Free scoping read. A written assessment of your manuscript in two business days, at no cost.

Start one
Contents

Academy · Publication support · Finalizing MeSH terms

Finalizing MeSH (Medical Subject Headings) Terms: Guidelines for Searching Articles

MeSH is a versioned, hierarchical thesaurus with five kinds of record. Every descriptor on this page was read from the MeSH Browser and links to its own entry.

Pubrica AcademyWritten by a research writing specialist Reviewed bya research methodology reviewer Published15 October 2026 Last checked15 October 2026 Reading time14 minutes Sources11 primary

Medical Subject Headings, or MeSH, is the National Library of Medicine’s controlled vocabulary for the biomedical literature — “a controlled and hierarchically-organized vocabulary produced by the National Library of Medicine.”1 Finalising a set of MeSH terms involves selecting the headings that exist, checking them against the hierarchy, pairing them with allowable qualifiers, and determining the free text required alongside them.

Sources of the terms on this page

Every descriptor string, unique identifier and tree number on this page was read from the MeSH Browser, and each one links to its own record so that it can be verified. Where a figure changes from year to year, the year is given with it.

Definition of MeSH

MeSH is a thesaurus, not a keyword list. NLM describes it as “The Medical Subject Headings (MeSH®) thesaurus is a controlled vocabulary produced by the National Library of Medicine.”2, maintained by subject specialists who “continually updated by subject specialists” it, with “Each year hundreds of new concepts are added.”2

Two release schedules run in parallel, and the difference matters more than it first appears. Descriptors and qualifiers “MeSH Descriptors and MeSH Qualifiers are released annually to coincide with the calendar year.”3 Supplementary concept records “are updated continually and released on a daily basis”3

So the vocabulary has a year. If you are recording a search strategy for a systematic review — or for anything you may be asked to reproduce — the MeSH year belongs in the record alongside the date you ran it. A heading that existed in 2019 may have been changed or removed since, and a heading you need may not have existed yet.

That total is for the 2021 vocabulary, the most recent figure that the National Library of Medicine (NLM) publishes on an accessible page. No current-year total is given here, because the number changes every year and an outdated figure presented as current would be misleading. For the scale of annual change, NLM's technical bulletin for the 2024 vocabulary recorded 311 new descriptors, 1,044 modified and one deleted.

The five kinds of MeSH record

Most of the confusion about MeSH comes from treating it as one kind of thing. It is five, and they behave differently in a search.

Table 1. MeSH record types and the tag that retrieves each

Record typeSearch tagWhat it is
Descriptor, or main heading[mh]The subject term itself — what the article is about. These are the records arranged in the hierarchy.
Entry term—A synonym or closely related term that cross-refers to a descriptor. Entry terms are not searched as headings; they are the means by which the appropriate heading is found.
Qualifier, or subheading[sh]Narrows a descriptor to an aspect of it — therapy, diagnosis, epidemiology. There are 77, and each descriptor has its own allowable list.
Supplementary Concept Record[nm]Chemicals, drugs and some rare diseases. These sit outside the tree hierarchy and map to a descriptor rather than appearing in it.
Publication type[pt]What the item is rather than what it is about — Randomized Controlled Trial, Review, Case Reports.

NLM’s own framing of the distinction that matters most: descriptors “characterize the subject matter or content” of an article, while publication types “indicate what the indexed item is, i.e., its genre, rather than what it is about”4 That is why Randomized Controlled Trial[pt] and Randomized Controlled Trials as Topic[mh] are different searches. The first finds trials. The second finds articles about trials — methodology papers, commentary, guidance. Confusing the two is a recurring error in search strategies, and it is a large one, because the second will not find you the evidence.

Entry terms are the other piece worth understanding properly. They are “synonyms or closely related terms that are cross-references to descriptors.”4 They are not headings. You search them to find the heading, and then you search the heading.

Term look-up and inverted descriptor forms

The MeSH Browser is the authority5. Three real records, read from it, show what you are looking for and why guessing does not work:

Table 2. Three MeSH descriptors as they actually appear

DescriptorUnique IDTree numbersDate introducedEntry terms
Diabetes Mellitus, Type 2D003924C18.452.394.750.149
C19.246.300
(both)
1984/01/01Type 2 Diabetes; NIDDM; Diabetes Mellitus, Non-Insulin-Dependent
(3 of 17)
Diet, MediterraneanD038441E02.642.249.278.250
G07.203.650.240.278.250
(both)
2003/01/01Mediterranean Diet
(the only one)
Glycated HemoglobinD006442D09.400.205.625.500.250
D12.776.124.400.354
(2 of 6)
1982/01/01Hb A1c; Hemoglobin, Glycosylated; Glycohemoglobin
(3 of 23)

The most common error

Note the inverted forms. The descriptor is Diet, Mediterranean and not “Mediterranean Diet”; Glycated Hemoglobin and not “Hb A1c”;Diabetes Mellitus, Type 2 and not “Type 2 Diabetes”. Those natural phrasings are entry terms, through which the heading is found; the heading itself is what is searched. Entering a descriptor from memory rather than looking it up is the most common reason a MeSH search retrieves nothing without any warning.

Note also that each descriptor carries a date introduced. Diet, Mediterranean was introduced in 20035. An article about Mediterranean diets published in 1998 cannot carry that heading, because the heading did not exist when the article was indexed. For any search reaching back more than a few years, the date introduced tells you where the controlled vocabulary stops being able to help you and free text has to take over.

Two things in that table are worth reading as warnings rather than as reference data. First, Glycated Hemoglobin sits at six tree numbers, of which two are shown5: a substance that is a blood protein, a biomarker and an analyte appears in several branches at once, and exploding it from one branch is not the same search as exploding it from another. Second, that record carries 23 entry terms, among them Hb A1c and Hemoglobin, Glycosylated5. A descriptor with that many synonyms is one where your own phrasing is unlikely to be the heading — and where the free-text half of your search has a lot of variants to cover.

The MeSH hierarchy and the effect of exploding a term

Descriptors are “organised into a numbered tree structure or hierarchy.”4 Those tree numbers in Table 2 are addresses in that hierarchy, and a descriptor can have more than one — Diabetes Mellitus, Type 2 appears in two places, under endocrine conditions and under nutritional and metabolic diseases5.

When you search a descriptor, PubMed “PubMed automatically includes narrower Descriptors indented under it in the MeSH Tree Structures.”6 “This type of search is sometimes called 'exploding' the Descriptor,”6 and it happens by default. NLM's own example: a search for Pneumoconiosis also retrieves Asbestosis and Silicosis, because both sit beneath it in the tree.6

To prevent it, “the search is limited with the tag '[mh:noexp]'.”6 That is occasionally what you want — if the narrower terms are a different clinical question — but it is almost always the wrong choice in a systematic review, where recall is the priority.

Major topics. A MeSH Major Topic is “A MeSH Major Topic is one of the main topics discussed in an article”7, and it “is denoted by an asterisk (*) on the MeSH term or MeSH/Subheading combination”7 on the record. “To search for a MeSH Heading as a MeSH Major Topic, use the [majr] tag.”7 One nuance NLM flags that is easy to miss: where the asterisk sits on a heading-and-subheading combination indicates which concept is a main topic, but “This placement does not affect PubMed searching.”7

Qualifiers and their allowable combinations

“There are 77 topical Qualifiers (also known as Subheadings)”4, and “Qualifiers can be used in conjunction with appropriate Descriptors,”6 with the syntax adverse effects [sh] AND Anticoagulants [mh]6.

The allowable qualifiers above are those listed on the Diabetes Mellitus, Type 2 record itself.5

Table 3. Allowable qualifiers on Diabetes Mellitus, Type 2 (D003924)

therapy (TH)diagnosis (DI)drug therapy (DT)
diet therapy (DH)epidemiology (EP)prevention & control (PC)
blood (BL)complications (CO)etiology (ET)
genetics (GE)metabolism (ME)mortality (MO)
surgery (SU)urine (UR)

Non-interchangeability of qualifiers

Note what is not on that list: adverse effects. It is a valid qualifier, but it is not allowable on a disease descriptor — it attaches to drugs and procedures. Each descriptor carries its own allowable list, so a qualifier that is valid with one heading may be invalid with another, and an invalid pairing retrieves nothing rather than returning an error.

Processing of untagged searches in PubMed

Typing a phrase into PubMed does not search that phrase. “PubMed optimizes searches using a process known as Automatic Term Mapping (ATM).”9 The process “compares terms from the search query with lists of terms”9 in a sequence of translation tables, and where it “includes an untagged term that maps to a MeSH term”9, “the term will be searched as the MeSH term as well as a Text Word”9

That is usually helpful, and it is why a plain search in PubMed often works better than people expect. But it is also why three common habits silently switch the controlled vocabulary off:

Table 4. Three things that turn automatic term mapping off

What you typedExampleWhat happens
A field tagdiabetes[tiab]Restricts to the tagged field only, so no MeSH heading is added.
A wildcarddiabet*Switches mapping off entirely, including the MeSH expansion.
Quotation marks"type 2 diabetes"Searches the phrase as written, with no MeSH heading added.

On field tags, the user guide is explicit: “Search field tags turn off Automatic Term Mapping (ATM), limiting your search to the specified term only”8 On wildcards, “Wildcards turn off Automatic Term Mapping (ATM) and the process that includes the MeSH term.”8 And a quoted phrase “will not perform automatic term mapping that includes the MeSH term”8

Each of these is sometimes appropriate, but none is neutral. A search built entirely from quoted phrases and field tags is a free-text search with no controlled vocabulary in it at all, whatever the author intended; a strategy must therefore pair the two deliberately rather than assume that PubMed adds headings automatically.

Limitations of MeSH as the sole search vocabulary

The Cochrane searching chapter’s advice is that “searches should comprise a combination of subject terms selected from the controlled vocabulary or thesaurus” “with a wide range of free-text terms” 10 — and it gives its reasons, which are worth reproducing accurately because they are not the ones usually offered.

It advises that “the extent to which subject terms are applied to references should be viewed with caution”10, because “indexers are not always experts in the subject areas”10 and because “the available indexing terms might not correspond to the terms the searcher wishes to use”10

A correction regarding the rationale for free-text searching

The usual explanation of why free text is necessary does not precisely reflect the sources. The Cochrane searching chapter's reasons are that indexing is applied inconsistently and that the available terms may not match those a searcher requires — not that recent articles are unindexed. It makes the opposite temporal point, that earlier articles are harder to identify than recent ones. The indexing lag is real and is documented separately by NLM: records enter PubMed from publishers and as in-process citations before MeSH headings have been applied. Both reasons hold; they are distinct reasons and derive from different sources.

The lag itself is documented by NLM. Publisher-supplied and in-process citations are in PubMed before indexing has happened: “they are not indexed with MeSH and the other value-added data entered”11, and in-process records “are preliminary (not yet quality controlled) citations”11 The causes include “the time lag in receipt of the journals after their actual publication as well as the time needed to index”11

The practical upshot for anyone finalising a search: the controlled vocabulary and the free text are doing two different jobs. MeSH gives you the articles an indexer recognised as being about your topic, including ones whose authors used words you did not think of. Free text gives you the articles that are not yet indexed, the ones indexed inconsistently, and the ones published before your heading existed. Dropping either leaves a predictable gap.

Summary of PubMed search tags

Table 5. MeSH-related search tags in PubMed

TagFieldWhat it retrieves
[mh]MeSH TermsThe descriptor, with its narrower terms included — exploded by default.
[mh:noexp]MeSH Terms, no explosionThe descriptor only, without the narrower terms beneath it.
[majr]MeSH Major TopicOnly where the descriptor is one of the article's main topics.
[sh]MeSH SubheadingA qualifier, combined with an appropriate descriptor.
[pt]Publication TypeWhat the item is — Randomized Controlled Trial, Review, Case Reports.
[tiab]Title/AbstractFree text in the title and abstract. Not MeSH, and the other half of a thorough search.

Finalising a set of MeSH terms

Practical application

  1. Look up every heading in the MeSH Browser rather than entering it from memory. The inverted forms — Diet, Mediterranean, Glycated Hemoglobin — are not what a searcher would predict.
  2. Record the MeSH year with the strategy, not only the date on which it was run. Descriptors are released annually, so the vocabulary is versioned.
  3. Check the year of introduction of each descriptor. If the search extends to earlier years, free text must cover those years.
  4. Read the allowable qualifier list on the descriptor’s own record before pairing a subheading with it. An invalid pairing returns nothing rather than an error.
  5. Decide explode or [mh:noexp] deliberately for each heading. Exploding is the default and is usually appropriate for a review.
  6. Keep [majr] out of a systematic review search. It sacrifices recall for precision, and recall is the property a systematic search must protect.
  7. Distinguish [pt] from [mh] for study designs. Randomized Controlled Trial[pt] finds trials; Randomized Controlled Trials as Topic[mh] finds articles about trials.
  8. Pair every concept with free text in the title and abstract, and report this in the methods. Controlled vocabulary alone has documented gaps.
  9. Check in the Search Details how the query was executed. A quoted phrase or a wildcard will have switched off the MeSH expansion.

Key points

MeSH is a versioned, hierarchical thesaurus with five kinds of record, not a keyword list. Headings should be looked up rather than guessed, because the forms are inverted and each descriptor has a year of introduction and its own list of allowable qualifiers. Exploding is applied by default. Quotation marks, wildcards and field tags all switch off the MeSH mapping. Controlled vocabulary alone is insufficient: indexing is inconsistent, the available terms may not match those of the searcher, and the newest records are not yet indexed. Every concept therefore requires free text alongside its heading.

Medical Subject Headings. It is the National Library of Medicine’s controlled vocabulary for indexing and searching the biomedical literature, and NLM describes it as “a controlled and hierarchically-organized vocabulary produced by the National Library of Medicine”. It is what PubMed and MEDLINE are indexed with.

Descriptors and qualifiers are released annually, timed to the calendar year, so the vocabulary has a year attached to it. Supplementary concept records — chemicals, drugs and some rare diseases — are updated continually and released daily. This difference is important when a search is documented: a descriptor search is reproducible against a stated MeSH year, and a supplementary concept search is reproducible only against a date.

The descriptor is the heading itself, and it is the term that is searched. An entry term is a synonym or closely related phrase that cross-refers to a descriptor, and it is how searchers find the heading. “Mediterranean Diet” is an entry term; the descriptor is Diet, Mediterranean. Searching the entry term as though it were a heading is one of the most common reasons a search retrieves nothing without warning.

Searching a descriptor together with all the narrower descriptors indented beneath it in the tree. PubMed does this by default, so a search for a broad heading already includes its children. NLM’s own example is Pneumoconiosis, which also retrieves Asbestosis and Silicosis. It is switched off with the [mh:noexp] tag.

One of the main topics discussed in an article, as judged by the indexer and marked with an asterisk on the record. Searching with the [majr] tag restricts retrieval to articles where the heading is a main topic rather than a passing mention. It raises precision and loses recall, so it belongs in a focused search rather than in a systematic review.

There are two separate reasons, from two different sources. The Cochrane searching chapter advises combining controlled vocabulary with a wide range of free-text terms, because indexing is applied inconsistently by indexers who are not always subject experts, and because the available terms may not match those the searcher requires. Separately, NLM documents that records enter PubMed from publishers and as in-process citations before MeSH headings have been applied, so the most recent literature is reachable only by free text.

No. There are 77 topical qualifiers, but each descriptor carries its own list of allowable ones. Adverse effects, for instance, is not allowable on a disease descriptor — it attaches to drugs and procedures. An invalid pairing does not produce an error message; it produces no results, which is harder to detect. The allowable list on the descriptor’s own record should be checked.

The figure changes every year, which is why a number without a year attached to it should be treated with suspicion. The most recent total NLM publishes on a generally accessible page is 29,917 descriptors for the 2021 vocabulary, alongside more than 270,000 supplementary concept records. As an indication of the scale of annual change, the 2024 vocabulary added 311 descriptors, modified 1,044 and deleted one.

References

  1. U.S. National Library of Medicine. Medical Subject Headings (MeSH). NLM. Page last reviewed 24 September 2026. read 15 October 2026. View on U.S. National Library of MedicineNLM's own home page for the vocabulary. The source for the definition and for the current MeSH year.
  2. U.S. National Library of Medicine. Preface to Medical Subject Headings. NLM. Page last reviewed 9 July 2025. read 15 October 2026. View on U.S. National Library of MedicineThe source for how the vocabulary is maintained, and for the only descriptor total NLM publishes on a reachable page, which is for 2021.
  3. U.S. National Library of Medicine. Introduction to MeSH. NLM. read 15 October 2026. View on U.S. National Library of MedicineThe source for the release schedule: descriptors and qualifiers annually, supplementary concept records daily.
  4. U.S. National Library of Medicine. MeSH Record Types. NLM. read 15 October 2026. View on U.S. National Library of MedicineThe authority for what each kind of MeSH record is and which search tag retrieves it. The source for the count of 77 topical qualifiers.
  5. U.S. National Library of Medicine. MeSH Browser. NLM. read 15 October 2026. View on U.S. National Library of MedicineThe authoritative place to look a descriptor up. Every descriptor string, unique ID and tree number on this page was read from it. The browser header confirms the current vocabulary year.
  6. U.S. National Library of Medicine. MeSH in retrieval. NLM. read 15 October 2026. View on U.S. National Library of MedicineThe source for exploding, for the no-explode tag, and for how a qualifier is combined with a descriptor in a search.
  7. U.S. National Library of Medicine. MeSH Major Topic. NLM PubMed training module. read 15 October 2026. View on U.S. National Library of MedicineThe source for what a major topic is and how the asterisk works. Note its useful nuance: where the asterisk sits indicates which concept is a main topic, but that placement does not itself affect searching.
  8. U.S. National Library of Medicine. PubMed User Guide. NLM. read 15 October 2026. View on U.S. National Library of MedicineThe current user guide. Cited for the three things that switch automatic term mapping off, which is the practical consequence authors most need to know.
  9. U.S. National Library of Medicine. Automatic Term Mapping. NLM Technical Bulletin, 10 December 2004. read 15 October 2026. View on NLM Technical BulletinNLM's explanation of how PubMed maps an untagged query. Cited with its date because it is a 2004 bulletin: the current PubMed user guide carries an equivalent section, but that page exceeds what our tooling can read in one pass.
  10. Lefebvre C, Manheimer E, Glanville J. Chapter 6: Searching for studies, section 6.4.5. In: Higgins JPT, Green S (editors). Cochrane Handbook for Systematic Reviews of Interventions, version 5.1.0. The Cochrane Collaboration, 2011. read 15 October 2026. View on Cochrane (archived)The source for combining controlled vocabulary with free text, and for the reasons it gives: indexer inconsistency and vocabulary mismatch. Cited from the archived fifth edition, because the current chapter's text sits in a PDF we did not open.
  11. U.S. National Library of Medicine. MEDLINE, PubMed, and Record Types. NLM Technical Bulletin, May–June 2000. read 15 October 2026. View on NLM Technical BulletinCited specifically for the indexing lag: publisher-supplied and in-process citations are in PubMed before MeSH headings have been applied. It is a 2000 bulletin and is cited with its date.

Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.

Related: MeSH in medico-legal literature searching · building a search from a PICO question.

How Pubrica can help

Our information specialists build and document database search strategies for systematic reviews and other evidence syntheses — controlled vocabulary and free text per concept, translated across databases, with the MeSH year, the run date and the yield at each step recorded so the search is reproducible and reportable.

Get a free quote

Regulations

Check your project against this rule

  1. 1Your work
  2. 2Your details
  3. 3Send
What do you need?

Material you send is held under signed confidentiality, with access limited to the assigned team and encrypted transfer. It is not used to train anything and is not passed to third-party tools. Ask for the agreement before you send anything you would rather cover first.