Many medical device manufacturers pursuing EU CE-MDR certification get stuck at clinical evaluation (CEP / CER), facing repeated questions and multi-round re-assessments. Even with early submission, certification timelines keep stretching. One project went through 4 review rounds over 2 years, with 30 non-conformities initially identified. After the first re-submission, 18 non-conformities remained, and 60% of the raised issues were not genuinely closed after responses.
This open class at Beijing CMEF breaks down Notified Body review requirements and countermeasures for EU CE clinical evaluation, combining real-world case studies, for enterprises planning to enter the EU market.
一、High-frequency Pitfalls Identified by Notified BodiesBased on de-identified review records from multiple Notified Bodies, five categories of recurring issues are summarised:(一)Confusing technical performance with clinical benefitThis is the most frequent error.Wrong approach:Clinical benefit = compliance with ISO standards, low connection failure rate, qualified bonding strength.MDR requirement:Clinical benefit must be realised on patients, representing positive outcomes meaningful to patients or patient management, rather than factory test results or hardware technical parameters.Classic reviewer feedback: "This describes technical performance rather than measurable patient clinical endpoints."Numerous projects have been flagged for vague, general descriptions of clinical benefit without measurable indicators or supporting data.Practical tip: Every statement in the intended purpose creates corresponding evidence obligations.
Sort out the logical chain: Intended purpose → Indications → Target population → Use scenario → Clinical benefit → Clinical performance → Safety → External claims.
Do not substitute clinical benefit with design or physical-chemical test results.
(二)SOTA(State-of-the-Art) written as a general literature reviewMany manufacturers treat SOTA as an introduction to disease background or technology development history, which is a critical misunderstanding.SOTA is not a popular science summary. It serves as the benchmark for evaluating your device, addressing three core questions::- What is the current clinical practice? Treatment regimens, alternative devices, clinical guidelines
- What outcomes are achievable? Quantified performance, benefits, complications and adverse events with cited literature sources
- What benchmarks must our device meet? Deliver actionable safety & performance criteria
- Only comparing against one reference product, ignoring alternative therapies and unmet clinical needs
- No valid justification for comparator selection; mixing SOTA literature retrieval and equivalent device retrieval
- Conflicting conclusions between cited literature and CER claims; simply deleting unfavourable literature to avoid discussion
Key takeaway:Deleting conflicting literature does not resolve issues. Conflicting or negative publications must be analysed and discussed explicitly within the document, instead of being removed.Qualified SOTA deliverable:A benchmark comparison table with figures and references. It is recommended to build search strategies using the PICO framework, and fully retain search protocols, keywords, databases, inclusion/exclusion criteria and search screenshots so reviewers can reproduce your search process.(三)Postponing clinical acceptance criteria; substituting clinical objectives with design requirements"The performance targets specified, such as bonding strength, bending resistance and sealing performance, are design requirements, not clinical performance targets derived from SOTA."— Notified Body Review CommentNumerous review comments converge on the same problem:- Acceptance criteria without numerical values, vague wording such as "no worse than similar devices"
- Criteria sourced from internal factory inspection, not derived from SOTA clinical data
- Criteria only explained in response letters to Notified Bodies, without being recorded in controlled CEP / CER files
Reviewers assess controlled documents, not your response emails.
- Extract all external claims from instructions for use and intended purpose statements one by one
- Match each claim to corresponding clinical performance / safety parameters
- Trace parameter values back to SOTA benchmarks, with figures, units and calculation logic clearly stated
- Define supporting evidence types (literature / clinical investigation / PMCF)
- Embed all content into the main body of CEP and CER, not only within response documents
It is recommended to build a traceability matrix:Claim — Parameter — SOTA Source — Acceptance Criteria — Evidence Source, to expose gaps at a glance.(四)Confusing similar devices with equivalent devices; equivalence assessment failsUnder MDR, similar ≠ equivalent! Only after a complete equivalence assessment (clinical, technical and biological comparability all satisfied), can data from an equivalent device be used for your clinical evaluation. Similar appearance and overlapping intended use alone do not permit direct reuse of third-party clinical data.Many projects cite extensive literature from D3 (similar devices) and treat them as D2 (equivalent device) evidence, which gets rejected directly by reviewers, requiring a full change of clinical evaluation strategy.
(五)Superficial literature appraisal, only selecting favourable evidenceLiterature appraisal cannot be limited to stacking references. Each publication requires three-dimensional assessment:Quality × Relevance × Contribution- Quality: Scientific validity of the study, sample size, endpoints, bias, statistical methodology
- Relevance: Alignment with the device’s indications, target population and duration of use
- Contribution: Whether it serves as key evidence or only supplementary support
Even for self-cited literature, negative findings cannot be ignored. After appraisal, clear conclusions must be formed: Pass / Fail / Insufficient evidence. For evidence gaps, propose solutions: supplementary literature, clinical investigation, PMCF plan, or narrowing the product claims.
Do not include functions with insufficient evidence in the intended purpose. Restrict your claims to the boundary covered by available evidence.