SUDEP Mechanisms: The pathway to prevention

Introduction

Most patients with epilepsy lead a normal life. As long as they take their medication, a majority can remain seizure free, work, and be active. If they do have a seizure, the risk of injury is usually minimal as long as they are not driving or in another dangerous situation. However, in recent years it has become increasingly appreciated that patients with epilepsy have a risk of sudden death that is more than 20 times greater than the general population1. This SUDEP is estimated to cause 2000-3000 deaths per year in the U.S. alone², which is probably an underestimate for reasons described below, and accounts for as many as 15% of all epilepsy related deaths³. The single risk factor identified most consistently across multiple studies is the frequency of generalized tonicclonic seizures (GTCS)⁴.

Despite such a high incidence, in Canadian pediatricians who treat children with epilepsy, only 56% reported that they knew their patients were at higher risk of sudden death. Furthermore, only 33% of the surveyed pediatricians had heard of the term SUDEP⁵. These numbers are likely similar in the US, and are probably higher among non-neurologists. However, these numbers will likely change rapidly, because SUDEP is becoming more salient among patients and their families, and as they educate themselves doctors will become more aware. The pace of publications on SUDEP is rapidly increasing⁶, yet there remains a gap in our knowledge of the mechanisms involved, in part because the unpredictable occurrence of SUDEP has hindered the gathering of high quality data at the time of death. In particular, there have been human SUDEP cases in which the EKG was recorded and respiratory rate was estimated by visual observation, but no blood pressure or tidal volume data were obtained. Here we discuss data from SUDEP cases, experiments monitoring human peri-ictal physiology and recordings during seizure-induced death in animal models. We then interpret these data and their relevance to the mechanisms of SUDEP.

SUDEP is “the sudden, unexpected, witnessed or unwitnessed, non-traumatic, and non-drowning death of patients with epilepsy with or without evidence of a seizure, excluding documented status epilepticus, and in which postmortem examination does not reveal a structural or toxicological cause of death”⁷. There is no diagnostic test for SUDEP, and no finding on autopsy that is pathognomonic. Therefore, ascertainment of cases is difficult, because the term SUDEP is not always used by coroners or medical examiners. There is also significant variability in the patient populations included in epidemiological studies. These factors contribute to the wide variation in estimates of the incidence of SUDEP, ranging from as little as 0.09 per 1000 persons per year among unselected incident cases, all the way to 9.3 per 1000 persons per year in surgical resection candidates 3. Misclassification may also contribute to the wide range of published estimates for SUDEP incidence. For example, among refractory epilepsy patients SUDEP has been determined to account for as little as 10% to as much as 50% of all deaths ³. A recent meta-analysis from the CDC of refractory patients has placed their lifetime risk for SUDEP at 37%². Although the total number of SUDEP deaths is not as large as that of many other neurological disorders, such as Parkinson's Disease, Alzheimer's Disease and motor neuron disease, the peak incidence of SUDEP occurs at the age of only 30 (Fig. 1), so the “years of potential life lost” is second only to stroke among all neurological diseases².

Pathophysiology

Determining the pathophysiological mechanisms underlying SUDEP is critical in order to develop effective preventive treatments. It has been well known for many years that seizures can induce cardiac arrhythmias (Fig. 2a), changes in blood pressure, and asystole, and this led to the assumption that SUDEP was due to a cardiac mechanism⁸. However, respiratory depression and apnea occuring during and after a seizure^9-11 have been well documented, and these abnormalities can be severe enough to cause oxygen desaturation (Fig. 2b)^{12, 13}. These observations led to a debate about whether the primary, or initiating, event in SUDEP is more frequently an abnormality of breathing, cardiovascular function, or both. Some reports have hypothesized that pulmonary edema is a possible cause of death, as mild pulmonary edema is regularly found in SUDEP cases upon autopsy^14-16. However, this mild edema is not considered severe enough to be the primary cause of death³. Here we will focus the effect of seizures on breathing and cardiac function. We will then describe how the loss of arousal mechanisms impairs the protective reflexes that can be essential for survival during recovery from a seizure. Finally, we will discuss insights obtained about mechanisms from SUDEP cases that were monitored at the time of death.

Seizures and respiratory dysfunction

Lack of oxygenation during seizures was first reported in 1899 when Hughlings Jackson noted his human patients and monkeys “turning blue” during seizures⁹. Since then, others have noted similar findings, and have documented apnea^{10, 17} and oxygen desaturation^11-13 during and after seizures. Using long-term video-EEG monitoring and respiratory recordings, Nashef and colleagues¹² found that 10 of 17 patients (59%) developed apnea and 6 patients experienced oxygen desaturation (SaO₂) ranging from 55-83%, both concurrent with seizures¹². Investigating 304 seizures in 56 patients with intractable localization-related epilepsy, Bateman and colleagues¹³ reported SaO₂ values decreasing below 90% in 33.2% of seizures, below 80% in 10.2% of seizures, and below 70% in 3.6% of seizures. Desaturation was common and severe even with seizures that did not generalize. In 100 seizures for which data were available including airflow measurements, central apnea or hypopnea occurred in 50%, whereas mixed or obstructive apneas occurred in 9% of seizures¹³. In seven patients with 19 seizures for which end-tidal carbon dioxide (ETCO₂) data were available, SaO₂ below 85% was accompanied by an increase in ETCO₂ of 18.6 ± 17.7 mmHg, indicating hypoventilation¹³. Oxygen desaturation was not unique to GTCS, as SaO₂ decreased below 90% in 34% of partial seizures¹³. These data provide important evidence that respiratory depression could be severe enough to cause SUDEP, but they raise the question of why death is not even more common given how often seizures induce hypoventilation.

Peri-ictal hypoventilation can lead to secondary cardiac failure. When SaO₂ drops below 90%, it has been shown that there is a greater than four-fold likelihood of QT prolongation as compared to seizures without decreased SaO₂¹⁸. Moreover, in seizures with oxygen desaturatinon there is a greater than two fold increase in likelihood that the QT interval is shortened. These findings raise a cautionary note, because cardiac changes observed as terminal events in SUDEP cases may be due to hypoventilation that would go unnoticed without proper monitoring.

Animal models are a valuable way to determine underlying mechanisms of seizure-induced death. DBA/2 mice are a well-established mouse model of audiogenic seizures, which lead to respiratory arrest and death, but the mice can be resuscitated using a ventilator¹⁹. A caveat is that these mice do not have spontaneous seizures so they would not meet the definition of SUDEP, which requires that the subject has epilepsy. However, there are no data to show that epilepsy is required for a seizure to be fatal in humans, rather than death being due to the seizure itself. Data from DBA/2 mice indicate that seizures inhibit the ponto-medullary respiratory control network responsible for control of breathing^{20, 21}, which leads to apnea followed by asystole and death. Seizure-induced respiratory arrest and death also occurs in DBA/1 mice, HTR2C knockout mice, and others^{22, 23}. Seizures can produce a strong inhibitory effect on breathing, and in mouse models this can be fatal, but it remains unclear whether this can occur in humans.

Seizures and cardiac dysfunction

In 1906 cardiac asystole during a seizure was reported by Russell²⁴. Since then many ictal cardiac abnormalities have been reported such as a prolonged QT interval, tachycardia, bradycardia, and Torsades de Pointes^{8, 25}. Seizures often induce tachyarrhythmias, with 76% of epilepsy patients experiencing tachycardia during at least one seizure, and 57% of seizures are accompanied by tachycardia. This is closely correlated with seizure generalization^{26, 27}, and may be due to seizure spread to the hypothalamus with activation of first order sympathetic neurons. The same studies found that only 2% of seizures result in bradycardia, while only 0.5% resulted in asystole. Moreover, in a study with over 250 seizures in 56 patients, one case of peri-ictal bradycardia was reportedly followed by asystole¹³. This case had concurrent O₂ saturation that fell below 50% suggesting that bradycardia and asystole may have been secondary to hypoventilation. Two additional reports demonstrate that hypercapnia and hypoxia can both lead to tachycardia in subjects with no cardiac abnormalities^{28, 29}, consistent with known effects of the respiratory system on autonomic output. These can be due to changes in blood gases or due to interactions between neurons with the brainstem.

Repolarization abnormalities such as QT prolongation are common during seizures and it is proposed that these could contribute to SUDEP pathophysiology. Recent studies have found that there is prolongation of the QT interval during seizures. In a study of 76 patients, significant ictal prolongation of the QT interval was found in 4.8% of seizures whereas short QT intervals were observed in 3.8% of seizures³⁰. However, no SUDEP risk factors were associated with altered QT intervals.

Heart rate variability (HRV) is due to modulation of the sinoatrial node by the autonomic nervous system³¹. Patients with temporal lobe epilepsy (TLE) have reduced HRV when compared to the general population, and this reduction is more pronounced during the night^{32, 33}. Reports show abnormalities in HRV and cerebrovascular autoregulation could be due to seizure activity, since they improve after epilepsy surgery^{34, 35}. HRV defects are correlated with surgical outcome – patients with the poorest outcome have more significant HRV impairment³⁶. Reduced HRV has been associated with increased cardiac mortality and sudden cardiac death^{31, 37}. Therefore, it would be expected that HRV is involved in SUDEP mechanisms. Others have reported an inverse correlation between high frequency HRV and risk of SUDEP³⁸. However, a study comparing 7 SUDEP cases and 7 age-matched controls showed no difference in HRV³⁹. Therefore, the role HRV plays in SUDEP pathophysiology remains unclear.

Post-ictal arousal deficits

GTCS^{40, 41} and complex partial seizures (CPS)^{42, 43} are both associated with behavioral arrest and impaired consciousness. Patients with these seizures are unresponsive to commands and questions during the post-ictal state, however patients with CPS are able to grasp a ball and visually track⁴⁴. These data suggest that although consciousness is reduced in CPS, they are closer to a minimally conscious state than a post-ictal coma as occurs following GTCS, which could increase risk of SUDEP.

Work in the 1930s and 1950s by Wilder Penfield and Herbert Jasper led to the theory that control of consciousness involved subcortical structures based in the diencephalon and upper brainstem^{45, 46}. These subcortical structures responsible for arousal and consciousness are collectively known as the ascending arousal system (AAS)^{47, 48}. This system is composed of discrete neurotransmitter-specific nuclei, including the raphe nuclei⁴⁹, the locus coeruleus⁵⁰, the ventral tegmental area⁵¹, the pedunculopontine and laterodorsal tegmental nuclei⁵², and the parabrachial complex⁵³. If a seizure spreads into the midbrain and medulla and affects these nuclei it could alter a patient's level of consciousness. Blumenfeld⁵⁴ proposed the network inhibition hypothesis, which posits that consciousness is maintained through interactions between the cortex and subcortical structures. Once aberrant discharges start in the cortex, they travel along known anatomical connections to subcortical structures and activate GABAergic interneurons that inhibit the AAS. With the AAS inhibited, cortical activity is suppressed resulting in loss of consciousness. Data supports this theory, as SPECT imaging has shown increased blood flow starting in temporal lobe structures and descending into the midbrain and brainstem in patients with temporal-lobe CPS⁵⁵. Furthermore, recordings from cholinergic neurons in the pedunculopontine tegmental nucleus demonstrate that acetylcholine neurons in the AAS are inhibited during limbic seizures⁵⁶. Loss of consciousness due to AAS inhibition leaves a patient with epilepsy at greater risk for SUDEP, as protective reflexes are suppressed during the post-ictal coma.

SUDEP: Respiratory, cardiac or both?

Our current understanding of SUDEP comes from a variety of sources. There have been measurements of the cardiorespiratory response to seizures of patients in epilepsy monitoring units (EMUs), but it remains unclear whether the observed effects are relevant to the fatal events of SUDEP. There are also data from seizure-induced death in animal models, but it is also not known if these mechanisms are the same as those in human SUDEP. The most relevant source of data comes from human SUDEP cases, including information from death scene investigations, deaths witnessed by nonmedical personnel, and deaths of patients while monitored in the hospital.

Genetics of SUDEP

There are no genetic mutations identified that cause SUDEP, but there are a variety that have been associated with an increase in risk (Table 1). It is possible that genetic screening of families of patients who have died of SUDEP could uncover mutations and polymorphisms in unidentified genes that make it more likely for cardiorespiratory dysfunction to occur during seizures or in response to hypoxia⁶³. There are also genes that confer an increased risk of SUDEP for reasons that are not yet clear. Examples include those for long QT syndrome (LQTS) and Dravet syndrome (DS).

Cellular mechanisms

The majority of work on the mechanisms of SUDEP have focused on cardiac electrophysiology and on cortical seizure spread. More recently, clues from animal experiments and from brain slice work point to possible brainstem mechanisms involving serotonin and adenosine, and to abnormalities of sympathetic innervation.

Parallels to SIDS

SUDEP and sudden infant death syndrome (SIDS) share many similarities^{6, 90}, and both are linked to defects in the 5-HT system. SIDS is defined as the death of an infant in the first year of life, which is often associated with sleep and the cause of death is not identified upon autopsy, death scene investigation, or complete review of the clinical history¹⁰⁶. This definition is very similar to that of SUDEP. Both are diagnoses of exclusion and there are no pathognomonic features. SIDS is thought to be caused by arousal deficits and cardiorespiratory dysfunction and much of the current research is focused on investigating the role of 5-HT in arousal and breathing^{107, 108}. Available data points to a defect in the 5-HT system in SIDS cases, including: increased number of 5-HT neurons of immature appearance¹⁰⁹; reduced 5-HT_1A ligand binding in the raphe nuclei¹⁰⁹; and reduced 5-HT in the brainstem^{110, 111}. The many similarities between the two syndromes makes it more convincing that 5-HT dysfunction may also play a role in SUDEP. Moreover, one study reported that infants who had a prolonged QT interval in the first week of life had a greater than 40-fold increase in risk of SIDS than controls¹¹², which further links SIDS and SUDEP. Since most cases of SUDEP and SIDS are unwitnessed, it is possible that some SIDS cases are actually SUDEP^{6, 113}.

Summary and hypothesis

The mechanisms of SUDEP remain undefined. Cardiac and respiratory abnormalities both occur quite commonly during and after seizure activity, yet SUDEP incidence is only 2000-3000 cases per year². This disconnect between the large number of cardiorespiratory abnormalities after seizures and the much lower incidence of SUDEP could be attributable to severity of cardiorespiratory depression. In cases where the abnormality is mild, a patient would easily recover and mortality rates would be low. Conversely, in cases where a patient has prolonged cardiac or respiratory depression they would be unable to fully recover and mortality would be high. If either cardiac or respiratory dysfunction occurred in parallel with arousal deficits, the risk of SUDEP would be even higher.

We propose a hypothesis in which seizures activate neurons that project to the midbrain and medulla and inhibit monoaminergic and cholinergic neurons of the AAS, including the component that descends to the respiratory network in the medulla (Fig. 4). Activity of cardiovascular and respiratory neurons would be inhibited, at the same time that cortical activity is suppressed. At the end of the seizure, inhibition of the AAS would cause postictal unresponsiveness and PGES. Hypoventilation during the seizure would be followed by severe postictal hypercapnia and hypoxia. These blood gas derangements would then lead to bradycardia, asystole, and death (Fig. 4). Physiological monitoring of a theoretical patient during an episode of SUDEP is shown in Fig. 5, illustrating the sequence of events that might occur during a primary respiratory event, and consistent with data obtained in the MORTEMUS study. Mouse models of SUDEP may be valuable to investigate whether the same sequence of events occurs, and to then investigate cellular mechanisms such as a role of 5-HT or adenosine. Adopting new treatments will rely on identification of biomarkers for those at high risk of SUDEP and a greater understanding of underlying mechanisms.