The US Food and Drug Administration (FDA), Health Canada, and the UK's Drug and Healthcare Products Regulatory Agency (MHRA) have jointly identified 10 guiding principles that can inform the development of excellent machine learning practices (GMLPs). These guideline principles help promote safe, effective, and high-quality medical devices using artificial intelligence and machine learning (AI/ML).
Artificial intelligence and machine learning technologies can transform healthcare by deriving new and important insights from the vast amount of data generated during healthcare delivery every day. They use software algorithms to learn from their actual use, and in some situations, this information may be used to improve product performance. However, they also present their own considerations due to their complexity and the iterative, data-driven nature of development.
These 10 guideline principles aim to lay the foundation for developing excellent machine learning practices that address the unique nature of these products. They will also help foster future growth in this rapidly moving field.
Ten guideline principles identify areas where the International Medical Device Regulatory Authority Forum (IMDRF), International Standards Organizations, and other collaborative bodies may work to advance GMLP. The areas of collaboration include research, the creation of educational tools and resources, international harmony, and consensus standards. This could help inform regulatory policies and guidelines.
It is assumed that these guidelines principles may be used.
- Adopt good practices proven in other areas
- To ensure coordination practices from other sectors can be applied to the health technology and health care sectors.
- Create new practices specific to the health technology and healthcare sector
As the field of AI/ML medical devices evolves, GMLP best practices and consensus standards must also evolve. Strong partnerships with international public health partners are important to enable stakeholders to promote responsible innovation in this area. Therefore, we expect this initial collaboration to inform a wider range of international engagements, including the IMDRF.
Continuous feedback via Public Docket (FDA-2019-N-1185) at Regulations.gov is welcome. We look forward to being involved in these efforts. The Digital Health Center of Excellence is at the forefront of this work for the FDA. Please contact us directly at digitalhealth@fda.hhs.gov, software@mhra.gov.uk, and mddpolicy-politiquesdim@hc-sc.gc.ca.
Guide principles
- Interdisciplinary expertise is utilized throughout the entire product lifecycle
- Excellent software engineering and security practices are implemented
- Clinical study participants and datasets represent the intended patient population
- The training dataset is not dependent on the test set
- The selected reference dataset is based on the best available methods
- The model design is tailored to the available data and reflects the intended use of the device
- Focused on human team performance
- Tests show device performance in clinically relevant conditions
- Users will be provided with clear and essential information
- The deployed models are monitored for performance and risk retraining is managed
