Robust Statistical Methods to Identify and Use Surrogate Markers in Diabetes

点击翻译按钮获取中文摘要

Project Summary/Abstract Diabetes prevention studies often require long term follow-up of patients in order to observe a sufficient number of diabetes diagnoses to precisely estimate treatment effects. In such settings, the availability of a surrogate marker that could be used to estimate the treatment effect and could be observed earlier than the occurrence of a diabetes diagnosis would allow researchers to make conclusions regarding the effectiveness of a given treatment with less required follow-up time. That is, validated surrogate markers could enable shorter randomized clinical trials and require smaller sample sizes, thus accelerating acquisition of clinical information. Identifying such surrogate markers, determining when these markers should be collected, and developing tools to use these markers to test whether a treatment is effective in future studies would contribute significantly to research aimed at identifying effective preventative treatments for diabetes. Research on identifying useful surrogate markers has largely focused on estimation of the proportion of treatment effect explained by a surrogate marker since a valid surrogate marker should capture a large proportion of the true treatment effect on the primary outcome. However, current methods to estimate the proportion of treatment effect explained have a number of limitations. In particular, they often require restrictive model assumptions that may not hold in practice and they often only allow for the evaluation of single surrogate marker measured at a single point in time. In addition, current methods do not provide any guidance regarding how to actually use an identified valid surrogate marker to test for a treatment effect earlier in a future study. In this study, we aim to shift current research practice on surrogate marker evaluation away from restrictive model-based approaches towards robust estimation approaches that can evaluate complex surrogate marker information by proposing novel methods that allow for more flexible model assumptions. Specifically, we propose to develop novel statistical methods to estimate the proportion of treatment effect explained by surrogate marker measurements over time and by multiple surrogate markers, and identify how such surrogate marker information can be used to test for treatment effectiveness in a future study, thereby allowing for less required follow-up time and shorter trials. We additionally propose to develop methods to identify heterogeneity in the utility of a surrogate marker and a procedure to account for such heterogeneity when using the surrogate marker to test for a treatment effect in a future study. We will apply these methods to data from the Diabetes Prevention Program study to comprehensively evaluate and identify potential surrogate markers of diabetes and to produce tools such that identified surrogate markers could be used to test for effective treatments in future diabetes studies.