Refractory Epilepsy: Clinical Overview


Address correspondence and reprint requests to Dr. Jacqueline A. French, Department of Neurology, Hospital of the University of Pennsylvania, 3 West Gates, 3400 Spruce Street, Philadelphia, PA 19104 U.S.A. E-mail:


Summary:  The incidence of refractory epilepsy remains high despite the influx of many new antiepileptic drugs (AEDs) over the past 10 years. Epidemiological data indicate that 20–40% of the patients with newly diagnosed epilepsy will become refractory to treatment. Factors that may be used to predict whether or not a patient will respond favorably to AED therapy include the type of epilepsy, underlying syndrome, etiology, and the patient's history of seizure frequency, density, and clustering. Environmental factors, such as trauma and prior drug exposure, and genetic factors that predetermine the rate of absorption, metabolism, and uptake of a drug by target tissue may also uniquely impact an individual and influence their response to AED therapy. Treatment resistance is, therefore, a multifaceted phenomenon. Since individuals with refractory epilepsy do not share a common reason for their treatment resistance, the use of targeted drug therapies may be our best option for improving treatment outcomes in this patient population. Pharmacogeneticists are currently attempting to understand the genetic basis of refractory epilepsy so that they can identify subgroups of patients who share a common genetic background and then target drug therapies to meet their specific needs.

Management of patients with refractory epilepsy is challenging because we do not fully understand how or why pharmacoresistance develops in a particular individual. Clinicians are constantly encountering patients who seem to have identical seizure disorders. They have the same etiology and the same lesion type and location on magnetic resonance imaging (MRI) scans, and yet, one will respond to treatment and the other will not. This situation is best illustrated by the case histories of two sisters who developed idiopathic generalized epilepsy in their teens. Electroencephalogram (EEG) findings demonstrated that they both had primary generalized spike and wave activity at the time of diagnosis. Despite apparently similar etiologies, and underlying genetics, these sisters had vastly different clinical outcomes. One sister has had only two generalized tonic–clonic convulsions in her entire life and is well controlled on carbamazepine, even though this is the wrong medication for her epilepsy syndrome. In contrast, the other sister has refractory juvenile myoclonic epilepsy (JME) and has failed multiple antiepileptic drugs (AEDs), including valproic acid, lamotrigine, and topiramate. Why did these sisters respond so differently to treatment? Is it due to genetic differences, early environmental factors, or to some unknown factor that uniquely impacted one and not the other? This article will briefly address each of these questions and, in doing so, lay the foundation for the remaining topics of discussion in this supplement.


Refractory epilepsy can have devastating consequences. Many individuals experience prolonged seizures or status epilepticus and, as a result, suffer bodily injuries requiring hospitalization. Others have shortened life spans because of the increased risk of sudden unexpected death that is associated with uncontrolled seizures. Studies have shown that patients with refractory epilepsy have significant neuropsychological, psychiatric, and social impairments that limit employment, reduce marriage rates, and decrease quality of life (1).

Evidence to date suggests that the incidence of refractory epilepsy is high. In a landmark study, Kwan and Brodie (2) followed the treatment of 525 individuals (9–93 years of age) with newly diagnosed epilepsy of all types (e.g., generalized or partial—idiopathic, symptomatic, or cryptogenic). Cumulative results over a 13-year period revealed that 63% of the patients were treatment responsive and had attained seizure freedom for at least 1 year. The remaining 37%, however, had poorly controlled seizures and were classified as treatment resistant. In addition, study results demonstrated that nearly half of the patients became seizure free on their first AED and that those who failed their first AED were less and less likely to gain control with each successive drug trial (Fig. 1). Patients who failed the first drug because of ineffectiveness had the lowest probability of future success (11%) when compared to those who failed because of intolerable side effects or idiosyncratic reactions (41–55%). These results suggest that newly diagnosed patients are either easy to control or difficult to control right from the start and do not become treatment resistant over time. Other data, however, derived from a sample of 333 epileptic patients who were presenting for resective surgery, suggest that the time to fail a second drug, the most common definition of refractoriness, can be quite variable (3). Although the majority of patients became refractory in ≤8 years, many did not fail their second drug until 9–34 years after diagnosis.

Figure 1.

Nearly 50% of all patients with newly diagnosed epilepsy attained seizure freedom on their first drug. Those who failed their first drug were less and less likely to gain control with each successive drug trial. [from Reference (2)].


A number of observational studies have attempted to identify phenotypic markers that can be used to predict refractoriness in patients with epilepsy. These include the type of syndrome, underlying etiology, patient history of seizure frequency and density, and EEG findings. Semah and colleagues (4) evaluated the ability of epileptic syndrome to predict ease of seizure control and final treatment outcome in a sample of 2,200 adult outpatients with epilepsy. The sample was stratified by syndrome and then subdivided according to whether the patients were easy to control and became seizure free on low levels of medication; difficult to control and required high doses of AEDs or multiple treatment regimens to become seizure free; or refractory to treatment. Not surprisingly, patients with idiopathic generalized seizures were most likely to become seizure free (82%) for ≥1 year and those with generalized symptomatic or cryptogenic seizures were the least likely (≈26%). More than half of the seizure-free patients in each of these groups were easy to control. Although treatment success rates were slightly higher in patients with symptomatic (35%) and cryptogenic (45%) partial seizures, they were still considerably lower than those with idiopathic generalized seizures, and less than half of the patients in each group were easy to control. At the University of Pennsylvania, we took a slightly different approach and began our study with the identification of 246 adults with drug-refractory epilepsy (≥1 seizure/month and failed ≥2 drugs). We then compared syndrome frequencies within this sample (5). The results revealed that 80.5% of the patients had localization-related epilepsy (e.g., focal or partial epilepsy), 11% had Lennox-Gastaut Syndrome, and 6.9% had primary generalized epilepsy. Very little is known about the influence of syndrome on the development of intractable epilepsy in children. Nonetheless, the findings of at least one study replicate those seen in adults. Berg et al. (6) followed 599 children with newly diagnosed epilepsy. After 30 months, it was determined that a diagnosis of localization-related idiopathic epilepsy was associated with the lowest risk (1.7%) of intractability and symptomatic generalized epilepsy with the highest risk (55%).

We know that there are also certain genetic syndromes that are most frequently associated with a benign clinical course and others that are associated with a very malignant course. For example, most children who inherit the monogenic mutation in KCNQ2 or KCNQ3, two subunits that regulate voltage-gated potassium channel activity and hence neuronal excitability, have benign familial convulsions that typically remit before adulthood (7). Only rarely are these syndromes associated with a malignant course. In contrast, juvenile myoclonic epilepsy, which also may arise from single gene mutations, does not remit but are very treatment responsive. Other nonremitting, genetically linked epileptic syndromes include the X-linked West syndrome, GEFS+, and cortical dysplasias.

Lesion etiology and location may be additional risk factors in refractory epilepsy (4,8). Acquired epilepsies, such as those following stroke or due to the development of a vascular malformation or tumor, seem to be much more treatment responsive than those associated with cortical dysgenesis, hippocampal sclerosis, or dual pathologies (4). Patients who develop epilepsy after a stroke, for example, have a 50% chance of becoming seizure free; half of the time the seizures will be easy to control and half of the time they will be difficult to control. Patients with tumors and vascular malformations have a similar prognosis. Hippocampal sclerosis and dual pathologies, on the other hand, are associated with much poorer outcomes. Comparatively few patients in these categories become seizure free, and their epilepsy is most often difficult to control. It is important to keep in mind that the incidence of refractory epilepsy will vary depending upon the source of the epidemiological sample. We know, for instance, that the prognosis of a patient with newly diagnosed epilepsy is much better than one with a more chronic history. This is best illustrated in patients with mesial temporal sclerosis. Whereas these individuals have a 42% chance of becoming seizure free at the time of presentation, they have only a 10% chance once they have been referred to a large neurology center (4,8).

The presence of multiple seizures prior to treatment, multiple seizures after treatment (e.g., failure to get control early), and seizure clustering (9) are all potential predictors of pharmacoresistance and poor treatment outcome (2,9,10). Although somewhat surprising, current findings suggest that the apparent poorer prognosis in patients who have multiple seizures prior to seeking medical treatment is probably not due to a kindling effect. Results from a recent study in which patients were treated after their first or second seizure demonstrated that the incidence of refractoriness did not differ significantly between the groups (11). Moreover, individuals in underdeveloped countries who have had seizures for several years before coming to the attention of the medical community show approximately the same response rate seen in those who are treated at the time of presentation (12).

Nonetheless, a history of multiple seizures prior to treatment is a definite red flag for clinicians. Most importantly, it tells us that the seizures are probably not generalized convulsions. Rather, they are most likely more subtle partial or complex partial seizures that have a much poorer prognosis. Multiple seizures are also often associated with a high seizure density, and this too can be a negative predictor of treatment outcome.

We have already discussed how failure to gain control of seizure activity early in the course of therapy can negatively impact treatment outcome (10,13–15). Additional findings suggest that the presence of seizure clustering in patients with newly diagnosed epilepsy is also associated with a high probability of treatment failure (9). This is particularly interesting in light of the fact that clinicians frequently see seizure clustering in patients with mesial temporal sclerosis, a diagnosis that carries a poor prognosis.

Other predictors of treatment resistance include early age of onset, presentation in status epilepticus, abnormal neurological exam, partial seizures at diagnosis, and mixed seizure types associated with developmental delay (6,16,17). Although there is little definitive data on the usefulness of EEG findings in the prediction of refractoriness, some studies have suggested that abnormal EEG activity is associated with a poorer outcome (18).

To recapitulate, cumulative evidence suggests that a patient's response to treatment is determined by multiple factors that are specifically related to their seizure disorder. More global human, environmental, and genetic factors, however, may also have a significant effect. Good treatment outcomes are dependent upon the physician's ability to correctly diagnose the epilepsy and then choose an effective treatment regimen. Patients may be labeled as refractory because they are on the wrong drug, taking too low a dose, or not dosing the drug frequently enough. Inconsistent compliance can also compromise treatment effectiveness. The role of these human factors in the determination of treatment responsiveness cannot be overstated.

Environmental factors, such as trauma or prior drug exposure, may either increase or decrease the likelihood that an epileptic patient will be refractory to treatment. New animal data suggest that when some drugs are given early in the course of treatment may produce both tolerance and cross-tolerance to similar drugs (19,20).

This is particularly relevant to the treatment of refractory epilepsy because it suggests that if a patient fails the first drug they may fail additional trials with similar drugs because of cross-tolerance.

Genetic factors may predetermine the rate and extent of drug absorption, its metabolism, blood–brain barrier permeability, and efficacy at target sites by affecting the activity of transporter molecules in the gut wall and blood–brain barrier, hepatic enzymes, and receptors on target tissues. Genetically conferred metabolic factors can affect how a drug is broken down and what metabolites are formed, thereby significantly impacting its safety. Factors such as these may explain why the two sisters, discussed earlier in this article, had such divergent clinical outcomes despite apparently identical seizure disorders.


Treatment resistance is clearly a multifaceted phenomenon. The fact that patients with refractory epilepsy do not present as a homogeneous group makes it difficult not only to design clinical trials but also to interpret and compare their results. This is particularly true of studies designed to examine the genetic basis of refractory epilepsy. The failure to obtain consistent results across studies may be due, in part, to the use of vastly different cutoff points to define treatment responders and nonresponders. Treatment responders have typically been classified as anyone who is seizure free for >1 year to >2 years; nonresponders as anyone who has one seizure/year to >10 seizures/year. Given these widely different definitions, it is not surprising that some studies have found an association between specific genes and treatment resistance and others have not. Therefore, to begin to identify the mechanisms underlying refractory epilepsy we must first establish a research definition of intractability and apply this definition consistently across all clinical studies. This is easier said than done because it requires the setting of descriptive norms and clear break-off points to separate outliers from our designated population of refractory patients. Variables, such as seizure density, number of drugs failed, seizure reduction, and length of the seizure-free period, need to be defined. Is a patient refractory if he/she fails two drugs or 20 drugs? Is there a difference between the patient who has one to two seizures/month and one who has one to two seizures/year? Should the patient who goes from having 20 seizures/day to one seizure/day be classified as treatment responsive or nonresponsive? Once we have agreed on a core definition of intractability, additional factors must be considered to refine our categorization of patients as either responders or nonresponders. Was the diagnosis confirmed? How do we classify the patient who becomes seizure free only at supratherapeutic doses? Do we need serum drug concentrations and pharmacokinetic end points to separate patients who are noncompliant or fast metabolizers from those who are receiving adequate doses of AEDs? How many seizures are people having per unit time? Is a patient intractable if they have one breakthrough seizure? These are all important questions that need to be addressed.


New AEDs are continually being developed to address the needs of treatment-resistant individuals, which make up one-third of the total patient population (21). To date, there are more than 20 drugs available and eight of these were developed in the last 10 years. Despite the influx of these new drugs, however, treatment outcomes in epilepsy remain poor. Fig. 2 presents a compilation of data derived from studies assessing the efficacy of new AEDs used as adjunctive therapy in patients who were refractory to standard AEDs (22–26). It is apparent that even once placebo effect is subtracted, the highest doses of the most efficacious adjunctive therapies, topiramate (TPM), oxcarbazepine (OXC), levetiracetam (LVT), and pregabalin (PGB) produced a 50% reduction in seizure frequency in only 32–37% of refractory patients. Treatment responsiveness was defined as a 50% reduction in seizure frequency in these studies because it is highly unlikely that a refractory patient will become seizure free. Indeed, a recent report found that only 38 of 246 (15.5%) patients with drug-refractory epilepsy (≥1 seizure/month, failed ≥2 drugs) attained seizure freedom (≥6 months) during a 3-year observation period (5). Once attaining seizure freedom, however, all patients remained seizure free through study end point. Although these findings indicate that we are definitely making progress in reducing the number of refractory patients, a remission rate of 5% per year is still much too low. We need to greatly improve these statistics. The simple solution would be to develop more efficacious drugs that have fewer side effects and a wider margin of safety. Fig. 3 compares the risks and benefits of two potential drugs. The therapeutic effects of Drug 1 plateau and about 50% of the patients become seizure-free at doses well below the toxicity limit. The therapeutic effects of Drug 2, however, increase as a function of dose. Although this drug has the potential to produce seizure remission in 100% of the patients, it does so at the expense of toxicity. If we could improve tolerability to Drug 2 by widening its safety window, we could conceivably increase the number of patients who would become seizure free and thus potentially increase our remission rate in refractory epilepsy. Alternately, we could attempt to identify patients who would be tolerant to higher doses of Drug 2 and, in doing so, enhance efficacy and the probability of treatment success.

Figure 2.

The percentage of patients, refractory to standard AEDs, who achieved a 50% reduction in seizures (minus placebo) while taking one of the new AEDs as adjunctive therapy is very low. GTB, gabapentin; LTG, lamotrigine; TGB, tiagabine; TPM, topiramate; OXC, oxcarbazepine; LVT, levitiracetam; ZNS, zonisamide; PGB, pregabalin. [from References (22)–(26)].

Figure 3.

The potential efficacy and toxicity of two drugs can differ greatly as a function of dose. The therapeutic effects of Drug 1 plateau and half of the patients become seizure free at nontoxic doses. Drug 2 has the potential to produce seizure freedom in 100% of the patients, but it does so at the expense of toxicity.

It appears that there are at least three approaches to the treatment of refractory epilepsy. We can continue on our present course and accept a 5% per year remission rate, try to develop a blockbuster magic bullet that cures everyone, or attempt to target drug therapy for each individual patient. Most would agree that a 5% per year remission rate is unacceptable. And it is probably unrealistic to think that there will ever be a magic bullet that works for everyone. However, a magic bullet may exist for certain subsets of refractory patients, and all we need to do is find it. This is the goal of pharmacogenomics—to understand the role of genetics in the development of intractable epilepsy, identify subgroups of patients who share a common genetic background, and then target drug therapy to meet the specific needs of these groups.