# Review articles and meta-analyses: when to cite which in your thesis

> Reviews, systematic reviews and meta-analyses in your thesis: when to cite which, and how to keep the primary studies behind them in view, with examples.

Acurio · 2026-10-29 · Review articles, Meta-analysis, Literature search

**Short answer:** If your sentence carries a specific figure, method or interpretation from one study, cite the primary study; if it describes a consensus or a range, cite the review. If you know the primary study only from the review, use a secondary citation.  
Consensus: … under long periods of screen work (Weber et al., 2022, systematic review).  
Secondary: Smith (2018, as cited in Weber et al., 2022)

Review articles are one of the most frequently misused source types in theses. You cite an overview, pick up a one-sentence summary of 50 studies, and then face a question in the viva: "Which of those 50 studies did you actually read yourself?" This guide shows you when a review article is the right source, when you have to switch to the primary study, and how to cite both cleanly in text and in your reference list.

## Review ≠ review: the three types that matter

The label "review" covers very different kinds of text in the literature. For your thesis it is worth keeping the three main forms apart — otherwise you cite an opinion piece as if it were systematically collected data.

### Narrative review

Expert X has worked on a topic for 20 years and writes an overview from her own experience. The study selection is hers: there is no disclosed protocol. You find narrative reviews in journal sections like *Current Opinion*, *Perspectives*, or *Lead Article*.

For a thesis: good as an entry point to a field, poor as evidence for a specific number. A narrative review tells you how someone with experience sees the field, not how it looks when you survey it systematically.

### Systematic review

Here the protocol is the main point: databases, search string, inclusion/exclusion criteria, PRISMA flow diagram. Two independent coders, documented conflicts, often pre-registered on [PROSPERO](https://www.crd.york.ac.uk/prospero/) or OSF. A systematic review is — at least in principle — repeatable.

For a thesis: a solid source for breadth ("all available studies on X show …"). But a primary study stays a primary study: if you cite *that one number* from *that one study*, you cite the study, not the review it appeared in.

### Meta-analysis

A specific systematic review that pools effect sizes quantitatively. Forest plot, heterogeneity measures ($I^2$, $\tau^2$), publication-bias check (funnel plot, Egger's test). A meta-analysis without a preceding systematic review is a warning sign — there is no transparent record of which studies were pooled.

For a thesis: the strongest source when you want to cite an effect size from the literature — provided the topic and methodological decisions match your own question.

Alongside these three there are *scoping reviews* (mapping a field without quality appraisal), *umbrella reviews* (reviews of reviews), and *rapid reviews*. During your studies, though, you will mostly meet the three above.

## Primary study or review — the decision rule

Two questions settle it:

1. **Does your sentence contain a concrete number, method, or interpretation from a single study?** → Cite the primary study.
2. **Are you describing a consensus or a range in the literature?** → Cite the review.

A fictional example from a psychology thesis (authors and figures invented):

- "Smith (2018) found a 12 % drop in concentration after four hours of screen work." → Cite **Smith 2018**, not the review you found Smith in.
- "Several studies consistently report a drop in concentration under long periods of screen work (Weber et al., 2022, systematic review)." → **Weber et al. 2022** is enough.

What you do **not** do: know Smith only through Weber, but cite Smith as if you had read the original. That is citation forgery, and it is formally resolved through [secondary citations](https://acurio.ch/en/blog/secondary-citations-done-right) when you really cannot get the primary study.

## Review as evidence for consensus — the sentence patterns

When you cite a review, make the type of evidence visible in the sentence. That helps readers and supervisors, because it makes the difference between "I read a study" and "I rely on a summary" transparent.

Three patterns you can use (fictional examples, authors and values invented):

- "A meta-analysis by Hoffmann et al. (2023) reports a pooled effect of *d* = 0.34 (95 % CI [0.21, 0.47]) across 28 primary studies."
- "A systematic review (Weber et al., 2022) identifies 42 studies on … The majority report …"
- "Narrative reviews of the topic (Mueller, 2021; Schulz, 2023) agree on … but differ on …"

The qualifiers *meta-analysis*, *systematic review*, *narrative review* are not filler — they tell the reader how robust the evidence is.

## When the review produces the number, not the primary study

Sometimes a meta-analysis computes something that is in none of the single studies: a pooled effect, a subgroup analysis, a moderator term. In that case the **meta-analysis is the primary source for that result**, even though it technically "only" synthesises other studies.

Fictional example (study and figures invented):

> The pooled prevalence of clinically relevant exam anxiety in European university students is 18.4 % (95 % CI [15.1, 21.9]; Fischer et al., 2024).

Fischer et al. *computed* that figure. You cite Fischer, not the 23 single studies that were pooled. The latter would be factually wrong — none of the 23 contains that exact 18.4 %.

## Citing Cochrane, Campbell, and the like

**Cochrane Reviews** (medicine, health sciences) and **Campbell Reviews** (education, social policy) follow their own format conventions. Two points that regularly go wrong:

- Cochrane reviews are updated continuously and receive a new version number on substantial updates. You cite the **version you used** with its DOI, not just "Cochrane Review".
- In-text in APA 7: `First author et al. (year of the version used)`. In the reference list the *Cochrane Database of Systematic Reviews* (CDSR) replaces the usual journal field.

Example of a reference-list entry in APA 7, a real Cochrane review in its current version:

> Clegg, A. J., Hill, J. E., Mullin, D. S., Harris, C., Smith, C. J., Lightbody, C. E., Dwan, K., Cooney, G. M., Mead, G. E., & Watkins, C. L. (2026). Exercise for depression. *Cochrane Database of Systematic Reviews*, 2026(1), Article CD004366. https://doi.org/10.1002/14651858.CD004366.pub7

The `.pub7` suffix on the DOI shows that you are citing the seventh published version. If you worked with the previous version, cite that one instead: Cooney et al. (2013), DOI `10.1002/14651858.CD004366.pub6`. As the example shows, the author list and even the first author can change between two versions. Equivalent conventions apply to the *Campbell Systematic Reviews*.

## Red flags in your own text

Two patterns that reliably come up in a viva:

**"Studies show …" without a reference to a review.** If you want to summarise several studies in one sentence, you need either a review that has done that systematically, or a list of the individual studies. "Studies show …" followed by a single primary study in parentheses is sloppy: either you only have that one study (then write "Smith (2018) shows …"), or you have several and should cite them.

**Citation chains through a review.** You find Smith in Weber, cite Smith directly, but only know Weber's summary of Smith. If you use a central claim, a number, or a method from Smith, you need to have read Smith. If that is not possible: either mark it as a [secondary citation](https://acurio.ch/en/blog/secondary-citations-done-right), or cite Weber instead and align your sentence with what Weber actually says.

## Finding a review — the practical workflow

1. **Search with a filter.** Google Scholar has no filter by study type. Use PubMed (filter *Article Type → Review / Meta-Analysis / Systematic Review*), Web of Science (*Document Type*), or in the social sciences PsycINFO (*Methodology → Systematic Review*). A comparison of the databases is at [Web of Science, Scopus, and PubMed](https://acurio.ch/en/blog/academic-databases-compared).
2. **Check the year.** A review from 2015 may have missed ten years of research. In fast-moving fields (AI research, mRNA vaccines) even a review that is two or three years old can miss important new studies. Look for newer primary studies or an updated version.
3. **Check the method.** A good systematic review has a PRISMA flow diagram, an open protocol, and a quality appraisal (RoB 2, GRADE, etc.). If those are missing, it is probably a narrative review — which is fine, but it should be labelled as one.
4. **Use the review as a starting point for the [snowball method](https://acurio.ch/en/blog/snowball-literature-search).** Studies that appear in two or three relevant reviews are often the key studies of your field.

## Checklist before submission

- Every meta-analysis in the text is labelled as a meta-analysis.
- Every systematic review is labelled as such — not just as "a study".
- Primary studies whose results you cite you have actually read (or marked as a secondary citation).
- Concrete numbers from a pooled effect are clearly attributed to the meta-analysis, not to one of the single studies inside it.
- Cochrane reviews are cited with their version number and DOI.

## Where Acurio helps — and where it doesn't

What Acurio checks: whether the claim you attribute to a review really appears in that review. That is often where the error sits — the review is cited correctly, but the pooled effect is reported too high, or a subgroup analysis is passed off as the overall effect.

> **Check the entries first:** Upload your thesis as DOCX or PDF. Every entry in your reference list is matched against Crossref, OpenAlex, DataCite and arXiv (books without a DOI: DNB and K10plus), and you see which sources cannot be found. [Upload your thesis and check citations](https://app.acurio.ch/quick-check?lang=en&utm_source=blog&utm_medium=body_cta&utm_campaign=reviewartikel-metaanalyse-zitieren)

What Acurio does not check: whether you classified the review correctly (narrative vs. systematic vs. meta-analysis), whether its methodological decisions fit your question, whether you should have cited a primary study instead. That stays your call as the author.

---

Review articles are tools for an overview, not shortcuts for citing. If you use them to map a field, you walk into your thesis with a strong literature base. If you use them to cite primary studies you did not read, you create citation errors that surface on the first question in the viva. Use reviews to read, not to avoid reading.


## Frequently asked questions

### Can I cite a primary study I only know from a review?

Formally yes, but only as a secondary citation, for example (fictional example) 'Smith (2018, as cited in Weber et al., 2022)'. Only Weber enters the reference list, not Smith. The better path is to read the primary study whenever you take a specific number, method, or interpretation from it — that is exactly where examiners probe during the viva.

### What is the difference between a systematic review and a meta-analysis?

A systematic review follows an open protocol (databases, search string, PRISMA flow diagram) and describes the included studies qualitatively. A meta-analysis is a specific type of systematic review that pools effect sizes quantitatively (forest plot, pooled effect, heterogeneity measure). Every meta-analysis is a systematic review, but not the other way around.

### How do I cite a Cochrane review correctly?

With the version number and DOI of the version you used, not simply as 'Cochrane Review'. In APA 7: first author et al. (year of the version used). In the reference list, 'Cochrane Database of Systematic Reviews' replaces the usual journal field, including the article number and a DOI with a version suffix like '.pub7' for the seventh published version.

---

Source: https://acurio.ch/en/blog/citing-review-articles-and-meta-analyses
