Most of the clinical evidence behind a CE-marked device comes from published literature. That makes the search that finds it more important than it looks. If the search is thin, biased or impossible to repeat, everything built on it, including your clinical evaluation report, is weaker.
What a literature search is in this context
A literature search is the structured process of finding, screening and appraising published and registered evidence relevant to your device. It feeds the clinical evaluation, which Article 61 and Annex XIV of the MDR require. It is not a casual keyword search. It is a documented method that another person should be able to repeat and get the same results.
It usually serves three purposes:
- finding clinical data on your own device
- finding data on equivalent devices where you claim equivalence, and
- establishing the state of the art, meaning what current good practice and comparable devices achieve.
Why it matters
The MDR asks for clinical evidence that is sufficient, relevant and honestly appraised. A thorough search is the proof that you looked properly, including for evidence that does not favour your device.
Reviewers generally check for three things. Did you search widely enough? Did you document the method so it can be reproduced? Did you treat all the evidence fairly, rather than keeping what helps and discarding what does not? A search that cannot answer those questions invites findings about the whole clinical evaluation.
Start with a protocol
Write the search plan before you run any search. A good protocol defines:
| Element | What to define |
|---|---|
| Objective | The questions the search must answer, for example safety, performance, state of the art |
| Scope | Device, equivalent devices, indications, patient groups and outcomes |
| Sources | Databases, registries and other sources, with the reason for each |
| Search strings | The exact terms, synonyms and operators for each source |
| Date range | Start and end dates, with the date the search was run |
| Selection criteria | Inclusion and exclusion rules, set in advance |
| Appraisal criteria | How you will judge relevance and quality, and how you will weight results |
A framework such as PICO (population, intervention, comparator, outcome) helps turn a clinical question into search terms. For a reusable dental instrument, for instance, the population might be patients undergoing restorative procedures, and the outcomes might include procedure-related complications and infection.
Where to search
A single database is not enough. Reviewers generally expect several sources, chosen for the device type.
- Bibliographic databases: PubMed/MEDLINE and Embase are the usual core. The Cochrane Library and Web of Science can add systematic reviews and wider coverage.
- Clinical trial registers: ClinicalTrials.gov, the EU registers and the WHO ICTRP, to find ongoing and unpublished studies.
- Safety databases and field safety notices: FDA's MAUDE database and notices from European competent authorities, for adverse events and recalls on similar devices.
- Other sources: Standards, guidelines and professional society guidance, for the state of the art.
Say why you chose each source. A search that uses one database and one set of keywords is difficult to defend.
Building search strings that work
The quality of the string decides what you find. A few principles:
- Use synonyms, alternative spellings and trade names, not only the generic term
- Combine terms with Boolean operators and test the string before committing
- Search for the equivalent device and the state of the art separately from your own device
- Avoid strings so narrow that they miss relevant papers, or so broad that nobody can screen the results
- Save the exact string, the date and the number of results for every source
Ask a second person to review the strings. Another pair of eyes finds missing terms quickly.
Screening and appraisal of literature
Once you have results, you work through them in stages.
- Remove duplicates.
- Screen titles and abstracts against your inclusion and exclusion criteria.
- Screen full texts of the papers that pass.
- Record the reason for every exclusion at full-text stage.
- Extract the data needed: study design, population, outcomes, device, follow-up and findings.
- Appraise each paper for relevance to your device and for methodological quality, then weight it accordingly.
A PRISMA-style flow diagram showing how many records were found, screened and included is the clearest way to display this. Having two reviewers screen independently or at least checking a sample, adds credibility. If you use software to help with screening, document how it works and how you checked it.
Documenting it so it can be repeated
Annex XIV requires the clinical evaluation to be well documented, and the search report is part of that record. It should include the protocol, the final search strings by source, the dates, the numbers at each stage, the excluded papers with reasons, and the appraisal results.
Keeping the search up to date
The search is not finished at certification. Your clinical evaluation plan and post-market surveillance plan should say how often it is repeated. Many Notified Bodies expect regular updates, often annually for higher-risk devices. New studies, new safety signals and shifts in the state of the art can change your benefit-risk conclusion.
When you update, keep the original protocol unless you have a documented reason to change it. Changing methods quietly between searches is a red flag.
Common mistakes
- Searching only one database
- Searching your own device name and nothing else
- Setting inclusion criteria after seeing the results
- Discarding unfavourable papers without recording why
- Failing to record the search date and the exact strings
- Skipping the state-of-the-art search
- Letting the search go stale between report updates
Need a search that holds up?
We design and run systematic literature searches, screen and appraise the results, and document everything so it can withstand review. Tell us about your device and we will outline the approach.
Next blog →Vigilance Reporting under EU MDR: What Every Manufacturer Needs to Know