Why restrict to open access
Library access to paywalled journals varies by institution and ends when your enrolment does. A review built only on openly available sources can be checked by your supervisor, your examiner, and a reader with no affiliation at all. The trade-off is coverage, and you handle that by stating it in your limitations rather than hiding it.
Where to search
- PubMed Central open-access subset — full texts deposited under open licences.
- Europe PMC — broad biomedical coverage including preprints and patents, with full text where available.
- DOAJ — journals that are fully open access.
- ClinicalTrials.gov, EU CTR, ISRCTN, WHO ICTRP — trial registries, for ongoing and unpublished studies.
- medRxiv and bioRxiv — preprints, clearly labelled as not yet peer reviewed.
- Cochrane open-access content — reviews and protocols available without a library login.
- Regulator archives — FDA, EMA, MHRA and ICH documents, often the only public source of full trial detail.
- OpenAlex, Crossref and Unpaywall — for locating an open version of a known record and resolving metadata.
Build the search, then record it
Translate each PICO element into search terms: controlled vocabulary where the database offers it, plus free-text synonyms and spelling variants. Combine synonyms with OR inside each concept, and combine concepts with AND. Keep the comparator out of the search string unless it genuinely narrows the set — it usually loses relevant studies.
For every source, record:
- The database or registry name and the interface used.
- The exact search string, copied rather than retyped.
- Any filters — date range, language, study design.
- The date the search was run and the number of records returned.
This block goes in an appendix. If it cannot be pasted into the same database and rerun, it is not yet reproducible.
Deduplicate on identifiers
Match on DOI first, then registry number, then on title and first author as a fallback. Records for the same trial in different registries are common; so are a preprint and its later journal version. Keep one record per study and note the linked versions.
Mark what you could actually read
Distinguish full text available from abstract only in your table. Abstract-only records can support screening decisions and can be reported as identified, but they rarely support extraction. Where a study was potentially eligible but no open full text existed, list it as such — that list is part of your limitations and shows the boundary of the review honestly.
Update before you submit
Rerun your searches shortly before submission with the same strings and note any new records. Examiners routinely check whether a relevant study published during your write-up was considered.