Reproducibility in biomedical research
Why a substantial share of published findings do not replicate — and what reduces the risk.
Attempts to replicate published preclinical work have repeatedly failed at high rates, in efforts spanning oncology, psychology, and biology broadly. The causes are mostly mundane rather than fraudulent: small samples, flexible analysis, selective reporting, and publication incentives that reward novelty over verification.
Cell-based regenerative work is particularly exposed. MSC preparations differ by donor, tissue source, passage number, and culture medium, so two laboratories using "the same" cells may not be — a reason results diverge, discussed in The stem cell niche. Exosome preparations are less standardised still.
The correctives are unglamorous and effective: pre-registration, adequate powering, blinded outcome assessment, reporting of negative results, and detailed methods including passage number and characterisation. When reading, treat a single unreplicated finding as a hypothesis. Systematic reviews and meta-analysis and Peer review and the evidence hierarchy help, but neither substitutes for independent replication.
Related articles
- MethodsPeer review and the evidence hierarchyWhat peer review does, what it does not do, and how evidence is ranked.
- MethodsSystematic reviews and meta-analysisSynthesising a whole literature rather than trusting a single study.
- CellsMesenchymal stem cellsThe most widely studied cells in regenerative medicine — and a reassessment of what they actually do.
- MethodsPreclinical modelsWhat animal models can and cannot tell you about human regeneration.
- CellsExosomes and the secretomeCell-free therapy: using what stem cells secrete rather than the cells themselves.
- MethodsBench to bedsideThe pathway from laboratory observation to clinical use, and where it usually breaks.