The prevalence of epilepsy and pharmacoresistant epilepsy in adults: A population-based study in a Western European country


Address correspondence to Dr. Marie-Christine Picot, C.H.U. Montpellier - Hopital Arnaud de Villeneuve, Unité recherche clinique et épidémiologie, Département de l'Information Médicale, 291, Avenue du Doyen G. Giraud, 34 295 Montpellier Cedex 5, France. E-mail:


Purpose: To determine the prevalence of epilepsy in a defined adult population and identify the frequency and principal features of pharmacoresistant epilepsy.

Methods: From a population over 15 years of age residing in a medium-sized French city, all patients with epilepsy on June 30, 1995 were identified from multiple sources. Pharmacoresistance was defined as failure to control epilepsy by at least two first-line antiepileptic drugs, with a seizure frequency of at least one per month for 18 months. Collected data were examined by experts in epileptology, and responding patients were reexamined using a standardized diagnostic questionnaire. ILAE definitions and classifications were used.

Results: The age-adjusted prevalence of active epilepsy was 5.4 per 1,000 (95% CI: 4.7–6.0) and was higher for males (7.8) than for females (5.2). For epilepsy in remission under treatment, this rate was 0.7 per 1,000 (95% CI: 0.5–0.95). Age-specific prevalence was highest in age groups 25–49 years and declined in the oldest age groups. Localization-related seizures represented 61.1% of cases and generalized seizures 30.9%. The proportion of noncontrolled epilepsy (seizure-frequency at least one per month for 18 months) was 15.6%, corresponding to a prevalence of 0.94 per 1,000. In this group, the mean age at onset was lower (p = 0.0007) and localization-related epilepsy more frequent (p = 0.01).

Conclusion: The findings support previous epidemiological estimates of the prevalence of epilepsy in developed countries. For approximately one patient in eight, epilepsy was not adequately controlled.

Prevalence studies provide pertinent epidemiological data for a rational approach to planning health-care provision, as well as providing a basis for hypothesis generation in clinical research or analytical epidemiological studies. Many studies have been performed in the past to determine the prevalence of epilepsy and these have yielded prevalence rates of active epilepsy varing from 4 to 10 in 1,000 (Crombie et al., 1960; Gudmundsson, 1966; Zielinski, 1974; Granieri et al., 1983; Haerer et al., 1986; Joensen, 1986; Keränen et al., 1989; Hauser et al., 1991; Maremmani et al., 1991; Forsgren, 1992; Giuliani et al., 1992; Cockerell et al., 1995; Nakashima et al., 1996; Reggio et al., 1996; Olafsson & Hauser, 1999; Oun et al., 2003). This variation can largely be explained by differences in case evaluation methods as well as in the definition of epileptic seizures and syndromes, since many of these studies date from several decades. In addition, improvements over time in diagnosis, prevention, and management of epilepsy may influence prevalence rates (Berg et al., 1996).

The publication of guidelines for epidemiological studies by the International League Against Epilepsy (ILAE) in 1993 (Commission on Epidemiology and Prognosis and International League Against Epilepsy, 1993) provided standardized definitions of epilepsy. However, there have been very few population-based prevalence studies performed since these guidelines were introduced. In Western Europe, no such population-based studies have been performed since then. The last major study was an incidence study of nonprovoked first epileptic seizures in South-West France (Loiseau et al., 1990).

The proportion of patients with pharmacoresistant epilepsy in the general population is currently unknown. Some studies have examined the proportion of patients with a high frequency of seizures but without specifying explicit criteria for pharmacoresistance (Haerer et al., 1986; Juul-Jensen, 1986; Keränen & Riekkinen, 1988; Hauser & Hesdorffer, 1990; Forsgren, 1992, 1995; Hart & Shorvon, 1995). Patients with pharmacoresistant epilepsy are the target for new antiepileptic drugs (AEDs) and, in case of failure, epilepsy surgery may be considered in certain circumstances. Therefore, it is important to determine the prevalence of pharmacoresistant epilepsy in order to plan resource allocations for epilepsy surgery centres.

The objectives of this study were, first, to assess the prevalence of epilepsy in a Western European country using a population-based sampling method and the ILAE guidelines and, second, to determine the frequency and features of pharmacoresistant epilepsy.


Study population

The sample-size required to estimate a prevalence of 1% with an accuracy of 0.1% and an alpha risk of 5% was 38,000. Therefore, the survey was carried out in Béziers, a medium-sized city in the South of France, with 59,407 people over 15 years of age (INSEE data, 1990). This town was also chosen for its low migration rate (demographic turnover 1990–1999: –0.29%), the presence of a hospital with a neurological department and a university hospital with an epileptology unit 80 kilometres away. Most of the patients were therefore evaluated and treated locally, making registration easier. The prevalence day was determined as the June 30, 1995.

Case definition and classification

Index cases were individuals alive on the prevalence day, residing in the study area and diagnosed with active epilepsy or epilepsy in remission under treatment. Only patients aged 16 or over were considered. This age cutoff corresponds to the age when adolescents leave the pediatric care sector in France, and has been used in previous epidemiological studies on epilepsy in adults (Keränen et al., 1989; Forsgren, 1992). Standard ILAE definitions of epilepsy were used (Commission on Epidemiology and Prognosis and International League Against Epilepsy, 1993). Epilepsy was defined as the occurrence of at least two unprovoked seizures a minimum of 24 hours apart. Active epilepsy was defined as the occurrence of at least one unprovoked seizure during the previous 5 years and, epilepsy in remission under treatment as freedom from seizures for the previous 5 years in patients still being treated with AEDs on the prevalence day.

Otherwise, subjects with an isolated seizure associated with characterized epileptiform EEG abnormalities or with an epileptogenic lesion on neuroimaging clearly related to the seizure and treated with an AED because of a high risk of seizure recurrence were also ascertained. Patients who received treatment after only one seizure were systematically excluded if there was any suspicion of a provoked seizure. Prevalence was calculated separately for active epilepsy, patients with treated isolated seizures, and epilepsy in remission under treatment. For the descriptive analysis, these groups were combined.

Since there is currently no agreed definition of pharmacoresistance, this was operationally defined as failure of at least two first-line AEDs to control seizures. We considered two levels of lack of control: Definition 1 - the occurrence of seizures at an average frequency of at least one per month for 18 months (Berg et al., 2001) and a less restrictive threshold, Definition 2, corresponding to at least one seizure per year (Loiseau & Jallon, 1995). A drug was considered to have failed if it did not control seizures or if it was discontinued due to unacceptable side effects. A drug was not considered to have failed if the dose used, for reasons of tolerance, was subtherapeutic or if there was documentation of poor compliance. It should be noted that only three of the twelve newer AEDs were available at the time of the survey.

Case ascertainment

Multiple sources were screened to identify all suspected and established cases, including local general practitioners, neurologists, local and university hospitals, emergency services, community clinics, institutions, and nursing homes.

All four neurologists, all the psychiatrists (N = 22), all the pediatricians (N = 10) as well as 20% of the general practitioners (N = 93) from this area agreed to refer patients with epilepsy to the study. A preliminary study showed that a small proportion of patients not monitored in Béziers were monitored almost exclusively by the university hospital in Montpellier. In the department of neurology at the hospital in Béziers, all outpatient records and hospitalization reports over the previous 10 years were reviewed. Interrogation of the Hospital Information System (HIS) using ICD codes for epilepsy or seizure enabled us to identify patients with epilepsy hospitalized in other, short-stay departments over the previous 3 years. Institutionalized patients in psychiatric or long-stay departments were identified directly from the department records. The records from the emergency department, available for the previous 3 years, were also reviewed in order to identify those patients brought in for epileptic seizures but not hospitalized. EEG records from the two laboratories in the area were reviewed for all persons who had been referred for an EEG because of a seizure or suspicion of seizures. Nonhospital records were searched manually, and all information was cross-checked between sources.

At the university hospital in Montpellier, the outpatient file from the epileptology unit, the EEG-laboratory and the HIS enabled us to identify those patients with epilepsy living in Béziers.

In community practice, physicians (neurologists, psychiatrists, pediatricians, and general practitioners) were asked to complete a case registration form containing their patients' identity, date of birth, date of the index and last seizure, and the whereabouts of the specialist consultation for epilepsy. Patients in nursing homes and psychiatric institutions who were residents of the district of Béziers were also identified.

Finally, a media campaign was initiated to inform and recruit patients directly through advertising posters, the press, local radio, and television. After completing a census of all probable epilepsy cases, the status of each patient at the date of prevalence (dead or alive, still living in Béziers or not, seizure frequency and AED prescription over the previous 5 years) was updated. The death registry at Béziers town hall was also consulted. For those patients who had been lost track of, the civic records of the place of birth were also consulted. The electoral roll for the previous 5 years was examined to identify any patients who had been struck off the list.

Procedure for diagnosis and classification

Diagnoses were initially established by the patient's neurologist or epileptologist. For the purposes of the study, two epileptologists validated the diagnosis on the basis of all the medical information collected, and classified independently epileptic seizures, syndromes, and etiologies according to ILAE criteria (Commission on Classification and Terminology of the International League Against Epilepsy, 1981; Commission on Classification and Terminology of the International League Against Epilepsy, 1989; Commission on Epidemiology and Prognosis and International League Against Epilepsy, 1993). In patients with multiple seizure types, the principal seizure type was retained for seizure classification.

All patients identified as still being alive were contacted by their neurologist to offer them reexamination by the epileptologist of the study using a French adaptation (Picot et al., 1999) of a standardized diagnostic questionnaire, the Semi-structured Interview for Seizure Classification (SISC) (Ottman et al., 1990). After two attempts to contact the patients, the participation rate of eligible patients was 35% (135 out of 388 contacted patients agreed to participate). These did not differ from the total population in terms of age, sex, duration of epilepsy, and etiology. However, the reinterviewed patients tended to have a lower age of onset, (23 years old vs. 28, p = 0.04), were more likely to be severe (28.8% vs. 19.5%, p = 0.05) and less likely to have an unclassifiable seizure type (1.9% vs. 11.0%, p = 0.03).

Other parameters assessed included AED use, surrogate markers of epilepsy severity (Duncan & Sander, 1991), disabilities, and diagnostic evaluations (EEG, computerized tomography [CT], and magnetic resonance imaging [MRI]). At least one EEG was available for 99.1% of patients, a CT scan was performed on 217 patients (63.1%) and an MRI on 55 patients (16%). The associated disabilities considered were motor, intellectual, difficulties related to a disabling chronic pathology and psychiatric disorders (principally anxiety and depression requiring treatment) diagnosed by the patient's physician.

Statistical methods

Prevalence rates were adjusted for age and sex by the direct method to the 1990 French (census 1990 - INSEE), U.S. (U.S. Census Bureau, the Official Statistics. Statistical Abstract of the United States: 1998), and world population over 15 years of age. The 95% confidence intervals (CI) of the adjusted rates were then calculated (Breslow & Day, 1987).

For categorical variables, comparisons made using Pearson's chi-square test. For quantitative variables, the mean and standard deviation were reported if the distribution was normal, the median and interquantiles (IQ 25–75%) if not. All analyses were two-tailed, with a p-value of <0.05 being considered statistically significant. The statistical analyses were carried out using SAS statistical software (Cary, NC, U.S.A.).


The study was approved by the Commission National de l'Informatique et des Libertés (French National IRB). Each subject received information on the study before giving consent to the follow-up evaluations.



In all, 903 patients with suspected epilepsy were initially identified. The flow chart is presented in Fig. 1. Nearly half the patients (47.6%) were identified from multiple sources. Only 10 patients were identified from GPs alone. For 49 patients, the diagnosis of epilepsy was not maintained after the files had been examined by the epileptologists or after the diagnostic interview. The diagnoses reassigned were vagal syndrome, cardiac syndrome, cerebrovascular event and symptoms related to chemotherapy of a cerebral tumor (four patients each), pulmonary embolism, narcolepsy, Parkinson's disease, hypoglycemia and Korsakoff's syndrome (one case each). Nearly half of the patients originally classified as presenting generalized epilepsies were reclassified as localization-related epilepsies with secondary generalization at this stage.

Figure 1.

Procedure for the identification of prevalent cases.

For the calculation of the prevalence of epilepsy, provoked seizures (N = 118) and isolated unprovoked nontreated seizures (N = 16) were excluded from the remaining 490 cases. The final number of prevalent cases was thus 360.

The proportion of men was 55.8% (N = 201). The median age was 41 years (interquartiles: 30–58 years) and the median age at onset 19.5 years (range: birth to 88 years; interquartiles: 12–40). The median epilepsy duration was 13.5 years (interquartiles: 5–28 years).


The age-adjusted prevalence of epilepsy is presented in Table 1. Respective age-adjusted prevalence for men and women were 7.76 (95% CI: 6.63–8.89) and 5.20 (95% CI: 4.32–6.07), with a sex-ratio equal to 1.5. Prevalence by age and gender is presented in Fig. 2. The proportion of patients in remission with treatment is 12.5% (n = 45; age-adjusted prevalence rate: 0.71 per 1,000 (95% CI: 0.47–0.95). The prevalence was 0.84 in men and 0.66 in women, respectively. The proportion of patients with treated isolated seizures was 5.5% (n = 20; age-adjusted prevalence: 0.30 per 1,000 [95% CI: 0.15-0.45]), with a prevalence of 0.40 in men and 0.28 in women. This leaves an age-adjusted prevalence of active epilepsy in the strict sense of 5.39 per 1,000 (95% CI: 4.74-6.04), 6.52 for men and 4.24 for women, respectivly.

Table 1.  Age-specific prevalence per 1,000 persons over 15 years
Age (years)Total
PopulationNPrevalence (95% CI)
  1. Data are presented with 95% CI. aAdjusted to the total 1995 French population. bAdjusted to the total 1990 U.S. population.

16–193,891123.08 (1.34–4.83)
20–244,94538 7.68 (5.25–10.12)
25–295,13241 7.99 (5.55–10.42)
30–344,44143 9.68 (6.80–12.56)
35–394,42936 8.13 (5.48–10.77)
40–444,79938 7.92 (5.41–10.43)
45–493,743277.21 (4.50–9.92)
50–543,55827 7.59 (4.74–10.44)
55–594,145153.62 (1.79–5.45)
60–644,617194.12 (2.27–5.96)
65–694,744204.22 (2.37–6.06)
70–743,164175.37 (2.83–7.92)
75 and over7,799273.46 (2.16–4.76)
Total59,407 360 6.06 (5.43–6.68)
France age adjusteda 6.48 (5.80–7.17)
U.S. age adjustedb 6.40 (5.70–7.11)
Figure 2.

Age- and gender-specific prevalence per 1,000 persons over 15 years.

Seizure type

The distribution and prevalence of specific seizure types are presented in Table 2. Only the principal seizure type in a given patient is reported. Multiple seizure types were reported in 12% of patients. Predominant typical absence seizures were associated with generalized tonic–clonic (GTC) seizures in 11 patients and with myoclonic seizures in 10 patients. Predominant GTC seizures were associated in five cases with absences and with myoclonic seizures in five other cases. The three types of seizures were observed in three patients.

Table 2.  Prevalence and distribution of seizure types
Seizures typeNPrevalence/1,000Distribution %
  1. GTC, generalized tonic–clonic seizures.

 Simple localization-related (SLR) 300.50 8.3
 Complex localization-related (CLR)1212.0433.6
 SLR with secondarily generalization  90.15 2.5
 CLR with secondarily generalization 130.22 3.6
 SLR to CLR with secondarily generalization  30.05 0.8
 SLR or CLR with secondarily generalization 330.50 9.2
 Localization-related (undetermined) 110.19 3.1
 Tonic–clonic (GTC) 771.3021.4
 Tonic  20.03 0.5
 Myoclonic (with or without GTC) 150.25 4.2
 Typical absence (with or without GTC) 150.25 4.2
 Atypical absence (with or without GTC)  20.03 0.5
Undetermined focal or generalized 290.49 8.1

Epileptic syndromes and risk factors for symptomatic epilepsies

According to the classification of epileptic syndromes (Table 3), 30.8% of epilepsies were generalized (n = 111), 63.6% localization-related (n = 229), and 5.6% undetermined (n = 20). For the localization-related epilepsies, 72.5% were symptomatic and 27.5% cryptogenic; for the generalized epilepsies 94.5% were idiopathic and 5.5% cryptogenic or symptomatic. Among the idiopathic generalized epilepsies (N = 105), 24 absence epilepsies, 22 juvenile myoclonic epilepsies, three juvenile absence epilepsies, two suffering from epilepsy with grand mal on awakening were identified. For 54 patients, the idiopathic generalized epilepsy could not be classified more precisely. Six cases of Lennox-Gastaut syndrome were identified among the generalized symptomatic or cryptogenic epilepsies. The prevalence of these broad etiopathogenic categories varied according to sex (Table 3). The etiologies of symptomatic epilepsies (n = 171) are presented by gender in Table 4.

Table 3.  Prevalence of epileptic syndromes by sex (rate per 1,000)
1. Localization-related
 1.2 Symptomatic1662.81053.8611.92  
 1.3 Cryptogenic631.1230.8401.20.7
2. Generalized
 2.1 Idiopathic1051.8582.1471.51.4
 2.2 Crypto. or Sympto.  
3. Undetermined200.3120.480.22  
Table 4.  Distribution and prevalence (per 1,000) of symptomatic epilepsy by sex and aetiology
N% Preval.;N% Preval.N% Preval.
Remote symptomatic
 Head injury3028.01.1034.70.093318.70.56
 CNS Infection87.50.29710.90.22158.80.25
 Pre- or perinatal factor1615.00.591421.90.443017.50.50
 CNS malformation54.70.1834.70.0984.70.13
 Chronic alcoholism54.70.1823.10.0674.10.12
 Posttoxic encephalopathy10.90.0411.60.0321.20.03
 Multiple causes65.60.2200.00.0063.50,10
 Other static10.90.0400.00.0010.60.02
Progressive symptomatic
 Cerebral tumor1413.10.511523.40.472917.00.49
 Slow infection00.00.0011.60.0310.60.02
 Degenerative disease10.90.0434.70.0942.30.07
 Autoimmune disease10.90.0423.10.0631.80.05

Pharmacoresistant epilepsy (PR)

Fifty-six patients with epilepsy (15.6%) met criteria for PR epilepsy (Definition 1), corresponding to a prevalence of 0.94 per 1,000 (95% CI: 0.69-1.19). Using the alternative definition of seizure control (Definition 2: seizure frequency of at least one per year), the proportion of PR epilepsy was 22.5% (n = 81), corresponding to a prevalence of 1.36 per 1,000 (95% CI: 1.07–1.66). Of these 81 subjects, 57.1% had failed two first-line AED treatments; 23.8% three treatments, and 16.7% four treatments and more. The seizure frequency was more than one per week in 32% of patients, between one per month and one per week in 42.7% and less than one per month in 25.3%.

The proportion of PR epilepsy was not significantly different between men versus women. The age at onset was significantly lower (p = 0.0007) for PR (mean: 19.6 years ± 19) compared to pharmacoresponsive epilepsy (mean: 28.1 years ± 21). Prevalence of PR epilepsy by age class are reported in Table 5.

Table 5.  Prevalence of pharmacoresistant epilepsy by age and by syndrome (per 1,000)
TotalDefinition 1Definition 2
N% PrevalenceN%Prevalence
561000.94 (0.69–1.19)811001.36 (1.07–1.66)
Age group
 16–1935.40.77 (0–1.64)   4 4.91.03 (0.02–2.03)
 20–393053.61.58 (1.02–2.15)3846.92.01 (1.37–2.64)
 40–591526.8 0.92 (0.46–1.39)2935.81.79 (1.14–2.43)
 >60814.30.39 (0.12–0.67)10 12.350.49 (0.19–0.80)
 1. Localization-related
 1.2 Symptomatic2646.40.443948.10.66
 1.3 Cryptogenic1628.60.272227.20.37
2. Generalized
 2.1 Idiopathic814.30.131113.60.19
 2.2 Crypto. or Sympto.47.10.075 6.20.08
3. Undetermined23.60.034 4.90.07

The proportion of localization-related PR epilepsy was 75.0% using Definition 1 and 79.2% using Definition 2 (Table 5), significantly higher (p = 0.01) than in pharmacoresponsive epilepsy (63.9%). In pharmacoresistant localization-related symptomatic epilepsy (Definition 2, n = 39), the most common etiologies identified were pre- or perinatal factors (33.3%; N = 13), cerebral tumours (20.5%; n = 8) and head injuries (15.4%; n = 6).

A motor disability was present in 17.4% of patients with epilepsy, and a chronic disabling pathology in 16%. These were not significantly more frequent in patients with PR epilepsy (18.9% and 15.1%, respectively, for definition 1) but were associated with symptomatic epilepsies in almost all cases (motor disability in 97% of cases, chronic disabling pathology in 81%). On the other hand, a significantly higher frequency of psychiatric disorder (p = 0.001), mainly anxiety or depression, was diagnosed in patients with PR epilepsy: 37% versus 18.9% in patients with controlled epilepsy (globally 24% in patients with epilepsy). Intellectual disability, observed in 21% of patients with epilepsy, is also more frequent in PR epilepsy (27.2%) than in controlled epilepsy (17.8%; p = 0.06). But, the intellectual disabilities are more specifically related to symptomatic epilepsy with 75% of cases (n = 54) versus 18.7% in idiopathic and 5.3% in cryptogenic epilepsy (p < 0.0001).

A third of patients with PR epilepsy (n = 18) were using three or more AEDs on the prevalence day, while 39.3% were using two. Only 28.6% of the patients with PR epilepsy were being prescribed new AEDs.


This study is the first population-based study on the prevalence of epilepsy in Western Europe, and the only epidemiological study to have used the current classification criteria for seizures and syndromes, together with the ILAE guidelines for epidemiological studies published in 1993. We obtained an adjusted prevalence of 5.39 per 1,000 for active epilepsy in adults. The study's second goal was to determine the prevalence of pharmacoresistant epilepsy in the general population, for which there is no specific data. This prevalence was 1.4 per 1,000, corresponding to a proportion of 22.1% if lack of control is defined by a seizure frequency of at least one seizure per year. With a more restrictive definition, a seizure frequency of at least one per month, this rate is 0.9 per 1,000 and the proportion is 15.3%.

The principal limitations of the study include the retrospective nature of the study, which meant that information on epilepsy type could not always be ascertained, the use of operational definitions of pharmacoresistance, in the absence of an agreed definition, and difficulties related to diagnosis and classification of seizures, in particular in the elderly, which may have biased the observed age-distribution of seizure prevalence (see below).

The two main difficulties encountered with prevalence studies are ensuring complete ascertainment and diagnostic accuracy. We used every possible means to ensure that few cases went undiagnosed and those patients ascertained did indeed have epilepsy. A U.S. study conducted in the general population using a screening questionnaire found that around 7% of subjects with apparent epilepsy had not been diagnosed (Haerer et al., 1986). Considering the current provision of medical care in France and improved awareness of epilepsy, the rate here is probably lower. Furthermore, such cases remain insufficient to be taken into account when planning healthcare resources.

The prevalence of 4–10 per 1,000 obtained is comparable with results from studies carried out in industrialized countries, in particular two Northern European studies carried out on adults in Finland (Keränen et al., 1989) and Sweden (Forsgren, 1992), which used a similar methodology. The definitions of active epilepsy were somewhat different; at least one seizure or ongoing treatment for epilepsy over the previous 5 years in the Finnish study, and at least one unprovoked seizure during the last 5 years or ongoing treatment for epilepsy over the previous year in the Swedish study. The prevalence obtained in these studies was 6.3 per 1,000 (Sweden) and 5.6 per 1,000 (Finland). Unlike these two Scandinavian studies, our study included patients with isolated treated seizures, although their contribution to overall prevalence (0.3%) was minimal.

In our study, epilepsy was most prevalent in 20- to 50-year-olds, with a secondary peak for subjects aged 70- to 74-year-old. Similar age distributions are reported in other prevalence studies (Beran et al., 1982; Granieri et al., 1983; Keränen et al., 1989; Maremmani et al., 1991; Forsgren, 1992; Oun et al., 2003). Although the notion that the prevalence of epilepsy rises with increasing age is a common finding for epidemiological studies (Brodie & Kwan, 2005), the extent of that rise is unclear. Hauser et al. (1991) reported a prevalence of 14.8 per 1,000 for subjects over 75 in Rochester, U.S.A. in 1980. A prospective study carried out in The Netherlands (de la Court et al., 1996) yielded a prevalence of 9.9 to 12.1 per 1,000 in subjects aged 65–74 to 85–94 years, but provoked seizures were included. The lower prevalence for the elderly reported in our study is mainly due to the way seizures following a cerebrovascular event were classified, this event being a major risk factor for seizures in the elderly (Cloyd et al., 2006). Such seizures, generally occurring soon after an ischemic episode and immediately treated, were considered as provoked seizures in our study and were thus excluded. The prevalence of provoked seizures in subjects aged 70–74 was around 3 per 1,000; including these yielded a prevalence which is consistent with the findings of the Netherlands study.

Complex localization-related seizures represent the most common seizure type, as other authors have previously noted (Zielinski, 1974; Keränen et al., 1989; Hauser et al., 1991; Forsgren, 1992; Olafsson & Hauser, 1999). The proportion of unclassifiable seizures was 8.4%, which is consistent with other studies (18% and 8% in the Finnish and Swedish studies, respectively).

When classified by syndrome, 64% of epilepsies were focal, again in agreement with the Finnish and Swedish studies (56% and 60%, respectively). These results were obtained with the use of the SISC and validated by epileptologists. Half of epilepsies initially considered as generalized were reclassified as localization-related epilepsies with secondary generalization. In addition, we observed a high prevalence of symptomatic epilepsies in men, mainly attributable to head injuries, which might explain the higher epilepsy prevalence in men with an overall sex ratio of 1.5.

The second objective of this study was to determine the proportion of pharmacoresistant epilepsies in adults in the general population. Previous studies indicate that 20–30% of patients who suffer from epilepsy will go through at least one phase of pharmacoresistance (Sander, 1993). The variations observed in the literature are mainly due to the absence of a consensus on the definition of pharmacoresistance. In the proceedings of a workshop on “Prevention of refractory epilepsy” (Arroyo et al., 2002), the authors consider that it would be more appropriate to separate patients into “easily controlled” and “difficult to control” according to their response to the first-line AED.

At the time of our study, only 20% of patients with uncontrolled seizures were taking new AEDs. Only vigabatrin was generally available and widely used, although some patients received lamotrigine or gabapentin. Nonetheless, the introduction of the newer drugs has clearly not resolved the problem of pharmacoresistant epilepsy. Response rates to these drugs (i.e., the proportion of patients whose seizure frequency is reduced by at least 50%) are typically lower than 50% (LaRoche & Helmers, 2004). In practice, the proportion of patients with pharmacoresistant epilepsy who become seizure-free with the new AEDs is around 10% (Forsgren et al., 2005; Gazzola & French, 2005). Furthermore, Mohanraj and Brodie have shown that overall response rates with the third treatment schedules and further drug trials is 3.1% (Mohanraj & Brodie, 2006).

Many patients with pharmacoresistant localization-related epilepsies may benefit from surgery. Given a prevalence of 0.71 cases per thousand (Definition 1), we can estimate that there are around 34,000 such cases in France. This represents an upper limit to the estimation, since only one-third of these patients, about 11,000 adult patients, would be eligible for surgery (Engel & Shewmon, 1993). Currently, such surgery is proposed in 12 hospital departments in France with about 300 patients operated on each year, which is probably less than the potential demand.

In conclusion, this study has identified the prevalence of adult epilepsy in Western Europe and found that one-eighth of subjects with epilepsy could be considered pharmacoresistant, principally those with localization-related epilepsies. These results could be used to estimate the target population for new AEDs or healthcare provision for surgery or alternative therapies in pharmacoresistant epilepsy.


We wish to thank Anne Berg and Dorine Neveu for their critical and helpful comments on the manuscript as well as Adam Doble and Teresa Sawyers for his precious help in the English revision of the manuscript. This study was supported by the University Hospitals of Montpellier and Nîmes (PHRC 1995) and the Fondation Française pour la Recherche sur l'Epilepsie (FFRE - Etude Action du ministère de la santé).

Conflict of interest: 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. The authors have no conflict of interest.