Population-based study of the incidence of sudden unexplained death in epilepsy

Sudden unexplained death in epilepsy (SUDEP) is a well-recognized entity in the literature and has been reported to be responsible for 7.5 to 17% of all deaths in epilepsy. Early studies of the causes of death in patients with epilepsy identified a group of patients who died without any apparent reason.^1-4 Later studies of epilepsy patients who died suddenly and unexpectedly reported certain risk factors and provided estimates of the incidence of SUDEP.^5-16 There is at least a tenfold difference in the reported incidence of SUDEP in these various studies (table 1). The range of incidence is associated with the composition of each cohort. A well-designed population-based study of the epidemiology of SUDEP has never been performed. It has been presumed, but never shown, that patients with epilepsy are at a higher risk than the general population of dying suddenly and unexpectedly.

The purpose of this study is to determine the population-based incidence of SUDEP and to determine whether the risk of sudden death is increased in the epilepsy population. To our knowledge, ours is the only study to contrast directly the rates of sudden death in persons with epilepsy with those in the general population.

Methods. Population characteristics. Rochester, MN, is an urban community that has experienced a population increase from approximately 25,000 in 1940 to approximately 71,000 in 1990. Approximately 95% of the population receives some or all of their care at the Mayo Medical Center. Each year, more than half of the population is seen at the Mayo Clinic or one of its affiliated facilities. All autopsies in the community, including those under the jurisdiction of the coroner, are performed by pathologists from the Mayo Clinic. The annual autopsy rate in the community varied during the study period. It was higher between 1950 and 1970, with rates fluctuating between 55 and 75%. The rate was generally 45% before 1950, and has declined since 1970 to approximately 40% in 1984.²⁴

Source of data. Since the early 1900s, our institution has maintained a unified medical record for each patient seen at the Mayo Clinic. Each record documents all aspects of the patient's medical care, including outpatient visits, hospital visits, laboratory and radiologic results, pathologic results, and medical correspondence. Each diagnosis is coded to allow efficient identification and retrieval of records. In addition, the causes of death are coded, and death certificates and autopsy results, if applicable, are maintained in the patient's record. This study was approved by the institutional review board.

Cohort identification and follow-up. A database of epilepsy patients is maintained through our ongoing study of the epidemiology of seizure disorders in the Rochester population. Subjects from the database were eligible for this study if their epilepsy was diagnosed during the period of January 1, 1935, to December 31, 1994, while residing in Rochester. These subjects were followed until the earliest occurrence of any of the following: death, migration out of Rochester, last contact, or June 30, 1996.

Terminology. Determination of SUDEP was similar to that used in a study of SUDEP in clinical drug trial subjects.²⁰ Definite SUDEP was defined when all of the following criteria were met: 1) history of epilepsy; 2) death or cardiorespiratory compromise was sudden, and not due to status epilepticus; 3) death was unexpected (no life-threatening conditions); and 4) death remains unexplained after review of all evidence, including autopsy. Probable SUDEP was defined by criteria 1, 2, and 3, but data are insufficient because of lack of autopsy, and no alternative explanation for death exists. Possible SUDEP was defined by criteria 1, 2, and 3, but an alternative explanation of death is present. Patients that did not meet these criteria were classified as non-SUDEP deaths. Patients with insufficient data were termed as insufficient data (ID). Terminology regarding epileptic seizure disorders was according to that established in earlier publications on the epidemiology of epilepsy in Rochester.^25,26 Classification of seizure type was according to the International Classification of Epileptic Seizures.²⁷

Study data and method of collection. All of the causes of death codes were reviewed in the Rochester epilepsy database. Any patient whose listed cause of death was neurologic (i.e., seizure-related, status epilepticus, cerebrovascular), cardiac (i.e., myocardial infarction, cardiac arrhythmia, coronary artery disease), pulmonary (i.e., pulmonary embolus, asphyxiation), accidental, suicide, or unknown was reviewed independently by three of the authors (D.M.F., E.L.S., and J.F.A.) and assigned a classification of death (i.e., definite, probable, possible, non-SUDEP, or ID). If the investigators disagreed on classification, then a majority vote was used to determine classification. A profile of the circumstances at the time of death was reconstructed by noting the following: time, location, activity, and awake versus asleep. The histories of all patients ascertained as SUDEP were reviewed for the following potential risk factors: age, gender, seizure type, frequency of seizures, etiology of epilepsy, presence of brain lesion, recent antiepileptic drug (AED) use and substance abuse, and most recent serum concentrations.

Statistical analysis. The incidence of SUDEP was derived from the total number of definite and probable SUDEP patients and the total person-year follow-up for the whole epilepsy cohort. A separate study²⁸ at our institution had determined the basal rate of sudden, unexplained death in 20- to 40-year-old persons between 1960 and 1989 in the Olmsted County general population, where Rochester, MN, is located. The standardized mortality ratio (SMR) of sudden death in the epileptic population was determined by comparing the basal rate with the observed rate in epilepsy patients in the same age group (20 to 40 years) and year of death (between 1960 and 1989).

Results. SUDEP patients. A total of 1,535 patients in our epilepsy cohort contributed to 25,939.70 person-years of follow-up. There were 535 deaths in this cohort. Review of the details of death revealed seven cases of definite SUDEP and two cases of probable SUDEP. In addition, one death was classified as possible SUDEP and six patients had ID for classifying death. Thus, 1.7% (9 of 535) of the deaths in the entire cohort were due to SUDEP (definite and probable cases). However, SUDEP accounted for 8.6% (7 of 81) of the deaths in the age group 15 to 44 years.

Incidence of SUDEP. The incidence of SUDEP in persons with epilepsy in Rochester, MN, was 0.35 per 1,000 person-years of follow-up. Table 2 shows the incidence of SUDEP by gender and age groups.

Clinical and pathologic details of SUDEP patients. Average age at death was 31.6 years (range, 7 to 49 years). Eight of the nine patients were found dead at home. Two patients were found dead in bed, two died in a chair, and five were found elsewhere in the house. Only two of the nine patients had a seizure witnessed before being found dead.

Autopsy was performed in seven patients. Three had a brain lesion: partial microcephaly, cerebral infarction, and a small hippocampal low-grade astrocytoma. Four patients had postmortem AED levels performed, which were either subtherapeutic or absent.

Seven patients had either an idiopathic or cryptogenic cause of epilepsy and two patients had epilepsy associated with developmental CNS abnormalities. Four patients had primary generalized epilepsy and five had partial epilepsy (information regarding anatomic localization was not available). Eight of the nine SUDEP patients had generalized tonic-clonic (GTC) seizures. Seizure frequency in the 6 months preceding death was unknown in three patients. One patient was seizure free, three patients had monthly seizures, and two patients had daily seizures. No patient underwent surgical treatment for epilepsy. Three patients had an active history of alcohol use (two of these patients also had a prior history of substance abuse).

EEGs were performed in all patients. Abnormalities were generalized epileptiform in one, focal epileptiform in one, and multifocal epileptiform abnormalities in three patients. Four patients did not have epileptiform abnormalities on EEG.

Five patients had either brain CT or MRI. These were normal in four patients, and one patient had generalized atrophy on imaging.

SMR ratio for SUDEP. Between 1960 and 1989 there were five SUDEP deaths in persons 20 to 40 years of age (one man and four women). All five underwent autopsy. Table 3 shows the SMR for sudden unexplained death in persons with epilepsy compared with the expected rates in the general population.²⁸ The risk for SUDEP persons is nearly 24 times higher than in nonepilepsy persons.

Discussion. Retrospective studies^5-8 have identified patients, usually from coroners' case files, who died suddenly or unexpectedly. Autopsies usually did not reveal a cause of death. Generally, patients were found dead in bed. Patients were typically young men, had GTC seizures, and had low or absent AED levels. Other studies also suggested that low AED levels may be a risk factor for SUDEP.^10,11 One study⁷ reported that most epilepsy patients whose deaths were sudden and unexplained were black men or had abused alcohol, but this may reflect the population studied (Cook County). The incidence rate was estimated to be between 0.5 to 1.9 per 1,000 person-years, but this rate was based on an estimation of the incidence of epilepsy in the community.

Leestma et al.¹² identified prospectively 60 epilepsy patients who over a 1-year period died suddenly without any demonstrable cause. Autopsies often revealed heavier than normal weights of the heart, liver, and lungs. Patients were often black men with an average age of 35 years, with infrequent GTC seizures, some of whom had brain lesions that were responsible for their seizures. AED levels were often low or absent, and there was often evidence of alcohol abuse. The attributes of these 60 patients likely reflect those of the population in Cook County, which is a large, urban industrial community. The investigators estimated the incidence of SUDEP to be 0.9 to 2.7 per 1,000 person-years.¹² The basal prevalence of epilepsy in Cook County has never been determined. Hence, only an estimation was used to derive the incidence of SUDEP reported by the authors. Moreover, the clinical history of these 60 patients was often incomplete and second-hand. The accuracy of the information regarding seizure type and frequency could neither be verified nor vouched. Autopsy-based coroner studies may overestimate the true incidence of SUDEP because the decision for autopsy is influenced by circumstances surrounding death, and a higher proportion of unexplained deaths may be autopsied.

Many studies have evaluated the incidence of sudden death in highly select cohorts of patients with epilepsy (see table 1). Some studies reported the incidence of SUDEP in patients seen at a tertiary care center^9,14,23 or in patients who are institutionalized for severe epilepsy or other disabilities.^13,14,17 Others have determined the rate of SUDEP in specific clinical epilepsy drug trials.^19-22 These findings are not applicable to persons with epilepsy in the general population, but the higher SUDEP rates they reported suggest that patients with epilepsy that is more difficult to control may be at higher risk for SUDEP.

Tennis et al.¹⁵ conducted a population-based study and found the incidence of SUDEP to be 0.54 to 1.35 per 1,000 person-years. However, the study at the outset was flawed in its method of determining eligibility for the study. Persons in the community were presumed to have epilepsy if they were taking AEDs. They were identified via computerized pharmacy records of AED prescriptions. Therefore, persons taking AEDs for nonepileptic conditions may have been included in the study, whereas epileptic patients who were noncompliant with medications may have been excluded. In fact, the authors admitted that 8.6% of the cohort may not have epilepsy. Another study¹⁶ used a similar method of identification via computerized pharmacy records. Derby et al.¹⁸ also used prescription drug files to identify patients with "refractory epilepsy" (using two or more AEDs), and determined a slightly higher SUDEP incidence than had been reported previously in population-based studies. Again, this study also presumes that patients taking AEDs had epilepsy. In addition, no information was provided regarding seizure frequency to determine if the patients were truly refractory.

Annegers et al.²⁹ evaluated the risk for sudden cardiac death in the Rochester epilepsy population, but autopsy was not required and no set criteria for SUDEP were used. Sudden death was defined in patients without prior ischemic heart disease who died within 24 hours of onset of symptoms suggestive of acute coronary insufficiency. This definition included patients with sufficient autopsy evidence of significant coronary artery disease to explain death that in the current study would have been classified as non-SUDEP. Thus, the incidence of sudden, unexplained death in the epilepsy population was likely overestimated in this study.

To our knowledge, our study is the first well-designed, population-based study of the incidence of SUDEP. The incidence of SUDEP in the Rochester, MN, epilepsy population is 0.35 per 1,000 person-years for all ages, and 0.5 per 1,000 person-years for patients age 15 to 44 years. We were able to ascertain all epilepsy patients in the Rochester population because of the unified record system and computerized coding of each patient's diagnoses. Thus our results likely represent the true incidence of SUDEP in the general epilepsy population. The low rate of SUDEP in our study may be due to the use of a population-based cohort, rather than groups of patients in referral practice or in clinical drug trials.

It has been presumed that the risk for sudden death is increased in the epilepsy population by comparing the results of SUDEP studies indirectly in selected populations with the published risks for sudden death in the general population. However, no study has compared the two rates in the same population. We determined that the overall risk for sudden, unexplained death in persons with epilepsy markedly exceeds that of the general population by approximately 24 times.

We initially intended to perform regression analyses to determine risk factors for SUDEP. However, the low total number of SUDEP patients made such analyses difficult. Most of our SUDEP patients experienced GTC seizures, as noted in other studies.^5-8,12 Postmortem AED levels were absent or subtherapeutic in our subjects, as suggested in other studies.^10,11

In our study the magnitude of the relative risk of SUDEP was based on the incidence of the condition in young adults. The relative risk in younger and older age groups could not be determined because the general population rates for sudden death in these age groups are unavailable. However, the relative risk determined in our study is most relevant to persons with epilepsy because our study, as well as others,^{8,9,12-14,16,30,31} shows that SUDEP is most often observed during young adulthood. As our study has revealed, close to one of every 10 deaths in young adults age 15 to 44 years was due to SUDEP, whereas only one in approximately 60 deaths in the whole epilepsy population was attributable to SUDEP. Nonetheless, we cannot exclude the possibility that this observation is in part the result of a greater likelihood of attributing sudden deaths in older epilepsy patients to ischemic heart diseases.³²

Theoretically, the incidence and the relative risk of SUDEP can only be ascertained fully if all sudden deaths in epilepsy persons and in the general population had postmortem examinations. This expectation is unlikely ever to be met, even in a prospective cohort study of epilepsy patients.³³ To minimize the misclassification of deaths, autopsy should remain a standard requirement in determining the risk of sudden death in epilepsy persons relative to the general population. The autopsy rates in our community during the study period were much higher than the national average rate of 12%.²⁴