03 · What You Need to Know
What each review approach is designed to do
Start with the question, not the review label
Before deciding on a methodology, finish this sentence:
After completing this review, I want to be able to say...
Your answer might be:
- how researchers currently understand a particular debate;
- whether an intervention improves an outcome;
- what types of evidence exist on an emerging topic;
- how a concept has been defined;
- which populations have and have not been studied;
- how research methods differ across a field; or
- what the combined quantitative evidence estimates about an effect.
These objectives imply different forms of evidence synthesis.
What is a narrative review?
A narrative review synthesizes and interprets a body of literature primarily through structured scholarly discussion rather than through the full prespecified procedures associated with a systematic review.
It may be appropriate when you want to:
- provide a broad overview of a field;
- explain how theories or ideas have developed;
- compare competing perspectives;
- integrate diverse forms of evidence;
- discuss a topic whose literature is methodologically heterogeneous;
- develop a conceptual argument; or
- provide an expert interpretation of a research area.
“Narrative” should not mean arbitrary. A strong narrative review still needs a clear purpose, defensible source selection, critical evaluation, accurate citation, and genuine synthesis.
The key difference is that the review does not necessarily claim to have identified evidence through the same prespecified, comprehensive, reproducible process expected of a systematic review.
What is a systematic review?
A systematic review uses explicit methods to identify, select, appraise, and synthesize evidence addressing a defined research question.
Cochrane describes systematic reviews as attempts to collate empirical evidence fitting prespecified eligibility criteria, using explicit systematic methods intended to minimize bias. Protocols document important methods in advance.
Typical elements include:
- a clearly defined review question;
- prespecified eligibility criteria;
- a documented search strategy;
- systematic study selection;
- structured data extraction;
- assessment of risk of bias or methodological limitations;
- a planned approach to synthesis; and
- transparent reporting of the review process.
PRISMA 2020 provides reporting guidance for systematic reviews, including explaining why the review was conducted, what methods were used, and what results were found.
Systematic reviews are especially appropriate when the objective is to answer a focused question using all relevant evidence that meets defined criteria rather than selecting studies informally to illustrate an argument.
What is a scoping review?
A scoping review systematically maps a body of evidence rather than necessarily attempting to produce one focused estimate or answer about effectiveness.
JBI identifies several reasons for conducting a scoping review, including identifying the types of available evidence, identifying and analysing knowledge gaps, clarifying concepts or definitions, examining how research is conducted, and identifying key characteristics or factors related to a concept.
A scoping review may therefore be useful when:
- the topic is broad or emerging;
- terminology is inconsistent;
- many different study designs exist;
- you want to map populations, concepts, methods, or settings;
- you want to understand how much evidence exists before developing a narrower systematic review;
- you want to identify research gaps systematically; or
- the primary objective is describing the landscape of research rather than estimating one effect.
Scoping reviews still use explicit methods. JBI's methodological framework includes protocol development, search strategies, data extraction, and presentation of results.
What is a meta-analysis?
A meta-analysis statistically combines quantitative results from two or more separate studies.
That definition immediately reveals why “systematic review or meta-analysis?” is often the wrong question.
Systematic review
A review methodology for systematically identifying, evaluating, and synthesizing evidence addressing a defined question.
Meta-analysis
A statistical synthesis method that may be used within a systematic review when appropriate quantitative study results can meaningfully be combined.
Cochrane explicitly describes meta-analysis as statistical combination and cautions against jumping directly to it before addressing the review question, eligibility criteria, study identification, risk of bias, comparisons, and suitability of the available data.
A systematic review can therefore contain:
- one or more meta-analyses;
- other quantitative synthesis methods;
- structured synthesis without meta-analysis; or
- different approaches for different outcomes.
The review remains systematic because of how the evidence was identified and handled, not because a forest plot appears in the paper.
How the four approaches compare
| Feature |
Narrative review |
Systematic review |
Scoping review |
Meta-analysis |
| Primary purpose |
Interpret and explain a body of literature |
Answer a defined question systematically |
Map the nature, extent, concepts, or gaps in a body of evidence |
Statistically combine quantitative results |
| Typical question breadth |
Variable |
Usually clearly bounded |
Often broader and exploratory |
Depends on the review question and studies being combined |
| Prespecified eligibility criteria |
May vary with review approach |
Expected |
Expected |
Inherited from the review in which the analysis occurs |
| Systematic search |
Can be structured but need not claim comprehensive systematic retrieval |
Core methodological component |
Core methodological component |
Does not itself define how studies were found |
| Critical appraisal |
Should critically engage with evidence, but methods vary |
Normally integral to evaluating included evidence |
Depends on purpose and methodology; not automatically identical to systematic-review appraisal |
Risk of bias must be considered when interpreting pooled results |
| Typical output |
Interpretive synthesis |
Systematic synthesis of evidence |
Map or characterization of evidence |
Pooled quantitative estimate and uncertainty, when appropriate |
| Meta-analysis required? |
No |
No |
No |
It is the statistical method itself |
A systematic review does not automatically require a meta-analysis
This distinction is important enough to repeat.
After eligible studies have been identified, reviewers must determine whether their quantitative results are sufficiently appropriate to combine. Studies may differ substantially in populations, interventions or exposures, comparators, outcomes, designs, measurements, follow-up periods, and risk of bias.
Cochrane recommends examining study characteristics before statistical synthesis and determining which studies can appropriately be grouped.
If statistical pooling would produce a misleading summary, the correct decision may be to conduct a systematic review without meta-analysis.
Watch Out
Do not perform a meta-analysis merely because you have several numerical studies. Cochrane warns that meta-analysis can seriously mislead when differences among studies, within-study bias, and reporting bias are not adequately considered.
A scoping review is not a preliminary systematic review with lower standards
Scoping reviews and systematic reviews have different objectives.
A scoping review might ask:
What types of interventions have been studied for reducing researcher burnout, in which populations and settings, and using which outcomes?
A systematic review might instead ask:
Among early-career researchers, do structured mentoring interventions reduce validated burnout scores compared with usual institutional support?
The first maps a landscape. The second asks a more focused comparative question.
JBI's current decision guidance explicitly distinguishes scoping-review purposes from evidence syntheses intended to provide evidence directly informing practice or policy, for which a systematic evidence-synthesis methodology may be more appropriate.
A broad question does not automatically mean you need a scoping review
Question breadth is only one consideration.
Ask what output you want.
If you want to map how a broad concept is defined, what methods have been used, and where research exists, a scoping review may fit.
If you want to make a critical scholarly argument integrating several traditions, a narrative review may fit better.
If you want a systematic answer to a question that can still be broad but has explicit eligibility criteria and a defined synthesis objective, a systematic review may be appropriate.
The label should follow the methodology and purpose, not simply the number of keywords in the question.
Critical appraisal differs according to the review's purpose
Systematic reviews generally need structured evaluation of the methodological limitations or risk of bias of included evidence because conclusions depend on how trustworthy those studies are.
Scoping reviews may have a different objective. If the purpose is to map what research exists rather than determine the effectiveness of an intervention, formal critical appraisal may not always be necessary. Whether it should be included depends on the scoping review's question and methodology.
Narrative reviews should still evaluate evidence critically. The absence of a formal risk-of-bias instrument does not justify treating every source as equally convincing.
Meta-analysis particularly requires attention to study validity because statistical precision cannot repair systematic bias in the studies being combined.
Review type does not determine whether the writing is narrative
The word narrative can create another source of confusion.
A systematic review without meta-analysis may synthesize findings in prose. That does not automatically make it a narrative review.
The distinction concerns the methodology used to conduct the review, not whether sentences rather than equations are used to present the results.
A systematic review can use synthesis methods other than meta-analysis. Cochrane's Handbook contains a dedicated chapter on synthesis using other methods for precisely this reason.
Do you need a protocol?
For systematic reviews, developing methods in advance is a central safeguard against making important decisions after seeing which studies or results are available. Cochrane emphasizes prespecified questions and methods documented in protocols.
Scoping reviews should also be planned systematically. JBI's guidance explicitly includes development of a scoping-review protocol.
A traditional narrative review may not use the same protocol structure, although documenting its purpose, scope, search approach, selection rationale, and synthesis process can improve transparency.
Do you need PRISMA?
PRISMA is a reporting guideline, not a method for turning an informal review into a systematic one.
PRISMA 2020 primarily guides reporting of systematic reviews, and separate extensions address other forms or aspects of evidence synthesis, including scoping reviews.
Using a PRISMA flow diagram does not by itself make a review systematic. The underlying search, selection, appraisal, and synthesis methods must justify the label.
When should you consider meta-analysis?
Meta-analysis becomes a reasonable possibility when several conditions align:
- the review question supports quantitative synthesis;
- multiple eligible studies provide relevant quantitative results;
- the studies estimate sufficiently related effects or quantities;
- outcomes can be represented using appropriate effect measures;
- differences among studies have been considered;
- risk of bias has been assessed appropriately; and
- combining results would produce an interpretable summary rather than obscure important differences.
Even then, methodological choices remain. Cochrane notes that most meta-analysis methods operate as forms of weighted averages and that heterogeneity across studies requires careful consideration.
When should you not conduct a meta-analysis?
Do not force statistical pooling when the studies are too different for the combined estimate to have a meaningful interpretation.
Potential problems include:
- fundamentally different interventions or exposures;
- incompatible outcomes;
- substantially different study designs answering different questions;
- important clinical or methodological heterogeneity;
- insufficient information for an appropriate effect estimate;
- serious risk-of-bias concerns that make the pooled number misleading; or
- a question for which mapping or qualitative interpretation is more useful than one summary estimate.
The absence of a meta-analysis is not evidence that a systematic review failed. Sometimes refusing to calculate a misleading pooled estimate is the methodologically stronger choice.
Consider feasibility before choosing the most demanding review
A systematic review can require substantial work: protocol development, complex searches, duplicate screening, data extraction, risk-of-bias assessment, synthesis, and reporting. Cochrane recommends that systematic reviews be conducted by teams with relevant domain and methodological expertise.
Scoping reviews can also become large projects because broad eligibility criteria may retrieve enormous numbers of records.
Do not label an ordinary course literature review a systematic review merely because “systematic” sounds more rigorous. If you cannot perform the methods that justify the label, choose a review approach you can conduct properly.
Check whether the review already exists
Before launching a substantial systematic or scoping review, search for existing reviews and registered protocols addressing the same question.
JBI's scoping-review decision guidance specifically recommends checking for existing syntheses or protocols and considering whether the proposed review offers a meaningful point of difference or whether enough time has passed to justify an update.
An existing review does not automatically prevent another one. A new review may be justified by newer evidence, a different population, a different methodology, important deficiencies in the earlier review, or a genuinely different question.
But discovering after months of work that a nearly identical high-quality review was recently completed is avoidable.
