| Stage | Systematic review | Meta-analysis | Planning use |
|---|---|---|---|
| Starting point | A research question and a protocol or research plan | Results from two or more primary studies | Ask whether the question and plan were fixed before the search. If not, treat the “review” with caution. |
| Selecting studies | Retrieved studies are screened against inclusion and exclusion criteria, by at least two people independently | Checks heterogeneity, meaning whether differences between studies skew the result | Favour reviews that describe who screened what, and how similar the pooled studies were. |
| Combining findings | Overview of primary studies using explicit, reproducible methods; may synthesize in words | Mathematical synthesis that can give a more precise estimate of effect than any single study | A pooled number is a bonus, not a requirement for a sound review. |
| How results appear | Study tables and a written synthesis | A forest plot with a summary measure | Find the summary measure and see how far individual studies spread around it. |
| Main risk | Descriptive reviews can cherry-pick; defining methods before searching reduces this | Misleading if the data are not sufficiently similar or are of poor methodological quality | Weigh method quality before the headline number. |
The stages above follow the Harvard Medical School library guide and the definitions in a Rutgers University library guide.
Systematic review vs meta analysis comes down to method versus arithmetic: a systematic review gathers and summarizes studies with explicit, reproducible methods, while a meta-analysis mathematically combines results from two or more studies. Many reviews include none.
The distinction matters most when a supplement label or advert cites “a review” as proof. For the wider checklist on how to tell if a health study is reliable, start with the hub, then use this page to judge the review itself.
Reading supplement claims with both terms in mind
A claim that rests on a review is only as strong as the way that review was built. Systematic methods are defined before searching, and that is what separates a systematic review from a descriptive review and reduces cherry-picking bias, according to the Harvard Medical School library guide. A descriptive review can quietly select the studies that flatter a product.
- Look for a protocol. A systematic review starts with a research question and a protocol or research plan. A review that never states its question or plan gives you little reason to trust that the studies were chosen fairly.
- Find the summary measure. If the review includes a meta-analysis, the results are displayed in a forest plot with a summary measure. That single figure is what marketing copy usually quotes, so check what it summarizes.
- Do not stop at the label. “Review” on its own says little. Ask whether the authors describe a systematic method or simply discuss the studies they happened to know.
Pooled results can still inherit the weaknesses of the studies behind them. If you are unsure what is a confounding variable and how it could distort single studies, read that guide before trusting any combined estimate.
Which is stronger evidence
Systematic reviews sit at the top of the evidence hierarchy, a position the Harvard Medical School guide describes as widely agreed. That ranking is qualitative: it does not put a number on how much stronger they are, and it applies to well-conducted reviews.
A meta-analysis is not a separate, higher tier. It is a statistical step that a review may or may not include. Its value is precision, because it can produce a more precise estimate of a treatment effect than individual studies. That value depends on the review behind it. As the Rutgers guide puts it, readers need to be sure the methods used for the review were valid and reliable before relying on a meta-analysis result.
In practice, a carefully run systematic review without pooling can be more trustworthy than a meta-analysis built on a loose search and mismatched studies. Judge the method first and the pooled number second.
When a review has no meta-analysis
Yes, a systematic review can exist without a meta-analysis. The Harvard Medical School guide notes that a meta-analysis is not always appropriate and that many systematic reviews do not include one. Reviews that do not contain a meta-analysis can still synthesize study data.

The usual reason is heterogeneity. CASP UK explains that evaluating heterogeneity is central to a meta-analysis: if the studies differ too much, combining them could lead to skewed or misleading conclusions. The Harvard guide adds that a meta-analysis can mislead when its data are not sufficiently similar or are of poor methodological quality.
- Trust a restrained review. Authors who decline to pool studies that differ widely are applying the safeguard the method calls for, not hiding a weak result.
- Be wary of forced pooling. A neat summary figure from studies that tested different things in different ways looks precise but may not mean much.
- Expect a different kind of conclusion. Without a pooled estimate, the answer arrives in words and tables, so read how the authors weighed the studies.
How the two methods work
Building a systematic review
The team writes a question and a protocol, then searches for studies to answer it. Retrieved studies are screened for eligibility using the inclusion and exclusion criteria, with at least two people working independently. Because the rules are set out in advance, another team could repeat the process, which is the point of calling the methods explicit and reproducible.
Adding a meta-analysis
A meta-analysis is a mathematical synthesis of the results of two or more primary studies. Typical effect sizes are the odds ratio, the risk ratio, the weighted mean difference and the standardized mean difference. To interpret ratio-type results, see Relative Risk vs Absolute Risk: How To Read Medical Statistics And Make Better Health Decisions (2026 Guide) – Testpledge.
The Cochrane Handbook, as quoted in the Rutgers guide, says that combining information from all relevant studies lets meta-analyses provide more precise estimates, and that they also facilitate investigation of the consistency of evidence across studies. A meta-analysis typically includes a comparison of different studies, as a university library FAQ notes, so inconsistencies or discrepancies between them can show up.
Results appear in a forest plot. When the studies scatter widely around the summary measure, that is the visible sign of the heterogeneity the review should have discussed.
Finding reviews in Cochrane and PubMed
The Cochrane Library holds a large collection of systematic reviews on physical and mental health topics, according to a university library FAQ, which makes it a practical first stop when you want a review on a health question. Search by the ingredient or condition and check whether the review states its question and methods.
PubMed can restrict results to systematic reviews only. That option is useful when a general search returns mostly single studies. After filtering, open the abstract and look for the signs covered above: a stated question, described search and screening, and either a pooled result with a forest plot or a clear explanation of why studies were not combined.

Frequently asked questions
What is narrative synthesis, and is it the same as a narrative review?
No. Narrative synthesis combines the findings of multiple studies using words, and it can sit inside a systematic review. A narrative review is non-systematic, with no systematic attempt to address bias.
Why should subgroup and meta-regression plans be set in advance?
The Harvard Medical School guide says such plans should be described and justified before looking at the data. Choosing subgroups after seeing results makes it easier to find a pattern that suits the authors.
Does a meta-analysis prove the review behind it was sound?
No. The Rutgers guide stresses checking that the review methods were valid and reliable before trusting the meta-analysis. A precise pooled estimate does not repair a weak search or poorly matched studies.



