EU MDR
MDR Article 61: what 'systematic' actually requires from your literature review
Direct answer: MDR Article 61 requires the clinical evaluation to follow "a defined and methodologically sound procedure", and Annex XIV Part A requires manufacturers to identify available clinical data and gaps in clinical evidence "through a systematic scientific literature review". The regulation never defines "systematic" — but systematic review methodology does: planned before execution, reproducible by a third party, and documented record by record. PRISMA 2020 is the closest thing to an operational definition, and mapping your literature review onto it is the most defensible way to satisfy the word.
What Article 61 actually says
Article 61(1) of Regulation (EU) 2017/745 requires confirmation of conformity with the relevant general safety and performance requirements to be based on clinical data providing sufficient clinical evidence, and obliges manufacturers to "plan, conduct and document a clinical evaluation" in accordance with Article 61 and Part A of Annex XIV. Article 61(3) then specifies that the clinical evaluation shall follow a defined and methodologically sound procedure based on, among other things, "a critical evaluation of the relevant scientific literature currently available relating to the safety, performance, design characteristics and intended purpose of the device".
Annex XIV Part A makes the literature requirement explicit: manufacturers shall "identify available clinical data relevant to the device and its intended purpose and any gaps in clinical evidence through a systematic scientific literature review". Three loaded words — systematic scientific literature review — and no definition anywhere in the regulation.
"Systematic" has an operational definition — borrow it
The MDR drafters did not invent the term. "Systematic review" has a settled meaning in evidence-based medicine, codified in reporting standards like PRISMA 2020 and in the methodology chapters of MEDDEV 2.7/1 rev 4. Distilled to its operational core, a literature review is systematic when it is:
- Planned — a protocol exists, dated before the search ran, stating the question, sources, search strings, and inclusion/exclusion criteria.
- Reproducible — a third party with the protocol and the search report could re-run the search and reach substantially the same included set.
- Documented — every identified record is accounted for: screened, included, or excluded with a stated reason, with counts that reconcile.
If your review meets those three tests, you can defend the word "systematic" in front of any notified body. If it fails any one of them, no volume of included papers compensates.
Why CER literature sections fail notified body review
Clinical evaluation findings are consistently among the most common sources of notified body nonconformities under the MDR, and within them the literature review is a recurring soft spot. The typical patterns:
- No protocol, or a protocol written after the search — the review reads as a justification of a pre-selected bibliography.
- Single-database searches — PubMed alone, despite MEDDEV 2.7/1 rev 4 Appendix A4 stating it is not sufficient (see our MEDDEV 2.7/1 rev 4 workflow guide).
- Missing exclusion records — the report lists what was included but cannot show what was rejected or why, so the flow of records does not reconcile.
- Irreproducible strategy — no dated query strings, no per-source hit counts, no deduplication record.
- State of the art skipped — the search covers the device but not the clinical condition and alternative treatments, so the benefit-risk benchmark is unsupported.
Mapping PRISMA 2020 onto MDR expectations
PRISMA 2020 is a reporting standard for systematic reviews, not a regulatory document — but its items map remarkably cleanly onto what an MDR reviewer needs to see. This mapping is the practical bridge between the academic and regulatory worlds:
| PRISMA 2020 concept | MDR / guidance counterpart | Evidence a reviewer expects |
|---|---|---|
| Eligibility criteria (item 5) | Selection criteria in the search protocol (MEDDEV 2.7/1 rev 4, App. A5) | Inclusion/exclusion criteria dated before execution |
| Information sources (item 6) | App. A4 source requirements; Annex XIV "systematic ... review" | Named databases and registers with dates of coverage and search |
| Search strategy (item 7) | Documented, reproducible search (App. A5) | Full query strings per source, with limits and filters |
| Selection process + flow diagram (items 8, 16a) | Literature search report; screening documentation | Identified → screened → excluded (with reasons) → included, numbers reconciling |
| Risk of bias / certainty (items 11, 15) | Appraisal of suitability and contribution (rev 4 §9.3, App. A6) | Per-study appraisal with criteria, by a qualified evaluator |
| Limitations (item 23) | Gaps in clinical evidence (Annex XIV Part A) | Explicit statement of what the evidence does not cover — this feeds the PMCF plan |
Note the last row: under the MDR, the "gaps" output of the literature review is not a formality. It is the formal input to post-market clinical follow-up — the gaps you declare are the gaps your PMCF literature reviews must then monitor.
Software as a medical device: the MDCG 2020-1 angle
For software (MDSW), MDCG 2020-1 structures clinical evidence into three components: valid clinical association (is the software's output clinically accepted?), technical performance, and clinical performance. The first component is typically demonstrated largely through literature — which makes the systematic review even more load-bearing for SaMD manufacturers than for hardware devices. The same three tests (planned, reproducible, documented) apply.
What to document, concretely
A literature review package that survives scrutiny contains: the protocol (versioned, dated); the search log (per-source queries, run dates, hit counts); the deduplication record; the screening log with one exclusion reason per rejected record; the flow diagram generated from those counts; the appraisal table; and the citation list in which every reference is verifiable. That last point is newly urgent: AI-assisted drafting has made fabricated references a documented industry problem (see the 2026 data on fabricated citations), and a citation in a CER that does not resolve is a credibility wound an auditor will not overlook.
AutoSearch produces this package as a by-product of running the review: the search executes across 12 source families under a stored protocol, every screening decision is logged, the PRISMA flow diagram is generated from the actual run counts, and every citation's DOI is verified against Crossref before it can appear in the output. The methodology page documents each step; the approach to PRISMA specifically is covered in our PRISMA 2020 methodology post.
FAQ
Does the MDR require PRISMA compliance?
No. The MDR requires a "systematic scientific literature review" without defining the method. PRISMA 2020 is the most widely accepted operational standard for what "systematic" means, which makes it the strongest available benchmark — voluntary, but persuasive to reviewers.
Is a literature review alone sufficient clinical evidence?
That depends on the device, its class, its claims, and the equivalence route (Article 61 sets specific conditions, and MDCG 2020-6 addresses legacy devices). The literature review is always required; whether it suffices without clinical investigation is a separate, device-specific determination.
Can AI tools run the review?
AI can execute the search, screening, logging, and flow-diagram generation — the parts where humans drift and software does not. Appraisal, benefit-risk judgement, and the conclusions of the clinical evaluation must remain with qualified evaluators, and the tool must expose an audit trail rather than a black-box answer.
Disclaimer. This article is general information, not regulatory advice. Responsibility for the clinical evaluation and conformity with Regulation (EU) 2017/745 remains with the manufacturer and its person responsible for regulatory compliance (PRRC, MDR Article 15). Verify all requirements against the current regulation, MDCG guidance, and your notified body.
Make "systematic" the easy part
See how AutoSearch handles the full workflow on the EU MDR/IVDR use case page, or start a systematic literature review for your clinical evaluation now.