Proton pump inhibitors (PPIs) have become an indispensable part of the armoury of treating gastro-oesophageal reflux disease (GORD). They are more effective than H₂ receptor antagonists and prokinetic agents in oesophagitis and endoscopy negative reflux disease.1 This efficacy comes at a price however and PPIs are the most expensive class of drug in the UK, costing nearly £300 million in 1998.2 There have been attempts to curb PPI expenditure and the UK National Institute for Clinical Excellence has issued guidelines on appropriate prescription.2

The problem is that there is a dearth of health economic data to inform those making health care decisions. Economic models have suggested that PPIs are cost effective therapy for GORD but there have been criticisms of this approach3 and they do not address the root of the problem. Economic decisions are simple when a drug that is more expensive and less effective is compared with a cheaper more effective agent. The latter is said to “dominate” the former and it requires little health economic expertise to realise that the cheaper drug should be used. Problems arise when a more expensive therapy is more effective than the cheaper option. The choice is less obvious in these circumstances and the decision depends on how much patients would be willing to pay to cure their symptoms. If they are willing to pay more than the extra cost per extra cure conferred by the more expensive therapy then it should be instituted, otherwise the cheaper drug should be prescribed. The choice of therapy is likely to depend on the severity of symptoms and once these interfere sufficiently with quality of life PPIs will represent value for money. There have been approximately 5000 publications evaluating PPIs and yet none has assessed how severe GORD symptoms need to be before patients are willing to pay for these drugs compared with cheaper alternatives.

The imbalance between the clinical and health economic information relating to PPIs has been partially redressed by Myrvoldet al in this issue ofGut (see page 488).4 They have performed a rigorous randomised controlled trial comparing omeprazole with open antireflux surgery. There was no statistically significant difference in relapse rates between the PPI and surgery arm so the two interventions were considered to have a similar efficacy. If both treatments work equally well, the amount patients are willing to pay for cure of their symptoms becomes less relevant and the cheapest strategy should be the most efficient. This type of economic evaluation is termed a cost minimisation analysis. Myrvold et al followed patients for five years and found that direct medical costs were statistically significantly higher in the surgery than the medical arm in Sweden, Denmark, and Norway but not in Finland. These data suggest long term PPI therapy is more cost effective than antireflux surgery in most countries studied.

These results should however be interpreted cautiously. The main concern is that a cost minimisation analysis was performed. This is rarely an appropriate form of economic analysis5 and in this study the absence of a statistically significant difference between the two interventions does not imply they have the same efficacy. Antireflux surgery may be more effective than PPI therapy but the trial did not have the power to detect the size of the effect. Indeed, heartburn scores were statistically significantly lower in the surgery arm compared with the omeprazole arm and the lack of difference between the two groups could be due to the artificial method the authors used to define treatment failure.6 An analysis that looked at costs and benefits with a confidence interval around the incremental cost effectiveness estimate would be the most appropriate method of presenting the data.7

There are other concerns about the analysis of the data. Parametric tests were used to compare the costs between the two interventions. Cost data are usually highly positively skewed and parametric statistics are often inappropriate.8 There was also a moderate drop out rate in the treatment groups and these patients were assumed to have incurred no costs. There are accepted methods of dealing with incomplete follow up in economic analyses9and it would be unusual to assume that these patients cost nothing.

There is also debate about the appropriateness of conducting an economic analysis alongside a randomised controlled trial evaluating the clinical efficacy of different treatments.10 The power of the trial is usually centred on the detection of differences in clinical outcomes and the sample size is rarely adequate for detecting differences in cost benefit.5 Economic trials need to be as pragmatic as possible while assessment of clinical outcome usually requires many more follow up visits and inves- tigations than would occur in clinical practice.11 Myrvoldet al have tried to control for this by not including protocol driven visits in their cost calculations. These extra visits may change patients' attitudes to their symptoms however and may also alter clinician management decisions. This may have an influence on the costs patients incur and the economic analysis may not reflect clinical practice.

Myrvold et al report one of the few economic analyses of PPI therapy using randomised controlled trial data. Economic evaluation of health care interventions present new challenges in trial design and analysis and this is a rapidly evolving discipline.7 The results are difficult to interpret but the authors should still be congratulated for providing data in this important field. Rationing is inevitable in health care and until we have more studies addressing the cost effectiveness of PPI therapy in GORD there will remain a tension between clinicians wanting to give the best drug to their patients and the health care payers wanting to cut costs.