Real-world data (RWD) is increasingly used to complement clinical trial data to assess the effectiveness and safety of therapeutic products. However, the potential of confounding from selection- and time-related biases, as well as unobservable patient characteristics, can pose challenges. Casual inference and casual assumptions provide a framework for addressing this – leveraging machine learning applications for powerful tools to improve the rigor and robustness of estimates.
The panel will explore the foundations and advances in causal machine learning methods, focusing on deriving patient insights and improving treatment decision-making.
1. Dr. Patrice Verpillat will discuss the regulatory perspective on AI, including examples of experimentation at the regulatory level. He will focus on EMA’s AI workplan, and initiatives from regulators to provide guidance to industry, as well as to drive awareness and stakeholder engagement.
2. Dr Katja Hakkarainen will provide the epidemiological perspective on the causal inference methods, especially the adoption of the target trial emulation framework in observational research. She will discuss the principles of target trial emulation, motivations for using the framework, and examples of applications of the framework in observational research.
3. Dr Andy Wilson will provide an overview of the statistical science and implementation of casual inference and describe emerging machine learning-based methods, such as targeted maximum likelihood aimed at addressing confounding. He will also present a case study on estimating vaccine effectiveness from observational data to illustrate the applications of these methods. He’ll conclude with an outline of promising innovative dynamic methods (including the concept of the modifiable ‘treatment policy’).
4. Ipek Özer Stillman will present a manufacturer’s assessment, focusing on evidence-generation activities to shape early-phase decisions, for example, protocol optimization and patient selection. Ipek will explore the potential impact of machine learning in causal inference, and how methods could be applied.
Sponsor: Parexel
Conference/Value in Health Info
2024-11, ISPOR Europe 2024, Barcelona, Spain
Code
115
Topic
Methodological & Statistical Research