TBD: A review of Artificial Intelligence and Machine Learning (AI/ML) - Enabled Medical Devices approvals in the US
Poster Presenter
Antoine Manson
Regulatory Policy and Innovation Specialist
Bayer AG France
Objectives
To analyze the AI/ML enabled medical devices regulatory approvals landscape, based on regulatory precedents, to help inform regulatory strategies and showcase important aspects that supported the regulatory decisions.
Method
For all AI/ML-based medical devices listed on the FDA database, the following characteristics were collected in an external file, where the analysis took place: device name, review panel, regulatory pathway, time for approval, MD classification, etc. Time range goes from 03/11/1997 to 06/17/2021
Results
AI-ML enabled medical devices are being developed and cleared at an exponential rate / Existing regulatory pathways (510(k), De Novo, PMA) are being utilized for AI/ML enabled medical devices. / Most of these devices are cleared under the 510(k) pathway (95%) with a median time for review of 168.8 days. / The number of cleared / approved AI/ML-enabled medical devices has increased substantially since 2015. / Only one of the analyzed AI/ML-enabled medical devices was classified as high-risk device (class III), whereas all the others were class II.
Conclusion
AI-ML enabled medical devices are being developed and cleared at an exponential rate, even if FDA’s traditional paradigm of medical device regulation was not designed for adaptive AI-ML technologies. FDA has highlighted its intention to develop an update to the existing regulatory framework to embrace the iterative improvement power of artificial intelligence that would allow for modifications to be made from real-world learning and adaptation. The regulatory frame in Europe is more heterogenous and doesn’t have the same public disclosure available now. A publicly accessible database for CE-marked medical devices in Europe would help assessing the regulatory landscape. European Commission targets spring of 2024 for fully functional Eudamed database.
