Before your first patient sees your device in a clinical investigation, MDR Annex XV Chapter I requires a complete pre-clinical evidence stack — biocompatibility, mechanical, electrical, software, and where applicable animal testing. For a typical Class IIb or III device, that is realistically 600 to 3,000 individual test results across 40 to 120 protocols. "We'll test it on patients" is not a regulatory strategy. It is a rejected Annex XV submission.

By Tibor Zechmeister and Felix Lenhard.

TL;DR

  • MDR Article 62 and Annex XV Chapter I require substantial pre-clinical evidence before a clinical investigation can begin.
  • The evidence stack typically includes biocompatibility, bench testing, electrical safety and EMC, software verification and validation, usability, sterility, stability, and where applicable animal testing.
  • For a Class IIb or III device, founders routinely underestimate the pre-clinical effort by an order of magnitude — the real test count is 600 to 3,000 results.
  • Risk management per EN ISO 14971:2019+A11:2021 drives which tests are required; you cannot skip a test just because it is expensive.
  • Competent authorities review Annex XV submissions and will reject applications where pre-clinical evidence is insufficient to justify first-in-human exposure.
  • Planning pre-clinical testing in parallel with clinical investigation design is the only realistic path for startups.

Why this matters

We once reviewed a pitch deck that contained this exact sentence: "We will validate the device through a 30-patient clinical study starting in Q3." The timeline showed the study starting ten months from prototype freeze. No mention of biocompatibility. No mention of bench testing. No mention of electrical safety. No mention of software verification. Just: prototype, then patients.

This is not malice. It is a genuine gap in how founders — especially those coming from software or consumer hardware — picture the regulatory path. In those worlds, you build, you release to users, you iterate. In MedTech under MDR, you build, you test exhaustively against the state of the art in non-clinical conditions, you document that the device is safe enough to put in front of a patient, and only then do you go to a clinical investigation. The competent authority reviewing your Annex XV submission is not impressed by your burn rate. They are looking for one thing: is the pre-clinical evidence sufficient to justify exposing humans to this device?

This post is a map of what "sufficient pre-clinical evidence" actually looks like. It is not light.

What MDR actually says

MDR Article 62 governs clinical investigations. It sets out the requirements for conducting a clinical investigation, the application procedure, the role of the sponsor, and the role of the ethics committee and competent authority. But the substantive content of what must be submitted is in Annex XV.

Annex XV is structured in three chapters. Chapter I is the general requirements for documentation of a clinical investigation. Chapter II is the clinical investigation plan. Chapter III is other documentation including the Investigator's Brochure.

Annex XV Chapter I requires, among other documentation, the following pre-clinical evidence (close paraphrase):

  • A summary of the pre-clinical evaluation based on relevant pre-clinical testing and experimental data, in particular design calculations, in vitro tests, ex vivo tests, animal tests, tests on simulated use or populations, or other tests.
  • The results of verification and validation testing of the design, including pre-clinical in vivo and animal testing where applicable.
  • The results of biological safety testing.
  • The results of electrical safety testing and electromagnetic compatibility testing where applicable.
  • Software verification and validation testing where applicable.
  • Information on the risk management process, including the results of the risk analysis per EN ISO 14971:2019+A11:2021.
  • A justification for the appropriateness of the pre-clinical data to support the proposed clinical investigation design and to justify the benefit-risk ratio.

The phrase that matters most is the last one: justification for the appropriateness of the pre-clinical data. The competent authority does not just want a list of tests you ran. They want evidence that the tests you ran adequately characterise the device's safety profile in every relevant dimension, that the residual risks are acceptable, and that the expected clinical benefit justifies exposing patients to those residual risks.

This connects directly to MDR Annex I (the GSPRs) and to EN ISO 14971:2019+A11:2021, which is the referenced standard for risk management. Every identified risk in the risk management file must have a corresponding control measure, and the effectiveness of that control measure must be verified. For most control measures, verification means testing. Testing means protocols, samples, test reports, and statistical justification for sample sizes.

A worked example

A Class IIb orthopaedic implant — let us say a titanium spinal cage. Here is a realistic pre-clinical evidence stack:

Biocompatibility per EN ISO 10993-1:2025 (biological evaluation plan driven by contact type and duration — implanted, long-term, tissue and bone contact): - Cytotoxicity (ISO 10993-5): one test, triplicate samples. - Sensitisation (ISO 10993-10): one test, multiple samples. - Irritation or intracutaneous reactivity (ISO 10993-23): one test. - Material-mediated pyrogenicity (ISO 10993-11): one test. - Acute systemic toxicity (ISO 10993-11): one test. - Subacute/subchronic toxicity (ISO 10993-11): one test. - Genotoxicity (ISO 10993-3): multiple assays (Ames, mammalian cell, in vivo). - Implantation (ISO 10993-6): one test with histopathology at multiple time points. - Chemical characterisation (ISO 10993-18): extractables and leachables analysis. - Biological risk assessment synthesising all of the above.

That is roughly 40-80 individual test results, conducted over 9-18 months at a GLP lab, costing 150,000-400,000 EUR.

Mechanical bench testing: - Static compression per ASTM F2077. - Dynamic compression fatigue per ASTM F2077 (typically 5 million cycles, multiple samples at different load levels). - Subsidence per ASTM F2267. - Expulsion testing (if applicable). - Torsion (if applicable).

Fatigue testing alone typically means 18-36 samples tested to millions of cycles each. That is 100-200 individual test data points before you even get to the other mechanical protocols.

Sterilisation validation: - Sterilisation process validation per ISO 17665 (gamma or EO). - Bioburden testing. - Bacterial endotoxin testing. - Sterile barrier system validation. - Shelf-life of sterility.

Packaging and shelf-life: - Accelerated aging per ASTM F1980. - Real-time aging (started in parallel). - Transport simulation per ASTM D4169 or ISTA.

Risk management file per EN ISO 14971:2019+A11:2021: - Risk analysis with typically 80-250 identified hazards. - Risk control verification for each hazard. - Benefit-risk analysis.

Animal testing (where required): - Large animal implantation study (typically sheep or pig spine for spinal devices), 6-12 animals, 12-26 week follow-up with histopathology.

Add it all up. For a Class IIb implant, you are looking at 600-1,500 individual test results, 60-100 protocols, 18-30 months of pre-clinical work, and 500,000 to 1,500,000 EUR of testing spend. For a Class III implantable with active components, the upper bound easily reaches 3,000 test results.

The founder who budgeted "10 months to first-in-human" is now staring at "24 months to pre-clinical complete, then clinical investigation application, then ethics committee and competent authority review, then patient enrolment."

The Subtract to Ship playbook

You cannot subtract your way out of pre-clinical testing. The evidence is not optional. But you can subtract the waste — the tests run in the wrong order, the protocols drafted without statistical planning, the samples scrapped because design changes invalidated them. Here is the lean approach:

  1. Build the risk management file first, not last. EN ISO 14971:2019+A11:2021 is not a documentation exercise you do at the end to satisfy an auditor. It is the upstream driver of everything else. Every hazard identified in the risk analysis generates a testing requirement. Build the risk file early, update it continuously, and use it to scope the test plan.
  2. Write a pre-clinical evaluation plan. Analogous to a clinical evaluation plan but for non-clinical evidence. It lists every test required (driven by the risk file and the GSPRs in MDR Annex I), the applicable standard, the acceptance criteria, the sample size justification, and the status. This document becomes the spine of your Annex XV pre-clinical submission.
  3. Freeze the design before you run pivotal tests. Running an ISO 10993 suite on one design iteration and then changing the surface coating means running it again. Most pre-clinical budget overruns come from running expensive tests on a design that is still moving.
  4. Sequence tests by dependency and cost. Run cheap bench tests early to fail fast. Run biocompatibility after material selection is stable. Run animal studies after mechanical bench testing has demonstrated basic function. Run the full ISO 10993 suite on the final manufacturing process, not the prototype process.
  5. Use a GLP lab for biocompatibility, a accredited lab for electrical safety, and qualified internal testing (with validated equipment) for routine bench tests. The competent authority looks for credible test reports with proper quality systems behind them. Skimping on lab credibility is false economy.
  6. Plan the clinical investigation application in parallel, not afterwards. The Annex XV submission package — Investigator's Brochure, clinical investigation plan, pre-clinical summary, risk management file, justification of benefit-risk — takes months to assemble even when the tests are done. Start the writing while the testing is running.
  7. Budget honestly. For a Class IIb implant, budget 500,000-1,500,000 EUR for pre-clinical testing alone, plus 200,000-600,000 EUR for the clinical investigation itself. This is not the cheap phase. This is the expensive phase. Investors who are told otherwise are being misled and will lose confidence when reality hits.

The Subtract to Ship point: pre-clinical evidence is not where you save money. It is where you spend deliberately and in the right order. The savings come from not running tests twice, not scrapping samples after design changes, and not being rejected by the competent authority because your submission was incomplete.

Reality Check

  1. Do you have a pre-clinical evaluation plan listing every test required, with sample size justification and acceptance criteria?
  2. Is your risk management file per EN ISO 14971:2019+A11:2021 actively driving the test plan, or is it a document you update retrospectively?
  3. Have you identified which tests must be run at a GLP-compliant or accredited lab, and budgeted accordingly?
  4. Is your design frozen, or are you running expensive tests on a design still in flux?
  5. Have you mapped each identified hazard in the risk file to a specific verification test and acceptance criterion?
  6. Does your Annex XV pre-clinical summary explicitly justify why the pre-clinical evidence is sufficient to support first-in-human exposure?
  7. Is your clinical investigation application being drafted in parallel with testing, or waiting until tests finish?
  8. Does your budget reflect the realistic pre-clinical cost for your device class, or does it assume an order of magnitude less?

Frequently Asked Questions

Can I start my clinical investigation before all pre-clinical tests are complete? No. MDR Article 62 and Annex XV Chapter I require the pre-clinical evidence to be complete and summarised in the submission to the competent authority. The competent authority will review the pre-clinical package before authorising the investigation. Incomplete pre-clinical evidence is one of the most common reasons for rejection or request for information.

Do I need animal testing for every device? No. Animal testing is required only where pre-clinical in vitro and bench testing cannot adequately characterise the safety profile of the device, particularly for implants, devices in contact with the central nervous system, or devices with novel materials or mechanisms of action. EN ISO 10993-2 and the ISO 10993-1 biological evaluation framework drive the decision. For many software devices and non-invasive devices, no animal testing is required.

What counts as "sufficient" pre-clinical evidence for the competent authority? Sufficiency is judged against the risk profile of the device and the proposed clinical investigation design. A first-in-human study with a Class III implantable requires far more pre-clinical evidence than a post-market clinical follow-up study on an already-marketed device. The Annex XV submission must include an explicit justification of why the pre-clinical data is appropriate to the proposed investigation.

How do I justify sample sizes for bench testing? Sample sizes for mechanical bench testing are typically driven by statistical considerations — confidence levels, power, variability of the test method. Many ISO and ASTM test methods specify minimum sample sizes. For custom protocols, sample size justification should reference standard statistical approaches and be documented in the test plan before testing begins. Retrospective justification is not credible to an auditor.

Can I use published literature to substitute for some pre-clinical testing? In limited cases, yes. Literature on equivalent materials or design features can reduce the scope of some testing, but it rarely eliminates it. The MDCG guidance on clinical evaluation equivalence (MDCG 2020-5) gives useful framing even though it is aimed at clinical rather than pre-clinical evidence. For novel devices, literature substitution is minimal.

How long does pre-clinical testing actually take? For a Class IIb or III device, typically 18-30 months from design freeze to complete pre-clinical package. Biocompatibility alone (especially implantation with chronic histopathology) is often 12-18 months. Fatigue testing to millions of cycles can be 6-12 months. Running tests in parallel shortens calendar time but increases cost. Founders who plan for 6-9 months are setting themselves up for a serious timeline miss.

Sources

  1. Regulation (EU) 2017/745 on medical devices, consolidated text. Article 62, Annex XV Chapter I, Annex I.
  2. EN ISO 14971:2019+A11:2021 — Medical devices — Application of risk management to medical devices.
  3. EN ISO 14155:2020+A11:2024 — Clinical investigation of medical devices for human subjects — Good clinical practice.
  4. EN ISO 10993-1:2025 — Biological evaluation of medical devices — Part 1: Evaluation and testing within a risk management process.