Article Text


New drug treatments for epilepsy
  1. A G MARSON,
  1. Department of Neurological Science, Clinical Science Centre for Research and Education, Lower lane, Fazakerley, Liverpool L9 7LJ, UK
  1. Professor D W Chadwick

Statistics from

After the introduction of valproate in 1973, excluding benzodiazepines, there was a 19 year gap before the introduction of vigabatrin, the first in a series of new antiepileptic drugs to be developed and licensed. At the time of writing gabapentin, lamotrigine, oxcarbazepine, tiagabine, and topiramate have a licence in the United Kingdom. Zonisamide and levetiracetam are licensed in some countries outside the United Kingdom. In this article, we review some of the evidence for the effectiveness and tolerability of these new drugs. We also attempt to draw attention to lessons learned and the conflict between the needs of the pharmaceutical industry, the clinician, and the patient.

New antiepileptic drugs as add-on treatments


New antiepileptic drugs are licensed as add-on treatments in the first instance, efficacy and safety having been demonstrated in randomised placebo controlled add-on trials, with additional safety data predominantly derived from follow on studies. Results from follow on studies are often difficult to interpret because comparisons are no longer randomised.

Trials usually recruit patients with drug refractory partial seizures. Patients are first of all observed during a prerandomisation baseline period of 8–12 weeks duration. Provided they have a sufficient a number of seizures (usually four a week), they are randomised to have either active drug or placebo added to their regime, and are followed up for 12–16 weeks. Efficacy is assessed by comparing individual patient's seizure frequency during the treatment period with the baseline period. Median reduction in seizure frequency and the number of patients with a 50% or greater reduction in seizure frequency (responders) are usually reported as efficacy outcomes.

The table summarises the results of intention to treat analyses from a systematic review of placebo controlled add on studies investigating gabapentin, lamotrigine, tiagabine, topiramate vigabatrin, and zonisamide in patients with drug refractory localisation related seizures,1 2 and in addition shows data for levetiracetam (UCB Pharma) and remacemide (AstraZeneca Pharmaceuticals) and oxcarbazepine (Novartis). For gabapentin, tiagabine, topiramate, and vigabatrin, the table shows estimated proportion of 50% responders/dose derived from regression models. For lamotrigine, levetiracetam, and zonisamide, the table shows odds ratios for a 50% reduction in seizures compared with placebo across a range of doses.

Efficacy of new drugs as add-on therapy

Regression models were not able to distinguish between doses of 400–1000 mg/day of topiramate, suggesting a plateauing of therapeutic response. Results were similar for vigabatrin, where regression models were unable to distinguish between doses of 3000–6000 mg/day. Results for gabapentin and tiagabine show no plateauing of therapeutic response and for gabapentin in particular, it is likely that optimal doses have not been tested. For remacemide, estimates of efficacy are low, and an odds ratio across doses shows no significant effect. A significant effect was found for higher doses (800–1200 mg/day)


There is some evidence for efficacy of topiramate against generalised tonic clonic seizures in drug refractory patients,3 but a trial comparing 1200 mg/day of gabapentin with placebo found no significant difference between treatment groups.4 There is no evidence from randomised clinical trials to support or refute the efficacy of the other new drugs in drug resistant generalised epilepsies, other than in Lennox-Gastaut syndrome (see below). It is therefore surprising that in the absence of such evidence, lamotrigine is considered by some to be a first choice add-on treatment for patients with generalised epilepsy syndromes such as juvenile myoclonic epilepsy.


There is some evidence of efficacy of lamotrigine5and topiramate6 for patients with this difficult epilepsy syndrome; however, trials have been of short duration (16 weeks), and treatment effects small. For example, in a trial comparing lamotrigine with placebo, 33% of the lamotrigine versus 16% of the placebo group had a 50% or greater reduction in seizure frequency.

Limitations of existing evidence for add-on use of new antiepileptic drugs

Gowers warns, when discussing evidence for the efficacy of bromides, that: “Indications for special treatment in epilepsy is a subject of greatest importance. There is no point in therapeutics however more open to fallacy, or on which more generalisations have been published, which subsequent observation has proved inaccurate.”7

His comments are still of relevance today. Although these trials were predominantly undertaken for the purpose of getting a license for add-on use, care is needed when trying to use these results in clinical practice, and many important clinical questions remain unanswered:


There is convincing evidence that reducing a patient's seizure frequency is the most important contributor to a change in quality of life.8 However, the true meaning of a 50% reduction in seizure frequency to the individual patient is difficult to put into context. One hundred per cent reduction in seizure frequency (seizure freedom) would be easier to interpret; however, this is an unrealistic outcome for existing treatments, given that seizure free rates in these trials are typically less than 10%, and there are no magic bullets on the horizon.


To choose between these drugs, the clinician needs to know how they compare with each other, and ideally needs the results of head to head randomised clinical trials that have used clinically meaningful outcomes. However, few comparative trials have been undertaken, and those that have lack the power to inform clinical practice.9

For drug refractory partial seizures, indirect comparisons can be made using the results of the systematic review already quoted.1 2 Overlapping confidence intervals around summary estimates indicate that no significant differences were found; however, the existence of important therapeutic differences has not been excluded either. Others have used these data to infer that significant differences do exist, but have used flawed methods.10 11


The trials reviewed and hence any meta-analysis included a heterogeneous population of patients with localisation related epilepsy, and the trials and meta-analysis provide overall estimates of effect for this heterogeneous population. It may be that certain drugs will be more effective for certain epilepsy syndromes, but current data do not provide evidence to support or refute this hypothesis. Any evidence to support this hypothesis from existing trials would need to come from subgroup analyses. However, the trials and meta-analyses are of insufficient size to allow reliable results from subgroup analyses and this issue will also need to be considered in future head to head trials.


Rational polytherapy is currently in vogue for the treatment of epilepsy.12 The theory is that combining drugs with differing mechanisms of action (for example, combining drug with a sodium channel action and a GABAergic drug) will be more effective and cause fewer side effects than combining drugs with similar modes of action. There is, however, no hard evidence from randomised control trials to support this approach.


Although a significant proportion of patients benefit in the short term, the proportion achieving long term benefit remains uncertain. Longer term follow up studies indicate continued benefit, but their interpretation is difficult as they are no longer dealing with a randomised comparison, and are confounded by selective drop out of non-responders.

A significant proportion of patients have these drugs withdrawn in the long term, despite responding in the short term. For example, an audit of patients recruited into placebo controlled add-on studies with lamotrigine or vigabatrin showed that 86% of those still alive had had the trial drug withdrawn after 6–8 years.13 This is despite 21%-31% of patients in the original trials having a 50% or greater reduction in seizure frequency. A clinic based audit of less refractory patients found that 45% of patients treated with add-on topiramate had had this drug withdrawn at 12 months.14


Although some studies have been conducted in children,15-17 the great majority of trials have been conducted in adults. Even for children with the same epilepsy syndromes as adults recruited into individual studies, great caution is needed when trying to extrapolate results from adults to children. In addition, children present with a unique range of epilepsy syndromes that need to be investigated in randomised controlled trials. One example is infantile spasms, for which there is evidence for the efficacy of vigabatrin,18 which has become the treatment of choice for this condition.


As most reported trials of new antiepileptic drugs are of relatively short duration, they are unable to inform us about long term toxic effects. They are also unable to inform us of the risks of rare side effects such as aplastic anaemia, a side effect that led to the withdrawal of felbamate in the United Kingdom shortly after it was given a licence. For example, if the incidence of aplastic anaemia on a drug were one in 5000, a total of 15 000 patients would need to be exposed to that drug to give a 95% chance of seeing a single case.19 However, nowhere near this number of patients are exposed to new antiepileptic drugs before they are marketed. A rigorous system of postmarketing survailence is required to pick up these rare but life threatening side effects as soon as possible; however, no such system is in place in the United Kingdom or elsewhere.

The visual side effects of vigabatrin are another example of a serious side effect picked up after marketing. The first reports were in 199720 and authors report concentric visual field constriction with relative temporal sparing. In 1995, Kalviainan et al 21 reported a monotherapy study comparing carbamazepine and vigabatrin. They have recently revisited this cohort and found that 41% of the vigabatrin group had concentric visual field abnormalities, compared with none in the carbamazepine group.22 These visual field abnormalities seem irreversible, and have persisted despite vigabatrin withdrawal.

As a result, vigabatrin cannot be recommended other than a treatment of last resort for patients with drug refractory partial seizures, but may still be given as monotherapy for children with infantile spasms. Hoechst-Marion-Rouselle (maker of vigabatrin) recommends that visual fields be assessed before starting treatment for any patient with a developmental age of 9 or older. Fields should then be reassessed at 6 monthly intervals for 3 years and annually thereafter. Static perimetry (Humphrey or Octopus), or kinetic perimetry (Goldmann) are recommended (Hoechst letter). This is more problematic for patients with a developmental age of less than 9, as there is no established method for detecting visual field abnormalities in this group, and is a particularly difficult issue for those managing children with infantile spasms for whom vigabatrin may be the treatment of choice.


Monotherapy licenses for new antiepileptic drugs tend to lag behind licenses for add-on treatment. This is due to the general agreement that monotherapy studies should only be undertaken once there is proof of efficacy as add-on.23


In view of the fact that head to head monotherapy trials have failed to find convincing evidence for differences in efficacy of antiepileptic drugs, it is worth considering whether future trials should be designed to detect a difference or equivalence.24 In 1998, The Commission on Antiepileptic Drugs of the International League Against Epilepsy25concluded that comparative monotherapy trials should be designed to detect equivalence. In other words, trials should be designed to generate confidence intervals around efficacy estimates that are narrow enough to exclude the possibility of important differences existing. This of course needs an a priori definition of the smallest important clinical difference. The Commission also suggests that a new drug could be considered a first line agent if it is shown to have equivalent efficacy to but is better tolerated than a standard drug. A new drug showing equivalent efficacy and tolerability to a standard drug would be considered a second line treatment.

For regulatory purposes, European authorities are willing to accept evidence of equivalence as adequate for licensing an antiepileptic drug for monotherapy. The Food and Drug Administration (FDA) in the United States, however, will not accept evidence of equivalence in epilepsy trials, and demand that trials show a statistical difference, and the arguments for such a stance are outlined by Leber.26 This stance has resulted in a considerable divergence in the needs of the pharmaceutical industry and the needs of the clinician. To gain a licence, the industry is required to design and undertake trials that demonstrate that their product is better than something. The resulting trials that use placebo or pseudoplacebo comparisons do little to inform clinical practice, and raise a number of ethical concerns. One approach is to compare monotherapy with a high dose of a new drug with a low dose (sometimes referred to as a pseudoplacebo) in patients who continue to have seizures despite conventional antiepileptic drug therapy.27 This trial does not consider a useful clinical question, and there are ethical concerns about giving low dose monotherapy to patients whose seizures have not been controlled by standard antiepileptic drugs. A more worrying trial design uses what has been called the surgical paradigm.28-30 Patients undergoing seizure recording as part of a surgical work up, who have had their antiepileptic drugs withdrawn are randomised to receive either placebo or new antiepileptic drugs. Admittedly, these trials are undertaken in controlled environments, in which patients are monitored continuously, and the duration of the trials is short with patients protected by exit criteria. However, the questions answered by this type of trial are of no clinical relevance, and the potential risk to the patients from status epilepticus cannot be ignored.



Lamotrigine is licensed for monotherapy in the United Kingdom, and has been compared with carbamazepine in three randomised clinical trials.31-33 The largest of these recruited 260 patients with partial onset seizures or generalised tonic-clonic seizures, and was of 42 weeks duration. Lamotrigine faired significantly better than carbamazepine for the global outcome time to treatment withdrawal, and significantly more patients had carbamazepine withdrawn due to side effects. There was no difference, however, for the efficacy outcome “proportion of patients seizure free in the last 24 weeks of the study”. The study, however, lacked the power to exclude the possibility of important differences in efficacy existing, the duration of the trial was short given that epilepsy is a chronic condition, and the efficacy outcomes have marginal clinical meaning and lack statistical power.


Gabapentin and carbamazepine have been compared in one head to head monotherapy trial recruiting patients with partial seizures.34 This trial was designed primarily to meet the licensing needs of the pharmaceutical industry and was of 24 weeks duration, with 292 patients randomised in a 1:1:1:1 ratio to 600 mg/day carbamazepine, 300 mg/day gabapentin, 900 mg/day gabapentin, and 1800 mg/day gabapentin. The primary outcome was time to exit and reasons for exit included a single tonic-clonic seizure, three complex partial seizures, and status epilepticus. For this outcome, no statistical difference was found between carbamazepine and 900 or 1800 mg gabapentin/day. Adverse events were more common in the carbamazepine group. Although this trial has met the regulatory needs of the pharmaceutical industry by finding a difference in efficacy between differing doses of gabapentin, the protocol deviates significantly from everyday clinical practice, and the results do little to inform clinicians


Vigabatrin has been compared with carbamazepine in three monotherapy trials.21 35 36 The largest of these trials recruited 459 patients who were randomised to vigabatrin or carbamazepine, and was of 52 weeks duration. The trial was double blind, and treatment dose was increased to meet clinical need in an attempt to mirror clinical practice. The results showed no significant difference between drugs for the primary end point time to treatment failure, or for the outcome time to 6 month remission. However, confidence intervals around these estimates failed to meet the authors' generous definition of equivalence. Patients taking vigabatrin had significantly earlier first seizures postrandomisation, and were significantly more likely to have vigabatrin withdrawn due to lack of therapeutic effect, whereas carbamazepine was significantly more likely to be withdrawn because of side effects. This trial failed to find evidence to support the use of vigabatrin as monotherapy, and in view of the associated visual field abnormalities, it is unlikely that it will ever gain a monotherapy licence.


As already mentioned, topiramate has been tested in a monotherapy trial comparing 100 and 1000 mg/day.27 Although this trial provides evidence of efficacy as monotherapy, it does not inform us how topiramate compares with standard treatments.


Oxcarabazepine has recently been licensed for monotherapy in the United Kingdom. Initial monotherapy trials of oxcarbazepine compared it with carbamazepine37 and due to the results of these trials and its chemical similarity to carbamazepine, oxcarbazepine was licensed in the 1990s for monotherapy in some European and Scandinavian countries. Close scrutiny of these trials, however, shows that they do not provide firm evidence on which to recommend oxcarbazepine for use as monotherapy. Although no therapeutic difference was found between oxcarbazepine and carbamazepine, these trials lacked the power to exclude the possibility of important differences existing. Also, results of the largest trial were confounded by the exclusion of 70 of the 235 randomised patients from efficacy analyses.37

More recently oxcarbazepine has been compared with phenytoin in two38 39 and valproate in one40 monotherapy randomised controlled trial. All three trials were of similar design and of 56 weeks duration. Primary efficacy outcomes were the proportion of patients seizure free, and the primary tolerability outcome was time to withdrawal due to side effects. None of these three trials found a difference between oxcarbazepine and its comparators in the proportion of patients seizure free, and no difference was found when compared with valproate for time to treatment withdrawal due to side effects.40 When compared with phenytoin, oxcarbazepine fared significantly better for time to treatment withdrawal due to side effects in both adults38 and children.39Although no difference was found for the primary efficacy outcomes, these trials also lacked the power to exclude the possible existence of important difference, and do not prove equivalence.


In this article, we have drawn attention to evidence pertaining to the efficacy and tolerability of some of the new antiepileptic drugs. Most of this evidence comes from randomised clinical trials, but with the growing culture of evidence based medicine, it is important to realise that the randomised controlled trial is not the only vehicle that provides valuable information about treatments. Other sources may be more appropriate, particularly for long term safety, and picking up rare but important adverse events.

The pharmaceutical industry has been the major driving force in generating data from randomised clinical trials, and most of the those referenced in this editorial were sponsored by the industry. The prime aim of the industry, however, is to provide data that satisfy the requirements of the licensing bodies. These requirements and particularly those of the Federal Drug Administration (FDA) have resulted in the serious schism between trials that satisfy licensing authorities and those that inform clinical practice. We find this position difficult to reconcile, and urge the licensing authorities to ensure that their requirements better reflect the needs of the clinician and patient.

We clearly need reliable evidence on which to base our choice of antiepileptic drugs, particularly for monotherapy, given that up to 70% of patients are managed on monotherapy. Given the increasing number of published randomised clinical trials, there is a clear need for systematic reviews of existing trials, a task being addressed by the Cochrane Epilepsy Group.41 However, as we have highlighted in this article, much of the evidence regarding the new antiepileptic drugs has been acquired from trials that do not reflect day to day clinical practice, or use clinically meaningful outcomes. There is therefore a clear need for pragmatic trials comparing new antiepileptic drugs with each other and with standard drugs. In response to this need, NHS Health Technology Assessment Programme has funded the Standard And New Antiepileptic Drugs (SANAD)42trial which is a pragmatic trial comparing carbamazepine, gabapentin, lamotrigine, topiramate and valproate.


We thank Astra Pharmaceuticals, Novartis, and UCB Pharma for their kind provision of unpublished data. The department has undertaken contracted work through the Clinical Trials Unit for all companies with new antiepileptic drugs discussed in this article. AGM has received hospitality from Janssen Cilag (makers of topiramate), Sanofi-Synthelabo (makers of tiagabine), Parke-Davis (makers of gabapentin), and GlaxoWellcome (makers of lamotrigine), and speaker fees from Jannsen Cilag, sanofi-Synthelabo, and Hoechst-Marion-Roussel (makers of vigabatrin). DWC has received hospitality and speaker's fees for all companies with new antiepileptic drugs discussed in this article.


View Abstract

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.