Which clinical evidence counts

Clinical evidence for a medical device can come from more sources than the manufacturer's own investigation. Which sources count, how they rank against each other and what they have to establish decide whether an evidence route is sufficient for a given device.

What clinical evidence has to establish

Clinical data are information on safety or performance generated from the use of a device. ¹ They come from clinical investigations of the device or of an equivalent device, from published peer-reviewed literature on either, or from clinically relevant post-market surveillance information, especially post-market clinical follow-up. ¹ They are needed to confirm compliance with the general safety and performance requirements of MDR Annex I and to evaluate undesirable side effects and the acceptability of the benefit-risk ratio. ¹ The clinical evaluation includes a thorough and objective assessment of favourable and unfavourable clinical data alike, which forms the clinical evidence for the device. ¹

The MDR defines clinical evidence as the clinical data and clinical evaluation results pertaining to a device of a sufficient amount and quality to allow a qualified assessment of whether the device is safe and achieves its intended clinical benefits when used as the manufacturer intends. ¹ Its Article 61 bases conformity with the relevant general safety and performance requirements on sufficient clinical evidence, but the MDR does not define sufficient. ¹ Sufficient clinical evidence is understood as the present result of the qualified assessment that has concluded the device is safe and achieves its intended benefits, and clinical evaluation is a process in which that assessment has to be made continuously. ¹ MDCG 2020-6, the Medical Device Coordination Group's guidance endorsed in accordance with Article 105 of the MDR, sets out the clinical data a legacy device needs to demonstrate conformity with the MDR, a legacy device being one placed on the market under the Medical Device Directive or the Active Implantable Medical Devices Directive before the MDR came into force. ¹

In one practitioner's view, writing for medical devices is straightforward even where the device carries an ancillary drug, because the same things have to be shown whether the device stands alone or is part of a combination: sufficient evidence that the device is safe, that it does what it is supposed to do, that its benefit outweighs its risk, and that it is at least as good as the state of the art. ^{2, 3}

Where clinical evidence comes from

MDCG 2020-6 holds that it is important to identify all available sources of clinical data from both the pre-market and the post-market phases, including all the clinical data the manufacturer generates and holds as well as clinical data for equivalent or similar devices. ⁴ In March 2026, one speaker pictured clinical evidence as a hub fed by five main sources: clinical investigation, scientific literature, bench testing, in vitro or ex vivo studies, and computational modelling where the work involves artificial intelligence. ⁵ Clinical investigation gives the gold-standard primary data, both before and after market entry. ⁵

Published literature

The published literature plays a very important role in clinical evaluation alongside clinical investigations, and data from the literature on other, similar products matter as well as data on the device itself, which is why the systematic review of the literature is very important for medical devices. ⁶ As of 2017, unlike in pharmaceutical development, few clinical investigations were conducted during the development of a medical device, so the published literature was an important source of clinical data for equivalent devices during CE marking and renewal, and for the subject device itself during CE renewal. ⁷

The manufacturer's own data

As of 2017, the clinical data the manufacturer generated and held included data from pre-clinical studies such as bench testing, from pre-market clinical investigations, and from risk management and post-market surveillance. ^{8, 9} All of them should be made available, not just the data generated in Europe, and should be summarised, appraised, analysed and referenced in the clinical evaluation report. ⁸

Examples of risk management and post-market surveillance data are: ¹⁰

• post-market clinical follow-up studies and device registries sponsored by the manufacturer
• post-market surveillance reports, including vigilance reports and trend reports
• the literature search and evaluation reports for post-market surveillance
• incident reports sent to the manufacturer
• complaints regarding performance and safety sent to the manufacturer
• analysis of explanted devices, as far as available
• details of all field safety corrective actions
• use as a custom-made device
• use under compassionate-use or humanitarian-exemption programmes
• other user reports

^{7, 10}

Risk management and post-market surveillance reports are usually large documents containing spreadsheets of quality control reports, complaints and sales figures, and information from external national databases such as MAUDE in the United States and the MHRA's device alerts in the United Kingdom. ⁸ Obtaining them means liaising with various groups within the manufacturing company so that the reports are available in time for the clinical evaluation report and its submission timelines, especially for a CE-mark renewal with a specific timeline. ⁸

The hierarchy of clinical evidence

MDCG 2020-6 sets out in its Appendix III a hierarchy of clinical evidence in twelve levels, ranked roughly from strongest to weakest, with variations depending on the device and on the quality of each data source. ^{1, 11, 12} The levels are: ¹³

1. Results of high-quality clinical investigations covering all device variants, indications, patient populations, duration of treatment effect and so on.
2. Results of high-quality clinical investigations with some gaps.
3. Outcomes from high-quality clinical data collection systems such as registries.
4. Outcomes from studies with potential methodological flaws but where data can still be quantified and acceptability justified.
5. Equivalence data, reliable and quantifiable.
6. Evaluation of the state of the art, including evaluation of clinical data from similar devices.
7. Complaints and vigilance data, and curated data.
8. Proactive post-market surveillance data, such as that derived from surveys.
9. Individual case reports on the subject device.
10. Compliance with the non-clinical elements of common specifications considered relevant to device safety and performance.
11. Simulated use, animal or cadaveric testing involving healthcare professionals or other end users.
12. Pre-clinical and bench testing, and compliance with standards.

^{13, 1, 11}

Similar devices can provide "soft" data, ranking roughly with the clinical evidence provided by the state of the art and below the clinical data from the manufacturer's own or equivalent devices. ⁴

The level of evidence a device needs

The level of clinical evidence required depends on the device class, its risk profile and its marketing history. ¹³ Class III legacy devices and implantable legacy devices are not well-established technologies and should have at least level 4 clinical data. ¹

A well-established technology has to meet these criteria: ¹

• a relatively simple, common and stable design with little evolution
• a generic device group known to be safe and not associated with safety issues in the past
• well-known clinical performance characteristics, with a generic device group of standard-of-care devices showing little evolution in indications and in the state of the art
• a long history on the market

¹

Screws and plates are typical examples of well-established technologies that the MDR mentions. ¹⁴

A well-established technology may be able to confirm conformity with the general safety and performance requirements from cumulative evidence at levels 5 to 12, though it cannot rely on complaints and vigilance data alone. ¹

In one practitioner's account, class III devices and implantable legacy devices must have evidence of at least level 4, whereas a legacy device that is a well-established technology may be able to justify a lower level of evidence, such as level 8 surveys. ¹⁵

As of 2019, for innovative high-risk devices, CE certification without data from clinical studies was nearly impossible. ¹⁶

In 2024, one practitioner reported that notified bodies were also providing some guidance on the minimum level of clinical evidence for certain devices. ^{12, 17}