The generalisability of pharmacoeconomic studies

Abstract

Authors of pharmacoeconomic analyses understandably want their findings to apply as broadly as possible. Also, decision-makers may have to interpret the results of analyses conducted in healthcare settings other than their own. The validity of transferring or generalising results from one setting to another raises important issues for health-economic evaluation. Pharmacoeconomic analyses attempt to model the costs and benefits of alternative treatments in normal clinical practice. Usually, no single clinical study directly provides all the required information, and a variety of data sources is generally included in each analysis. Different data sources present different problems in terms of their relevance to decision-makers. At one extreme, an analysis based purely on trial outcomes and resource use may be precise, but not reflect normal practice; at the other extreme, an analysis using practice data may appear relevant, but be exposed to biases and confounding. Reviews of published studies suggest that general standards have been inadequate in the past. Reapplying such analyses in different localities may simply replicate inadequate findings. The 'perfect' should not become the enemy of the merely 'good'. Models can be helpful in decision-making, provided that they accurately communicate uncertainties in modelling and data. Even so, there will be limits to the generalisability of pharmacoeconomic models, since the required analysis differs between jurisdictions, and because of variations in normal clinical practice. The transferability of research findings re-opens the issue of credibility in pharmacoeconomics. Methodological standardisation, reporting standards and researcher independence are recognised as important factors for enhancing credibility. Where possible, pharmacoeconomic analyses should reflect the findings of systematic reviews of health outcomes to avoid the risk of biased selection of the evidence. In addition, the application of findings to individual healthcare settings must be considered, since cost effectiveness may vary markedly by setting and perspective.