How to Choose a Meta-Analysis Topic That Can Be Published
Start from a real clinical question
The best topics come from uncertainty doctors face on the ward: two treatments with no clear winner, a new drug with mixed early results, or a guideline that says "evidence is limited". New trials published in the last two or three years are a strong signal, because they can change the answer of an older review.
Check whether it has been done
- Search PubMed for your topic with the filter "Systematic Review" or the words "meta-analysis" in the title.
- Search PROSPERO for registered reviews that are still in progress.
- If a meta-analysis exists, check its search date. If several new studies have been published since, an update can still be worth doing. If it is recent and complete, change your angle: a different population, outcome, comparator or study design.
Count the studies
Run a quick search and screen the first results roughly. As a rule of thumb, you want at least five studies that report the same outcome in a way you can pool. Fewer than that, and you may only be able to write a narrative synthesis, which is harder to publish in a good journal.
Check that studies report the numbers you need: event counts for binary outcomes, or means and standard deviations for continuous ones. Medians can be converted; see our median and IQR conversion guide.
Write it as a PICO question
- Population: who exactly? Adults, a disease stage, a procedure.
- Intervention or exposure: the treatment or factor of interest.
- Comparator: what it is compared with.
- Outcome: one clear primary outcome, measured the same way across studies.
Example: "In adults undergoing laparoscopic cholecystectomy (P), does fluorescent cholangiography (I) compared with standard white light (C) reduce bile duct injury (O)?"
Red flags
- A topic so broad it would return thousands of studies.
- Outcomes measured with many different scales.
- Almost all evidence from case reports or tiny case series.
- Three or more recent meta-analyses on the same question.
Sources
Higgins JPT et al. Cochrane Handbook for Systematic Reviews of Interventions, chapters 2 and 3. Page MJ et al. The PRISMA 2020 statement. BMJ 2021;372:n71.
Frequently asked questions
How many studies do I need for a meta-analysis?
Technically two, but most journals and readers expect more. Aim for at least five comparable studies on the primary outcome.
Can I do a meta-analysis of observational studies?
Yes, if they are comparable and you assess risk of bias carefully, for example with ROBINS I or Newcastle Ottawa.
Is updating an old meta-analysis publishable?
Often yes, when enough new studies have appeared since the last search date to change or strengthen the conclusion.
Related
Register your review on PROSPERO · Search builder · All guides
