Literature Search Protocol for EU MDR Clinical Evaluation: Databases, Appraisal, and CER Evidence

Why the Literature Search Protocol Matters Under EU MDR

Under EU MDR 2017/745, clinical evaluation is a continuous, lifecycle-driven obligation. Article 61 and Annex XIV Part A require manufacturers to follow a "defined and methodologically sound procedure" for clinical evaluation — and the literature search protocol is the cornerstone of that procedure. Without a compliant, well-documented search protocol, the entire clinical evaluation report (CER) can be called into question during a Notified Body review.

MEDDEV 2.7/1 Rev. 4, though written for the prior Medical Device Directive, remains the primary technical guidance for literature search methodology under the MDR. MDCG 2020-6 explicitly references MEDDEV 2.7/1 Rev. 4 sections as still valid for clinical evaluation under the regulation. The protocol must be pre-defined, systematic, unbiased, and reproducible — not a retrospective documentation of ad hoc searches.

Regulatory Framework

EU MDR Requirements

The MDR addresses literature search in several provisions:

• Article 61(1): Requires clinical evaluation to be based on clinical data adequate to demonstrate conformity with relevant General Safety and Performance Requirements (GSPRs)
• Annex XIV Part A(1): Specifies that clinical evaluation must follow a defined and methodologically sound procedure, starting with a clinical evaluation plan
• Annex XIV Part A(2): Requires the plan to identify and appraise all relevant clinical data, including from literature

MEDDEV 2.7/1 Rev. 4 Requirements

MEDDEV 2.7/1 Rev. 4 Appendix A5 sets out the elements of a compliant literature search and review protocol. The protocol should:

• Address the objectives of the review
• Specify methods for identification, selection, collection, and appraisal of relevant publications
• Be objective, non-biased, and systematic
• Use established methodologies such as the Cochrane Handbook, PRISMA Statement, or PICO framework

MDCG Guidance Documents

Several MDCG documents provide additional context:

• MDCG 2020-5: Equivalence — requires literature on equivalent device data
• MDCG 2020-6: Sufficient clinical evidence — references MEDDEV 2.7/1 Rev. 4 methodology as still applicable
• MDCG 2020-1: Clinical evaluation of medical device software — adapts literature search principles for SaMD
• MDCG 2024-10: Clinical evaluation of orphan medical devices — allows adjusted literature search scope for rare conditions

Building the Literature Search Protocol

Step 1: Define the Scope and Objectives

Before executing any search, define what you are looking for and why. The scope is set by the Clinical Evaluation Plan (CEP) and typically addresses two distinct categories:

Category A — Device-specific data: Clinical data relating to the subject device or an equivalent device (if equivalence is claimed). This includes safety data, performance data, and adverse event information.

Category B — State of the art (SOTA) data: Current knowledge including applicable standards, guidance documents, data on benchmark devices, medical alternatives, and the specific medical conditions and patient populations the device is intended to manage.

Each category requires its own search strategy because the objectives differ. Device-specific searches aim to find evidence of safety and performance. SOTA searches aim to establish clinical benchmarks and context.

Step 2: Formulate the Research Question Using PICO

The PICO framework is the most widely accepted method for structuring literature search questions under MEDDEV 2.7/1 Rev. 4:

Adaptations include PICOS (adding Study design), PICOT (adding Time frame), and PEO (Population, Exposure, Outcome — useful for SOTA searches).

Step 3: Select Databases

MEDDEV 2.7/1 Rev. 4 emphasizes that relying on a single database is insufficient. PubMed alone does not adequately capture European journal content or studies reporting negative results.

Required databases for most devices:

Supplementary databases (use as applicable):

• Google Scholar — Grey literature, conference proceedings, theses
• ClinicalTrials.gov — Ongoing and completed clinical investigations
• WHO ICTRP — International clinical trial registry
• NIH RePORTER — US-funded research
• ScienceDirect / Web of Science — Broader coverage of engineering and materials science
• National guideline databases (e.g., NICE, AWMF) — Clinical practice guidelines for SOTA

Step 4: Develop Search Strings

Each database requires its own search syntax. Construct search strings using:

Boolean operators:

• AND — narrows results (all terms must appear)
• OR — broadens results (any term may appear)
• NOT — excludes terms (use sparingly to avoid excluding relevant results)

Truncation and wildcards:

• * (asterisk) — captures word variations (e.g., stent* finds stent, stents, stenting)
• ? — single character wildcard

Controlled vocabulary:

• PubMed: MeSH terms
• Embase: Emtree terms

Example search string for PubMed (device-specific):

("drug-eluting balloon" OR "drug-coated balloon" OR "paclitaxel balloon")
AND ("peripheral artery disease" OR "critical limb ischemia" OR "femoropopliteal")
AND ("patency" OR "revascularization" OR "safety")
Filters: Clinical Trial; Meta-Analysis; 2015-2026

Example search string for Embase (SOTA):

('drug coated balloon'/exp OR 'drug eluting balloon') 
AND ('peripheral arterial disease'/exp OR 'peripheral occlusive arterial disease') 
AND [humans]/lim AND [2015-2026]/py

Step 5: Define Inclusion and Exclusion Criteria

Pre-specify criteria to ensure systematic, unbiased selection. Document these in the protocol before reviewing any results.

Typical inclusion criteria:

• Published in peer-reviewed journals or authoritative sources
• Published within a defined time window (justify the period — e.g., last 10 years for SOTA, or since the device entered the market for device-specific data)
• In English or with English abstract available (justify language restrictions)
• Directly relevant to the device, equivalent device, or SOTA topic
• Human clinical data (for device-specific searches)
• Adequate methodological quality for data extraction

Typical exclusion criteria:

• Editorials, letters, or opinion pieces without original data
• Conference abstracts without full-text publications (unless no other data exists)
• Studies with populations outside the intended use
• Duplicate publications of the same study
• Animal or in-vitro studies (unless supporting specific safety claims)
• Case reports with fewer than a defined threshold of patients

Step 6: Execute the Search and Document Results

Run the searches on each selected database and document:

1. Date of search
2. Database name and version
3. Complete search string used
4. Number of results returned
5. Number of articles screened by title/abstract
6. Number of full-text articles reviewed
7. Number of articles included after applying inclusion/exclusion criteria

Use a PRISMA flow diagram to visualize the search and selection process:

Records identified through database searching (n = ___)
                     │
Records after duplicates removed (n = ___)
                     │
Records screened by title/abstract (n = ___)
                     │
        ┌────────────┴────────────┐
Records excluded          Records assessed
(n = ___)                for full-text (n = ___)
                                │
                   ┌────────────┴────────────┐
           Full-text articles         Studies included
           excluded (n = ___)         in the review (n = ___)
           with reasons:
           - Wrong population (n = __)
           - Wrong intervention (n = __)
           - Insufficient quality (n = __)
           - Other (n = __)

Critical Appraisal of Literature

Appraisal Framework

MEDDEV 2.7/1 Rev. 4 requires that all included data be appraised for methodological quality and relevance. Use validated assessment tools:

Appraisal Criteria

For each included publication, evaluate:

Methodological quality:

• Study design appropriateness for the research question
• Sample size and power calculation
• Selection bias (inclusion/exclusion criteria)
• Confounding factors and how they were controlled
• Blinding (single, double, or open-label)
• Duration of follow-up
• Statistical methods and analysis

Relevance to the device:

• Does the study population match the intended patient population?
• Is the device or intervention comparable to the subject device?
• Are the outcomes measured relevant to the claims being made?
• Are the conditions of use consistent with the intended use?

Data quality weighting:

• Level of evidence ranking (e.g., MDCG 2020-6 Appendix III provides a 12-level hierarchy)
• Source credibility and journal impact factor
• Consistency with other published evidence
• Whether findings have been replicated

Documenting the Appraisal

Create a structured appraisal table for each study:

Critical Process Requirement: CEP Before Literature Search

The Clinical Evaluation Plan (CEP) must be written, reviewed, and approved before the literature search begins. Producing a CEP that mirrors the completed CER — clearly written after the fact — is treated by most Notified Bodies as a major nonconformity. The CEP is the methodology declaration: if it does not pre-date the evaluation, it has no methodological value.

The correct sequence is:

1. Write and approve the CEP (including the literature search protocol as an appendix or referenced document)
2. Execute the literature search according to the protocol
3. Appraise and analyze the results
4. Write the CER based on the findings

A CEP that is version-controlled and dated prior to data retrieval is the single most important procedural evidence for demonstrating a systematic approach.

Emerging Trend: Living Systematic Reviews

A growing practice under EU MDR is the adoption of "living systematic reviews" — continuously updated evidence reviews that align with the MDR's lifecycle approach to clinical evaluation. Rather than conducting static, point-in-time literature searches at each CER update, manufacturers maintain an ongoing surveillance process that identifies new evidence as it is published. This approach is particularly useful for high-risk devices requiring annual CER updates and for PMCF literature monitoring.

Common Pitfalls and How to Avoid Them

1. Relying on a Single Database

Using PubMed alone is the most frequently cited deficiency in Notified Body audits. MEDDEV 2.7/1 Rev. 4 explicitly states that studies with negative results or regional clinical outcomes may be underrepresented in PubMed. Embase must be included for adequate European journal coverage.

2. Failing to Document the Search Strategy

The protocol must be documented before the search begins. Notified Bodies expect to see:

• The original search protocol document (dated and version-controlled)
• Screenshots or exported search logs showing the actual queries run
• PRISMA flow diagrams for each search category

3. Using Vague or Overly Broad Search Terms

Search terms that are too broad return thousands of irrelevant results, making the review impractical. Search terms that are too narrow risk missing relevant evidence. Test your search strings iteratively and document refinements.

4. Not Justifying the Search Period

The time window for the literature search must be justified. Common justifications include:

• "Search period covers the time since the device was first placed on the market"
• "Search period covers the last 10 years to capture current SOTA"
• "Extended from previous search date of [date] to present"

5. Ignoring Unfavorable Data

MEDDEV 2.7/1 Rev. 4 requires an unbiased search. Excluding studies that report negative results or adverse events without valid justification is a major nonconformity. Document the reason for excluding every study, not just the ones you include.

6. Insufficient Evaluator Qualifications

MEDDEV 2.7/1 Rev. 4 raised the bar for who can conduct and appraise literature reviews. Evaluators should have:

• A relevant clinical or scientific degree
• Demonstrable expertise in the device technology or clinical area
• Training in systematic review methodology
• Understanding of regulatory requirements

Linking Literature Search to the CER

The literature search protocol outputs feed directly into several CER sections:

1. State of the Art section — Data from SOTA searches establish clinical benchmarks
2. Clinical data analysis — Device-specific data from literature is appraised and synthesized
3. Benefit-risk assessment — Weighted evidence supports conclusions about safety and performance
4. PMCF planning — Gaps identified during literature review inform post-market clinical follow-up activities
5. Data sufficiency assessment — The volume and quality of evidence is evaluated against MDCG 2020-6 criteria

The CER must reference the literature search protocol and report as controlled documents within the quality management system. Update the literature search at each CER update cycle (typically annually for high-risk devices, every 2-5 years for lower-risk devices, per MDR Article 61(11) and Annex XIV Part B).

Literature Search for PMCF Activities

The literature search protocol is not only for the initial CER. Under MDR Annex XIV Part B, PMCF activities must also include literature searches. MDCG 2020-7 (PMCF Plan Template) and MDCG 2020-8 (PMCF Evaluation Report Template) describe specific literature review expectations for PMCF:

• PMCF literature searches focus on new published data about the device, equivalent devices, and SOTA since the last evaluation
• Search frequency should be defined in the PMCF plan (typically annual)
• Scope may be narrower than the initial CER search but must still be systematic
• Findings must be integrated into the CER update

Practical Checklist

Use this checklist to verify your literature search protocol before submission to a Notified Body:

• Protocol written and approved before conducting searches
• Two or more databases used (PubMed + Embase minimum)
• PICO or equivalent framework used for research question formulation
• Search strings documented with Boolean operators, controlled vocabulary, and filters
• Inclusion and exclusion criteria pre-specified and justified
• Search period justified
• PRISMA flow diagram completed for each search category
• All searches are reproducible (documented with dates, databases, exact strings)
• Validated appraisal tools used for critical appraisal
• Appraisal tables completed for all included studies
• Evaluator qualifications documented
• Excluded studies documented with reasons
• Protocol and report are controlled documents in the QMS
• Protocol aligns with the Clinical Evaluation Plan (CEP)
• Findings linked to CER sections and PMCF planning