Glossary · Methodology
Meta-analysis
In short
A meta-analysis combines the results of several studies on the same question into one overall result by calculation. That makes the estimate more precise — but not better than the studies it is made of.
Individual small studies fluctuate strongly; pooled together, random swings partly cancel out. That is why a good meta-analysis of randomized trials sits at the top of the evidence hierarchy.
The weaknesses are inherited: if the individual studies are small, unblinded or almost all come from one source, the overall result stays shaky too. In addition, negative studies are published less often. If they are missing, the pooled effect comes out too large.
A meta-analysis is usually part of a systematic review that specifies in advance which studies are included. If it mixes in observational studies, it also inherits their confounders.
Many weak studies do not add up to a strong one.
Where this comes up in the fact checks
- Berberine – Pooled across more than 30 studies, 0.19 percentage points of HbA1c remain — less than well-known single studies suggest, and the studies are considered methodologically weak.
- Turmeric – A network meta-analysis of 23 studies on knee osteoarthritis compares several preparations in one statistical model.
- Cutting down on alcohol – A meta-analysis of 107 cohort studies compared only with lifelong non-drinkers — and the apparent advantage of light drinking was no longer significant.
Sources
Related terms: Systematic review · Effect size · Confidence interval
Information only, not medical advice. The BK-Score is a subjective assessment according to disclosed rules – not a scientific evaluation. Scoring rules