Mortality of Epilepsy in Developed Countries: A Review


  • The commission wishes to acknowledge the support of UCB Pharma for logistical support in the organization of this workshop.

Address correspondence and reprint requests to Dr. Lars Forsgren at Department of Neurology, Umeå University Hospital, SE-901 85 Umeå, Sweden. E-mail:


Summary:  Mortality in people with epilepsy has been studied in many different populations. In population-based incidence cohorts of epilepsy with 7–29 years follow-up, there was up to a threefold increase in mortality, compared to the general population (standardized mortality ratios [SMR] ranged from 1.6 to 3.0). When studies include selected epilepsy populations where patients with frequent and severe seizures are more common, the mortality is even greater. Relative survivorship (RS) following the diagnosis of epilepsy was 91%, 85%, and 83% after 5, 10, and 15 years, respectively. In a population with childhood-onset epilepsy, RS was 94% and 88% after 10 and 20 years.

The level of increased mortality is affected by several factors. In idiopathic epilepsy where the causes of seizures are unknown, the results are conflicting. There was no significant increase in mortality in studies from Iceland, France, and Sweden, a barely increased risk in a study from the United Kingdom, and a significantly increased risk in a study from the United States. In contrast, all studies report a significant increased mortality in remote symptomatic epilepsy (standardized mortality ratios [SMRs] ranging from 2.2 to 6.5). The highest mortality is found in patients with epilepsy and neurodeficits present since birth, including mental retardation or cerebral palsy (SMRs ranging from 7 to 50).

Mortality is also affected by age, with the highest SMRs in children, the combined effect of low mortality in the reference population, and high mortality in children with neurodeficits and epilepsy. The highest excess mortality is found in the elderly, ≥75 years. A pronounced increase in mortality is found during the first year following the onset of seizures due to underlying severe diseases. The increased mortality remains in different studies 2–14 years following diagnosis.

Most of the factors responsible for the increased mortality are related to the underlying disorder causing epilepsy with pneumonia, cerebrovascular disease, and neoplastic disorders (risk remains elevated when primary brain tumors are excluded), as the most frequently recorded causes. The most common direct seizure-related cause of death in adolescents and young adults is sudden unexpected death, which is 24 times more common than in the general population.

During the last decade, and especially during the last 5 years, several studies on mortality in epilepsy have been published and still others have been initiated with results expected over the coming years. These studies provide important information on an aspect of epilepsy that has largely been ignored in epileptology during the 1980s and early 1990s. During this period, population-based studies showed the prognosis of epilepsy to be much better than previous studies from tertiary referral centers and from epilepsy hospitals with an accumulation of patients with severe epilepsy. This new optimistic picture of epilepsy has been very important and has provided a positive and hopeful view on epilepsy to patients, their relatives, and the general public. However, it must be remembered that a large subgroup of patients with severe epilepsy exists, with increased risk of additional impairments and increased mortality.


The epilepsy population selected for studies of mortality should be representative of the general population with epilepsy. From a global perspective, it is impossible for a single study to be representative since epilepsy populations around the world differ in many aspects (e.g., frequency of risk factors and age composition). While the effect of a factor such as age can be adjusted for in the analysis to compare mortality across studies, such adjustment is impossible for many other factors (e.g., the distribution of risk factors).

There are many potential sources from which epilepsy populations can be identified—diagnostic registries of in- and outpatients at hospitals, registries in EEG laboratories, registries of groups with increased risk for epilepsy (e.g., persons with mental retardation), and other registries. When these registries, which are partly overlapping and partly complementary, are thought to have identified the vast majority of persons with epilepsy in the study area, the study is considered to be population-based and thus representative of the general epilepsy population in the study area. The use of hospital registries as the sole source for identification can render a biased sample with overrepresentation of patients with severe epilepsy. Such a population may overestimate the mortality of all people with epilepsy.

Prognostic studies, such as studies of mortality should include all people with newly diagnosed epilepsy. This is best achieved in incidence studies where all new cases during a specified period are identified. An incidence cohort will contain a larger proportion of cases with mild epilepsy than a prevalence cohort where a proportion of mild cases are missed because they enter remission.

The chances for a more complete ascertainment of cases increases when cases are identified prospectively since the identification net in prospective studies is not entirely built on conventional registries. In prospective studies, case identification can also include key persons working in places where diagnostic registries are incomplete or not up to date (e.g., in wards for old and demented persons).

Thus, epidemiological investigations of mortality in epilepsy should preferably be based on population-based incidence studies and this review mainly reports the results from such studies. However, results from large hospital-based studies are also included with emphasis on studies published during the last decades.


Measures of mortality

The mortality rate estimates the risk of dying and is calculated as the number of deaths during a specified time period divided by the person-years at risk. Mortality rates of epilepsy reported in this review are often supported by additional information to improve the validity of the data.

A special subtype of mortality studies in epilepsy has used death certificates to identify deaths related to epilepsy for the numerator of the mortality rate. However, causes of death are unreliable on death certificates because both false positives and false negatives occur (1).

Mortality in epilepsy is most often presented as the ratio of the observed and expected numbers of death—the standardized mortality ratio (SMR). Expected deaths are calculated by applying the death rates of an external reference population to the age distribution of the study population. If there is no difference in mortality between the study- and the reference-population then the SMR is 1. The 95% confidence interval (CI) provides an estimate of the statistical significance of the calculated SMR. The SMR cannot be compared between studies when there is a difference in age distribution of the study populations, or when age- and sex-specific death rates differ.

The proportionate mortality (PM) gives the percentage of deaths in persons with epilepsy that are due to any one cause. PM studies provide information on the frequency of various causes of death in an epilepsy population. However, in PM studies, the distribution of deaths is not provided for a reference population, making it difficult to know if certain causes of death are under- or overrepresented in the epilepsy population.


Epilepsy is defined as a condition characterized by recurrent (two or more) epileptic seizures, unprovoked by any immediate identified cause (2). Other groups (i.e., single seizures or all seizures except those provoked by fever) are included in some studies. Active epilepsy means that a seizure has occurred during the last 5 years, and the epilepsy is inactive—in remission—when no seizure has occurred during the last 5 years (2). A person can be in remission with or without antiepileptic treatment. In cohort studies of mortality in epilepsy, all members of the cohort have active epilepsy at the time of inclusion in the cohort. When mortality is analyzed after many years of follow-up, the members of the study cohort will, depending on the different prognosis of epilepsies, be a mixture of persons with active and inactive epilepsies, with or without treatment.

Developed countries

Categorizing countries as developed and developing is misleading since all countries change and in that sense are developing. Other terms for developed countries are industrialized countries and countries with established market economies (3). In this review, developed countries include the countries of Europe, North America, Japan, Australia, and New Zealand. Studies on mortality in epilepsy have only been reported from a few developed countries.


Population-based studies

There are five population-based studies of mortality in the entire epilepsy populations (4–8) and one limited to the adult population (9) where mortality was presented as SMR (Table 1). The study from Poland (4) was based on a prevalence population while the remaining studies (5–9) were based on newly diagnosed populations. The studies from the United Kingdom (6), Iceland (7), France (8), and Sweden (9) also included persons who had only had a single unprovoked seizure at the time of the formation of the study cohort. In the U.K. study, 15% of the patients had acute symptomatic afebrile seizures. The SMR in this subgroup was 2.9, similar to the whole group. An extended follow-up of the U.K. study, up to 14 years, reported an SMR of 2.1 (95% CI 1.8–2.4) (10). There are two publications on epilepsy mortality in the Rochester, Minnesota, U.S.A. population (5,11), where the latter study included more patients and had a longer follow-up than the previous study. SMRs range from 1.6 to 4.1 (Table 1) in these studies, which means that in unselected epilepsy populations, the excess mortality is 60–310% greater than the reference population.

Table 1. Mortality in population-based studies of epilepsy
Country, year
(95% CIa)

Study design
  1. SMR, standardized mortality ratio.

  2. aConfidence interval.

  3. bThis study is also referred to as an inception study because the catchment area where patients were identified was not specified.

Warsaw, Poland, 1974 (4)1.8Restrospective, prevalence
Rochester, Minn., U.S.A., 1980 (5)2.3 (1.9–2.6)Historic cohort, incidence
Rochester, Minn., U.S.A., 1984 (10)2.1 (1.9–2.5)Historic cohort, incidence
United Kingdom, 1994 (6)3.0 (2.5–3.7)Prospective, incidenceb
Iceland, 1997 (7)1.6 (1.2–2.2)Historic cohort, incidence
France, 1999 (8)4.1 (2.5–6.2)Prospective, incidence
Sweden, 2000 (9)2.5 (1.6–3.8)Prospective, incidence, adults

The mortality rate for patients with known epilepsy in Warsaw was 7.8 per 100,000 person-years, but it fell to 2.3 per 100,000 person-years if information was only collected from death certificates that reported epilepsy (4). Based on death certificates, the mortality rate in Rochester was 0.68 per 100,000 person-years when epilepsy was listed as the underlying cause of death, and 2.4 if based on any mention of epilepsy on the death certificate (12).

Mortality in populations with childhood-onset epilepsy has been reported from three studies (13–15). In Finland, a population-based cohort of children with epilepsy (245 cases, 61% incident, 39% prevalent) was identified during the period 1961–1964. At follow-up 30 years later, 90% (220 cases) of the cohort could be traced and 44 were found to be dead yielding a mortality rate of 6.2 per 1,000 patient-years (95% CI 5.7–6.7) (13). Several deaths occurred when the cohort had reached adulthood. The mortality rate was 4.8 (95% CI 4.4–5.3) in incident cases and 8.4 (95% CI 7.8–9.0) in prevalent cases. This difference in mortality between incident and prevalent cases was significant with a relative risk of death among incident cases of 0.53 (95% CI 0.31–0.90, p = 0.03) (13).

An Australian study reported on mortality in children with epilepsy where deaths occurred during childhood (14). From a database of all children in the study area, aged 1–14 years, were identified who died and had a diagnosis of epilepsy or seizures, or another neurological diagnosis known to be associated with epilepsy, or died accidentally or unexpectedly. Their medical files were evaluated to confirm or refute the diagnosis of epilepsy. Of 1,095 children who died, 93 had a history of epilepsy that was active or inactive. Assuming a prevalence of childhood epilepsy of 7 per 1,000, the authors estimated that the mortality rate was 3.1 per 1,000 patient-years (95% CI 2.0–4.8) in children with epilepsy compared to a mortality rate of 0.23 per 1,000 (95% CI 0.22–0.25) in children without epilepsy (relative risk = 13.2; 95% CI 8.5–21).

In a study from Nova Scotia, Canada, 686 children up to the age of 16 years were followed for a median of 14 years from onset of epilepsy (15). The SMR relative to reference populations from the 1980s and 1990s were 5.3 (95% CI 2.3–8.3) and 8.8 (95% CI 4.2–13.4), respectively.

The 100-fold difference in mortality rates between the Warsaw and Rochester studies (4,12) and the childhood studies from Finland and Australia (13,14) is explained by whether the total population (4,11) or the epilepsy population (13,14) was used in the denominator. The number produced by using the epilepsy population in the denominator is also called case-fatality.

Selected epilepsy populations

In studies of mortality in selected epilepsy populations, the selection bias differs widely from studies of very selected populations with severe epilepsy to populations that approach the composition of the general epilepsy population. SMRs from selected epilepsy populations are presented in Table 2. The SMRs in these selected populations are generally higher than in population-based studies with SMRs of three or more in six (18,20–24) of the nine studies (16–24). These selected epilepsy populations contain a larger proportion of persons with severe epilepsy than the unselected general epilepsy population. The higher SMRs found in these selected populations indicate that mortality is related to seizure frequency and severity. This is supported by one study in which the SMR was higher in persons with moderate to severe epilepsy than in persons with few seizures or those who were seizure free (17).

Table 2. Mortality in selected populations with epilepsy
Country, year (reference)Type of populationSMR (95% CIa)Age (yr)
  1. SMR, standardized mortality ratio.

  2. aConfidence interval.

Sweden, 1950 (16)Hospital-based2.4 (2.0–2.8)Mainly adults
Denmark, 1970 (17)Hospital-based2.7 (2.3–3.2)Adults
England, 1972 (18)Epilepsy residential care3.0 (2.8–3.3)Median 26
England, 1993 (19)Epilepsy residential care1.9 (1.6–2.3)18–91
England, 1995 (20)Tertiary referral center5.1 (3.3–7.6)10–80
England, 1995 (21)Special school 15.9 (10.6–23)  5–24
Sweden, 1996 (22)Mentally retarded with epilepsy5.0 (3.3–7.5) 1–79
Sweden, 1997 (23)Hospital-based3.6 (3.5–3.7)15–70
Netherlands, 1999 (24)Epilepsy center, incident cases3.2 (2.9–3.5)0.5–70 


According to the presumed etiology of unprovoked seizures, most studies have classified patient in one of the following categories: (a) unknown etiology (idiopathic) (6–9) and (b) seizures due to conditions resulting in static encephalopathy (remote symptomatic) (6–9). The term idiopathic includes all with epilepsy of unknown cause and is not used in the more limited way as defined by the Commission on Classification and Terminology where idiopathic is reserved for certain epileptic syndromes with particular characteristics and specific EEG findings (25). In some studies, remote symptomatic etiology includes patients with seizures and a CNS lesion presumed to be present at birth (i.e., cerebral palsy or mental retardation) (7,9). In other studies, specific classifications have been used for the latter group—neurodeficit (5) or congenital deficit (6). The term postnatal acquired secondary epilepsy is synonymous with remote symptomatic etiology with the neurodeficit group excluded (5,11). Cases with progressive CNS disorders, i.e., brain tumors and neurodegenerative disorders, have often been included in the remote symptomatic group (5–9). In one study (8) and according to guidelines (2), these cases were classified as a separate entity of progressive symptomatic. This study was also the only study to use cryptogenic epilepsies in the classification (8). Cryptogenic epilepsies are presumed to be symptomatic but the etiology is unknown (25). They are usually included in the idiopathic group in epidemiological studies.

Population-based studies of mortality by etiology are shown in Table 3. All studies find lower mortality in the idiopathic group than in the remote symptomatic group. Most studies find either no increase or borderline increase in mortality in the idiopathic group (6–9). Children with epilepsy and without severe neurological deficit do not have increased mortality (15). In the remote symptomatic group, mortality is increased two to five times compared to the mortality of the general population. This range of increased mortality may be due in part to different lengths of follow-up (see below). The highest mortality is found in the group with neurodeficits, with SMRs of 7–50 (5,6,11). An increased mortality in children with neurodeficits and epilepsy was reported in a Swedish cohort that included both incident and prevalent cases of epilepsy who were followed up for 12 years (26). Eleven percent of children with neurodeficit and epilepsy died compared with 2.5% of children with epilepsy, but without neurodeficit (26).

Table 3. Standardized mortality ratios (SMRs) in population-based studies of epilepsy by etiology
Country, year (reference)Etiology (SMR [95% CIa])
IdiopathicRemote symptomaticNeurodeficit
  1. a95% confidence interval.

  2. bIncludes all cases with seizures of unknown etiology: idiopathic and cryptogenic.

Rochester, Minn., U.S.A., 1980 (5)1.8 (1.4–2.3)2.2 (1.8–2.7)11.0 (6.9–16.4)
Rochester, Minn., U.S.A., 1984 (11)1.6 (1.3–1.9)2.8 (2.4–3.4) 7.0 (4.6–10.2)
U.K., 1994 (6)1.6 (1.0–2.4)4.3 (3.3–5.5)50.0 (10–146) 
Iceland, 1997 (7)1.3 (0.8–1.9)2.3 (1.4–3.5) 
France, 1999 (8)  1.5 (0.4−3.9)b 6.5 (3.8–10.5) 
Sweden, 2000 (9)1.1 (0.5–2.4)3.3 (2.4–4.5) 
Range (SMR)1.1–1.8     2.2–6.5      


In most studies (4,5,7,9), mortality was higher in males with epilepsy than in females (Table 4). Further studies are needed to address the reasons for this gender difference in mortality.

Table 4. Mortality in populations with epilepsy by gender
Country, year (reference)Males
SMR (95% CIa)
SMR (95% CIa)
  1. NR, not reported; SMR, standardized mortality ratio.

  2. a95% Confidence interval.

  3. bIdiopathic cases.

Warsaw, Poland, 1974 (4)2.0 (NR)   1.4 (NR)   
Rochester, Minn., U.S.A., 1980 (5)2.1 (1.5–2.8)1.6 (1.1–2.2)
United Kingdom, 1994 (6)2.7 (NR)   3.4 (NR)   
Iceland, 1997 (7) 1.4 (0.9–2.2)b 1.0 (0.4–2.2)b
Sweden, 2000 (9)2.7 (1.8–3.9)2.3 (1.4–3.7)

At all ages, the SMR is increased for people with epilepsy compared to referent populations. Most studies have found an inverse relation between SMR and age. The highest SMRs are found in children, while decreasing SMRs are found with increasing age (Table 5, Fig. 1). The high SMR in children and young adults with epilepsy is a reflection of both the low mortality in the reference population and the high mortality in children with epilepsy and neurodeficits (1,26).

Table 5. Mortality in populations with epilepsy by age
Age (yr)Warsaw (4)
Rochester (5)
SMR (95% CIa)
United Kingdom (6)
SMR (95% CIa)
Denmark (17)
SMR (95% CIa)
  1. SMR, standardized mortality ratio.

  2. a95% Confidence interval.

0–24 8.5 (5.4–12.9) 
0–49 7.6 (4.2–12.5) 
10–30 3.3 (2.1–4.9)
25–44 7.7 (5.1–11.0) 
30–50 5.5 (4.3–7.2)
45–54 3.5 (2.0–5.7)  
50–592.5 8.6 (4.7–14.1)1.5 (1.0–2.2)
55–64 3.0 (2.0–4.5)  
60–691.8 3.6 (2.2–5.5)  
65–74 1.5 (1.0–2.2)  
70–79 1.9 (1.2–2.8)  
≥701.5 0.7 (0.1–2.0)
≥75 1.4 (1.1–1.9)  
≥80 2.6 (1.8–3.6)  
Figure 1.

Standardized mortality ratios (SMR) of epilepsy by age in four studies: Warsaw, Poland (4); Rochester, Minn., U.S.A. (5); United Kingdom (6); and Denmark (17).

Although the highest SMRs are found in children, the highest excess mortality is found in the elderly. From the Rochester study (5), the excess mortality due to epilepsy was calculated and the lowest excess mortality—6 per 1,000—was found in children, where the SMR is highest (1). The highest excess mortality—47 per 1,000, or 8 times higher than in children—was found in the oldest age group, 75 years and older (1).

In the study from Iceland, 81% of the study population had idiopathic epilepsy. When mortality was analyzed by age at onset in idiopathic epilepsy, there was no increase in mortality in any age group (7).


The increase in mortality is most pronounced during the first years following diagnosis. In the Rochester study, the SMR for all epilepsy was significantly increased during the first 10 years after diagnosis, and after 25 years (Fig. 2). In the National General Practice Study of Epilepsy (NGPSE) in the United Kingdom (6), the SMR was significantly increased during the first 4 years following diagnosis, and especially so during the first year (SMR = 6.6, 95% CI 4.8–8.7) (Fig. 3). The increased mortality during the first year after diagnosis is comparable to the French study, reporting an SMR of 4.1 (95% CI 2.5–6.2) (8). In the study from Iceland (7), the mortality was significantly increased during the first 14 years, with most of the increase occurring in the first 4 years (Fig. 4). In the Swedish study (9), a significantly increased mortality was found during the first 2 years, and again, the increase was most pronounced during the first year (SMR = 7.3, 95% CI 4.4–12.1) (Fig. 5). A late increase in mortality after 9–11 years was also found (9).

Figure 2.

Standardized mortality ratios (SMR) and 95% confidence intervals (CI) of epilepsy in Rochester, Minn., U.S.A. by year since diagnosis.

Figure 3.

Standardized mortality ratios (SMR) and 95% confidence intervals (CI) of epilepsy in United Kingdom by year since diagnosis.

Figure 4.

Standardized mortality ratios (SMR) and 95% confidence intervals (CI) of epilepsy in Iceland by year since diagnosis.

Figure 5.

Standardized mortality ratios (SMR) and 95% confidence intervals (CI) of epilepsy in Sweden by year since diagnosis.

The increase in mortality during the first years following diagnosis was found in incidence studies where only newly diagnosed patients with epilepsy were included. In studies primarily or partly based on cases with established epilepsy of several years duration, there is an underrepresentation of patients with life-threatening causes of epilepsy causing increased mortality during the first year(s) following onset of seizures. Despite this a greater increase in mortality has been reported from the majority of studies of selected populations with cases of longstanding epilepsy (Table 2) as compared with population-based incident populations (Table 1). The more marked increase in mortality in populations of cases with longstanding epilepsy, with less life-threatening etiologies, support the assumption that factors other than the etiology of epilepsy affect the level of mortality. As previously suggested, seizure frequency and seizure severity could be some of these factors.

Relative survivorship

The two- to threefold increase of overall mortality in the epilepsy population can produce an impression of death being a more common outcome in epilepsy than it actually is. Another view is provided by the relative survivorship (RS). RS is defined as the proportion of observed to expected number of survivors. The decline in RS is highest during the first years after diagnosis, i.e., during the period with highest mortality. The decline becomes successively less pronounced and RS remains high 20–30 years following onset of epilepsy, especially in idiopathic epilepsy (7,13,27) (Table 6).

Table 6. Relative survivorship (proportion observed to expected number of survivors [%]) of epilepsy
Country, year (reference)Interval since diagnosis (years)
  1. aAfter 40 years.

Rochester, 1975 (27)
 Total 1945–1967 918583 
Males all 848074 
Males idiopathic 908983 
Females all 918789 
Females idiopathic 949191 
Age at diagnosis <1 736961 
(years) 1–19 989899 
20–59 938682 
≥60 8676  
Partial seizures 878483 
Generalized tonic–clonic 928682 
Iceland, 1998 (7)
Remote symptomatic 817666686666
Finland, 1998 (13)
All 94 88   75a

The importance of etiology for survivorship in people with epilepsy is illustrated by a study of childhood-onset idiopathic epilepsy where the RS to the age of 40 years was 87% (95% CI 77–96), but only 73% (95% CI 65–81) in those with remote symptomatic epilepsy (13).


Generalized tonic–clonic seizures (GTCS)

A significant increase in mortality was reported from Rochester for idiopathic GTCS (SMRs of 3.5 and 2.4 for 5 and 30 years following diagnosis), but mortality was not affected in the same group in Iceland (SMR = 1.0) (5,7). In Swedish males with GTCS, there was a statistically significant risk for death (SMR = 3.9, 95% CI 2.3–6.8) in those with idiopathic or remote symptomatic epilepsy (9).

Partial seizures

Mortality was not increased in Rochester patients with complex partial seizures (PS) with or without generalization (SMR = 1.5, ns) (5). In Iceland, idiopathic cases with PS (all types combined) did not have an increased mortality (SMR = 1.5, 95% CI 0.7–2.8) (7). In Sweden, PS were associated with an increased mortality (SMR = 2.1, 95% CI 1.2–3.6) when all etiologies were considered together (9).

Other seizure types

A significantly increased mortality was reported for myoclonic seizures (SMR = 4.1), however, no deaths were observed in patients with absence seizures with or without secondary generalization (5).


Population-based studies with reference population

Population-based studies have found that persons with epilepsy have an increased mortality due to cerebrovascular disease (CVD) (5,6,9), neoplastic disorders (5,6,9), and pneumonia (5,6) (Table 7). In two large population-based studies, the increased risk for death due to neoplasm in people with epilepsy remained after exclusion of primary brain tumors (5,6) (Table 7). The Rochester study also reported an increased mortality due to accidents and noncardiac/cerebrovascular circulatory disorders (5) (Table 7). While suicide is very uncommon in population-based studies of epilepsies (5,6,9), a population-based study from Iceland (28) found a 5.8-fold increased risk for suicide in men (95% CI 1.6–14.8).

Table 7. Cause-specific mortality in epilepsy. SMR and confidence interval (95% CI)
DiseaseStudy (SMR [95% CIa])
United States (5,11)United Kingdom (6)Sweden (9)United Kingdom (18)Sweden (23)
  1. SMR, standardized mortality ratio.

  2. a95% Confidence interval.

  3. bSignificant increase in ages 25–44 years (SMR 5.7, 1.8–13.3) and 45–64 years (SMR 2.5, 1.4–4.1).

Cerebrovascular2.6 (1.8–3.6)3.7 (2.4–5.4)4.2 (2.2–8.0)1.8 (1.2–2.4)5.3 (4.9–5.8)
Heart disease 1.1 (0.8–1.5)b 1.1 (0.9–1.5) 
Ischemic 1.2 (0.6–2.1)1.5 (0.7–3.2) 2.5 (2.3–2.7)
Myocardial insufficiency 1.8 (1.2–3.0) 
Other circulatory7.1 (3.4–13) 
Neoplasms2.9 (2.1–3.9)4.8 (3.4–6.4)3.4 (1.9–5.8)1.5 (1.2–1.9)2.6 (2.4–2.8)
Exclusion of primary brain tumors1.8 (1.1–2.6)3.4 (2.3–4.8) 1.4 (1.1–1.8)2.0 (1.9–2.2)
Pneumonia3.5 (1.6–6.6)10.3 (6.1–16)   7.9 (6.2–9.9)4.2 (3.6–4.8)
Accidents2.4 (1.3–3.7) 4.04.7

Studies on selected populations with a reference population

In a large hospital-based study, where an estimated 80% of the patients on which calculations were based had true epilepsy, mortality was increased for all causes examined (Table 7) (23). Suicide was significantly more common than expected (SMR = 3.5, 95% CI 2.6–4.6). Except for heart disease, mortality was also increased for all causes examined in a long-term residential care unit for patients with epilepsy (18). A later study in the same institution found an increased mortality for epilepsy due to neoplasms (SMR = 2.0, 95% CI 1.3–2.9) (19). Only one of the 29 neoplasms was due to a brain tumor. The SMR was not increased for death due to circulatory diseases, including ischemic heart disease and CVD (SMR = 0.8, 95% CI 0.5–1.1). No suicides occurred.

Studies of proportionate mortality

The distribution of various causes of death has been reported in many papers. The study populations differ widely, from being populations-based (4,5,6,8) or ascertained from general hospitals (23) to ascertained in highly specialized institutions (18,19,29). The results of these studies are therefore very difficult to compare.

As seen in Table 8, CVD accounts for 12–17% of epilepsy mortality except in patients from a long-term residential center for persons with epilepsy where CVD was a less common cause (5–6%) of death (18,19). There is also a substantial difference in the proportion of different causes of death in population-based studies (4,5,6,8) (Table 8). The difference can partly be due to different length of follow-up at the time of analysis. This is best illustrated by a study where deaths due to brain tumors accounted for 53% of the mortality, 1 year following the epilepsy diagnosis (8). Other studies of incident cases with a longer follow-up have found only 9–15% of deaths due to brain tumors (4,5,6). Also lower proportions of deaths due to brain tumors (0.6–0%) were also reported from other populations (18,19,23,30).

Table 8. Causes of death in epilepsy
Cause of deathStudy (proportionate mortality [%])

Warsaw (4)
States (5,11)
Kingdom (6)a

France (8)b

Sweden (23)
Kingdom (18)
Kingdom (19)
  1. a24% of deaths had possible epilepsy.

  2. bUnprovoked seizures + progressive symptomatic.

  3. cSudden unexpected death in epilepsy.

Cerebrovascular1214131714 6 5
Heart disease1619 8 161019
Neoplasm, all242035 161226
Brain tumors15 8 953 4   0.6   0.8
Pneumonia 8 812  51225
Suicide 7   1.6   0.6    1.3   3.3 0
Accidents 4 6  3 7 3
Seizure-related10 6   1.3 3112
SUDEPc 4  0  6
Other152530304118 5


Deaths in patients with epilepsy may be unrelated to epilepsy, related to the underlying cause of epilepsy, or related to seizures. Seizure-related deaths include deaths during status epilepticus, accidents and drowning caused by seizures, and sudden unexpected death in epilepsy (SUDEP). Suicide in persons with epilepsy is sometimes considered to be epilepsy-related. Since suicide is not uncommon among persons who do not have epilepsy, the relation between suicide and the epilepsy disorder must be evaluated on a case-by-case basis.

The pathophysiological mechanisms in SUDEP are presently unknown but are believed to be seizure-related (31,32). Several studies on SUDEP have been published during the last decade and these will not be reviewed in detail here. Suffice it to say that yearly rates of SUDEP vary across different epilepsy populations, from 0.35 per 1,000 person-years in an unselected population-based study (33), to around 1 per 1,000 person-years in somewhat unselected epilepsy populations (34–36). Higher rates of 1.5–6 per 1,000 person-years have been reported in more selected populations (23,37–42), and rates as high as 1 per 100 annually have been reported in patients referred for epilepsy surgery (43) or in patients with continued seizures following surgery (44). In the only study of SUDEP with an unselected population, SUDEP accounted for 1.7% of all deaths in the epilepsy cohort and 8.6% of deaths in the 15–44 year old age group (33). In the 20–40 year old age group, SUDEP exceeded the expected rate of sudden death in the general population nearly 24 times (SMR = 23.7, 95% CI 7.7–55).

Two studies report on the mortality in persons with epilepsy, who were patients in a long-term residential care unit in the United Kingdom (18,19). Mortality was considered to be due to epilepsy in 20–31%. The 20% mortality reported in one of these studies appears to have been seizure-related deaths (19). In an epilepsy hospital in Finland, 35% of patients had a possible seizure-related death (29). A large proportion of these deaths were due to drowning (because the hospital was next to a lake), with 19% of seizure-related deaths remaining when drowned cases were excluded. Two other studies, from Finland and Australia with childhood-onset epilepsy, have reported a large proportion (22–45%) of seizure-related deaths (13,14). The study from Finland was population-based and a 45% seizure-related mortality was also found in the incident subgroup of that study (13). This is very high compared to other population-based studies of incidence cohorts where 1.7–5.5% of deaths were seizure-related (6,11). In a study of mortality in patients with prevalent epilepsy and mental retardation, only 7% of deaths were seizure-related, despite higher mortality in patients with frequent and severe seizures (22). This may indicate that seizure frequency and seizure severity are markers of the severity of the underlying etiology in patients with neurodeficits and contribute to mortality; these factors are less often events directly responsible for death. A recent study from California of persons with mild developmental disabilities also found higher mortality in patients with frequent and severe seizures (45). The proportion with seizure-related deaths was not reported.

The cumulative evidence suggests that the two- to threefold increase in mortality in population-based incidence cohorts with epilepsy are mainly the effect of the underlying causes of epilepsy and, less often, due to direct seizure-related events. Thus, new treatments to reduce or eliminate seizures may have limited effect on mortality in all people with epilepsy, but may have a substantial effect in adolescents and younger adults with intractable epilepsy where the most common direct seizure-related cause, SUDEP, most often occurs (33,44).

The cause for the late rise in mortality found in the Rochester study 25–29 years after onset of epilepsy is unknown (5). While several factors may contribute (e.g., seizure-related, underlying diseases, late effects of antiepileptic treatment), all factors may be partly responsible. An increased mortality during the first 5 years following remission of seizures have also been reported and indicate the importance of factors other than seizure-related factors for mortality in epilepsy of many years duration (5).


One would expect that the profound improvement in the standard of living and in the standard and availability of medical care, including introduction of new antiepileptic drugs and epilepsy surgery, over the past 70–80 years should have resulted in decreased mortality in epilepsy. This question is difficult to answer, because it is difficult to find reliable data to study time trends in epilepsy mortality. Published studies, evaluating trends in epilepsy, are based on death certificates with all the attendant methodological problems (see definitions).

The annual mortality rates for epilepsy have declined from the 1950s to the 1970s (46). A study of epilepsy mortality in England, Wales, and the United States between 1959 and 1994 found secular trends to be similar for both sexes and in both countries (47). Epilepsy mortality declined steeply after 1950 among people younger than 20 years of age. A less pronounced decline was found for adults, while in the elderly mortality declined between 1959 and 1974 but then increased (47). There were pronounced birth cohort effects with a fall with each successive birth cohort indicating that etiological/risk factors varied by generation (47).

The birth cohort effect could be due to a decrease in incidence of epilepsy. A significant decrease in incidence of children younger than 10 years of age was reported from Rochester, U.S.A. (12). A similar trend was reported from the United Kingdom, where the cumulative incidence of epilepsy by the age of 5 years declined from 4 per 1,000 in children born in 1946 (48) to 2.9 per 1,000 in children born in 1958 (49). The reason for the decreased incidence in children is unknown but may partly be due to improved antenatal and perinatal care. The increase in mortality during the last decades in the elderly with epilepsy (47) can be an effect of the increasing incidence of epilepsy in the elderly over time (50–53), which almost doubled during the 50-year study interval in Rochester (51). This increased incidence in the elderly may be partially due to better survivorship from stroke, a group with increased risk for epilepsy (54,55).