• Helicobacter pylori
• decision analysis
• dyspepsia
• proton pump inhibitors

Decision analysis can be used to bridge the gap between evidence and practice. It may be used to highlight the best choice between competing management strategies where no direct trial evidence is available or add additional information, for example cost data, to extrapolate from available trial evidence.¹ The use of sensitivity analysis, changing the parameters such as the prevalence of Helicobacter pylori, is particularly helpful in establishing an “envelope” within which a particular strategy is cost effective. Trials, meta-analyses, and modelling can be seen as working together to create a practical quantified evidence base.

Cochrane reviews on the management of dyspepsia,² therapies for non-ulcer dyspepsia,^3,4 and the treatment of endoscopy negative and uninvestigated gastro-oesophageal reflux disease⁵ have created a sufficiently rich evidence base upon which to build sophisticated models incorporating alternative approaches to both investigative and empirical treatment strategies in dyspepsia. Health service researchers in both the USA⁶ and the UK⁷ have now published models that in spite of methodological differences (discrete event simulation v Monte Carlo simulation)⁸ come to similar conclusions as to the cost effectiveness of the H pylori “test and treat strategy” compared with endoscopy and empiric proton pump inhibitor (PPI) treatment in young patients without alarm symptoms.

As dyspepsia is a chronic relapsing disorder, the management of patients with persisting symptoms is as important as the initial management strategy. This is difficult to capture in randomised trials where complex algorithms and planned follow up may take trials too far from everyday practice. The effect of possible “follow on” strategies can be explored by modelling. Both the UK and US models allowed for endoscopy of treatment failures. The UK model did not use the same base case as the US model, “test and treat” followed by endoscopy for treatment failures, as it was assumed that in UK primary care a trial of PPI would always be interposed before endoscopy referral. The US analysis is useful in showing that this early use of endoscopy is indeed less cost effective than interposing a trial of PPI therapy.

The message from both models is that endoscopy adds little to the management of dyspeptic patients under the age of 55 years without alarm symptoms for malignancy, and that the most important uncertainty is whether to test and treat for H pylori initially, or after PPI therapy, or not at all. The balance depends on the prevalence of H pylori related and acid related disease in the population and the relative costs of endoscopy, eradication therapy, and PPIs.

Recent trials confirm that “test and treat” is as effective and less costly than endoscopy based management in patients referred to a dyspepsia clinic.⁹ A recent Canadian trial has also shown that “test and treat” in primary care is more effective than initial PPI therapy at preventing recurrence of dyspeptic symptoms at one year in H pylori positive patients (NNT=7).¹⁰ A significant difference in costs was not shown but exclusion of H pylori negative patients makes the economic comparison incomplete.

In the UK, the lower cost of H pylori eradication therapy relative to endoscopy and PPIs should lower the threshold for a “test and treat” approach. A large primary care based MRC funded trial (CUBE) which has just started aims to determine the cost effectiveness of initial “test and treat” compared with initial empirical acid suppression. Until there is more evidence, either approach is justifiable.