Center weighting factors and treatment effects in multicenter clinical trials

Abstract

The statistical evaluation of multicenter clinical studies is subject to ongoing debates. While the inclusion of several hospitals into a trial is often the only possibility to achieve the required sample size for the demonstration of a treatment effect, a standard solution for the statistical evaluation of such data is not yet available.

After investigating current methods (fixed and mixed effect models) as well as performing sensitivity analyses for the examination of their robustness, a new weighting factor is introduced. It rests upon a fixed effects model. Heterogeneity of treatment effects is however also incorporated by means of a "standard center" which is determined in advance of the trial. Important concomitant variables that the patients of each center should have are utilized to define the "standard sample of patients" and therefore the "standard center". The integral part of the new weighting factor is the deviation of a participating center from the "standard center" which is performed with the help of the Matusita distance measure.