A point estimate such as a risk ratio of 0.80 is only the centre of a range. The confidence interval around it, for example 0.68 to 0.94, expresses the precision of that estimate: it is the range of effect values that are compatible with the observed data at the chosen confidence level, conventionally 95%. A narrow interval means the study had enough information to pin the effect down; a wide interval means the data are compatible with effects ranging from trivial to large. The position of the interval relative to the null value of 1.0 matters for statistical significance: an interval that excludes 1.0 corresponds to a statistically significant result at the 0.05 level, and an interval that includes 1.0 does not. But the position of the interval relative to a clinically meaningful threshold matters for a different judgment. An interval of 0.95 to 0.99 excludes the null and is statistically significant, yet it also excludes any clinically worthwhile benefit, so the result is precise but unimportant. An interval of 0.60 to 1.05 includes the null and is not statistically significant, yet it is also compatible with a substantial benefit, so the study is inconclusive rather than negative. Reading the interval means asking both questions at once: does it cross 1.0, and does it cross the threshold that would change practice?
Learn MotionHow Doctors Judge Whether a Medical Study Can Be Trusted
Reading the Result: Effect Size, Uncertainty, and Significance
Reading a Confidence Interval
2 / 4
Look at the horizontal axis with the null value marked at one point zero, and a clinically meaningful threshold marked further to the left. Each interval is a bar. Ask two questions of every bar. First, does it cross the null line? If it does, the result is not statistically significant. Second, does it cross the clinically meaningful threshold? If it does, the study cannot rule out a benefit large enough to matter. The interval from zero point nine five to zero point nine nine does not cross the null, so it is significant, but it also never reaches the meaningful threshold, so the benefit is too small to act on. The interval from zero point six zero to one point zero five crosses the null, so it is not significant, but it reaches well past the meaningful threshold, so the study is inconclusive rather than negative. Significance and clinical importance are answered by different parts of the same bar.
0:00 / 0:00