Clinical Evaluation for Medical Devices

This is a first draft (v0.1). We are publishing it early on purpose, so that everyone can read it, comment on it, ask questions and suggest changes. It has not yet been reviewed by the community, and it will change as your comments come in.

The text of Regulation (EU) 2017/745 (the MDR) is what binds you. Where this page and the Regulation differ, the Regulation wins; please tell us so we can fix the page.

1. Purpose and scope

This guidance is for people who write, review or manage clinical evaluations at medical device manufacturers: regulatory affairs and quality (QA/RA) specialists, clinical specialists, and the engineers who own the design and the risk file. It covers devices placed on the EU market under the MDR.

It does not cover in vitro diagnostic devices, which follow Regulation (EU) 2017/746 and the performance evaluation route. It does not cover how to design or run a clinical investigation. It is not legal advice.

Article references in this page point to the consolidated text of the MDR published on EUR-Lex. The sources panel links the official texts.

Source of the MDR text quoted or paraphrased here: Regulation (EU) 2017/745, consolidated text of 19 July 2026, EUR-Lex, © European Union. Quoted extracts are unchanged; the rest is paraphrased.

2. Key definitions

The MDR defines the core terms in Article 2. Getting them exactly right matters, because a word used loosely weakens a report: a technical test result presented as a clinical benefit, or a literature search presented as the whole evaluation.

Three things that are easy to mix up

Clinical benefit is an outcome for the patient. Less pain, restored function, fewer reinterventions, a correct diagnosis that changes treatment. If you cannot say what changes for the patient, you have not yet described a clinical benefit.

Clinical performance is the device doing its job in the patient. It is the medical effect that achieves the intended purpose and leads to the benefit. For a wound dressing, keeping the wound moist is a technical characteristic; the reduction in wound area at a stated timepoint is clinical performance; complete closure, less pain or fewer infections are clinical benefits.

Technical performance is not clinical performance. The MDR does not define technical performance as a term, but the distinction follows from Article 2(52): clinical performance is the medical effect that stems from the technical characteristics, not the characteristics themselves. Tensile strength, flow rate, accuracy on a bench or software verification results are technical performance. They support the evaluation but do not demonstrate a clinical benefit on their own.

The intended purpose (Article 2(12)) is the use stated in the manufacturer's label, instructions for use, promotional or sales material and the clinical evaluation. Write it once, precisely, and reuse the same wording everywhere. The state of the art is used in the MDR (for example Annex I, Section 1) but not defined in it; see Section 7.

3. When it is required and how often it is updated

Every device needs one

Article 10(3) requires manufacturers to conduct a clinical evaluation in accordance with Article 61 and Annex XIV, including a PMCF. A class I device needs a clinical evaluation too; what changes with the class and the risk is the depth and the amount of evidence, not whether you do one.

Article 61(1) sets the basis: conformity with the relevant general safety and performance requirements, the evaluation of undesirable side-effects and the acceptability of the benefit-risk ratio must be based on clinical data providing sufficient clinical evidence. You must also specify and justify the level of clinical evidence your device needs, in view of its characteristics and intended purpose. Annex XIV, Section 2 adds that the depth and extent of the evaluation must be proportionate to the nature, classification, intended purpose and risks of the device, and to your claims.

When clinical data are not the basis

Article 61(10) allows an exception where demonstrating conformity on the basis of clinical data is not deemed appropriate. It opens with "Without prejudice to paragraph 4", so it does not remove the requirement for clinical investigations that Article 61(4) sets for implantable and class III devices. You then need an adequate justification based on your risk management, on how the device interacts with the human body, on the clinical performance intended and on your claims, and you must substantiate in the technical documentation why non-clinical testing alone (performance evaluation, bench testing, pre-clinical evaluation) is adequate. You still produce a clinical evaluation; it documents why the evidence is non-clinical. MDCG 2020-6 states that Article 61(10) cannot be applied to class III or implantable devices.

When clinical investigations are required

For implantable devices and class III devices, Article 61(4) requires clinical investigations. It exempts a device designed by modifying a device you already market when three conditions are all met: you have demonstrated equivalence to the marketed device under Annex XIV, Section 3, the notified body has endorsed that demonstration, and the clinical evaluation of the marketed device is sufficient to demonstrate conformity of the modified device. Article 61(6) adds two cases: (a) devices lawfully placed on the market or put into service under Directive 90/385/EEC or Directive 93/42/EEC whose clinical evaluation is based on sufficient clinical data and complies with any relevant product-specific common specification; and (b) the device types listed in Article 61(6)(b), such as sutures, staples, screws, plates and pins, under the same conditions on sufficient clinical data and common specifications. Relying on another manufacturer's device is covered in Section 5.3.

How often you update it

What the MDR says:

• The clinical evaluation and its documentation must be updated throughout the life cycle of the device with clinical data from your PMCF plan and your post-market surveillance plan (Article 61(11)).

• For class III and implantable devices, the PMCF evaluation report and, if indicated, the summary of safety and clinical performance (SSCP) must be updated at least annually with such data (Article 61(11)).

• Data from your post-market surveillance system must be used, among other things, to update the clinical evaluation (Article 83(3)(c)).

• For devices assessed under Annex IX, the notified body considers whether specific milestones need to be defined so that it can review updates to the clinical evidence that result from post-market surveillance and PMCF data (Annex IX, Section 4.7).

• The periodic safety update report (PSUR), which sets out the main findings of the PMCF and the conclusions of the benefit-risk determination, is updated at least annually for class IIb and class III devices, and when necessary and at least every two years for class IIa devices (Article 86(1)).

What the MDR does not say: it fixes no interval for the CER as a whole. The PMCF evaluation report in the second bullet is part of the CER (Annex XIV, Part B, Section 7), and its conclusions are taken into account in the clinical evaluation (Annex XIV, Part B, Section 8). MDCG 2020-13 also asks the notified body to identify when updates to the CER will be assessed and at what frequency. Beyond that, you decide, and justify in your procedure, which events trigger an update: a new PMCF evaluation report, a PSUR that changes the benefit-risk picture, a new safety signal, a change to the device or intended purpose, or a change in the state of the art.

4. Clinical evaluation plan

Annex XIV, Section 1(a) requires you to establish and update a clinical evaluation plan (CEP). Its minimum content, in plain terms:

• the general safety and performance requirements that need support from relevant clinical data;

• the intended purpose of the device;

• the intended target groups, with clear indications and contra-indications;

• the intended clinical benefits to patients, with relevant and specified clinical outcome parameters;

• the methods you will use to examine the qualitative and quantitative aspects of clinical safety, with clear reference to residual risks and side-effects;

• an indicative list and specification of the parameters you will use, based on the state of the art in medicine, to determine whether the benefit-risk ratio is acceptable for each indication and intended purpose;

• how you will address benefit-risk issues tied to specific components, such as pharmaceutical substances or non-viable animal or human tissues;

• a clinical development plan, from exploratory investigations to confirmatory ones and to PMCF, with milestones and a description of potential acceptance criteria.

A plan contains no results

The plan says what you will look for, where, how you will judge it and what would count as acceptable. It does not contain the answer. If your CEP quotes complication rates from your own studies, or acceptance criteria that happen to match the results you already have, an assessor will read it as a plan written after the fact. Keep results in the report.

A few habits make the plan easier to defend:

• Write the outcome parameters before you search. For each clinical benefit, name the endpoint, how it is measured, the scale or unit, and the timepoint you will compare at. A pain score at 12 months is not comparable with a pain score at 6 weeks. A score reported on 0 to 10 is not directly comparable with one on 0 to 100 mm until the scale is confirmed, and an NRS is not a VAS.

• Tie each safety question to the risk file. The plan should point to the residual risks and side-effects you need clinical data on, so the report can close each one.

5. Identifying data

Annex XIV, Section 1(b) asks you to identify the available clinical data relevant to the device and its intended purpose, and any gaps in clinical evidence, through a systematic scientific literature review. Article 61(3) adds that the evaluation must follow a defined and methodologically sound procedure based on a critical evaluation of the literature, on all available clinical investigations, and on a consideration of currently available alternative treatment options. The literature route in Article 61(3)(a) is tied to two conditions: the device is demonstrated to be equivalent to the device the data relate to, and the data adequately demonstrate compliance with the relevant general safety and performance requirements.

There are three families of data. An evaluation can draw on more than one.

5.1 Scientific literature

The literature review has two jobs, best run as two searches. The first finds data on your device, or on a device you claim is equivalent. The second describes the state of the art: the clinical condition, the alternative treatments and the similar devices, with the outcomes they achieve.

Make the search reproducible. Record the databases, the date of each search, the exact search strings, the limits, and the number of records found, screened, excluded (with reasons) and included. Someone repeating your search should get the same list when applying your strings and date limits. Include unfavourable publications as well as favourable ones; Annex XIV, Section 2 requires the evaluation to take both into account.

Appraise from the full text. An abstract rarely gives you the population, the follow-up, the scale or the denominator you need to appraise a result.

5.2 Your own data

Your own clinical data come from clinical investigations of the device, from post-market surveillance and from PMCF (all listed as sources in Article 2(48)). Article 61(3)(b) requires a critical evaluation of the results of all available clinical investigations, taking into account whether they were performed under the MDR's clinical investigation requirements (Articles 62 to 80, any acts adopted pursuant to Article 81, and Annex XV).

Complaints, vigilance reports, registry data, user surveys and PMCF studies all feed the evaluation, but they need the same discipline as literature: a defined source, a defined period, and an honest denominator. A complaint count without the number of devices used is not a rate.

5.3 Data from an equivalent device

Annex XIV, Section 3 lets you base a clinical evaluation on clinical data from a device for which equivalence can be demonstrated. Equivalence is assessed on three groups of characteristics:

The characteristics must be similar to the extent that there would be no clinically significant difference in safety and clinical performance. The demonstration needs proper scientific justification, and you must clearly show that you have sufficient access to the data on the device you claim equivalence with.

For implantable and class III devices, if you want to avoid a clinical investigation by relying on equivalence to another manufacturer's device, Article 61(5) adds two conditions to those of Article 61(4): a contract that explicitly allows you full access to the technical documentation on an ongoing basis, and clear evidence to the notified body that the original clinical evaluation complied with the MDR. Without such a contract you can still use the other device's data, if equivalence and sufficient access are demonstrated, but the clinical investigation remains required unless the device falls under Article 61(4) or 61(6). MDCG 2023-7 explains that the contract applies only to the Article 61(5) case.

Two MDCG documents are the usual reference here: MDCG 2020-5 (Guidance on clinical evaluation – Equivalence) and MDCG 2023-7 (Guidance on exemptions from the requirement to perform clinical investigations pursuant to Article 61(4)-(6) MDR and on ‘sufficient levels of access’ to data needed to justify claims of equivalence).

A practical test before you commit to equivalence: list every difference between the two devices and, for each, write why it makes no clinically significant difference, with evidence. Without the other device's material and design data, that comparison cannot be completed.

6. Appraisal and analysis

Appraisal

Annex XIV, Section 1(c) requires you to appraise all relevant clinical data by evaluating their suitability for establishing the safety and performance of the device. Appraisal comes before any conclusion and is done source by source.

For each source, ask two questions and record the answers:

• Is it relevant? Is it the device under evaluation or a demonstrated equivalent? The same indication, population, site and user? Does it report the outcomes in your plan, at the timepoints in your plan?

• Is it reliable? What is the design, how many patients, how complete is the follow-up, how are outcomes measured, are there obvious sources of bias or conflicts of interest?

Define your appraisal criteria in the plan, apply them the same way to every source, and keep the record. A source can be relevant and weak, or strong and only partly relevant, and both facts belong in the report.

Analysis

Annex XIV, Section 1(e) requires you to analyse all relevant clinical data to reach conclusions about the safety and clinical performance of the device, including its clinical benefits.

Work outcome by outcome. For each clinical benefit, performance and safety outcome in the plan, show the data from each source, state the unit, the scale and the timepoint next to every value, and compare with the acceptance criterion set from the state of the art (Section 7). Only pool or compare values that were measured the same way at a comparable timepoint. When sources differ, say how you handled it, or keep them apart.

If the analysis shows a gap, Annex XIV, Section 1(d) points to generating new or additional clinical data through properly designed clinical investigations in line with the clinical development plan. PMCF can address residual uncertainties once the evidence already suffices to demonstrate conformity, such as long-term performance or rare side-effects. It cannot fill a gap that prevents that demonstration: an indication without sufficient clinical evidence is either supported before CE marking or removed from the claims. MDCG 2020-6 says this explicitly for legacy devices.

7. State of the art and acceptance criteria

What the state of the art is for

The MDR refers to the generally acknowledged state of the art in Annex I, Section 1, when it requires risks to be acceptable when weighed against benefits, and Annex XIV, Section 1(a) asks for benefit-risk parameters based on the state of the art in medicine. Article 61(3)(c) requires the evaluation to consider currently available alternative treatment options. The state of the art is therefore the benchmark your device is judged against: what current medical practice achieves for the same condition, with similar devices and with the alternative treatments.

A useful state of the art section describes the clinical condition and its natural course, the available treatment options (including non-device and conservative options where relevant), the similar devices on the market, and, for each relevant outcome, what those options achieve, with sources.

Where acceptance criteria come from

Acceptance criteria are the values that tell you whether your device's results are acceptable: a minimum success rate, a maximum complication rate, a minimum improvement in a score. They come from the state of the art: the outcomes reported for similar devices and for alternative treatments. They never come from your own device's study results. A criterion derived from your own data cannot fail, so it shows nothing.

Some practical rules that keep the criteria defensible:

• Set them before you analyse your own data, and record them in the plan.

• Make them traceable. Each criterion cites the sources it was derived from and states how it was derived (for example, the range or a summary statistic of the values reported for similar devices).

• Compare like with like. The same endpoint, the same instrument, the same unit and scale, the same timepoint. If the sources report a score on different scales, or most of them do not say which scale they used, record that as a gap rather than guessing, and decide openly how you will handle it.

• Keep one benchmark. If the plan, the report and the state of the art section quote acceptance criteria, they must be the same numbers from the same pool of sources.

• Do not confuse the criterion with a minimal clinically important difference. A clinically meaningful change for the patient helps interpret an outcome, but it is a property of the instrument, not a statement of what similar devices achieve.

8. Benefit-risk conclusion

Annex I, Section 1 requires that any risks associated with use constitute acceptable risks when weighed against the benefits to the patient, taking into account the generally acknowledged state of the art. Annex I, Section 8 requires all known and foreseeable risks and undesirable side-effects to be minimised and acceptable when weighed against the evaluated benefits from the achieved performance of the device.

Build the conclusion from the analysis, not from a template sentence:

• State the clinical benefits shown, with the evidence and how it compares with the acceptance criteria.

• State the residual risks and side-effects, with their observed frequency and severity, and how they compare with the state of the art.

• Link each residual risk to the risk management file. Annex XIV, Part B, Section 6.2(d) and Section 8 tie the clinical evaluation and PMCF to the risk management referred to in Annex I, Section 3.

• Say for which indications, populations and users the benefit-risk ratio is acceptable. If the evidence supports only part of the intended purpose, say so, and either narrow the claims or plan the data that will close the gap.

• List what remains uncertain and how PMCF will address it.

9. Clinical evaluation report

Article 61(12) requires the clinical evaluation, its results and the clinical evidence derived from it to be documented in a clinical evaluation report (CER), as referred to in Annex XIV, Section 4. Except for custom-made devices, the CER is part of the technical documentation (Annex II, Section 6.1(c)). Annex XIV, Section 4 adds that both favourable and unfavourable data considered in the evaluation must be included in the technical documentation.

The MDR prescribes no table of contents for the CER. A structure that follows the logic of the evaluation:

• the device, its intended purpose, indications, contra-indications, target population and claims;

• the scope of the evaluation and the clinical evaluation plan it follows;

• the state of the art and the acceptance criteria derived from it;

• the data identified, with the search method and its results, and the appraisal of each source;

• the analysis, outcome by outcome, against the acceptance criteria;

• the benefit-risk conclusion, the residual uncertainties and their link to PMCF;

• where applicable, the equivalence demonstration, the justification for not performing a clinical investigation in the Article 61(6) cases (Article 61(7)), the consideration given to the views of an expert panel, if you consulted one on your clinical development strategy (Article 61(2), open to class III and certain class IIb devices), and the justification for relying on non-clinical data, which Article 61(10) requires in the technical documentation referred to in Annex II;

• the qualifications of the evaluators and the dates of the evaluation.

Under Annex IX, Section 4.8, the notified body documents the outcome of its assessment of your clinical evaluation in a clinical evaluation assessment report. MDCG 2020-13 is the clinical evaluation assessment report template. Use it to check your report against the questions an assessor will ask, not as a template to fill. For class III implantable devices and class IIb active devices intended to administer and/or remove a medicinal product (Article 54(1)), the notified body also follows the clinical evaluation consultation procedure, unless one of the cases in Article 54(2) applies.

For implantable and class III devices other than custom-made or investigational devices, Article 32(1) requires a summary of safety and clinical performance, and Article 32(2)(f) lists a summary of the clinical evaluation among its contents. MDCG 2019-9 rev.1 (Summary of safety and clinical performance) is the guidance on it. Keep the two consistent: any figure in the SSCP should be traceable to the CER.

Use the same words for the device description, intended purpose and claims as in the instructions for use and the labelling.

10. Link to PMCF

Annex XIV, Part B, Section 5 describes PMCF as a continuous process that updates the clinical evaluation and is addressed in the manufacturer's post-market surveillance plan. Through it you proactively collect clinical data from the CE-marked device used as intended, to confirm safety and performance over its expected lifetime and to detect emerging risks.

The links run both ways:

• From the CER to the PMCF plan. The residual uncertainties and gaps you listed in the CER become the specific objectives of the PMCF plan. Annex XIV, Part B, Section 6.2 sets the plan's minimum content, including its methods and their rationale, its specific objectives, a reference to the relevant parts of the CER and to the risk management, and a justified time schedule.

• From PMCF back to the CER. You analyse the PMCF findings in a PMCF evaluation report, which is part of the clinical evaluation report and the technical documentation (Annex XIV, Part B, Section 7). Its conclusions are taken into account in the clinical evaluation and in the risk management, and any preventive or corrective measure it identifies must be implemented (Annex XIV, Part B, Section 8).

If you consider PMCF not applicable to your device, the technical documentation must contain a justification instead of a PMCF plan and report (Annex II, Section 6.1(d)). Make that justification specific to the device and its evidence.

For implantable and class III devices exempted from clinical investigations under Article 61(4), the notified body checks that the PMCF plan is appropriate and includes post-market studies to demonstrate safety and performance. MDCG 2020-7 (PMCF plan template) and MDCG 2020-8 (PMCF evaluation report template) are the usual references for structure.

The post-market surveillance plan (Article 84, Annex III, Section 1) covers the PMCF plan or the justification why PMCF is not applicable.

11. Open questions for contributors

Practice varies on these points and this draft does not settle them. If you work on clinical evaluations, please comment or suggest changes.

• How much literature is enough? The MDR sets no number of studies, patients or years of follow-up. How do you decide that the evidence is sufficient for a given device and risk class, and how do you document that decision?

• Access to data for equivalence. Annex XIV, Section 3 requires sufficient levels of access to the data for every equivalence claim, and a contract is needed only when an implantable or class III device relies on another manufacturer's device to avoid a clinical investigation (Article 61(5)). What have you seen accepted as sufficient access without a contract, and what contract terms have worked when one was needed?

• Legacy devices. Devices placed on the market under the previous Directives can still be placed on the market until the dates and under the conditions set in Article 120(3) to (3c). How much weight do you give to long market experience, and where does MDCG 2020-6 (sufficient clinical evidence for legacy devices) leave room for judgement?

• Acceptance criteria when the literature is thin or inconsistent. When few similar devices report the outcome, or the sources use different scales and timepoints, how do you set a criterion without inventing one?

• When to update the CER. What triggers and intervals do you use for each risk class, and how do you align them with the PSUR and the PMCF evaluation report?

• Software and orphan devices. MDCG 2020-1 covers the clinical evaluation of medical device software and MDCG 2024-10 covers orphan devices. What has worked when little clinical data exists?

• What is missing from this page? If a topic you deal with every week is not here, propose a section.