Traditionally, new antiepileptic drugs (AEDs) are tested initially as adjunctive treatment in patients who are refractory to available agents. Unlike the situation in other therapeutic areas, in the case of AEDs, regulatory authorities do not regard evidence of efficacy obtained under adjunctive use as sufficient to grant a monotherapy license. Instead, regulators require for monotherapy approval that specific studies be conducted to demonstrate that efficacy is retained when the AED in question is used as sole medication.
In the European Union, the European Medicines Agency (EMEA) recommends that monotherapy studies be conducted in patients with newly diagnosed epilepsy, using “randomised, double-blind positive controlled trials aiming to demonstrate at least a similar benefit/risk balance of the test product as compared to an acknowledged standard product at its optimal use” (CPMP, 2000). Therefore, monotherapy approval in Europe generally requires demonstration of noninferiority to an already established comparator, a paradigm first implemented in 2007 to obtain a monotherapy license for levetiracetam (Brodie et al., 2007).
In the United States, a noninferiority approach to demonstrate efficacy of AEDs is not regarded as acceptable by the U.S. Food and Drug Administration (FDA). The FDA argues, justifiably, that no sufficient evidence exists to determine that established AEDs can be consistently differentiated from less effective or ineffective treatments under any specific study design, and, therefore, the finding of equivalence or noninferiority does not allow to exclude that, in the particular population and under the specific conditions in which the trial was conducted, both treatments could have been similarly ineffective (Leber, 1989). Based on these premises, the FDA has insisted that evidence of efficacy be obtained by demonstrating superiority over a comparator. Because it is unrealistic to expect that a new AED will be superior to a standard treatment used at optimal dosages, a number of designs have been developed whereby patients in the control group are randomized to receive a suboptimal (low-dose) treatment and required to exit if seizure deterioration occurs (Perucca, 2008). Under these conditions, efficacy is demonstrated when the investigational AED used as monotherapy at a fully effective dose is found to be superior to a low-dose comparator in end points such as time to exit, or time to first seizure.
The design most widely used to demonstrate superiority is the conversion to monotherapy design, in which patients with refractory epilepsy taking up to two AEDs are randomized to receive, under double-blind conditions, a full dose of the investigational AED and a low-dose comparator (Perucca, 2008). Background medications are then discontinued, and patients are required to exit the trial when prespecified criteria for seizure deterioration are met. Proof of efficacy is obtained by demonstrating that exit rates in the full-dose group are lower than in the group randomized to the low-dose comparator. This approach has allowed the efficacy to be demonstrated of several established AEDs, including felbamate, topiramate, lamotrigine, and oxcarbazepine, but not gabapentin and tiagabine (Beydoun & Kutluay, 2003). This design, however, has also met with increasing criticism from the epilepsy community, largely because randomizing patients with active epilepsy to a deliberately suboptimal treatment is ethically questionable (Perucca & Tomson, 1999).
In this issue, French et al. (2010) report important findings that can eliminate the need to include suboptimally treated controls in future conversion to monotherapy trials. The authors reviewed previous trials using this design and found that of the 10 existing trials, 8 used not only the same primary outcome (time to exit) but also very similar entry and exit criteria. Outcomes in all groups randomized to suboptimal treatment in these trials were carefully scrutinized, leading to the discovery that exit rates were substantially similar across groups. In other words, patients randomized to the control group in all trials using a similar design behaved relatively consistently in terms of their primary outcome. This finding set the conditions for the construction of a dataset of external (historical) controls, against which response to investigational AEDs could be compared in future trials. To obtain proof of efficacy, it would then be sufficient to demonstrate that the exit rate in a group of patients converted to monotherapy with a full dosage of the AED of interest remains significantly below the exit rates previously observed in historical controls. A critical point in this approach is to determine how significantly should be defined, and the authors propose to set the cutoff at a 65.3% exit rate, which represents the lower bound of the 95% prediction interval for the combined exit rates of historical controls. A less conservative cutoff could apply should the sponsor elect to conduct two separate trials, both of which would have to yield exit rates below the cutoff. It is important to note that these arguments have been accepted by the FDA, which now considers as valid the use of a historical control group in conversion-to-monotherapy AED trials (French JA, personal communication, 2009).
The acceptance by regulators of historical controls represents a major turning point in AED development. Indeed, several conversion to monotherapy trials using the new paradigm have already been initiated for many second- and third-generation AEDs, including levetiracetam (UCB, 2009a), lacosamide (UCB, 2009b), eslicarbazepine acetate (Sepracor, 2009), and brivaracetam (UCB, 2009c,d).
The main advantage of the historical control design is that all patients receive a promising AED at dose(s) expected to be fully effective, making the study more attractive to patients and to physicians (ICH, 2001). Moreover, this design removes the need for a parallel group on suboptimal treatment, which is ethically problematic in view of the risk of injury and even mortality associated with epileptic seizures. The design, however, is not free from concerns, some related to conversion-to-monotherapy per se and some related specifically to historical controls.
Concerns with conversion-to-monotherapy trials per se are manyfold (Perucca, 2008): First, these trials are not conducted in the population in which monotherapy is primarily used, that is, people with newly diagnosed epilepsy, and provide no information on the most important outcome measure in this population, that is, long-lasting seizure freedom. Second the end point used (time of exit) is a measure of seizure deterioration, rather than improvement in seizure control. Third, no information is provided on the optimal dose, or the range of effective dosages. In fact, partly because of the refractory nature of the population and partly because of the need to maximize the probability of demonstrating a difference versus controls, these trials typically investigate doses in the very high range. For example, 1,000 mg/day topiramate, 2,400 mg/day oxcarbazepine, and 300–500 mg/day lamotrigine, which were investigated in conversion-to-monotherapy trials of these AEDs, are doses far greater than needed in most patients with previously untreated epilepsy (Perucca et al., 2001). Finally, the principle that these trials truly investigate efficacy under monotherapy conditions can be questioned. As shown in Figure 2 of French’s article, historically a large proportion of patients in the control groups exited the study during the first few weeks of conversion, before concomitant medication could be fully withdrawn.
Concerns that are specific for the use of historical controls relate to the difficulties involved in minimizing potential bias (ICH, 2001). French et al. (2010) go some way toward assessing the influence of possible confounders, such as the withdrawal rate of concomitant AEDs and changes in background AEDs over time. Studies in other disorders, however, have shown an increase in placebo response in more recent years (Walsh et al., 2002; Stolk et al., 2003; Sysko & Walsh, 2007; Gallahan et al., 2010). As far as epilepsy is concerned, a meta-analysis of 27 published adjunctive-therapy trials of AEDs in adults with refractory partial seizures found a clear trend for placebo response rates to increase with time over the period 1996–2008 (Guekht et al., 2010). Although this was of borderline statistical significance (p = 0.064) in the study by Guekht et al., a more recent meta-analysis that included 63 randomized adjunctive-treatments in adults with refractory epilepsy identified a highly significant (p = 0.001) time-dependent increase in responder rates on placebo between 1990 and 2009 (Ryvlin P and Rheims S, personal communication, 2010). If an increase in placebo response over time also occurs in conversion to monotherapy trials, the use of historical controls in this setting could be invalidated. Although French et al. (2010) found no evidence of reduced exit rates in patients allocated to suboptimal treatment in the two most recent studies, the limited statistical power associated with the small number of trials prevents meaningful testing for a potential effect of time. Other concerns relate to the feasibility of replicating adequately the experimental conditions used in the historical trials. Like historical trials, currently ongoing trials using historical controls incorporate a double-blind design and randomization to two groups (Sepracor, Inc, 2009; UCB, Inc., 2009a,b,c,d). However, although in the historical trials one group was allocated to suboptimal treatment, in currently ongoing trials both groups receive doses expected to produce maximal or near maximal effects. Apparently, the consent forms used in the historical trials did not inform patients that they could be allocated to a suboptimal dose (French et al., 2010), a revelation that raises serious concerns about the appropriateness of the ethical review processes in clinical drug trials. Yet, physicians were informed that a suboptimal dose was used, and it is also possible that they may have discussed it with their patients. Whether this could have influenced patients’ and physicians’ expectations in a way that could confound the trial outcome is unknown.
French et al. (2010) should be commended for their scholarly work in analyzing response rates in control groups in earlier trials, and in setting the stage for a new conversion-to-monotherapy design in which patients are no longer required to receive suboptimal and, therefore, potentially hazardous treatments. Despite this important advance, however, currently applied monotherapy trial designs in epilepsy are still far from ideal. Regulatory requirements for the monotherapy indication differ widely between the two sides of the Atlantic, but at neither side are currently used study designs exempt from methodologic concerns.