Mortality in a Population-based Cohort of Epilepsy Surgery Patients


  • Lena Nilsson,

    1. Department of Neurological Rehabilitation, Stora Sköndal Hospital, Sköndal;
    2. Department of Clinical Neuroscience, Division of Neurology, Karolinska Hospital, Stockholm; and
    Search for more papers by this author
  • Anders Ahlbom,

    1. Division of Epidemiology, Institute of Environmental Medicine, Karolinska Institute, and Division of Epidemiology, Stockholm Center of Public Health, Stockholm, Sweden
    Search for more papers by this author
  • Bahman Y. Farahmand,

    1. Division of Epidemiology, Institute of Environmental Medicine, Karolinska Institute, and Division of Epidemiology, Stockholm Center of Public Health, Stockholm, Sweden
    Search for more papers by this author
  • Torbjōrn Tomson

    1. Department of Clinical Neuroscience, Division of Neurology, Karolinska Hospital, Stockholm; and
    Search for more papers by this author

Address correspondence and reprint requests to Dr. L. Nilsson at Department of Neurological Rehabilitation, Stora Sköndal Hospital, 128 85 Sköndal, Sweden. E-mail:


Summary:  Purpose: To investigate mortality and especially the incidence of sudden unexpected death in epilepsy (SUDEP) in a population-based cohort of epilepsy surgery patients.

Methods: All patients who underwent epilepsy surgery treatment between January 1990 and December 1998 (surgery patients) or whose presurgical evaluation started, although not leading to an operation, during the same period (nonsurgery patients) were identified through the Swedish National Epilepsy register. All subjects were followed up through the Cause of Death Register until December 1998. Standardized mortality ratios (SMRs) for all causes of death and incidence of SUDEP were calculated.

Results: During the study period, 651 surgical operations were carried out on 596 patients (316 male). Of those, 14 patients died (six in SUDEP), rendering a total SMR of 4.9 [95% confidence interval (CI), 2.7–8.3]. SUDEP incidence was 2.4 per 1,000 person years. No major differences were found in SMRs or SUDEP rates between subgroups when stratifying for type of operation and for seizure outcome 2 years after surgery. SMR and SUDEP rates were higher in right-sided temporal lobe resections for gliosis than in left-sided, but the number of deaths was small. Among 212 nonsurgery patients, five died (four in SUDEP). The SMR for all causes was 7.9 (2.6–18.4), and SUDEP incidence, 6.3 per 1,000 person years.

Conclusions: In this large and strictly population-based cohort, SMR for all causes and SUDEP incidence among surgery patients were similar to those of other studies. No differences in overall mortality emerged by seizure outcome, but none of the SUDEP cases was seizure free at the time of death. Four of five deaths in the nonsurgery group occurred during the surgery evaluation period. Mortality appeared to be lower for surgery than for nonsurgery patients, and the interpretation of this finding is discussed.

People with epilepsy have an increased mortality rate, in particular, those with more severe epilepsy (1–10). This excess mortality may be related to the underlying cause of epilepsy, as well as being more directly related to the epilepsy and seizures. Sudden unexpected death in epilepsy (SUDEP) is the most important directly epilepsy-related cause of death. The incidence of SUDEP varies between fewer than one per 1,000 person years in population-based studies of incidence cases (3,4) and as high as 10 per 1,000 person years in cohorts of patients with chronic refractory epilepsy, such as epilepsy surgery candidates (7,8,10). This indicates that poor seizure control may increase the risk of SUDEP, a hypothesis confirmed in recent studies of risk factors for SUDEP (10–15). High seizure frequency, polytherapy with antiepileptic drugs (AEDs), and frequent changes of AED dosages have all been shown to increase the risk of SUDEP (11,13). Given that mortality and especially SUDEP seems to be seizure related, it is reasonable to assume that treatment resulting in improved seizure control also would decrease mortality.

A few recent studies demonstrated a substantially lower mortality rate after epilepsy surgery (16,17); in one study (16), this was strongly associated with seizure relief. However, these studies were based on cohorts from specialized epilepsy centers, and none of them was population based. In Sweden, each of the six epilepsy surgery centers serves a catchment area of 0.9–1.8 million inhabitants, the recruitment being exclusively dependent on a general insurance system and not influenced by the socioeconomic status of the individual patient. All six centers collaborate in a national database established in 1990 with the support of the Swedish National Board of Health and Welfare. The aim of this study was to investigate mortality and especially the incidence of SUDEP in a population-based cohort of epilepsy surgery patients by using this national epilepsy surgery register.


The study cohort

The Swedish Epilepsy Surgery Register is prospective from January 1, 1995, but all data from procedures between September 1990 and December 1994 were collected on a retrospective basis. Every patient who enters a preoperative epilepsy surgery assessment program in any of the six operating centers in Sweden is included in the register. The local centers submit completed case report forms to the common database at the time of decision for operation or if surgery evaluation is discontinued. After surgery, a report form is submitted at 3-month and at 2-year follow-up after surgery. The data held in the register include the seizure characteristics, age at onset, education, social security status, data on presurgical investigations, surgical procedures, tissue pathology, complications, and results at follow-up in terms of the seizure and social outcome. The validity of the data collected from the centers is regularly checked by several systems further described elsewhere (18).

This study cohort consists of all subjects present in the Swedish epilepsy surgery register, who were either operated on during 1990–1998 or whose presurgical evaluation was started during 1990–1998, although not leading to therapeutic surgery.


By using the personal identification numbers, we linked all subjects to the National Cause of Death Register to identify those who had died up to December 31, 1998. Through this linkage, we obtained information on the date of death and underlying cause of death, coded according to the International Coding of Diseases, version 9 or 10 (ICD9, ICD10), as reported on the death certificate.


The number of person years was ascertained individually for all subjects. The date of study entry was the date of first operation or, for those not surgically treated, the starting date of the patient's presurgical evaluation. Date of completion of follow-up was the date of death or the end of follow-up on December 31, 1998.

The observed number of deaths in the cohort was compared with expected numbers based on death rates in Sweden obtained from the Cause of Death Register, and on the number of person-years at risk among members of the cohort. In the analysis, calendar year, sex, and age (by 5-year increments) were taken into account. The standardized mortality ratio (SMR), which is the ratio of the observed to the expected number of cases, was used as a measure of the increased or decreased risk of death (all causes).

SMR of SUDEP could not be calculated, as reliable estimates of numbers of SUDs in the general population are lacking (9). For the dead patients, necropsy reports and medical records from the operating center and from the patient's home district were reviewed by two of the authors (L.N., T.T.), to identify patients whose death could be classified as definite or probable SUDEP. We used the following criteria, as suggested by Nashef and Brown (19,20). Definite SUDEP was defined as sudden, unexpected, witnessed or unwitnessed, nontraumatic and nondrowning death in patients with epilepsy, with or without evidence of a seizure, and excluding documented status epilepticus and where the necropsy examination did not reveal any toxicologic or anatomic cause of death. Cases fulfilling these criteria but in which postmortem examination was not available or difficult to interpret because of postmortem decomposition were classified as probable SUDEP (21).

SMRs for mortality in all-cause mortality as well as the incidence of SUDEP were calculated for the whole cohort (nonsurgery and surgery patients) and in subgroups: by sex, localization of resection, side of temporal lobe resection, and seizure outcome.


A total of 212 of the patients in the register had entered a presurgical assessment that did not lead to a therapeutic operation (nonsurgery patients), for one of three reasons. First, they were rejected based on the results of the workup; second, they declined the operation; or third, because they died during the assessment before surgery. Five of these patients died, four of definite or probable SUDEP, giving an SMR (all causes of death) of 7.9 [95% confidence interval (CI), 2.6–18.4] and a SUDEP incidence of 6.3 per 1,000 person years (Table 1). Further data concerning the deceased nonsurgery patients are shown in Table 2. SUDEP cases are presented in Fig. 1.

Table 1.  Standardized mortality ratio and 95% confidence interval for all cause death and incidence of SUDEP in
nonsurgery patients by sex
 Number of
at risk
Observed number
of deaths
Expected number
of deaths
(95% CI)
Observed number
SUDEP per 1,000
person-years (95% CI)
  • SMR, standardized mortality ratio; CI, confidence interval.

  • a

     Accounted for calendar year, sex, and age by 5-year increments.

All21263550.67.9 (2.6–18.4)46.3 (1.7–16.1)
Males10029830.38.8 (1.8–25.7)26.7 (0.8–24.2)
Females11233620.36.8 (0.8–24.6)25.9 (0.7–21.5)
Table 2.  Characteristics of dead nonsurgery patients
SexAge at epilepsy
onset (yr)
Type of
Types of
Reason for discontinuing
preoperative evaluation
Age at
death (yr)
of death
  1. LRS, localization-related symptomatic; CPS, complex partial; MS, multiple sclerosis; LRC, localization related cryptogenic; PSG, partial secondarily generalized; LR, localization related unspecified; SUDEP, sudden unexpected death in epilepsy.

Males41LRSPSGBilateral not localized seizure onset53SUDEP
 25LRCPSGDeath during evaluation period41MS
 15LRSCPS + PSGDeath during evaluation period36SUDEP
Females16LRCCPS + PSGProbably bilateral temporal foci49SUDEP
 2LRCPS + PSGDeath during evaluation period32SUDEP?
Figure 1.

Characteristics of sudden unexpected death in epilepsy cases, nonsurgery patients.

During the study period, 651 therapeutic operations had been carried out on 596 patients (316 men, 280 women); 106 procedures were performed as a reoperation, and 555 (84.2%) resective surgery procedures and 76 (11.5%) corpus callosotomies were performed. Further data are presented in Table 3.

Table 3.  Characteristics of surgery patients
  • a

     In eight procedures, more than one type of operation was performed.

Age at epilepsy onset, per patient (yr)  
 Mean (SD)10.9(11.2)
Age at surgery, per operation (yr)  
 Mean (SD)25.9(14.4)
Type of epilepsy per operation  
 Localization-related cryptogenic16625.5
 Localization-related symptomatic36956.7
 Localization-related unspecified345.2
 Generalized idiopathic20.3
 Generalized symptomatic or cryptogenic385.8
 Generalized unspecified91.4
 Special syndromes264.0
 Missing data30.5
Type of operationa  
 Temporal lobe resection37256.5
 Selective amygdalohippocampectomy101.5
 Extratemporal resection14722.3

At the time of our analyses, 2-year postoperative follow-up data on seizure control were available for only 500 of 651 operations. Our calculations of mortality ratios and SUDEP rates by seizure control are thus based on this information. Forty patients were given repeated surgery within 2 years, and their treatment outcome after the original operation could not be assessed. Although not included in the present analyses, subsequent attempts to retrieve information on 2-year seizure outcome was successful after 65 additional operations. In the remaining 46, follow-up data were not available, because nine were explorative surgeries; eight were for malignant tumors; seven were dead; in six, the patients or their parents declined follow-up; six had moved abroad; one was not possible to evaluate; two were operated on too late in the period for a 2-year assessment; and in seven (1.1%), missing data were unexplained. Two hundred fifty-nine (46%) patients were classified as seizure free (Engel classification 1 A–D), and another 155 (26%) had >50% seizure reduction 2 years after surgery.

During the study period, four men and 10 women died (Table 4). SUDEP was the cause of death in six (Fig. 2). SMR for all causes of death in the surgery group was 4.9 (95% CI, 2.7–8.3), 2.2 (95% CI, 0.6–5.5) for men and 10.1 (95% CI, 4.8–18.5) for women. SUDEP incidence was 2.4 (95% CI, 0.9–5.3) per 1,000 person years. Stratifying for type of operation, side of temporal lobe resection and seizure outcome revealed no major differences in SMR or SUDEP incidence between subgroups (Table 5). Stratifying for side of operation in patients whose tissue pathology at temporal lobe resection was found to be gliosis showed SMR of 2.0 (95% CI, 0.1–11.0) in left-side and 5.1 (95% CI, 1.0–14.8) in right-side operations. Three SUDEPs occurred in the right and no SUDEPs in the left temporal gliosis group. A high SMR of 19.1 (95% CI, 3.9–55.8) was found in subjects with unknown seizure outcome. This is a small subgroup with few person years at risk, but with three deaths, none of which were SUDEP (see Table 5).

Table 4.  Characteristics of dead surgery patients
SexAge at
surgery (yr)
of death
  1. SUDEP, sudden unexpected death in epilepsy.

Male453.5 yrSUDEP
 209 moSUDEP
 462 yrAstrocytoma-glioma
 343 yrDrowning
Female363 moSUDEP
 381 yrSUDEP
 123 yrSUDEP
 61.5 yrSUDEP
 6224 hIntracerebral postoperative
 504 yrPulmonary carcinoma
 501 yrAstrocytoma-glioma
 471 yrGlioma
 253 yrAstrocytoma-glioma
 112 yrHemorrhage in cerebral tumor
Figure 2.

Figure 2.

Characteristics of sudden unexpected death in epilepsy cases, surgery patients.

Figure 2.

Figure 2.

Characteristics of sudden unexpected death in epilepsy cases, surgery patients.

Table 5.  Standardized mortality ratio and 95% confidence interval for all-cause death and incidence of SUDEP by sex, seizure outcome, and type and side of surgery
 No. of
at risk
of deaths
of deaths
(95% CI)
1,000 person-
years (95% CI)
  • PAD, pathological anatomical diagnosis.

  • a

     Accounted for calendar year, sex, and age by 5-year increments.

  • b

     At 2 years after surgery.

  • c

     Engel 1A–D.

All patients5962455142.94.9 (2.7–8.3)62.4 (0.9–5.3)
 Male316129941.92.2 (0.6–5.5)21.5 (0.2–5.6)
 Female2801156101.010.1 (4.8–18.5)43.5 (0.9–8.9)
Surgical procedure       
 Temporal lobe resection355140671.84.0 (1.6–8.2)32.1 (0.4–6.2)
 Left16666030.83.6 (0.7–10.5)34.5 (0.9–13.3)
 Right16770430.93.3 (0.7–9.7)00.0 (0.0–9.5)
PAD: gliosis left9036210.52.0 (0.1–11.0)00.0 (0.0–10.2)
PAD: gliosis right8636130.65.1 (1.0–14.8)38.3 (1.7–24.3)
Extratemporal resection16065850.86.4 (2.1–14.9)23.0 (0.3–11.0)
Other operations9442220.36.5 (0.8–23.3)12.4 (0.1–13.2)
Seizure outcomeb       
 Seizure freec228110251.33.8 (1.3–9.0)32.7 (0.6–8.0)
 Recurrent seizures255113361.44.3 (1.6–9.4)32.6 (0.6–7.7)
 Unknown11322030.219.1 (3.9–55.8)00.0 (0.0–16.8)


This is the largest follow-up of mortality in an epilepsy surgery cohort published. It is of particular interest because, in contrast to previous reports, it represents a complete coverage of all epilepsy surgery patients in a country during a specified period. The Swedish health care system linked to the unique personal identification numbers enabled us to track all individuals over the study period. However, despite the large cohort, the number of deaths was small, which made some estimates imprecise and limited subgroup analyses. A limitation in our study is that the information concerning seizure control is confined to the report 2 years after surgery, and furthermore, that such information is missing in a proportion of the cohort we analyzed. All medical records of the deceased patients were, however, reviewed in detail to determine seizure frequency as well as cause of death. Some of the deceased patients who had been seizure free 2 years after surgery were found to have had relapses later when the medical records were reviewed (see Fig. 2.)

Our analyses included 212 patients who were present in the registry because they had been accepted for epilepsy surgery workup. This included people awaiting a decision about surgery and people who, for various reasons, were never given therapeutic epilepsy surgery. Although efforts have been made to adopt common criteria for inclusion of such patients in the national epilepsy surgery register, their application is likely to differ somewhat between centers and over time, so this comparison group is probably incomplete and heterogeneous. In addition, this is not an ideal control group for the surgery patients; the fact that most of them were rejected for surgery demonstrates that they were different from the surgery group in ways that might affect mortality.

Investigating cohorts of epilepsy surgery candidates does not provide information on mortality that is representative of the epilepsy population in general. However, it provides the opportunity to study patients with severe and chronic epilepsy of whom a substantial proportion of patients will have a radically improved seizure situation after surgery. If mortality were demonstrated to be lower in patients rendered seizure free by surgery, this would additionally support the hypothesis that occurrence of seizures is an important risk factor for death.

In the present study we took advantage of the unique qualities of the Swedish epilepsy surgery register to assess mortality and in particular the incidence of SUDEP in patients subjected to epilepsy surgery workup and treatment.

SMR and SUDEP incidence in surgery patients and in the total cohort

We found an SMR of 4.9 (95% CI, 2.7–8.3), a fivefold higher mortality rate among surgery patients than in the general population. We also found that the incidence of SUDEP was 2.4 per 1,000 person years among these patients. This corresponds with results of the few previously published studies in which number of person years is given and calculations of SMR and incidence are possible (16,17,22,23). In a cohort of patients treated with temporal lobe resections, Hennessy et al. (17) found an SMR of 4.5 and a SUDEP incidence of 2.2 per 1,000. Sperling et al. (16) demonstrated SUDEP incidence to be four per 1,000 person years after epilepsy surgery. SMR for all causes was 4.7 among patients with recurrent seizures, but there were no deaths in patients becoming seizure free (16). SUDEP incidence among these surgery patients, including the present study, was lower than the estimated mortality rate before surgery among epilepsy surgery referrals, as reported by Dasheiff (10). Dasheiff (10) described seven SUDEP cases, of which four occurred during the evaluation period before operation, giving a total SUDEP incidence of nine per 1,000 person years. In the present study, the closest corresponding figure [SUDEP incidence in the total cohort (i.e., both surgery and nonsurgery patients)] was 3.2 per 1,000 person years, still considerably lower than that in the Dasheiff cohort.

In the present study, six (43%) of 14 deaths among surgery patients and four (80%) of five deaths among nonsurgery patients were due to SUDEP, obviously the single most important cause of death in this population. As the total number of deaths other than SUDEP was only nine, calculations of cause-specific SMRs were not meaningful. Five deaths were due to brain tumor. Tissue pathology at surgery in these patients was ganglioglioma in three patients, astrocytoma grade 1–2 in one, and astrocytoma grade 3–4 in one patient.

Stratified analysis

Among the 14 dead patients in the surgery sample, 10 were women, giving a considerably higher SMR for women (see Table 5). This has not been reported in previous studies (22–24), and we consider it most likely that the sex difference in mortality is a chance phenomenon (the number of deaths is small, and we have no other explanation for the finding).

We compared temporal lobe resections (TLRs) with resections outside the temporal lobes and with other types of operations (callosotomy, hemispherectomy). In the different types of surgical procedures, SMRs for all causes was within the same range, in keeping with previous studies (16,22,23) (see Table 5).

Hennessy et al. (17) followed up 299 patients after TLR and found no difference between surgery of the left compared with the right temporal lobe. However, when confining the comparison with patients in whom tissue pathology showed mesial temporal sclerosis (MTS), the SMR was 10 times and SUDEP incidence >7 times higher in right compared with left MTS, based on 13 deaths. In the present study, we compared side of operation in patients who had TLR and gliosis and found 2.5 times higher SMR in right- than in left-side operations. SUDEP incidence was 8.3 of 1,000 in right, but there were no SUDEP deaths among patients with left temporal lobe gliosis (see Table 5). Although our data seem to support the findings of Hennessy et al. (17), it is important to point out that our results are based on only four deaths, of which three were SUDEP, and therefore the data must be interpreted with caution. The explanation for this potential difference also is unclear.

We could not demonstrate any difference in the SMR for all causes of mortality or in SUDEP incidence between seizure-free surgery patients and those with recurrent seizures at the 2-year follow-up assessment (see Table 5). Sperling et al. (16) followed up 393 patients after epilepsy surgery and found that, of 199 patients who had seizure recurrence, 11 died (six patients in SUDEP), whereas no patient died among 194 who were seizure free. Sperling et al. were able to follow up all patients' seizure outcomes; as soon as a seizure occurred, the patient's status changed from “seizure free” to “recurrent seizures.” In the present study, we had access to information concerning seizure control 2 years after surgery, whereas the study period in some cases lasted several more years, during which time, seizures could recur. In such cases, the patients' data related to the seizure-free period would still contribute to the calculation of the seizure-free group, thus making our seizure-outcome data less accurate. Record review of our SUDEP cases revealed that one patient continued to have auras, and all other SUDEP patients had experienced seizure recurrence before the time of death, regardless of 2-year seizure-outcome status in the register.

Surgery versus nonsurgery patients

Nonsurgery patients in this study had an SMR for all causes of death of 7.9 and a SUDEP incidence of 6.3 per 1,000 person years. These rates are similar to those of other cohorts of epilepsy surgery candidates (10) (see Table 1) and seemingly higher than those of surgery patients in our study. One may speculate that this indicates that surgery decreases mortality in epilepsy. Vickrey et al. (24) compared epilepsy patients who were operated on with nonsurgery patients and found that surgery was associated with a lower mortality. However, as mentioned earlier, persons participating in epilepsy surgery assessment not leading to surgical treatment are not ideal controls for surgery patients. Moreover, the inclusion of nonsurgery patients in the Swedish epilepsy surgery register is not complete, and the selection is partly uncontrolled. However, scrutinizing the medical records of the dead nonsurgery patients revealed that the majority—three of five deaths, three of four SUDEPs—died during the evaluation period, thus before actually being classified as unsuitable for surgery.


This is the first follow-up study of mortality in an epilepsy surgery cohort covering all epilepsy surgery patients in one country in a recent period. Overall mortality and SUDEP incidence of the surgery patients were similar to those of other epilepsy surgery studies and in the lower range of corresponding data from cohorts of patients with refractory and chronic epilepsy. SUDEP was the dominating cause of death among our patients. However, despite the size of the cohort, the limited number of deaths complicated further subgroup analyses. Some of the results, such as the higher mortality of women compared with men, and the higher mortality of patients with right- compared with left-side temporal gliosis, must be interpreted with caution and might be spurious findings. With the methods used, we could not demonstrate any association between mortality rates and seizure outcomes 2 years after surgery. However, record review revealed that none of the SUDEP cases was seizure free at the time of death regardless of seizure outcome 2 years after surgery.

Patients undertaking epilepsy surgery evaluation that did not lead to an operation appeared to have a higher mortality rate than did patients who underwent therapeutic surgery. This might be an effect of the groups not being entirely comparable, but the influence of surgery treatment on mortality cannot be ruled out. Eighty percent of SUDEPs (60% of all deaths) in the nonsurgery group died during presurgical evaluation (i.e., before a decision on therapeutic surgery was made).

Acknowledgment: This work was supported by grants from the Carnegie and the Follin Foundations and the Karolinska Institute. We thank Drs. Kristina Malmgren and Bertil Rydenhag for valuable comments and the epilepsy centers participating in the Swedish Epilepsy Surgery Register: Lund University Hospital in Lund, Sahlgrenska University Hospital in Göteborg, Linköping University Hospital in Linköping, Karolinska Hospital in Stockholm, Uppsala University Hospital in Uppsala, and Umeå University Hospital in Umeå.