Using AI for IVDR Compliance: Where it Helps, Where it May Hurt, and Where Humans Still Matter

The transition to the European Union’s In Vitro Diagnostic Medical Devices Regulation (IVDR) marks a pivotal moment for the in vitro diagnostics (IVD) sector. For more than 20 years, in vitro diagnostics were covered not by a regulation, but by a less restrictive EU directive–the IVDD. With the implementation of the IVDR, manufacturers must comply with stricter and more standardized regulations to ensure that their IVD devices are safe, effective, and high-quality. 

The IVDR has ushered in a new era of regulatory compliance: many previously self-certified devices now require conformity assessments, and each EU member state is allowed to define its own language requirements. Some allow English to be used for documentation aimed at professionals, but others require translation (into one or more of the EU’s 24 official languages) for everyone. For more on this, see the European Commission’s overview of language requirements for manufacturers of medical devices. 

This increase in documentation has prompted many manufacturers to wonder if, or how, artificial intelligence (AI) tools might help streamline the IVDR application and compliance process. The truth? AI tools can help, they can hurt, and subject-matter experts are still key to a finished product that complies with the IVDR and ensures patient safety. Let’s take a closer look at AI’s role in IVDR compliance. 

AI as a potential response to IVDR-related complexity and costs 

As compared to the In Vitro Medical Diagnostic Devices Directive (IVDD), adopted in 1998 and in place since the year 2000, the IVDR has dramatically increased documentation volume, regulatory scrutiny, and ongoing reporting obligations. The result: many manufacturers are feeling the strain on their already limited internal resources. We’ve seen five primary reasons that manufacturers may think about turning to AI for IVDR compliance: 

1. Increases in documentation volume and frequency under the IVDR. Manufacturers are feeling squeezed by the IVDR’s requirement for more extensive technical files, ongoing post-market surveillance (PMS), periodic safety update reports (PSURs), and performance evaluations. Producing all of this in one language is a tall order, let alone the potential that user-facing documentation may be needed in up to 24 languages. AI can help manage this by automating multiple steps in the process. Particularly if a manufacturer uses AI agents, it may be possible to automate the organization of at least some of the data, automatically check for updates, and flag any elements that are missing or incomplete. This can reduce time-intensive “busy work,” without cutting corners on the more complex aspects of the project that humans handle best. 
2. “Notified body” bottlenecks. Notified bodies, defined in this European Commission document, are an important component of the IVDR, and manufacturers must use the services of an approved notified body for all but Class A devices. At present, there are 19 notified bodies recognized by the European Commission; obviously a small number, as compared to the huge volume of applications. This has resulted in a significant backlog; manufacturers can’t solve this issue on their own, but they can use AI-supported workflows to submit cleaner and more consistent documentation, reducing the risk of queries and requests for changes. 
3. The IVDR makes compliance into a continuous process, rather than a one-time checkbox. As compared to the IVDD, the IVDR requires ongoing monitoring and reporting after a device is initially approved. Handing the tracking aspect of this over to an AI tool can help manufacturers stay ahead of ongoing deadlines instead of scrambling as a target date approaches, or worse, after it’s already passed. 
4. Compliance costs are on the rise, especially for smaller manufacturers. The IVDR’s expanded requirements for clinical evidence, reporting, and multilingual documentation can rapidly get expensive. Identifying tasks that are a good fit for an AI tool can help control costs while still complying with the IVDR’s requirements and ensuring patient safety. 
5. As guidance evolves, AI can often shift more nimbly than humans can. Implementation of the IVDR has been marked by delays, deadline extensions, and changing requirements. If a manufacturer needs to re-analyze large data sets or comply with updated formatting or terminology requirements, an AI tool may work faster and more accurately than a human team. 

Where AI adds real value in IVDR workflows 

We’ve identified four IVDR-related tasks where AI can reduce the risk of human error, and smooth out the often-complicated regulatory approval process. 

The limits of AI in the IVDR’s regulatory and language contexts

As seen in the table above, AI has the potential to transform many aspects of the IVDR compliance workflow. At the same time, it’s important to take a realistic view of AI’s capabilities and limitations. Following are four frequently asked questions from IVD manufacturers who are trying to decide where to implement AI and where to stick with human experts. 

1. Why don’t European regulators automatically accept AI output? The output of AI systems must be validated to prove that it is accurate, consistent, and fit-for-purpose. Regulators expect that IVD manufacturers will demonstrate that any AI tools used in the IVDR compliance workflow are valid, reliable, and traceable. Obviously, this takes time and costs money, which must be included in any cost/benefit analysis of AI usage. 
2. Can AI fully understand the IVDR’s complex requirements?  AI excels at analyzing structured data, but it struggles with nuance, context, and interpretation. AI on its own is not the best option for any decisions (IVDR-related or other) that are complex and potentially ambiguous. Human expertise is still the key when it comes to interpreting these rules, making decisions related to them, and taking final accountability for the outcome. 
3. Are there data privacy concerns related to AI in IVDR workflows? Yes, definitely. Particularly when manufacturers must comply with the GDPR–the European Union’s General Data Protection Regulation–potentially in combination with cloud-based or third-party AI tools, data privacy is a significant issue. As with U.S. healthcare privacy laws such as HIPAA, data privacy requirements can often dictate the choice of vendors for any information technology services, including AI. 
4. Is AI a cost-effective option for smaller device manufacturers? Particularly for smaller companies, AI integration costs may pay off over time, but integration may be a significant up-front expense. It’s best to have a clear strategy and prioritize where AI can have the biggest impact, so that your company has realistic expectations about the return on investment. 

What works: A balanced approach, integrating automation and expertise 

When you think about AI-leveraged success in the IVDR landscape, think about a collaborative ecosystem. Human-assisted AI, continuous learning and adaptation, and proactive risk management all play a critical role. As with all things AI, the landscape is changing quickly and it’s important to keep up to date. If you’re researching best practices for AI usage in general, a good starting point is the European Medicines Agency (EMA) news release, EMA and FDA set common principles for AI in medicine development. This is not specific to the IVDR, but provides some overarching principles that can be helpful. Additionally, we recommend that IVD manufacturers: 

• Envision a collaborative ecosystem. Think of a cooperative process, with defined roles for AI and for human experts, working toward a common goal of regulatory compliance and patient safety. 
• Define roles for humans, AI, and human-assisted AI. AI is clearly the right tool for the job, when it comes to analyzing large volumes of data; what’s changed in the most recent iteration of the IVDR guidance, for example. This frees humans up to make strategic decisions, interpret nuances, and make complex decisions with a high cost of failure. Still other tasks may benefit from an AI “first pass” with a review by human experts. 
• Choose solutions that learn and adapt. AI is not an off-the-shelf solution; it needs to evolve along with IVDR-related regulations. Manufacturers must consider this when choosing AI tools; it’s important that they incorporate machine learning and can adapt as compliance goals shift. 
• Proactively manage risk. The best time to manage risk is before something goes wrong. AI excels at predictive analytics: anticipating bottlenecks and challenges based on existing data and patterns. By making the most of AI’s abilities, manufacturers can invest time and money where it matters most–upstream of compliance risks. 

To make sure that your IVDR compliance efforts succeed, we recommend four key strategies:

• Invest in professional language translation services to create multiple language versions of mandatory documents in a streamlined way. 
• Establish good relationships with notified bodies. This will help you create a realistic timeline and prioritize your highest-risk devices. 
• Have any AI-translated documents validated by human experts. This avoids errors (that may be propagated into other languages) that can derail your application. 
• Follow the IVDR-related news, so that you’re aware of how requirements and timelines may be changing. 

With these types of best practices in place, you’ll stack the IVDR compliance deck in your company’s favor, given the new environment that the IVDR represents.

Check out our IVDR approach to all your translation needs.

Or Contact Us to learn more.