Mortality in epilepsy is increased two- to threefold (see Chapter 10). Much of the excess mortality is due to associated or underlying disease, but there is also a smaller amount of excess related to the epilepsy itself, due to accidental injury, drowning, status epilepticus or sudden death. Sudden unexpected (“unexplained”) death in epilepsy (SUDEP), in which an otherwise well person with epilepsy dies unexpectedly in benign circumstances with no cause found at autopsy, is the most important category of epilepsy-related deaths. As discussed in this chapter, we believe that such deaths are most likely related to epileptic convulsions. In recent years, some significant risk factors have been recognized through case–control studies. Nonetheless, it is not understood why some individuals are more at risk than others with a similar profile.

It has long been recognized that otherwise well individuals with epilepsy can die suddenly. Suffocation or asphyxia during a convulsion was considered a likely mechanism, as described in 1854 by Delasiauve¹³ in his treatise on epilepsy, under the section Terminaisons:

This description focuses on extrinsic mechanisms. It was also recognized that intrinsic mechanisms could account for such deaths. Early in the twentieth century, based on experience at the Craig Colony, Spratling⁶¹ described epilepsy as

In Munson’s mortality series³⁶ from the same colony, death from status epilepticus/series of seizures was relatively common and had declined compared to earlier years at the colony, unlike the category of sudden death. Nocturnal deaths were considered accidental and potentially preventable. Of note was “a definite and fairly large group where neither accident of any kind nor suffocation can be assigned as the cause of death which seemed to be intrinsic rather than extrinsic.” He concluded “that death is imminent at the time of seizures, unless help is at hand. The cause may be traumatic, suffocation may take place, or deaths may occur without any apparent cause…. The epileptic should be by himself as little as possible.” Many of the observations made by these earlier physicians are reflected in current knowledge. Despite a period of denial from some quarters in the second part of the twentieth century, SUDEP is now acknowledged as a real phenomenon and one that engenders concern. Not all that long ago, however, its existence was disputed. The efforts of investigators in the 1970s and 1980s, on both sides of the Atlantic, in drawing due attention to it should be acknowledged. These include Terrence, Hirsch, Brown, Earnest, Lathers, Schraeder, Leestma, Schwender, their colleagues, and others. SUDEP, however, is not a “cause.” It is but a convenient category in which to classify such deaths. Different risk factors may operate through a final common pathway of cardiorespiratory compromise.

This has been the subject of debate. To be classified as SUDEP, a death occurring in benign circumstances has to be sudden, unexpected, and without a clear cause. The latter requirement differentiates SUDEP from sudden death in general, in which autopsy may reveal a cause for death. Some investigators exclude observed peri-ictal deaths. However, most SUDEP cases are unwitnessed, with the person found dead, more often in bed, having been well the evening before. Evidence suggestive of a terminal epileptic seizure, such as tongue biting or incontinence, may be found at autopsy, but such evidence cannot be regarded as conclusive, nor indeed does its absence exclude an epileptic seizure. Thus, it is often not possible to be certain that the death occurred during or shortly after an epileptic seizure.

The following is a pragmatic, workable, if rather cumbersome, definition of SUDEP: 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 post mortem examination does not reveal a (structural or toxicologic) cause for death.³⁷ This definition makes no assumption of mechanisms(s) and allows similar deaths to be conveniently grouped. However, autopsies are not always performed, and, when they are, may be limited.⁴² Annegers and colleagues¹ suggested that deaths in epilepsy should be classified into (a) definite (with autopsy), (b) probable (suggestive circumstances but no autopsy), or (c) possible SUDEP cases or (d) cases with a clear cause of death. The definition above has been helpful in case–control studies. It has a major limitation, however. By excluding cases with coexisting pathology that predisposes to sudden death, it does not allow us to assess any increased risk of sudden death from epilepsy in such cases.

This needs to be viewed with reference to the background risk of sudden death in the general population of around 0.05 to 0.1/1,000 for those <45 years of age and 3/1,000 for those older.¹ Reported rates of SUDEP vary depending on the cohort studied, as addressed in previous reviews.^46,63 Broad estimates are presented below. Most cohorts are selected. Population-based studies of adequate size are difficult to achieve. The U.K. prospective population-based study of newly diagnosed epilepsy³² observed only 1 SUDEP case among 792 patients in 11,400 person-years of follow-up. One needs to consider, however, the excess mortality due to underlying or associated disease and the observation that 70% developed lasting remission. In a population-based study based on 9 SUDEP cases from Rochester, New York,¹⁸ the rate observed was 0.35/1,000 person-years. Two SUDEP cases were observed in 5,000 person-years of follow-up in the Medical Research Council antiepileptic drug (AED) withdrawal study³⁵ among patients with controlled epilepsy randomized to continued treatment or gradual drug withdrawal after a minimum of 2 years of seizure freedom. A multicenter unselected hospital series from the United States reported a rate of 1.21/1,000.⁷³ Rates of 1 per 200 to 300 person-years of follow-up were reported for selected cohorts with epilepsy and other disability or cohorts seen at specialist centers.^46,63 These reflect more intractable epilepsy or associated morbidity. Higher rates are suggested for series considered for epilepsy surgery and for failed surgery cohorts as discussed in more detail later. Another group with severe epilepsy is that of patients undergoing vagal nerve stimulation. A sudden death rate of 4.5/1,000 for definite/probable SUDEP cases was reported (6/1,000 if possible cases were included). This is comparable to that expected in an intractable cohort. The rate seemed to decrease with longer follow-up, but this has not been studied further.¹ More publications reflect risk for younger adults rather than children or the elderly, and extrapolation to other age groups is not appropriate. In the elderly, sudden death as defined earlier is less amenable to study. Attribution is difficult in the presence of competing comorbidity, and epilepsy-related excess mortality is “lost” within much higher overall death rates. SUDEP is known to occur in children more often in symptomatic than idiopathic epilepsy and is thought to be rare, but studies are limited and sometimes small.^{9,14,21,58,74} Mortality is considered more often related to the underlying neurologic disorder or associated deficit than to seizures. A long-term study of outcome in children with epilepsy, one that separates deaths in a seizure from SUDEP cases, observed an overall mean age of death of 18.6 years (range of 1–41 years), suggesting a greater risk as children enter adulthood.⁵⁸

There has been increasing interest in SUDEP in relation to epilepsy surgery, discussed here in more detail. The risk of SUDEP appears to be particularly high among patients with refractory epilepsy considered as candidates for epilepsy surgery, with an estimated incidence of 1/100 patient-years.¹² Follow-up studies after epilepsy surgery have initiated a discussion on whether epilepsy surgery could reduce mortality and lower the incidence of SUDEP in this high-risk group.⁵⁵ Stavem and Gudlvog⁶² reported survival in patients operated for focal epilepsy in Norway from 1948 to 1988. Operations for known brain tumors were excluded, but the pathology was otherwise not provided. With an average postoperative follow-up of 25 years, there were 34 deaths among 139 surgery patients (4 considered SUDEP and another 4 with “epilepsy” as cause of death). The risk ratio for death in relation to the general population was 6.2 (95% confidence interval [CI] 3.1–12.6). The survival of the epilepsy surgery group was not significantly different from that of matched controls with medically treated intractable epilepsy, although there was a trend toward a lower death risk ratio, 0.6 (95% CI 0.4–1.1). Patients who were seizure free 2 years after surgery did not differ from patients with recurrent seizures, but data on longer-term seizure control were not available. Unfortunately, the report does not clarify whether the postoperative SUDEP cases occurred among the patients rendered seizure free. The comparatively small number of epilepsy surgery patients included in this study may also have contributed to the failure to demonstrate a difference between surgery patients and controls. Furthermore, the fact that this cohort was operated on before the introduction of modern neuroimaging in presurgical workup calls for caution in generalization.

Outcomes have also been reported in 248 patients who had diagnostic evaluations for epilepsy surgery at UCLA from 1974 to 1990.⁷¹ Of these, 202 underwent surgery (anterior temporal resection in 175) and 46 did not. On follow-up, 14 (7%) of the surgery patients had died compared to 9 (20%) in the nonsurgery cohort. Among those who died, 81% had seizures, whereas only 47% of the surviving patients had seizures. Unfortunately, no detailed account is given on the causes of death, and there is no mention of SUDEP in this report, although the observations indicate a lower mortality in patients with refractory epilepsy that undergo surgery, in particular among those who become seizure free. There were 11 deaths in a follow-up of 215 patients treated surgically for refractory temporal lobe epilepsy at another U.S. center.⁵⁶ Three patients died during seizures, another 3 were classified as SUDEP, and 2 patients died in accidents. Mortality was lower (2%) among patients rendered seizure free than among those who continued to have seizures after the operation (11.9%). Survival was assessed among 305 patients who had surgery for temporal lobe epilepsy from 1975 to 1995 at the Maudsley Hospital in London.²² There were 20 deaths (6 SUDEP), resulting in a standardized mortality ratio (SMR) of 4.5 (3.2–6.6) and a SUDEP incidence of 2.2/1,000 person-years. Only 2 of the 6 cases of SUDEP were reported to be seizure free postoperatively. A population-based nation-wide Swedish study followed 596 patients who had undergone epilepsy surgery and 212 patients referred to epilepsy surgery evaluation that did not lead to an operation.⁴³ There were 14 deaths (6 SUDEP) in the surgery group, yielding an SMR of 4.9 (95% CI 2.7–8.3) and a SUDEP incidence of 2.4/1,000 person-years, very similar to the U.K. data. Among the nonsurgery patients, 5 died (4 SUDEP), resulting in an SMR of 7.9 (2.6–18.4) and an incidence of SUDEP of 6.3/1,000 person-years. Seizure outcome data for the entire cohort were available only at 2 years after surgery and did not seem to affect mortality or risk of SUDEP. However, none of the 6 cases of SUDEP were seizure-free.

A follow-up of 393 patients who had epilepsy surgery in the United States identified 11 deaths, of which 7 were epilepsy related (6 SUDEP).⁵⁹ None of the 199 patients who became seizure free after surgery died, whereas the SMR for those with recurrent seizures was 4.7 (2.3–7.9). This series has been extended to include 583 patients prospectively followed after epilepsy surgery.⁶⁰ In total, 19 deaths (10 SUDEP) were observed, 18 of which were in patients with recurrent seizures. The only death in the seizure-free cohort was due to breast cancer, and all 10 SUDEP deaths were among patients with recurrent seizures. The SMR was 5.8 (3.5–9.3) for the patients with seizures despite epilepsy surgery, whereas the mortality rate for those who stopped having seizures after surgery was indistinguishable from that of the general population.

Hence, available follow-up data after epilepsy surgery strongly suggest that SUDEP preferentially occurs among those with recurrent seizures, whereas the risk is remarkably low in patients that are rendered seizure free. The lower SUDEP risk is likely, at least in part, to be a consequence of the successful surgery, but other interpretations also need to be considered. Because these observations are from nonrandomized studies, it is possible that there may be underlying preexisting biologic differences between patients who respond favorably to surgery and those with recurrent seizures. Such difference, for example, in the spread of the epileptogenic zone, could be linked to the risk of SUDEP. The reported divergence in preoperative HRV between patients with good and poor outcome of surgery for temporal lobe epilepsy lends support to this latter interpretation.⁵⁰

The previous definition, by being inclusive, acknowledges the difficulties of separating SUDEP cases directly related to a “terminal” epileptic seizure and those occurring independent of such a seizure. In an unwitnessed case, circumstances that suggest, but do not prove, an epileptic seizure include fresh tongue, cheek or lip biting, secretions, incontinence, disrupted environment or bedding, fall off the bed, contorted facial expression, or timing and triggers of per habitual seizures. The absence of such clues does not exclude a terminal epileptic event. Whereas a number of studies suggest that SUDEP is frequently a peri-ictal event, the evidence is mostly indirect. The proportion of witnessed cases varies among studies from 7% to 38%. Of those witnessed, between one third and all were reported to be related to convulsions.²⁹ A series of witnessed SUDEP cases,³⁰ among the first 135 ascertained for a case–control study, reported convulsions in 12 of 15 cases. An older study, based on detailed interviews of bereaved relatives or partners, addressed circumstances of death in 26 SUDEP cases (2 witnessed), and reported evidence suggestive of an epileptic seizure around the time of death in 23.³⁹ Signs of preceding seizures were reported in 67% in a case–control study of 42 SUDEP cases from Norway.²⁸ Monitored and reported SUDEP cases are rare. One such case occurred in a video telemetry unit during a secondarily generalized epileptic seizure.⁶ Birnbach et al.,⁷ in an early case–control study, reported an association with convulsive seizures. Observations from more recent case–control studies report a higher relative risk in people with more frequent seizures and in those with a history of generalized convulsive seizures (Table 1). This supports the premise that most SUDEP cases are seizure related. We believe that the majority of SUDEP cases are likely to be related to an epileptic seizure occurring close to the time of death.

The association with convulsive epileptic seizures accounts for some of the evidence only, and other risk factors need to be considered. Some individuals might be more at risk because of social factors, lifestyle, suboptimal management, and lack of adherence to treatment. Others might have additional biologic susceptibility. Descriptive cohorts suggested risk factors for study. These included youth, male gender, remote symptomatic epilepsy, structural findings on neuropathology, severe epilepsy, unwitnessed seizures, alcohol abuse, abnormal electroencephalograms (EEGs) with epileptiform changes and greater variations, mental disability, psychotropic medication, African American race, lack of adherence to treatment, abrupt medication changes, and low AED levels.

Certain risk factors have been shown to be associated with SUDEP in case–control studies. These studies have different methodologies and are not directly comparable. Some studies, for example, only include patients on long-term treatment for epilepsy or in specified age ranges. Not all studies are nested. Control groups vary and include living patients known to have epilepsy or patients with epilepsy who died of other causes. SUDEP cases are identified from a variety of sources, which could result in different data being available. Table 1 lists significant findings in some more recent case–control studies. Negative findings and older studies have not been listed.