Start with the question and study type
Write down the question the authors tested, the material they studied and the experimental model. A study of an isolated compound does not automatically establish what a blend will do. A paper about a compound also does not verify a supplier’s inventory.
Check whether you are reading original research, a review or a commentary. Original research commonly follows introduction, methods, results and discussion. Reviews bring together other work, so examine how the included literature was selected. ICMJE manuscript guidance.
Read the methods before the headline
For animal research, ARRIVE 2.0 provides a useful framework for checking whether enough detail is reported to assess the study. Ask:
- Which groups were compared, and what was the control or comparator?
- What was the experimental unit, and how many units were included?
- How were units assigned to groups? Were relevant study stages blinded?
- Were inclusion and exclusion criteria described, along with any exclusions?
- Which outcomes were measured, and which outcome was primary?
- Were the animals, procedures and statistical methods described clearly?
Missing information limits your ability to evaluate a result. Reporting a checklist item is not, by itself, proof that a conclusion is correct. ARRIVE 2.0 explanation and elaboration.
Separate observations from interpretation
Look for group-level results, variability, effect sizes and uncertainty, where reported. Compare the actual outcome with the language used in the abstract. Note whether a proposed mechanism was tested directly or offered as a possible explanation.
Record the authors’ limitations, funding and competing interests. Then ask which findings apply only to the model and conditions studied. ARRIVE’s reporting guidance includes results and the context needed to interpret them. Read the guidance.
Worked example: separate tissue evidence from a trial
Campbell and colleagues’ 2012 GHK paper is useful for practicing this distinction. Read the original study (Genome Medicine 4:67; DOI 10.1186/gm367).
| Study component | What to record | Boundary |
|---|---|---|
| Lung tissue | 64 regions from eight lungs; six COPD explants and two donor lungs. | 64 regional samples are not 64 independent people. |
| Computational analysis | Gene-expression patterns and a Connectivity Map comparison. | A computational match generates a hypothesis. |
| Cell experiments | GHK was tested in cultured human fibroblasts; the paper reports gene-expression and collagen-contraction observations. | Human-derived cells do not make this a human treatment trial. |
The paper does not establish clinical safety, efficacy or an administration regimen. It does not test Peptyed inventory. Preserve the material name used by the authors—GHK—rather than treating every similarly named catalog material as equivalent.
Reading exercise: compare the abstract with the sample-acquisition methods and Figure 1. Write down the experimental unit for each analysis before interpreting its sample size. These notes are an educational example, not a systematic evidence review.
Continue with our comparison of GHK and GHK-Cu preclinical research to examine how material identity, experimental timing and reporting details affect a comparison between papers.
Keep a short evidence note
Use this format when reviewing a paper:
- Source: full citation, publication date and DOI or stable link.
- Question: what was tested, in which model, with which material?
- Design: groups, experimental units, sample size and bias controls.
- Result: the measured outcome and reported uncertainty.
- Limits: what the study does not establish, and what remains unanswered.
For terminology, open the research glossary. For the difference between a laboratory finding and clinical evidence, read preclinical vs. clinical research.