MDR Articles 61–82 · Annexes XIV and XV
A clinical investigation is not the clinical evaluation
One asks what the complete evidence means. The other generates human data when the answer is still missing.
Confusing the two creates opposite errors: commissioning a study before defining the evidence gap, or trying to write around a gap that only new clinical data can close. The better route begins with the claims and ends with a study only when the logic requires one.
Separate the two concepts ↓Can the claimed safety, performance and benefit be demonstrated?
The rule in one sentence
Evaluate first, investigate when a material clinical question remains unanswered
01The clinical evaluation defines the evidence needed, identifies what already exists and tests whether it is sufficient.
02A clinical investigation should be designed around a specific unresolved question, not used as a substitute for an evidence strategy.
Two connected processes
Same evidence system, different jobs
The MDR definitions make the boundary clear. The clinical evaluation is continuous and comprehensive. A clinical investigation is a defined investigation involving human subjects.
The decision framework
Clinical evaluation
A systematic and planned process that continuously generates, collects, analyses and assesses clinical data to verify safety, performance and clinical benefit when the device is used as intended.
- Required for every medical device under the MDR
- Covers all relevant favourable and unfavourable data
- Connects evidence to claims, GSPRs and benefit-risk
- Is documented in the Clinical Evaluation Report
- Continues throughout the device lifecycle
A justified conclusion on whether the clinical evidence is sufficient
The data-generation method
Clinical investigation
A systematic investigation involving one or more human subjects, undertaken to assess the safety or performance of a device.
- Required when the legal route or evidence gap demands it
- Answers pre-specified safety, performance or benefit questions
- Requires a sponsor, CIP, ethical review and regulatory pathway
- Produces data that must still be critically appraised
- May occur before or after CE marking
A clinical investigation report and data for the wider clinical evaluation
The MDR uses “clinical investigation” for medical devices. “Clinical trial” is commonly associated with medicinal products and should not be used as the default device term.
The investigation decision
Start with the device situation, then test evidence sufficiency
Select a situation to see the governing logic. A device can move between these situations when its intended purpose, design, claims or evidence base changes.
Investigation is the starting expectation
Check the Article 61 exemption routes precisely
For implantable and Class III devices, Article 61(4) states that clinical investigations shall be performed. The exceptions in Article 61(4), (5) and (6) have distinct conditions. MDCG 2023-7 explains that these routes are independent unless the Regulation explicitly links them.
- Do not merge conditions from different exemption cases
- Demonstrate equivalence under Annex XIV where relied upon
- Show sufficient access to data when the selected route requires it
- Document the justification in the CER and PMCF strategy
Evidence sufficiency drives the decision
There is no automatic study requirement, but there is no automatic waiver
For devices that are neither implantable nor Class III, the manufacturer defines and justifies the necessary clinical evidence level. A clinical investigation becomes necessary when reliable existing evidence cannot answer a material safety, performance or benefit question.
- Consider novelty, claims, risk and clinical context
- Appraise relevance and quality, not just data quantity
- Separate clinical gaps from technical or biological gaps
- Document why a new investigation is or is not needed
The old evidence must survive the new question
A familiar device can create an unfamiliar evidence gap
New indications, patient populations, anatomical sites, operating principles, algorithms, materials, accessories or stronger clinical claims may reduce the relevance of the previous evidence package. The change assessment should identify exactly which clinical conclusions remain transferable.
- Map each change to affected claims and risks
- Reassess literature and equivalence assumptions
- Consider whether bench evidence can answer the question
- Generate new human data where clinical uncertainty remains
The lifecycle can generate new questions
Use PMCF to close defined residual uncertainties
PMS signals, rare events, long-term performance, new user groups or limited pre-market follow-up may justify a post-market investigation. PMCF is not a generic promise to obtain missing evidence later, and it does not make an inadequate pre-market case acceptable.
- Define the residual question and decision threshold
- Select the correct Article 74 or other applicable route
- Feed results into risk management and CER updates
- Escalate new safety signals without waiting for routine review
The evidence chain
Do not begin with a protocol
A well-designed investigation can still be strategically useless if it measures endpoints that do not support the intended purpose, claims or unresolved GSPRs.
- 01Scope
Fix the intended purpose
Define patients, users, indications, contraindications, setting, operating principle, variants and duration of use.
- 02Claims
Convert claims into questions
State the clinical performance, benefit and safety outcomes that the evidence must demonstrate.
- 03Existing evidence
Inventory every relevant source
Include own-device data, literature, equivalent-device data, PMS, PMCF and relevant alternatives.
- 04Appraisal
Test relevance and quality
Examine population, device version, design, bias, endpoints, follow-up, statistics and access to underlying data.
- 05Gap
Name the unresolved uncertainty
Write the gap as an answerable question and explain why existing evidence cannot close it.
- 06Generation
Choose the right method
Use a clinical investigation only when the unresolved question requires data from human use.
- 07Integration
Return results to the CER
Appraise the study, update benefit-risk, GSPR mapping, labelling, risk controls and residual gaps.
What counts as clinical data
The MDR recognises several routes into the evidence base
Article 2(48) identifies clinical data generated through use of the device. Each source can be valuable, but none becomes sufficient merely because it is labelled “clinical”.
Bench testing, software verification, biocompatibility testing and animal data are important evidence, but they are not clinical data. They answer different questions within the overall conformity case.
Investigations of the device concerned
The closest source to the actual design, intended purpose and population, provided the investigation is methodologically sound and relevant.
Investigations or studies of an equivalent device
Usable for conformity only when technical, biological and clinical equivalence is demonstrated and the data are adequately accessible.
Peer-reviewed reports of clinical experience
Literature may support the device or an equivalent device, while similar-device data can also help define state of the art and known risks.
Clinically relevant PMS and PMCF information
Complaints, registries, PMCF studies, vigilance and real-world use can confirm conclusions or reveal that the evidence strategy must change.
Gap triage
Not every evidence gap belongs in a clinical investigation
The fastest way to overspend is to ask human subjects to answer a question that should have been resolved by engineering, laboratory or usability work.
Does the device perform as designed?
Accuracy, strength, reliability, interoperability, cybersecurity, software verification, transport or shelf-life questions normally require technical evidence.
Bench, analytical or software verification and validation
Is the biological interaction acceptable?
Material characterisation, toxicology, biocompatibility and preclinical questions should be addressed within the biological evaluation strategy.
Biological evaluation and justified preclinical testing
Can intended users operate it safely?
Interface comprehension, critical tasks, training and foreseeable use error often require usability engineering rather than a clinical investigation.
Formative and summative usability evaluation
Does it achieve the intended clinical result safely?
Unresolved questions about clinical performance, clinical benefit, undesirable side-effects or benefit-risk may require new data from human use.
Clinical investigation or another justified clinical-data route
Choose the regulatory route
“Clinical investigation” is not one administrative pathway
The device status, study purpose, intended use and additional procedures determine the applicable MDR article. National requirements remain relevant.
Conformity assessment
Pre-market or pre-expansion evidence generation
Article 62 applies where an investigation is carried out as part of the clinical evaluation for conformity assessment purposes to establish or verify intended performance, clinical benefit or clinical safety.
- Authorisation and ethical review apply
- Annex XV governs planning and conduct
- The device may be pre-CE or used outside its certified scope
Post-market clinical follow-up
Further assessment within the CE-marked intended purpose
Article 74 addresses PMCF investigations of CE-marked devices used within their intended purpose. Article 74(1) provides a specific notification route where subjects undergo additional invasive or burdensome procedures.
- The investigation must align with the certified intended purpose
- Additional procedures affect the regulatory route
- National notification requirements must be checked
Other clinical investigations
Investigations outside the Article 62 purposes
Article 82 sets minimum MDR protections for other investigations involving devices. Member States may impose additional requirements, so the national pathway must be established before submission.
- Scientific and ethical safeguards still apply
- The study purpose must be classified correctly
- National provisions can differ between Member States
A study conducted after CE marking is not automatically a PMCF investigation under Article 74. Its purpose, device use, intended-purpose boundary and study procedures still need to be assessed.
Design backwards from the claim
A compliant study can still produce unusable evidence
Authorisation is not proof that the design will answer the CER question. The CIP should make the connection between intended purpose, clinical claims, risks, objectives, endpoints and analysis visible before the first subject is enrolled.
If the primary endpoint is met, which exact clinical claim or GSPR conclusion becomes supportable? If that cannot be answered, the design is not finished.
- 01
Representative population
Match the intended patients, relevant subgroups, users and use environment.
- 02
Meaningful endpoints
Measure outcomes linked to performance, benefit, safety and the state of the art.
- 03
Defensible comparator
Use the current clinical alternative or justify why another design answers the question.
- 04
Adequate follow-up
Capture the timing of benefits, foreseeable side-effects and relevant device failures.
- 05
Sample-size logic
Base assumptions on clinically meaningful effects and realistic event rates.
- 06
Bias control
Plan randomisation, blinding, adjudication or other controls where feasible and relevant.
- 07
Missing-data strategy
Predefine handling of withdrawals, protocol deviations and incomplete follow-up.
- 08
Safety visibility
Define events, reporting, monitoring, stopping rules and the analysis population.
After the database is locked
A study result does not enter the CER by copy and paste
The clinical investigation report is a source. The clinical evaluation must still decide how much weight that source deserves and what it changes.
Compare actual conduct with the approved CIP and account for deviations.
Assess bias, precision, completeness, relevance and statistical uncertainty.
Distinguish statistical findings from clinically meaningful benefit and risk.
Revise CER conclusions, risk controls, GSPRs, labelling, PMS and PMCF.
Support only what was actually demonstrated
A successful primary endpoint does not automatically support every population, indication, duration, variant or promotional claim.
Do not make unfavourable evidence disappear
Analyse the cause, affected claims, benefit-risk impact, corrective actions and whether the intended purpose remains defensible.
Uncertainty is itself a regulatory finding
Decide whether the limitation can be resolved through analysis, further investigation, restricted claims, added controls or withdrawal of the claim.
Three evidence stories
The device name does not decide whether a study is needed
The answer changes with intended purpose, novelty, risk, claims, available data and residual uncertainty.
Established reusable surgical instrument
Assume a conventional instrument with a narrow intended purpose, mature technology, well-characterised risks and no novel clinical claim.
Relevant literature, state of the art, verification, reprocessing validation, usability, risk management and market experience may provide a sufficient package.
A clinical investigation is not added merely to make the file look stronger. The manufacturer documents why the available evidence is sufficient.
Software with a new clinical recommendation
Assume an algorithm changes from displaying information to recommending a patient-specific clinical action in a new population.
Software verification can confirm implementation, but it may not demonstrate clinical performance, generalisability, benefit or new use-related risks.
If existing clinical data do not represent the new intended use, a prospective investigation or another justified human-data strategy may be necessary.
Novel Class III implant
Assume a new implant design and material combination with long-term performance claims and limited relevant human experience.
Article 61(4) starts from the requirement to perform clinical investigations, while novelty and long-term uncertainty also make transferable evidence difficult.
The development programme should generate device-specific evidence and plan post-market follow-up for residual long-term questions.
Where evidence programmes fail
Seven expensive shortcuts
Most failures begin before study start or after study completion, at the points where strategy and evidence should connect.
- 01
Writing the CIP before defining the exact CER gap, intended claims and decision the study must support.
- 02
Assuming every piece of human-use information is relevant, reliable and sufficient clinical evidence.
- 03
Using “similar” and “equivalent” as interchangeable terms without completing the Annex XIV equivalence analysis.
- 04
Trying to solve technical, biological or use-related questions through a costly clinical investigation.
- 05
Selecting convenient endpoints that cannot substantiate the intended clinical performance or benefit claim.
- 06
Treating PMCF as a future repair plan for evidence that should have been available before conformity assessment.
- 07
Summarising the final report in the CER without updating benefit-risk, GSPRs, risk management, labelling and residual gaps.
Practical questions
Clinical investigation and evaluation FAQ
Open each question for the distinction that usually changes the regulatory answer.
Does every medical device need a clinical investigation?+
No. Every device needs a clinical evaluation, but a new investigation is not automatically required for every device. For implantable and Class III devices, Article 61(4) starts from a mandatory investigation requirement subject to defined exceptions. For other devices, the need follows from the clinical evaluation objectives and the sufficiency of existing clinical evidence.
Can literature replace a clinical investigation?+
Sometimes existing literature contributes enough evidence, but only when it is relevant, methodologically sound and appropriately linked to the device. Reliance on another device's data for conformity requires a defensible equivalence demonstration. Similar-device literature can still inform the state of the art without proving conformity for the device under evaluation.
Is a PMCF study always a clinical investigation?+
No. PMCF includes different methods, from literature and registry analyses to surveys and formal investigations. When a PMCF activity meets the MDR definition of a clinical investigation, the applicable investigation requirements must be followed.
Can bench testing support a clinical evaluation?+
Yes. Non-clinical evidence is part of the overall conformity case and can support clinical conclusions when it answers the relevant technical or biological question. It should not be relabelled as clinical data, and it cannot replace human data where the unresolved question is inherently clinical.
Does a successful investigation guarantee a positive CER conclusion?+
No. The study must be appraised for relevance, quality, bias, precision and consistency with the complete evidence base. A positive result may support a narrow claim while leaving other safety, population, duration or lifecycle questions unresolved.
Can one investigation support several device variants?+
Potentially, if the variants are sufficiently represented or a scientifically justified bridge shows that the differences do not affect clinical safety, performance or benefit. The CER should make the applicability boundary explicit rather than treating a product family as automatically homogeneous.
Primary regulatory sources
Build the decision on the Regulation, then use guidance to test the reasoning
MDCG documents support a common application of the MDR but are not legally binding. Device-specific facts and applicable national requirements still control the final route.
Regulation (EU) 2017/745
Articles 2, 61–82 and Annexes XIV and XV
MDCG 2021-6 Rev. 1
Questions and answers regarding clinical investigation
MDCG 2024-3
Content of the Clinical Investigation Plan
MDCG 2023-7
Article 61(4)–(6) exemptions and equivalence
MDCG 2020-5
Clinical evaluation and equivalence
MDCG 2020-6
Sufficient clinical evidence for legacy devices
From evidence gap to usable data
Need to decide whether your device requires a clinical investigation?
MDRC can review the clinical evidence strategy, define the unresolved questions, select the regulatory route and align the CIP, endpoints and analysis with the intended CER conclusions.
