A confidence interval is a range around an estimate that shows how much statistical uncertainty remains after a study measures an effect.
What the range means
A point estimate gives one number. The interval gives the plausible neighborhood around that number under the study's model and sampling assumptions. Narrow intervals suggest more precision. Wide intervals suggest the study cannot pin the effect down tightly.
For an effect size, the interval may include values that would change the practical interpretation. A result can look promising at the point estimate and still have an interval that includes trivial or null effects.
Why it beats a p-value alone
A p-value asks whether the data are unusual under a null model. A confidence interval shows the size of effects still compatible with the data. That is usually the more useful question for supplement decisions.
For example, a trial can clear a statistical threshold with an effect too small to matter. Another trial can miss the threshold yet still be compatible with a meaningful benefit if it was underpowered.
How to read the null
The null value depends on the metric. For mean differences and Cohen's d, the null is usually 0. For ratios such as a hazard ratio, the null is usually 1.
If the interval crosses the null, the estimate is statistically uncertain in direction. If it stays on one side of the null, the direction is clearer, though the size may still be too small to matter.
How this affects ranking
Confidence intervals help separate stable evidence from noisy evidence in recommendation ranking. A precise small effect can be more reliable than an imprecise large effect.
They also explain why a single small study should rarely dominate an evidence tier.
Safety note
Intervals around benefit do not describe all safety uncertainty. Harm data often need separate capture and may be too sparse for precise intervals.