Targeted Learning for Causal Inference Using Real-World Data

PURPOSE: This workshop will aim to discuss the application of novel methods of estimating causal inference using real world data.

DESCRIPTION:

Real-world evidence (RWE) is valued for its reflection of actual clinical use and diverse patient populations. However, challenges arise due to confounding factors, intercurrent events, and informative drop-outs. To address these challenges doubly robust causal inference methods, such as the targeted maximum likelihood estimator (TMLE), are proposed. These methods follow a causal inference roadmap (e.g., Eyler et al. , 2023) involving careful formulation of the causal estimand, aligned with FDA guidance, and the statistical estimand that identifies the causal estimand under specified causal assumptions. TMLE incorporates state-of-the-art machine learning through the super-learning algorithm, enhancing performance. For example, in the context of dealing with treatment by indication, these double robust methods incorporate models for the propensity score and outcome regression, and require only one of the two models to be correctly specified. As the use of RWE for decision making is gaining traction with regulators, it is important to demonstrate that these advanced causal inference methods can generate robust and more reliable real-world evidence. This is particularly important for complex diseases with multiple potential confounding factors.

In this workshop we will:

1. Introduce the problem and limitation of current approaches
2. A description of the roadmap for targeted learning and causal inference
3. Present a case study applying advanced methods to real-world COP cohort.
4. Discuss the acceptance of these methods from a regulatory perspective.

Real-time polling will gauge participants' experience with causal inference methods, potential applications of advanced methods, and likely acceptance by decision-makers. This interactive engagement aims to identify areas for further advancement of these methods and enhance stakeholders' understanding. Overall, the workshop seeks to contribute to the robustness and reliability of real-world evidence in decision-making processes.