Multistate Models as a Framework for Estimand Specification in Clinical Trials of Complex Processes

Abstract

Intensity-based multistate models provide a useful framework for characterizing disease processes, the introduction of interventions, loss to followup, and other complications arising in the conduct of randomized trials studying complex life history processes. Within this framework we discuss the issues involved in the specification of estimands and show the limiting values of common estimators of marginal process features based on cumulative incidence function regression models. When intercurrent events arise we stress the need to carefully define the target estimand and the importance of avoiding targets of inference that are not interpretable in the real world. This has implications for analyses, but also the design of clinical trials where protocols may help in the interpretation of estimands based on marginal features.

Surrogacy validation for time‐to‐event outcomes with illness‐death frailty models

“…Within the principal stratification framework, unobserved outcomes due to truncation by death can be addressed by defining strata based on survivorship cohorts; causal effects are defined for individuals who would survive regardless of the assigned treatment (Comment et al, 2019;Xu et al, 2020;Zhang & Rubin, 2003). Models do exist that are suitable for these data by accounting for semicompeting risks, though are there complications in defining proper estimands (see Bühler et al, 2022). Specifically, whereas the previous CEP validation quantities suggest conditioning on counterfactual surrogate outcomes, in our setting the surrogate outcome may not be well-defined if it is not observed before the primary outcome; it may not be possible to condition on strata defined by 𝑆(0) and 𝑆(1) occurring or not by time 𝜏.…”