Peer review and the evidence hierarchy
What peer review does, what it does not do, and how evidence is ranked.
Peer review submits a manuscript to independent specialists who assess methods, analysis, and conclusions before publication. It is a filter, not a guarantee: reviewers rarely see raw data and cannot detect careful fraud. Peer-reviewed work is regularly corrected or retracted. "Peer-reviewed" means it cleared a threshold, not that it is true.
Evidence is conventionally ranked: systematic reviews and meta-analyses of randomised trials at the top, then individual randomised trials, then observational studies, then case series, then mechanistic and animal work. The hierarchy concerns confidence in causal claims, not usefulness — preclinical studies sit low on it while remaining essential to understanding mechanism.
The practical habit is to ask what kind of evidence supports a claim before assessing the claim. A striking in vitro finding and a large randomised trial are not equivalent, however similar the headlines. See How to read a research paper and Reproducibility in biomedical research.
Related articles
- MethodsSystematic reviews and meta-analysisSynthesising a whole literature rather than trusting a single study.
- MethodsHow to read a research paperA practical order of operations for judging a study without taking its abstract at face value.
- MethodsReproducibility in biomedical researchWhy a substantial share of published findings do not replicate — and what reduces the risk.
- MethodsPreclinical modelsWhat animal models can and cannot tell you about human regeneration.
- MethodsBench to bedsideThe pathway from laboratory observation to clinical use, and where it usually breaks.
- MethodsClinical trial phasesWhat each stage of clinical testing is actually designed to answer.