Literature review for a CER (MDR): the step-by-step method

By Bopha Suybeng, PharmD, founder of NOÉ. Twelve years of experience in the clinical evaluation of medical devices.

In short: under Regulation (EU) 2017/745 (MDR), the literature review in a clinical evaluation report (CER) must be planned in advance, reproducible and traceable. It follows eight steps, spread across stages 0 to 4 of MEDDEV 2.7/1 rev. 4: define the scope, write the protocol, search the databases, screen the articles, appraise each article, extract the data, analyse and synthesise, then write the CER.

Why the literature review is at the heart of the CER

Article 61 and Annex XIV, Part A, of the MDR require the clinical evaluation to follow a defined and methodologically sound procedure. Scientific literature serves two purposes: documenting the performance and safety of the device (or of a demonstrated equivalent device), and establishing the state of the art in the medical field. MEDDEV 2.7/1 rev. 4, still used as the methodological reference, details how to identify, appraise and analyse these data.

Step 1: define the scope

Everything starts from the clinical evaluation plan (CEP): device, claimed indications, target population, expected clinical performance and benefits, risks to document. Clinical questions are framed in the PICO format (population, intervention, comparator, outcomes). A poorly framed question leads to a search that is too broad or, worse, incomplete.

Step 2: write the protocol

The protocol is written before the search. It sets out:

• the databases searched (for example PubMed/MEDLINE, Embase, Cochrane Library) and other sources (trial registries, vigilance databases);

• the search strings, the period covered and the languages;

• the inclusion and exclusion criteria;

• the method used to appraise and weight the data.

Step 3: search the databases

Each search string is run exactly as written in the protocol, with the date, the database and the number of results recorded. Duplicates are then removed. A notified body must be able to rerun the same search and get the same results.

Step 4: screen the articles

Screening happens in two rounds: on title and abstract, then on full text. Each exclusion is justified by a protocol criterion. The PRISMA flow diagram summarises this process, from the number of records identified to the number of articles included.

Step 5: appraise each article

Not all included studies carry the same weight. MEDDEV 2.7/1 rev. 4 requires each article to be appraised for the relevance of its data to the device and indication, and for its methodological quality. Appraisal criteria are defined in advance and applied the same way across the whole corpus.

Step 6: extract the data

Relevant data are recorded in a predefined grid: population, follow-up duration, performance and safety outcomes, adverse events. Each data point stays linked to its source so it can be checked.

Step 7: analyse and synthesise

Data are weighted according to their relevance and quality, then analysed against the state of the art. The analysis covers safety, performance and the benefit-risk ratio. It also highlights gaps to be addressed by post-market clinical follow-up (PMCF).

Step 8: write the CER

The report describes the method, the results, the limitations and well-argued conclusions. Full traceability of the search (search strings, dates, screening decisions) is included or appended, so that a notified body can follow every step.

What notified bodies check

• Consistency between the CEP, the protocol and the report.

• Reproducible searches.

• A justification for every exclusion.

• Data weighting applied consistently.

• Regular updates of the review, linked to post-market clinical follow-up (PMCF).

Where a tool like NOÉ comes in

NOÉ assists the evaluator with the most repetitive steps: applying eligibility criteria across the whole corpus, keeping the justification for every decision, the PRISMA flow diagram and data extraction. The expert stays in control of every decision and of the clinical analysis.