• Incidence;
  • Epilepsy;
  • Unprovoked seizures;
  • Review;
  • Epidemiology


  1. Top of page
  2. Abstract

Summary:  Purpose: To evaluate the methodology of incidence studies of epilepsy and unprovoked seizures and to assess the value of their findings by summarizing their results.

Methods: A Medline literature search from January 1966 to December 1999 was conducted. In each selected study, key methodologic items such as case definition and study design were evaluated. Furthermore, a quantitative meta-analysis of the incidence data was performed.

Results: Forty incidence studies met the inclusion criteria. There was considerable heterogeneity in study methodology, and the methodologic quality score was generally low. The median incidence rate of epilepsy and unprovoked seizures was 47.4 and 56 per 100,000, respectively. The age-specific incidence of epilepsy was high in those aged 60 years or older, but was highest in childhood. Males had a slightly higher incidence of epilepsy (median, 50.7/100,000) than did females (median, 46.2/100,000), and partial seizures seemed to occur more often than generalized seizures. Developing countries had a higher incidence rate of epilepsy (median, 68.7/100,000) than did industrialized countries (median, 43.4/100,000). Similar results were found for unprovoked seizures. The incidence of epilepsy over time appears to decrease in children, whereas it increases in the elderly.

Conclusions: The age-specific incidence of epilepsy showed a bimodal distribution with the highest peak in childhood. No definitive conclusions could be reached for the incidence of unprovoked seizures and other specific incidence rates of epilepsy. More incidence studies with an adequate study methodology are needed to explore geographic variations and time trends of the incidence of epilepsy and unprovoked seizures.

Incidence studies provide important information regarding the natural history of epilepsy and its risk factors. During the last three decades, several incidence studies of epilepsy have been performed. These studies indicate that the incidence of epilepsy varies considerably with age and that epilepsy is slightly more common in males. Moreover, recent studies suggest a secular trend in the age-specific incidence, with a decline in children and an increase in elderly people (1,2).

However, the available data are difficult to interpret and compare because of methodologic differences (3,4). So far, only one systematic study on the epidemiology of epilepsy has been done (5). In this study (5), the results of incidence and prevalence studies of epilepsy have been described without conducting a meta-analysis of the incidence data. We performed a systematic review and a quantitative meta-analysis of incidence studies of epilepsy and unprovoked seizures (USs) to study the effect of methodologic quality on the incidence reported, on the magnitude of possible geographic differences in incidence, on the epidemiology of different types of seizures, and the possible changes of incidence over time.


  1. Top of page
  2. Abstract

Study selection

We identified previously published incidence studies by conducting a Medline literature search from January 1966 to December 1999. The precise structure of the Medline search strategy is shown in Table 1. Additional relevant articles were found by tracking references. We considered studies eligible for inclusion if they recorded the incidence of epilepsy or of USs and were published in English, French, German, or Dutch. We excluded studies that included only acute symptomatic seizures, certain seizure patterns, or epileptic syndromes such as absence epilepsy, reflex epilepsy, and Lennox–Gastaut syndrome. We also excluded studies restricted to selected populations because we were interested in the incidence of epilepsy in the general population and not in certain patient groups (e.g., patients with gliomas or a stroke). Furthermore, we excluded hospital-based studies and studies published as a review, a letter, or in abstract form only. If one study had several reports, we used the most recent and complete data for analysis.

Table 1. Results of Medline search
1Epidemiology and publication year 1966–1999460177
21 and epilep*2487
32 near incidence203
43 not acute symptomatic seizures and human  studies198
54 and language English167
64 and language French8
74 and language German2
84 and language Dutch1

Methodologic quality

Based on earlier published proposals for reporting meta-analyses of observational studies (6), we developed a checklist (Appendix 1) to evaluate the methodologic quality of all included studies. Each incidence study was scored on the following items: adequacy of case definition, type of study design, description of demography and selection criteria of the study population, type of described incidence rates, and type of epilepsy or epileptic seizure of which the incidence was defined. The total quality score of each study is given in round numbers from 0 to 30. The maximal score on each specific item depended on the quality of the information provided by the included studies. In accordance with the guidelines for epidemiologic studies on epilepsy of the ILAE (International League Against Epilepsy) (7,8), we considered the items adequacy of case definition and type of study design as important quality items; therefore these items received the highest scores (5 for clear definitions and 9 for a prospective design). Two authors (T.vM. and I.K.) independently evaluated all the incidence studies and reached a consensus with the aid of a third author (F.K.) in cases of disagreement. The degree of agreement was defined by using the Spearman correlation coefficient.

Meta-analysis of results

From each included study, we extracted the country, year of publication, incidence rate of epilepsy or USs, upper or lower age limits of the study, gender- and seizure-specific incidence rates, and total quality score. Furthermore, we recorded whether age-adjusted or crude incidence rates were reported.

The median and ranges of the incidence rate of epilepsy and USs were determined for three different age categories (children, 0–14 years; adults, 15–59 years; elderly, 65 years or older) as well as by gender and seizure type. We estimated incidence rates from figures if a study did not present them in numbers and from original data if a study provided detailed demographic data (denominator) and the number of cases included (numerator). Wherever possible, we used crude incidence rates, as the methods used to adjust the results varied from study to study. Partial and generalized seizures were pooled, and simple partial, complex partial, and secondarily generalized partial seizures were combined. We examined differences in the distribution of the several incidence rates by using the Mann–Whitney or the Kruskal–Wallis tests.

As a sensitivity analysis, the pooling process was repeated after the successive removal of incidence studies with a low quality score. All statistical tests were two-sided, with a p value < 0.01 for statistical significance.


  1. Top of page
  2. Abstract

Selected studies

The Medline search yielded 161 English publications, eight French, one German, and one Dutch (Table 1). Application of the exclusion criteria resulted in 30 incidence studies, and reference tracking yielded another 10 studies. Only four studies (9–12) of 40 were conducted earlier than 1980. Seven (13–19) of the 40 incidence studies were carried out in developing countries, and two (20,21) in eastern European countries. Three studies (22–24) reported incidence rates of other diseases as well.

Methodologic quality

The quality score ranged from 6 to 30 (median, 16). Table 2 shows the methodologic quality scores of the 40 included studies. The interrater agreement on the scale was good (r = 0.98; p = 0.0001). Most disagreement was caused by slight differences in interpretation and was easily resolved. Consensus was reached in all cases. Quality score and publication year of the study did not correlate (r = 0.20; p = 0.21). Subgroup analysis provided similar results for industrialized and developing countries and showed that recently published studies were not associated with a high quality score.

Table 2. Methodological quality scores of the incidence studies
StudyCountryPopulationDefinitionType of studyDemo- graphySelection criteriaType of incidenceScore
  1. med rec and reex, review of medical records and reexamination; med rec, review of medical records; survey and reex, survey and reexamination; US, unprovoked seizures

Sidenvall (38)SwedenchildrenclearprospectiveyesclearUS30
Forsgren (26)SwedenadultsclearprospectiveyesclearUS27
Jallon (39)SwitzerlandgeneralclearprospectiveyesclearUS27
Beilmann (20)Estoniachildrenclearprospectiveyesclearepilepsy27
Jallon (13)Martiniquegeneralclearprospectiveyesnot clearUS26
Forsgren (25)SwedenadultsclearprospectiveyesclearUS25
Verity (40)UKchildrenclearprospectiveyesnot clearboth23
Ellenberg (41)USAchildrenclearprospectiveyesnot clearUS22
Hauser (2)USAgeneralclearmed rec and reexyesnot clearboth22
Braathen (42)Swedenchildrenclearmed recyesnot clearepilepsy22
Olafsson (43)Icelandgeneralclearmed rec and reexyesclearepilepsy22
Hauser (11)USAgeneralclearmed recyesclearepilepsy21
Annegers (44)USAgeneralclearmed recyesclearboth21
Lavados (14)Chilegeneralclearmed recyesclearepilepsy21
Doerfer (45)Germanychildrenclearprospectiveyesnot clearepilepsy20
Granieri (46)Italygeneralclearmed rec and reexyesnot clearepilepsy20
Tsuboi (47)Japanchildrennot clearmed recyesnot clearboth19
Keränen (48)Finlandadultsnot clearmed rec and reexyesclearepilepsy19
Pavlovic (21)Serbiachildrennot clearmed recyesclearepilepsy18
de Graaf (10)Norwaygeneralclearmed recyesnot clearepilepsy16
Blom (12)Swedenchildrenclearmed recyesnot clearepilepsy16
Placencia (15)Ecuadorgeneralclearsurvey and reexyesnot clearepilepsy16
Tekle-Haimanot (16)Ethiopiageneralclearsurvey and reexyesnot clearepilepsy16
Mani (17)Indiageneralclearsurvey and reexyesnot clearepilepsy16
Juul-Jensen (49)Denmarkgeneralnot clearmed recyesnot clearboth15
Doose (50)Germanychildrennot clearmed recyesclearepilepsy15
Cockerell (51)UKgeneralnot clearmed recyesclearepilepsy15
Joensen (52)Denmarkgeneralnot clearmed recyesnot clearepilepsy14
Cockerell (22)UKgeneralnot clearmed recyesnoepilepsy14
Camfield (53)Canadachildrennot clearmed recyesclearepilepsy14
Rwiza (18)Tanzaniageneralnot clearsurvey and reexyesnot clearepilepsy13
Gissler (23)Finlandchildrennot clearmed recyesnot clearepilepsy12
Kurtz (54)UKchildrennot clearmed recyesnot clearepilepsy12
Rantala (55)Finlandchildrennot clearmed recyesclearepilepsy12
Lühdorf (56)Denmarkelderlynot clearmed recyesnot clearepilepsy11
Li (19)Chinageneralnot clearsurvey and reexyesnot clearepilepsy11
vd Berg (9)USAchildrennomed recyesnoUS10
Ishida (57)Japanchildrennot clearmed recyesnot clearepilepsy10
Cavazutti (58)Italychildrennomed recyesnot clearepilepsy9
Okkes (24)The Netherlandsgeneralnomed recnonoepilepsy6

Results from the incidence studies

Tables 3 and 4 summarize the results from the incidence studies. One study (23) did not provide the total incidence of the study population. Figure 1 shows the distribution of the several incidence rates of epilepsy on a logarithmic scale. Because of the small number of incidence studies of USs, we could not reasonably determine the distribution of the specific incidence rates of USs.

Table 3. Reported incidence rates of epilepsy (/100/000/year)
StudyTotalChildren (0–14 yr)Adults (15–59 yr)Elderly (≥60 yr)Male totalFemale total
  • a

     divided the cumulative incidence reported in the study by the length of the observation period in years

Beilmann (20) 45.0    
Verity (40) 430.0    
Hauser (2)
Braathen (42) 53.0    
Olafsson (43)46.567.528.0107.355.936.6
Hauser (11)48.7   51.846.2
Annegers (44)35.5     
Lavados (14)113.0124.4103.4 125.3100.9
Doerfer (45) 52.1    
Granieri (46)33.1  7.539.127.3
Tsuboi (47) 430.0    
Keränen (48)24.0
Pavlovic (21) 650.0    
De Graaf (10)32.8     
Blom (12) 82.3    
Placencia (15)190.0219.0158.0150.3173.0207.0
Tekle-Haimanot (16)
Mani (17)49.361.
Juul-Jensen (49)a124.3     
Doose (50) 72.4    
Cockerell (1)48.360.933.475.233.358.0
Joensen (52)42.871.529.015.0  
Cockerell (22)23.0     
Camfield (53) 41.0    
Rwiza (18)73.393.243.339.768.977.5
Kurtz (65) 840.0    
Rantala (55) 130.0    
Lühdorf (59)   87.0  
Li (19)25.0   32.019.0
Ishida (57) 145.0    
Cavazutti (58) 82.0    
Okkes (24)100.0     
Table 4. Reported incidence rates of unprovoked seizures (/100,000/year)
StudyTotalChildren (0–14 yr)Adults (15–59 yr)Elderly (≥60 yr)Male totalFemale total
  • a

     divided the cumulative incidence reported in the study by the length of the observation period in years

Sidenvall (38) 94.7    
Forsgren (26)56.0 37.0139.055.056.0
Jallon (39)45.6     
Jallon (13)64.1     
Forsgren (25)33.6   38.528.7
Verity (40) 570.0    
Ellenberg (41) 124.0    
Hauser (2)61.0     
Annegers (44)50.9     
Tsuboi (47) 900.0    
Juul-Jensen (49)a198.3     
Lühdorf (59)   104.0  
vd Berg (9) 93.0    

Figure 1. Median and ranges of the incidence of epilepsy.

Download figure to PowerPoint

The age-specific incidence of both epilepsy and USs shows a U-shaped pattern with higher rates for children and the elderly than for adults. Overall, the distribution of the age-specific incidence rates of epilepsy differed significantly (p = 0.008). In particular, children differed significantly from the other age categories (p = 0.002). Furthermore, the incidence of epilepsy and USs is slightly higher for males than for females. However, this difference was not significant.

Generally, the median seizure-specific incidence rates are higher for partial than for generalized seizures, but this difference was not significant (Table 5). The median incidence rate of partial seizures is higher in children (30) than in adults (7). Correspondingly, the median incidence rate of generalized seizures is 45.2 in children and decreases to 21 in adults.

Table 5.  Seizure-specific incidence rates (/100,000/year)
StudyPartial totalGeneralized total
Forsgren (26)38.09.0
Forsgren (25)24.25.6
Hauser (2)40.017.0
Olafsson (43)14.432.1
Hauser (11)30.422.0
Lavados (14)60.843.2
Granieri (46)10.619.6
Joensen (52)21.916.8
Rwiza (18)39.772.1

Geographic differences and changes over time

The incidence of epilepsy appears to differ from country to country and to change over time. These two aspects of the incidence of epilepsy, the geographic and time pattern, are very important to our understanding of the distribution of risk factors for epilepsy. In comparison with the industrialized countries (median, 43.4), developing countries (median, 68.7) have a higher median incidence rate of epilepsy, but this difference was nonsignificant. Moreover, in these studies, higher rates are found in children and adults than in the elderly.

Two studies (11,25) were repeated in the same area by using the same methodology, allowing an assesment of the incidence of epilepsy and USs over time.

In the studies of Hauser et al. (2,11), there was no significant difference in total incidence of epilepsy over time. However, they (2) observed significant changes in the age-specific incidence over time. The incidence in children younger than 10 years decreased by 40% over the 10-year interval between the early study and the most recent one, whereas the incidence in the elderly population almost doubled during the same period.

Forsgren et al. (26) determined the incidence of USs in adults prospectively, and they found significantly higher incidence figures than did a previous study (25). According to the authors (26), the most likely explanation for the increased incidence is a better case ascertainment in the latest study.

Sensitivity analysis

As a sensitivity analysis, incidence studies with a quality score lower than 15, 20, and 25 were removed successively, resulting in a decrease of the median incidence of epilepsy and USs (from 47.6 through 46.5 to 45, and from 58.5 through 53.4 to 50.8, respectively). Comparably, there was a decrease in the median incidence of generalized seizures (from 19.6 through 19.5 to 7.3, respectively) and in the median incidence of epilepsy for females (from 46.2 to 43.6), whereas the median incidence of partial seizures and the median incidence of epilepsy for males slightly increased (from 30.4 through 34.2 to 31.1, and from 50.7 through 51.3 to 53.8, respectively). The median gender-specific incidence for USs remained constant.


  1. Top of page
  2. Abstract

Our meta-analysis shows that the incidence studies of epilepsy and USs differ considerably in their study methods, resulting in a wide variation of the incidence rates. The methodologic quality was generally low, and despite the guidelines for epidemiologic studies offered by the ILAE (8) in 1993, recently published studies did not significantly correlate with a high quality score.

Because planning and implementation of incidence studies may take several years, there may be a lack of correlation between the quality score and the year of publication of studies launched prior to the ILAE publication in 1993.


The incidence of epilepsy and USs showed a bimodal distribution, with the highest peak in childhood. In particular, this distribution was significant for the incidence studies of epilepsy. The information from studies using selected age groups is consistent with the incidence data found in total population studies when age-specific incidence rates are evaluated. Possible causes of epilepsy in childhood are congenital, developmental, and genetic conditions (4,27). The most common cause of epilepsy in the elderly population is cerebrovascular disease (4,27,28).

With respect to the incidence studies of childhood epilepsy, there were large differences in the incidence rates, which may partly be explained by the unclear description of the inclusion criteria of the studies. A few studies did not indicate clearly the inclusion of single seizures, neonatal seizures, or febrile seizures, resulting in high incidence rates. The median incidence of epilepsy in children was 82.2/100,000, which is higher than the annual incidence of epilepsy reported in a recent study of Sillanpää et al. (29). This study reported the long-term outcomes of a population-based cohort of children with epilepsy who were prospectively followed up for several decades. In this study the estimated annual incidence of epilepsy in children was 35/100,000.

Overall, the gender-specific incidence was slightly higher for males than for females, which may be the result of the higher incidence in males of definite risk factors for epilepsy (i.e., head injury, stroke, central nervous system infection). However, a similar pattern is seen when only cases with idiopathic or cryptogenic epilepsy are considered (4,27). Partial seizures seem to be the predominant seizure type in incidence studies. In contrast to studies that have reported that generalized seizures are more common in childhood, this review shows that partial seizures occur more often than generalized seizures both in children and in adults. Studies that reported higher rates of generalized seizures may have, therefore, underreported partial seizures. A recent community-based study of Berg et al. (30) also demonstrated that even in children, partial seizures are more common than generalized seizures. In this study, the generalized seizures occurred in 45.4% of the cohort, and partial seizures, in 55.1%.

Geographic differences

The few studies that have been performed in developing countries have found higher incidence rates than in developed countries, and demonstrated a decrease of the incidence of epilepsy with age. It is likely that this different pattern of age-specific incidence of epilepsy reflects a different etiologic profile of epilepsy. In general, life expectancy in developing countries is shorter, resulting in fewer elderly people with strokes or cerebral tumors, which are common causes of epilepsy in developed countries. Furthermore, a number of specific tropical risk factors, such as infections (i.e., onchocercosis, cysticercosis) and trauma (perinatal or lifelong), are associated with epilepsy, but the exact extent of this contribution is unknown (31).

Changes over time

There is some evidence that the incidence of epilepsy over time decreases in children, whereas it increases in the elderly (4,27,32). Several factors may be involved in this pattern, such as improved antenatal and prenatal care for children, and an increase in life expectancy in the elderly, associated with an increased risk for causes of epilepsy common in old age. The geographic and time pattern of the incidence of epilepsy provide insight into the distribution of risk factors for epilepsy. Additional information regarding the risk factors for epilepsy may be provided by studies of epilepsy in certain patient populations. For instance, recent studies of epilepsy in Iceland studied the risk of epilepsy in specific patient groups (i.e., long-term survivors of surgery for aneurysmal subarachnoid hemorrhage, patients with multiple sclerosis) (33,34). Both studies (33,34) showed that in these specific patient groups, there is an increased risk for epilepsy when compared with that expected in the general population. Acute symptomatic seizures and persistent neurologic impairment seem to be associated with an increased risk for epilepsy among the survivors of a subarachnoidal hemorrhage, whereas among the patients with multiple sclerosis, the reason for this increased risk for epilepsy is unclear.

Methodology of incidence studies

Our meta-analysis of incidence data is accompanied by some limitations related to several methodologic issues of the selected incidence studies.

Most of the incidence studies had a retrospective design and did not provide enough data for a further subgroup analysis of the incidence data (in separate age categories and by gender and seizure type). Furthermore, incidence studies often used different age-adjustment methods and defined age limits differently. However, some studies did define similar age limits (0–15 years for children, 16–59 years for adults, and 60 years or older for the elderly), or the results could be synthesized and analyzed as such.

Another limitation is the issue of classification of epileptic seizures. Recent studies have classified epilepsy and epileptic seizures differently than did older studies. Older studies often used the International Classification of Epilepsies and Epileptic Syndromes by the ILAE in 1989 (35). This classification appeared to be difficult to apply in retrospective studies or field surveys (4). Moreover, the extent to which an epileptic seizure is diagnosed and classified depends on a meticulously obtained seizure description and neurologic investigation (3,4). A further difficulty concerns the definition of epileptic seizures. Many incidence studies did not provide thorough definitions of epilepsy and epileptic seizures, of the selection criteria and demography of the population studied, and did not fully explain their methods, introducing potential uncertainties into the results. In addition, incidence studies may differ in the inclusion of single seizures, neonatal seizures, febrile seizures, and seizures in acute illness resulting in high incidence rates (3). In this meta-analysis, we excluded the aforementioned seizures from the analysis. Moreover, although nonfebrile seizures can be distinguished in provoked and unprovoked seizures, a few studies, especially incidence studies in children, categorized all nonfebrile seizures as USs.

In this study we used a scale to evaluate the overall quality of the incidence studies. Although composite scales may provide a useful overall assessment when comparing different populations or studies published in different languages, their use may be problematic. Therefore, Jüni et al. (36) suggested that first several methodologic aspects should be identified and then assessed individually.

Finally, another limitation is that our assessment of the quality of a study depended on the quality of the report. Taking into account all these limitations, we carried out a sensitivity analysis. This resulted in lower median incidence rates of epilepsy and USs, reflecting a better case ascertainment method, classification, and definition of epileptic seizures. Hauser (27) reported that in incidence studies in France and Rochester, Minnesota, at least half of the newly occurring afebrile seizures did not fulfill the criteria for epilepsy.


The results of our meta-analysis were significant only for the age-specific incidence rates of epilepsy. For the other specific incidence rates of epilepsy and the incidence rates of USs, our findings did not reach significance, probably due to the small number of incidence studies. Similarly, our conclusions regarding the time trends of both the incidence of epilepsy and US, and the pattern of age-specific incidence of epilepsy in developing countries, cannot be definitive due to the low power of the studies. There is an urgent need for further incidence studies with appropriate methodology to document reliably the incidence of epilepsy and US worldwide, noting possible geographic differences and changes in the incidence over time. In particular, these studies should take into account the classification of seizures, epileptic syndromes and risk factors in order to provide valuable information regarding the prognosis of epilepsy and its risk factors. Therefore, the use of the guidelines of the ILAE (7,8) should be emphasized, and incidence studies should clearly state the demography of the studied population, the selection criteria and case ascertainment method. Preferably, in such studies, multiple overlapping search strategies for the identification of incident cases should be used (e.g., including review of neuroradiology reports and EEG files) to avoid selection bias by referral (37).


  1. Top of page
  2. Abstract