Economic evaluations alongside Stepped Wedge Trials: the need for appropriate statistical methods
A new study by The George Institute for Global Health has reviewed statistical methods used when conducting economic evaluation alongside stepped wedge trials (SWTs). We asked lead author and Head of the Statistics Division in Australia, Gian Luca Di Tanna, to tell us about the significance of this study for future clinical trials.
What is a stepped wedge trial?
An SWT is a multi-arm cluster randomised design, with a unidirectional cross-over: clusters switch from the control or routine care condition to the active intervention. In the most common design, all clusters begin in the control condition before crossing over at regular intervals to receive the intervention in a randomised order. This process continues until all clusters are exposed to the intervention. Due to their pragmatic design, SWTs have been used to assess the effect of new interventions in real-world settings where its not possible to apply the intervention across all clusters at the same time for logistical, financial or ethical reasons.
Why is this review important?
Stepped wedge trial design has become increasingly popular in health services research. Many health services studies also include a health economic evaluation to assess the economic benefit of the intervention being investigated. However, adequate documentation of appropriate methods of economic evaluation in an SWT is generally lacking in the literature.
What did you do?
We conducted a methodological systematic review of studies (protocols, trials results, cost effectiveness/utility analyses) that described an economic evaluation alongside an SWT. We assessed their eligibility, findings, reporting of statistical methods and quality of reporting.
What did you find?
We looked at 69 studies that were eligible for our review. In around half of these, the reporting of statistical methods associated with the economic evaluation was incomplete. Around half the remaining studies did not use appropriate statistical methods, failing to consider critical elements of an SWT such as correlation between costs and outcomes, clustering and time patterns.
What are the implications of this study?
This is the first attempt to review the statistical methodology of planned and conducted economic evaluation of SWTs to date. We intend to propose appropriate methods and perform a simulation study to assess the performance of SWTs to better inform methodologists and applied researchers interested in cost-effectiveness analyses of trials that are based on this design.
