A study reports a difference, but its confidence interval crosses the point representing no difference. “Nothing happened” is a tempting summary. It may discard the most useful information: the range of effects the analysis remains compatible with.
Hypothetical example: mean difference, 2 points; 95% confidence interval, −1 to 5.
Zero lies inside this interval. The result does not establish a precise two-point benefit, but it also does not demonstrate that the true difference must be zero.
The Cochrane Handbook’s discussion of interpretation explains why estimates and uncertainty should be considered together rather than reduced to a significant-or-not label. The relevant no-effect value depends on the measure: zero for a difference, one for a ratio.
Ask what the range would mean in practice
Are effects near either end large enough to matter for the outcome being studied? How reliable is the study design? Those questions remain even when an interval is narrow or excludes no effect. Statistical precision does not remove bias or establish clinical importance.
When making a reading note, copy the measure, units, estimate, interval and comparison direction. Do not convert a frequentist 95% interval into a statement that this particular fixed parameter has a 95% probability of lying inside it. For a personal treatment decision, discuss the evidence with a qualified clinician; this example is a guide to reading results, not choosing care.
AI-generated editorial illustration; not a documentary photograph.