• Epilepsy;
  • Incidence;
  • Epidemiology


  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

Purpose:To describe and report initial findings of a system for prospective identification and follow-up of patients with newly diagnosed single unprovoked seizures and epilepsy in Stockholm, Sweden, the Stockholm Incidence Registry of Epilepsy (SIRE).

Methods:From September 2001 through August 2004, a surveillance system has been in use to identify incident cases of first unprovoked seizures (neonatal seizures excluded) and epilepsy among residents of Northern Stockholm, an urban area with approximately 998,500 inhabitants. Potential cases are identified through multiple mechanisms: Network of health care professionals, medical record screening in specific hospital units, including outpatient clinics, emergency room services, and review of requests for electroencephalography (EEG) examination. Potential cases are classified 6 months after the index seizure based on review of medical records.

Results:After screening approximately 10,500 EEG requests and 3,300 medical records, 1,015 persons met the criteria for newly diagnosed unprovoked seizures (430 single seizures; 585 epilepsy). The crude incidence for first unprovoked seizures and epilepsy was 33.9/100,000 person years, (the same adjusted to the European Standard Million), highest the first year of life (77.1/100,000) and in the elderly. No cause could be identified in 62.4%.

Conclusions:We have established a sustainable system for prospective identification of new onset epilepsy cases in Stockholm. Despite a possible under-ascertainment, the registry provides a useful starting point for follow-up studies.

Prospective population-based studies of incident cases are the ideal for an unbiased assessment of the incidence of unprovoked seizures and epilepsy in the population, as well as for exploration of risk factors and assessment of prognosis (ILAE Commission, 1997). However, such studies are uncommon (Keränen et  al., 1989; Forsgren, 1990; Sander et  al., 1990; Hauser et  al., 1993; Sidenwall et al., 1993; Forsgren et  al., 1996; Olafsson et  al., 1996; Jallon et  al., 1997; Annegers et  al., 1999; Zarrelli et  al., 1999;  Jallon et al., 2001; Öun et  al., 2003; Forsgren et  al., 2005; Olafsson et al., 2005; Christensen et  al., 2007). In these studies, incidence rates have ranged from approximately 20 to 80 per 100,000 person years, the variation accounted for by differences in the population at study, ascertainment rate, methodology, and the criteria used (Sander & Shorvon, 1987). To promote conformity in design and thus facilitate comparison, the International League Against Epilepsy (ILAE) has issued guidelines for epidemiological studies on epilepsy  (Commission, 1993, 1997). However, with a few exceptions (Olafsson et  al., 2005), these recommendations have not been strictly adhered to in previous publications. A limitation of most previous population-based incidence studies of epilepsy is the comparatively small number of included cases. This is important, considering the heterogeneity of epilepsy, and hampers the evaluation of subgroups of, for example, different seizure types, specific etiologies, or age groups.

We set out to establish a system for a prospective identification of all patients with newly diagnosed unprovoked first seizures and epilepsy in Northern Stockholm, Sweden—the Stockholm Incidence Registry of Epilepsy (SIRE). The primary objective was to set up a population-based surveillance system based on data available in medical records that could function over time and thus permit longitudinal epidemiological studies in a large cohort of incident cases. In this first report from the registry, we describe the methodology and the incidence of unprovoked seizures and epilepsy over the first 3 years of SIRE. In the classification of cases, we have applied the ILAE guidelines for epidemiological studies (Commission, 1993; ILAE Commission, 1997) as well as the epilepsy syndrome classification suggested by the ILAE (Commission, 1989).


  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

Setting and study population

The study was conducted in the northern part of  Stockholm, an urban region with 998,500 inhabitants. The median age of the population in this area was 37 years. There are three hospitals serving the inhabitants of the region, but only one has departments of neurology, neurosurgery, and pediatrics. The other two have outpatient  clinics for adult neurological care. There is one central electroencephalography (EEG) laboratory reading all EEGs from the region.

Patients and definitions

Index cases in this report were all residents of the defined region in Stockholm who were first diagnosed with an unprovoked seizure or epilepsy from September 1, 2001 through August 31, 2004. Patients with acute symptomatic seizures or neonatal seizures (onset before 28 days of age) were not included. The index seizure was the seizure that prompted the patient to seek medical advice and that eventually led to their identification by the SIRE. This was not necessarily the patients' first seizure. Patients with index seizure before September 1, 2001 were not included.

Epilepsy was defined as more than one unprovoked seizure occurring less than 5 years apart, according to the guidelines for epidemiological studies of epilepsy issued by the ILAE (Commission, 1993). Multiple seizures in a 24-h period were considered a single event as was an episode of status epilepticus.


A network of reporting physicians and other health care professionals was established to identify potential cases. This consisted of all neurologists (private and public), pediatricians, and geriatricians working in the area, as well as nurses in nursing homes. All were given written and oral information about the project before it was initiated and through newsletters repeatedly throughout the study. For any potential case, the members of this network were asked to submit to the project coordinator a brief note including the unique patient identification number (the 12 digit social security number used in medical records and all health registries in the country), date, their own name, and institution. Additional methods to identify cases included review of all EEG requests to the central EEG lab for investigation on suspicion of new onset seizures, screening of medical records of all new referrals to the neurooncology section of the Karolinska University Hospital, review of medical records of all patients discharged from the department of Neurology or the Department of Pediatrics at the Karolinska University Hospital for the first time with an International Classification of Disease (ICD) code of G40, G41, or R56.8, and review of pediatric emergency room records to evaluate possible cases not reported otherwise.

Six months after first notification of a potential case, the coordinator collected all relevant medical records. All such information generated during the first 6 months after the index seizure was used in the subsequent classification of the case.

All potential cases were evaluated by a panel, which consisted of a neurologist (T.T.), neuropediatrician (P.Å.), a resident in neurology (C.A.), a resident in pediatrics (E.Å.), and the study coordinator, a trained nurse (E.H.). Classification was made by consensus, and the consistency in application of classification criteria over time was ascertained by rereview of all cases. The classification was thus based solely on data that was available in the medical records, was generated as part of routine investigations of the patients, and was completed within 6 months. Each potential case was classified as definite first unprovoked seizure or definite epilepsy. In addition, there were two categories for boarder-line cases: (1) Those with definite seizures, but uncertain whether or not they were provoked and (2) those uncertain whether a seizure occurred, but it was definitely unprovoked.

The index seizure was classified according to the proposal of the ILAE (Commission, 1981). This classification was applied based on three different levels of information. The first level relied only on the semiology of the index seizure as described in the available medical records. Accordingly, a tonic–clonic seizure with an aura or other focal signs or symptoms would qualify for designation as partial, whereas all other tonic–clonic seizures would be labeled “generalized tonic–clonic uncertain if partial or generalized onset.” On the second level, the index seizure was reclassified taking into account information on any unprovoked seizures preceding the index seizure and on those occurring within 6 months after it. On this level, a tonic–clonic index seizure without focal symptoms or signs would be classified as partial secondary generalized if accompanied by partial seizures, and as primary generalized in the event of myoclonic or absence seizures on other occasions. The third level takes into account all available information in the medical records 6 months after the index seizure in the classification of the index seizure. A generalized tonic–clonic seizure without focal symptoms or signs, with normal or unspecific EEG abnormalities, and no focal lesions on neuroimaging, was classified as generalized tonic–clonic uncertain if focal or generalized onset. The same seizure semiology with focal epileptiform EEG abnormalities or a focal lesion on neuroimaging considered likely to be an etiological factor qualified for classification as partial onset. A similar seizure with generalized epileptiform activity on EEG (e.g., polyspike-wave or 2–4 Hz spike-wave activity) would be classified as primary generalized.

We classified the etiology of our cases into three broad categories: (1) Symptomatic, cryptogenic, or idiopathic; the latter reserved for cases consistent with certain epileptic syndromes with particular clinical characteristics, specific EEG findings, and a presumed genetic basis (Commission, 1993, 1997). Symptomatic cases were subdivided into those with a static or progressive etiology (ILAE Commission, 1997). Among the symptomatic cases, those with neurological deficits from birth were identified separately. Cryptogenic were cases where an underlying cause or lesion was suspected, but not identified (Commission, 1993).

All cases were assessed and classified on two levels for more specific types of risk factors and etiologies. On the first level, data in the medical records indicating any condition that might be of relevance as a risk factor was recorded. These were conditions present before the onset of the unprovoked seizure/epilepsy and that had been identified as potential risk factors in previous studies. On this level, individual cases could have more than one risk factor indicated. On the second level, these factors were evaluated with respect to the causal relationship to the seizures, and the presumed etiology in each case was established. This would be a condition preceding seizure onset, known to be an etiological factor for epilepsy, and compatible with the type of seizures/epilepsy of the case in question. Only the most likely etiology was selected.

Mental retardation and cerebral palsy were not considered as causes of epilepsy, but rather as manifestations of an earlier brain insult. These underlying pathologies were used in our etiological classification and included cortical malformations, chromosomal abnormalities, hypoxic ischemic encephalopathies, and injuries to the immature brain. The group “other specified” was used for specific conditions with known association to epileptic seizures, such as cerebrovascular malformations, tuberous sclerosis complex without known chromosomal defects, mesial temporal sclerosis, a history of a cerebral abscess, or tuberculoma.

Statistical analysis

Data were analyzed using SAS 9.1. The number of cases of newly diagnosed epilepsy and first unprovoked seizure within the specified population over the 3-year study period was assessed.

Estimates of the number of person years were made with data from Statistics, Sweden. We added the official populations on December 31 of each year of the 3 years of case registration. The total number of person years was estimated at 2,995,553. Age and sex-specific incidence rates were calculated as well as age-adjusted rates using the European Standard Population as standard population (National Cancer Institute, U.S. National Institutes of Health, 2008). Calculation of the 95% confidence intervals was done using the Poisson distribution (Morris & Gardner, 1988).

Ethical approval was granted by the Ethics Review Board at the Karolinska Institute, Stockholm.


  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References


After screening approximately 3,500 EEG referrals and 1,100 medical records per year, a total of 1,015 patients (566 males) met the inclusion criteria during the study period. Of these, 430 (42.4%) had a first single unprovoked seizure, whereas 585 (57.6%) had recurrent seizures and, thus, a diagnosis of epilepsy. In addition, there were 67 patients in the border-line groups for which the available data did not allow a definitive classification of unprovoked seizures.

The crude incidence rate of all unprovoked seizures (first unprovoked and epilepsy) was 33.9 per 100,000 person years, 37.1 for males, and 30.5 for females. The age adjusted incidence rate according to the European Standard Million (National Cancer Institute, U.S. National Institutes of Health) was also 33.9 per 100,000 person years. The incidence rates by age and sex are summarized in Table 1. The incidence rates were highest among the very young, 77.1 per 100,000 person years under 1 year of age, and in particular among males >85 years of age, 96.9/100,000 person years. The age-adjusted incidence rate was 32.3 the first year of study, 36.7 the second year, and 36.4 for the third year of the study.

Table 1.   Age- and sex-specific incidence rates (per 100,000 person years) of unprovoked seizures/epilepsy in Northern Stockholm September 1, 2001 through August 31, 2004
 Single unprovoked TotalIncidence95% Confidence Incidence95% Confidence Incidence95% Confidence
Age (years)seizuresEpilepsynumberrateintervalNumberrateintervalNumberrateinterval
<1 11 314277.153.8–100.42466.339.8–92.91898.352.9–143.7
 1–4 40 7911951.742.4–61.06640.030.4–49.75381.159.3–102.9
 5–9 47 8413177.764.4–91.06271.653.8–89.46984.164.3–103.9
10–14 39 519049.439.2–59.65356.941.6–72.23741.628.2–55.0
15–19 30 346441.231.1–51.34354.137.9–70.22127.715.8–39.5
20–24 29 204930.021.6–38.43036.923.7–50.01923.212.8–33.7
25–29 11 243516.110.8–21.51917.8 9.8–25.81614.57.4–21.6
30–34 19 163514.5 9.7–19.32218.210.6–25.81310.84.9–16.7
35–39 17 163313.9 9.2–18.71714.1 7.4–20.91613.77.0–20.4
40–44 25 143919.113.1–25.12524.014.6–33.41414.06.7–21.4
45–49 15 153016.110.4–21.91515.9 7.9–24.01516.48.1–24.6
50–54 19 284725.418.2–32.73033.021.2–44.81718.19.5–26.7
55–59 27 305728.020.7–35.22626.116.1–36.13129.819.3–40.3
60–64 30 316140.230.1–50.33647.932.3–63.62532.619.8–45.4
65–69 13 314441.229.0–53.32753.033.0–73.01730.416.0–44.9
70–74 17 183538.825.9–51.61947.526.2–68.91631.816.2–47.3
75–79 18 294755.439.6–71.22469.441.6–97.12345.827.1–64.5
80–84 11 132433.320.0–46.61141.216.8–65.51328.713.1–44.3
>85 12 213353.135.0–71.21796.950.8–142.91635.918.3–53.5

Seizure types

Seizure classification using the three levels of information is presented in Table 2. The proportion with generalized tonic–clonic seizures, uncertain whether primary or secondary generalized, was reduced from 47.4% at the first level to 27.8% at the third level of classification. Generalized-onset seizures occurred in 9.9% (100 of 1,015) of all cases. The age-specific incidence rates of generalized onset, partial onset, and unclassified seizure types are shown in Fig. 1.

Table 2.   Classification of the index seizure based on three levels of information
 Level of information
Seizure typen (%)n (%)n (%)
  1. A, exclusively the semiology of the index seizure; B, all seizures up to 6 months after the index seizure; and C, all available information in medical records up to 6 months after the index seizure.

Generalized seizuresGeneralized tonic–clonic seizure20.2171.7414.0
 Myoclonic seizure161.6161.6161.6
 Tonic/atonic seizures00.000.000.0
Partial seizuresPartial without secondary generalization22622.322622.322422.1
 Partial with secondary generalization24524.130029.640539.9
UnclassifiedGeneralized tonic–clonic seizure, uncertain48147.441040.428227.8
  whether primary or secondary generalized      
Multiple seizure types
Total 1,015100.01,015100.01,015100.0

Figure 1.   Age-specific incidence rates of generalized/partial and unclassified seizure types, based on all available information in medical records up to 6 months after the index seizure.

Download figure to PowerPoint


Symptomatic static seizures accounted for 21.7% and symptomatic progressive seizures for 15.9%. Idiopathic seizures occurred in 7.9%, whereas those with cryptogenic seizures constituted the largest group at 54.4%. In total, the cause of the unprovoked seizures was unknown in 62.4% (633 of 1,015). Specific risk factors and presumed etiologies are summarized by age and gender in

Table 3. A presumed etiology was identified in 28.3% (108 of 382) of patients under 15 years, 37.4% (185 of 494) of those aged 15 to 69 years, and in 64.0% (89 of 139) of those 70 years and older. The most commonly identified etiology was stroke followed by brain tumors (Table 3). Cortical malformations, chromosomal abnormalities, hypoxic ischemic encephalopathies, and congenital brain injury taken together accounted for 5.3% of all cases. Neurological deficits from birth were noted in 10.3%. EEG was performed in 85.2% of the patients, computerized tomography in 77.4%, and magnetic resonance imaging in 20.9%, out of which 46.1%, 61.5%, and 45.8% were normal.

Table 3.   Distribution of risk factors and presumed etiology in patients with newly diagnosed unprovoked seizures/epilepsy in Northern Stockholm September 1, 2001 through August 31, 2004
  Presumed etiology
 Risk factors MalesFemales
 TotalTotal0–14 years15[RIGHTWARDS ARROW]90 years0–14 years15[RIGHTWARDS ARROW]90 years
 n (%)n (%)n (%)n (%)n (%)n (%)
  1. Column percent (%) is out of a total of 1,015 patients.

  2. aIncludes diagnoses such as arteriovenous malformation, periventricular leucomalacia, demyelinating disease, tuberous sclerosis, mesial temporal sclerosis, brain abscess, and tuberculoma.

  3. bIncludes diagnoses such as degenerative CNS disease and malformation syndromes.

Traumatic brain injury313.1222.210.1141.400.070.7
Cortical malformation212.1181.890.910.170.710.1
Chromosomal abnormality212.1191.960.630.370.730.3
Hypoxic ischemic encephalopathy141.4141.460.600.070.710.1
Congenital brain injury414.030.320.
Other specifieda393.8333.3121.260.680.870.7
Brain tumor, primary888.7828.100.0474.610.1343.3
Brain tumor, secondary90.990.900.
Brain tumor, unknown origin10.
Drug addiction121.
Epilepsy among close relatives686.700.
No known causes44944.265964.915515.321220.913913.715315.1


  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

The methodology of SIRE was selected to allow for an enduring surveillance of a fairly large population using limited resources and without any interference with the management of the patients. Although the data are derived from multiple sources, they have been assessed in a uniform manner following predefined operational criteria and applying the recommendations of the ILAE. This has allowed us to establish one of the largest population-based incident cohorts of newly diagnosed unprovoked seizures so far.

Analyzing single and recurrent unprovoked seizures together, the crude incidence was 33.9 of 100,000 person years, with the highest incidence in the first year of life and among the elderly. Adding the borderline cases that could not be definitively classified would yield a crude incidence rate of 36.1 of 100,000 person years. This incidence is in the lower range of the incidence rates of 24 to 69 of 100,000 reported from Europe and the U.S. (Joensen, 1986; Keränen et  al., 1989; Loiseau et  al., 1990; Hauser et  al., 1993; Forsgren et  al., 1996; Jallon et  al., 1997; Olafsson et  al., 2005; Christensen et  al., 2007). In accordance with most previous population-based studies (Joensen, 1986; Keränen et  al., 1989; Hauser et  al., 1993; Olafsson et  al., 1996; Jallon et  al., 1997; Öun et  al., 2003; Christensen et  al., 2007), the incidence was in general higher among males than females, especially above 60 years of age, while at 1–9 years of age, the highest incidence was observed among girls (Table 1). An underlying presumed etiology was identified in 37.6% of the cases, while only 7.9% were found to have idiopathic epilepsy syndromes.

While our methodology has enabled us to register a large number of incident cases, there are also limitations with the design of SIRE. First, the registry is based entirely on data included in medical records, and their quality set the limit. Additionally, the SIRE does not include a standardized work-up. Availability of EEG or neuroimaging depended on if the patient's physician initiated such investigations. Hence, EEG was performed in 85.2% of the cases and neuroimaging in 83.7%, which is slightly less than in some other studies from Europe (Forsgren,1990; Jallon et  al., 2001; Olafsson et  al., 2005). Second, the 6 months follow-up limit after the index seizure can affect the classification. In the Rochester study (Hauser et  al., 1993), time from a first afebrile seizure to diagnosis exceeded 6 months in 50% of the patients and extended beyond 2 years in more than 30%. Third, the size of the surveyed population and the complexity of the health care involved make it difficult to keep the network of reporting collaborators efficient. It is likely that an under-ascertainment contributes to the comparatively low incidence in our study. Although this is probably rare, patients may seek medical advice for their seizures in regions outside Northern Stockholm. Failure to report a potential case or to identify symptoms as possible seizures could also contribute. It has not been possible to quantify this potential under-ascertainment, but it does not seem to change over time, since the incidence rates were similar over the 3 years under study.

The incidence rates in our study seemed to be lower in all age groups compared to most previous studies (Joensen, 1986; Lavados et al., 1992; Hauser et al., 1993; Sidenvall et al., 1993; Braathen & Theorell, 1995; Forsgren et al., 1996; Olafsson et al., 1996; Jallon et al., 1997; Tekle-Haimanot et al., 1997; Annegers et al., 1999; Jallon et al., 1999; Zarrelli et al., 1999; MacDonald et al., 2000; Freitag et al., 2001; Medina et al., 2005), although this difference was most pronounced among the elderly. However, the distribution among our cases by gender, seizure type, and etiology are in keeping with previous studies, indicating that there is no pronounced selection bias in these respects (Forsgren, 1990; Loiseau et al., 1990; Hauser et al., 1993; Forsgren et al., 1996; Jallon et al., 1997; Annegers et al., 1999; Jallon et al., 1999; Zarrelli et al., 1999; Öun et al., 2003).

Many previous studies have failed to account for detailed operational criteria for the seizure classification, although in general, following the ILAE classification (Kotsopoulos et al., 2002), we found it useful to classify the index seizure based on three levels of information (Table 2). The proportion with unclassified seizures was reduced from 47.4% to 27.8% when all available information was taken into account rather than only the semiology of the index seizure, which demonstrates the importance in describing the basis for seizure classification in studies of this kind. The 27.8% considered to have unclassified seizures is still higher than in most other studies (Sander et al., 1990; Manford et al., 1992; Zarrelli et al., 1999). This can be explained by our dependence on medical records, the short follow-up, and the comparatively low proportion with neuroimaging, but also by our application of strict criteria for classification. With our criteria, onset of generalized seizures was confined to childhood and early adulthood, whereas partial onset, and to a lesser extent unclassified seizures, accounted for all cases from 50 years of age (Fig. 1).

The comparatively low rate of neuroimaging could be expected to affect the reliability of the  etiological classification. Nevertheless, the presumed etiologies among our cases were similar to those reported in other population-based studies of epilepsy (Hauser et al., 1993; Olafsson et al., 1996; Jallon et al., 1997, 1999; Olafsson et al., 2005). Stroke was the most common cause followed by brain tumors  (Table 3). We found slightly fewer patients with stroke, 11.3% of presumed etiologies to compare with 14%–20% in other studies from the U.S. and Europe (Sander et al., 1990; Forsgren et al., 1996; Olafsson et al., 1996; Zarrelli et al., 1999; Öun et al., 2003), and 2.2% of presumed causes with dementia, to compare with 3.5%–14.8% (Hauser et al., 1993; Forsgren et al., 1996; Jallon et al., 1997; Olafsson et al., 2005). A higher proportion with dementia and stroke could be expected with a more efficient case ascertainment among the elderly, but our strict criteria for presumed etiology versus risk factors could also contribute.

In conclusion, we have established a surveillance system to identify patients with newly diagnosed unprovoked seizures and epilepsy in Stockholm. Our methods are probably associated with an under-ascertainment of cases, in particular among the elderly. Apart from that, we have no indications of selection bias as the distribution of cases by gender, seizure type, and etiology is comparable to other population-based studies. The large number of newly diagnosed patients with unprovoked seizures and epilepsy in SIRE appears to be suitable for further studies exploring in more detail comorbidities and risk factors as well as long-term follow-up of prognosis utilizing the unique patient identification number that is used in all medical records and public registries in the country.


  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

This study was supported by a grant from the Stockholm County Council (ALF).

Conflict of interest: The authors of this manuscript confirm that we have read Epilepsia's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines. The authors have no conflicts of interest to disclose.


  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References