- Frank Krummenauer: Institut für Medizinische Biometrie, Epidemiologie und Informatik der Universität Mainz. krummi@imsd.uni-mainz.de
Confidence intervals combine the ideas of clinical relevance and statistical significance by using one instructive expression, which enables remarkable reduction of table structures and result sections in scientific publications. Confidence interval based conclusions can be transferred from a clinical trial to its underlying study population with respect to a residual statistical error probability, i.e. the significance concept is retained. However, their presentation using the original unit of the clinical endpoint under consideration allows for immediate interpretation of the results' clinical impact. For example, the comparison of two therapy groups based on a binary endpoint becomes feasible using the relative risk's confidence interval. If "1" is not contained in the interval, the therapy groups significantly differ concerning this endpoint. The larger the interval turns out, the less precise the characterisation of the "real" risk value based on the study risk estimate. The larger the risk estimate turns out, the more clinical relevance.