Erythromycin for Improved Upper Endoscopy

10 for decreased need for second endoscopy

Benefits in NNT

1 in 10 were helped (decreased need for second endoscopy)
10% were helped by decreased need for second endoscopy (26% control versus 16% erythromycin)

Harms in NNT

1 in 33 were harmed (GI discomfort)
None were harmed (increased transaminases or cardiac dysrhytmias)
3% were harmed by GI discomfort
None were harmed with increased transaminases or cardiac dysrhytmias
View As:

Efficacy Endpoints

Visualization of gastric mucosa, need for 2nd endoscopy, hospital length of stay, and units of pRBC transfused

Harm Endpoints

GI symptoms and cardiac arrythmias


Upper gastrointestinal bleeding is a medical emergency that accounts for more than 500,000 hospital admissions each year and has a significant mortality of 5-10% with standard medical therapy8. Urgent endoscopy with adequate visualization of the gastric mucosa is essential for the identification of the source of bleeding and for treatment of these lesions. Traditionally, nasogastric (NG) lavage has been used to improve visualization, but this technique carries with it the significant burden of patient discomfort (pain and nausea) and risk (misplaced tube, risk of clot dislodgment). Erythromycin, a macrolide antibiotic which also acts as a motilin receptor agonist, facilitates the motility of the gastric antrum and duodenum and thereby accelerates gastric emptying. For this reason, erythromycin has been proposed as a pre-endoscopy aid in visualization and treatment of bleeding lesions.
This meta-analysis includes seven trials (n = 558) involving adult patients with upper GI bleeding. All were randomized, controlled trials (RCT) comparing erythromycin administration versus no erythromycin administration prior to endoscopy All seven of these studies evaluated the primary endpoints of adequate visualization of the gastric mucosa and need for a second-look endoscopy. Erythromycin treatment resulted in a statistically-significant increase in visualization of gastric mucosa and a statistically-significant decrease in the need for second-look endoscopy. Secondary outcomes demonstrated statistically-significant decreases in packed red blood cells (PRBC) transfusions (5 trials, n = 504, 0.4 units of pRBCs) and hospital length of stay (4 trials, n = 335, 1.5 days). No statistically significant difference was observed in duration of endoscopy (4 studies). The meta-analysis did not explore harm outcomes, but several individual studies monitored for adverse outcomes. The medication was generally well tolerated, with only a few minor adverse reactions reported (GI discomfort).


Two of the studies were not blinded, which introduces a risk of bias. However, this was addressed by the authors with a sensitivity analysis in which both of these studies were removed; the results remained similar in all measured outcomes. There was also considerable dose variation among the included trials, although studies have suggested effect in as low as 70mg6, which is well below even the lowest doses used in these trials. In addition, as the authors point out, given the small number of patients in the studies, mortality could not be fully assessed in this meta-analysis.

It is unclear whether improved gastric visualization could be considered a patient-important outcome, and is therefore not included in the summary above. However, need for second-look endoscopy, as well as secondary outcomes such as length of stay and need for transfusion are more directly relevant to a patient-centered approach. Statistically significant heterogeneity was found in two outcomes (gastric visualization and procedural duration). However, sensitivity analysis did not change the overall magnitude of results. This heterogeneity may have been due to differences in study population, particularly in the percentage of the population with acute variceal bleeding.

It is suggested in the discussion that erythromycin may be less effective in variceal bleeding, although this was not specifically addressed in the meta-analysis. Finally, two of the studies utilized NG lavage, a significant potential confounder, before the administration of erythromycin. One prospective, randomized, multicenter trial5 compared intravenous erythromycin infusion without NG lavage, NG lavage without erythromycin, and the combination of both erythromycin and NG lavage. No statistically-significant difference was seen between the three groups. This suggests that erythromycin and NG lavage may be similar in efficacy, and that the combination of the two adds no statistically-significant benefit.

Given these limitations, there is definite need for further research to examine the efficacy of erythromycin in this setting. However, given the reported benefits of decreased need for second look endoscopy, decreased PRBC units transfused, and decreased hospital length-of-stay, with rare, minor adverse events, we recommend this intervention in adult patients with acute upper GI bleeding, particularly in those for whom NGT is contraindicated or less desirable, and give it a rating of Green.


Brian Resler, MD; Gary Green, MD


May 1, 2015