Review articles
A review article brings existing evidence together to explain a subject, answer a question and show what the evidence supports.
Contribution and readership
Reviews present no new data but assess published research, making them secondary research; abundant information, divergent views and lack of consensus create a need for synthesis. They can be original through innovative thinking, practical insight and new approaches to familiar problems. Condensing substantial evidence into a succinct clinical review that can influence practice is difficult, and success depends on finding, weighing, organising and presenting the evidence.
Readers include clinicians, students, residents, fellows, health-policy specialists and attorneys; they often choose reviews by compelling titles. Repeated publication on a topic makes an author’s name familiar and can bring specialist referrals. Reviews need useful, skilful writing to sustain attention and are generally peer reviewed for medical content and carefully edited for comprehensibility.
Clinical reviews can explain diagnosis, treatment or prevention, follow symptoms or signs across specialties, describe innovations or changes in practice, or summarise a field’s current state. Literature reviews can bridge research and patient care, support future research and academic rigour, and inform health policy; beginning authors should seek experienced mentorship for an authoritative state-of-the-art review.
Review forms
Narrative reviews summarise literature, compare studies, discuss data and develop hypotheses; status quo reviews contain only the most recent research. A narrative review may suffice for background or typical practice, while a systematic review tends to cite more comprehensive research, be less biased and address explicit clinical questions. An evidence-based clinical review is a systematic review focused on a clinically relevant question and the quality of its evidence. The boundaries between literature reviews, systematic reviews and evidence-based clinical reviews are not always clear from labels alone.
Narrative and systematic reviews can both include randomised trials, observational studies and case reports, and both are vulnerable to bias because they look back at existing research. Their principal differences are:
| Aspect | Narrative review | Systematic review |
|---|---|---|
| Features | Describes and appraises published articles, sometimes without explaining selection methods. | Defines a question, secondary questions or subgroups; specifies selection, extraction and synthesis methods; searches comprehensively; and critically appraises studies against standards. |
| Uses | General debate, appraisal of previous research and gaps, rationales for future research and speculation on new interventions. | Identifies, assesses and synthesises evidence for a question, establishes what is known and its basis, and exposes assumptions and methods to external examination. |
| Limitations | Assumptions, planning and selection or evaluation bias may be unknown; not reproducible. | Scope is bounded by the question, search terms and criteria; readers often need to reformulate questions left unanswered. |
Narrative reviews can preserve the historical development of a principle or clinical concept and allow broader exploration, while systematic rigour is needed to evaluate diagnostic or treatment efficacy and exposure outcomes. Neither form meets every need; meta-narrative reviews and realistic syntheses were among alternatives discussed in 2015.
Question, scope and journal
Ideas can come from daily practice and questions worth sharing with colleagues, but an early search should check whether a similar review exists; an empty result can reflect publication delays or unindexed journals. Colleagues or local medicines-management groups can test whether the idea adds value; in 2007, Drug and Therapeutics Bulletin (DTB) used twice-yearly specialist and generalist editorial-board meetings to critique and prioritise ideas from publications, promotions, readers and practice.
Writing down one or two key questions clarifies the purpose and saves time, alongside six planning questions:
- Why review this topic?
- What should the review achieve?
- What is its scope?
- Who is its audience?
- How long should it be?
- What are its key messages?
The review’s scope should distinguish a single issue from its wider management context, retaining relevant information and excluding distracting detail. A 2003 nicorandil review removed background on angina management, beta-blocker underuse and nitrate costs because it added nothing about nicorandil. Available space may determine scope, but even unlimited space calls for a fixed target length and approximate section allocations.
Specialists may need less background, while mixed readerships may need two or three paragraphs on clinical features, epidemiology, diagnosis or established treatments; jargon and undefined abbreviations should be avoided, while shared professional terminology can remain.
Writers should study comparable reviews in the target journal and read its instructions before drafting for length, line spacing, language, displays and submission checklists. A 2018 assessment saw specialty and controlled-circulation journals as realistic outlets for new authors, while leading international journals sought leading experts.
Early contact with an editor can reveal an overlapping commission and establish a useful relationship; a telephone conversation was preferred to email in 2018 guidance. An editorial inquiry should cover:
- The topic and proposed treatment.
- Relevant qualifications.
- Planned tables and figures.
- Completion date.
- Possible conflicts of interest.
Encouragement is not acceptance of an unseen manuscript; an editor may first request an outline with draft abstract and major headings, accompanied by an abbreviated curriculum vitae.
Outline and organisation
Three familiar models are:
- Approach to disease: diagnosis, treatment or prevention.
- Clinical manifestation or diagnosis: a symptom or sign followed towards diagnosis.
- List: a bounded set such as indications, methods or adverse effects.
A clinical review can be a journal article or book chapter and follow a predictable, flexible sequence; one migraine outline used:
- Pathophysiology.
- Clinical manifestations.
- Diagnosis: history and physical findings; diagnostic imaging.
- Therapy: lifestyle and diet; management of the acute headache; prophylactic therapy.
A category or list title can suggest its outline: a case for and against statins in primary prevention suggests current issues, the case for, the case against and a conclusion. Other organising titles include a drug-class history, myths about a therapy, three new drugs, four common evaluation errors or five reasons to write. Classic reviews resemble book chapters in their introduction, thematic development, conclusion and references.
Narrative-review organisation can use Introduction, Methods, Results and Discussion, chronology or a conceptual frame based on variables and their relationships, respecting journal style and disciplinary conventions. A general framework is:
| Part | Contents |
|---|---|
| Introduction | Rationale, organisation of collected information, objectives and scope. |
| Literature search | Databases and keywords; inclusion and exclusion criteria for studies, languages and dates; availability of selected studies; citations and reference list. |
| Central body or discussion | Sections for key concepts, each discussed, evaluated and summarised against the question. |
| Conclusions | Main points from each section, links to research needs and implications for research design. |
| Abstract | The journal’s descriptive or structured format. |
Evidence selection and assessment
Narrative reviews can borrow systematic search discipline to reduce selection bias, with explicit questions, inclusion and exclusion criteria and relevance-based study selection. A methods section is not mandatory for a narrative review, depending on journal style, but including one can clarify key messages.
One 2015 narrative-review search used PubMed, EMBASE and Google Scholar and the terms “medical literature review writing”, “medical narrative review writing” and “medical systematic review writing”; it included all article types, PubMed-indexed and human-related articles, excluded unavailable full texts, non-English articles and grey literature, and added references found in retrieved papers.
Evidence policies differ: DTB in 2007 normally used publicly available full peer-reviewed trials and syntheses, using abstracts or company-held data only exceptionally, identifying their limitations and never basing conclusions on them alone; 2016 commissioned-review advice similarly favoured primary, high-quality and pivotal studies, with few other reviews and unpublished data as exceptions. Less accessible evidence can be crucial for systematic reviews or new drugs, but may distract other reviews.
Relevance depends on the question: mode of action matters for a first-in-class drug, while physicians may find pharmacokinetic and pharmacodynamic studies useful for a new formulation; older studies with valuable information should not be forgotten. A methodological approach matters because inadequate reporting affects interpretation, translation and application of research.
Drafting and displays
One drafting approach first puts key messages and basic information down without concern for style or accuracy, then checks accuracy and develops arguments and structure. Its drafting reminders are:
- Set the focus in a brief introduction.
- Give brief relevant data and comments on quality and clinical relevance.
- Demarcate information so messages are easy to grasp.
- Use headings and subheadings as signposts.
- Summarise unequivocal evidence concisely.
- Conclude faithfully, linking points to practice.
- Highlight controversies and research needs.
Another approach starts with retrieved data displayed in figures and tables, with every section referring to that evidence, then writes the introduction after results and discussion; working backwards helps remove irrelevant or redundant points. The discussion reassesses results against the original question and explains the meaning and validity of conclusions. A 2016 estimate allowed one to two weeks for a first draft, with reading notes on organisation and interesting points producing an early rough draft.
The introduction should establish topic, purpose and relevance to practice; how often a practitioner meets a problem may engage more than conventional epidemiology, while a new-drug review can identify established comparators. The body should follow a logical rather than necessarily chronological order, reporting relevant study methods, populations and interventions and primary outcomes with statistical significance, preferably confidence intervals. Numbers needed to treat or harm may aid interpretation, while secondary endpoints should be selective and unequivocal findings can be summarised without detailing every study. A 2005 trial summary illustrated the level of detail: a double-blind randomised trial compared aspirin 325 mg daily with clopidogrel 75 mg daily in 19,185 patients with prior ischaemic stroke (6,431), myocardial infarction (6,302) or peripheral arterial disease (6,452); after mean follow-up of 1.91 years, annual rates for the combined endpoint of ischaemic stroke, myocardial infarction or vascular death were 5.32% with clopidogrel and 5.83% with aspirin (P=0.043). Discussion should address methodological strengths, weaknesses, clinical significance, achievements, controversies and outstanding questions rather than merely listing findings.
Headings, subheadings, text boxes and displays can highlight themes and key information. Tables should supplement rather than repeat prose, and figures explaining complex drug mechanisms call for a graphic designer experienced in medical illustration.
Title, abstract and conclusion
Titles and keywords matter for indexing and reaching readers; titles should convey the article’s essence and attract readers, with informative titles preferred to merely indicative ones. Generic labels such as “A review”, “Clinic review”, “Updated review” or “Clinical evidence” add no value, while “literature review”, “narrative review” or “review of the literature” clarifies design. A title can name both subject and scope, such as soccer injuries and their epidemiology and aetiology.
The abstract should be written last, with recurring and descriptive keywords checked against Medical Subject Headings (MeSH), the National Library of Medicine’s controlled vocabulary. The conclusion should stand alone, summarise what is known and unknown, leave messages for action and introduce no new data or ideas.
Opinion may help where reliable evidence is missing, but its basis and status should be clear and its soundness, objective support and breadth of acceptance checked; artificial consensus should not be imposed on complex or specialised disagreements.
Revision and checking
Seven mistakes to check for are:
- An unimportant topic: test its interest with colleagues.
- A stale rehash: search, consult experienced clinicians and ask an editor whether the proposed contribution is new.
- An existing review using the intended approach: change the angle or topic.
- Losing direction: maintain the major-heading outline.
- Excessive length: a 2018 suggestion was 16–20 double-spaced pages including references, with longer state-of-the-art reviews as exceptions.
- Too many or too few references: compare similar articles in the journal.
- A wrong journal: avoid outlets that do not publish reviews or accept only internationally recognised experts.
Editing should address style, clarity, consistent abbreviation definitions and display placement, removing repetition and unnecessary text without adding filler to meet length. Even one critical colleague can identify missing evidence, disputed interpretation, unsupported assumptions and errors; in 2007, DTB used up to 100 specialist and generalist commentators over three stages and validated each statement against its references. Commentators can bring insights from providing or receiving treatment and suggest solutions.
Four final questions are:
- Is it right?
- Is it relevant?
- Is it readable?
- Is it ready?
Commissioned reviews
A commissioned review can run from an advisory-board brief through author input and peer-review responses over several months. A commercial rationale in 2016 was to strengthen a product’s scientific profile, create interest before approval, establish it after approval, renew interest in an older product, discuss adverse-event management or present real-world performance; the provider’s view was that reviews carried greater credibility and attention than brochures.
After the initial meeting, the writer circulates a brief containing stakeholders’ opinions for expert and sponsor approval, and often an executive summary for the sponsor; balancing sponsor wishes with what experts will say can be difficult, but a clear agreed brief reduces later corrections. Target-journal discussion belongs at the initial meeting, balancing impact factor against relevance and realistic acceptance; open access can widen dissemination despite sometimes lower impact factors.
Stakeholder review can take weeks or months and several revisions, with the experts having the final say on content. Distribution can use the sponsor’s website, trade-fair offprints and sales-force delivery to physicians; dissemination plans should inform journal choice and the use of open access or purchased offprints.