We read with interest the meta-analysis by Sung et al (Gut 2007;56:1364–73) who conclude that successful application of haemoclips is superior to injection alone for patients with non-variceal upper gastrointestinal bleeding (NVUGIB). We are concerned with their conclusions, which we do not believe are justified due to several methodological errors in their study.

The authors combined six randomised controlled trials for peptic ulcer bleeding (PUB) and two studies for Dieulafoy lesions in the analysis comparing haemoclips and injection. The two Dieulafoy studies significantly favoured the outcomes of interest (initial and definitive haemostasis, rebleeding and surgery). It is known that Dieulafoy lesions have unique endoscopic and histopathological characteristics and are more likely to respond to mechanical endoscopic haemostasis than PUD, due to protrusion of the arterial vessel …