Machine Learning-Enabled Medical Devices Authorized by the US Food and Drug Administration in 2024: Regulatory...

Area of Science:

• Medical Device Regulation
• Machine Learning in Healthcare
• Regulatory Science

Background:

• Over 690 machine learning (ML)-enabled medical devices were authorized by the US Food and Drug Administration (FDA) between 1995 and 2023.
• New 2024 FDA guidance introduced Predetermined Change Control Plans (PCCPs), aiming to enhance transparency, equity, and safety within the Good Machine Learning Practice (GMLP) framework.

Purpose of the Study:

• To evaluate regulatory pathways, predicate lineage, demographic transparency, performance reporting, and Predetermined Change Control Plan (PCCP) utilization for FDA-authorized ML-enabled devices in 2024.

Main Methods:

• A cross-sectional analysis of all FDA-authorized ML-enabled devices in 2024 was performed.
• Data extracted from FDA summaries included regulatory pathway, predicate genealogy, performance metrics, demographic disclosures, PCCPs, and cybersecurity statements.
• Descriptive and nonparametric statistical methods were employed.

Main Results:

• The FDA authorized 168 ML-enabled Class II devices in 2024, predominantly via 510(k) clearance (94.6%), with radiology being the leading field (74.4%).
• Non-US sponsors accounted for 57.7% of clearances; predicate reuse was infrequent (9.9%), with a median predicate age of 2.2 years.
• Predetermined Change Control Plans (PCCPs) were present in 16.7% of summaries, and demographic data was reported by only 15.5% of devices.

Conclusions:

• 2024 saw record approvals and internationalization of ML-enabled medical devices, yet the uptake of PCCPs and transparent reporting of performance and demographic data was limited.
• Standardized disclosures and enhanced post-market surveillance are crucial policy objectives to fully realize the benefits of Good Machine Learning Practice (GMLP).