What evidence shows the device is safe and works, and how is it kept current?

Introduction#

Before a device can be sold in the EU, the company has to show with clinical data that it is safe and works as claimed. In legal terms it must meet the general safety and performance requirements (GSPR), which require among other things that the device performs as intended with risks acceptable against its benefits. Showing this has to rest on clinical data providing sufficient clinical evidence. Under the Medical Device Regulation, Regulation (EU) 2017/745 (MDR), that evidence comes from a clinical evaluation, a planned and continuing process of gathering and assessing clinical data on the device. Under the In Vitro Diagnostic Medical Device Regulation, Regulation (EU) 2017/746 (IVDR), which covers tests run on samples taken from the body, it comes from a performance evaluation. For an implantable device or a class III device, the highest risk class, Article 61(4) also decides whether a clinical investigation, a study of the device in people, has to run first.

The chapter covers what each evaluation has to show, when the company can rely on data from an equivalent device instead of its own, and when an investigation is required. It also covers the advice available before an investigation, and how follow-up after launch keeps the evidence current. Four illustrative products run through the book: a wearable cardiac monitor from a US company and triage software enabled by artificial intelligence (AI) from a German company. The others are a spinal implant certified under the former directive and a near-patient cardiac troponin test from a Swiss company. The law is stated as consolidated on 19 July 2026 for the MDR and 10 January 2025 for the IVDR. The sources were checked on 29 September 2026.

1. What do clinical and performance evaluation establish?#

Manufacturers conduct a clinical evaluation under MDR Article 61 and Annex XIV, including PMCF. For IVDs the counterpart is a performance evaluation under IVDR Article 56 and Annex XIII, including PMPF. (MDR Art. 10(3); IVDR Art. 10(3))

(MDR Art. 2(44), 2(48), 2(51), 2(53))

Annex I of the MDR sets the GSPRs. Devices have to achieve the performance their manufacturer intends, and any risks have to be acceptable when weighed against the benefits to the patient. (MDR Annex I s. 1, 8)

Under Article 61(1), confirming conformity with the relevant GSPRs, and evaluating side-effects and the benefit-risk ratio, have to rest on clinical data providing sufficient clinical evidence. The manufacturer specifies and justifies the level of clinical evidence, appropriate to the device's characteristics and intended purpose. (MDR Art. 61(1))

The evaluation has to be thorough and objective, and take into account both favourable and unfavourable data. Its depth and extent are proportionate to the device's nature, class, intended purpose and risks, and to the manufacturer's claims. Our reading, which marks the authors' own interpretation, is that each claim added to the label or sales material widens what the evaluation has to support. (MDR Annex XIV s. 2)

What guidance and notified-body papers are worth#

The Medical Device Coordination Group (MDCG) is made up of members appointed by each Member State to represent its competent authorities. A Commission representative chairs it. (MDR Art. 103(2), (5))

Each MDCG guidance document states that its views are not legally binding, and that only the Court of Justice of the European Union can give binding interpretations of Union law. (Commission MDCG page)

Team-NB, the European association of medical device notified bodies, publishes position papers. Its Best Practice Guidance (BPG) on technical documentation calls them the interpretation of Team-NB and affiliated notified bodies. Our reading is that these papers show what assessors expect, and carry no legal status. Team-NB publishes current versions on its website. (Team-NB BPG V4, pp. 1 and 5)

2. What do the clinical evaluation plan and report contain?#

Annex XIV Part A requires a clinical evaluation plan (CEP), established first and kept updated. It includes at least the GSPRs that need clinical data, the intended purpose, and the target groups with indications and contra-indications. It describes the intended clinical benefits "with relevant and specified clinical outcome parameters". (MDR Annex XIV, Part A s. 1(a))

The plan also specifies the methods for examining safety, including residual risks and side-effects. It sets the parameters used to decide whether the benefit-risk ratio is acceptable in light of the state of the art in medicine. Its clinical development plan, with milestones and potential acceptance criteria, runs from exploratory investigations to confirmatory ones, and then to PMCF. (MDR Annex XIV, Part A s. 1(a))

The Annex names first-in-man, feasibility and pilot studies as exploratory, and gives pivotal clinical investigations as its example of a confirmatory one. MDCG 2020-6 defines the state of the art as the developed stage of current technical capability and accepted clinical practice. (MDR Annex XIV, Part A s. 1(a); MDCG 2020-6, pp. 5 and 6)

A systematic literature review identifies the relevant clinical data and any gaps, and all relevant data are appraised for suitability. The manufacturer generates new data through properly designed clinical investigations where outstanding issues need them, and analyses all relevant data to reach conclusions. (MDR Annex XIV, Part A s. 1(b) to (e))

The results go into a clinical evaluation report (CER), which supports the conformity assessment and forms part of the technical documentation. That documentation includes the favourable and unfavourable data considered. (MDR Art. 61(12); Annex XIV s. 4)

The notified body, the conformity assessment body, records its conclusions on the clinical evidence in a clinical evaluation assessment report (CEAR). MDCG 2020-13 is the template, and its review covers equivalence, benefit-risk, the PMCF plan and whether milestones are needed. (MDCG 2020-13, p. 3)

MDCG 2020-13 asks whether the literature search terms are wide enough to find data on the state of the art, risks and side-effects. A search restricted to the manufacturer's own product or its chosen equivalent is not acceptable, and the assessor looks for benchmarks drawn from aggregate data where they exist. (MDCG 2020-13, pp. 14 and 16)

3. When can the evidence come from an equivalent device?#

A clinical evaluation may rest on clinical data from a device shown to be equivalent. Technical, biological and clinical characteristics are all considered. They have to be similar to the extent that there would be no clinically significant difference in safety and clinical performance. (MDR Annex XIV s. 3)

(MDR Annex XIV s. 3)

Annex XIV also requires a clear demonstration that the manufacturer has sufficient levels of access to the data on the device it claims equivalence to. MDCG 2020-5 states that without sufficient access, equivalence claims cannot be made for conformity assessment. Our reading is that access should be checked first, because comparing the characteristics is wasted effort if the manufacturer cannot obtain sufficient access. (MDR Annex XIV s. 3; MDCG 2020-5, pp. 11 and 12)

MDCG 2023-7 interprets sufficient access as access to the data needed to establish the three groups of characteristics, and does not require access to the complete technical documentation. It expects the manufacturer to justify its level of access in the CER, and the notified body has to accept that justification. (MDCG 2023-7, pp. 7 and 8)

MDCG 2020-5 adds four points on how the comparison is made. (MDCG 2020-5, pp. 6 to 13)

• Some characteristics have to be the same, and the MEDDEV exceptions allowing materials that are not the same are not acceptable under the MDR.
• Software algorithms have to be similar in functional principle, clinical performance and intended purpose. Equivalent code is not demanded, provided it was developed in line with international standards for safe design and validation of medical device software.
• A manufacturer may name more than one equivalent device, but may not combine parts of different devices. The guidance allows an exceptional deviation from that rule for a device within a system of stand-alone devices.
• The comparison should emphasise differences, and consider the additive effect of many small ones.

For devices that are neither implantable nor class III, MDCG 2020-5 states that no contract is required with another manufacturer, though sufficient access is. It allows equivalence to a device certified under the directives or the MDR and, exceptionally, to one that is not CE marked. (MDCG 2020-5, p. 13)

4. When does an implantable or class III device need a clinical investigation?#

An implantable device is intended, by clinical intervention, to be totally introduced into the body or to replace an epithelial surface or the surface of the eye, and to remain after the procedure. A device partially introduced and intended to remain for at least 30 days is also deemed implantable. (MDR Art. 2(5))

Implantable and class III devices need clinical investigations, subject to the exceptions in Article 61. MDCG 2023-7 numbers four exemption cases and states that they are independent: the criteria of one do not apply to another unless directly referenced. Where Article 61(4) makes an investigation mandatory, the guidance interprets the requirement as calling for at least a pivotal investigation, which is the confirmatory stage of the clinical development plan. (MDR Art. 61(4); Annex XIV, Part A s. 1(a); MDCG 2023-7, pp. 4 to 7)

(MDR Art. 61(4) to (7); MDCG 2023-7, pp. 5 to 7)

Common specifications are adopted by Commission implementing acts and can cover clinical evaluation. (MDR Art. 9(1))

On Case 1, MDCG 2020-5 adds that the marketed device's CE marking should be valid and rest on an updated clinical evaluation. (MDCG 2020-5, pp. 12 and 13)

For Case 3, Delegated Regulation (EU) 2026/1451 replaced the Article 61(6)(b) list of well-established technologies. It was published on 29 June 2026 and entered into force on the twentieth day after publication. On our count, group (a) of the new list keeps twelve types from the former list, among them sutures, screws, wedges, plates, wires and pins. (Delegated Regulation (EU) 2026/1451, Arts. 1 and 2)

Group (b) adds many further types, among them spinal posterior fixations and guidewires. The consolidated MDR carries the amended list. MDCG 2023-7 predates the amendment and prints only the twelve types that group (a) keeps. (Delegated Regulation (EU) 2026/1451, Art. 1; MDR Art. 61(6)(b); MDCG 2023-7, pp. 5 to 7)

For Case 4, paragraph 5 sets its conditions "in addition to what is required in that paragraph", meaning paragraph 4. MDCG 2023-7 reads this as carrying over the conditions of paragraph 4 into Case 4, except the condition that the device is a modification of the manufacturer's own device, which cannot apply to another manufacturer's device. (MDR Art. 61(5); MDCG 2023-7, pp. 5 to 7)

MDCG 2020-5 reads the Case 4 condition that the original clinical evaluation complies with the MDR as meaning that the equivalent device is certified under the MDR. On that reading, a comparator certified under the directives cannot be used for Case 4. Team-NB asks for the signed contract and evidence of MDR certification. (MDCG 2020-5, pp. 12 and 13; Team-NB BPG V4, p. 73)

For Cases 2 and 3, the guidance asks for more than the MDR does. MDCG 2020-7 states that for implantable and class III devices without clinical investigations under Article 61(4), the PMCF plan shall include post-market studies, and the MDCG 2020-13 template asks the same. Our reading is that post-market studies should be budgeted for whichever of the four cases applies. (MDCG 2020-7, pp. 7 and 8; MDCG 2020-13, p. 20)

The exemptions concern the requirement to perform clinical investigations, and Article 10(3) still requires a clinical evaluation on every route. (MDR Arts. 10(3) and 61(6))

When can non-clinical data be enough?#

Article 61(10) allows an exception where demonstration of conformity based on clinical data is not deemed appropriate. The justification rests on risk management, the device's interaction with the body, its intended clinical performance and the claims. The manufacturer then substantiates in the technical documentation why non-clinical testing alone is adequate. (MDR Art. 61(10))

MDCG 2020-6 states that Article 61(10) cannot be applied to class III or implantable devices, and for other classes only in exceptional cases. Team-NB calls it an exceptional pathway that changes the type of data, and leaves the purpose of the evaluation unchanged. (MDCG 2020-6, pp. 8 and 10; Team-NB Article 61(10) paper V1, pp. 1, 2 and 4)

5. What does a clinical investigation involve?#

A clinical investigation is a systematic investigation involving one or more human subjects, undertaken to assess the safety or performance of a device. Where it is part of the clinical evaluation for conformity assessment, Articles 62 to 80 and Annex XV apply. (MDR Art. 2(45), 62(1))

The sponsor is the individual, company, institution or organisation responsible for initiating the investigation and for managing and setting up its financing. A sponsor not established in the Union needs a legal representative established in the Union. Where the investigation runs only in one Member State, or there and outside the Union, that Member State may accept a contact person on its territory instead. (MDR Art. 2(49), 62(2))

Under Article 62(3), the subjects' rights, safety, dignity and well-being prevail over all other interests, and the data have to be scientifically valid, reliable and robust. An ethics committee under national law, including at least one lay person, performs the ethical review. (MDR Art. 62(3))

The sponsor applies to each Member State where the investigation is to run, with the Annex XV documents, through the electronic system in Article 73. Article 33 makes that system part of the European database on medical devices (EUDAMED). (MDR Art. 33(1), (2)(e), 70(1))

While the clinical investigation module of EUDAMED is not fully functional, MDCG 2021-8 offers facilitative templates, and says sponsors should check each Member State's own requirements. The module's current status is on the Commission's EUDAMED overview page. (MDCG 2021-8, p. 3; Commission EUDAMED overview)

(MDR Art. 70(7))

Annex XV requires a plan reflecting the latest scientific and technical knowledge, defined to confirm or refute the manufacturer's claims, with enough observations to guarantee valid conclusions. The endpoints address intended purpose, benefits, performance and safety, and the primary endpoint has to be clinically relevant. (MDR Annex XV, Chapter I s. 1, 2.1, 2.5 to 2.7)

EN ISO 14155:2020 with its amendment A11:2024 is the harmonised standard cited for clinical investigations under the MDR. Conformity with a cited standard gives a presumption of conformity with the MDR requirements it covers. (Implementing Decision (EU) 2021/1182, Annex, entry 39; MDR Art. 8(1))

MDCG 2024-3 calls ISO 14155:2020 strongly recommended but not mandatory, and says the MDR prevails where the two conflict. Current citations are on the Commission's harmonised standards page. (MDCG 2024-3, p. 5; Commission harmonised standards page)

Whatever the outcome, the sponsor submits a clinical investigation report within one year of the end, or three months of an early termination or temporary halt. A summary the intended user can understand goes with it. Where scientific reasons prevent the one-year deadline, the report is submitted as soon as it is available, and the investigation plan says when, with a justification. The report includes any negative findings. (MDR Art. 77(5); Annex XV, Chapter I s. 2.8)

Health data are a special category under the General Data Protection Regulation (GDPR), processed only under an Article 9(2) exception. Member States may maintain or introduce further conditions on them, so on our reading the conditions differ by country. (GDPR Art. 9(1), (2), (4))

Who can advise before the investigation starts?#

(MDR Art. 61(2); Regulation (EU) 2021/2282, Art. 3(1), 16(1) to (3); MDCG 2019-6 rev.5, pp. 4 to 6)

The European Medicines Agency (EMA) provides the panels' secretariat. Its guide limits applications to manufacturers established in the Union or their authorised representatives. (EMA guide EMA/23357/2025, s. 1.1)

The questions put to the panel have to concern clinical aspects only, and investigations that have not yet started. The final briefing document can be submitted only on EMA's published timetable. The guide counts the start of the procedure as day 1 and delivers the advice on day 60. EMA publishes the current timetable with its yearly dates. (EMA guide EMA/23357/2025, s. 1.1, 1.3, 2; EMA advice page; EMA 2027 timetable)

Joint clinical assessment under Regulation (EU) 2021/2282 reaches selected class IIb and III devices that had an expert panel opinion under Article 54, and selected class D IVDs that had the panels' views. For a device, the joint scientific consultation may run in parallel with the expert panel. Chapter 12 covers eligibility and timing. (Regulation (EU) 2021/2282, Arts. 7(1)(c), (d), 16(5) and 17)

6. How is the evaluation kept current after launch?#

PMCF is a continuous process that updates the clinical evaluation, addressed in the post-market surveillance plan. The manufacturer proactively collects and evaluates clinical data from use of the CE-marked device as intended. The aim is to confirm safety and performance over the expected lifetime and detect emerging risks. (MDR Annex XIV, Part B s. 5)

The PMCF plan specifies general methods, such as user feedback and literature screening, and specific ones, such as registers or PMCF studies. It gives a rationale, objectives and a justified time schedule. The post-market surveillance plan covers a PMCF plan, or a justification of why PMCF is not applicable. (MDR Annex XIV, Part B s. 6.2; Annex III s. 1(b))

The findings go into a PMCF evaluation report, part of the CER and the technical documentation. Its conclusions are taken into account in the clinical evaluation and in risk management. (MDR Annex XIV, Part B s. 7, 8)

Article 61(11) requires the clinical evaluation to be updated throughout the device's life cycle and, for class III and implantable devices, the PMCF evaluation report to be updated at least annually. (MDR Art. 61(11))

Team-NB notes that the MDR sets no update schedule for the clinical evaluation. It expects an update at least annually for devices with significant risks or not yet well established. (Team-NB BPG V4, p. 74)

Team-NB says specific PMCF is usually required for novel technologies, higher-risk devices, and devices approved on equivalent-device data. (Team-NB BPG V4, pp. 81 and 83)

For implantable and class III devices, other than custom-made or investigational devices, the manufacturer also draws up a summary of safety and clinical performance (SSCP). The notified body validates it, and it is made public through EUDAMED. MDCG 2019-9 rev.1 expects it to be objective and balanced, covering favourable, unfavourable and inconclusive data. (MDR Art. 32(1); MDCG 2019-9 rev.1, pp. 6 and 16)

7. What does an IVD's performance evaluation require?#

Performance evaluation is the assessment of data to establish or verify scientific validity, analytical performance and, where applicable, clinical performance. It has to demonstrate all three, and the data and conclusions are the device's clinical evidence. (IVDR Art. 2(44), 56(3))

(IVDR Art. 2(38), 2(40), 2(41), 56(4); Annex XIII s. 1.2.1 to 1.2.3)

The IVDR draws no class distinction in requiring clinical performance studies. (IVDR Art. 56(4))

The manufacturer establishes a performance evaluation plan (PEP) which, as a general rule, has at least thirteen elements, on our count. The plan justifies any element it omits. (IVDR Annex XIII, Part A s. 1.1)

MDCG 2022-2, the guidance on IVD clinical evidence, asks for a search protocol written before searching, as part of the PEP. (MDCG 2022-2, p. 21)

The scientific validity, analytical performance and clinical performance reports, together with their assessment, form the performance evaluation report (PER), part of the technical documentation. For class C and D devices the PER is updated when necessary and at least once a year. (IVDR Annex XIII s. 1.3.1, 1.3.2; Art. 56(6))

PMPF updates the performance evaluation, and its plan can use registers and post-market studies, among other methods. Class C and D devices also need a summary of safety and performance (SSP). For a class D device where no common specifications are available and the device is the first of its type to be certified, the notified body consults the expert panel on the PER. (IVDR Annex XIII, Part B s. 4, 5.2; Art. 29(1), 48(6))

Which performance studies need authorisation?#

A performance study is undertaken to establish or confirm analytical or clinical performance. Article 57 requires performance studies, including those using left-over samples, to comply with data protection law. MDCG 2025-5 reads Article 57 as applying to every study, whoever the sponsor. (IVDR Art. 2(42), 57; MDCG 2025-5, pp. 9 and 10)

(IVDR Art. 2(46), 58(1); MDCG 2025-5, p. 18)

MDCG 2025-5 defines left-over samples as remnants of specimens collected for other purposes, after all intended analyses, that would otherwise be discarded. EN ISO 20916:2024 is the cited standard for clinical performance studies. (MDCG 2025-5, p. 21; Implementing Decision (EU) 2021/1195, Annex, entry 15)

How is software evidenced?#

Software sits across both regulations. MDCG 2020-1 names three components of clinical evidence for medical device software: valid clinical association, technical performance and clinical performance. It maps them to scientific validity and analytical performance under the IVDR. It expects the manufacturer to consider clinical performance validation at each new software release, or to justify not doing so. (MDCG 2020-1, pp. 10, 11 and 13)

8. What happens to unfavourable evidence?#

Both Regulations require favourable and unfavourable data to be considered and included in the technical documentation. Investigation and study reports include negative findings. (MDR Annex XIV s. 2, 4; Annex XV, Chapter I s. 2.8; IVDR Annex XIII s. 1, 1.3.3, 2.3.3)

MEDDEV 2.7/1 rev.4, the 2016 guidance on medical devices (MEDDEV) written for the directives and legally not binding, says a literature search should identify all relevant favourable and unfavourable data. It notes that negative results may appear outside high-impact journals, and MDCG 2020-6 lists both passages as still relevant under the MDR. (MEDDEV 2.7/1 rev.4, pp. 1, 19 and 36; MDCG 2020-6, p. 18)

MDCG 2020-6 is written for devices previously certified under the directives. For those devices, where the evidence does not support the intended purpose, it states that manufacturers shall narrow it until the evidence does. (MDCG 2020-6, p. 16)

Class III implantable devices, and class IIb active devices that administer or remove a medicinal product, go through the clinical evaluation consultation procedure. Where the expert panel finds their clinical evidence insufficient, the notified body shall, if necessary, advise restrictions, such as on the intended purpose, and specific PMCF studies. (MDR Annex IX s. 5.1(g))

9. What would the Commission's proposal change?#

The Commission's proposal COM(2025) 1023, published on 16 December 2025, would amend the MDR and the IVDR. It would remove the Article 61(5) contract requirement and replace the Article 61(6)(b) list with a definition of well-established technology devices. It would also let PMCF findings go directly into the updated CER, without a separate PMCF evaluation report. (COM(2025) 1023, recitals (39) and (40), Art. 1, point (52))

The proposal is not law, and its progress is on the European Parliament's Legislative Observatory. (Legislative Observatory, 2025/0404(COD))

10. How the answer is reached#

The plan comes before any study, because Annex XIV requires new clinical data to be generated as the clinical development plan sets out. The steps run in this order. (MDR Annex XIV, Part A s. 1(d))

1. The intended purpose and class come first, worded exactly as in the analysis that determined the device's qualification and class. Whether the device is implantable or class III decides whether Article 61(4) applies. (MDR Art. 2(12), 61(4); IVDR Art. 2(12))
2. The requirements needing data are identified: each GSPR that needs clinical data, or each Annex I requirement of an IVD. (MDR Annex XIV, Part A s. 1(a); IVDR Annex XIII s. 1.1)
3. Each claimed benefit gets an outcome parameter and an acceptance criterion. Our reading is that a claimed benefit with no outcome parameter is either given one that can be measured or removed from the claims. (MDR Annex XIV, Part A s. 1(a))
4. The literature search covers the state of the art and unfavourable data, and identifies the gaps. (MDCG 2020-13, pp. 14 and 16; MEDDEV 2.7/1 rev.4, p. 19; MDCG 2022-2, p. 21)
5. The route is chosen. For an implantable or class III device each of the four exemption cases is tested, and for an equivalence claim access to the other device's data is checked first. (MDCG 2023-7, pp. 5, 7 and 8)
6. New studies are designed to support the claims, since pilot data are in general insufficient, and the data protection conditions are checked in each country. (MDR Annex XV, Chapter I s. 2.1, 2.5; MDCG 2021-6 rev.1, p. 11; GDPR Art. 9(4))
7. Advice is sought before the protocol is fixed, from the expert panel and through a joint scientific consultation where either is open to the device. A class D IVD has no Article 61(2) route, but a joint scientific consultation can be open to it if it is likely to face joint clinical assessment. (MDR Art. 61(2); Regulation (EU) 2021/2282, Arts. 7(1)(d), 16(2) and 17; EMA advice page)
8. The follow-up plan is written beside the evaluation plan, with the device's expected lifetime stated as a figure and the events that trigger an update of the report. (MDR Annex XIV, Part B s. 6.2, 8; Annex III s. 1(b); Team-NB lifetime paper V1, p. 11; Team-NB BPG V4, p. 74)
9. The report includes unfavourable data and justifies any exemption. Our reading is that a claim the evidence does not support is narrowed, as MDCG 2020-6 states for devices certified under the directives. (MDR Annex XIV s. 4; Art. 61(7); Team-NB BPG V4, pp. 77 and 78; MDCG 2020-6, p. 16)
10. The draft report is tested against the questions the notified body's assessor will ask. Our reading is that answering the MDCG 2020-13 template before submission finds gaps while the plan can still change. (MDCG 2020-13, p. 3)

11. The four running cases#

The monitor: a wearable cardiac monitor from a US company#

The monitor records heart rhythm continuously for later review by a clinician, and its companion application displays the recording. On our reading the hardware is class IIa under Rule 10. The application is classified separately: class IIa if it only records, class IIb if it analyses the rhythm to guide a physician's diagnosis. The planning assumption is that it records only. The monitor is already cleared and selling in the United States.

As a further illustrative assumption, the company holds a completed US study of the monitor. Its results are clinical data, and the evaluation takes into account whether an investigation was performed under Articles 62 to 80. On our reading, the CER states where the US protocol differs from Annex XV, and whether its population transfers to Europe. (MDR Art. 2(48), 61(3)(b))

The triage tool: AI software from a German company#

The triage tool suggests how soon each patient should be seen. Triage nurses in adult urgent care centres use it for patients they have already assessed as having no life-threatening condition, and the nurse confirms or changes the suggestion. Classes IIa to III are all arguable under Rule 11, and class IIb is the planning assumption.

Rule 11 puts the tool in class III if a triage decision based on its output may cause death or an irreversible deterioration of health. Our reading is that this class question is settled before the protocol is fixed, because class III changes the investigation, the advice and the summary required. (MDR Annex VIII, Rule 11)

MDCG 2020-1 says a prospective study may be required where software output affects patient management decisions. Our reading is that a priority suggestion the nurse acts on affects patient management decisions in this sense, and that a prospective study is planned unless the notified body accepts otherwise. (MDCG 2020-1, pp. 14 and 15)

Our reading is that the tool falls outside the Article 54 consultation procedure at class IIa, IIb or III. That procedure covers class III implantable devices and class IIb active devices that administer or remove a medicinal product. As a result, joint clinical assessment, and with it a joint scientific consultation, is unlikely to be open to the tool. On our reading, one PMCF objective should be to compare the tool's performance in routine use with its performance in the validation population. (MDR Art. 54(1); Regulation (EU) 2021/2282, Art. 7(1)(c), (4))

The implant: a spinal implant from a European company with a directive certificate#

The implant was CE marked under Directive 93/42/EEC as a non-active implant and now seeks its first MDR certificate under the Article 120 transition. The cage is class III, and the screws and plates class IIb. Hooks are class IIb on MDCG 2021-24 rev.1's reading, and the class of rods, wires and pins is open, because Rule 8 does not name them.

The Case 2 condition on common specifications applies only where a CS for spinal interbody devices is available, so the CER states whether one is. (MDR Art. 61(6)(a))

MDCG 2020-6 states that complaints and vigilance alone are not sufficient for class III and implantable devices. It expects class III devices certified under the directives that are not well-established technologies to have clinical data at rank 4 of its hierarchy as a minimum. Rank 4 is studies with potential methodological flaws whose data can still be quantified and justified. (MDCG 2020-6, pp. 20 and 21)

The guidance also expects a gap analysis. It says new PMCF studies started under the MDR should not be relied on to bridge gaps such as unsupported indications. (MDCG 2020-6, pp. 13 and 16)

On our reading, if Case 2 fails, a prospective investigation is the remaining route, and it has to be planned against the date until which Article 120(3a) allows the device to remain on sale. Sale continues only while the Article 120(3c) conditions are met, among them no significant change in design and intended purpose. The date is in the consolidated MDR, Article 120(3a). (MDR Art. 120(3a), (3c))

Before any such investigation, the manufacturer may seek expert panel advice on the cage, since it is a class III device. As a class III implant, it goes through the consultation procedure unless a CS on the clinical evaluation of its type applies under Article 54(2)(c). If the panel then gives an opinion, the cage becomes eligible for joint clinical assessment, subject to selection. On our reading a joint scientific consultation may then be open. (MDR Art. 54(2)(c), 61(2); Regulation (EU) 2021/2282, Arts. 7(1)(c) and 16(2))

In that procedure the expert panel has 21 days to say whether it will give an opinion on the CEAR, and 60 days to give one. Continued sale under Article 120(3c) required a written agreement with the notified body signed by 26 September 2024. As an illustrative assumption, the company signed one in time. On our reading, because that agreement predates 25 February 2027, Articles 1 to 3 of Implementing Regulation (EU) 2026/977, on quotations, timelines and interruptions, do not apply. (MDR Annex IX s. 5.1(c) to (f); Art. 120(3c)(e); Implementing Regulation (EU) 2026/977, Art. 8(1))

The near-patient test: a cardiac troponin test from a Swiss company#

The near-patient test measures cardiac troponin in blood near the patient in hospital emergency departments, on an analyser the same company supplies. Serial measurement is class C under Rule 3(j). Single measurement is class B under Rule 6 on the guidance's reading, which is contested, so both are planned as class C until a notified body confirms otherwise. The answers below are for serial measurement.

The Swiss company has no Union entity. Through distributors, it already sells the test in Germany, the Netherlands, Belgium and Austria as a legacy device, under a declaration of conformity drawn up under Directive 98/79/EC with no notified body. Chapter 2 sets out the conditions for staying on the market.

For a near-patient device, IVDR Annex I requires the performance obtained in relevant environments, such as emergency units, to be specifically checked where use may affect it. Performance studies are performed, where appropriate, in circumstances similar to normal conditions of use. Our reading is that the clinical performance parameters should be chosen to suit serial measurement, and that a study using only left-over samples would need a second study at the point of care. (IVDR Art. 57; Annex I s. 9.4(b))

12. When specialist help is worth paying for#

• The device is implantable or class III, since each of the four exemption cases has its own conditions. (MDCG 2023-7, p. 5)
• The plan rests on equivalence, because access to the comparator's data and the comparator's certification are argued case by case. (MDR Annex XIV s. 3; MDCG 2020-5, pp. 10 to 13)
• The plan relies on non-clinical data alone, an Article 61(10) route that guidance reads as exceptional and MDCG 2020-6 excludes for class III and implantable devices. (MDCG 2020-6, p. 8; Team-NB Article 61(10) paper V1, pp. 1 to 4)
• An investigation or performance study is being designed, since the application, ethics review and data protection conditions vary by Member State. (MDR Arts. 62(3) and 70; GDPR Art. 9(4))
• The device may be an orphan or breakthrough device. MDCG 2024-10 and MDCG 2025-9 describe when more limited pre-market data may be accepted, and each expects a defined PMCF plan. (MDCG 2024-10, pp. 9 and 10; MDCG 2025-9, pp. 13 and 14)

Conclusion#

The clinical evaluation plan comes first, naming each clinical benefit the company intends to claim and how it will be measured, before any data are gathered. The device's class then decides whether Article 61(4) requires a clinical investigation, a study in people: implantable and class III devices need one unless one of four exemption cases applies.

Every route ends in a clinical or performance evaluation, so an exemption never removes the need for evidence, and follow-up after launch keeps it current, unfavourable findings included. Exempted implantable and class III devices still face post-market studies, which the MDR requires for two exemption cases and guidance expects for the other two. On our reading they should be budgeted for.

The wearable heart monitor from a US company is class IIa, or IIb if its application analyses the rhythm. Being below class III and, on our reading, non-implantable, it needs no investigation under Article 61(4). On our reading its claims, and so its evidence, cover monitor and application together.

The AI triage software from a German company needs no mandatory investigation at its planned class IIb, although on our reading a prospective study, following patients forward, is planned because nurses act on its suggestions. Its class is to be settled before the protocol is fixed, as class III would add a mandatory investigation, access to expert panel advice and a public summary of safety and clinical performance.

For the spinal implant certified under the former directive, the exemption for well-established technologies listed in the MDR covers the screws, plates, wires and pins, which still need sufficient clinical data. The class III cage cannot rely on being a modification of the company's own device, nor on equivalence to another manufacturer's device. Its route to test is the exemption for devices sold under the directive, which turns on whether its clinical data are sufficient. On our reading, if that fails a prospective investigation is the remaining option, planned against the date until which the MDR's transition rules allow sale.

The near-patient cardiac troponin test from a Swiss company needs evidence that cardiac troponin is linked to the condition, that the test measures it correctly, and that results match the patient's state. An interventional performance study, whose results may influence patient care, needs authorisation, and on the guidance's reading so does one that draws blood only for the study. On our reading, a study using only samples left over from routine testing would need a second study at the point of care.

Chapter 2 covers the market sequence and the US evidence, chapter 5 the risk management file the evaluation feeds, and chapter 7 the technical documentation. Chapter 8 covers what a notified body, the body that certifies the device, may discuss before and after an application. Chapter 10 covers what the AI Act adds for the triage tool, chapter 11 post-market surveillance and vigilance, and chapter 12 joint scientific consultation and joint clinical assessment.