What Are the Clinical Trial Requirements for AI Devices?

By Rasit Dinc

Artificial intelligence (AI) is rapidly transforming the healthcare landscape, offering innovative solutions for diagnosis, treatment, and patient management. As AI-powered medical devices become more prevalent, ensuring their safety and effectiveness is paramount. This has led to the development of specific regulatory frameworks and clinical trial requirements for these devices. This article provides a high-level overview of the clinical trial requirements for AI devices in the United States and the European Union.

The U.S. Food and Drug Administration (FDA) Approach

The FDA has adopted a risk-based approach to the regulation of AI-enabled medical devices, which are typically classified as Software as a Medical Device (SaMD). The regulatory pathway for these devices depends on their level of risk to patients. The main premarket submission pathways include:

• Premarket Notification (510(k)): For devices that are substantially equivalent to a legally marketed device.
• De Novo Classification Request: For novel, low-to-moderate-risk devices that do not have a valid predicate.
• Premarket Approval (PMA): For high-risk devices that require a more stringent review process.

Recognizing the unique challenges posed by the adaptive nature of AI and machine learning (ML) algorithms, the FDA has issued the "Artificial Intelligence and Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) Action Plan" [1]. This plan outlines the FDA's commitment to developing a regulatory framework that is tailored to the specific characteristics of AI/ML-based SaMD.

A key component of this framework is the concept of a Predetermined Change Control Plan (PCCP). A PCCP is a plan that a manufacturer submits to the FDA that describes anticipated modifications to an AI/ML-based SaMD and the methodology for implementing and validating those changes. If the FDA agrees to the PCCP, the manufacturer can make the specified changes without submitting a new 510(k) for each modification. This allows for more efficient iteration and improvement of AI algorithms while still ensuring patient safety.

The FDA also emphasizes the importance of Good Machine Learning Practice (GMLP), transparency, and lifecycle management for AI-enabled devices. GMLP principles promote the development of high-quality AI/ML models and include recommendations for data management, model training, and performance evaluation [3]. Manufacturers are expected to provide clear information about the device's intended use, its performance, and the data used to train and validate the algorithm. They are also expected to have a robust plan for monitoring the device's performance in the real world and for managing any risks that may arise.

The European Union (EU) Framework

In the EU, medical device software (MDSW) is regulated under the Medical Device Regulation (MDR) (EU 2017/745) and the In Vitro Diagnostic Medical Device Regulation (IVDR) (EU 2017/746). The Medical Device Coordination Group (MDCG) has published a guidance document, MDCG 2020-1, which provides a framework for the clinical evaluation of MDSW [2].

The guidance outlines three key components for generating the necessary clinical evidence:

1. Valid Clinical Association / Scientific Validity: This involves demonstrating that there is a sound scientific basis for the MDSW's output and its association with the targeted clinical condition. This can be established through a review of the scientific literature, professional guidelines, or by conducting new clinical studies.
2. Technical Performance / Analytical Performance: This demonstrates the MDSW's ability to accurately, reliably, and precisely generate the intended output from the input data. This is typically demonstrated through verification and validation activities.
3. Clinical Performance: This demonstrates that the MDSW yields a clinically relevant output that has a positive impact on the health of an individual or on patient management. This can be demonstrated through clinical investigations, usability studies, or by analyzing real-world data.

The EU framework also emphasizes a lifecycle approach to the clinical evaluation of MDSW. Manufacturers are expected to continuously monitor the performance of their devices and to update their clinical evaluation as new data becomes available.

Challenges and Future Directions

The development of a robust regulatory framework for AI-enabled medical devices is an ongoing process. Some of the key challenges include the 'black box' nature of some AI algorithms, the potential for bias in the data used to train these algorithms, and the need to ensure the security and privacy of patient data.

In the future, we can expect to see further convergence of regulatory requirements across different jurisdictions. There will also be a greater emphasis on the use of real-world evidence to support the clinical validation of AI devices. As AI technology continues to evolve, regulatory frameworks will need to adapt to ensure that they remain fit for purpose.

Conclusion

The regulatory landscapes for AI-enabled medical devices in the U.S. and EU are continuously evolving. Both the FDA and the European Commission are working to develop frameworks that can accommodate the unique characteristics of these devices while ensuring patient safety and promoting innovation. As AI continues to play an increasingly important role in healthcare, it is essential for developers, clinicians, and regulators to work together to ensure that these powerful tools are used in a safe and effective manner.

References

[1] U.S. Food and Drug Administration. (2021). Artificial Intelligence and Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) Action Plan. https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-and-machine-learning-software-medical-device-action-plan

[2] Medical Device Coordination Group. (2020). MDCG 2020-1: Guidance on Clinical Evaluation (MDR) / Performance Evaluation (IVDR) of Medical Device Software. https://health.ec.europa.eu/system/files/2020-09/md_mdcg_2020_1_guidance_clinic_eva_md_software_en_0.pdf

[3] You, J. G., et al. (2025). Clinical trials informed framework for real world evaluation of artificial intelligence in healthcare. Journal of the American Medical Informatics Association, ocae096. https://pmc.ncbi.nlm.nih.gov/articles/PMC11832725/