James A. Rogers, PhD: No financial relationships to disclose
For a biomarker to be at least a “level 3 surrogate” that is “reasonably likely to predict clinical benefit for a specific disease and class of interventions” (1), it must be either a mediator on the causal pathway between treatment and response or else be causally downstream of such a mediator (1, 2). Causal mediation analysis is an attempt to assess the statistical evidence for or against such causal hypotheses. At present, causal mediation analysis is less well developed for dynamic models, in which a time-varying biomarker process may interact (perhaps iteratively) with a time-varying clinical outcome process. In some cases, however, a reasonable “approximate answer to the right question” may be obtained via “landmarking” the biomarker process at a particular timepoint t, and modeling the clinical outcome data after time t (3). We reprise the existing theory that supports this approach, formalizing estimands of interest in terms of natural direct and natural indirect effects, and demonstrate a very general estimation methodology(4) that may be used in the context of models familiar to the pharmacometrics community, specifically nonlinear mixed effects models for the biomarker process and fully parametric time-to-event models for the clinical outcome process. In clarifying and demonstrating these key concepts and methodologies, we hope to set the stage for future work on mediation analysis with dynamic (e.g., PKPD) models. (1) Fleming, T.R. and Powers, J.H. Biomarkers and surrogate endpoints in clinical trials. Stat. Med. 31 (2012):2973–2984. (2) Joffe, M.M. and Greene, T. Related causal frameworks for surrogate outcomes. Biometrics 65 (2009):530–538. (3) Putter, H. and van Houwelingen, H.C. Understanding landmarking and its relation with time-dependent Cox regression. Stat. Biosci. 9 (2017):489–503. (4) Imai, K., Keele, L. and Tingley, D. A general approach to causal mediation analysis. Psychol. Methods 15 (2010):309–334.
