Article Text

Download PDFPDF

Research paper
Treatment changes in a cohort of people with apparently drug-resistant epilepsy: an extended follow-up
  1. Aidan Neligan1,
  2. Gail S Bell1,
  3. Muaz Elsayed1,
  4. Josemir W Sander1,2,
  5. Simon D Shorvon1
  1. 1Department of Clinical and Experimental Epilepsy, UCL Institute of Neurology, Queen Square, London, UK
  2. 2SEIN-Epilepsy Institute in the Netherlands Foundation, Heemstede, The Netherlands
  1. Correspondence to Professor Simon D Shorvon, Department of Clinical & Experimental Epilepsy, UCL Institute of Neurology, Box 5, Queen Square, London WC1N 3BG, UK; s.shorvon{at}


Background The seizure response to the addition of a previously unused antiepileptic drug in a cohort of 155 people with refractory epilepsy was previously reported after a median of 18 months follow-up.

Methods The authors followed 139 (90%) of the original cohort for a median follow-up of 6.9 years to determine the longer term outcome in people with refractory epilepsy.

Results During the 6.9 year follow-up period, a total of 448 medication changes were made. Eight per cent of these resulted in 12 months or more of seizure freedom and a further 17% of changes resulted in at least 50% improvement in seizure frequency. At the last follow-up, 26 (19%) of individuals had been seizure-free for 12 months or more, and 41 (29%) had 50%–99% improvement in seizure frequency. Terminal seizure freedom was correlated with having no seizures at the time of the previous report (p=0.03), a lower number of previous antiepileptic drugs taken (p=0.052) and a lower number of concomitant antiepileptic drugs (p=0.03). In those who entered remission the probability of remaining seizure-free 5 years later was 0.48 (95% CI 0.32 to 0.63).

Discussion This suggests that about half of people with apparent drug-resistant epilepsy can have significant improvements in seizure control with further drug changes. Some will subsequently relapse, but long periods of seizure freedom or significantly improved seizure control in the absence of complete seizure control can occur. Such valuable improvements suggest that the recently proposed International League against Epilepsy definition of refractory epilepsy may be too restrictive.

  • Epilepsy
  • refractory
  • relapse
  • treatment
  • prognosis
  • epidemiology
  • anticonvulsants
  • channels
  • neurophysiology
  • neuromuscular
  • motor neuron disease
  • neuropathy

Statistics from


Pharmacoresistance is a serious problem which is said to affect approximately a third of people with epilepsy.1 Despite this, there has until recently been very little data on the course and seizure prognosis in those in whom a pattern of medically intractable epilepsy has already been established.

The International League against Epilepsy (ILAE), the main epilepsy professional body, recently proposed a standardised definition of drug resistant or chronic epilepsy.2 In this, drug-resistant epilepsy is pragmatically defined as continuous seizures despite effective trials of at least two appropriately selected antiepileptic drugs. This has been criticised as being restrictive.3 Three recent studies have suggested that up to 16% (4%–5% per year) of people with apparently drug-resistant epilepsy will later have a significant period of seizure freedom.4–6 Such findings, while encouraging, are limited by the relatively short periods of follow-up (from 18 months to <4 years).

We report an extended follow-up of one of these cohorts4 giving a more complete appraisal of seizure prognosis. Our aims were:

  1. To determine the response to the addition of a previously untried antiepileptic drug in a cohort of people with chronic refractory epilepsy.

  2. To determine the seizure outcome of these individuals at the end of the extended follow-up period (median 6.9 years), and the factors correlated with this outcome.

  3. To determine the proportion of people in whom seizure remission is followed by subsequent relapse, and to estimate the time to relapse.


The methodology was described previously.4 All adults with refractory epilepsy in a single epilepsy clinic who had been commenced on a new antiepileptic drug (265 drug changes in 155 people) during a 24-month period (1 January 2002 to 31 December 2003) were identified. Inclusion criteria were: (1) active epilepsy (defined as at least one seizure per month in the previous 3 months); (2) a history of epilepsy for at least 5 years and a history of treatment with at least two appropriate antiepileptic drugs; (3) age 16 years or older at the time of inclusion; (4) a minimum follow-up in the clinic of 12 months after the drug change; and (5) a response to treatment that was easily quantifiable by means of an accurate seizure diary. Individuals were excluded if there was a suspicion of non-epileptic seizures or a history of poor drug adherence.

All were treated according to a standard treatment protocol,4 which involved assessing the seizure response to the addition of a previously untried antiepileptic drug. If an adequate response was not obtained, a further drug addition was made. The choice of drug (and dose) was made by an experienced neurologist (SDS) based on the seizure classification, the clinical context, prior history of drug response and personal choice. If the first-choice drug trial succeeded, the individual was maintained on the drug at the appropriate dose. The duration of each trial varied depending on the baseline seizure frequency and was generally at least as long as the period over which three to five seizures would have occurred at the baseline frequency. If the first-choice drug trial failed, the drug was usually withdrawn and the individual was then switched to a second (and so on). In some individuals, concomitant drugs were withdrawn or doses changed.

Follow-up in this cohort was previously reported up to the end of June 2005.4 We have now extended the follow-up to the end of December 2009 (or time of death). Those who subsequently underwent epilepsy surgery (excluding vagal nerve stimulator (VNS) implantation) were assessed up to the time of surgery.

The recorded monthly seizure frequency in the 3-month period prior to starting the first new antiepileptic drug in 2002–2003 was used as baseline seizure frequency. Seizure outcome was categorised into three groups: (1) seizure-free (for 12 months or more at last follow-up); (2) 50%–99% reduction in seizure frequency; and (3) <50% reduction in seizure frequency in the last 12 months of follow-up, compared with baseline. The period of observation was a minimum of 12 months from the time of the last drug change or shorter if the drug was discontinued because of lack of efficacy or side-effects.

Correlation was made with duration of epilepsy, number of previous antiepileptic drugs tried, epilepsy syndrome, current age, age of onset of epilepsy, seizure type, sex or number of concomitant antiepileptic drugs. Analysis was carried out using STATA (version 10, Statacorp LP, Texas, USA). Categorical data were compared using the χ2 test or, where appropriate, the Fisher's exact test.

Seizure outcomes at the end of the initial and later follow-up periods were compared using the Stuart–Maxwell test. Kaplan–Meier analysis was used to estimate the probability of seizure recurrence in those who became seizure-free for at least 1 year and Cox proportional hazard regression was used to compare time to recurrence in these people. HRs were calculated with 95% CIs. Time to event was defined as the duration from the date of starting seizure freedom to the date of seizure relapse or to the date of the last clinic visit if seizure relapse had not occurred.


Of the 155 people in the original cohort, further follow-up was available on 139 (90%). The mean duration of follow-up was 6.7 (SD 1.1, median 6.9) years. During this period, 448 drug changes (table 1) were made in 139 individuals. Over half of these additions (228, 50.9%) were accompanied by the withdrawal of another drug (drug substitutions). The outcome of the 448 medication changes are shown in table 2.

Table 1

Antiepileptic drugs additions in 139 people

Table 2

Response to individual antiepileptic drug changes

Final outcome at the time of last follow-up

At the time of last follow-up, 26 (19%) were seizure-free and 41 (29%) had a 50%–99% reduction in seizure frequency. Thus, 67 (48%) had a >50% reduction in seizure frequency. Two people died, five underwent epilepsy surgery and 12 had VNS implanted during the follow-up period. Those who later died, had surgery or had VNS implantation are included until censored at time of death or surgery.

The 12 people who had a VNS implanted had a mean duration of epilepsy of 30.7 (SD 13.0, median 30.5) years, a median of 13.5 antiepileptic drugs tried (range 10–16) and a median of two current concomitant antiepileptic drugs (range 1–4). VNS resulted in a 50%–99% reduction in seizure frequency in one person while another reported some symptomatic improvement (seizures less severe and shorter in duration) without a significant reduction in seizure frequency.

Of the two people who died, one had a 50%–99% reduction in seizure frequency up to death. The other person continued to have seizures with no improvement in frequency, up to the time of death. Four of the five people who had surgery had less than a 50% reduction in seizure frequency at the last follow-up prior to the surgery.

Factors correlated with final outcome

Seizure outcome at the end of the original follow-up was predictive of outcome at the end of the current follow-up (Stuart–Maxwell p=0.03). In particular, 64% of those who were seizure-free at the end of the original follow-up were seizure-free at last follow-up. Similarly, 69% of those with less than 50% improvement remained so at last follow-up. Outcomes are summarised in table 3. The number of concomitant antiepileptic drugs at the time of the last drug change (0, 1 or 2 or more concomitant antiepileptic drugs) was associated with seizure prognosis (χ2=10.5, df=4, p=0.03), with those taking two or more concomitant antiepileptic drugs less likely to attain seizure freedom. Those with a lower number of previous antiepileptic drugs tried at entry (<five antiepileptic drugs) were slightly more likely to attain seizure freedom (χ2=5.9, df=2, p=0.052). None of the other clinical factors analysed correlated with the final outcome.

Table 3

Seizure outcomes in 2004 and 2009 in 139 people with epilepsy

Relapse after a period of remission

In addition to the 35 people seizure-free at the start or the end of the current follow-up, there were an additional 13 (9%) who had a significant period of seizure freedom of 6 months (range 6–10 months, N=3) or one or more years (range 1–3.5 years, N=10) during the follow-up period but were not seizure-free at either end point. Consequently, 45 (32%) had at least 1 year of seizure freedom during follow-up. Three with IGE (two of whom also had at least 1 year of complete seizure freedom) became free of generalised tonic-clonic seizures as a result of a medication change but continued to have myoclonus. Similarly, two people with focal epilepsy, both of whom had significant periods of seizure freedom (one with a seizure-free period of 6 months and the other with two separate periods of 1 year and 18 months of seizure freedom), but were not seizure-free at either end point, had no further convulsive seizures (but continued to have complex partial seizures). In total, of the 26 people seizure-free at the last follow-up, 20 had remained seizure-free since remission onset with a mean period of seizure freedom of 4.9 (SD 2.1) years. Five of the 26 people seizure-free had six previous drug failures and a further two had tried seven or more previous antiepileptic drugs.

Time to relapse in those who had a period of seizure freedom is shown in figure 1.

Figure 1

Kaplan–Meier analysis of the probability of remaining seizure-free.

Of the 45 people who achieved at least 1 year seizure freedom during the course of follow-up, 25 (56%) subsequently relapsed.

Using survival analysis, the probability of remaining seizure-free at 5 years after entering seizure freedom was 0.48 (95% CI 0.32 to 0.63). Neither epilepsy syndrome (p=0.57) nor the number of previous antiepileptic drugs taken (p=0.71) was predictive of subsequent relapse. In contrast, those with a duration of epilepsy of more than 10 years were less likely to relapse than those with shorter durations (HR 0.35, 95% CI 0.15 to 0.82).


Of the 139 people in whom prolonged follow-up was available, 67 (48%) had a greater than 50% reduction in seizure frequency compared with 51 (37%) at the previously published follow-up. This supports the notion of persevering with treatment as eventually almost half had a sustained period of significant improvement in seizure control. We found, however, that a significant proportion of people who did achieve initial remission subsequently relapsed on extended follow-up. This proportion is similar to that found by others.6–8 In the extended follow-up period in our cohort there were some people who became seizure-free who had not been seizure-free at the time of the first follow-up.4 Indeed, six of the 25 who relapsed after becoming seizure-free during the observation period were in remission again at the end of follow-up. It is also possible that some of the people lost to follow-up are no longer attending a tertiary referral centre because their seizures remain under complete control or are significantly improved; so, we may have underestimated or overestimated the proportion of people in whom there is a good outcome.

In the original analysis,4 three factors (epilepsy duration, number of previous antiepileptic drugs and epilepsy syndrome) were found to correlate with a final outcome of seizure freedom. In this analysis, seizure outcome at the end of follow-up (in 2009) was correlated with seizure outcome at the end of the earlier follow-up (in 2004). A correlation with a smaller previous number of antiepileptic drugs (<five antiepileptic drugs) tried also approached statistical significance, and a lower number of concomitant antiepileptic drugs at the time of drug change were statistically significant in predicting a better long-term seizure prognosis. This underlies the importance of the number of previous drug failures in predicting seizure prognosis, a finding replicated in some studies5 ,9 but not in others.7 ,10 In one of these studies, of 478 people seen sequentially in an epilepsy clinic, 62% of people with newly treated epilepsy responded to the first antiepileptic drug compared with 42% with one previous antiepileptic drug tried and 17% with two to five previous antiepileptic drugs given. Significantly, no one became seizure-free after failure of six or more antiepileptic drug trials,9 leading to the suggestion that absolute drug-resistance occurs after failure of six antiepileptic drugs. We found that 27% (seven of 26) of people seizure-free at last follow-up had six or more previous drug failures.

About a quarter (six of 26) of people seizure-free at last follow-up had relapsed after attaining initial seizure remission before becoming seizure-free again. Similarly, we previously reported 32 individuals with chronic epilepsy who had entered late terminal remission. In that group the median number of antiepileptic drugs was 6 (IQR 4.5–8) suggesting that there is always a chance of becoming seizure-free even if a large number of antiepileptic drugs had been tried previously.11

Here we showed that neither epilepsy duration nor the epilepsy syndrome correlated with long-term seizure outcome. Having a shorter duration of epilepsy was the only factor identified which predicted subsequent relapse once a person had become seizure-free; this finding is counterintuitive, and we do not have any explanation apart from, possibly, an effect due to the small numbers.

The number of previous antiepileptic drug failures is central to the recently proposed ILAE definition of drug-resistant epilepsy.3 It was recommended that epilepsy be classified as treatment resistant when an individual continues to have seizures despite trials of at least two appropriately selected antiepileptic drugs, assessed for a sufficient length of time to allow an accurate evaluation of seizure frequency. Treatment success was defined as subsequent seizure freedom of 1 year or a period of seizure freedom three times the duration of the previous inter-seizure interval, whichever is longer.3

One rationale for identifying early predictors of potentially refractory epilepsy is so that the evaluation of alternative means of treatment (eg, surgery) can be initiated earlier in the course of epilepsy. Only a small proportion of people with treatment-resistant epilepsy, however, are suitable candidates for epilepsy surgery.12 The option of epilepsy surgery was discussed with people in whom epilepsy surgery was felt to be of potential benefit. Many were worked up for epilepsy surgery but most were found to be unsuitable or were not willing to proceed. In the end, <4% of the cohort underwent epilepsy surgery. Consequently, further medication changes offer the only real prospect of seizure freedom for the majority of people with treatment-resistant epilepsy.

In considering multiple antiepileptic drug changes, it is important to recognise that seizure aggravation can occur as a result of inappropriate antiepileptic drug choice in people with idiopathic generalised epilepsy,13 ,14 underlying the importance of accurate syndromic classification. Seizure control typically improves following cessation of the inappropriate antiepileptic drug.

Seizure aggravation as a result of antiepileptic drug changes may also occur in focal epilepsy.15 It is, however, likely that often an apparent deterioration in seizure control is due to coincidental change in seizure pattern when starting an ineffective drug,16 rather than a true antiepileptic drug-induced deterioration of seizure control. Seizure aggravation can deter people from considering further antiepileptic drug trials, making the prospect of improving seizure control more remote.

The heterogeneity of chronic epilepsy is demonstrated here. More importantly, the suggestion in this cohort is that persistence with medication changes in the face of apparent intractability may result in nearly half of people having a significant reduction in seizure frequency, while approximately a third will become seizure-free for one or more years. A commonly expressed extrapolation from the ILAE definition of drug-resistant epilepsy is that the prognosis of epilepsy will be poor if two or more antiepileptic drugs have failed to control the seizures. This, however, does not universally apply. Prognosis is a complex phenomenon with various influencing factors and to label such people as therapy resistant is sometimes inaccurate and potentially harmful (carrying the risk of reticence to initiate further drug changes). We have shown that, taking an active approach to treatment and with further drug changes, many attain terminal remission even after several previous failed therapies, and others have great improvements in seizure control.


We wish to thank Dr Anna Luciano for compiling the original database.



  • Funding This work was carried out at the University College.

  • Competing interests AN has received travel grants and honoraria from Eisai, UCB and Janssen. GSB's husband works for, and has shares in, GSK. ME reports no disclosures. JWS has received research grants, honoraria or consultancy fees from various companies including UCB, Eisai, Janssen, Viropharma, Medtronic and GSK. SDS has received consultancy fees from Janssen, UCB and Eisai, and has received speaker's honoraria from GSK, Janssen and UCB. 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.

  • Ethics approval Approval provided by the Joint Research Ethics Committee of the National Hospital for Neurology and Neurosurgery and the UCL Institute of Neurology.

  • Provenance and peer review Not commissioned; externally peer reviewed.

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.