Legacy device manufacturers have had a lot on their plate under the MDR.
Legacy devices have had to go through considerable re-evaluation and in many cases rearrangements to meet the stringent MDR requirements.
As a result, quite a few legacy device makers have faced financial and operational challenges. Some have been forced to take their devices off the market due to the high compliance costs.
AI developers, especially those designing AI systems for medical purposes, can learn valuable lessons from legacy device manufacturers. Not only about working in a heavily regulated environment. But also about how to develop products and processes that can adapt to the ever-evolving changes.
So let’s check out three takeaways from legacy devices to AI developers.
What are legacy devices?
Legacy devices refer to medical devices, which had a valid CE mark under the old legislation (MDD) and that are kept on the Union market* after the MDR date of application in 2021, and for which the manufacturer pursues a new CE marking under the MDR.
The manufacturer of these devices has to meet certain conditions in exchange for a transition period. The transition time allows the devices to be supplied to the market until 2027 or 2028 with the old MDD CE mark, while their new MDR CE mark certification is processed by a notified body.
* Union market is explained here. It’s in Finnish. With browser translation, you can get a fairly decent English translation.
Why legacy devices have had a hard time under the MDR?
1.) Stricter clinical rules
Many legacy devices were launched to the market in the pre-MDR era with modest clinical data.
The MDR didn’t allow these devices to be grandfathered in, even if they hadn’t had any safety issues.
Legacy devices have had to meet tougher clinical evidence requirements under the MDR to secure the CE mark on the devices.
Even manufacturers of many standard care devices have had to step up their clinical evidence.
-> Lesson for AI developers:
Plan clinical evidence generation early
Developing Medical Device Artificial Intelligence (MDAI) involves several stages.
Start the generation of clinical evidence while building the product. Generating strong clinical evidence takes time.
Especially regarding high-risk AI systems, notified bodies and authorities want to see sufficiently representative and robust data.
And speaking of data, don’t forget the dual compliance. The MDAI needs to meet both the MDR clinical data requirements and AI Act-specific data requirements.
So bring together clinical, technical, regulatory and other relevant experts early to help you design and conduct audit-proof MDAI clinical data creation.
2.) Notified body bottlenecks
All legacy devices need to go through CE mark re-certification by notified bodies for the MDR.
Fewer notified bodies have been designated under the MDR than under the MDD. As a result, notified bodies’ services are in high demand.
In addition, many legacy device manufacturers have put off the re-certification until the last minute.
These have created long queues for notified bodies.
Surveys of notified bodies have indicated that it still often takes over a year to complete the MDR QMS or MDR QMS & product certification process.
Legacy devices with incomplete documentation are part of the problem. Notified bodies are experiencing a heavy workload reviewing unclear and insufficient documentation. This leads to delays and even rejections of legacy device re-certification applications.
-> Lesson for AI developers:
Integrate compliance into MDAI life cycle
Once the AI Act is in full force, providers of high-risk AI systems might experience bottlenecks too in the conformity assessments conducted by notified bodies.
To prevent delays with your assessment, build compliance into your AI system from the beginning.
For example:
- Have comprehensive information on the development of the AI system
- Keep clear logs and records including model versions and training data sets
- Ensure traceability and explainability of the AI system
Also, get in touch with your notified body early.
This will help you prepare for the conformity assessment, and to avoid the rush hour when everyone else starts panicking.
3.) Assuming legacy = safe and compliant
Many self-certified legacy devices pulled off years on the market with pre-product launch testing, random post-market monitoring and minimal documentation.
These manufacturers assumed market presence, “working just fine”, was enough to prove their devices’ compliance and safety.
They thought the rules would stay like this forever.
But the MDR changed all that.
The old requirements got tougher and new requirements were introduced too.
The legacy device makers needed to prove their marketed devices align with the latest standards. They also had to start planning how to conduct continuous surveillance on their established devices.
Many self-certified legacy devices were pushed to a higher risk class under the MDR. That translated into heavier burden of proof. The manufacturers had to let a notified body be the judge of whether there’s enough evidence of the devices’ compliance and safety.
-> Lesson for AI developers:
Expect rules to evolve
The regulatory complexity is growing and the requirements are evolving.
To avoid brutal wake-up calls, make sure to keep up to date with regulatory developments.
If you have AI systems on the market already, don’t assume they’ll be exempt from new regulations and standards.
If you’re developing new AI systems, design systems that can evolve too. Adaptable, explainable and traceable from the start.
Set up, implement and maintain a robust post-market monitoring system. And document it properly, as the AI Act requires. This is essential for meeting future audit trail needs.
Legacy devices have faced major challenges under the MDR.
The lessons learned can help AI developers do things differently, and to achieve a strong position for funding, partnerships and market access.
If you have any questions about legacy devices, developing AI systems for medical purposes, or the regulations (the MDR and the AI Act), feel free to get in touch.
I’m always happy to listen to what’s on your mind and help you take the next step.
