Address correspondence to Nancy R. Temkin, Ph.D., University of Washington, PO Box 359924, 325 9th Avenue, Seattle, WA 98104-2499, U.S.A. E-mail: firstname.lastname@example.org
Posttraumatic epilepsy presents an ideal target for prevention efforts. Traumatic brain injury (TBI) is common, characteristics that put people at high risk such as penetrating injury or subdural hematoma or provoked seizures are easily identified, and the latency between the injury and the onset of epileptic seizures is frequently short. Several drugs have been tested for their ability to prevent provoked seizures and epilepsy after TBI. We describe the design of those studies and their results. Phenytoin and carbamazepine significantly reduce the incidence of provoked seizures. Phenobarbital and the combination of phenobarbital and phenytoin also look promising for reducing provoked seizures, but small sample sizes in the studies evaluating these drugs do not allow definitive conclusions. None of the drugs studied (phenytoin, phenobarbital, their combination, carbamazepine, valproate, or magnesium) have shown reliable evidence that they prevent, or even suppress, epileptic seizures after TBI. For most of the regimens tested (the phenytoin/phenobarbital combination being the exception), the best estimate of effect is under a 25% reduction in posttraumatic seizures, well less than the 50% reduction most studies were designed to detect. The evaluation of the tested drugs has serious limitations, however, and antiepileptic drugs (AEDs) developed since 1980 and other compounds have barely been tested at all. Better understanding the process of epileptogenesis, testing treatments that demonstrate antiepileptogenic effects in the laboratory, and performing thorough preclinical and phase II evaluations before attempting definitive trials should greatly improve the chance of identifying ways to prevent posttraumatic epilepsy, providing the ultimate cure for this condition.
As indicated in earlier articles in this supplement, posttraumatic epilepsy is common. About 6% of epilepsy is attributed to traumatic brain injury (TBI). More than a million people in the United States experience a TBI each year (Langlois et al., 2006). Most are mild, but even mild TBI increases the subsequent risk of epilepsy by about 50%. Moderate TBI increases the risk by about 3-fold, whereas severe TBI is associated with about a 17-fold increase in risk, which translates into more than 15% of people with severe TBI developing epilepsy (Annegers et al., 1998). For some subgroups—such as those with penetrating injury, a depressed skull fracture, a subdural hematoma, or early (provoked) seizures—more than 20% develop posttraumatic epilepsy (Temkin, 2003). About one-third of those who ultimately develop posttraumatic epilepsy (PTE) have their first unprovoked seizure within 3–4 months, and two-thirds have their first seizure within 24 months (Jennett, 1975).
Conceptually, the TBI begins an epileptogenic process that culminates in unprovoked seizure after a period of weeks or months or years. The hope is that somewhere in this silent period is a window of opportunity in which an appropriate treatment will stop the process, so that the person will never develop PTE. In this paper, I will discuss the human experience to discover and evaluate such a regimen. I will not discuss treatment of PTE other than to indicate that I know of no studies to rigorously evaluate a treatment specifically for PTE and that the treatment recommendations are to treat PTE the same as one would treat seizures of the same type of any etiology (Langendorf et al., 2008).
Quite a few clinical trials have been carried out to evaluate whether a regimen prevents the development of PTE. The designs have been roughly similar. Most include adults with a high risk of developing PTE. Time from injury to the start of treatment varied among studies with some requiring treatment within 12 hours of injury and others not starting for up to a month after injury. The participants are monitored for the occurrence of clinical seizures for between 6 months and 3 years. Most studies include a period when there is monitoring for seizures after the drug has been stopped. It is only in this period when one can determine whether a drug has truly affected the epileptogenic process. If the drug under evaluation is an antiepileptic drug (AED), during the period of treatment, one cannot tell if a reduction in epilepsy is due to seizure suppression or an effect on the epileptogenic process. Reports of the trials have not generally separated out the period of treatment from the untreated observation period. This is particularly problematic in positive trials, that is, those showing an effect. Because the regimens are expected to have a seizure suppressing effect, a regimen that has no effect during treatment and after can probably be safely assumed to not be preventing or postponing epileptogenesis. Most trials considered the first (unprovoked) seizure to be the endpoint, not waiting for the two unprovoked seizures required for the usual definition of epilepsy. If a regimen truly stopped the epileptogenic process, one would expect to see that reflected in the percentage of patients developing a first unprovoked seizure as well as the percentage developing epilepsy.
Phenytoin (PHT) is the drug that has been most frequently tested for an antiepileptogenic effect. It was first tested in the 1940s (Hoff & Hoff, 1947), with early trials reporting a positive effect. Hoff reported that 6% of cases receiving PHT developed seizures, whereas 51% of those without treatment did. Not unexpectedly, the study did not meet the current standards for rigor. The method of assigning cases to treatment is unclear, the treatments were not masked, and early seizures (provoked seizures in the first week after injury) were not separated from late seizures (unprovoked or epileptic seizures occurring after the first week). Subsequent studies used more rigorous designs, with random assignment to treatment (McQueen et al., 1983; Young et al., 1983; Temkin et al., 1990, 1996), identical-appearing placebos for the control group allowing blinding or masking of treatment, and separation of the early and late periods (Young et al., 1983; Temkin et al., 1990; Pechadre et al., 1991; Temkin et al., 1996). Some studies monitored serum concentrations and attempted to keep them in the therapeutic range for epilepsy treatment (Young et al., 1983; Temkin et al., 1990). Fig. 1 shows results of the largest study (Temkin et al., 1990), with a substantial effect of PHT on early seizures (panel A), and a lack of a positive effect on late seizures, even during the first year when participants were receiving treatment (panel B). A meta-analysis of all the trials (Temkin, 2001) shown in Fig. 2 yields a similar conclusion.
Phenytoin and Phenobarbital
The combination of PHT and phenobarbital (PB) was evaluated in five studies [(Penry et al., 1979; Popek & Musil, 1969) and unpublished studies led by Brackett, Marshall, and Locke summarized in (Temkin et al., 1996)]. These trials had substantial problems with recruitment and retention, and several were stopped before reaching their planned sample sizes. Although the best estimate of effects is promising (Fig. 2), the small sample sizes led to confidence intervals so wide as to allow no meaningful conclusions.
Phenobarbital alone has been evaluated in two studies [(Manaka, 1992) and the study by Locke summarized in (Temkin et al., 1996)]. Confidence intervals are wide and results for late seizures at least are not encouraging (Fig. 2).
Carbamazepine (CBZ) has been evaluated in one study (Glötzner et al., 1983), with results similar to those for PHT (Fig. 2)—a reduction in early seizures, but no effect on epilepsy even during the period of treatment.
Valproate (VPA) has been evaluated in only one study (Temkin et al., 1999), and the comparator was 1 week of PHT, making the effect on early seizures difficult to interpret. The early seizure rate was higher on VPA than on PHT, although with the small number of early seizures, the difference is not statistically significant. VPA showed no positive effect on late seizures.
Magnesium (MG), the only drug tested that is not frequently used to treat epilepsy, was tested in one trial (Temkin et al., 2007). Most subjects received PHT clinically for the first week, so both groups had few early seizures. The confidence interval is very wide for early seizures, but the results give no suggestion of a positive effect of 5 days of MG on either early or late seizures (Fig. 2).
Although five drugs have been rigorously tested for an antiepileptogenic effect, none has been shown to exert such an effect. PHT and CBZ suppress early seizures, but none of the tested regimens show a positive effect on late seizures, even during treatment. For most of the regimens tested (the phenytoin/phenobarbital combination being the exception), the best estimate of effect is under a 25% reduction in posttraumatic seizures, considerably less than the 50% reduction most studies were designed to detect. Interestingly, for three of the regimens, the best estimate of effect is an increased rate of epileptic seizures in those treated with the active drug. This raises the intriguing question—if these drugs are effective in treating posttraumatic epilepsy, why isn’t there a positive effect at least during the period of treatment?
The trials that have been performed to date have substantial limitations. None has done electroencephalography (EEG) monitoring to evaluate subclinical seizures. All studies treated during the early period, making it impossible to separate suppression of early seizures from an effect on the process of developing early seizures. Most studies had small samples, with only four studies including more than 200 cases. Most studies had relatively short periods of observation after the drug was stopped, although the lack of overall effects seen makes moot the limited ability to separate a seizure suppression effect from antiepileptogenesis. Most studies did not monitor compliance or test drug concentrations. Many studies had high rates of loss to follow-up. Finally, some studies were not blinded.
The studies that have been done had little laboratory work to inform their design. Laboratory models of PTE were not available when the studies were being designed. Therefore, decisions on when to start the drug, what dose to use, and what duration of treatment to use were made without benefit of knowing what worked in the laboratory.
The range of drugs tested has been narrow. Most of the drugs tested are ones known to suppress seizures once a person has epilepsy, and only the older drugs approved before 1980 have been tested so far. Newer AEDs need to be evaluated in the laboratory at least, and, if promising, in clinical trials. And the drugs that stop the process of epileptogenesis may be different from those that suppress seizures in someone who has already developed epilepsy.
The mechanisms of epileptogenesis are not well understood. We need more laboratory work to better understand the process of epileptogenesis and to identify targets and therapies that might stop the process before epilepsy develops. Then we need translational work to determine the window of opportunity and doses and durations of treatment needed. Thorough phase II clinical trials are needed to determine how to best translate laboratory findings into treatments for humans. Some of these topics are discussed in later articles in this issue. This additional work should greatly improve the chance of identifying ways to prevent posttraumatic epilepsy, providing the ultimate cure for this condition.
Supported by 5 U01 HD042653 from NIH/NICHD/NCMRR.
Disclosure: The author declares no conflicts of interest.