Incidence and risk factors in sudden unexpected death in epilepsy

Methods.

We prospectively enrolled patients evaluated at three upper midwestern epilepsy centers (MINCEP Epilepsy Care, Minneapolis, MN; Mayo Clinic Epilepsy Division, Rochester, MN; and Marshfield Clinic Epilepsy Section, Marshfield, WI). After consent was obtained, demographic information, seizure type, and current anticonvulsants were recorded. The great majority of patients underwent extensive evaluation (see below). A surveillance system was set up to identify deaths in this prevalence cohort. All reported deaths were investigated as soon as possible after death in a systematic manner designed to distinguish between SUDEP and other causes of death. Investigation included interviews with relatives and witnesses, and review of medical records, death certificates, and autopsy reports when available. Unreported deaths were detected and reported deaths confirmed by comparing all enrolled patients against the Social Security Death Index (Epidemiology Resources Inc., Newton Falls, MA) annually. This index lists all deaths reported to the US Social Security Administration.

A committee, which included investigators at each center one other expert in SUDEP, evaluated all deaths. Committee members independently assessed and classified each death using previously reported criteria.5 Briefly, definite SUDEP required the following; 1) a history of epilepsy (more than one epileptic seizure during a period of less than 5 years); 2) that the death occur suddenly; 3) that the death was unexpected (i.e., there was no life-threatening illness); and 4) that the death remained unexplained after all investigative efforts, including autopsy. Probable SUDEP required criteria 1 through 3. Autopsy was not available in these patients, but SUDEP was considered likely because there was no alternative explanation for death. Possible SUDEP required criteria 1 though 3 above. However, an explanation for death other than SUDEP was found. In these patients, committee members could not establish with certainty whether SUDEP or the other possible cause was responsible for the death. In non-SUDEP cases, the cause of death was thought due to documented injury or medical illness. In insufficient data cases, information allowing classification into the above categories was insufficient. Agreement among committee members after an initial independent review was excellent. In 13/20 cases all four observers agreed on the ultimate classification, in 4/20 cases three of four observers agreed on the ultimate classification, and in the remaining 3/20 cases two of four observers agreed on the ultimate classification. Differences were easily resolved after open discussion by the four committee members.

We analyzed all deaths occurring in this cohort between the first enrollment (June 1, 1991) and December 31, 1996. Definite and probable SUDEP in this group of deaths were determined and grouped together. Four controls were randomly selected for each SUDEP case from among the patients enrolled during the same month at the same epilepsy center. Potential risk factors in SUDEP cases and controls were identified by chart review.

We studied the following risk factors: age at enrollment, gender, number of seizures of any sort in the year preceding the last visit (calculated as seizures per month), number of tonic-clonic seizures in the year preceding the last visit, and duration of epilepsy (age at onset to age at enrollment). Mental retardation was defined as Full-Scale Wechsler Intelligence Scale Quotient of less than 70. Patients who had been severely impaired since birth, unable to communicate, and had not been considered testable were also classified as mentally retarded. Three SUDEP cases and four control cases met these criteria and were included with patients that had a measured IQ of less than 70. Epileptogenic structural lesions were defined as focal areas of cortical abnormality that were sufficient to cause seizures and were found on brain CT or MRI or at autopsy. Anticonvulsant compliance was evaluated by reviewing anticonvulsant serum levels at the last visit. Many patients were treated with more than one anticonvulsant. Results were classified as all levels therapeutic, at least one level therapeutic, or no levels therapeutic.

The occurrence of risk factors in SUDEP patients and controls was compared. We used logistic regression analysis with logistic procedure in SAS (version 7.0, SAS Institute, Cary, NC) to obtain the maximum likelihood estimates of the OR for duration of epilepsy, Full-Scale IQ, presence of epileptogenic lesions, use of psychotropic drugs, seizure frequency, and number of generalized tonic-clonic seizures. Using the same statistic, the OR estimates were adjusted for seizure frequency, number of generalized tonic-clonic seizures, and number of antiepileptic medications when indicated. Age- and gender-specific incidences were calculated by comparing SUDEP cases to the remainder of the cohort.

Results.

Discussion.

Our study identified three apparently independent risk factors for SUDEP: 1) exposure to tonic-clonic seizures, 2) mental retardation, and 3) the number of anticonvulsant drugs used. Nillson et al.4 reported that a higher seizure frequency increases risk of SUDEP; however, the impact of tonic-clonic seizures was not distinguished from that of other seizure types. Our study also found that increased seizure frequency is associated with SUDEP. However the risk appeared to be associated with tonic-clonic seizures rather than with seizures of any type. This finding strongly supports the idea that tonic-clonic seizures are an important proximate cause of SUDEP. This concept is further supported by three lines of evidence. The great majority of reported cases of SUDEP experienced tonic-clonic seizures before death.9,11-13,15,16⇓⇓⇓⇓⇓ The majority of SUDEP cases in which circumstances of death are carefully examined show evidence of tonic-clonic seizure preceding death.11-13,15,17⇓⇓⇓⇓ Finally, death has been directly related to generalized convulsive seizures in an animal model of SUDEP.18,19⇓

Our study has clearly demonstrated mental retardation to be a risk factor for SUDEP, even after accounting for seizure frequency. Previous reports have provided only weak evidence for the association between SUDEP and mental retardation. Mental retardation was thought to be more common in medical examiners’ series of SUDEP.11,13⇓ Two large, population-based studies did not find mental retardation to be a risk factor.3,20⇓ One found a trend toward increased SUDEP risk.21 Mental retardation was determined by observer impressions rather than by IQ examination in all previous studies.

Increased SUDEP risk in mentally retarded patients can be explained in terms of one theory of the pathophysiology of SUDEP. The animal model of SUDEP,18,19⇓ clinical observations,11,17,22,23⇓⇓⇓ and video-EEG incidentally recorded during SUDEP24,25⇓ suggest that a combination of postictal central apnea, prone position, and neurogenic pulmonary edema causes death in this entity. There is some evidence that mentally retarded patients are more likely to experience prolonged postictal encephalopathy,26 which may decrease postictal respiratory drive. Furthermore, the neurologic abnormalities often associated with mental retardation may preclude movement and righting reflexes if the patient is prone or supine following a generalized tonic-clonic seizure. For both of these reasons, mentally retarded patients may be more susceptible to the postictal central apnea and positional asphyxia that may cause SUDEP. This idea is supported by observations at a boarding school for epilepsy and learning difficulty. No SUDEP cases occurred while patients were on the school premises, where epileptic seizures initiated immediate treatment and appropriate positioning. All SUDEP cases occurred while pupils were off the school premises, and usually while they were alone.22

A previous study3 reported that simultaneous treatment with more than one anticonvulsant drug was associated with SUDEP and concluded that the number of anticonvulsants used was a surrogate for the severity of epilepsy. However a more recent study4 and this report have found that treatment with multiple anticonvulsants is a risk factor even after adjustment for seizure frequency. Uncontrolled reports27 have suggested that carbamazepine may play a role in SUDEP; however, neither our study nor the other controlled study of SUDEP4 have implicated any particular drug or drug combination. Treatment with multiple anticonvulsants is likely to cause sedation. Pharmacologically induced sedation may prolong recovery from the postictal state and cause greater susceptibility to the postictal central apnea and positional asphyxia that may play a role in SUDEP. Determining whether postictal apnea is more common or is of longer duration with mental retardation or following treatment with multiple anticonvulsants would provide information to support this idea. Studies based at video-EEG monitoring units have found that postictal central apnea is common.28,29⇓ However, whether mental retardation or treatment with multiple anticonvulsants increases the prevalence or severity of postictal apnea is unknown.

We did not find male gender to be associated with SUDEP. In fact, SUDEP appeared to be more common in females and the SUDEP risk associated with increased seizure frequency was significant only in females. One previous study9 also found SUDEP-related mortality to be elevated only in females. Another3 found incidence to be increased in males, though the diagnosis of epilepsy in this study was based on anticonvulsant use and may not have been accurate. A third study4 found increased seizure frequency to be a more powerful risk factor for SUDEP in males, though other risk factors were more powerful in females. This last study was not designed to assess sex as a risk factor for SUDEP. Our analysis was based on a firm diagnosis of epilepsy in SUDEP cases and complete information regarding age and sex for the entire cohort. Though the issue is not settled, the weight of the evidence argues that neither sex nor any particular age group is a strong risk factor for SUDEP.

The idea that cerebral structural lesions are a risk factor for SUDEP is based on medical examiners’ case series. These observations were not controlled and probably reflect selection bias. The current study is the only properly controlled assessment of structural lesions in SUDEP and did not find an increased frequency of structural lesions in this entity.

Speculations that patients with SUDEP are less compliant with anticonvulsant treatment began with the observation that postmortem anticonvulsant levels in medical examiners’ case series of SUDEP were frequently subtherapeutic.10-12⇓⇓ Again, these were uncontrolled observations and selection bias may have played a substantial role. One controlled study of postmortem anticonvulsant levels30 found that subtherapeutic levels were significantly more common in SUDEP cases than in epilepsy patients dying of other causes, but another study14 found no difference. Multiple factors such as redistribution and continuing metabolism can affect postmortem anticonvulsant levels and it is not clear whether levels obtained at various times after death accurately reflect levels shortly before death.31 Until this study, no reports of antemortem anticonvulsant levels in SUDEP were available. The great majority of both SUDEP cases and controls in this study had therapeutic anticonvulsant levels at the last visit. Thus, noncompliance appears not to be a risk factor, at least in this cohort of closely followed patients. Analyzing all available anticonvulsant blood levels and their variation over serial visits would further support this conclusion. Compliance with treatment clearly influences the frequency of tonic-clonic seizures, which we found to be a strong risk factor for SUDEP. Therefore, compliance may prevent SUDEP in an epilepsy population that is less closely followed and should be encouraged. However, the evaluation of noncompliance as a risk factor for SUDEP must take into account the finding that the frequency of tonic-clonic seizures is also a risk factor.

This study has several limitations. Only half of the cases underwent complete autopsy. However, we used a consistent definition of SUDEP, multiple reviewers independently analyzed information for each death, and interobserver agreement was high. Secondly, the number of SUDEP cases was relatively low. Although our results are statistically valid, internally consistent, and consistent with other reports, the small number of cases did not provide adequate power to detect less powerful risk factors for SUDEP. This may explain, for example, why we did not find psychotropic drug use to be a risk factor for SUDEP. Finally, we studied patients presenting to epilepsy centers. Such patients may have more severe epilepsy than that encountered in a general clinical practice.

Information regarding risk factors for SUDEP remains limited and further controlled studies are needed. However, the results reported by Nillson et al.4 and this report are consistent and help to define a population at increased risk for SUDEP. We believe that it is reasonable to discuss the possibility of SUDEP with patients at high risk. This potentially distressing topic should be discussed in a sensitive manner taking into account the individual situation. It is important to emphasize that the great majority of patients tolerate the great majority of seizures without much difficulty. The discussion of SUDEP can be used to emphasize the serious nature of severe epilepsy and the need for regular follow-up and compliance so that seizures can be brought under the best possible control.

Unfortunately, the current state of information does not allow firm recommendations for preventing SUDEP. Certainly, every effort should be made to prevent tonic-clonic seizures with the fewest anticonvulsants possible. The possibility of SUDEP can be considered when epilepsy surgery is being contemplated. Epilepsy surgery appears to decrease mortality,8 though it is not yet clear that surgery decreases the rate of SUDEP. Ideally, caregivers should be able to deliver appropriate first aid after epileptic seizures, position the patient properly, and assure that the airway remains open. Vigorous stimulation after a tonic-clonic seizure may be reasonable because some believe that this decreases the duration of postictal apnea. It is also reasonable to encourage epilepsy patients to sleep in the supine rather than the prone position. It is not clear that these practices will prevent SUDEP, but they may be reasonable measures to suggest when discussing this issue with patients.