, n (%) 6 (12%) 17 (31%) 0.02 Mean number ofblood units transfused in 24 h Mean hospital stay (days) Patients with cirrhosis (n) Mean endoscopic score Empty stomach, n (%) 12 (92%) 4 (21%) Mean duration ofendoscopy (min) Need for second-look endoscopy n (%) 1 (8%) 9 (47%) 0.02 bleeding and endoscopy was hours and did not differ between the 2 groups. The distribution of the gastro-duodenal lesions responsible for the bleeding was similar to the usual frequency of GI bleeding seen in a daily practice, as reported by The National ASGE Survey on Upper Gastrointestinal Bleeding. 2 Each cause of bleeding was equally distributed between the 2 groups ( 2, P 0.05; Table 1). When the subgroup of patients with cirrhosis was analyzed, those with cirrhosis had lower hemoglobin values than those without cirrhosis at admission ( vs g/dl; P 0.03). In this subgroup, the source of bleeding was caused by variceal rupture in 29 cases and by gastro-duodenal ulcers in 3 cases. Primary End Points Erythromycin significantly increased the number of patients with an empty stomach, as assessed by both the objective mean endoscopic score and the subjective assessment by the endoscopist (empty vs. full stomach; Table 2). It is interesting to note that the 2 scoring systems correlated well with each other ( 2 test; P 0.001). In addition, the mean duration of the endoscopic procedure was shorter in the erythromycin group (Table 2). The need for sclerotherapy was not different between the 2 groups (Table 2). Therapeutic intervention was not feasible in 3 patients in the erythromycin group (immediate surgery) and in 1 in the placebo group (immediate surgery). In the erythromycin group, a diagnosis was reached in 42 patients (82%) at the time of the first endoscopy. In this group, the stomach was found to be full of blood in 9 patients (18%), all of whom required either immediate surgery (3 cases of bleeding ulcer) or a second-lookendoscopy (6 cases) for gastro-duodenal ulcer (4 cases), obstructive pyloric tumor (1 case), or esophageal variceal rupture (1 case; Table 2). In the placebo group, a diagnosis was reached in 36 patients (65%) at the time of the first endoscopy (18 patients with an empty stomach and 18 with a full stomach), although the stomach was found to be full of blood in 36 patients (65%). Among these 36 patients, 18 required either surgery (1 case of bleeding ulcer) or second-lookendoscopy for gastro-duodenal ulcer (8 cases), esophageal variceal rupture (7 cases), or Mallory Weiss syndrome (2 cases; Table 2). Secondary End Points Second-lookendoscopy was required more often in the placebo group than in the erythromycin group (17 vs. 6 cases; P 0.02; Table 2). The main reasons for second-lookendoscopy were either poor visibility during the initial endoscopy or bleeding recurrence. These causes were equally distributed in the erythromycin and placebo groups (Table 3). No endoscopy-related complication was reported throughout the study. No death occurred within the first 24 hours after the endoscopic procedure. The mean total number of blood units transfused during the first 24-hour period was similar in the 2 groups (Table 2). The duration of hospital stay was similar in both groups (Table 2). The number of patients with clinically relevant adverse events were as follows in the erythromycin group: 3 patients had an urgent bowel movement after the infusion, and 1 patient presented with moderate abdominal cramps. In the placebo group, 1 patient developed headache 30 minutes after the infusion. When we analyzed the effects of erythromycin in the subgroup of patients with cirrhosis, we found that erythromycin also increased the number of patients with an empty stomach, as assessed by the mean endoscopic score

5 July 2002 ERYTHROMYCIN FOR UPPER GI BLEEDING 21 Table 3. Reasons for Second-Look Endoscopy Within 24 Hours of Index Bleeding Variable Erythromycin Placebo P value No. second-look endoscopies Poor visibility, n (%) 4 (67%) 11 (65%) Bleeding recurrence, n (%) 2 (33%) 6 (35%) Previous endoscopic intervention, n (%) 2 (33%) 6 (35%) Quantity ofsclerosant at first endoscopy (ml) Mean hemoglobin at admission (g/dl) Mean number ofblood units transfused in 24 h and the endoscopist s judgment (empty vs. full stomach; Table 2). However, the duration of the endoscopic procedure did not differ in the 2 subgroups of patients with cirrhosis (Table 2). Additionally, erythromycin reduced the need for second-lookendoscopy in patients with cirrhosis (P 0.02; Table 2). Finally, we performed a logistic regression analysis to determine the factors that were independently associated with an empty stomach, the first end point in our study. Erythromycin administration and delay between the onset of bleeding and endoscopy were the 2 factors that were independently associated with an empty stomach in patients with recent hematemesis (Table 4). Discussion This study was designed to determine the effects of intravenous erythromycin infusion on emergency endoscopic yield in patients with recent hematemesis. Among prokinetic drugs, erythromycin, when given intravenously at a dose of 3 mg/kg, has the most powerful effects on gastric clearance. This has been established in healthy subjects as well as in patients with diabetic gastroparesis and critically ill patients. 28,29 As a motilin agonist, erythromycin stimulates antral contractions similar to phase III of the interdigestive migrating motor complex. 30 Although 80% 90% of bleeding episodes cease spontaneously at the time of admission, 31 the stomach may still be full of blood, which obscures the visual field and thereby decreases the endoscopic yield. Table 4. Logistic Regression Analysis ofvariables Associated With an Empty Stomach Variable P value Erythromycin Delayed bleeding/endoscopy 0.01 Hemoglobin at admission 0.08 Systolic blood pressure at admission 0.24 Bleeding source 0.33 Sex 0.39 Age 0.51 Cirrhosis 0.52 In this trial, we showed that intravenous erythromycin before endoscopy in patients with recent hematemesis shortened and improved the efficacy of the endoscopic procedure, resulting in a reduction of the need for second-lookendoscopy (Table 2). These findings held also true in the subgroup of patients with cirrhosis (Table 2). Second to erythromycin administration, the most important predictor of an empty stomach was a longer delay between hematemesis and endoscopy (Table 4). It should be noted that this delay was in fact slightly shorter in the erythromycin group, although this difference did not reach statistical significance ( hours vs hours; P 0.17). This even strengthens the results of this study, because longer spontaneous gastric emptying in the placebo group might have contributed to an underestimation of the effects of erythromycin. Although an imbalance between groups was noted, age was not a predictor of an empty stomach as assessed by the logistic regression analysis (Table 4) and does not seem to be a confounding factor in this trial, because ageing has no or very limited effects on gastric emptying. 32,33 A difference in hemoglobin concentrations was present in the 2 groups (8.8 g/dl in the erythromycin group vs. 8.0 g/dl in the placebo group; P 0.07) but was not statistically significant in univariate analysis; we wondered whether this difference might have contributed to influence our reported outcomes by favoring the erythromycin group. However, the results of the logistic regression analysis clearly ruled out such a possibility. To assess the yield of the endoscopy, we relied on both subjective and objective scoring systems. The subjective method we used was derived from the method reported by Avgerinos et al., 22 which measures the ease of an endoscopic procedure and ranges from easy to difficult. Good reproducibility of this subjective score contributed to establish the use of octreotide infusion at the early phase of upper GI bleeding in patients with portal hypertension. 22 Regarding the objective scoring system, we chose to use a simple method and scored the visibility of each area of the stomach, with a maximal score of 8 for this scoring system. The objective method was designed

6 22 FROSSARD ET AL. GASTROENTEROLOGY Vol. 123, No. 1 to provide a more precise and detailed pattern of blood accumulation within the stomach. In this trial, we gave the total score for the fundic, corporeal, antral, and bulbar areas. However, detailed analysis of the results obtained with this objective score showed that the fundic area was the worst area to be visualized during endoscopy, even in patients receiving erythromycin (data not shown). One possible explanation for this feature could be that erythromycin is known to preferentially change the antroduodenal motility and to reduce the proximal gastric tone. 28 These 2 methods were easy to use in a daily practice and provided the same information at the end of the endoscopy. No patient in this trial had a nasogastric tube or gastric lavage before endoscopy. This was required to meet the end points of this trial and was made possible by the availability of emergency endoscopy in a tertiarycare center. Indeed, the mean delay between hematemesis and endoscopy was very short in both study groups. It might be speculated that systematic intravenous erythromycin use might reduce the need for nasogastric tube placement, a method with a 79% sensitivity and 59% specificity to determine whether a patient is bleeding, 34 and the risks associated with nasogastric tube placement and gastric lavage, such as perforation or broncho-aspiration. 18 Although there was no attempt to remove blood in either group, removing blood by aspiration might presumably improve the endoscopic performance in a more efficient way in the erythromycin group than in the placebo one. The quantity of blood transfused and the length of hospital stay may be considered as indirect but robust indicators of the efficacy of interventional endoscopy. Although there was a tendency to reach significance, these 2 indicators were similar in both groups. In fact, sample size was designed to reach significance only for the primary end point of this trial. How can we account for the apparent lackof efficacy of intravenous erythromycin infusion in the remaining 18% of our patients (9 cases)? First, it should be noted that 4 of these patients had active ongoing bleeding, leading to surgery in 3 cases. Second, failure of erythromycin may also be explained in some patients by the infusion of a suboptimal dose. Indeed, we found that only 2 patients with full stomach in the erythromycin group had received a suboptimal dose of antibiotic. Because of the randomized, controlled design of this trial, a standardized dose of erythromycin (250 mg) was chosen, although it has been shown that the optimal dosage is 3 mg/kg. Finally, failure of erythromycin may also be explained in some patients because stress-induced hyperglycemia could have affected the gastric response to erythromycin, as was recently shown by Jones et al. 28 Our trial confirmed that in this population, erythromycin had an excellent safety profile. Indeed, allergic drug reactions to macrolides are extremely rare, 35,36 a feature that was also confirmed in this study. Because erythromycin is a macrolide antibiotic, we were concerned by the potential riskof inducing bacterial resistance. Although we are not aware of such cases in our cohort, most observers believe that the riskof inducing bacterial resistance strongly depends on the duration of administration and on the systemic half-life of the compound. 37 The design of this study is considered safe in terms of inducing bacterial resistance because erythromycin has a short systemic half-life (3.5 hours) and is administered as a single low dose. Throughout the study, no endoscopy-related complications were noted, and no deaths were directly related to the GI bleed. We had 2 delayed deaths; 1 occurred 2 weeks after the index bleeding as a result of postoperative death (surgery for gastric ulcer in a 74-year-old patient), and 1 patient died from infection 2 months after liver transplantation (3 months after the index bleeding). In conclusion, this is the first randomized, controlled trial that clearly establishes the clinical benefit of a motilin agonist, erythromycin, used before endoscopy in patients with recent hematemesis. Intravenous erythromycin infusion, 20 minutes before endoscopy, should be considered as an effective means to optimize the yield of the emergency interventional procedure for upper GI bleeds. Systemic intravenous erythromycin use might reduce the need for and the risks associated with gastric lavage. References 1. Morgan A, McAdam W, Walmsley G. Clinical findings, early endoscopy and multivariate analysis in patients bleeding from the upper gastrointestinal tract. Br Med J 1977;2: Silverstein F, Gilbert D, Tedesco F, Buenger N, Persing J. The national ASGE survey on upper gastrointestinal bleeding. Gastrointest Endosc 1981;27: Freeman M, Cass O, Peine C, Onstad G. The non-bleeding visible vessel versus the sentinel clot: natural history and risk ofrebleeding. Gastrointest Endosc 1993;39: Laine L. Refining the prognostic value ofendoscopy in patients presenting with bleeding ulcers. Gastrointest Endosc 1993;39: Laine L, Peterson W. Bleeding peptic ulcer. N Engl J Med 1994; 331: Cook D, Guyatt G, Salena B, Laine L. Endoscopic therapy for acute nonvariceal upper gastrointestinal hemorrhage: a metaanalysis. Gastroenterology 1992;102: Sacks H, Chalmers T, Blum A, Berrier J, Pagano D. Endoscopic hemostasis: an effective therapy for bleeding peptic ulcers. JAMA 1990;264: Morissey J, Reichelderfer M. Gastrointestinal endoscopy. N Engl J Med 1991;325:

7 July 2002 ERYTHROMYCIN FOR UPPER GI BLEEDING Zuccaro G. Bleeding peptic ulcer: pathogenesis and endoscopic therapy. Gastroenterol Clin North Am 1993;22: Besson I, Person B, Boutroux D, Heresbach D, Bernard P, Hochain P, Larricq J, Gourlaouen A, Ribard D, Kara N, Legoux J, Pillegand B, Becker M, Di Costanzo J, Metreau J, Silvain C, Beauchant M. Sclerotherapy with or without octreotide for acute variceal bleeding. N Engl J Med 1995;333: Choudari C, Rajgopal C, Elton R, Palmer K. Failures ofendoscopic therapy for bleeding peptic ulcer: an analysis of risk factors. Am J Gastroenterol 1994;89: Brullet E, Calvet X, Campo R, Rue M, Catot L, Donoso L. Factors predicting failure of endoscopic injection therapy in bleeding duodenal ulcer. Gastrointest Endosc 1996;43: Charles L, Segreti J. Choosing the right macrolide antibiotic. Drugs 1997;53: Lin H, Sanders S, Gu Y, Doty J. Erythromycin accelerates solid emptying at the expense ofgastric sieving. Dig Dis Sci 1994;39: Weber F, Richards R, McCallum R. Erythromycin: a motilin agonist and gastrointestinal prokinetic agent. Am J Gastroenterol 1993; 88: Itoh K, Nakaya M, Suzuki T, Arai H, Wakabayashi K. Erythromycin mimics exogenous motilin in gastrointestinal contractile activity in the dog. Am J Physiol 1984;247: Peeters T, Matthijs G, Depoortere I, Cachet T, Hoogmartens J, Vantrappen G. Erythromycin is a motilin receptor agonist. Am J Physiol 1989;257: Savides T, Jensen D. Therapeutic endoscopy for nonvariceal gastrointestinal bleeding. Gastroenterol Clin North Am 2000;29: Brandiss M, Richardson W, Barold S. Erythromycin-induced QT prolongation and polymorphic ventricular tachycardia (torsades de pointes): case report and review. Clin Infect Dis 1994;18: Paris D, Parente T, Bruschetta H, Guzman E, Niarchos A. Torsades de pointes induced by erythromycin and terfenadine. Am J Emerg Med 1994;12: Periti P, Mazzei T, Mini E, Novelli A. Pharmacokinetic drug interactions ofmacrolides. Clin Pharmacokinet 1992;23: Avgerinos A, Nevens F, Raptis S, Fevery J. Early administration of somatostatin and efficacy of sclerotherapy in acute oesophageal variceal bleeds: the European acute bleeding oesophageal variceal episodes (ABOVE) randomised trial. Lancet 1997;350: Burroughs A, Alexandrino P, Calès P, Fleig W, Grace N, Minoli G, Siringo S. Sore points. A review ofthe points where there was disagreement at Baveno I and an attempt to reach consensus. In: de Franchis R, ed. Portal hypertension II. Oxford: Blackwell Science, 1996: Rollhauser C, Fleischer D. Nonvariceal upper gastrointestinal bleeding: an update. Endoscopy 1997;29: Hsu P, Lin X, Chan S, Lin C, Chang T, Shin J, Hsu L, Yang C, Chen K. Bleeding peptic ulcer risk factors for rebleeding and sequential changes in endoscopic features. Gut 1994;35: Khuroo M, Yattoo G, Javid G, Khan B, Shah A, Gulzar G, Sodi J. A comparison ofomeprazole and placebo for bleeding peptic ulcer. N Engl J Med 1997;336: Lau JY, Sung JJ, Lee KK, Yung MY, Wong SK, Wu JC, Chan FK, Ng EK, You JH, Lee CW, Chan AC, Chung SC. Effect of intravenous omeprazole on recurrent bleeding after endoscopic treatment of bleeding peptic ulcers. N Engl J Med 2000;343: Jones K, Kong M, Berry M, Rayner C, Adamson U, Horowitz M. The effect of erythromycin on gastric emptying is modified by physiological changes in the blood glucose concentration. Am J Gastroenterol 1999;94: Chapman M, Fraser R, Kluger M, Buist M, De Nichilo D. Erythromycin improves gastric emptying in critically ill patients intolerant ofnasogastric feeding. Crit Care Med 2000;28: Janssens J, Peeters T, Vantrappen G, Tack J, Urbain J, De Roo M, Muls E, Bouillon R. Improvement ofgastric emptying in diabetic gastroparesis by erythromycin. Preliminary studies. N Engl J Med 1990;322: Silverstein F, Feld A, Gilbert D. Upper gastrointestinal bleeding. Arch Intern Med 1981;141: Gainsborough N, Maskrey V, Nelson M, Keating J, Sherwood R, Jackson S, Swift C. The association of age with gastric emptying. Age Ageing 1993;22: Nakae Y, Onouchi H, Kagaya M, Kondo T. Effects of aging and gastric lipolysis on gastric emptying oflipid in liquid meal. J Gastroenterol 1999;34: Cuellar R, Gavaler J, Alexander J, Brouillette D, Chien M, Yoo Y, Rabinovitz M, Stone B, Van Thiel D. Gastrointestinal tract hemorrhage: the value ofa nasogastric aspirate. Arch Intern Med 1990;150: Amsden G. Erythromycin, clarithromycin and azithromycin: are the differences real? Clin Ther 1996;18: Rubinstein E. Comparative safety of the different macrolides. Int J Antimicrob Agents 2001;18: Periti P, Mazzei T, Mini E, Novelli A. Adverse effects of macrolide antibacterials. Drug Saf1993;9: Received March 18, Accepted March 21, Address requests for reprints to: Jean Louis Frossard, M.D., Division of Gastroenterology, Geneva University Hospital, 1211 Geneva 14, Switzerland. jean-louis.frossard@hcuge.ch; fax: (41)