Read the prediction interval as the range in which the effects of future studies may fall, while the confidence interval describes uncertainty about the pooled mean effect. The Cochrane Handbook treats both as part of careful interpretation of pooled results, so record them together with the null value rather than reading the pooled estimate alone.
What each interval tells you
- The pooled estimate is the combined effect across included studies. Its confidence interval expresses uncertainty about that combined mean.
- The prediction interval addresses a different question: where the effects of future studies may fall, given the heterogeneity observed across the included studies.
- The null value (for example, no difference) is the reference point. Compare both intervals with it to see whether the pooled mean and the spread of future effects are consistent with no effect.
How to read them together
- Locate the pooled estimate and its confidence interval on the forest plot, and note whether the interval crosses the null. A confidence interval that excludes the null indicates the pooled mean is estimated away from no effect; one that includes it leaves the mean effect unresolved.
- Locate the prediction interval and compare its width and position with the null. A wide prediction interval that crosses the null indicates future study effects could plausibly include no effect, even where the pooled mean appears clear.
- Compare the two intervals directly. A prediction interval much wider than the confidence interval signals substantial heterogeneity among study effects; similar widths indicate the studies agree more closely.
Record a plain-language statement
Write two separate sentences in your evidence record. First, state the certainty about the mean, such as the pooled estimate and whether its confidence interval includes the null. Second, state the spread of future effects, such as the prediction interval and whether it includes the null. Keep the two statements distinct so a reader can see which conclusion rests on the average effect and which rests on the variability between studies.
Cochrane guidance on interpreting results also calls for exact P values, absolute and relative effects, and applicability when drawing review conclusions. Use those alongside the intervals when the decision depends on the size of the effect or on whether the included populations match your setting.
