Why this matters

If you make medical devices and want to sell them in the EU, you cannot skip the Clinical Evaluation Report (CER). The CER is one of the documents Notified Bodies look at most closely, and the literature search sits at the heart of it. Get this part wrong and you can end up with deficiencies that delay your CE mark by months. Get it right and you have a defensible, auditable record that supports your safety and performance claims.

This guide walks through how to plan, run, and document a literature search that will actually hold up under review. It is written for regulatory and clinical affairs professionals who are either preparing their first CER or trying to improve a recurring one.

What clinical evaluation actually is

Clinical evaluation is a continuous process. You collect clinical data, appraise it, analyze it, and use it to demonstrate that your device is safe and performs as intended throughout its lifecycle (GHTF Study Group 5, 2007; MDR Annex XIV).

The literature review is one of three main sources of clinical data, alongside clinical investigations and post-market clinical follow-up (PMCF). For most legacy or low-to-medium risk devices, published literature carries a lot of the weight (MEDDEV 2.7/1 rev. 4, 2016).

A search that does not follow a structured methodology will not survive a Notified Body review. The criteria you need to meet are reproducibility, traceability, and scientific validity (Pati & Lorusso, 2017).

Step 1: Define the clinical question

You cannot write a good search strategy until you know exactly what question you are asking. Many CERs go wrong here because the question is either too broad (“is this catheter safe?”) or too vague to operationalize.

The PICO framework helps. PICO stands for Population, Intervention, Comparator, and Outcome (Ecker & Skelly, 2010). For a clinical evaluation, you usually need at least two PICO questions: one focused on your device or an equivalent device, and one focused on the state of the art, meaning the broader clinical context, the standard of care, and alternative therapies.

A worked example. Suppose you are evaluating a balloon-expandable peripheral stent for use in the superficial femoral artery:

• P: adult patients with symptomatic peripheral arterial disease (Rutherford 2 to 4) of the superficial femoral artery
• I: balloon-expandable nitinol stent
• C: drug-coated balloon angioplasty
• O: primary patency at 12 months, target lesion revascularization, major adverse limb events

That question is searchable. It also tells you which comparator literature you need to pull, which is the area where reviewers most often find gaps.

Step 2: Build a search strategy you can defend

A good search strategy is something a second person could run and get the same results from. That means you need to write down, before you search:

• the exact keywords and their synonyms, including MeSH terms where applicable
• the Boolean logic linking them
• the databases and registries you will query
• the date range and any language restrictions
• your inclusion and exclusion criteria
• the number of reviewers and how disagreements will be resolved

On databases, PubMed alone is not enough. MEDDEV 2.7/1 rev. 4 expects multiple sources. In practice, most CERs use a combination of PubMed/MEDLINE, Embase, and the Cochrane Library. For some device classes, you should also search:

• gov and the EU Clinical Trials Register, for ongoing or unpublished studies
• FDA MAUDE and EUDAMED, for adverse event signals
• national regulator databases such as TGA DAEN or Health Canada MDPL, for international post-market data
• conference proceedings and grey literature for emerging evidence

Keep an audit trail. Save the raw search strings, the date of each search, the number of hits, and the filters applied. Notified Bodies routinely ask to see them. If you cannot reproduce the search, the rest of the CER becomes harder to defend.

Step 3: Identify which clinical data to include

Clinical data can come from clinical investigations, scientific literature, post-market data, or equivalent devices (GHTF Study Group 5, 2007; MDR Article 61).

Each source has its strengths and limits:

• Randomized controlled trials are the strongest individual studies but are rare for many device categories.
• Systematic reviews and meta-analyses integrate findings across studies and are generally preferred over single trials (Pati & Lorusso, 2017).
• Registry data and real-world evidence are increasingly accepted, especially for long-term safety signals (Valla et al., 2023).
• Case series and case reports are weak on their own but useful for adverse event detection.

If you are relying on equivalence to another device, the literature for that device counts, but only if you can actually demonstrate equivalence on technical, biological, and clinical characteristics in line with MDR Annex XIV Part A. The bar is higher under MDR than it was under MDD. Many manufacturers underestimate this.

Step 4: Appraise the evidence

Finding studies is only the start. Each one has to be appraised for methodological quality, relevance, and risk of bias.

There is no single mandated appraisal tool, but several are widely accepted:

• Oxford Centre for Evidence-Based Medicine levels of evidence, for quick ranking
• GRADE, for assessing the certainty of evidence across a body of literature
• AMSTAR 2, for systematic reviews
• Cochrane RoB 2, for randomized controlled trials
• ROBINS-I, for non-randomized studies of interventions

What matters most is consistency. Use the same appraisal criteria across all included studies, document the rationale for each judgment, and have a second reviewer agree.

Common reasons studies get downgraded:

• small sample sizes without power calculations
• surrogate endpoints rather than clinically meaningful outcomes
• selection bias, particularly in retrospective registry analyses
• industry funding without independent oversight
• outcomes that do not match your device’s intended purpose or claims

Step 5: Analyze and synthesize the data

After appraisal, you put the body of evidence together and ask whether it actually supports the claims you make in your Instructions for Use, your marketing, and your risk management documentation.

The synthesis section of the CER is where the benefit-risk balance is established. You need to:

• compare the clinical performance reported in the literature against your performance claims
• identify residual risks that the literature does not adequately address
• flag any signals of harm, especially device-specific ones
• compare your device’s results against current standard of care, not against placebo

If the literature does not cover something, say so. A weakness identified and addressed, for example through a planned PMCF activity, is much easier to defend than a weakness left unmentioned.

The conclusions go into the CER, which sits inside your technical documentation under MDR Annex II.

Step 6: Keep the review alive

The MDR treats clinical evaluation as a lifecycle obligation, not a one-off submission. The literature search has to be updated regularly. The frequency depends on the device’s risk class and on what you committed to in your PMCF plan, but for higher-risk devices, annual updates are typical, and for implantable or Class III devices, more frequent updates are often expected.

Things that should trigger an earlier update:

• a safety signal from your post-market surveillance system
• a new publication on a similar device or technique that changes the state of the art
• a new clinical guideline from a relevant medical society
• a change in your device’s intended purpose, claims, or design
• a regulatory action (recall, field safety notice) on a similar device

Real-world evidence, registry data, and patient-reported outcomes are increasingly being used to supplement traditional literature, and Notified Bodies are paying more attention to how manufacturers integrate these (Valla et al., 2023).

Common deficiencies Notified Bodies cite

If you read enough Notified Body opinions, the same problems show up over and over:

• Search strings are not reproducible or not documented at all.
• The literature does not cover the comparator the manufacturer claims equivalence with.
• The state-of-the-art search is missing or treated as an afterthought.
• Inclusion and exclusion criteria are defined after the fact, not before.
• Appraisal is inconsistent or done by a single reviewer.
• The conclusion does not actually follow from the data presented.
• There is no plan to update the search and no integration with PMCF.

Most of these are process problems, not data problems. The fix is to write the search protocol before you start searching and to treat the CER as a living document.

Practical recommendations

A few things that make the process easier in practice.

Run the search with two reviewers from the start. The cost in time is small compared to redoing it after a Notified Body finding.

Use reference management software such as Zotero, EndNote, Mendeley, or a dedicated systematic review tool like Covidence or Rayyan. Manual tracking does not scale once you cross a few hundred hits.

Build a PRISMA flow diagram even if you are not publishing a systematic review. It is the cleanest way to show identification, screening, eligibility, and inclusion in a CER.

Save the raw output of each database search as a timestamped PDF or screenshot. This is the easiest piece of evidence to produce when a Notified Body asks for it.

Maintain a deviation log. If something changed between your search protocol and your actual search, document why.

References

Ecker, E. D., & Skelly, A. C. (2010). Conducting a winning literature search. Evidence-Based Spine-Care Journal, 1(1), 9 to 14.

European Commission. (2016). MEDDEV 2.7/1 Revision 4. Clinical Evaluation: A Guide for Manufacturers and Notified Bodies.

European Parliament and Council. (2017). Regulation (EU) 2017/745 on Medical Devices (MDR).

GHTF Study Group 5. (2007). Clinical Evaluation (SG5/N2R8). Global Harmonization Task Force.

Pati, D., & Lorusso, L. N. (2017). How to write a systematic review of the literature. Health Environments Research & Design Journal.

Valla, V., Tzelepi, K., Charitou, P., Lewis, A., Polatidis, B., Koukoura, A., et al. (2023). Use of real-world evidence for international regulatory decision making in medical devices. International Journal of Digital Health, 3(1).