How to build a PubMed search strategy for a systematic review
1. Choose your concepts
Use the PICO elements of your question. Search the population and the intervention; leave comparators and outcomes out unless they are essential, because they are often missing from abstracts. Example question: metformin versus insulin for gestational diabetes. Concepts: gestational diabetes; metformin.
2. Find MeSH headings
MeSH (Medical Subject Headings) is the controlled vocabulary the US National Library of Medicine uses to index MEDLINE. Look up each concept in the MeSH database. Here: "Diabetes, Gestational"[Mesh] and "Metformin"[Mesh]. In PubMed, [Mesh] includes narrower headings automatically.
3. Collect free-text synonyms
Not every record is indexed with MeSH, especially recent ones. Add the words authors actually use, searched in the title and abstract with [tiab]: synonyms, abbreviations, brand names, UK and US spellings, and hyphenated variants. Put phrases in quotes. Use truncation (*) carefully, only after a distinctive stem, to catch word endings.
4. Combine with Boolean logic
OR the terms within each concept, then AND the concepts together. For the example question, the free search strategy builder produces:
("Diabetes, Gestational"[Mesh] OR "gestational diabetes"[tiab] OR "gestational diabetes mellitus"[tiab] OR "pregnancy-induced diabetes"[tiab] OR "pregnancy induced diabetes"[tiab]) AND ("Metformin"[Mesh] OR metformin[tiab] OR glucophage[tiab] OR dimethylbiguanidine[tiab])5. Add a study design filter if needed
If your review includes only randomised trials, a validated filter keeps the search sensitive while removing obviously irrelevant designs. The builder can add the widely used Cochrane highly sensitive search strategy for PubMed:
AND ((randomized controlled trial[pt] OR controlled clinical trial[pt] OR randomized[tiab] OR placebo[tiab] OR drug therapy[sh] OR randomly[tiab] OR trial[tiab] OR groups[tiab]) NOT (animals[mh] NOT humans[mh]))Check the current Cochrane Handbook technical supplement for the latest version of this filter, and do not use design filters in databases where they are not validated.
6. Test, document and peer review
- Known-paper test: list 3–5 papers you know are eligible and confirm the search retrieves all of them. When we ran the example search above in PubMed in October 2026 it retrieved all three known trials we tested.
- Read the search details: PubMed's Advanced search history shows how each term was interpreted. Fix anything unexpected.
- Translate, do not copy: Embase, Scopus, Web of Science and CENTRAL use different syntax and vocabularies (Embase uses Emtree, not MeSH).
- Document: record the database, platform, date, the exact strategy and the number of results for every search. The builder's search log exports this as CSV for your methods section.
- Peer review: ask an information specialist to review the strategy, ideally using the PRESS 2015 checklist.
Sources
Lefebvre C, Glanville J, Briscoe S, et al. Chapter 4: Searching for and selecting studies. In: Cochrane Handbook for Systematic Reviews of Interventions. McGowan J, Sampson M, Salzwedel DM, et al. PRESS Peer Review of Electronic Search Strategies: 2015 guideline statement. J Clin Epidemiol 2016;75:40–46. PubMed User Guide.
Frequently asked questions
What is the difference between [Mesh] and [tiab] in PubMed?
[Mesh] searches the Medical Subject Heading assigned by indexers and includes narrower headings by default. [tiab] searches words in the title and abstract. A sensitive search combines both, because new or poorly indexed records may not have MeSH headings yet.
Should I use AND or OR between my terms?
Use OR between synonyms within one concept, and AND between different concepts. For example (diabetes terms) AND (metformin terms).
Should I include outcomes in my search?
Usually not. Outcomes are often poorly described in titles and abstracts, so adding them can miss relevant studies. Most systematic review searches combine population and intervention, sometimes with a study design filter.
How do I know my PubMed search is good enough?
Check that it retrieves a set of known relevant papers, review the search details PubMed shows for each term, and ask an information specialist to peer review it, for example with the PRESS checklist.