Primary Efficacy Endpoint in Clinical Trials of Antiepileptic Drugs: Change or Percentage Change

Abstract

In randomized clinical trials of antiepileptic drugs (AEDs), the seizure frequency per x days during baseline and treatment phase periods are recorded to evaluate efficacy of a drug. The seizure frequency data are often nonnormal, and hence an appropriate mathematical transformation is necessary for a statistical analysis. The most commonly used transformations in AED development research are (a) log-transformation of the seizure frequency data, and (b) calculation of percentage change (PC) from baseline in seizure frequency. The log-transformed postbaseline seizure frequency data are analyzed using a parametric ANCOVA model including the log-transformed baseline data as a covariate and treatment group as a factor in the model. The PC data are analyzed using either a Wilcoxon rank-sum test on the PCs, or ANOVA/ANCOVA analysis on the ranks of PCs including treatment groups as a factor and any covariate of interest in the model. A limited number of research works is available in the literature regarding a choice of log-transformed or PC of seizure frequency data in the statistical analyses. In this research, an attempt is made to evaluate the impacts of choosing either of the two transformations on seizure frequency data of AED trials.