Synthetic Data for Drug Development: Epistemic and Ethical Challenges
Friday, October 13, 2023
3:15 PM – 4:30 PM ET
Location: Dover C (Third Floor)
Synthetic data generated by artificial intelligence (AI) are an alternative to real-world data that are being proposed as a means for improving drug development. Synthetic data can be used to replace some or all real-world data collection to facilitate the completion of clinical trials, increasing the efficiency of clinical translation (and therefore potential to realize benefits to the public), reducing the costs of drug development and new medicines, and minimizing potential harms to research participants. In particular, synthetic data are proffered as a solution to the problem that over 80% of clinical trials in the US fail to meet patient enrollment criteria, and thus cannot yield benefit. Synthetic data are also seen as a potential solution to research disparities caused by lack of representation of marginalized and underserved populations, including those with rare diseases. To date, the Food and Drug Administration has already approved several drugs based on the use of synthetic control arms and it is envisioned that over half of all drug approvals will be based on synthetic data in the not-too-distant future. However, it is not yet clear whether the promise of synthetic data to facilitate trial completion or to achieve justice goals has been realized, nor what epistemic and ethical challenges are raised by its use. The current paper will identify ethical and epistemic issues that are unique to the use of synthetic data in the context of drug development, including the conduct of clinical trials and the development of AI technologies for health care.
Nicole Martinez-Martin – Departments of Pediatrics & Psychiatry – Stanford University; Mildred Cho – Departments of Pediatrics & Medicine – Stanford University