--- title: "Bringing Your US-Authorised AI Diagnostic to Europe: What Transfers and What Doesn't" description: "Moving an FDA-cleared AI diagnostic to Europe? Learn what evidence transfers, what IVDR requires and how a gap assessment can guide CE marking." url: https://qbdgroup.com/en/blog/fda-to-ivdr-ai-diagnostics-evidence-gap-assessment type: "Blog post" language: en published: 2026-10-14 author: "Pieter Smits & Pieter Bogaert" category: "Software Solutions & Services" publisher: "QbD Group" citation: "QbD Group, \"Bringing Your US-Authorised AI Diagnostic to Europe: What Transfers and What Doesn't\", https://qbdgroup.com/en/blog/fda-to-ivdr-ai-diagnostics-evidence-gap-assessment" --- # Bringing Your US-Authorised AI Diagnostic to Europe: What Transfers and What Doesn't > Moving an FDA-cleared AI diagnostic to Europe? Learn what evidence transfers, what IVDR requires and how a gap assessment can guide CE marking. You have obtained FDA 510(k) clearance or De Novo authorisation for your AI diagnostic. Now you want to bring the device to Europe and achieve CE marking under the In Vitro Diagnostic Regulation (IVDR). The good news is that much of the work already completed for the US market can provide a valuable starting point. The catch is knowing which evidence can be reused, which conclusions need to be reassessed and which EU-specific deliverables still need to be developed. The guiding principle is simple: reuse the data, not the regulatory conclusion.

In This Blog Post:

## What Can Transfer From an FDA Submission to IVDR? An existing FDA submission can contain substantial technical and clinical work. Much of that underlying evidence may be useful for the European submission, provided it is relevant to the device's EU intended purpose and can be adequately justified.

Potentially reusable materials include:

Reuse does not mean simply inserting the same documents into a new technical file. The manufacturer still needs to establish that the evidence supports the IVDR claims, intended users, population and conditions of use. ## Clinical Performance Studies May Need an Additional Gap Assessment A clinical performance study performed for the US market may remain relevant, but it should be reviewed against applicable IVDR expectations, including ISO 20916 where relevant. Any differences in study design, documentation or conduct need to be assessed rather than assumed away. Population applicability matters too. Manufacturers should justify whether the studied population is sufficiently representative of the population covered by the intended European use. A study conducted outside Europe is not automatically unusable, but its applicability must be supported. The question is therefore not only whether you have clinical performance data. It is whether those data answer the clinical questions posed by your IVDR intended purpose. ## What Does Not Transfer: FDA Regulatory Conclusions For devices cleared through the FDA 510(k) pathway, the substantial equivalence argument does not carry over as an IVDR conformity route. The IVDR does not offer an equivalent predicate-device pathway. Likewise, De Novo authorisation is not itself evidence of CE conformity. Under the IVDR, the manufacturer must demonstrate that the device meets the applicable requirements and supports its intended purpose, performance and safety with appropriate evidence. This often means revisiting the intended purpose statement and the risk management file. Even when the underlying product is unchanged, the regulatory framing and evidence mapping may need to be different. ## Which IVDR-Specific Deliverables Need Attention?

Several elements require particular attention when translating an FDA package into an IVDR technical documentation set:

The technical documentation should also be prepared in line with applicable IVDR requirements and the submission instructions of the relevant notified body. These activities are not necessarily a complete restart. A gap assessment should identify where existing evidence is sufficient, where it needs reframing and where new work is genuinely required. ## Why AI and Machine-Learning Diagnostics Face Additional Questions If the device uses machine learning, manufacturers should expect particular scrutiny of how performance was established, how the model behaves in clinical use and how changes will be controlled over time. ### Reference Standards and Independent Assessment Manual-versus-machine comparison studies need a defensible reference standard. Where human assessors are involved, their qualifications and impartiality should be considered and documented. Without a credible reference, apparently strong performance figures may be difficult to interpret. ### Predefined Acceptance Criteria Acceptance criteria should be established before evaluating performance, including when comparable historical data are limited. This helps ensure that the evidence is assessed against a justified standard rather than criteria selected after results are known. ### Human Oversight and Explainability Manufacturers need to consider how intended users understand, interpret and act on software outputs. Clinical evidence and usability engineering can help demonstrate that the intended workflow and any necessary human oversight are workable in practice. The appropriate evidence depends on the specific device and its claims. ### Data Governance and Bias The origin and suitability of datasets matter. Training and test data should be appropriately separated, and manufacturers should consider representativeness, provenance and potential bias. A model that performs well on a narrow dataset may not necessarily perform equally well in its intended population. ### Post-Market Monitoring and Model Changes A plan for monitoring performance after launch is especially important for AI-enabled diagnostics. Where retraining or model updates are contemplated, the criteria, thresholds and change-control approach should be defined and justified before deployment. These questions connect directly to the IVDR evidence strategy: they help establish whether the software can reliably achieve its intended purpose throughout its lifecycle. ## Start With a Structured FDA-to-IVDR Gap Assessment The most efficient starting point is to compare the existing FDA evidence package with the proposed IVDR intended purpose and applicable requirements.

A practical assessment can examine:

This gives teams a more useful plan than either assuming the FDA package will transfer unchanged or rebuilding everything from scratch. ## Your FDA Package Is a Head Start, Not a Finished IVDR File An FDA submission can give AI diagnostic manufacturers a substantial head start on European market entry. But the underlying data must be assessed against a different regulatory framework, and some IVDR-specific evidence and documentation will still be needed. The objective is to preserve the value of work already completed while identifying exactly what must change to support CE marking. A structured gap assessment provides that clarity. --- Source: https://qbdgroup.com/en/blog/fda-to-ivdr-ai-diagnostics-evidence-gap-assessment — © QbD Group. Quote freely with attribution and a link back.