PHASTAR helped to design and analyse a placebo-controlled crossover trial using an anti-epileptic to assess cognitive function in children. The study had a large number of outcome measures, covering cognitive function but also seizure frequency, an ambulatory EEG assessment, IQ and a behavioural assessment using the Conners Rating Scales. The study was also attempting to determine the relationship between transient ischemic attacks and short-term impacts on cognitive function.
Patients initially entered a single blind phase, then had the study treatment added to their existing medication regime in a blinded fashion. After a treatment-free period of five weeks, the second treatment was added. PHASTAR created the randomization scheme using blocked stratification.
PHASTAR worked on designing the study, looking at the expected impacts of treatment on the various outcome measures. We proposed that a measure of cognitive function based on the first principal component of the cognitive function tests, after reaction times had been log transformed, be used as the primary outcome. This is a technique that was researched and developed by PHASTAR’s statistical research team.
All data was prepared according to CDISC standards, with the client’s SDTM and ADaM standards considered. We prepared an analysis plan and table shells that were subject to a review process with the client. As there were a large number of outcome variables that would be analysed in a similar manner, we decided at an early-stage programming strategy meeting to create a SAS macro that would analyse a large number of outcome measures, either using parametric (log transformed or not) or nonparametric methods, taking into account the crossover design of the study. We proposed a non-parametric estimator based on the van-Elteren test to allow treatment effect estimates to be extracted appropriately. There was an observed a learning effect in the cognitive function tests, where the average responses improved on each assessment. This was accounted for in the analysis as a period effect using mixed models. Graphical presentations of the results were prepared.
A regulatory compliant data package was prepared, including a define.xml documentation set. We subsequently supported the production of the Clinical Study Report and publication, with the trial results showing no adverse effects on cognitive function with the new treatment.