A systematic review and a meta-analysis are closely related but distinct methodological approaches to evidence synthesis. A systematic review is a structured, transparent process for identifying, evaluating, and synthesizing all relevant research on a specific question. A meta-analysis is a statistical technique that quantitatively combines the results of multiple studies to produce a pooled effect estimate. Every meta-analysis is conducted within a systematic review, but not every systematic review includes a meta-analysis.
Understanding when and how to use each approach is fundamental for researchers in evidence-based medicine, public health, education, psychology, and other disciplines that rely on synthesized evidence for decision-making.
Defining Systematic Reviews
A systematic review follows a rigorous, pre-specified methodology to answer a defined research question. The process includes comprehensive database searching, explicit eligibility criteria, systematic study selection, quality assessment, and structured synthesis of findings. The entire process is documented transparently, typically using a PRISMA 2020 flow diagram.
The key features of a systematic review are:
- Pre-registered protocol (ideally on PROSPERO or similar registry)
- Comprehensive, reproducible search strategy across multiple databases
- Explicit inclusion and exclusion criteria applied by at least two independent reviewers
- Standardized data extraction and risk of bias assessment
- Structured synthesis of findings (qualitative or quantitative)
- PRISMA 2020 compliant reporting
For a complete walkthrough, see our guide on how to conduct a systematic review step by step.
Defining Meta-Analysis
A meta-analysis is the statistical component that may be conducted within a systematic review. It uses mathematical methods to combine quantitative results from multiple independent studies, producing a single summary effect estimate with a confidence interval. Meta-analysis increases statistical power, improves precision of effect estimates, and can resolve conflicting findings across individual studies.
Key statistical outputs of a meta-analysis include:
- Pooled effect estimate (odds ratio, risk ratio, mean difference, standardized mean difference)
- Confidence interval around the pooled estimate
- Forest plot visualizing individual study effects and the pooled result
- Heterogeneity statistics (I², Q, tau²) quantifying variability between studies
- Funnel plot for publication bias assessment
Key Differences
| Aspect | Systematic Review | Meta-Analysis |
|---|---|---|
| Definition | Structured evidence identification and synthesis | Statistical pooling of quantitative results |
| Always included? | Yes (in a meta-analysis) | No (only when appropriate) |
| Output | Narrative or quantitative synthesis | Pooled effect estimate with CI |
| Requires statistics? | Not necessarily | Yes, always |
| Can stand alone? | Yes | No (requires systematic review) |
| PRISMA required? | Yes | Yes (reported within the SR) |
| Handles heterogeneity | Through narrative synthesis | Through statistical models |
When to Conduct a Meta-Analysis
A meta-analysis is appropriate when:
- Two or more studies report quantitative outcomes that can be combined
- Studies are sufficiently similar in population, intervention, comparison, and outcome (clinical homogeneity)
- The same or comparable effect measures can be calculated across studies
- Statistical heterogeneity is manageable (assessed via I² and Q statistics)
A meta-analysis is NOT appropriate when:
- Studies are too heterogeneous in design, population, or outcomes to combine meaningfully
- Only one study exists for an outcome
- The data reported are insufficient to calculate effect estimates
- Studies measure fundamentally different constructs
Learn more about assessing variability in our guide on understanding heterogeneity in meta-analysis.
Meta-Analysis Statistical Methods
Fixed-Effect vs Random-Effects Models
- Fixed-effect model: Assumes all studies estimate the same underlying true effect. Appropriate when studies are very similar and heterogeneity is low (I² < 25%).
- Random-effects model: Assumes the true effect varies between studies. Accounts for both within-study and between-study variability. More commonly used because it produces more conservative, generalizable estimates.
Forest Plots
A forest plot is the standard visualization for meta-analysis results. Each horizontal line represents one study's effect estimate and confidence interval. The diamond at the bottom represents the pooled effect. The vertical line represents no effect (null).
Subgroup Analysis and Meta-Regression
When heterogeneity is present, subgroup analysis stratifies studies by a categorical variable (e.g., study design, population age group) to explore sources of variability. Meta-regression models the relationship between a continuous moderator and the effect size.
Sensitivity Analysis
Sensitivity analyses test the robustness of meta-analysis results by systematically removing studies (leave-one-out analysis), changing the statistical model, or restricting to low risk-of-bias studies.
The Role of PRISMA Flow Diagrams
Both systematic reviews and meta-analyses report their study selection process using a PRISMA 2020 flow diagram. The diagram documents how many studies were identified, screened, and included regardless of whether a quantitative synthesis was performed. Create your PRISMA flow diagram using our free online PRISMA flow diagram tool.
The flow diagram is particularly important for meta-analyses because it transparently shows which studies were available for inclusion and why some were excluded, allowing readers to assess potential selection bias in the pooled estimate.
Quality Assessment in Both Approaches
Both systematic reviews and meta-analyses require risk of bias assessment of included studies. The choice of assessment tool depends on study design. For a comprehensive comparison, see our guide on quality assessment tools for systematic reviews.
In meta-analyses, quality assessment results can be incorporated into the analysis through sensitivity analyses (restricting to low-risk studies) or by using quality scores as moderators in meta-regression.
Frequently Asked Questions
Can a systematic review exist without a meta-analysis?
Yes. Many systematic reviews use narrative synthesis instead of meta-analysis when studies are too heterogeneous to combine statistically. A systematic review is always the broader methodology; meta-analysis is an optional statistical component within it.
Is a meta-analysis more rigorous than a systematic review?
Not necessarily. A well-conducted systematic review with narrative synthesis can be just as rigorous as one with meta-analysis. The choice depends on data availability and clinical homogeneity, not on rigor. A poorly conducted meta-analysis that inappropriately combines heterogeneous studies is less useful than a careful narrative synthesis.
What is a "systematic review and meta-analysis"?
This phrase indicates that the authors conducted a systematic review (the overall methodology) that included a meta-analysis (quantitative synthesis) as part of its evidence synthesis. The title should reflect whether meta-analysis was performed.
Do I always need a forest plot?
If you conduct a meta-analysis, a forest plot is the standard way to present results and is expected by most journals. For systematic reviews without meta-analysis, forest plots are not required but can still be useful for visualizing individual study results without pooling.