Chapter 15 Sudden Unexpected Death in Epilepsy
Patients with epilepsy have an increased mortality rate compared to the general population, to which sudden unexpected death in epilepsy (SUDEP) is a major contributor. Some groups of patients are at greater risk of SUDEP than others, although the reasons for this are only partly understood. A large number of risk factors have been proposed, and there are numerous theories as to the pathophysiological basis of SUDEP. This chapter explores the evidence behind the “recognized” risk factors to determine who is most at risk of SUDEP and examines theories surrounding proposed mechanisms of SUDEP and how these mechanisms are integrated with substantiated risk factor data. Areas of possible future research are also discussed.
All-cause mortality rates in patients with epilepsy are approximately two to three times higher than the general population and are age dependent.1–3 The major contributors of increased mortality are the causes of epilepsy, for example, traumatic brain injury, cerebrovascular disease, brain tumors, and accidents (mostly falls and drowning) or status epilepticus and SUDEP. In contrast, treatment-related mortality, from epilepsy surgery or the adverse effects of antiepileptic medication, is very rare.4
SUDEP is defined as the sudden, unexpected, witnessed or unwitnessed, nontraumatic, and nondrowning death in patients with epilepsy with or without evidence for a seizure and excluding documented status epilepticus, in which postmortem examination does not reveal a structural or toxicologic cause for death.5 Where autopsy is not performed, and for the purpose of epidemiological studies, sudden death occurring in benign circumstances with no known competing cause for death is classified as “probable SUDEP.” Cases classified as “possible SUDEP” are usually not included in epidemiological studies of SUDEP incidence. These are cases in which SUDEP cannot be excluded because either the information regarding circumstances of death is too limited to confirm classification as a probable or definite SUDEP or that adequate or complete documentation is available but there is a plausible competing explanation for the death.3 It must be stressed that this defines a category of sudden death in epilepsy, and not a condition, a concept that is sometimes overlooked. Despite an applicable definition, and clear guidance where there is uncertainty, significant variability in use has hampered efforts to integrate findings from multiple studies on epidemiological and risk factor data and hence establish common relevant factors.6,7
Sudden unexpected death in the general population is extremely rare in young adults with an incidence of 5 to 10/100, 000 person-years, whereas the rate climbs steeply with advancing age to approximately 300/100, 000 person-years in the elderly.8 The incidence of sudden death in patients with epilepsy is significantly higher and varies markedly with the population studied.9 For example, in population-based studies, the incidence has been reported to be 0.35 and 2.7/1000 person-years, depending on the methodologies employed.10,11 This increases to between 2 and 5.9/1000 person-years in cohorts of patients attending specialist epilepsy clinics,12–14 3.4/1000 person-years in pupils with epilepsy enrolled in a special residential school,15 and up to 9.3/1000 person-years in epilepsy surgery candidates.16,17 The incidence of sudden death in young adults with intractable epilepsy is, therefore, many times that of the general population, with a peak between the ages of 20 and 40 years.18 In older age groups, the relative increased incidence of SUDEP is too small to measure and is confounded by the occurrence of comorbidity such as cardiovascular, respiratory, or cerebrovascular disease. This makes it difficult to ascribe a sudden death to a “pure” SUDEP category but does not exclude the possibility that there is no additive risk from the epilepsy. There is limited data available on the incidence of SUDEP in children. An incidence rate of approximately 0.1 to 0.2/1000 person-years has been estimated from a number of earlier studies.19,20 Significant methodological limitations exist, however, including inadequate follow-up periods, difficulties with case-ascertainment, and assumptions regarding the prevalence of epilepsy in community-based studies. A more recent retrospective study with an 18-year follow-up period confirmed a low incidence rate of approximately 0.4/1000 person-years, despite following a cohort of children with refractory epilepsy and learning disability.21 This is approximately 10 times lower than the incidence rate for similarly affected adults; the rate for children with uncomplicated epilepsy will be lower still.22 The reason for the discrepancy between children and adults is unknown and is unlikely to be entirely due to methodological constraints, but may be as a result of a different cardiorespiratory response from a developing brain compared to a mature brain, a more intensively supervised environment, the inclusion of specific self-limiting pediatric epilepsy syndromes, lack of comorbidity, or possibly, greater autonomic stability.23
There is significant debate regarding risk factors for SUDEP. A large number of variables that may influence the risk of SUDEP have been proposed, and the significance of each has been discussed at length without clear consensus. Relevant and independent risk factors are difficult to establish given the nonindependence of patient, syndrome, seizures, and treatment characteristics. Multiple logistic regression analyses require large cohorts of patients to achieve statistical significance for each of the variables evaluated, and this is difficult to attain.24 Furthermore, the high variability between studies in terms of patient cohorts, definition, choice of control group, methodology, and overall study quality precludes not only a valid meta-analysis, but even a simple meaningful comparison. In a critical review of the literature regarding risk factors for SUDEP, Tellez-Zenteno noted that a clear definition of SUDEP was stated in only 65% of relevant studies. Furthermore, the low frequency of postmortem data was evident in many studies. This lack of consistency significantly undermines the suitability of such studies to comment on SUDEP causality and risk.7 The preliminary studies of risk factors in SUDEP were descriptive, providing direction for more meaningful case-control studies. The value of these observations, otherwise, is limited, however, because, in general, these studies lack suitable control groups and comprise small numbers of an often highly selected patient population.
In case-control studies, the choice of control group is important and is dependent on the risk factors to be evaluated. A control group of matched living patients with epilepsy tends to favor the assessment of risk factors exploring lifestyle and clinical issues such as seizure frequency and antiepileptic drug use. A control group of non-SUDEP deaths in epilepsy is more suitable for exploring circumstances of death, such as body position, place of death, and seizures immediately prior to death.
A number of studies assessing risk factors used a case-control design (Table 15-1), this often being the only option for the study of relatively uncommon occurrences, such as SUDEP. However, it also has important limitations. For example, case-control studies traditionally evaluate the probability of being exposed to a risk factor, not the risk of developing a condition. Moreover, case-control studies are subject to several sources of bias, such as selection bias of cases and controls and information and recall bias when collecting data on exposure. These biases are particularly important where an objective definition of the condition of interest is not consistently applied, as is the case in the SUDEP literature.
An attempt has been made to resolve the SUDEP risk factor literature and disentangle the often contradictory results on the basis of extensive literature reviews and the implementation of a study-validity scoring system and calculation of relative risk factor ratios.6 Monte and colleagues stratified studies evaluating risk factors on the basis of fulfillment of a number of variables that were considered to be markers of good quality studies. However, the scoring system used was unvalidated and potentially misleading, with, for example, equal credence attributed to controlled and descriptive studies. Papers not achieving an arbitrary threshold were excluded, and the possibility of the same study population being used more than once was not adequately addressed.6 A more simplistic approach was used by Tellez-Zenteno, in which percentages of studies with a positive risk factor were reported. Risk factors that were reported to be significant in more than 50% of studies were a terminal seizure, subtherapeutic antiepileptic drug (AED) levels, high seizure frequency, and a high number of AEDs.7
Descriptive studies have almost universally reported that patients with SUDEP are young adults.10,11,13,14,25–29 A number of biases exist, however, including, as discussed earlier, the exclusion of patients with significant comorbidity associated with increasing age, such as ischemic heart disease or cerebrovascular disease, identified on postmortem examination.13,25,29 Other examples of bias include case identification through self-referral by bereaved relatives, most commonly parents,27 and studies with only small numbers of patients.10,13 Case-control studies are less conclusive. Some studies only included defined age groups and can draw no conclusions regarding other age groups. Nevertheless, it is interesting to note that 70 to 80% of the studied population in a number of case-control studies were less than 45 years old.18,30 Data regarding age, however, is not available from a number of large studies due to age matching of control subjects.18,30,31
Of the remaining studies, the use of a cohort of non-SUDEP deaths as a control group may bias the patient group toward a younger age due to exclusion of comorbid conditions more commonly associated with advancing age, such as atherosclerosis, thromboembolism, and metastatic carcinoma,32–35 although young age as an independent risk factor has not been universally reported.36,37 The likelihood of selection bias is corroborated by finding significantly less comorbidity in the SUDEP group than the non-SUDEP group.35 In studies using living control subjects, younger age was not seen more frequently in the SUDEP group, although numbers of SUDEP patients were small.12
Although a large number of descriptive studies have suggested that male gender is a significant risk factor for SUDEP,11,13,26,29,38 this has not been confirmed by the vast majority of case-control studies.30–33,35–37,39 In addition, a small number of both descriptive and case-control studies have reported a significantly increased standardized mortality rate in female patients, which may be attributable to a lower background rate of death in the female non-SUDEP control group.10,34
In summary, although not universally accepted, the weight of evidence at the present time suggests that sex is not a strong risk factor for SUDEP. With regard to age, incidence appears lower in children than young adults. Methodological constraints, selective definition, coexisting pathology, the absence of suitable control subjects, and lack of controlled prospective data do not allow any conclusions to be drawn regarding incidence of sudden death related to epilepsy in older age groups.
A number of case-control studies have suggested that early onset of epilepsy is a significant risk factor for SUDEP.30,33,35,36 For example, Nilsson reported an eightfold higher SUDEP risk in patients with an onset of epilepsy between the ages of 0 and 15 years, compared to patients with seizure onset after 45 years of age.30 However, whereas this may reflect a different etiological basis for the epilepsy, it may also merely be a surrogate marker for an increased cumulative lifetime risk of having seizures for a longer period of time, as suggested by other studies.18,28,29 Conversely, several reports give a shorter duration of epilepsy being associated with an increased risk of SUDEP, although this is most likely as a result of comparison with an older control population.32,33,35 Furthermore, following conditional multiple logistic regression analysis, Walczak showed that a long duration of epilepsy (>30 years) was no longer a risk factor after adjustment for seizure frequency.37
One would expect epilepsy syndrome to be a key factor in defining the risk of SUDEP. Yet, there is only limited evidence to support the association of epilepsy syndrome with an increased risk of SUDEP.30,36 Discordant results from the relatively few case-control studies that assessed this risk factor and low numbers of patients in each group preclude detailed evaluation or definitive conclusions.34 In the study reported by Nilsson, 7 out of 57 (12%) SUDEP cases had primary generalized epilepsy compared to 12 out of 171 (7%) control subjects. Statistical comparison revealed that there was a higher risk of SUDEP in patients with primary generalized epilepsy compared to patients with focal, symptomatic epilepsy, although this was only significant in men.30 Nevertheless, although idiopathic primary generalized epilepsy (IGE) is usually less refractory to treatment, individuals with IGE are well represented in SUDEP cohorts. It is possible that specific epilepsy syndrome subtypes carry an increased risk of sudden death due to phenotypic expression in other cerebral and possibly cardiac structures. For example, Rett syndrome, typically due to mutations in the MECP2 gene, is associated with brain stem immaturity and autonomic, particularly respiratory, instability in association with epilepsy. In addition to respiratory abnormalities, which include apneustic breathing and hyperventilation, patients with Rett syndrome may also present with a prolonged QT interval and reduced heart rate variability.40 The coexistence of cardiac arrhythmia and central apnea may act synergistically in the development of sudden death. Severe myoclonic epilepsy in infancy (SMEI) also appears to be associated with an increased risk of sudden death, although the pathophysiological mechanism remains unclear. A recent report of a pedigree with an SCN1A mutation and two cases of SUDEP is of interest in this respect.42 Although some of these conditions represent the most severely affected group of patients and the high mortality rate may be multifactorial, it is possible that more subtle genetic abnormalities, such as channelopathies in, for example, IGE, may also predispose patients to SUDEP. In this regard, it is important to note that a number of functional cardiac conduction abnormalities, such as long QT syndrome, are also channelopathies, and genetic susceptibility is also well documented in sinus node dysfunction and bradyarrhythmias.43 However, no epidemiologic data indicate a higher incidence of epilepsy among relatives of patients with inherited susceptibility to arrhythmia, and a family history of early sudden cardiac death is not reported in SUDEP series. Furthermore, malignant tachyarrhythmias are relatively uncommon in seizures, and ictal respiratory changes have been documented in the absence of cardiac abnormalities. The proposed cardiorespiratory mechanisms of SUDEP will be discussed in more detail later in the chapter. Clearly, scope exists for potentially useful epidemiologic studies looking at the incidence of epilepsy in families of sudden cardiac death victims; additionally, it may be helpful to study the incidence of syncope in patients with idiopathic epilepsy and their relatives.
Controversy exists on whether high seizure frequency is an independent risk factor for SUDEP. Several descriptive and large case-control studies have reported an increased risk of SUDEP in patients experiencing frequent seizures.28,30,31,33,37,38 This increased risk is most marked for convulsive seizures10–12,27,28,31,37 rather than nonconvulsive episodes, such as complex partial seizures.33 Moreover, on logistic regression analysis, Walczak noted that only the frequency of convulsive seizures was relevant, and not the frequency of all seizures combined.37 Conversely, high seizure frequency was not an independent risk factor in a number of other reports, although a number of methodological issues exist.12,29,34,36 For example, in a retrospective case-control study of 42 patients with SUDEP, reported by Kloster and Engelskjon, there was no reported difference in seizure frequency between the SUDEP and non-SUDEP control group. The study was undertaken at a tertiary referral center with both groups having chronic refractory epilepsy and frequent seizures.36 Other negative studies may have been similarly influenced.34 Intuitively, the severity of convulsive seizures may also be important in SUDEP, but this is more challenging to quantify and hence has not been evaluated as a risk factor.
In summary, with regard to epilepsy characteristics, the most important factor, on the basis of currently available data, is the frequency of convulsive seizures. Data with respect to seizure severity does not exist at the present time, whereas data regarding the influence of the pathogenesis of the underlying epilepsy and syndromic diagnosis are also limited to a small number of specific syndromes and requires further study.
The number of antiepileptic medications (AEDs) taken concomitantly have been reported to be an independent risk factor for SUDEP,38 even after correction for seizure frequency.30,37 This is not universally reported, however,12,18,34–36 although small numbers of patients and a high frequency of polytherapy in control subjects may be contributory in these negative studies. Langan found that the risk of SUDEP increased with the number of AEDs previously taken, despite correction for seizure frequency, perhaps a surrogate for epilepsy severity. In addition, risk was also increased in those who had never been on AEDs. This is potentially an important group and needs further clarification. It may include those in whom the epilepsy is considered mild and treatment is not recommended, those who decline treatment, or those with recent onset epilepsy who have not yet been assessed or offered treatment. Risk of SUDEP is also increased in those whose treatment history was unclear, which may reflect the risk associated with the lack of treatment and uncontrolled seizures, although the reason for this was not objectively assessed.31
Despite several descriptive studies suggesting that subtherapeutic levels of AEDs are a risk factor for SUDEP,10,11,25,29 this has not been corroborated by the majority of case-control studies,34,44,45 as it is difficult to study as an independent factor. Of note is that postmortem levels of AEDs may not accurately reflect antemortem levels possibly due to, for example, redistribution and continuing metabolism.46 In a postmortem study reported by George and Davis, however, so-called subtherapeutic drug levels were detected in 69% of the 52 cases of SUDEP, in 75% of the eight cases where a seizure precipitated an accident causing death, and in 34% of the control population.47 This suggests that the increased likelihood of a seizure associated with “subtherapeutic” AED levels, rather than the levels per se, drives the observed elevated mortality rate. Compliance with AED treatment was first proposed as a risk factor for SUDEP in an uncontrolled study by Leestma, who found subtherapeutic AED levels in 68% of SUDEP cases.25 Therapeutic drug monitoring has traditionally been considered a surrogate for medication adherence, although due to the existence of a number of confounding factors, it is clear that the two terms are not interchangeable. For example, in patients with uncontrolled seizures, changes of dose and type of medication are commonplace, and serum levels will not be stable and may frequently be subtherapeutic, despite excellent compliance. More recent studies have attempted to address this by integrating additional clinical information, such as a previous history of noncompliance29 or an arbitrary judgment regarding the degree of compliance made by the treating physician.34 Despite this, conflicting results have been obtained (Table 15-2). Of paramount importance, however, is the understanding that the evaluation of drug levels or noncompliance as independent risk factors for SUDEP must take into account the presence, frequency, and severity of seizures. The issue of variability of AED use was recently addressed in a study by Williams and colleagues comparing hair AED concentration variability in patients with SUDEP, non-SUDEP epilepsy-related deaths, epilepsy outpatients and epilepsy inpatients. The SUDEP group showed greater hair AED concentration variability than either the outpatient or the inpatient groups, reflecting variable AED ingestion over time. However, these variations cannot distinguish prescribed changes from poor compliance or identify consistent noncompliance over time. Second, it does not provide information on drug-taking behavior immediately before death, as it takes about 5 days for a drug sequestrated into the follicle to appear at the scalp; therefore short-term noncompliance immediately before death is not assessed by this study and may have been overlooked.48
In summary, therefore, despite the existence of some evidence for serum or hair sample AED variability in SUDEP patients, the issue of AED compliance is far from being satisfactorily resolved. Perhaps one useful conclusion that can be drawn from the studies available is that none has shown a higher risk from therapeutic levels compared to those with subtherapeutic levels.
Despite a number of descriptive and controlled studies, no specific AED has been clearly associated with an increased risk of SUDEP,18,29,32,36,37,44,49 although a small number of studies have implicated treatment with carbamazepine as an independent risk factor.31,50,51 For antiepileptic medication in general, proposed mechanisms include perturbed heart rate variability, lengthening of the Q-T interval on the electrocardiogram combined with a mild proarrhythmic effect of epileptic seizure discharges, or excessive postseizure brainstem inhibition producing a blunting or transient abolition of the central hypoxic and hypercarbic respiratory drive, with consequent postictal respiratory arrest.50–52 Elevated serum levels of carbamazepine have been associated with an increased risk of SUDEP, even after adjustments for seizure frequency have been made. Frequent drug changes and multiple concomitant AEDs, conventional markers of severe, and unstable epilepsy increased this risk synergistically.45 On this basis, it is difficult to know whether a high carbamazepine level is an independent risk factor or is merely representative of challenging epilepsy.
In summary, AED-related factors are likely to be important in relation to SUDEP risk, but the evidence available does not allow us to differentiate risks from uncontrolled seizures and challenging epilepsy from those related to specific drugs, drug combinations, and drug changes, either prescribed or otherwise.
There is evidence from both descriptive and controlled studies that a terminal convulsive seizure,11,14,25,27,29,34,36,53 being found alone in bed,14,26–28,34,36 and being in the prone position29,36 are independent risk factors for SUDEP. Whereas a small number of descriptive studies have not found an association (Table 15-2), all case-control studies that have evaluated these factors have found a positive relationship with the risk of SUDEP. For example, the descriptive study reported by Ficker failed to find an association between being in bed and SUDEP or evidence for a terminal seizure, although only nine patients were evaluated.10 In a report by Nashef, following interviews with bereaved relatives, evidence for a terminal seizure was found in 24 of 26 cases, but it is of interest that only two were witnessed. The observation that, in most studies, unwitnessed cases far outnumber those witnessed suggests that enhanced surveillance of patients with epilepsy may be protective.27 This is corroborated by a study of young patients with epilepsy at a special residential school. All sudden deaths that occurred during the period of the study were when the pupils were not under the close supervision of the school, and most were unwitnessed.15 Similar findings of a protective effect of enhanced supervision at night were also found in a large controlled study, where supervision was defined as the presence in the bedroom of an individual of normal intelligence and at least 10 years old or the use of special precautions, such as checks throughout the night or the use of a listening device.31
In some cases when a prone position was not observed, other factors that might compromise breathing were identified. For example, Nashef noted that only five of 26 people were found facedown in the pillow and a sixth with the head in carpet pile. In total, however, in 11 of 26 cases, an extrinsic or intrinsic positional obstruction to breathing amenable to intervention may have contributed.27 Moreover, it is possible that this may be an underestimate, as obstructive apnea can occur in an apparently benign position.54
There is limited evidence for an independent relationship between learning disability and an increased risk of SUDEP. Early descriptive and population-based studies, in which learning disability was determined by observer impressions rather than by formal IQ examination, provided only weak support for this association.11,55 Most recent studies have found no clear correlation.27,34–36,39 However, Walczak identified an IQ of less than 70 to be a risk factor for SUDEP, even after accounting for seizure frequency.37 Observational studies are likely to be biased. Nashef found that five of 11 patients with SUDEP had a low IQ, but the study was undertaken at a tertiary referral center with a higher-than-average background incidence of learning disability.14 Similar bias was introduced in an observational study by Opeskin, where all patients in the study region who died in institutions for the disabled were mandatory coroner’s cases, thus artificially elevating SUDEP cases with learning disability.28 It has been postulated that patients with learning disability are more susceptible to central apnea and positional asphyxia that may cause SUDEP as a result of prolonged postictal encephalopathy,56 decreased postictal respiratory drive, and impaired movement and righting reflexes.37 This is likely to reflect associated pathology, rather than the learning difficulty per se. Despite early reports of an increased incidence of structural lesions in patients with SUDEP,11,25,57 this has not been confirmed by more recent controlled studies.30,36,37 Although there is evidence that psychotropic medication can influence the risk of sudden death in general, there is no convincing evidence of this being particularly relevant in SUDEP. Logistic regression analysis on 18 patients with SUDEP suggested an independent association of the risk of SUDEP with the number of concomitant psychotropic medications; however, the study was observational and uncontrolled, and the principal inclusion criteria for the study included any patient prescribed regular antiepileptic medication. Thus, the study population excluded those with epilepsy on no medication and included patients without epilepsy who were prescribed AEDs for an unrelated condition, for example, migraine, neuropathic pain, or psychiatric disorders.38 A single case-control study has also reported an increased risk of SUDEP in women on neuroleptic medication and in men prescribed anxiolytic medication. The explanation for this is unclear and most likely represents an effect of confounding factors.30 A number of more recent case-control studies have failed to identify a similar association.34,35,37
In summary, although not universally accepted, the weight of evidence at present suggests that children are at less risk but that neither sex nor a particular adult age group are strong risk factors for SUDEP, with the absence of suitable control subjects and lack of controlled prospective data significant methodological constraints. Seizure frequency appears to be an independent risk factor, particularly with respect to convulsive seizures, and the number of previous or concomitant antiepileptic medications is, most likely, interrelated. SUDEP mostly occurs after a convulsive seizure and is more common in unsupervised patients more often found in bed in the prone position. Supervision is protective. There is no convincing evidence of an independent association between SUDEP and subtherapeutic AED levels or specific AEDs. Structural cerebral lesions and co-prescribed psychotropic medication are also, for the most part, likely to be unrelated to the risk of SUDEP. A number of epilepsy syndromes are associated with an increased risk of sudden death, although evidence for an association with the more common syndromes, such as idiopathic generalized epilepsy, is limited at the present time, and more research is required. Methodologically robust multicenter controlled prospective studies of risk factors for SUDEP are required to definitively evaluate the large number of proposed variables in an independent fashion using, for example, logistic regression analysis. This will be further facilitated by standardization of definitions, case ascertainment, and analysis methods.