Efficacy trials must compare two treatment groups in a blinded fashion. It is not typical to use a population as its own control, comparing a baseline pretreatment epoch with an epoch after treatment has been initiated, because of placebo effect. In other words, patients who have an expectation that they may improve will demonstrate improvement with or without new treatment.

Physicians may have difficulty accepting the reality of the placebo effect in intractable epilepsy patients, who have been switched from treatment to treatment without benefit, yet time and again patients in the placebo arm of multicenter epilepsy studies will demonstrate a substantial seizure reduction. The degree of placebo effect will vary from study to study for unclear reasons. In recent studies, 0% to 36.5% of patients in the placebo arm of blinded trials showed a 50% or greater seizure reduction over a standard 3-month exposure period.¹⁴ The degree of a placebo effect also varies between regions and individual centers within one large multicenter trial.

For some purposes it is considered acceptable or even desirable to choose two active treatment arms as blinded comparison groups. Some of the best early examples of this type of trial were two VA cooperative studies,^46,47 which randomly assigned patients with newly diagnosed seizures to different antiepileptic drug therapies. This allowed direct comparison of these drugs in terms of both efficacy and safety. A large number of active control comparison trials have been performed in the last decade, comparing standard drugs such as carbamazepine and phenytoin to the newer drugs.^{10,12,15,17,18,39,41,55,68} This type of trial may be extremely informative for clinical purposes. In Europe, they have contributed to the registration of a number of new antiepileptic drugs (AEDs) as monotherapy. In the United States, however, the FDA does not accept active controlled trials as proof of monotherapy efficacy.⁴³ These active-control trials, including the VA cooperative studies, were able to demonstrate equivalence between drugs. In order for the FDA to accept a trial as proof of efficacy, superiority of the new drug must be demonstrated.

There are inherent disadvantages to adjunctive, or add-on, trials in which active treatments or placebo is added to the patient’s baseline antiepileptic drugs. It is more difficult to prove efficacy in a patient who is already partially treated. Side effects may be magnified by combining drugs, and pharmacokinetic interactions may alter baseline or experimental drug levels. Despite these drawbacks, most efficacy trials use an adjunctive therapy design because of the ethical issues raised by monotherapy parallel trials involving placebo in patients with epilepsy. New trial designs were developed in the 1990s that attempted to circumvent ethical issues and permit monotherapy studies. The two most popular designs were described by Pledger and Kramer.⁵³ The first, which involves randomization to drug or placebo in inpatients who have had their background antiepileptic drugs withdrawn for presurgical evaluation, has been deemed too short to be used for registration purposes and not relevant for extrapolation to a general clinical population, and is now used primarily for a proof-of-principle trial.^13,28,63 A second design is performed in outpatients.^{8,29,38,53,60,62} Patients are randomized to treatment with an experimental drug or placebo, after which baseline therapy is withdrawn over 2 to 8 weeks. A modified, or “ethical,” placebo is utilized rather than a true placebo to reduce the likelihood of status epilepticus or secondary generalization. This can consist of a minimally effective dose of either the same investigational drug or of any other therapy presumed to be less effective than the test drug. A starting dose of valproic acid (15 mg/kg) has been employed in a number of trials for this purpose. Outcome is assessed in terms of “failures” and “completers.” Failure is determined on the basis of escape criteria, such as doubling of seizure frequency, occurrence of generalized tonic–clonic seizures, or increase in seizure severity. If more patients receiving the experimental drug at a therapeutic dose in monotherapy can complete the trial, without fulfilling escape criteria, than patients receiving modified placebo in monotherapy, the treatment is considered effective in monotherapy. This trial design has been successfully used to obtain a monotherapy indication from the FDA for oxcarbazepine and lamotrigine (withdrawal to monotherapy in refractory patients). A drawback of this design is that antiseizure effects are evaluated in an AED withdrawal situation only.

Two other monotherapy designs have been employed to gain FDA monotherapy approval. These have been performed in patients with newly diagnosed epilepsy. One study compared oxcarbazepine to placebo in patients having frequent seizures (average 5.5 per month) at baseline.⁵⁹ The end-point was time to third seizure. Another study compared 50 mg of topiramate to 400 mg in newly diagnosed adolescents and adults with partial-onset seizures.¹ Outcome measures included time to first seizure as well as seizure-free rate at 6 months and 1 year. Although all these monotherapy trials led to registration of at least one AED as monotherapy, the trials continue to raise ethical as well as pragmatic issues. For this reason, efforts are under way to devise new methods to perform monotherapy trials.^30,34

In parallel trials, patients are randomly assigned to one of two treatment groups for a period of time. Their seizure frequency during the treatment period is compared to a pretreatment baseline. Seizure outcome is compared between the two groups. In a crossover design, patients are also randomly assigned to two treatment groups, but after a period of time each group is crossed over to the other treatment, usually after a washout period. Outcome as compared to baseline is determined for each treatment. There are some advantages to a crossover design. Far fewer patients are required to perform the trial, provided
that the subjects complete all required crossover periods, including the washouts. Also, if two active treatments are used, this trial design is more like clinical practice, in which patients are usually crossed from one treatment to another. Each patient will receive both treatments, allowing more direct comparison of efficacy and tolerability. The disadvantages of a crossover trial include a much longer duration, risk of patients dropping early, and, more importantly, a potential unblinding of the trial. This is of the utmost concern in a trial that compares placebo to active drug. Patients may be able to discern a difference in side effects when switching from placebo to drug, or vice versa. There may be carryover effects of a drug, which would impact the initial portion of the second treatment phase. For these reasons, the FDA and the EMEA do not favor such trials.

The majority of phase II and phase III AED trials are performed in patients with partial epilepsy, specifically with complex partial seizures. These patients comprise the majority of adults with uncontrolled seizures, and the pharmaceutical companies need an indication from registering bodies to treat partial seizures, for market share. With rare exceptions, a drug that is unable to treat such patients would not be profitable to develop. Recently, there has been discussion regarding bringing drugs forward for niche indications or for orphan diseases such as Lennox-Gastaut syndrome. Nonetheless, to date no new AEDs have been approved without a partial seizure indication.

When a partial seizure indication has been obtained, further studies may be performed to assess efficacy in other syndromes. Studies in different syndromes offer their own potential obstacles and may impact trial design. One example can be seen in the Lennox-Gastaut population. Initially, it was felt that seizures in these patients were easily recognizable, and that placebo effect would not be an issue in this severely impaired population. Therefore, open trials were undertaken in patients treated with cinromide, a drug that was in development. The results were very promising, demonstrating a >50% response. However, when the study was repeated with a placebo control, it was found that the entire treatment effect could be attributed to a placebo response.²³ It was felt that this reflected not only a bias in parental observation, but also a difficulty in differentiating between seizures and abnormal behaviors, common in this population. As a result, subsequent trials incorporated video-electroencephalographic (V-EEG) monitoring to train parents on differentiating seizures. In addition, the primary outcome variable was reduction in motor (tonic, atonic, and generalized tonic–clonic) seizures, which are the most recognizable. This led to a successful trial design, and ultimately to approval of three of the new AEDs (felbamate, lamotrigine, and topiramate) for use in the Lennox-Gastaut syndrome.^26,49,61 This and other examples of potential populations, study limitations, and ways of circumventing them are given in Table 1.

Another patient characteristic that may impact AED trials is disease severity. It cannot be taken for granted that a drug that has been proven to be highly effective in intractable, severely affected patients will also be the most ideal drug for new-onset patients. It is feasible that some drugs may work preferentially in refractory patients, due to specific pathophysiologic alterations that are found in these patients. One possible example of this relates to the antiepileptic drug vigabatrin. This drug exhibits retinal toxicity in selected populations and is not available in the United States, but is approved in Europe. In placebo-controlled add-on trials in refractory patients, this drug appeared to be more potent than almost any other available antiepileptic drug.³¹ However, when tested in a head-to-head fashion against carbamazepine, it was not as effective.¹⁶

Once a drug has been proven safe, trials in new-onset patients are frequently performed. These trials have usually been performed as active-control comparisons, in which an investigational agent is compared to a standard agent, with the exception of oxcarbazepine, which was tested against placebo.⁵⁹ Both placebo control and active control may have drawbacks to confirm the efficacy of AEDs in newly diagnosed patients. Placebo-controlled trials are short by necessity, and only prove that a drug is better than “nothing.” On the other hand, if the population for an active-control equivalence trial is chosen poorly, if the trial is not designed properly, or if the sample size is not sufficient to have the power to show a difference when one exists, then equivalence may be demonstrated between two drugs for which there actually exists a clinically meaningful difference in efficacy or safety.³⁴

Women of childbearing potential also bear further consideration. In earlier epilepsy trials, women were only included in studies if they were postmenopausal or had undergone surgical sterilization. Recently, there has been pressure to include more women in trials at an earlier point in development.⁴² In order to do this, it is necessary to expose women of childbearing potential to drugs that do not have established efficacy and safety. Most protocols lay out strict guidelines for contraception that is considered acceptable during a trial based on potential hormonal interaction data. Occasionally, contraception will fail and a pregnancy will ensue. It is often the policy at present to discontinue the investigational drug immediately in such a situation. This may not always be the safest course for the mother or fetus, particularly if the individual had a very significant seizure reduction from that drug. Postmarketing pregnancy registries by sponsor companies and academic consortia have been set up to better investigate the comparative teratogenicity of AEDs.

Clinical testing of antiepileptic drugs considers children separately because of the age-related changes in both brain and overall physiologic and biochemical status that occur during childhood along with the age dependency of certain seizure types and epileptic syndromes. Most studies on AEDs have considered “children” to be those younger than 12 years of age, and have included those aged 12 years and over in trials designed primarily for adults. The population of patients younger than 12 years is often subdivided into neonates (<1 month postnatal age); infants (1 month up to 2 years), and children (from 2 years up to 12 years). The children group can be further subdivided into preschool (2 to 5 years) and school age (5 to 12 years). This last division is not arbitrary; there are differences in the type of epilepsies likely to present before and after 5 years⁵⁷ and there are different methods and scales used to monitor behavioral and cognitive side effects between these two age groups.