Statistical significance versus clinical relevance. Part II. The use and interpretation of confidence intervals

Most clinical studies are carried out to estimate some quantity, such as a difference in means or proportions. However, significance testing without controlling the power of the test is not an appropriate tool for this purpose, since practical conclusions are difficult (or even impossible) to formulate. The use of confidence intervals, which provide effective information in this regard, is recommended and illustrated by means of an example. With confidence intervals, it can clearly be stated in what range the population value (difference, proportion, or any other point estimate) is likely to lie. A clinically relevant interpretation can then be made without difficulty.