The standard designs for safety and efficacy trials of investigational antiepileptic drugs are placebo-controlled, add-on trials and active control equivalence studies. These designs, motivated by medical ethics, have serious evidential limitations. Add-on trials are frequently criticized as insensitive and difficult to interpret in the presence of drug interactions; active control equivalence studies are not probative of test drug activity. As an alternative, we describe two trial designs: a placebo-control design with inpatients who in undergoing a presurgery seizure evaluation have had all antiepileptic drugs discontinued; and an active-control design aimed at showing the test drug superior to the control treatment, thus avoiding the interpretational difficulties of no-difference outcomes. A critical feature of these new designs is the limitation of subject exposure to unacceptable treatments. This is accomplished through protocol criteria--corresponding to therapeutic failure--which both terminate a subject's trial participation and form the basis of efficacy comparisons.