Evaluation of Methods to Detect Legacy Effects in Cardiovascular Post-Trial Studies

Publication Type:
Issue Date:
Full metadata record
Post-trial follow-up studies after randomized controlled trials (RCT) are increasingly used to investigate the clinical effectiveness of an intervention in the long term. “Legacy effect”, which was proposed in the context of such studies, describes the effects of an intervention that are only observed after the end of trial and are not due to the direct effects observed during the trial period itself. Much of the clinical interest in legacy effects has been in the drug treatments for cardiovascular disease prevention, as the finding of such effect could provide support for earlier initiation of the intervention. However, limited attention has been paid to the methodological challenges of analysing post-trial data. In this thesis, I provide a summary of the methods used, and evaluate the potential for bias, in the cardiovascular post-trial studies. I also investigate how we might best analyse data from a matching RCT and post-trial follow-up study, specifically, the choice of time period and trial participants to include in analysis and the strategy to correct for potential selection bias and confounding. Simulations are conducted to compare the performance of different methods. I use data from the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial and its follow-up data to illustrate the application of different approaches. Analyses combining both the initial trial period and the post-trial follow-up period have often been incorrectly interpreted as evidence of a legacy effect, which is better assessed on the basis of separate post-trial analysis. To address the issues of selection bias and potential confounding requires appropriate study designs and rigorous methods of analysis. The choice of statistical methods should consider the availability of post-trial data, size of direct treatment effect and causal pathway of legacy effect. It is recommended to conduct a sensitivity analysis to check the robustness of the findings. Better reporting of legacy effects is needed to realize their full value in informing clinical practice and health policy.
Please use this identifier to cite or link to this item: