Ready or Not, Here Come Estimands! The Emerging Attention to Causal Estimands in Real-World Evidence Studies

Background: An estimand is a quantity that we would like to estimate using data. The so-called "estimand framework" was proposed in 2019 by the International Council of Harmonisation (ICH) to describe causal estimands other than the intention-to-treat effect in randomized trials. The goal is to better align a trial’s objective, design, analysis, and interpretation. This framework is endorsed by the European Medicines Agency and by the US Food and Drug Administration.

Target trial emulation was developed to improve causal inference in observational analyses of real-world data (RWD). Because this framework starts by specifying the causal estimands in the (randomized) target trial, it has played a crucial role in reconciling results of randomized trials and observational studies.

The causal estimands considered in target trials are different than those described by the ICH framework. Therefore, investigators who work with both randomized trials and observational studies find themselves having to master both frameworks. This situation is exacerbated by the emergent use of external comparators in randomized trials.

This symposium is aimed to researchers who work with RWD, particularly in studies used for drug development and regulatory purposes such as external control arms.

Objectives:
- To review the importance of well-defined causal estimands
- To review causal estimands used for target trials in RWE studies
- To review causal estimands proposed by the ICH framework
- To discuss the overlap of both approaches and their relative advantages and disadvantages of each approach for regulators and HTAs

Description:
1. Motivation, introductions. Xabier Garcia de Albeniz (RTI Health Solutions, Spain). (5')
2. The regulator point of view. Dr. Daniel Morales (Data Analytics Task Force of the European Medicines Agency) has preliminary agreed to present. (15')
3. Causal estimands for randomized trials in the ICH framework. Jason Roy (Rutgers University), US (15')
4. Causal estimands for randomized trials when emulating a target trial. Xabier Garcia de Albeniz (RTI Health Solutions, Spain) (15').
5. Commonalities and differences between both frameworks and applications. Prof. Miguel Hernán (Harvard University, Boston, US). (15')
6. Discussion. Led by Xabier Garcia de Albeniz, Costel Chirila (RTI Health Solutions, US) and Paul Kluetz (FDA, US). (25')