Defining Clinical Trial Estimands: A Practical Guide for Study Teams with Examples Based on a Psychiatric Disorder

Abstract

While the ICH E9(R1) Addendum on “Estimands and Sensitivity Analysis in Clinical Trials” was released in 2019, the widespread implementation of defining and reporting estimands across all clinical trials is still in progress and the engagement of non-statistical functions in this process is still to be fully experienced. Case studies are sought after, especially those with documented clinical and regulatory feedback. This paper describes an interdisciplinary process for implementing the estimand framework, devised by the Estimands and Missing Data Working Group (a group with clinical, statistical, and regulatory representation) of the International Society for CNS Clinical Trials and Methodology. This process is illustrated by specific examples using various types of hypothetical trials evaluating a treatment for major depressive disorder. Each of the estimand examples follows the same template and features all steps of the proposed process, including identifying the trial stakeholder(s), the decisions they need to make and the questions that would support their decision making. Each of the five strategies for addressing intercurrent events are addressed in at least one example; the featured endpoints are also diverse, including continuous, binary and time to event. Several examples are presented that include specifications for a potential trial design, key trial implementation elements needed to address the estimand, and main and sensitivity estimator specifications. Ultimately this paper highlights the need to incorporate multi-disciplinary collaborations into implementing the ICH E9(R1) framework.

Estimands in CNS trials – A review of strategies for addressing intercurrent events

“…For specific CNS disorders, some publications on the thinking process for defining an estimand [ 4 ], or on the methods of estimation aligned to certain strategies [ 5 , 6 ] exist, and some regulatory guidance that includes strategies to account for intercurrent events has been given [ 7 , 8 ]. However, the experience in constructing and reporting estimands is in general limited [ 9 ].…”