Is a Meta-Analysis a Primary or Secondary Source? How to Cite and Classify It

The question of whether a meta-analysis counts as a primary or secondary source comes up regularly in academic writing courses, manuscript preparation, and heated disagreements between co-authors at the revision stage. The honest answer is that the classification depends on what you are citing the meta-analysis for, and understanding that distinction matters more than memorizing a single label.

In most formal definitions, a meta-analysis is a secondary source. It does not collect new data from patients, experiments, or observations. Instead, it gathers, appraises, and statistically pools data from studies that did. The primary sources are those underlying studies. The meta-analysis re-analyzes their combined output to produce a more precise estimate than any single trial could provide. By the traditional library-science definition, anything that synthesizes or analyzes earlier work rather than reporting firsthand data is secondary literature, and a meta-analysis fits that description cleanly.

That is the tidy answer for a research methods course. In practice, the picture is more layered, and medical journal editors know it.

What Primary and Secondary Actually Mean in Research

The terms primary and secondary source do not have completely fixed meanings across all academic disciplines. In historical research, a primary source is a document or artifact from the period under study. In law, primary sources are statutes and case law. In biomedical research, the distinction is rooted in who collected the data and how.

A primary study in biomedicine reports data collected directly by the researchers. Randomized controlled trials, cohort studies, case-control studies, cross-sectional surveys, laboratory experiments, and case reports all generate original data and are therefore primary sources. The researchers interacted with the subject matter, measured outcomes, and recorded what they found. The paper is the first formal account of that data.

A secondary source, in this framework, takes existing data or existing literature as its input. Narrative reviews, systematic reviews, meta-analyses, editorials, practice guidelines, textbooks, and evidence summaries are all secondary. Their value lies in synthesis, interpretation, and organization rather than in original data collection. This does not make them less rigorous. A well-conducted Cochrane meta-analysis requires more methodological precision than most individual trials. But it is still working with borrowed material.

There is also a tertiary category, though it is rarely discussed explicitly in medical writing. Tertiary sources compile and organize secondary material: textbooks, encyclopedias, and clinical guidelines that draw on existing systematic reviews fall here. For most purposes in academic medical writing, you are navigating between primary and secondary.

Where Meta-Analyses Sit in the Evidence Hierarchy

The evidence hierarchy is a concept most researchers encounter early in their training and rely on throughout their careers. At the top, at least in questions about treatment effectiveness, sit well-conducted systematic reviews and meta-analyses of randomized controlled trials. This ranking reflects statistical power and bias control, not source classification. A meta-analysis of high-quality trials produces a more reliable estimate of treatment effect than any single trial, because it pools sample sizes and reduces the influence of chance findings in individual studies.

This positioning causes some of the confusion about source classification. If meta-analyses are the highest form of evidence, how can they be secondary? The answer is that evidence quality and source type are different dimensions. A meta-analysis ranks high on evidence quality precisely because it synthesizes multiple primary sources. Its strength comes from its secondary character, from the fact that it draws on many underlying trials rather than just one. Strength and originality are not the same thing.

The GRADE framework, which is used by Cochrane, many clinical guideline bodies, and journals including the BMJ and JAMA, formalizes this by rating evidence quality separately from study design. A meta-analysis of randomized trials starts at high quality and can be downgraded for risk of bias, inconsistency across studies, imprecision, indirectness, or publication bias. A meta-analysis of observational studies starts at lower quality. The starting point reflects study design within the primary literature that was synthesized, not the meta-analysis itself. Understanding this keeps you honest when you cite a meta-analysis as strong evidence: you are relying on the quality of the underlying trials as much as on the synthesis.

When a Meta-Analysis Functions as Your Primary Evidence

Here is where the practical nuance sits. When you write the introduction or discussion of your own paper, you regularly cite meta-analyses for their conclusions. You might write: a 2024 Cochrane review found a pooled reduction in systolic blood pressure of 8 mmHg with a particular drug class across 22 trials. You are not citing any of those 22 trials individually. You are citing the meta-analysis as the source of that estimate.

In this context, the meta-analysis is functioning as your primary evidentiary source. You are relying on it directly to support a claim. That is what makes it feel like a primary source even though it is technically secondary. The confusion arises because people conflate the type of a source with the role it is playing in a specific argument. A systematic review is secondary by type. But if it is the actual evidence you are using, it is primary to your argument.

Medical journals, for their part, generally expect authors to cite the best available evidence for any factual claim. If a well-conducted meta-analysis exists on the question, citing only a single supporting trial when you could have cited the review is poor practice. Reviewers sometimes push back: why are you citing one small observational study when a Cochrane review covers the same comparison? This is an argument for using the meta-analysis as your primary citation, even while classifying it correctly as secondary by design.

Is a Systematic Review a Primary Source? The Same Logic Applies

Systematic reviews without a quantitative meta-analysis follow the same classification. A systematic review that synthesizes evidence narratively is secondary literature. It collects, appraised, and organizes existing studies without generating original data. But a systematic review of qualitative studies, for example, may use thematic synthesis to develop a conceptual framework that did not previously exist in the literature. That framework is a new contribution. In that specific sense, the review generates something primary even though it is a secondary study type.

The formal distinction matters for one particular reason in academic writing: how you handle secondary use in your own methods section. If your literature review for a new trial draws on previous meta-analyses to characterize the current evidence base, you are using those meta-analyses appropriately as secondary literature. If you then run a new meta-analysis that incorporates the same primary trials those reviews included, you need to be transparent about overlap. PRISMA 2020 guidance, along with the 2022 update to PROSPERO registration requirements, addresses this directly: reviewers should list included studies, identify any overlap with previously registered reviews, and explain why a new review adds something distinct.

Journals increasingly ask authors of systematic reviews and meta-analyses to justify why a new synthesis is needed at all. That justification almost always involves describing the limitations of previous meta-analyses, which requires engaging with them as secondary literature rather than as settled facts.

How to Cite a Meta-Analysis Correctly in Your Manuscript

Correct citation of a meta-analysis depends on what you are attributing to it. The most common error is over-attributing. Authors sometimes write as if a meta-analysis proved a finding, when what it did was pool estimates from trials that individually tested the intervention. The meta-analysis refined the estimate. The trials did the proving.

A clean attribution names what the meta-analysis contributed: its pooled result, its confidence interval, its assessment of heterogeneity, or its conclusion about evidence quality. You might write: a 2023 meta-analysis of 15 randomized trials (n = 8,400) found a pooled risk reduction of 23% for the primary endpoint (RR 0.77, 95% CI 0.68 to 0.87). That phrasing attributes the estimate to the review without suggesting the meta-analysis ran the trials itself.

When your manuscript is itself a meta-analysis, the citation question becomes even more specific. You are required to cite the individual primary studies in your methods and results sections. Your references section will contain those trials, not just other meta-analyses. A common mistake among less experienced authors is to cite a prior meta-analysis in place of the trials it covered, which hides the primary evidence from readers and makes replication of your search and selection process impossible. PRISMA requires a full list of included studies, and the reference list should reflect that.

The Role of Pre-Registration in Distinguishing Primary from Secondary Work

One shift in contemporary medical publishing is that meta-analyses are increasingly expected to be registered prospectively, just as clinical trials are. PROSPERO, the international prospective register of systematic reviews maintained by the UK National Institute for Health and Care Research, allows authors to register a review protocol before data extraction begins. This makes the secondary character of the work explicit: you announce that you plan to synthesize existing evidence before you start, reducing the scope for selective reporting of which outcomes or analyses you emphasize.

The journals that require PROSPERO registration, which now includes many titles in the BMJ stable, JAMA, and Annals of Internal Medicine, are signaling that even secondary work should be transparent about its methods from the outset. This is the same logic that drives trial registration for primary studies. A well-conducted meta-analysis with a pre-registered protocol and a PRISMA-compliant report has more credibility than one that materializes without any public record of what was planned. The registration does not change the type of the source. But it raises its evidence quality by reducing the risk that the synthesis was shaped by what the authors found.

For readers encountering a meta-analysis in the literature, the presence of a PROSPERO registration number and a PRISMA checklist is a reasonable signal that the secondary work was conducted to a standard that supports treating its findings as reliable. The absence of those markers is not automatically a problem, especially for older reviews, but it should prompt a closer look at whether the outcome selection and analysis plan were transparent.

Using Meta-Analyses in Your Introduction and Discussion

The introduction of a clinical paper typically establishes what is known and identifies the gap that motivates the current study. Meta-analyses are the most defensible citations for "what is known" claims, because they represent synthesized evidence rather than a single study that may or may not be representative. When you write that prior work shows a consistent association between a risk factor and an outcome, citing a meta-analysis of 20 well-conducted cohort studies is stronger than citing one of those cohort studies in isolation.

The discussion section is where the source type distinction becomes important again. If your primary study finds a result consistent with an existing meta-analysis, you should say so and specify what the meta-analysis found. If your result diverges, you should discuss why. Possible explanations include different patient populations, different outcome definitions, different follow-up durations, or a longer recruitment period that captured a different phase of practice. That analysis requires engaging with the meta-analysis as secondary literature whose methods you scrutinize, not simply as settled fact you quote.

One place authors get into trouble is when they cite a meta-analysis with high heterogeneity to support a single-direction claim. High heterogeneity, often flagged by an I-squared statistic above 50 to 75 percent depending on the field's conventions, means the underlying studies are not telling a consistent story. Citing the pooled point estimate from such a review as if it settles the question is misleading. A reviewer who knows the field will notice. The honest framing acknowledges the heterogeneity and notes that the review suggests a trend without resolving it.

When Instructors and Editors Mean Different Things

Much of the confusion around this question arises because researchers carry the definition they learned in a methods course into manuscript preparation, where it does not always apply the same way. In a research methods course, the distinction between primary and secondary literature is used to teach students about study design and the structure of scientific knowledge. In that context, the answer is clear: a meta-analysis is secondary.

A journal editor reviewing your manuscript is not applying that pedagogical framework. They are asking whether your claims are adequately supported. A reviewer asking you to strengthen your evidence base by citing better literature is not questioning source classification. They are asking whether you are using the best available evidence, which often means preferring a high-quality meta-analysis over a handful of individual studies. The answer to that question is the same regardless of whether you classify the meta-analysis as primary or secondary in a course assignment.

Students writing dissertations or theses sometimes face a different version of this question from their supervisors or institutional style guides. Some guidelines explicitly state that secondary sources should be avoided where possible, and students interpret this as a prohibition on citing meta-analyses. That interpretation misunderstands the rule. The prohibition on secondary sources in that context typically means: do not cite a review article that discusses a primary study when you could go and read the primary study yourself. It does not mean that a systematic review, which is the culmination of a formal methodological process, should be replaced with one of the studies it synthesized. A Cochrane review is not a secondary source in the pejorative sense that a textbook chapter or a popular science article is.

What Journals Ask When You Submit a Meta-Analysis

If you are submitting a meta-analysis as your paper rather than citing one in it, the classification question shifts from citation practice to reporting requirements. Journals that publish systematic reviews and meta-analyses have specific requirements, and those requirements are built around the secondary character of the work. You are expected to describe where you searched, how you selected studies, what quality assessment tools you used (such as the Cochrane Risk of Bias 2 tool or the Newcastle-Ottawa Scale), and what statistical methods you applied for the pooling.

PRISMA 2020, the Preferred Reporting Items for Systematic Reviews and Meta-Analyses, is the reporting standard used by the vast majority of journals that publish these study types. It includes a 27-item checklist and a four-phase flow diagram showing records identified, screened, assessed for eligibility, and included. Most journals require the PRISMA checklist at submission. Many also require a supplementary search strategy, showing the exact database queries and date ranges. These requirements exist because a meta-analysis is secondary work: its validity depends entirely on the transparency and completeness of how it handled its primary sources.

A practical implication for authors: if you are preparing a meta-analysis for submission, read the target journal's instructions carefully to determine whether it also requires a PROSPERO registration number, a protocol paper, or a completed PRISMA-S extension for literature search reporting. Different journals apply these requirements to different extents, and missing one can stall a submission at the administrative review stage.

A Short Summary for Practical Use

The underlying principle is that classification tells you something about how a paper was made, not about how reliable it is or how you should use it. A well-conducted meta-analysis of large, well-designed randomized trials is among the most valuable evidence in medicine precisely because it is secondary, because it combines what multiple research teams found rather than relying on any one of them. Knowing that it is secondary should make you more attentive to what it synthesized, not less likely to cite it.

When you write your next introduction and reach for the best available evidence on the question you are studying, the right citation will often be a meta-analysis. Use it confidently. Acknowledge its limitations honestly. And if a reviewer asks why you did not go back to the original trials, you now have a considered answer.

Further Reading