A study should be evaluated for the question it was designed to answer. Start with the original record, not a headline or product page.
Identify the record
Use the PubMed User Guide to understand citation fields, publication types, search filters, and links. Record the title, journal, authors, date, identifier, and any correction or retraction notice shown in the record.
Read the design
- What question was specified?
- Was the work laboratory, animal, observational, or interventional?
- Who or what was included, excluded, and compared?
- Were allocation, masking, and analysis methods described?
- Were outcomes defined before results were known?
FDA's clinical research overview describes the different questions addressed across clinical research phases.
Compare registration with reporting
For a registered study, use ClinicalTrials.gov's study-record guide to locate outcome measures, enrollment, dates, sponsor information, and submitted results. Compare those fields with the publication.
Examine the results without overextending them
- Distinguish the prespecified primary outcome from secondary or exploratory analyses.
- Look at the size and precision of an observed effect, not only whether a threshold was crossed.
- Check missing data, withdrawals, protocol changes, and whether all groups were reported.
- Read the authors' limitations and identify important limits they may not discuss.
- Check funding, declared conflicts, data availability, and independent replication.
Write a bounded conclusion
A careful summary names the study type, population or model, comparison, measured outcome, and limitations. It does not turn an early or indirect result into advice.
Source record
Authoritative sources
- National Library of Medicine PubMed User Guide
- ClinicalTrials.gov How to Read Study Records
- U.S. Food and Drug Administration Step 3: Clinical Research