PubMed Search Tutorial for Literature Reviews

PubMed is the most widely used free biomedical database in the world and, for most health-sciences literature reviews, the single most important database to search well. This tutorial walks through PubMed's interface, syntax, controlled vocabulary, and export workflow so you can build a reproducible search that stands up to peer review.

About PubMed

PubMed is a free search interface maintained by the U.S. National Library of Medicine (NLM) at the National Institutes of Health. It provides access to more than 35 million citations drawn from three primary sources: MEDLINE (the NLM's flagship indexed biomedical database), PubMed Central (PMC, an open-access full-text archive), and publisher-supplied records that have not yet been fully indexed for MEDLINE.

Coverage is strongest in clinical medicine, biomedical research, nursing, public health, dentistry, veterinary medicine, and the life sciences. Journals indexed for MEDLINE are hand-selected by the Literature Selection Technical Review Committee and are curated with MeSH (Medical Subject Headings) — a hierarchical controlled vocabulary maintained by the NLM. This indexing is what makes PubMed so powerful for systematic reviews: once a record is fully indexed, you can retrieve it by concept rather than by the specific terminology an author happened to use.

PubMed's strengths for reviewers include:

• Free, global access with no institutional login required
• Deep historical coverage of the biomedical literature
• Rich controlled-vocabulary indexing via MeSH
• Stable linking to PMC full text for open-access articles
• Robust filters for study type, species, age, and publication date
• Direct export to RIS, NBIB, and CSV formats

It is less strong for conference proceedings, grey literature, humanities and social sciences sources, and very recent records that have not yet been MeSH-indexed. For those, plan to complement PubMed with another database such as Scopus or CINAHL.

Getting started

PubMed is free at pubmed.ncbi.nlm.nih.gov. You do not need an institutional login, but creating a free NCBI account is worthwhile: it lets you save searches, set up email alerts, create collections, and keep a search history that persists across devices. If your institution has a subscription to a link-resolver such as LibKey or SFX, you can also configure PubMed to show "institutional access" full-text buttons under the Outside Tool settings.

The default search box at the top of PubMed performs Automatic Term Mapping (ATM), which silently translates your plain-language query into a more complex expression using MeSH terms and field tags. ATM is convenient but opaque — for a reproducible review, you should always inspect and, where necessary, override it. Click Advanced beneath the search box to open the Advanced Search Builder, where you can:

• Build queries field by field
• Inspect the full system-generated query string
• Combine previous searches from your History using AND, OR, NOT
• Export the search history for your review protocol

For systematic reviews, the Advanced Search Builder is the primary workspace. Your final search string and the exact run date should be recorded verbatim in your search strategy documentation.

Building your search

PubMed supports standard Boolean operators — AND, OR, NOT — in uppercase. Parentheses control precedence, and PubMed evaluates left to right within a precedence level, so always parenthesize OR groups explicitly.

Phrase searching uses straight double quotes: "type 2 diabetes". If PubMed does not recognize the phrase, it will drop the quotes and search the words separately, so check your Search Details panel to confirm it was honored.

Truncation uses the asterisk * at the end of a word stem: nurs* retrieves nurse, nurses, nursing. PubMed requires at least four characters before the * and expands to the first 600 variants. There is no mid-word wildcard, and truncation disables Automatic Term Mapping for that term — so a truncated keyword will not map to MeSH.

Field tags are appended in square brackets and restrict a term to a specific record field. The tags most relevant to reviewers are:

• [tiab] — title and abstract
• [tw] — text word (title, abstract, MeSH, other terms)
• [MeSH] — MeSH Major and Minor headings
• [MeSH:noexp] — MeSH heading without automatic explosion to narrower terms
• [majr] — MeSH Major Topic only
• [dp] — date of publication (e.g., "2015"[dp]:"2025"[dp])
• [pt] — publication type (e.g., "randomized controlled trial"[pt])
• [au] — author
• [la] — language

PubMed has no true proximity operator; the closest equivalent is phrase searching plus careful use of [tiab] to keep concepts in the same field.

Using controlled vocabulary

MeSH is what separates a keyword search from a properly indexed search. When an NLM indexer catalogs a MEDLINE record, they assign 10–15 MeSH headings that describe the article's content. Searching on a MeSH term therefore retrieves articles about that concept regardless of the author's word choice.

To find the right MeSH term, use the MeSH Database (linked from the PubMed homepage under "Explore"). Enter your concept, review the definition and tree structure, and inspect the Entry Terms — these are synonyms that map to the heading. The MeSH Database can also build a search for you with the correct field tag applied.

Two behaviors are worth understanding explicitly:

1. Explosion. By default, a MeSH search automatically includes all narrower terms in the hierarchy. Searching "Neoplasms"[MeSH] retrieves articles indexed under Neoplasms or any of its hundreds of narrower children. Use [MeSH:noexp] to disable this if you want a specific level only.
2. Major topic. Add [majr] to restrict to articles where the concept is a central focus, not just mentioned in passing.

For a comprehensive search, combine MeSH with free-text using OR: MeSH captures fully indexed records, and [tiab] captures recent or not-yet-indexed records. This belt-and-braces pattern is the standard for systematic reviews.

Example search strings

Filters and limits

The left sidebar of the results page exposes PubMed's filters. The most useful for reviewers are:

• Publication date — custom ranges using [dp] in the query are usually more reproducible than the sidebar sliders
• Article type — Randomized Controlled Trial, Systematic Review, Meta-Analysis, Review, Clinical Trial, Observational Study
• Species — Humans vs Other Animals (essential when pre-clinical work will dilute your results)
• Language — limit or exclude non-English records, depending on your protocol
• Age group — Infant, Child, Adolescent, Adult, Aged
• Sex — Female, Male
• Text availability — Abstract, Free full text, Full text

Apply filters after you have finalized the core Boolean query, and document every filter you enable. Activated filters are sticky across searches in the same session — remember to clear them before running a different concept.

Exporting results to reference managers

From a results page, use the Send to button above the list. For reference managers, select Citation manager to download a .nbib file (compatible with EndNote, Zotero, Mendeley, and RefWorks) or File and choose PubMed format for plain text. You can send up to 10,000 records at a time.

Typical workflows:

• EndNote desktop: Direct .nbib import via File > Import > File, selecting the PubMed (NLM) filter
• Zotero: With the Zotero Connector installed, click the folder icon in the browser toolbar and tick the records to save; alternatively import the .nbib file directly
• Mendeley Reference Manager: Use Add new > Import library > RIS after saving your results in RIS format

Once your citations are captured, move them into your screening workflow. If you are organizing hundreds of sources, plan ahead for how you will extract and compare them — a literature review matrix is the standard tool.

Tips and pitfalls

• Inspect the Search Details panel. PubMed's Automatic Term Mapping can silently change your query. Always confirm the translated query before trusting a result set.
• Combine MeSH with free text. MeSH alone will miss the most recent records (indexing lags 1–3 months); free text alone will miss older records where terminology has shifted.
• Parenthesize OR groups. A OR B AND C is not the same as (A OR B) AND C; PubMed evaluates left to right within precedence.
• Document run dates. PubMed adds records daily. Record the exact date and full query string in your protocol and PRISMA flow diagram.
• Do not rely on PubMed alone. For any formal systematic review, cross-check against at least one other database — Scopus for multidisciplinary coverage or CINAHL for nursing and allied health topics.