To write a narrative review, decide which question the review answers and for which readers, then choose one organising structure for the whole topic before drafting. Write each section as an argued synthesis, describe how you searched, and test the draft against the six items of the Scale for the Assessment of Narrative Review Articles (SANRA).
This guide covers the full-length review article on a broad topic; for one narrow question, the scope check in our guide to writing a mini review fits better. A narrative review is a summary of the literature on a topic with the authors' own interpretation and critique. It cannot claim to have collated all the evidence that meets pre-specified criteria, as a systematic review attempts to, because its search and selection are not set out in advance to be exhaustive. Our comparison of systematic reviews, meta-analyses and literature reviews sets out that limit. Still choosing a format? See the main types of research article.
Choose one organising structure for the whole narrative review
The introduction, methods, results and discussion (IMRaD) order of a research paper rarely fits the body of a review. Pautasso's ten simple rules for writing a literature review say two parts still make sense: an introduction to the context and, near the end, a recap of the main points with take-home messages. The order of everything between them is your decision.
Make that decision before drafting. A concept map helps, and Pautasso notes it can later become a figure. The five structures below are practical options, not a published standard, and they combine: a thematic review can run its themes in mechanism-to-management order.
| Structure | Order of the main sections | Fits when | Watch for |
|---|---|---|---|
| Thematic | One section per question or debate | The field splits into distinct questions, each with its own evidence | Overlapping themes that discuss the same studies twice |
| Mechanism to management | Mechanism, presentation and diagnosis, management, outcomes | A clinical topic where readers need the reasoning behind practice | Mechanism sections that never connect to the clinical ones |
| Chronological | Turning points in how the concept or technique developed | A definition or method has changed, and the change still matters | A history that never says where the field stands now |
| Methodological | One section per kind of evidence: laboratory, observational, trial, qualitative | The question is how far each kind of study can be trusted | The same findings repeated in every section |
| By population or setting | Adults, children, low-resource settings or other groups in turn | Evidence or practice differs between groups | Groups with too little evidence to fill a section |
Test the plan against Pautasso's account of what a reader should take from a review: the field's major achievements, its main areas of debate and its outstanding research questions. If no section carries the debates or open questions, add one. Within each section, synthesise rather than list, as the listing-versus-synthesis example in the mini review guide shows.
A chronological review is not a meta-narrative review, a distinct method that maps how a research tradition developed, with its own reporting standard on the EQUATOR Network.
Budget words, references and figures before you draft
A broad topic fills whatever space it is given, so budget each section from the target journal's limits. Protect the section on debates and open questions: Byrne, writing on the peer review of narrative literature reviews, notes that extending or developing ideas "is often the weakest section of a review". Leave some words and references spare too, because Pautasso suggests a fresh search for new work at the revision stage.
Copy the outline below and fill each slot. The example budget is invented to illustrate the structure; its numbers are not real data. It plans a thematic review against Directive's general Review limits, given at the end of this guide.
Title: [topic]: [the question or angle the review takes]
Abstract: as the target journal asks (at Directive, structured, 200–250 words).
1. Introduction and scope: why [topic] matters to [this journal's readers]; the questions answered; related topics left out, and where they are reviewed; the section order. Budget: [words], [references], [figures] (example: 700 words, 15 references, 1 concept map).
2. Literature search: sources, search terms, period covered, types of literature included. Budget: [words] (example: 250 words, no references).
3 to 5. Main sections in your chosen order: one section for each [theme, stage, period, kind of evidence or group] your structure sets, each written as a synthesis rather than a list. Budget for each: [words], [references], [figures or tables] (example: sections 3 and 4, 1,200 words, 22 references and 2 figures or tables each; section 5, 1,000 words, 18 references and 1 figure or table).
6. Debate and open questions: what is still disputed; the outstanding research questions and the kind of study that could settle each; a summary table. Budget: [words], [references], [tables] (example: 800 words, 12 references, 1 table).
7. Take-home messages: [number] specific points a reader can act on or test. Budget: [words] (example: 300 words, no references).
Back matter: declarations (for a review, the data availability statement can say that no new data were generated or analysed), references, tables as editable text, figure legends.
Totals: [words] of [limit] words; [references] of [limit]; [figures or tables] of [limit] (example: 5,450 of 6,000 words; 89 of 100 references; 7 of 8 figures or tables).
Count each reference only in the section that first cites it. The example leaves 550 words, 11 references and one figure or table spare. Directive's word limit excludes tables, so a table of study-by-study detail saves words for the argument but still uses one of the 8 figure or table slots.
Score the draft on SANRA's six items
The Scale for the Assessment of Narrative Review Articles (SANRA) was published in 2019 by Baethge, Goldbeck-Wood and Mertens. The SANRA paper in Research Integrity and Peer Review rates six items from 0 to 2, for a maximum of 12. Its authors built it for editors, reviewers and readers, and for authors writing narrative reviews.
Know its limits. In its authors' words, SANRA is "a critical appraisal tool and not a reporting guideline", and the EQUATOR library did not list it when we checked in October 2026. It does not judge originality, topicality, conflicts of interest or whether the content is correct. There are no established cut-offs; the authors say only that, in their experience, a score of 4 or below indicates very poor quality. Use the scores to find weak items, not as a pass mark.
The scale wording is quoted under its Creative Commons Attribution (CC BY 4.0) licence; the questions are ours. Score the whole manuscript, abstract included.
| SANRA item | Scale wording for a score of 2, then 1 | Ask of your draft | Your score |
|---|---|---|---|
| 1. Justification of the article's importance for the readership | 2: "The importance is explicitly justified." 1: "The importance is alluded to, but not explicitly justified." | Does the introduction name the clinical problem and the open questions for this journal's readers? | [0, 1 or 2] |
| 2. Statement of concrete aims or formulation of questions | 2: "One or more concrete aims or questions are formulated." 1: "Aims are formulated generally but not concretely or in terms of clear questions." | Can a reader underline the aims or questions? | [0, 1 or 2] |
| 3. Description of the literature search | 2: "The literature search is described in detail, including search terms and inclusion criteria." 1: "The literature search is described briefly." | Are search terms and types of literature included stated, in the text or a supplement? | [0, 1 or 2] |
| 4. Referencing | 2: "Key statements are supported by references." 1: "The referencing of key statements is inconsistent." | Is every key statement your argument depends on backed by a reference, apart from uncontroversial facts? | [0, 1 or 2] |
| 5. Scientific reasoning | 2: "Appropriate evidence is generally present." 1: "Appropriate evidence is introduced selectively." | Does each key argument name its study designs and, where available, the level of evidence? | [0, 1 or 2] |
| 6. Appropriate presentation of data | 2: "Relevant outcome data are generally presented appropriately." 1: "Data are often not presented in the most appropriate way." | Are absolute risks given, not only relative ones; are clinical end-points reported, not only surrogate or intermediate ones; do effect sizes carry confidence intervals? | [0, 1 or 2] |
| Total | Maximum 12 | Which items scored 0 or 1, and what will you change? | [total] of 12 |
Item 3 does not need systematic-review detail such as flowcharts, but a 2 is not possible without the search terms and the types of literature included. The mini review guide shows how to write that search paragraph. Not everyone agrees that narrative reviews need search rules: Greenhalgh, Thorne and Malterud call how-to guides that demand an explicit search strategy and strict inclusion criteria misleading. A short search description still costs little, and SANRA rewards it.
Directive's Review Article format and limits
Our article types page defines a Review Article as "a narrative, expert critical synthesis of a topic — distinct from a systematic review in that its search and selection are not exhaustively pre-specified". Its typical structure is "a structured argument organised into thematic sections, supported by a comprehensive, current reference list", so chronological, mechanism-to-management and other structures can still sequence those themes. Review Articles receive full external double-blind peer review, with at least two reviewers sought. The page assigns them no reporting guideline; PRISMA applies to systematic reviews. No Directive page asks editors or reviewers to score reviews with SANRA, so a SANRA self-check is for your own use.
The Author Guidelines set general limits for a Review: up to 6,000 words (excluding the abstract, references, tables and figure legends), a structured abstract of 200–250 words (Background, Methods, Results, Conclusion), up to 100 references and up to 8 figures or tables. The article types page gives a shorter indicative length but names the Author Guidelines as the definitive source, so use those figures unless a journal's own page states something more specific. Because review is double-blind, phrase citations of your own work in the third person, as our guide to anonymising a manuscript explains. When you submit, choose Review as the article type on the submission form. If you spot an error in this guide, see how to report a problem to us.