Addressing the problems of treatment failure in epilepsy: You cannot fix what you do not understand

"No seizures, no side effects" has been the mantra for epilepsy therapy discovery for the past several decades. Unfortunately, we are no closer to achieving that goal today than when it was first uttered. The prevalence of treatment failure remains virtually unchanged despite the many new epilepsy medications available, and side effects continue to be a major cause of failure.1-7 Perhaps one of the greatest obstacles that we face in reaching the "promised land" of seizure freedom devoid of interfering consequences is that we really do not know why the medications fail or why side effects are present for some patients but not for others. We are often told that we learn from our mistakes, but in the treatment of epilepsy, we move on from our failures without ever learning why a drug failed in some but not other patients. One of the contributing obstacles may be rooted in how the epilepsy community identifies new treatments. Fundamentally, epilepsy is a chronic disease that requires ongoing treatment to suppress the seizures that are unpredictable in occurrence. It is also a disease that has many different causes and types of pathophysiology. At present, the discovery process relies primarily on short-term or acute testing (single doses, or at most, several weeks of repeated dosing), with the end point of a measurable effect on the duration or severity of single evoked seizures at doses that do not produce consequences on gross motor behavior such as the rotorod test.8-10 Complete suppression of the seizures is rarely seen, and there are no consistent tests for subtle side effects. In the clinic, success is measured primarily in the self-reported reduction of the number of spontaneous epileptic seizures. From the patient's perspective, true success is an all-or-none phenomenon, which is determined over months and years of treatment. Significant but incomplete reduction has less of a meaning for patients, because their lives are still compromised (sometimes devastated) by uncontrolled seizures. There are reports of improved seizure control in Dravet syndrome or Lennox–Gastaut syndrome with specific drugs (e.g., clobazam, stiripentol, fenfluramine), but complete control is rarely seen.11 A statistically significant reduction in seizure frequency or duration is considered a scientific success, but patients with epilepsy want—or even need—total control of seizures, because even a single seizure will not only interfere with their lives but put them at increased risk for accidental injury or death (e.g., automobile accidents). Side effects of these treatments are mentioned, often generically, but the focus of the reports remains on seizure control. Although side effects or toxicity are certainly considered a "problem" for the patient, they are often not considered as a major contributor to treatment failure. The current definition of treatment resistance specifies the failure of two treatments that are otherwise well tolerated.12 The reality is that most treatments are adjusted until side effects interfere with quality of life, at which point the treatment is considered a failure.13, 14 If an absence of side effects is required to designate a compound as a treatment failure, then most treatments can never be called failures, because they are most commonly stopped as a result of side effects. In some trials, withdrawal from treatment may be as high as 40% when unblinded, open-dose adjustment is allowed.3-5 At present, we have no good explanation for treatment failure. Moving forward requires that we identify the basis for failure so that treatments can be identified that minimize the chances for failure over the longer term. The basis for treatment failure is a topic that has been the focus of cancer and microbiological research for many years.15, 16 It is time for it to become a critical focus of epilepsy research. There is nothing in this paper that should be considered a criticism of the current discovery process. This process is designed to identify the potential of a compound to suppress seizures, and it has been successful for this purpose, but it has not reduced the failure rate of drug treatment.2, 17 We are hypothesizing that improved outcomes will not come simply from the discovery of additional compounds that can suppress seizures (which should and will continue); rather, it will come from understanding why treatments fail, particularly from side effects. Until we understand the basis for our failures, progress will be much more difficult. New approaches to discovery are essential.18 Treatment failure is not a uniform entity; instead, there are several patterns that we see in the clinic.19 A common pattern is an incomplete reduction in seizure frequency or severity, but some seizures continue to interfere with the patient's life. In another pattern, the medicines have no discernable effect on the seizures. Still another response pattern is that seizures are controlled for varying periods before they return and are no longer controlled. These different observations have several potential mechanistic implications. To help understand the possible underlying causes, we propose that the different patterns of failure include both (1) the effect of the medicine on seizure occurrence and (2) the timeline for the reappearance of seizures following a period of seizure control. A potential confounding issue to our interpretation of the patterns of failure is the natural history of the disease; if untreated, would the seizures get better or worse? The mechanisms underlying treatment failure (patterns of response and the time to appearance of breakthrough seizures) are likely overlapping. The categories that we are proposing in this paper should be considered first steps in understanding failure and may well need adjustment as we develop a better picture of the failure patterns and the natural histories of the different epilepsies. These concepts may be applied to the response to a single treatment or to the collective response to several different treatments. Examples would be, "There was an incomplete reduction of seizures on carbamazepine, levetiracetam resulted in no change in the seizures, but the seizures were controlled by lamotrigine" or "Three different medications resulted in no change in the seizures." We will also suggest potential mechanisms as illustrations, but other causes that we do not present here are possible. These patterns of failure apply to an individual's response. A person with an incomplete response to one medication may respond very well to another.3-6, 20 Another person may have a very different response—or even the opposite response—to the same medications. One should also note that some of the patterns appear only after a period of treatment, suggesting that functional changes have occurred in the brain over time. Treatment may significantly reduce the frequency or severity of seizures, but seizures still occur. In an individual with multiple seizure types, some seizures (but not all) may be suppressed (Figure 1A). The pattern could be the result of variations in the distribution of the drug in the brain or alterations in the local pharmacokinetics of a drug so that local levels of the drug in the seizure circuits fall below the threshold for seizure control more quickly. Little is known about how drugs move through the brain and how pathology affects the movement. Neither seizure frequency nor the clinical severity of seizures is significantly affected by treatment (Figure 1B). The pattern suggests that the treatment is directed at the wrong target or is unable to reach the critical regions of the seizure circuit. Seizures are initially controlled, but over time the seizures return, in some cases returning to pretreatment frequency and severity with no change in dose (Figure 1C). In some cases, the seizures have been controlled for months to years. While the patient is still receiving the therapy, the seizures return and may not come under control again. The pattern could be related to ongoing changes in the brain, either by the natural history of the underlying disease or from treatment-induced changes in the brain. The changes could derive from alterations in the subtype or distribution of specific receptors or channels, or there could be changes in the access of the drug to key components of the seizure circuit. Change in generic source is becoming a more common cause of treatment failure and should be considered when seizures return after a period of control.21-23 This particular situation suggests a pharmacokinetic cause. Seizures that impair awareness and function are controlled by treatment, but seizures that do not interfere (e.g., auras) with a person's life continue despite treatment (Figure 1D). Although this response to treatment fits the current definition of resistance, this pattern is typically considered a treatment success even with the persistence of auras, and patients are often not considered for more aggressive treatments. However, this pattern may provide us with mechanistic information. As noted earlier, there may be variations in where the drug can go, so the drug goes to regions involved in seizure spread but not to where the seizure actually begins.24 It may also be considered as a variant of an incomplete response. In this situation, there is an underlying cause that affects the brain (e.g., autoimmune encephalitis or a metabolic disorder; Figure 1E). Seizures do not respond to standard medical treatment but resolve when the cause of the disease is addressed. This category suggests that conventional treatment cannot influence the seizure-causing aspect of the underlying disease. Some patients do not respond to the first medication(s) tried but do respond to subsequent medications.2, 3, 5, 25 The drug that successfully controls the seizures varies from person to person. This observation has many mechanistic implications, such as differences in the distribution of or sensitivity to the individual drugs. Some drugs may be getting to the main target and others not. In some cases, it is simply the wrong drug for that type of epilepsy (e.g., carbamazepine for generalized epilepsies). Side effects remain one of the most common causes for treatment failure. The drugs are unable to reach a therapeutic level in the seizure circuit but are surpassing toxicity levels in other regions of the brain (Figure 2). Although efficacy implies that an effective tissue level has been achieved in critical components of the seizure circuit, toxicity may occur because toxic levels are reached in regions outside the seizure circuit. These differential sensitivities could in part be the result of the many factors that control the distribution of drug in the brain or of the drug target (e.g., sodium channel) having a different sensitivity to the drug in different parts of the brain. The difference between the two thresholds for the patients on a given drug can be wide or narrow in either direction. When we talk about people failing a drug because of toxicity or side effects, as we do with patterns of failure in efficacy, we present that failure as a monolithic entity. "The patient couldn't tolerate three different drugs." Most commonly that observation is left at that binary level of classification. Failure is failure, and we move on to the next treatment trial. However, there are many symptoms that contribute to poor tolerance for a medication, and at present we have no classification of side effect patterns that might provide insight about potential mechanisms. The range of side effects is broad. Motor, balance, and coordination issues are commonly reported, as is blurring of vision. However, others include sedation, cognitive slowing, and alterations in behavior. Systemic or metabolic consequences, including significant weight gain or loss, are treatment-associated issues. Headaches have limited the usefulness of some medications. Allergic reactions may prevent or limit the use of a drug, but overall they contribute less to treatment failure and are not an issue related to the nervous system. Other than the broad classification of "stopped taking medication" in clinical trials, which has been reported at rates up to 40%, we do not really know what the patterns are for any given drug. The rate of side effects may actually be higher, as many patients will accept some side effects in exchange for seizure control. Just as we have found it useful to classify the epilepsies, we also need to classify the patterns of failure to model them better in preclinical testing. In this commentary, we have described several possible features of treatment failure in epilepsy. These patterns emphasize that failure is multifactorial. At a minimum, it should be defined in terms of (1) the pattern of failure (e.g., incomplete reduction, persistent noninterfering) and (2) the temporal pattern of when the failures appear (present from the beginning, or relapse after months). The different patterns of failure and the different time frames, along with the variable natural histories of the epilepsy syndrome, suggest several causes that require investigation, and we need to model them. Much of the current medication screening does not target actual seizures, but rather behavioral surrogates of seizures (e.g., hind-limb extension as a surrogate for the tonic component of a tonic–clonic seizure). Those screening protocols that do target electrographic seizures are generally aimed at acute, induced seizures. The studies generate consistent data on the reduction of some aspect of the seizures, but the effect on spontaneous seizures is rarely evaluated over the weeks or months that would reflect clinical reality. Time to appearance of treatment failure is usually recognized over longer periods after multiple doses of the medication. We are not modeling that aspect of failure. A much better understanding of the interplay between the clinical reality and the laboratory is critically needed. The laboratory must provide the data that establish potential mechanisms of treatment failure that can be tested in the clinic, and the clinic needs to provide much greater detail about the many features of failure to allow for the design of translational research more focused on the actual clinical problem, which would lead to a better alignment of the two worlds. Simply classifying patients as having drug resistance will not provide insight into the mechanisms. One step could be to examine patients who failed a first drug and compare them to patients with the same form of epilepsy who did well. Would studying patients with persistent auras provide more mechanistic insight into drug resistance, as these seizures fall into the category of persistent noninterfering/incomplete control? Animal models that have both patterns may allow us to develop testable hypotheses. Table 1 highlights some of the major issues that will need to be addressed going forward. Classify clinical patterns and timelines for the appearance of drug failure (epidemiology). Identify treatment targets and their locations. Define effects of drugs on key seizure circuits. Evaluate regional variations in drug distribution that might contribute to neurological side effects. Examine regional variations in drug distribution and the contributions of local pathology to drug distribution and local pharmacokinetics. Determine whether long-term exposure to repeated seizures and/or drugs alters brain structure and pharmacokinetics. Most importantly, understanding treatment failure in epilepsy will require a change in our approach to mechanistic studies because of the longer time that will be needed to perform them. It will require a significant change from our relying primarily on results from acute or short-term models. Perhaps the most important first step is for the epilepsy community to agree that there must be a much greater emphasis on the clinical aspects of failure, including toxicity. Laboratory studies of the mechanisms underlying failure and toxicity will need to involve more time- dependent changes into protocols, and clinical studies will be needed that better define classes of failure and the specific conditions where they occur. There are several studies that have suggested that different patterns of failure can be replicated. At least two studies have shown that with repeat dosing, levetiracetam rapidly loses its efficacy, even with high doses.30, 31 There are also studies that demonstrate that with longer term dosing, a pattern of incomplete response can be seen over several weeks of chronic treatment. Using existing models over the longer term may allow researchers to develop new insights into the effects of drugs on epilepsy and to analyze the time-dependent or chronic aspects of how epilepsy medications lose their efficacy or cause toxicity. Without these kinds of studies, solutions to the critical problem of failure will continue to evade us. Other than suggesting new epidemiological studies to classify patterns of failure and types of side effects, many of our points have been made before, albeit in different ways. Calls for new ways of identifying more effective therapies have been made for more than 2 decades.32, 33 It is unfortunate that we are still having to make those calls. We need to focus on the reasons for drug-treatment failure in epilepsy. None. E.H.B. receives support for services from the International League Against Epilepsy through the University of Virginia. F.E.D. has equity interest in and previously received discounted equipment and consultant fees from Epitel. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.