Chair: Yuan Wu, PhD (Duke)
Instructors:
Bie Verbist, PhD (J&J)
Wim Van der Elst, PhD (J&J)
Course Description:
The duration, complexity, and cost of a clinical trial are substantially affected by the endpoints used to assess treatment efficacy (Burzykowski, Molenberghs, and Buyse 2005; Alonso et al., 2016). In some situations, the most credible indicator of therapeutic response, the so-called true endpoint, may be distant in time (e.g., survival time in early cancer stages), rare (e.g., pregnancy in severe luteinizing hormone deficiency), ethically challenging (e.g., procedures that involve a nonnegligible health risk), or expensive (e.g., imaging data). An appealing strategy in these circumstances is to substitute the true endpoint by a “replacement endpoint” that can be measured earlier, occurs more frequently, is more ethically acceptable, and/or is cheaper. If such a replacement endpoint allows for the accurate prediction of the treatment effect on the true endpoint, it is called a surrogate endpoint.
Surrogate endpoints play an important role in oncology, as about two-thirds of the cancer drugs that were approved by FDA in recent years are based on surrogate endpoints such as tumor shrinkage or progression free survival (Walia et al., 2022). The statistical evaluation of surrogate endpoints is not a trivial endeavor, and various initial proposals turned out be fundamentally flawed. In particular, the so-called single-trial methods (where the data of a single clinical trial are used to evaluate a surrogate) were shown to be inappropriate, because they do not allow to evaluate the extent to which the treatment effect on the true endpoint is related to the treatment effect on the surrogate endpoint.
In this short course, the focus will be on the so-called meta-analytic surrogate endpoint evaluation approach (Buyse et al., 2000). This approach allows for evaluating surrogacy (i) at the level of the individual patients, i.e., how well can we predict a patient’s true endpoint based on his/her surrogate endpoint? and (ii) at the level of the clinical trial, i.e., how well can we predict the treatment effect on the true endpoint when we know the treatment effect on the surrogate endpoint in a new clinical tria
Instructors:
Bie Verbist, PhD
Translational Medicine & Early Development Statistics Discovery
Johnson & Johnson
Wim Van der Elst, PhD
Translational Medicine & Early Development Statistics Discovery
Johnson & Johnson
Bie Verbist holds a PhD in organic chemistry from Catholic University of Leuven, BE (2005) and a PhD in applied statistics from Gent University, BE (2014). Bie joined J&J immediately after where she initially performed modelling and multivariate statistics to support biomarker discovery. Gradually she took up increased responsibilities where she leads the discovery statistics oncology team. She has experience in various fields in support of oncology with some major focuses on mabel dose estimation, combination experiments and biomarker discovery.
Wim Van der Elst holds a PhD in psychometrics from Maastricht University, NL (2006) and a PhD in statistics from Hasselt University, BE (2016). Wim joined J&J immediately after where he joined the manufacturing statistics department. In 2019, he moved to the discovery statistics neuroscience team. Wim has a major interest in validation of surrogate endpoints, a field in which he published various methodological papers (especially in the causal-inference setting) and is the lead programmer of the Surrogate package in R.
