SW-4D - Why Treatment Strategies are a Sine Qua Non of the Causal Estimand

Background: Causal pharmacoepidemiology studies do not compare treatments, but treatment strategies. A treatment strategy is a set of rules to assign treatment at each time of follow-up. Without a sufficiently well-defined treatment strategy, some causal contrasts (e.g. the effect under “complete adherence”) will be incompletely specified and we will be missing a foundation for analytical decisions and interpretation of causal effect estimates

Objectives:
- To review the concept of consistency as one of the conditions for causal inference and define how the concept of treatment strategy is related to consistency

- To explore, via a case study, the taxonomy of treatment strategies and their interplay with causal contrasts

- To illustrate, via a case study, the relevance of defining treatment strategies in observational analyses that use existing clinical trials as benchmarks

- To delve into the role that defining treatment strategies play in the study of disparities

This symposium is intended for researchers involved in the design, analysis and interpretation of RWD studies.

Description:
1. Motivation (5’). Dr. Garcia de Albeniz (RTI Health Solutions, Spain)

2. Sufficiently well-defined causal questions and consistency (15’) (Sonja Swanson, University of Pittsburg, US).
This talk will elaborate the concepts of consistency and well-specified treatment strategies as one of the components of consistency in a non-technical manner. These concepts will be key to integrate the concepts of the next presentations.

3. Treatment strategies and causal contrast (15’) (Barbra Dickerman, Harvard University, US).
Well-defined treatment strategies are required for well-defined causal contrasts. Through a practical application, this talk dissects different types of treatment strategies and their comparisons under different causal contrasts. It also describes how misclassified treatment assignment can lead to immortal time and discusses a framework to avoid this bias.

4. A target trial with incomplete adherence (15’) (Xabier Garcia de Albeniz).
This talk will show an application of benchmarking an observational study against an existing trial with incomplete adherence. It will illustrate the relevance of treatment strategies to perform such analysis, as well as to estimate the effect under full adherence.

5. Treatment strategies in the study of disparities (15’) (John Jackson, Johns Hopkins University, US)
This talk will use a real-world data example to review ethical principles for choosing covariates to guide treatment strategies and how to extend the target trial framework to assess effects of treatments on disparities.

6. Discussion. Led by John Seeger (RTI Health Solutions, US), all speakers will contribute. (25’)