FULL-LENGTH ORIGINAL RESEARCH
Cause-specific mortality among patients with epilepsy: Results from a 30-year cohort study
Address correspondence to Eugen Trinka, Department of Neurology, Christian Doppler Klinik, Paracelsus Medical University Salzburg, Ignaz Harrer Strasse 79, A-5060 Salzburg, Austria. E-mail: firstname.lastname@example.org
Purpose: Death rates of patients with epilepsy are two to three times higher than expected. The aim of our study was to further delineate the causes and the patterns of premature death in patients with epilepsy.
Methods: We included all patients who were prospectively enrolled between 1970 and 1999 in our epilepsy outpatient clinical database. Patients were followed until death or December 31, 2003. Standardized mortality ratios (SMRs) were calculated using reference rates from the same region.
Key Findings: After 48,595 person years of follow-up, 648 of 3,334 patients had died, resulting in an overall SMR of 2.2 (95% confidence interval [CI] 2.0–2.4). The highest SMRs were for patients aged 26–45 years (6.8, 95% CI 3.8–11.2) and with symptomatic epilepsies (3.1, 95% CI 2.3–4.9); those for cryptogenic causes (2.2, 95% CI 1.6–3.1) were also elevated, whereas those for idiopathic causes were not increased (2.7, 95% CI 0.7–7.0) after 2 years of follow-up. SMRs for patients with persistent seizures (3.3, 95% CI 2.6–4.4) were higher than those for seizure-free patients (1.4, 95% CI 0.8–2.3). The highest cause-specific SMRs were for epilepsy (91.6, 95% CI 66.3–123.4), brain tumors (22.7, 95% CI 15.7–31.8), and external causes (2.4, 95% CI 1.8–3.3) at end of study period.
Significance: Epilepsy patients have a higher-than-expected risk of death throughout life and especially during the first 2 years following diagnosis. Standardized mortality rates were especially high in younger patients and in patients with symptomatic epilepsies. Persistent seizures are strongly related to excess mortality.
Epilepsy carries a risk of premature death that is two to three times that of the general population (Alstrom, 1950; White et al., 1979; Klenerman et al., 1993; Cockerell et al., 1994; Nilsson et al., 1997; Shackleton et al., 1999; Strauss et al., 2003; Bell et al., 2004). However, the results of studies vary substantially and depend essentially on the population studied. In particular, population-based cohorts differ from hospital-based cohorts; reference populations differ, sample size may be limited, and the diagnosis of epilepsy may not always be confirmed (Alstrom, 1950; White et al., 1979; Klenerman et al., 1993; Cockerell et al., 1994; Nilsson et al., 1997; Shackleton et al., 1999; Strauss et al., 2003; Bell et al., 2004; Mohanraj et al., 2006; Neligan et al., 2011). Population-based studies are generally regarded as gold standard, mainly because of their complete case ascertainment. Their main drawbacks are the lower diagnostic accuracy and the lack of detailed information on etiology and syndrome. To avoid these sources of bias, we studied a large hospital-based cohort of patients in Tyrol, Austria, with confirmed epilepsy, starting continuous enrollment in the 1970s. Some of the patients were therefore followed for 30 years. Standardized mortality ratios (SMRs) were computed using a standard population living in the same region, not the population of Austria.
Herein, we present the 30-year results of this study, which confirms that the risk of premature death is twice that expected and that it is highest in younger patients, patients with symptomatic epilepsies, and patients with persistent seizure activity.
All patients examined at the outpatient clinic for seizure disorders at the Department of Neurology, Innsbruck, Austria, between January 1, 1970 and December 31, 1999 were included in this analysis. Data were captured prospectively since 1970 and entered into a database by one adult neurologist (G.B.). During the 30-year study period, the outpatient clinic for seizure disorders was the only specialized facility for epilepsy patients in the Tyrol region. Patients were either referred to our services by general practitioners, other neurologists, and pediatric epilepsy clinic when discharged at the age of 18 years or through self-referral for a second opinion. Those not permanent residents in the area of Tyrol were excluded from the study.
Diagnosis was established according to the classification system proposed by the International League Against Epilepsy (ILAE) (Commission on Classification and Terminology of the International League Against Epilepsy, 1989). All patients were required to have had at least two unprovoked epileptic seizures. Patients with only acute symptomatic seizures were excluded from the cohort. All patients underwent electroencephalography (EEG) recording at the EEG laboratory of the Department of Neurology, Innsbruck. Epilepsies were classified as cryptogenic, idiopathic, or symptomatic epilepsies, according to the ILAE classification system for epidemiologic studies (1997; 1993) (Commission on Epidemiology and Prognosis, International League Against Epilepsy, 1993; ILAE Commission Report, 1997) Central nervous system (CNS) tumors are known to carry a high mortality risk; therefore, “symptomatic” was defined as “remote symptomatic” for SMRs as suggested by the ILAE (1997) to prevent bias (ILAE Commission Report, 1997). In addition to diagnosis, also recorded were demographic information, date of diagnosis, seizure outcome, and other clinical factors. Patients were followed until death or December 31, 2003, whichever occurred first. Seizure-free patients were scheduled for yearly follow-up visits to the department until end of observation period. For those lost to follow-up we applied last observation carried forward.
The Institute for Clinical Epidemiology (Tiroler Landeskrankenanstalten or TILAK) linked patients in the epilepsy database with the regional death registry of Tyrol, Austria (Oberaigner & Stühlinger, 2005) This linking identified all patients residing in Tyrol who died anywhere in Austria. If death occurred outside Austria (e.g., the patient had emigrated), the death would not be registered, and the patients would be lost to follow-up. Although the number of such deaths is likely to be small, they would nevertheless be missed. The method of probabilistic record linkage is an established and reliable way of identifying deceased patients through comparison of first name, last name, and date of birth (Oberaigner, 2007).
Cause-specific mortalities were analyzed using International Classification of Diseases, Ninth and Tenth Revisions (ICD-9 and ICD-10) codes, as registered with the National Institute for Statistics Austria (Statistik, Austria). The Institute registers the primary cause of death for each person as recorded on the death certificate. Causes of death were classified as the following: neoplasms (ICD-9: 140.0–239.9, ICD-10: C00-D48); neoplasms except CNS tumor (ICD-9: 140.0–190.9/192.0–208.9, ICD-10: C00-C68/C73-C48); CNS tumors (ICD-9: 191.0–191.9, ICD-10: C69-C72); all diseases of the respiratory system (ICD-9: 460.0–519.9, ICD-10: J00-J99); pneumonia (ICD-9: 480.0–486.0, ICD-10: J10-J22), diseases of the circulatory system (ICD-9: 390.0–459.9, ICD-10: I100-I52/I70-I99); cerebrovascular diseases (ICD-9: 430.0–438.9, ICD-10: I60-I69); and external causes such as traffic accidents, drowning, or injury (ICD-9: E800.0-999.0, ICD-10: S00-T98). Deaths from status epilepticus or sudden unexpected death in epilepsy could not be directly identified from ICD coding. Therefore, we defined a category called “direct epilepsy-related” based on codes ICD-9: 345.0–345.9 and ICD-10: G40.0-G41.9, which refer to hospitalizations for epilepsy.
SMRs were calculated using mortality rates for a standard population corresponding to the area of a patient’s last residence (identified by postal code), sex, birth year, and age. They were calculated for 2, 5, 10, 15, 20, 25, and 30 years after the first visit at the epilepsy outpatient clinic. The time point of first contact will be referred to as the entry date. Patients whose seizure onset date and the date of entry were within 24 months of each other were classified as new-onset cases and formed an incident cohort. This subgroup was analyzed separately, in addition to the overall cohort including all patients enrolled. Patients were divided into four groups on the basis of age at diagnosis: younger than 25, 26–45, 46–65 years, and older than 65 years.
SMR values and corresponding 95% confidence intervals (95% CIs) were calculated for each group and results were compared. Values were considered significantly elevated if 95% CIs were >1. Significant differences between groups were assumed if 95% CIs did not overlap. Analyses were performed using STATA (StataCorp, College Station, TX, U.S.A.).
The cohort consisted of 3,334 patients, of whom 1,603 were women and 1,731 were men (Table 1). Number of patients enrolled was number of patients analyzed. There were 48,595 person-years (py) of follow-up and 648 deaths (19.4% of the study population). The incidence cohort consisted of 1,530 patients (19,819 py) of which 851 (10,773 py) were men and 679 (9,046 py) were women.
Table 1. Epidemiologic characteristics of 3,334 patients with epilepsy in Tyrol, Austria, followed for 30 years
|All patients (n = 3,334; 1,731 men, 1,603 women)|| || || |
| Age at first seizure||29.5 (20.4)||24||1–93|
| Age at diagnosis||38 (18.6)||33||1–103|
| Time between seizure onset and first contact||9.5 (11.3)||4||1–70|
| Time between first contact and death||23 (14.3)||21||1–87|
| Age at end of the study period||51.5 (17.5)||48||9–106|
|The incident cohort (n = 1,530; 851 men, 679 women)|| || || |
| Age at first seizure||39.3 (20.1)||35||1–93|
| Age at diagnosis||39.9 (20.1)||35.5||1–93|
| Age at end of study||51.8 (18.5)||49||9–104|
| Survival||13.5 (7.5)||12||1–34|
Standardized mortality ratios by age group
In the overall cohort, 648 deaths were registered where only 297 were expected, resulting in an overall SMR of 2.2 (95% CI 2.0–2.4). The SMR was highest during the first 2 years after diagnosis at 2.7 (95% CI 2.2–3.3), but it remained significantly elevated over the next three decades. The ratios were elevated for men and women in all age groups, but the highest numbers were found for ages 26–45 (Table 2).
Table 2. Standardized mortality ratios for 3,334 patients with epilepsy in Tyrol, Austria, followed for 30 years, by age and time since diagnosis (the first outpatient visit)
| 2||2.7 (2.2–3.3)||2.3 (1.7–3.2)||3.0 (2.3–3.8)||7.7 (3.5–14.6)||6.8 (3.8–11.2)||4.0 (2.7–5.8)||1.8 (1.3–2.4)|
| 5||2.4 (2.1–3.3)||2.3 (1.9–3.2)||2.5 (2.1–2.9)||4.0 (2.1–6.8)||6.4 (4.6–8.6)||3.1 (2.4–3.9)||1.7 (1.4–2.1)|
|10||2.2 (2.0–2.5)||2.0 (1.7–2.4)||2.4 (2.1–2.7)||3.5 (2.2–5.3)||5.4 (4.3–6.8)||2.6 (2.2–3.1)||1.6 (1.4–1.9)|
|15||2.1 (2.0–2.3)||1.9 (1.6–2.2)||2.4 (2.1–2.6)||3.5 (2.4–5.1)||5.0 (4.1–6.1)||2.4 (2.0–2.8)||1.5 (1.3–1.7)|
|20||2.1 (1.9–2.3)||1.8 (1.6–2.1)||2.3 (2.1–2.6)||3.5 (2.3–4.7)||4.6 (3.9–5.5)||2.3 (2.0–2.6)||1.5 (1.3–1.7)|
|25||2.2 (2.0–2.3)||1.9 (1.7–2.1)||2.4 (2.1–2.6)||3.4 (2.4–4.7)||4.4 (3.7–.52)||2.4 (2.1–2.8)||1.5 (1.3–1.7)|
|30||2.2 (2.0–2.3)||1.9 (1.7–2.2)||2.4 (2.1–2.6)||3.6 (2.5–4.8)||4.5 (3.8–5.2)||2.4 (2.1–2.8)||1.5 (1.3–1.7)|
|35||2.2 (2.0–2.4)||1.9 (1.7–2.2)||2.4 (2.2–2.6)||3.5 (2.5–4.8)||4.5 (3.8–5.3)||2.4 (2.1–2.8)||1.5 (1.3–1.7)|
In the incidence cohort, 346 deaths occurred, and men had higher SMRs than did women for the entire follow-up period (2.9–3.1 vs. 2.0–2.7). Younger patients were at higher risk, with an SMR of 5.9 (95% CI 1.2–17.2) in those younger than 25 years; 10.7 (95% CI 5.1–19.7) in 26- to 45-year-olds; 4.8 (95% CI 2.7–7.8) in 46- to 65-year-olds; and 1.8 (95% CI 1.2–2.5) in those older than 65 years in the first 2 years after diagnosis.
Standardized mortality ratios by cause of epilepsy
The highest SMR was in patients with symptomatic epilepsies: 2.8 (95% CI 2.5–3.2) to 3.1 (95% CI 2.3–4.0). The highest SMR occurred during the first 2 years after diagnosis. Patients with cryptogenic epilepsies also had a high SMR of 1.7 (95% CI 1.5–2.0) to 2.2 (95% CI 1.6–3.1), whereas mortality in patients with idiopathic epilepsies was not elevated during the first 10 years after diagnosis, but it increased thereafter, with SMRs ranging from of 1.9 (95% CI 1.2–2.7) to 2.1 (95% CI 1.4–3.0). However, the number of person-years 664–3,072 was low during the early years of follow-up in this group, and therefore the statistical power may have been low (Table 3).
Table 3. Standardized mortality ratios for 3,334 patients with epilepsy in Tyrol, Austria, followed for 30 years, by cause of epilepsy
| 2||3.1 (2.3–4.0)||2.2 (1.6–3.1)||2.7 (0.7–7.0)||1.4 (0.8–2.3)||3.3 (2.6–4.0)|
| 5||2.9 (2.4–3.4)||2.0 (1.6–2.4)||1.1 (0.4–2.7)||1.6 (1.2–2.1)||2.8 (2.4–3.2)|
|10||2.9 (2.5–3.3)||1.7 (1.5–2.0)||1.5 (0.8–2.5)||1.6 (1.3–2.0)||2.5 (2.2–2.8)|
|15||2.8 (2.5–3.2)||1.7 (1.5–1.9)||1.9 (1.2–2.8)||1.5 (1.2–1.8)||2.4 (2.2–2.7)|
|20||2.8 (2.5–3.2)||1.6 (1.4–1.8)||1.9 (1.2–2.7)||1.5 (1.3–1.8)||2.4 (2.1–2.6)|
|25||2.9 (2.6–3.2)||1.7 (1.5–1.9)||2.0 (1.4–2.9)||1.6 (1.3–1.9)||2.4 (2.2–2.6)|
|30||2.9 (2.6–3.2)||1.7 (1.5–1.9)||2.0 (1.4–2.9)||1.6 (1.4–1.9)||2.4 (2.2–2.6)|
|35||2.8 (2.6–3.2)||1.7 (1.5–1.9)||2.1 (1.4–3.0)||1.6 (1.4–1.9)||2.4 (2.2–2.6)|
Results for the incidence cohort were even more pronounced, showing an SMR of 3.1 (95% CI 2.7–3.6) to 3.4 (95% CI 2.7–4.1) in the symptomatic epilepsy group, 1.7 (95% CI 0.9–2.6) to 2.1 (95% CI 1.5–2.7) in the cryptogenic epilepsy group, and nonsignificant levels for the entire idiopathic epilepsy group (SMR 1.7–2.5, 95% CI 0.06–13.87).
Risk in patients with and without abnormal neurologic status did not differ significantly; the SMR was 1.9 (95% CI 1.5–2.4) to 3.1 (95% CI 1.7–5.1) in patients with normal status and 2.2 (95% CI 2.0–2.4) to 2.7 (95% CI 2.2–3.3) in patients with abnormal status. The same finding was true for the incidence group (normal status 2.4–3.2, 95% CI 0.8–5.7; abnormal status 2.5–2.8, 95% CI 2.1–3.6).
Standardized mortality ratios by seizure type
Patients with focal seizures including those with secondary generalization had SMRs almost identical to those with generalized seizures of any type: 2.1 (95% CI 1.9–2.3) to 2.8 (95% CI 2.2–3.5) for focal seizures and 2.2 (95% CI 1.9–2.5) to 2.8 (95% CI, 1.8–4.1) for generalized seizures. Patients with primary and secondary generalized tonic–clonic seizures had the highest mortality, with an SMR ranging from 2.3 (95% CI 2.0–2.7) to 2.8 (95% CI 1.8–4.3). Seizure-free patients did not have elevated SMRs when seizure-free within the first 2 years after diagnosis, and values increased to only 1.6 (95% CI 1.3–2.0) when seizure-free within 10 years of diagnosis. In contrast, non–seizure-free patients had an SMR of 3.3 (95% CI 2.6–4.0) after the first 2 years, a value that declined to 2.4 (95% CI 2.1–2.6) after 20 years (Table 3).
Similar numbers were found for the incidence group; only patients with generalized seizures had higher values, with an SMR of 3.8 (95% CI 2.3–6.0) to 3.0 (95% CI 2.4–3.6).
Causes of death were grouped into nine categories: neoplasm excluding CNS tumors (23%), CNS tumors (5%), all diseases of respiratory system (3%), pneumonia (1%), cardiovascular diseases (26%), cerebrovascular diseases (15%), epilepsy (7%), external causes (9%), and other (11%).
Elevated SMRs were found for patients with neoplasms excluding CNS tumors (SMR 1.8 [95% CI 1.6–2.2] to 2.9 [95% CI, 2.0–4.2]), CNS tumors (SMR 22.7 [95% CI 15.7–31.8] to 30.6 [95% CI 9.9–71.5]), cerebrovascular diseases (SMR 2.4 [95% CI 1.9–3.0] to 3.6 [95% CI 2.1–5.8]), external causes (SMR 2.2 [95% CI 1.6–2.9] to 4.8 [95% CI 2.8–7.7]), and epilepsy (SMR 74.9 [95% CI 38.7–130.8] to 91.6 [95% CI 27.1–194.8]). However, death from cerebrovascular disease was not elevated during the first 2 years after diagnosis. Similar results were found in the incidence cohort. Deaths from pneumonia, other respiratory diseases, and cardiovascular diseases were not associated with higher SMRs. Further comparisons by sex or diagnosis in more specific subgroups were not possible as a result of low patient numbers.
We found that mortality rates were substantially increased among patients with a diagnosis of epilepsy in Tyrol, Austria, over a period of 30 years. This increase was most pronounced during the first 2 years after diagnosis, when it was almost three times higher than that of the standard population. However, even in patients followed for three decades, mortality was still elevated more than twofold. Mortality was higher in men than in women for all age groups. Those aged 26–45 years had the highest risk for premature death, after which the risk decreased with age. Furthermore, mortality was greatly influenced by the cause of epilepsy. Patients with symptomatic epilepsies had significantly higher risk of mortality than did patients with idiopathic epilepsies, whose risk was not increased. Patients who were seizure-free during the last year of follow-up had a lower risk of dying than those who experienced seizures.
Although prospective, population-based cohorts are considered the gold standard for analysing the prognosis in epilepsy patients, they have three major problems. First, a diagnosis of epilepsy is often not assured. Cockerell et al. (1997) reported that of 1,091 patients recruited in their prospective population-based cohort, epilepsy was confirmed in only 564. Second, follow-up periods are usually short, and third, the standard populations used do not allow correcting for regional difference in the mortality of the reference population. The mortality rates in the province of Tyrol are 10% lower than the overall Austrian average (Tiroler Gesundheitsbericht, 2002; http://www.tirol.gv.at/themen/gesundheit/gruppe-gesundheit-und-soziales/gesundheitsbericht/basisberichte/). Therefore, we compared the death rates of our cohort not with the general population of Austria, but with the regional population of Tyrol. To our knowledge this is the first study, which takes the variance of death rates within a county into consideration. This large hospital-based prospective cohort, including patient data of more than 30 years, and ensures high diagnostic accuracy, which is a definite advantage over population-based cohorts. However, over the last three decade the changes in coding the causes of death created some difficulties. During this time the death causes were classified in ICD-9 and ICD-10. The ICD-9 codes, which form the basis of our analysis, did not allow differentiation of direct and indirect epilepsy-related deaths, and only 6.6% of cases had epilepsy as a cause of death on their death certificate. Further comorbidities could not be analyzed because only one code is registered as the main cause of death. Therefore the results of the study may have been influenced by the decision of the physician who documented and coded the cause of death in a given patient. It has been reported, that the frequency of death from cardiovascular diseases has been overestimated by as much as 25% (Goldacre, 1977; Goldacre & Harris, 1980; Lloyd-Jones et al., 1998) and that newly diagnosed diseases were more often reported as cause of death than were the chronic and less “dramatic” disease, such as diabetes mellitus, hypertension, or epilepsy (Johansson & Westerling, 2000, 2002; Bell et al., 2004). In our cohort, the frequency of death from pneumonia was comparatively low, only 1.2% of cases, whereas previous studies reported a range from 3.9% to 18.1% by other authors (Nilsson et al., 1997; Lhatoo et al., 2001). This difference might be related also to differences in coding practice in other countries, (Goldacre, 1993; Johansson & Westerling, 2002) making further studies on this issue necessary.
The strongest risk factor for premature death in epilepsy patients is the underlying cause itself. Patients with a diagnosis of symptomatic epilepsies, associated with neurologic deficits from birth, have the highest risk of dying (Hauser et al., 1980). Nevertheless, mortality rates also appear to be elevated in patients with idiopathic or cryptogenic epilepsies (Hauser et al., 1980; Cockerell et al., 1997; Olafsson et al., 1998; Sillanpaa et al., 1998). Additional risk factors include the occurrence of status epilepticus, generalized tonic–clonic seizures, younger age, male sex, high number of seizures, and a recent diagnosis of epilepsy (Zielinski, 1974; Hauser et al., 1980; Cockerell et al., 1994; Forsgren et al., 1996; Cockerell et al., 1997; Sperling et al., 1999; Lhatoo et al., 2001). We found that mortality was higher for men, younger age groups, and patients with symptomatic epilepsies. We could not identify the specific causes of premature death for patients with symptomatic epilepsies, but it seems plausible that the causes of the epilepsy also increased the risk for premature death. On the other hand, symptomatic epilepsies encompass localization-related epilepsy syndromes, which are in many cases drug-resistant (Semah et al., 1998) and may also contribute to the premature death in the remote symptomatic group. As in population-based studies, SMR was not elevated in patients with idiopathic epilepsies during the first 10 years of follow-up. After 10 years, the SMR was elevated, which might be the result of a higher proportion of more refractory patients or of the recurrence of seizures in older adults after long, seizure-free intervals.
Several studies have assessed the influence of seizure types on mortality (Hauser et al., 1980; Lindsten et al., 2000; Lhatoo et al., 2001). However, the results were not conclusive. This uncertainty might be the result of increasingly smaller patient numbers in more detailed subgroups or by lumping together epilepsy syndromes of different causes when they share a specific seizure type. It combines essentially different syndromes, such as juvenile myoclonic epilepsy and Lennox-Gastaut syndrome, into the same group, since both are characterized by myoclonic seizures. The same explanation may apply to grouping benign rolandic epilepsy with severe therapy-resistant frontal-lobe epilepsy because both are characterized by focal motor seizures. In this study, we also compared focal to generalized seizures, which resulted in increased SMRs over all categories. Furthermore, we compared generalized tonic–clonic seizures regardless of their focal or generalized onset, to all other types of seizures and again found no significant differences in SMRs.
The only significant difference in SMRs was between patients who were seizure-free during the last year of follow-up (SMR 1.4 [0.8–2.3]) and patients who experienced seizures during the first 2 years after diagnosis (SMR 3.3 [2.6–4.0]). This difference remained during the entire follow-up period. Mohanraj et al. (2006) found an SMR of 2.54 among patients with ongoing seizure activity in a cohort of newly diagnosed epilepsy patients compared to 0.95 in those who entered remission in the same cohort. Patients referred to for difficult-to-treat epilepsy had an SMR of 2.04. These findings clearly emphasize the importance of seizure freedom as ultimate goal in the treatment of epilepsy.
The effect of seizure activity on mortality was also assessed by Forsgren et al. over 7 years in a hospital-based cohort of 1,478 patients with learning disabilities. Patients without epilepsy had an SMR of 1.6 (1.3–2.0), whereas those with epilepsy had a significantly higher SMR of 5.0 (3.3–7.5). Rates differed among seizure types, but the differences were not statistically significant (Forsgren et al., 1996). Strauss et al. (2003) also investigated patients with mild learning disabilities and found that the mortality rate was significantly higher in patients with persistent seizures. The highest rates were in patients who experienced status epilepticus during the last year of follow-up. However, all patients had mild learning disabilities, and mortality rates were not compared to the standard population but to other patients with similar learning disabilities but not epilepsy. Therefore, their results cannot be compared to ours. Nevertheless, all these results emphasize that being seizure-free is associated with a lower mortality rate in patients with a diagnosis of epilepsy.
The causes of death in patients with epilepsy fall into two groups: epilepsy-related and non–epilepsy-related. Epilepsy-related deaths can be further categorized into direct epilepsy-related deaths, such as status epilepticus and sudden unexpected death in epilepsy, and indirect epilepsy-related deaths, such as accidents or drowning during seizures. Non–epilepsy-related deaths occur independently without any relation to the person’s epilepsy or seizure activity (Gaitatzis & Sander, 2004). Causes for non–epilepsy-related deaths again vary, depending on the cohort investigated. In hospital-based cohorts, pneumonia is the most common cause of death in 12–25% of such deaths (Iivanainen & Lehtinen, 1979; White et al., 1979; Klenerman et al., 1993), whereas in population-based cohorts, cerebrovascular causes are the dominant reason for death in 12–17% (Hauser et al., 1980; Cockerell et al., 1994). Other frequent reasons for death are trauma, 1–16% (Krohn, 1963; Henriksen et al., 1967; Zielinski, 1974; Iivanainen & Lehtinen, 1979; Hauser et al., 1980; Klenerman et al., 1993; Cockerell et al., 1994) and neoplasms, 5–26% (Iivanainen & Lehtinen, 1979; White et al., 1979; Klenerman et al., 1993).
In our study, mortality for patients with CNS tumors (SMR 22.7–37.1, 95% CI 15.7–31.8, 21.6–59.4), even though patients with a known CNS tumor as cause for the epilepsy were excluded from analysis, and epilepsy (SMR 74.9–92.4, 95% CI 38.7–13.8, 62.0–125.2) was highly increased. Risk was also increased for patients with neoplasms excluding CNS tumors (SMR 1.8–2.9), cardiovascular diseases (SMR 1.3–1.6), cerebrovascular diseases (SMR 2.4–3.6), and external causes (SMR 2.2–4.8). Other causes of death were not associated with an increased SMR, such as pneumonia or other pulmonary diseases. The increased rate of death from external causes may be explained by the higher risk of accidents, especially with active epilepsy. Although other studies have found a higher rate of traffic and other accidents in patients with epilepsy, these accidents were usually less severe and not lethal. However, we did not investigate the risk of nonlethal accidents and so cannot comment on the frequency or severity of accidents in our cohort (van der Lugt, 1975; Nilsson et al., 1997; Mohanraj et al., 2006).
Our results are well in concordance with those from a recent study conducted by Mohanraj et al. in a hospital-based cohort of patients newly diagnosed with epilepsy. The SMRs were 2.6 in their population and 1.6–2.9 in ours for patients with respiratory diseases; 1.5 and 1.3–1.6, respectively, for cardiovascular diseases; and 4.8 and 2.4–4.8, respectively, for external causes. Only for malignant neoplasms excluding CNS tumors was no significant increase found, whereas SMR was elevated with 1.8–2.9 in our population (Mohanraj et al., 2006). More specific subgroups could not be analyzed as a result of low patient numbers. However, these results strongly suggest that underlying conditions carry the greatest risk for premature death, as well as causing epilepsy itself.
Limitations of the study
Although the study population comprises a large number of patients over more than three decades, our study design is retrospective and hospital-based. There is a trade-off between diagnostic accuracy, which is much higher in well-characterized hospital-based cohorts, and the good case-ascertainment, which is only possible in prospective population-based cohorts. Therefore, we regard both types of study complementary to gain further insight into the patterns of mortality in epilepsy as well as their causes.
On the basis of our single-center study, with a large number of patients with confirmed epilepsy followed for up to 30 years, we believe that the results support three conclusions:
- 1 Patients with a diagnosis of epilepsy have a higher-than-expected risk of death throughout life and especially in the first 2 years after diagnosis.
- 2 The number of deaths compared to the standard population is especially high in younger patients and in patients with symptomatic epilepsies.
- 3 Patients who are seizure-free consistently have the lowest risk of premature death at all ages, emphasizing the importance of eliminating seizures in this population.
Tom Ford contributed to this article as editor and was funded by ET.
ET has acted as a paid consultant to Eisai, Medtronics, Bial, Biogen-Idec, Böhringer Ingelheim, Everpharma, and UCB. He has received research funding from UCB, and speakers’ honoraria from Bial, Cyberonics, Desitin Pharma, Eisai, Gerot-Lannacher, GSK, Midas, and UCB. CAG has received travel support from UCB, Eisai, and Gerot-Lannacher. Klaus Seppi has received honoraria for speaking and/or consulting from Novartis, Astra Zeneca, Boehringer Ingelheim, Lundbeck, Schwarz Pharma, UCB Pharma, Teva, and GlaxoSmithKline. 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.