How do you read a confidence interval in a study?
Updated: 2026-09-09 · Safety and evidence
Short answer
A confidence interval is the range of values compatible with the observed data. The narrower it is, the more precise the estimate. A very wide interval means the study is compatible with quite different results. An interval that includes 'no effect' means the data cannot rule out no difference at all.
In plain English
- The central value is the best estimate; the interval shows how much uncertainty surrounds it.
- Wide intervals usually come from small studies or highly variable measurements.
- For differences, 'no effect' is zero; for risk ratios, it is one.
- Two studies can agree more than they appear to if their intervals overlap broadly.
What we know
- Confidence intervals convey magnitude and precision together, not just a yes/no.
- Reporting guidelines recommend presenting them alongside p-values.
What we don't know yet
- An interval does not fix design flaws: if the study measures badly, the range is biased too.
- It does not give the probability that a result will repeat in another population.