Including non-English studies in a systematic review

Why the English-language limit persists

Almost nobody applies an English-only restriction because they think it improves the review. They apply it because the alternative is operationally painful: unfamiliar database interfaces, incompatible query syntax, no shared export format, and abstracts the review team cannot read well enough to screen.

Cochrane guidance discourages language restrictions, and PRISMA 2020 requires you to report any you apply. The gap between the guidance and common practice is a tooling gap, not a disagreement about method.

Step 1 — search where the literature actually is

MEDLINE and Embase index a minority of the world's health journals. Which regional index matters depends on your question:

• LILACS — Latin America and the Caribbean; Portuguese and Spanish.
• J-STAGE — Japanese society journals, many with Japanese-only abstracts.
• ThaiJO — Thai journals across medicine and public health.
• KoreaMed — Korean biomedical journals.
• HAL — French open repository, strong on public health and epidemiology.
• DOAJ and OpenAlex — broad multilingual open-access coverage that catches journals the regional indexes miss.

Add the WHO ICTRP and regional trial registries when your question involves interventions.

Step 2 — translate at the right stage

You don't need a professional translation of every record. Work in two passes:

• Screening pass. Machine translation of title and abstract is sufficient to decide "possibly relevant / clearly not." Note in your protocol that screening used machine translation.
• Full-text pass. For records that survive screening and will contribute data, use a fluent reader or professional translation. Data extraction from a machine-translated methods or results section is not defensible.

Keep the original-language text alongside the translation throughout, so a reviewer can check any claim against the source.

Step 3 — report it under PRISMA

• List every database searched, with the exact strategy and the date of each search (PRISMA items 6 and 7).
• State explicitly that no language restriction was applied — or, if one was, state it and give the reason.
• Describe how non-English records were screened and extracted, including any use of machine translation.
• Report per-source record counts in the flow diagram so readers can see the regional contribution.

Does it change conclusions?

Sometimes yes, sometimes no — and that is the point. The published evidence on language bias is mixed and topic-dependent, which means the effect on your specific review is an empirical question you can only answer by running the search. Excluding those records by default answers it by assumption instead.

How Symmathy fits

Symmathy runs one query in parallel across PubMed, Europe PMC, Cochrane, ClinicalTrials.gov, J-STAGE, LILACS, DOAJ, HAL, OpenAlex, CrossRef, ThaiJO, and KoreaMed. Non-English titles and abstracts are translated inline for the screening pass, the original text and source link stay attached to every record, results are deduplicated across sources, and the set exports to RIS, BibTeX, or CSV for Covidence, Rayyan, DistillerSR, or your reference manager.

Symmathy does not judge, score, or rank studies, and offers no clinical guidance. It removes the retrieval barrier; the methodology stays yours.