We know very well what a significant change of (medical) devices is in the scope of the MDR since the publication of MDCG 2020-3 and the last revision of this document. Not every defined term will receive such detailed support in the form of an interpretation document. The reason for such a highly formalized interpretation is that the absence or non-implementation of a significant change has become a condition for the application of transitional provisions between MDD and MDR and thus a condition for the device to remain in the so-called legacy devices regime.
Deciding on the significance or insignificance of a change has thus become a relatively large science. The basic questions are: "Is this a change in the intended purpose?" and "Is this a change of the design?" In the case of the second question, it is still being investigated whether it is a software change, a change of substance or material, or a change of terminal sterilization or packaging design with impact on sterilization? If the answer to any of the last three questions is "yes", the change needs to be explored in more depth. In most cases, it will be a significant change, which mainly concerns changes in the area of material composition (for substance-based products) and changes in the area of final sterilization. On the other hand, changes such as the trade name of the manufacturer or change in the seat (address) of the manufacturer do not constitute a significant change. A simple rule can usually be used as a guide: If a change creates a new risk or, for existing risks, leads to an increase in the severity or probability of this risk, it is a significant change. And vice versa.
Significant change and software certification
Software can be contained in a device, such as Software in Medical Device (SiMD), or it can be a device in itself as Software as Medical Device (SaMD). So how to look at the significance of the change when assessing compliance and certifying the device, whether SaMD or SiMD? Changes to an already certified device always require approval by the notified body that issued the EU certificate (of the assessment of the technical documentation) if these changes could affect the safety and effectiveness of the device or the conditions given by manufacturer for the use of the device. If the manufacturer intends to make such changes, he must inform the notified body that issued the relevant certificate about this fact in advance. The notified body shall assess the intended changes and decide whether said changes require a new conformity assessment under Article 52 or whether they could be resolved differently.
AI Act
Many medical software manufacturers, some of them as SaMD or used to support the function or control of a physical medical device (SiMD), are today using artificial intelligence (AI) functions or planning to implement AI for the next generation of their device. So let's look at the area of change in design and development through the lens of AI. On June 13, 2024, Regulation (EU) 2024/1689 of the European Parliament and of the Council, commonly known as the AI Act (AIA), entered into force. The regulation applies to providers, deployers, importers, distributors, manufacturers of AI products, authorized representatives and affected persons located in the EU in relation to AI. The regulation introduces completely new terminology, including "manufacturers" and other entities. In order not to completely confuse the readers of this article, we will continue to use the MDR terminology in relation to the AIA. This is admittedly inaccurate, but hopefully the AIA enthusiasts will forgive it.
Definition of substantial modification
The AIA provides in Article 3 under point 23 the definition of the so-called "substantial modification", which it defines as a change to the AI system after it has been placed on the market or put into service, which the manufacturer did not foresee or plan during the initial assessment of conformity, and as a result of which affect the AI system compliance with the requirements set out in Chapter III Section 2 of the AIA or that results in a change to the intended purpose for which the AI system was assessed. Regarding this term, the recitals of the AIA state: “In line with the commonly established notion of substantial modification for products regulated by Union harmonisation legislation, it is appropriate that whenever a change occurs which may affect the compliance of a high-risk AI system with this Regulation (e.g. change of operating system or software architecture), or when the intended purpose of the system changes, that AI system should be considered to be a new AI system which shouldundergo a new conformity assessment”. So far, we are "on the same page" with the MDR, and we might think that we can equate the concepts of "significant change" and "substantial modification". Nevertheless, it is further stated in the recitals: “However, changes occurring to the algorithm and the performance of AI systems which continue to ‘learn’ after being placed on the market or put into service, namely automatically adapting how functions are carried out, should not constitute a substantial modification, provided that those changes have been pre-determined by the provider and assessed at the moment of the conformity assessment”.
So what's the point?
AIA introduces a decidedly more complicated conformity assessment procedure in relation to (medical) devices, as it will be necessary to take into account the requirements of two EU regulations for devices with AI. However, it can also provide some support for the violation of the hitherto binding principle that AI devices must be locked after training, and locked can be certified and placed on the market and put into service. And that their further learning will always require a new certification. In this context, point 17.1 of the GSPR should be recalled: “Devices that incorporate electronic programmable systems, including software, or software that are devices in themselves, shall be designed to ensure repeatability, reliability and performance in line with their intended use. In the event of a single fault condition, appropriate means shall be adopted to eliminate or reduce as far as possible consequent risks or impairment of performance”. We will monitor further developments for you. Perhaps AI will be able to provide repeatability, reliability and performance in ways other than state locking.
Author: Aleš Martinovský
Comments