Mortality by clinical characteristics in a tertiary care cohort of adult patients with chronic epilepsy


Address correspondence to Olli Nevalainen, School of Public Health, University of Tampere, Tampere FI-33014, Finland. E-mail:


The authors evaluated the contribution of various clinical characteristics to mortality risk and underlying causes of death among all adult patients with epilepsy seen at the Department of Neurology, Oulu University Hospital in Finland during 1996 and 1997. Hazard ratios (HRs) for mortality in 1998–2006 relative to a population-based reference cohort were estimated using Cox modeling, with adjustment for age and gender. The HR for total mortality was 2.66 (95% confidence interval [CI] 2.09–3.39). Infectious etiology of epilepsy (HR 5.77, 95% CI 2.52–13.2) and a seizure frequency of ≥1 per month (HR 4.42, 95% CI 3.00–6.52) related to high risks of death. Cancer (21%), ischemic heart disease (15%), and accidents (12%) caused most of the potential years of life lost. Despite recent advances in treatment of epilepsy and improved seizure control, chronic epilepsy still carries a substantially increased risk of death.

Premature death is considered the ultimate measure of disease impact, and all previous studies on mortality in patients with epilepsy (PWE) have found an increased risk of death relative to the general population. However, there are substantial differences in the risk estimates across subgroups of PWE by etiology and disease severity, with the risk estimates depending strongly on the source population (Shackleton et al., 2002; Neligan et al., 2010). We evaluated the contribution of various clinical features to mortality risk in a well-defined and comprehensive tertiary care clinic–based cohort.


In Finland, the clinical diagnosis of epilepsy is established by a board certified neurologist based on the International League Against Epilepsy (ILAE) criteria (Commission on Classification and Terminology of the ILAE, 1989). According to the national guidelines, all patients suspected of having an epileptic seizure should be referred to specialist medical care (Working group set up by the Finnish Medical Society Duodecim and the Finnish Neurological Society, 2008). The cohort included all 1,383 PWE seen at the neurology clinic of University Hospital of Oulu between January 1, 1996, and December 31, 1997. A randomly sampled, population-based reference cohort without epilepsy was identified from the Population Register Center and consisted of 1,483 individuals matched for age, gender, and municipality (within the region Northern Ostrobothnia). Underlying causes of death were obtained from the Statistics Finland and classified according to the International Statistical Classification of Diseases and Related Health Problems, 10th revision (ICD-10).

Patient charts were reviewed retrospectively to obtain all pertinent medical information. Median age at onset was 23 years (interquartile range [IQR] 14–41). Among the 704 (50.9%) male and 679 (49.1%) female PWE, median duration since diagnosis of epilepsy in years was 11 (IQR 5–21) and 13 (IQR 6–24), respectively. Seizure type was classified according to the recommendations of the ILAE (Commission on Classification and Terminology of the ILAE, 1981), and 1,198 (86.6%) had focal-onset epilepsy (FE), 132 (9.5%) primary generalized epilepsy (PGE), and 53 (3.9%) unclassified epilepsy. Baseline seizure frequency data was an estimate of 1 year preceding the hospital visit at baseline, and was available for 1,336 patients; 753 (54.5%) were seizure free, 377 (27.3%) had <1 seizure per month, and 206 (14.9%) had ≥1 seizure per month.

The follow-up started on January 1, 1998 and ended at death or the common closing date (December 31, 2006), whichever was first. From the original cohort (Ranua et al., 2005), we excluded three PWE and one referent who died before the start of the follow-up. Hazard ratios (HRs) on mortality and their 95% confidence intervals (95% CIs) were estimated for PWE in relation to the reference cohort without epilepsy using Cox proportional hazards analysis with adjustment for age and gender. In the multiple regression analysis, no significant deviations from proportionality assumption were found by plotting the log minus log survival estimate with total mortality as outcome. Excess mortality by cause of death was evaluated as the differences between observed and expected numbers of deaths, with the expected number obtained as the observed number divided by the HR. The number of excess deaths was then related to the overall deaths among PWE. To quantitatively assess premature mortality in PWE, we weighted potential years of life lost (PYLL) using the remaining life expectancy for each individual in 1998 by yearly age and gender based on vital statistics provided by the Statistics Finland.


During the median follow-up time of 9 years, 11,498 person-years of follow-up were accrued and 204 deaths occurred among PWE. For the reference cohort, the corresponding figures were 12,954 and 97. Heart disease was the most frequent cause of death (21%), but cancer accounted for the largest share of the premature mortality (21% of PYLL; Table 1). Age-adjusted HR for total mortality was 2.78 (95% CI 2.05–3.77) for males and 2.48 (1.66–3.70) for females. There was no significant difference in HR for overall mortality by gender (p = 0.57 for an interaction term epilepsy × gender, adjusted for age).

Table 1.   Numbers of deaths and measures of mortality by major causes of death classified according to the ICD-10a among patients with epilepsy (PWE) and referents without epilepsy. Cox model with age- and gender-adjusted hazard ratios (HRs) and 95% confidence intervals (95% CIs)
 Deaths, PWE (n)Deaths, referents (n)HR95% CIExcess mortality (%)bPYLL (%)c
  1. aThe 10th Revision of the International Statistical Classification of Diseases and Related Health Problems.

  2. bRelative excess mortality evaluates the differences between observed and expected number of deaths, with the expected number obtained as the observed number divided by the HR. The number of excess deaths was then related to the overall deaths among PWE.

  3. cPotential years of life lost (PYLL) based on remaining life expectancy in 1998, taking into account the yearly age and gender of each individual.

  4. dCodes: F102, I426, K703, K860, K8600, and X45.

Total mortality204972.662.09–3.391273586.6
Cancer35192.331.33–4.0820 (16)754.3 (21)
Ischemic heart disease43232.441.47–4.0525 (20)542.2 (15)
Cerebrovascular disease2793.991.87–8.5020 (16)317.4 (9)
Influenza and pneumonia110136.3 (4)
Alcohol-related diseases and accidental poisoning by alcohold871.010.37–2.79187.5 (5)
Accidents2192.871.31–6.2713.68 (11)420.3 (12)
Suicides and sequelae of intentional self-harm230.750.13–4.4862.1 (2)

Mortality among patients with symptomatic epilepsy was higher (HR 3.48, 95% CI 2.67–4.55) than among patients with idiopathic/cryptogenic epilepsy (HR 1.95, 95% CI 1.44–2.63; Table 2). By etiologic subtypes, mortality risk was of similar magnitude from intracranial hemorrhage (HR 4.10, 95% CI 2.53–6.66), disruption of cerebral blood flow (HR 4.03, 95% CI 2.66–6.10), neoplasm (HR 3.91, 95% CI 2.30–6.66), cerebral aneurysm (HR 2.83, 95% CI 1.69–4.74), and trauma (HR 2.40, 95% CI 1.43–4.02), whereas slightly higher from intracranial infections (HR 5.77, 95% CI 2.52–13.2).

Table 2.   Age- and gender-adjusted Cox model with hazard ratios (HRs) and 95% confidence intervals (95% CIs) for total mortality in patients with epilepsy by etiology in strata of seizure frequency, seizure type and duration of epilepsy relative to the reference cohort without epilepsy
 Idiopathic/cryptogenic epilepsySymptomatic epilepsy
DeathsHR95% CIDeathsHR95% CI
Seizure frequency at baseline year      
 No seizures381.601.10–2.33562.882.07–4.02
 <1 per month242.341.49–3.67383.292.26–4.79
 ≥1 per month112.741.45–5.16256.354.06–9.92
Seizure type      
 Primarily generalized32.240.68–7.3615.790.80–41.9
Duration of epilepsy      
 ≤10 years462.711.91–3.86804.073.02–5.48
 >10 years331.470.98–2.19432.711.89–3.88

In analyses by seizure type, both patients with PGE (HR 3.33, 95% CI 1.13–9.83) and those with FE (HR 2.68, 95% CI 2.10–3.42) had increased mortality that was comparable between the two patient groups (Table 2). Patients with less than one seizure per month had an almost threefold risk of death (HR 2.81, 95% CI 2.05–3.87) and patients with one or more monthly seizures had a fourfold risk (HR 4.42, 95% CI 3.00–6.52) in relation to the reference cohort without epilepsy. In addition, patients without seizures had an increased risk of death (HR 2.15, 95% CI 1.62–2.86).


This prevalence cohort of patients with a preexisting diagnosis of epilepsy is representative of adults seen by neurologists in tertiary care. In terms of PYLL, the leading causes of premature mortality were cancer, ischemic heart disease and accidents. Although previous studies have been somewhat inconsistent in demonstrating an increased risk of ischemic heart disease mortality among all PWE (Hauser et al., 1980; Lhatoo et al., 2001; Mohanraj et al., 2006), an excess risk of heart disease mortality (including ischemic heart disease and other cardiac diseases) in younger adults with epilepsy remains a concern (Hauser et al., 1980). In the present study, mortality from ischemic heart disease was twofold among PWE and caused a fifth of the excess mortality among PWE (Table 1). Increased mortality from external causes, especially accidents and suicides, has been reported among PWE (Hitiris et al., 2007), with a high proportion of the PYLL (14%) in the present study. Yet, there was no significant increase in suicides, based on only two events among PWE. Otherwise, the risk estimates of the present results on cause-specific mortality are in line with previous findings (Hitiris et al., 2007).

Etiology of epilepsy is among the strongest predictors of mortality. In general, symptomatic epilepsy is consistently associated with a higher risk of death than idiopathic (or cryptogenic) epilepsy (Hitiris et al., 2007; Sillanpää & Shinnar, 2010). Previous studies showed conflicting results regarding whether patients with idiopathic epilepsy have a higher risk of death relative to the general population (Hitiris et al., 2007). In the present study, higher seizure frequency clearly increased mortality risk in patients with symptomatic epilepsy, whereas seizure frequency had less impact among patients with idiopathic/cryptogenic epilepsy.

The patients were treated at a university hospital, which, in comparison to all PWE, probably resulted in a lower proportion of misdiagnosis of other paroxysmal conditions as epilepsy, more difficult-to-treat PWE and fewer patients who achieve remission, as well as many institutionalized subjects (severe congenital disorders and other major comorbidity) being excluded. Therefore, extrapolation of the results to the entire pool of PWE may be limited by these factors. Because follow-up did not start at diagnosis, the survival times of the PWE are truncated from the beginning (left-censoring). Therefore, some degree of bias is involved, in contrast to studies with incidence sampling, where all newly diagnosed patients are included and the immediate postdiagnostic period with higher mortality covered. Use of magnetic resonance imaging was relatively limited before the late 1990s (13.2% of PWE), whereas computerized tomography was performed for at least 85.5% of PWE. Despite recent advances in treatment of epilepsy resulting in improved seizure control in PWE, chronic epilepsy still carries a substantially increased risk of death.


We have no conflicts of interest to disclose. Olli Nevalainen has received financial support from the competitive research fund of the Tampere University Hospital (Grant No. 9F004 and 9H004) and a grant from the Finnish Epilepsy Research Foundation. 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.