OBJECTIVE : Previous research suggests that natural fluctuations in seizure rates within individuals have a quantifiable impact on therapeutic clinical trial outcomes.
METHODS : A trial simulator estimated the statistical power of clinical trials with a typical trial design with and without patients included who exhibited a range of means (1-15 seizures/mo) and standard deviations (1-15 seizures/mo) in their baseline seizure rates. Trial outcomes were evaluated using 50% responder rates, median percentage change, and time to prerandomization.
RESULTS : Patients with higher seizure frequencies and lower standard deviations during their baseline contribute more to the statistical power regardless of the method used to evaluate the trial. Power varied from -20% to 30% depending on baseline seizure characteristics.
SIGNIFICANCE : Patient-specific characteristics can predict the contributions to the statistical power of clinical trials for epilepsy treatments. It may be possible to characterize this contribution with baseline data, leading to more efficient clinical trials.
Romero Juan, Goldenholz Daniel M
clinical trials, deep learning, epilepsy, seizures, statistics