Statistics from Altmetric.com
Editor,—Eggleston et al(Gut 1998;42:13–16) concluded from their observational study that the “step up” approach, starting with a prokinetic or H2 receptor antagonist, represents the most cost effective initial therapeutic strategy for the treatment of patients with a diagnosis of uncomplicated gastro-oesophageal reflux disease (GORD). In studies of the effects of treatments, non-randomised comparisons can be affected by confounding factors1 and in our opinion this has resulted in the wrong conclusions being drawn from this study. Therapeutic interventions are commonly prompted by an indication. Primary care physicians selectively treat patients exhibiting more severe GORD with omeprazole and less severe symptoms with cisapride or ranitidine. Thus, by the very rationality of the decision to intervene, treatment differs between patients with GORD with respect to the outcome criterion in efficacy assessment.2
The authors minimised confounding by indication by excluding 511 of 790 patients with a diagnosis of dysphagia or complicated GORD or whose condition was sufficiently complex to require immediate specialist referral. Restricting subject selection indeed prevents confounding by indication, but introduces a change in the magnitude of the effect measured, which is called effect modification. By only evaluating patients with uncomplicated GORD, the difference in the effect measured between cisapride, ranitidine and omeprazole treatment will decrease. Many randomised clinical trails have shown that in patients with GORD omeprazole provides more rapid and complete healing of associated symptoms than ranitidine or cisapride.3 Therefore, it is misleading to minimise confounding by restricting subject selection, unless it can be assumed that there are no differences in treatment effect between patients. To control for confounding by indication, several other options are available. However, even when confounding factors are taken into account, the validity of findings on drug benefits in non-randomised studies may still be questionable.1
Besides confounding by indication and effect modification, the authors misused the term cost effectiveness.4 Cost effectiveness analysis is one form of full economic evaluation in which both the cost and consequences of treatments are examined.5 Egglestonet al assumed equivalence in therapeutic outcome between the cisapride, ranitidine and omeprazole therapies on the basis of retrospective analysis of prescription data. If this assumption is correct then the design of their study should be called a cost minimisation analysis. The term cost effectiveness should be restricted to studies in which the result can be described as having an additional benefit worth the additional cost. To determine which treatment strategy has additional benefit worth the financial effort, more information is required about symptom relief. Furthermore, the claim of substantial differences in cost cannot be verified without the use of statistical testing. Although statistical testing of cost and cost effectiveness ratios is not as obvious as statistical testing of outcomes, methods are currently being developed to overcome these limitations.6
In conclusion, the study by Eggleston et al has several methodological shortcomings which have led to unfounded and therefore incorrect conclusions. To avoid the limitations of this study, a prospective randomised trial with representative patients7must be carried out to determine which treatment approach, “step up” or “step down”, is best for treating patients with GORD and dyspeptic symptoms in general.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.