How precise was the measured difference?

The best way of expressing this is the Confidence Interval (CI). Although measurements of statistical significance with "p values" are important, they do not give a picture of the data spread like the CI does. The usual method is to calculate the "95% CI" which simply defines the range of values into which the true result (eg. RRR) will be found 95% of the time.

It is important to look at the results and especially the lower and upper confidence limits to see if the results make sense. These limits may indicate that not enough patients were enrolled in the study.

If the CI is not given, "p values" can be used. The standard deviation or standard error can also be multiplied by 2 to give an estimate of the 95% CI.

A better, way of presenting the results would be as an Absolute Risk Reduction (ARR) as it would better reflect the rarity of the outcome. In this case the AAA = (20 - 15) = 5%

Application to Pollock et al's thyroxine paper

In this study, there are very large confidence intervals that reflect the small sample size and make the results suspect.

Previous

EBM Home