A literature review (literature search with structured evaluation) is the systematic search, selection and summary of scientific and regulatory relevant sources for a clearly defined question. In clinical development, medical device compliance and medical writing, it serves to present the state of knowledge in a comprehensible manner, to identify evidence gaps and to justify decisions. Depending on the goal, a review can be carried out as a narrative summary, as a systematic review or as a component of a clinical evaluation according to the MDR.
Goals and applications in the CRO and sponsor environment
In projects, literature reviews are used among other things for the planning of clinical trials (choice of endpoints, active controls, expected effect sizes), for risk assessments, for writing clinical study reports and for marketing authorization dossiers. For medical devices, they are a central component of the clinical evaluation as well as the post-market clinical follow-up, because external clinical data can support the safety and performance of the device. A review is thus often the bridge between scientific literature, internal data and regulatory requirements.
In practice, the review quality determines whether notified bodies accept the argumentation as complete and reproducible. Unclear search strategies or selective choice can be considered bias and often lead to questions in the review process.
Methodology: Question, search strategy and documentation
A robust literature review begins with a precise question and clear inclusion and exclusion criteria (population, intervention, comparison, endpoints, study types, time periods, languages). Based on this, a search strategy for databases (e.g. PubMed/MEDLINE, Embase) and grey literature (guidelines, authority reports, registry data) is defined. It is important that search strings, search dates, filters and number of hits are documented so that an independent person can understand the process.
Two independent reviewers are often used to select publications in order to reduce misclassifications. For comprehensive reviews, screening tools and structured data extraction help. For regulated documents, audit trail thinking is also important: Why was a source excluded, and which version was used?
Assessment of evidence quality and bias
In addition to the content summary, the critical assessment of the evidence is also included. This includes questions such as: Are the studies randomized and blinded? How high is the risk for selection or publication bias? Are endpoints clinically relevant or surrogate? How consistent are the results across studies? Standardized evaluation frameworks are used for this in systematic reviews, while narrative reviews should at least contain transparent chains of argumentation.
Typical stumbling blocks are duplicate publications of the same study, incomplete reporting or overlooking unpublished registry data. Therefore, cross-checking with study registries and regulatory public assessment reports is often useful.
Another practical aspect is data extraction: Even with a few dozen sources, a standardized extraction table is worthwhile, in which population, study design, endpoints, follow-up duration, relevant safety events and central effect sizes are recorded. This makes it transparent why certain statements are made and what uncertainties exist. For regulatory documents, it is also important that citations can be unambiguously traced back to the original source.
If a review serves as input for a protocol, it should be explicitly shown how literature results were translated into design decisions – for example into the choice of a comparator, into assumptions about the event rate or into the definition of clinically relevant effect sizes. This increases internal comprehensibility and reduces the risk of later protocol amendments.
In everyday CRO work, timeliness also plays a role: For ongoing programs, reviews are often kept as “living” documents that are updated at defined intervals. Then versioning, search updates and the treatment of newly identified evidence must be clearly described so that the review remains reproducible.
Regulatory embedding (EU focus)
In the context of the Regulation (EU) 2017/745 (MDR), a systematic literature search is expected as part of the clinical evaluation. Manufacturers must show that relevant clinical data have been identified and appropriately evaluated, including the question of whether the literature concerns their own device or similar devices. Also in the pharmaceutical sector, literature reviews are established in dossiers and risk management documents, for example to evaluate known risks or to classify safety signals.
For clinical trials under Regulation (EU) No 536/2014 and within the framework of ICH E6(R3): Literature searches rarely replace the need for controlled trial data, but can underpin the rationale, justify endpoints and support the selection of safety parameters. In medical writing processes, consistency is also crucial: Citations, statements and conclusions must match the data status and be cleanly referenced.
Best practices for reproducible reviews
A clear structure has proven itself: Search strategy and flow (identified, screened, included), tabular extraction of the core data, and a synthesis that discusses strengths, weaknesses and relevance for the question. If quantitative syntheses are not meaningful, it should be justified why (heterogeneity, missing data, different endpoints). For sponsors, it is helpful to determine early on which statements in the dossier are claim-relevant and therefore must be supported particularly strictly.
Frequently Asked Questions (FAQ)
What is the difference between a systematic and a narrative review?
A systematic review follows a previously defined search and selection methodology and documents this transparently. A narrative review summarizes literature in a more descriptive way, but is more susceptible to selection bias if the methodology is not disclosed.
Which sources count as grey literature?
These include guidelines, authority reports, HTA documents, conference contributions or study registries. They are important because not all relevant results are published in journals.
How detailed must the search strategy be documented?
At least search terms, databases, search date, filters and number of hits should be stated. In regulated documents, a comprehensible justification for inclusions and exclusions is also required.
Regulatory references: Regulation (EU) 2017/745 (MDR), Regulation (EU) No 536/2014 (Clinical Trials Regulation), ICH E6(R3) Guideline for Good Clinical Practice.