When constructing estimands a key question that arises is how to handle intercurrent events and missing data. In a recent presentation in Cytel’s Advancing Innovation in Clinical Trial Design webinar, Frank Bretz and Jiawei Wei of Novartis introduce the estimand framework according to ICH E9(R1). They discuss in detail intercurrent events and common strategies for addressing intercurrent events when defining the clinical question of interest, with application to the COVID-19 pandemic.

What is an estimand?

An estimand is a precise description of the treatment effect reflecting the clinical question posed by the trial objective. It summarizes at a population-level what the outcomes would be in the same patients when compared under different treatment conditions.

According to ICH E9(R1), a properly defined estimand has the following five attributes:

1. Population of patients targeted by the clinical question
2. Variable (or endpoint) to be obtained for each patient that is required to address the clinical question
3. Treatment condition of interest and, as appropriate, the alternative treatment condition to which comparison will be made
4. Remaining intercurrent events occurring after treatment initiation that affect either the interpretation or the existence of the measurements associated with the clinical question of interest

Population-level summary for the variable should be specified, providing a basis for comparison between treatment conditions

What is an intercurrent event?

Intercurrent events are events occurring after treatment initiation that affect either the interpretation or the existence of the measurements associated with the clinical question of interest. It is necessary to address intercurrent events when describing the clinical question of interest in order to precisely define the treatment effect that is to be estimated.

Bretz & Wei provide the example of a trial for Dapagliflozin for glycemic control for adults with Type II Diabetes Mellitus. During the course of the trial, those adults whose glycemic levels were not adequately controlled by the intervention were provided with a rescue therapy. This led to the following conundrum: the sponsor tried to establish the treatment effect of the initially randomized treatments, where no patient had received rescue medication. The FDA compared the treatment policies ‘Dapagliflozin plus rescue’ versus ‘control plus rescue’. By asking two different questions, it had led to two different treatment effect estimates.

What to do about intercurrent events?

Bretz and Wei present five strategies to handle intercurrent events, when constructing appropriate estimands. They take the additional medication as an example.

• Treatment policy strategy: Compare treatment effects regardless of the intercurrent event (e.g., compare ‘Drug, plus additional medication as needed’ versus ‘Placebo, plus additional medication as needed’)
• Composite strategy: Include occurrence of the intercurrent event in the endpoint definition (e.g., set up separate endpoints that include treatment + rescue and treatment without rescue, along with place and placebo + rescue)
• Hypothetical strategy: Exclude intercurrent event in the endpoint definition (e.g. Use quantitative techniques to determine ‘Effect of Drug vs Placebo if additional medication had not been taken’)
• While on Treatment Strategy: Measure treatment effect before the occurrence of the intercurrent event (e.g., ‘Effect of Drug vs Placebo before additional medication is taken’)
• Principal stratum strategy: Stratify population without the intercurrent event (e.g., ‘Effect of Drug vs Placebo on patients who do not take additional medication’ used as a different subpopulation)

Share Via