All Times EDT
Keywords: surrogacy, neurodegenerative, causal inference
Surrogate endpoints are generally used when it is impractical to conduct a clinical trial using clinical endpoint, thus could substantially expediate the drug development programs. However, it is challenging to evaluate the surrogacy of biomarkers in clinical trials as well as its quantitative relationship between to clinical endpoint of interest. In this presentation, a model-based approach with casual inference component built-in is proposed to characterize the drug effect on clinical benefit through biomarker pathway. Given the underlying assumption that the drug has a positive effect on the biomarker, the model is composed of three components: natural disease progression trajectory, the drug effect through the biomarker pathway, and the effect through the non-biomarker pathway where could be beneficial or adverse. In a well-controlled randomized clinical trial, the natural disease progression is captured through the placebo arm, while the trajectory of the disease for patients on treatment would be altered due to the drug effect on biomarker and clinical endpoint. An example will be provided to illustrate the application of the constructed model. In the example, early improvement in biomarker level will be used to predict the clinical benefit on the clinical endpoint at a later time point. Maximum likelihood estimation is utilized to solve the parameters of interest in the constructed model with multiple regression equations. An R code for parameter estimation will also be demonstrated in the presentation.