Skip to content

Abstract Search

Primary Submission Category: Broken Randomized Studies

Estimand strategies for safety outcomes

Authors: Fabrizia Mealli, Alessandra Mattei, Veronica Ballerini,

Presenting Author: Veronica Ballerini*

Safety evaluation of new therapies is an essential aspect of clinical trials, with the primary focus of quantifying the incidence of adverse events (AEs) and comparing it to a standard treatment.
Several estimand strategies have been proposed for efficacy analysis of time-to-event outcomes in the presence of censoring and competing events, also in accordance to the ICH E9 Addendum. Safety analysis of adverse events, instead, is often rather simplistic: AE probabilities are estimated without explicitly defining the target causal comparison and neglecting assumptions on the censoring mechanisms leading to differential follow-up times (e.g., in oncology trials, patients may discontinue the control treatment earlier than the new treatment due to Progressive Disease).

Here, we explicitly define the assumptions under which estimators typically used in the literature, such as the Exposure-Adjusted Incidence Rate, Kaplan-Meier and Aalen-Johansen estimators, have a causal interpretation. We introduce new principal stratum and hypothetical estimand strategies for safety outcomes in the presence of censoring, competing events and varying follow-up times. We also propose identifying assumptions as well as estimators under these assumptions.
Our contribution will enhance interpretation of AE risks and has the potential of changing clinical trial practice with regard to safety analysis and risk-benefit assessment.