If you are doing any kind of literature search, "I searched PubMed" is the answer most students give and the answer most reviewers do not accept. PubMed is excellent, free, and the default for biomedical research, but it is not exhaustive. The five major databases each cover a different slice of the literature, and a serious search uses at least three. Here is what each one is good at and where the overlaps and gaps are.

PubMed / MEDLINE

Coverage: Over 35 million biomedical citations, going back to 1946. Indexes around 5,200 journals selected by the National Library of Medicine. Approximately 90 percent of MEDLINE content overlaps with Embase, but Embase has 2,000+ journals MEDLINE does not.

Strengths: Free, fast, and uses the curated MeSH controlled vocabulary. Best for US-published biomedical research, clinical medicine, life sciences, public health. Updated daily.

Weaknesses: Underrepresents European, Asian, and South American journals. Indexing takes weeks to months, so the most recent 6 months of papers are partly invisible to MeSH-based searches. Some conference abstracts and gray literature are missing.

Use it when: Always. It is the foundation of every biomedical search.

Embase

Coverage: Around 41 million records, including all of MEDLINE plus over 2,800 additional Embase-only journals. Particularly strong in pharmacology, toxicology, and European biomedical literature.

Strengths: Uses the Emtree controlled vocabulary, which is more granular than MeSH for drugs and devices. Excellent for systematic reviews on drug efficacy, adverse events, and clinical pharmacology. Includes conference abstracts back to 2009, which catches early reports of unpublished trials.

Weaknesses: Paywalled (Elsevier). Most universities subscribe; if yours does, access is via Ovid or Embase.com. Search syntax differs from PubMed — you cannot copy-paste your PubMed string.

Use it when: Any systematic review (it is required by most guidelines). Drug-related questions especially.

Cochrane Library / Cochrane CENTRAL

Coverage: The Cochrane Library has multiple components. CENTRAL (Cochrane Central Register of Controlled Trials) has around 2 million records of randomized and quasi-randomized trial reports. CDSR (Cochrane Database of Systematic Reviews) holds the Cochrane SR/MAs themselves.

Strengths: CENTRAL is the most complete database of clinical trial records, drawing from MEDLINE, Embase, trial registries, and hand-searches. CDSR provides the gold-standard evidence syntheses. Free to access in many countries via national licenses.

Weaknesses: Limited to controlled trials — not useful for observational, qualitative, or descriptive questions. CDSR is selective in topic coverage.

Use it when: Any SR/MA needs CENTRAL to capture the trial universe. Check CDSR first to confirm no existing Cochrane review answers your question.

Scopus

Coverage: Around 90 million records spanning sciences, social sciences, and humanities. Indexes more than 25,000 active titles. Broader subject coverage than PubMed/Embase, including engineering, computer science, and business.

Strengths: Excellent citation tracking — you can find papers that cited a key paper forward in time (forward citation searching). Useful for biomedical questions that intersect with other fields (e.g., AI in medicine, health policy, medical devices engineering). Provides h-index and journal-level metrics.

Weaknesses: Paywalled (Elsevier). Less granular controlled vocabulary than MeSH or Emtree for clinical questions. Coverage of pre-1996 literature is thinner.

Use it when: Your topic crosses disciplines, or you want to do forward citation tracking from a landmark paper.

Web of Science (Core Collection)

Coverage: Around 90 million records across the Science Citation Index Expanded, Social Sciences Citation Index, Arts and Humanities Citation Index, and Conference Proceedings. Selectively curated — only "core" journals.

Strengths: Stricter journal selection criteria mean less noise. Strong citation analysis (similar to Scopus). Useful for bibliometric analyses, identifying highly-cited works, and journals' impact factors.

Weaknesses: Paywalled (Clarivate). Selectivity means it misses some smaller specialty journals. Subject overlap with Scopus is substantial.

Use it when: Bibliometric or impact-focused questions, or as a complement when Scopus is unavailable.

Overlap and minimum coverage

For a systematic review submitted to most journals or to Cochrane, the minimum acceptable database list is:

  • PubMed / MEDLINE
  • Embase (via Ovid)
  • Cochrane CENTRAL

Add Web of Science or Scopus if your topic spans disciplines. Add CINAHL if nursing or allied health is central. Add PsycINFO for mental health. Add regional databases (CNKI for Chinese literature, LILACS for Latin American, KoreaMed for Korean) if the disease or intervention is regionally concentrated — this matters more than students think for diseases like H. pylori, TB, or schistosomiasis.

Why PubMed alone is not enough

Studies that have compared PubMed-only searches against multi-database searches consistently find that 15 to 30 percent of relevant studies for a typical systematic review are missed by PubMed alone. The miss rate is higher for non-English studies, drug studies, conference abstracts, and qualitative research. A reviewer who sees a methods section saying "we searched PubMed" will at minimum ask why other databases were not searched, and may reject the manuscript on that basis alone.

Practical tactics

  • Write each database search separately. Translating syntax across databases takes time. Use the Polyglot Search Translator from Bond University as a starting point but verify the translation.
  • Export everything as RIS or NBIB. Import to EndNote / Zotero / Rayyan / Covidence for deduplication.
  • Save the search strings. Document the exact string used in each database, the date searched, and the number of hits. Reviewers and PRISMA require this.
  • Use trial registries. ClinicalTrials.gov and the WHO ICTRP are not "databases" in the indexing sense but they capture unpublished and ongoing trials that biomedical databases miss.

If you are doing a quick novelty check rather than a formal SR, ResearchChecker queries PubMed (3 parallel searches), ClinicalTrials.gov, Europe PMC (which indexes Cochrane content and preprints), Semantic Scholar, and OpenAlex in parallel and gives you the deduplicated counts. That covers the lightweight screening case. For a formal review, you still do the proper multi-database search above.