Helicobacter pylori and peptic-ulcer disease
*J R Boulton-Jones, C J Hawkey

Their conclusions on the prevalence of endoscopically detected ulcers are based on four or five studies selected from 16 that met their inclusion criteria. Although they state reasons for excluding reports, six of those not selected do not support their findings. Furthermore, the conclusions of other well designed studies addressing this question, which did not meet their inclusion criteria, differ from their own. Moreover, in the selected studies, no control who had an ulcer was negative for H pylori and NSAID use. This zero rate required statistical modelling by adding a constant pseudo count. Therefore, data calculated by comparison are distorted pseudo estimates. Clinical trials, which provide a far larger, more rigorous, controlled, prospective data source were not analysed. In such studies, results have shown effects that range from increased risk of ulceration in H pylori-positive patients taking NSAIDs2 to no effect and greater effectiveness of acid-suppression,2 and overall do not support Huang and colleagues' conclusions. Selection bias may also exist in the analysis of ulcer bleeding. A large study that was not analysed shows an opposite effect to that of Huang and colleagues.3 Again, clinical trials in which clinically important ulcer complications were the primary endpoint do not support a role for H pylori,4 whereas a significant reduction in such events, in patients on NSAIDs or rofecoxib in endoscopic studies, has been reported.5 If these papers had been included, no additive risk between H pylori and NSAIDs on ulcer bleeding would have been evident. The aims and methods of the included studies are heterogeneous, which makes it debatable whether meta-analysis, defined as combined analysis of similar studies to produce a unifying conclusion, is an appropriate statistical tool. For reasons that are unclear, the primary conclusion of the nine studies of ulcer bleeding differ from Huang and colleagues' conclusions (reduced ulceration in two, reduced gastric ulceration in one, no effect in four, enhanced ulcer risk in one, and lowered subgroup risk in one). Prescribers need to know what to do when an NSAID user presents with a clinical event such as a bleeding peptic ulcer. Trial data show that 18·8% of such patients will have a further ulcer bleeding episode if they have H pylori eradication and continue taking an NSAID, compared with 4·4% if they continue taking NSAIDs and a protective proton-pump inhibitor and do not have H pylori eradication. Since there are no data, prescribers can make their own decisions about giving H pylori eradication as well as proton-pump inhibitor, but should bear in mind that H pylori eradication will reduce the effectiveness of proton pump inhibitors at lowering the concentration of intragastric acid by a hundred-fold.

The Lancet, published, and copyrighted © 2003. All rights reserved.