Summary:
Le 19 juin 2025, le Groupe de coordination des dispositifs médicaux (MDCG) a publié des directives sur l’interaction entre le Règlement sur les dispositifs médicaux (MDR), le Règlement sur les dispositifs médicaux de diagnostic in vitro (IVDR) et la loi sur l’intelligence artificielle (IA). Ces directives clarifient l’application jointe de ces règlements pour les systèmes d’IA utilisés à des fins médicales, en présentant les définitions, la portée d’application, la gestion du cycle de vie et la conformité des documents techniques.
Original Link:
Original Article:
On 19 June 2025, the Medical Device Coordination Group (MDCG) published new guidance on the interplay between the Medical Device Regulation (MDR), In Vitro Diagnostic Medical Device Regulation (IVDR), and the Artificial Intelligence (AI) Act (MDCG 2025-6). The document provides answers to frequently asked questions regarding the joint application of the AI Act and the MDR/IVDR for manufacturers of AI systems used for medical purposes (MDAI).
### Key Points in the Guidance:
#### Definitions and Roles
– **Manufacturer vs Provider and User vs Deployer:** References to “manufacturer” in MDR/IVDR align with “provider” under the AI Act. The term “deployer” in the AI Act is distinct from “user” (e.g., healthcare professionals or lay persons) under MDR/IVDR.
– **Annex XVI Products:** The guidance confirms that the AI Act may apply to Annex XVI products and accessories.
– **High-Risk Classification:** Classification of an AI system as a high-risk AI system under the AI Act does not imply higher risk classification of the medical device or IVD under MDR/IVDR.
#### Scope of Application
– A Medical Device AI system (MDAI) is classified as high-risk under Article 6(1) of the AI Act if it meets these two conditions:
1. It is a safety component, or the AI system itself is a medical device.
2. It is subject to third-party conformity assessment by a notified body under MDR/IVDR.
The MDCG provided a table outlining which devices fall under Article 6(1).
| **Classification** | **Notified Body Involved?** | **AI Act High-Risk (Art. 6(1)) Conditions Fulfilled?** |
|———————————–|—————————–|——————————————————-|
| MDR Class I (non-sterile, non-measuring, non-reusable surgical) | No | No |
| MDR Class I (sterile, measuring, reusable, surgical) | Yes | Yes |
| MDR Class IIa, IIb, III | Yes | Yes |
| MDR Annex XVI | Yes | Yes |
| IVDR Class A (non-sterile) | No | No |
| IVDR Class A (sterile) | Yes | Yes |
| IVDR Class B, C, D | Yes | Yes |
| In-house device according to Art. 5(5) MDR* | No | No |
*In-house devices are not considered high-risk AI systems unless assessed by a notified body.
#### Lifecycle Management
– Both MDR/IVDR and the AI Act necessitate continuous lifecycle and risk management.
– The AI Act enforces additional stipulations for post-market monitoring, including human oversight and data traceability capabilities.
– Manufacturers are encouraged to integrate the AI Act’s quality management system (QMS) requirements into their existing MDR/IVDR processes.
– **Data Governance:**
– High-quality training, validation, and testing datasets must account for representativeness and bias avoidance.
– Manufacturers should ensure transparency, integrity, and traceability in their data practices, using certified third parties where applicable.
#### Technical Documentation
– The MDCG recommends producing a unified set of technical documentation for MDAIs, adhering to both MDR/IVDR and AI Act requirements.
– Sampling rules laid out in prior MDCG guidelines (e.g., MDCG 2019-13) remain applicable.
#### Usability and Human Oversight
– **Human-Centric Design:** AI systems in healthcare require user-centric design principles. Oversight measures should consider risk levels and include provisions for intervention when necessary.
– **Transparency:** Documentation of usability engineering and safety risk assessments is required.
#### Specific Guidance Areas:
1. **Cybersecurity:** Cybersecurity measures should address both standard and AI-specific threats, such as adversarial attacks and model poisoning.
2. **Clinical Evaluation:** Clinical testing must meet MDR/IVDR and AI Act criteria, such as accuracy, robustness, and cybersecurity.
3. **Conformity Assessment Processes:** High-risk devices must comply with conformity assessments required under both frameworks.
4. **Modifications and Updates:** Significant post-market updates must be aligned with transitional periods, particularly for changes introduced before and after 2 August 2027.
5. **Post-Market Monitoring:** Manufacturers must implement systems to detect risks throughout the device lifecycle, including interactions with other systems.
This guidance underscores the need for a harmonized approach between the MDR/IVDR and the AI Act, particularly for high-risk MDAIs, ensuring safety, performance, and regulatory compliance across the lifecycle.