Search tips
Search criteria 


Logo of jrsocmedLink to Publisher's site
J R Soc Med. 2004 January; 97(1): 15–19.
PMCID: PMC1079258

When is drug treatment not necessary in epilepsy? Factors that should influence the decision to prescribe

Epilepsy is a common treatable disorder with an incidence in the UK of about 0.05% and a prevalence of 0.5%.1 The average secondary school has about 6 children with the condition. A paediatrician who has diagnosed epilepsy will feel pressure to prescribe medication—as indeed will a clinician in adult practice. In view of the efficacy of anti-epileptic drugs,2 this is understandable. The primary aim of therapy is of course to prevent seizures, but do anti-epileptic drugs alter the natural history of the condition, either in the short term or by enhancing the chance of long-term remission? Should the benign epilepsies of childhood be treated differently? How do medication and everyday life issues interact? This paper deals with the factors that should influence the decision to prescribe. The focus is on childhood and adolescence, but much of the published work relates to adults and many of the points are relevant to an adult population too.


The first, and crucial, issue in management of epilepsy is diagnosis. Although in most young people who are diagnosed as having epilepsy there will be supportive evidence from electroencephalograms (usually recorded interictally) there is no single diagnostic test and the diagnosis has often to be made on clinical grounds. Classification of seizure type, the definition of aetiology and identification of the epilepsy syndrome are integral to this process. The reliance on clinical information offers ample scope both for under-diagnosis and for overdiagnosis, and even paediatric neurologists encounter difficulties with diagnosis. In the Dutch Study of Epilepsy in Childhood3 a panel of three neurologists identified diagnostic error in 4.6% of children with recurrent paroxysmal events. The diagnosis had to be deferred in 5.6% where there had been multiple events and in 7.4% following a single event. Overall, one in fifteen children with epilepsy were initially not recognized to have epilepsy.

Once sufficient diagnostic certainty is reached, the news should be given to the family in a planned and structured way. Linked to the giving of that news should be access to information, presented in a way that will be readily understood. Information for the child will need to be geared to his or her age and ability. Several studies indicate that at least a quarter of people prescribed anti-epileptic drugs stop the therapy, either because of unwanted effects or because it is perceived not to work; moreover, between 20% and 70% deviate from the regimen suggested by their doctors.4,5 What are the reasons for this poor adherence? One likely contender is lack of understanding of the nature of the condition. Houston et al.6 showed that children with epilepsy had significantly less understanding of their condition than did their counterparts with asthma and diabetes, even though they had access to specialist nurses and the families had received written information about the condition. Much of the information for children makes use of analogy to help them understand, but a child of 8 or so tends to interpret such material very literally. In my own department we think a simple biological model potentially more useful than analogy in conveying the necessary information, decreasing the child’s reluctance to disclose the diagnosis to others and perhaps making him or her more inclined to accept the treatment. A leaflet was written to meet the needs of 8–12-year-olds, with particular attention to idioms and language, and was piloted with 240 children between school years 3 and 8. Even so, we found that several of the central teaching points were being misunderstood. An essential part of medical intervention is to develop age-appropriate information of this sort with expert help. If children better understood the rationale of treatment, adherence might well improve.

The decision to treat or not and the consultation process

Each consultation, as with any ‘contract’, consists of a meeting of unlike minds. Parents and children have their own beliefs, hopes, fears and expectations, and so do doctors (together, in their case, with an evidence base). This issue was dealt with well by Haynes, Devereaux and Gyatt,7 who made the point that evidence does not make decisions; people do. If we hope to alter people’s behaviour (by persuading them to take medication), the best strategy is to influence their beliefs and accept their preferences. Haynes’ term for this is not evidence-based medicine but research-enhanced healthcare. Children should not be treated with drugs until the family has a good understanding of the condition and the way treatment might help. More effective ways of achieving this aim need to be defined. Drug treatment, and the seizure suppression it brings, can be recommended on two main grounds—that it may alter the natural history of the condition by increasing the remission rate; and that it may enhance quality of life by augmenting psychosocial opportunity and reducing risk of harm. The evidence around these premises will be reviewed in the rest of this paper.


An Italian hospital-based study8 of 397 adults randomized to treatment or no treatment found the 24-month recurrence rate of tonic–clonic seizures to be 26% in the treated group and 51% in the untreated group. In other words, without treatment about half have another episode within 2 years. Similar data have been reported in children. For example, of 186 children with an initial afebrile unprovoked seizure 52% had a recurrence, and of those with a recurrence 79% had additional seizures.9


The National General Practice Study of Epilepsy10 looked at outcome 9 years from the index seizure. 275 general practitioners participated in the study and 1091 people with probable epilepsy were identified. Further assessment indicated that only 564 of these had definite epilepsy. At 9 years from the index seizure, 86% of those with definite epilepsy had achieved a 3-year remission, and 68% a 5-year remission. The detail in the paper does not allow us to draw conclusions about the approach to treatment, but the figure of 68% is remarkably similar to that identified in other studies (see below).

If it is possible for people with epilepsy to achieve spontaneous remission without treatment, then prevalence rates in developing countries where epilepsy is rarely treated should be very similar to those in the developed world. The differences between countries would be accounted for either by change in remission rate or by a change in death rate. A study from Ecuador centred on a locality with a population of 75 000.11 7212 people were screened with a cascaded questionnaire and 881 with epilepsy were identified. This gave a prevalence rate of 0.67 (compared with the UK’s 0.5%); 15% of these were currently being treated. Only 29% had ever been treated. 46% of the untreated appeared to be in long-term remission.

Osuntokun et al.12 did a population study of 18 954 people in Nigeria which revealed a 0.53% prevalence of epilepsy with 4% treated. This was a town-based (Igbo-Ora) study, and it was acknowledged that people in Aiyate rural community only 20 km away had shown a much higher prevalence of 3.7%. The authors argued that the town-dwellers had benefited from primary-care interventions to reduce the incidence of common epileptogenic diseases such as central nervous system infection. In rural Ethiopia13 the prevalence was 0.52% with only 1.6% of these ever treated.

In a Finnish community-based project Keranen and Riekkinen14 identified 33 individuals whose epilepsy had not been treated. Of these, 42% were epilepsy-free at 10 years and 52% at 20 years, though acquisition bias cannot be discounted.

A different perspective is given on the figure from developing countries by the remarkable work of Louise Jilek-Aall.15 She had established a rural community hospital and worked with people with epilepsy from 1963 through to 1971. She then returned after 30 years and obtained data on 164 of 200 people in the original group. She found that 60% of the people had died and half of these had died from a cause related to epilepsy, including prolonged status epilepticus, burns or drowning. Clearly, therefore, in the developing world the reported prevalence rates in the population may be lowered by the excess mortality in this group.


Experience in the developing world

On the matter of treatment timing, Feksi et al.16 looked at a group of people in Kenya with chronic epilepsy. 249 of 302 completed the study. 53% became seizure-free in the second six months of therapy and 26% had a more than 50% reduction in seizure frequency. That is, almost 80% benefited. These people had had epilepsy for years before the onset of the study, which suggests that drug efficacy is not greatly influenced by treatment delay, though selection bias in this study cannot be excluded.

Experience in Europe

This finding is contradicted somewhat by the observations of Reynolds.17 In data from 241 adults and 167 children he showed that the number (rather than the frequency) of seizures before starting treatment has a bearing on long-term remission rates. People in the study were stratified according to the numbers of seizures experienced in the year before study entry. The cohort was then followed for 2 years. For those who had had two seizures before study entry, remission at the endpoint was in excess of 80%; for those with twenty seizures it was about 70%; whereas for those who had experienced fifty or more seizures before study entry the remission rate was only 50%. Elwes, Johnson and Reynolds18 looked at intervals between untreated tonic–clonic seizures in adults. This was a retrospective study of 183 adults with two to five seizures. At one month 56 had had a recurrence, at three months 93 and at 1 year 159. The median interval between successive seizures fell steadily from 12 weeks (range 10–18) between seizures one and two, to 8 weeks (range 4–12) between seizures two and three, to 4 weeks (range 2–2) between seizures three and four, and finally to 3 weeks (range 1–4) between seizures four and five. Their data, they suggested, were evidence for an accelerating disease process.

This conclusion was not supported by the work of Van Donselaar et al.19 In a hospital-based study the Dutch group assessed the course of untreated tonic–clonic seizures prospectively. 204 children aged one month to 16 years were recruited and followed until treatment needed to be started (78 children), or until they had had four seizures (41) or until 2 years had passed (85). They looked for either an accelerating pattern or a decelerating pattern. A decelerating pattern included children who became seizure free. 110 children were ‘unclassified’, usually because drug treatment was started early (outside the protocol) after the first, second or third seizure. 3 of 41 children with more than four seizures showed a decelerating pattern, 8 of 41 an accelerating pattern. Overall, 47% showed a decelerating pattern. Their conclusion was that the notion that epilepsy is a ‘progressive disease’ could not be used to justify treatment with anti-epileptic drugs.

Animal models

Advocates of early treatment have argued that over-stimulation of excitotoxic cascades may lead to the influx of calcium ions, free radical formation with enhanced proteolysis and triggering mechanisms and either immediate cell necrosis or later programmed cell death. The ‘kindling’ animal model is often cited at this point—a dynamic phenomenon whereby repeated initial subconvulsive stimuli lower seizure threshold without producing spontaneous seizures. Koh and Jensen,20 working with rats, showed how anti-epileptic drugs may modulate the potential for epileptogenesis when neurons are exposed to noxious stimuli. Topiramate reversed the damaging effect of hypoxia on later-life seizure-induced neuronal injury; carbamazepine and phenytoin prevented neuronal damage but did not prevent kindled seizures. This knowledge, however, seems not to apply in man. Anti-epileptic drug prophylaxis following head injury does not alter seizure outcome.21 Indeed, kindling has never been shown to happen in human beings. In many, as we have seen, epilepsy is a non-progressive condition.20

Mechanisms other than kindling may be important in some people. Seizures may result from repeated seizure-induced neuronal loss and consequent changes in neuronal circuitry: for example, repeated seizures arising from one temporal lobe may be associated with the emergence of seizures in the other. This ‘mirroring’ was demonstrated by Hughes.22 Following the initial appearance of a unilateral focus he showed that, in 1% of affected people per annum, a bilateral focus appeared.

The MESS study

Results from the Multi-centre study of Early Epilepsy in Single Seizures (MESS study), due to be published soon, offer new evidence that the timing of treatment has negligible effect on remission rates.23 1426 complete data sets are available involving 6976 patient years with 440 individuals aged between 0 and 19 years. Participants were randomized to start treatment either early or after a delay. Preliminary data indicate that at 2 years the respective remission rates were 55% and 39% and at 5 years 68% in both groups (almost identical to that in the National General Practice Study of Epilepsy).10 Thus, a delay in treatment did not alter ultimate remission rates. We can conclude that, if kindling or similar seizure-induced models of nerve-cell injury do operate in human beings, the effects are small. Clinical and experimental data do not support the view that anti-epileptic drugs alter the underlying course of the condition.


In several epilepsy syndromes of childhood the outcome is nearly always favourable. These include benign neonatal convulsions, benign occipital and Rolandic epilepsy, benign myoclonic epilepsy of infancy and specifically activated seizure syndromes including febrile seizures. In view of the good outcome in a large majority of children with these conditions, treatment is not justifiable on the argument it will enhance the likelihood of natural remission. The decision should centre on whether treatment will improve the person’s quality of life while the epilepsy is active. For example, if in benign epilepsy of childhood with centro-temporal spikes (benign Rolandic epilepsy) seizures are infrequent, short and confined to sleep at night, it is perfectly reasonable not to use anti-epileptic drugs. If, however, a child with the same condition is having diurnal seizures with distressing sensations of the tongue and unilateral tonic–clonic movement, which in turn are attracting adverse comment from the child’s peer group, then treatment should be considered. Seizure frequency alone is not an absolute determinant of whether treatment should be used or not. Some families will calmly accommodate infrequent seizures, while the very same frequency can leave other families in constant fearful anticipation of the next recurrence. Consideration of the psychosocial issues in each family may well lead to different decisions about similar patients.

The commonest of these conditions, febrile seizures, has been studied in detail. In a meta-analysis of intervention studies with valproate and phenobarbitone, Newton24 showed that, when the data were analysed by intention to treat, the administration of anti-epileptic drugs did not alter seizure occurrence rates in this condition.


The reported adverse outcomes in epilepsy include excess mortality, sudden unexpected death and intellectual deterioration or poor academic performance largely reflecting the degree of underlying brain malfunction and the biological substrate. It is not yet proven whether medication, with optimum treatment and assessment, can reduce mortality rates. When deciding whether to treat or not the paediatrician needs to focus on the issues discussed earlier in this paper. The 5-year remission rate with use of anti-epileptic drugs seems to be about 68%, whether treatment is started early or delayed. Early treatment undoubtedly gives earlier remission. If treatment is started after a single seizure the recurrence rate that at most is 50% without treatment, means two people are treated to save one person a recurrence. After two seizures most people have a recurrence and the decision to treat is easier. At this early stage, consideration must be given to the family’s hopes, fears and attitudes to seizure recurrence on the one hand, and the unwanted medication effects on the other. We must bear in mind that even one or two seizures in a young person’s life can have serious adverse psychosocial consequences. This may lead to teasing from an unkind peer group, altered expectation on the part of teachers and a resulting lack of self-esteem. All these issues must be weighed carefully before the contract between unlike minds is settled. Whether development of clinical networks25 will improve standards in the care of people with epilepsy remains to be seen. In about 20% of people on anti-epileptic drugs, unwanted drug effects are troublesome, though these can be mitigated by slow progression to the target doses. About 70% of those treated with drugs will become seizure-free, with obvious advantage for other quality of life measures such as employment prospects, social adjustment and retaining a driving licence. In the MESS study23 preliminary data point to a non-significant increase in self-esteem in both treatment groups (early and delayed), unrelated to seizure or anti-epileptic drug status. Individuals with no further seizures had a higher sense of mastery (control of their own destiny) than those with recurrences.


Existing data indicate that long-term remission rates are not altered by the use of anti-epileptic drugs. A decision to treat should therefore be based on how the early achievement of remission of seizures might improve quality of life and avoid harm. After two seizures most people will have a further seizure and this is probably the pivotal point at which the decision to treat or to delay treatment ought to be made.


1. Hauser WA, Annegers JF, Anderson VE. Epidemiology and the genetics of epilepsy. In: Ward AA, Penry JK, Purpura D, eds. Epilepsy. New York: Raven Press, 1983: 267–94
2. Marson AG, Kadir ZA, Chadwick DW. New anti-epileptic drugs. A systematic review of their efficacy and tolerability. BMJ 1996;313: 1169–74 [PMC free article] [PubMed]
3. Stroink H, van Donselaar CA, Geerts AT, Peters ACB, Brouwer OF, Arts WFM. The accuracy of the diagnosis of paroxysmal events in children. Neurology 2003;60: 979–82 [PubMed]
4. Conrad P. The meaning of medications: another look at compliance. Soc Sci Med 1985;20: 29–37 [PubMed]
5. Trostle JA. Medical compliance as an ideology. Soc Sci Med 1988;27: 1299–308 [PubMed]
6. Houston EC, Cunningham CC, Metcalfe E, Newton R. The information needs and understanding of 5 to 10 year old children with epilepsy, asthma or diabetes. Seizure 2000;9: 340–3 [PubMed]
7. Haynes RB, Devereaux PJ, Gyatt GA. Physicians’ and patients’ choices in evidence based practice. BMJ 2002;324: 1350. [PMC free article] [PubMed]
8. First Seizure Trial Group. Randomized clinical trial on the efficacy of anti-epileptic drugs in reducing the risk of relapse after a first unprovoked tonic–clonic seizure. Neurology 1993;43: 478–83 [PubMed]
9. Camfield PR, Camfield CS, Dooley JM, Tibbles JA, Fung T, Garner B. Epilepsy after a first unprovoked seizure in childhood. Neurology 1985;35: 1657–60 [PubMed]
10. Cockerell OC, Johnson AL, Goodridge DMG, Sander JWAS, Shorvon SD. The remission of epilepsy: results from the National General Practice Study of Epilepsy. Lancet 1995;346: 140–4 [PubMed]
11. Placencia M, Sander JWAS, Shorvon SD, Alarcon F, Roman M, Cacante S. Anti-epileptic drug treatment in a community health care setting in Northern Ecuador: a prospective 12 month assessment. Epilepsy Res 1993;14: 237–44 [PubMed]
12. Osuntokun BO, Adeuja AOG, Nottidge VA, et al. Prevalence of epilepsy in Africans: a community based study. Epilepsia 1987;28: 272–9 [PubMed]
13. Tekle-Himanot R, Forsgren L, Abebe M, et al. Clinical and electroencephalographic characteristics of epilepsy in rural Ethiopia: a community based study. Epilepsy Res 1990;7: 230–9 [PubMed]
14. Keranen T, Riekkinen PJ. Remission of seizures in untreated epilepsy. BMJ 1993;307: 483. [PMC free article] [PubMed]
15. Jilek-Aall L, Jilek W, Miller JR. Clinical and genetic aspects of seizure disorders prevalent in an isolated African population. Epilepsia 1979;20: 613–22 [PubMed]
16. Feksi AT, Kaamugish J, Sanders JWAS, Jatiti S, Shorvon SD. Comprehensive primary health care anti-epileptic drug treatment programme in rural and semi-rural Kenya. Lancet 1991;337: 406–9 [PubMed]
17. Reynolds EH, Hellar AJ, Elwes RDC, et al. Factors influencing prognosis of newly diagnosed epilepsy. Epilepsia 1989;30: 648
18. Elwes RDC, Johnson AL, Reynolds EH. The course of untreated epilepsy. BMJ 1988;297: 948–50 [PMC free article] [PubMed]
19. Van Donselaar CA, Brouwer OF, Geerts AT, Arts WF, Stroink AH, Peters AC. Clinical course of untreated tonic–clonic seizures in childhood: prospective, hospital based study. BMJ 1997;314: 410–14 [PMC free article] [PubMed]
20. Koh S, Jensen PE. Topiramate blocks acute and chronic epileptogenesis in a rat model of perinatal hypoxic encephalopathy. Epilepsia 1999;40(suppl 7): 5–6 [PubMed]
21. Temkin NR. Personal communication cited in Sander JWAS, White HS. Disease modification in epilepsy. Prog Neurol Psychiatry 2000;5(2)
22. Hughes JR. Long-term clinical and EEG changes in patients with epilepsy: Arch Neurol 1985;43: 213–23 [PubMed]
23. Chadwick D. The MESS Study. Lecture delivered to 29th Annual Conference of the British Paediatric Neurology Association, Liverpool, 2003
24. Newton RW. Randomised controlled trials of phenobarbitone and valproate in febrile convulsions. Arch Dis Child 1988;63: 1189–91 [PMC free article] [PubMed]
25. Newton R. Neurological networks. Devel Med Child Neurol 2002;44: 795. [PubMed]

Articles from Journal of the Royal Society of Medicine are provided here courtesy of Royal Society of Medicine Press