The past, present and future challenges in epilepsy related and sudden deaths and biobanking

Past and current post-mortem practice in epilepsy related deaths

In suspected ERD and SUDEP cases, a ‘complete’ post-mortem examination is generally considered to be external and full internal examination, including examination of the whole brain by a neuropathologist or personnel with neuropathology training, and toxicology, including anticonvulsant levels [16, 17]. National audits of ERD highlighted inadequacy of post-mortem investigations in 87% in the UK in 2002 [17] and 41% in a Melbourne study in 2016 [16]. Insufficient or absent neuropathological investigations were common and in some regions access to neuropathology services is limited.

Recognizing that in some cases autopsies were either not done at all or incomplete (e.g., head only) or the cause of death was inconclusive, categories of definite, probable and possible SUDEP were devised [18] and subsequently revised [13] to facilitate classification of deaths for purposes of epidemiology and research. In ‘probable SUDEP’ cases, circumstances are most consistent with SUDEP but post-mortem examinations are lacking or incomplete. In ‘possible SUDEP’ cases, a potential competing cause of death (other than epilepsy) is identified, meaning that epilepsy and another unrelated cause are both plausible. Cardiac, pulmonary diseases and drug intoxication are the most common competing causes in the ‘possible SUDEP’ deaths as well as death in water without evidence of submersion or drowning [13]. In instances where the cause of death is considered to be the combined effect of epilepsy plus a concomitant condition, the term ‘definite SUDEP plus’ has been proposed [13] (Table 1).

Medicolegal aspects and death certification

International and regional autopsy practice in ERD vary and may be conducted by general pathologists or forensic pathologists, typically under jurisdiction of a Coroner or Medical Examiner. Determining cause of death in ERD cases is a challenge for the Medical Examiner or Coroner since they are often confronted with incomplete data (e.g. unwitnessed death, insufficient medical history, decomposed body), competing causes of death, or compelling circumstances without confirmatory anatomic or toxicological findings.

The cause of death is the aetiologically specific disease or injury that initiated the sequence of physiologic events leading directly to death. Examples of aetiologically specific causes of death include idiopathic epilepsy, atherosclerotic cardiovascular disease, drowning, and blunt force trauma. Mechanisms of death, the intervening steps that link the cause to the death, are the altered physiology or biochemistry through which the cause exerts its lethal effect and are not aetiologically specific (i.e. they can be due to many underlying aetiologies). Because the physiology of organ systems is complex and intertwined, there is never just one mechanism. Pulmonary oedema, cerebral hypoxia-ischaemia, arrhythmia and seizure are mechanisms of death. Each of these mechanisms may be happening during the death of a person with epilepsy. Pulmonary oedema is a finding that should be documented when present in ERD but is non-specific and it is recommended that this is not used in isolation as the cause of death in a patient with epilepsy.

Because there is often little evidence at post-mortem examination to determine when a seizure has occurred, it is often impossible to be certain that epilepsy was the cause of death in an unwitnessed case. When a person with epilepsy dies suddenly and is found to have another potentially lethal disease (e.g. Coronary artery disease), the Medical Examiner must weigh which aetiology was the more likely cause of death using details from the investigation and post-mortem examination. Pulmonary oedema and tongue bites, particularly lateral tongue bites, are considered suggestive but not pathognomonic for seizure since they may be due to other aetiologies [19, 20]. Even a description of terminal seizure-like activity is not specific since persons dying of other diseases may have such activity without established epilepsy. Therefore, medical history and circumstances of death are critical to the final evaluation in any case. Absolute certainty of the cause is not needed for competent death certification. Once the Medical Examiner or Coroner has explored and excluded potential causes unrelated to epilepsy, through detailed investigation and complete post-mortem examination, the sudden death of a person with epilepsy is most likely caused by epilepsy.

The choice of wording for the death certificate varies widely between jurisdictions and between individuals within jurisdictions. Inconsistencies and lack of standardisation of wording remain widespread [6, 21] limiting our ability to track SUDEP rates over time. Despite the ubiquitous reference to SUDEP as a cause of death in the literature, it is an acronym that describes the circumstances of death; sudden, in a person with epilepsy, not due to an unnatural cause (e.g. not trauma or drowning), not the result of status epilepticus, and in the setting of a negative post-mortem examination (i.e. no obvious toxicological or other anatomic cause identified) (Nashef, 1997). Some Medical Examiners and Coroners are resistant to using SUDEP as a cause of death statement and favour use of the specific aetiology (epilepsy) instead. Unfortunately, the sudden nature of the death is therefore often omitted from the cause statement, making identification and ascertainment of SUDEP cases for epidemiology and research difficult. Nevertheless although rarely used in previous decades [22], ‘SUDEP’ is now more commonly recognized as an entity and listed on death certificates [23] and accepted as a valid diagnosis [11, 22].

Neuropathology in SUDEP: Why it is needed

The primary objectives of the neuropathological examination in an ERD are to (i) exclude any unexpected cause of death, (ii) detect any underlying lesion causing the epilepsy and (iii) explore any neuropathological consequences of chronic or recent seizure activity. No neuropathological feature can establish a seizure as the mechanism of death. Furthermore, epilepsy is a clinical diagnosis supported by EEG and there are no specific neuropathological alterations that can categorically confirm epilepsy without confirmatory investigations undertaken during life. Neuropathological lesions that cause sudden death are often dramatic, macroscopically straightforward diagnoses; these include acute traumatic brain injury or spontaneous intracerebral haemorrhages [24]. More focal or subtle neuropathologies, such as fatal colloid cysts [25] or acute encephalitis [26, 27], require a systematic, careful approach to brain examination with confirmatory immunohistochemistry or molecular studies, for example for viral markers, if necessary.

Among lesions that cause epilepsy, published series do not suggest that any single type is over-represented in SUDEP (Table 2, Figure 2). The range of pathologies encompass those commonly encountered in surgical epilepsy series, including hippocampal sclerosis, cortical malformations such as focal cortical dysplasia (FCD) and low grade tumours [28]. Focal neuropathological abnormalities occur in ~50% of SUDEP autopsies ranging from 34 to 89% between larger SUDEP series (Table 2). Leestma, identified brain lesions in 60% of SUDEP cases compared to 11% in non-epilepsy autopsies [3], but not all SUDEP studies support this [23, 29] and there are few post-mortem non-SUDEP epilepsy studies for comparison (Figure 2). The variability in lesion detection may reflect the population demographics of the study as well as methodological differences between studies (e.g., macroscopically visible versus additional histological pathologies being reported). Identification of brain pathology pertaining to epilepsy is highly dependent on (i) the brain being examined in detail following a period of fixation [3, 30, 31], (ii) the number of histological regional samples taken [31], (iii) the level of scrutiny, for example if routine immunohistochemistry was carried out, as recommended in epilepsy surgical diagnostic protocols, [32], and (iv) the experience and training of the pathologist. In the forensic setting, the depth and detail of neuropathological investigation may be negatively affected by limited resources, including cost and access to special histologic techniques, neuropathology services and reporting time restrictions. Indeed, many epilepsy related pathologies are focal and visible only microscopically, some requiring specific immunohistochemistry to confirm diagnosis. Examples include milder or atypical forms of hippocampal sclerosis (ILAE type 3), granule cell dispersion [33] and hippocampal mossy fibre sprouting, considered to be a relatively specific secondary process in epilepsy (reviewed in [34]) (Figure 1c,d). Mild malformations of cortical development (MCD), in particular mild MCD type II (heterotopic neurones in the white matter), have been reported in SUDEP [35] but are recognised to require quantitative immunohistochemistry confirmation with comparison to validated control ranges [36]. Currently we lack international agreement on precise diagnostic criteria for Mild MCD type II in surgical TLE or other brain regions [37] and because of this its recognition and reporting in ERD may widely vary. Even well-defined epileptogenic pathologies likely go undetected at post-mortem examination due to under-sampling. For example, symptomatic FCD are not infrequently MRI ‘occult’ even with modern neuroimaging [38] (Figure 1e,f); such lesions are typically macroscopically less visible with absent hallmark features, as regions of cortical thickening or blurring of the grey-white matter interface [37]. Access to clinical notes and any localising investigations (e.g. EEG, seizure semiology, MRI, PET findings) is good practice and may help to focus the pathologist on the epileptogenic zone and anecdotally can aid in the identification of underlying pathology, although data is lacking to prove this increases overall lesion detection rate. Given these methodological approaches that may influence the sensitivity of the post-mortem, there is lack of clarity from the available neuropathological series as to whether structural brain lesions are more frequent in SUDEP than epilepsy without SUDEP and therefore if symptomatic epilepsy compared to idiopathic/genetic causes represents a higher risk.

Regional Neuropathology - insights into mechanisms in SUDEP

Repetitive seizures can initiate molecular signals and cellular processes that ultimately alter the brain structure in terms of neuronal loss, gliosis [39], microgliosis and inflammation [40], blood brain barrier breakdown [41], vascular changes and axonal reorganisation [42, 43]. These can occur in the area of seizure onset in focal epilepsies but widespread brain alterations in the cortex [44], thalamus and cerebellum can be observed at post-mortem [45]. Many of these sequelae represent a ‘pathological’ repair process that can further alter the seizure threshold, as exemplified by astrogliosis [39]. One key question is whether specific or strategic patterns of epilepsy-related brain injury occur in SUDEP which could increase vulnerability or mechanistically be relevant to the cause of death. Central autonomic networks and cardiorespiratory centres are obvious candidate areas to investigate [12, 46]. These include the insular cortex, prefrontal cortex, hippocampus and amygdala through their connections to the hypothalamus, pons and medulla [47]. Ictal autonomic phenomena, as urinary urge, piloerection, and heart rate changes arise due to excitation or inhibition of these brain regions and are particularly associated with limbic or TLE [48]. Neuropathology studies have focused on chronic structural alteration as well as acute pathology. Inherent caveats to SUDEP research include imprecise times of death as most are unwitnessed, the unanticipated nature of death that incurs logistical delays in brain removal, and lack of suitable controls groups such as non-epilepsy sudden deaths and non-SUDEP epilepsy cases of similar age and severity of epilepsy.

SUDEP – evidence for acute brain injury

SUDEP may be the consequence of extensive, irreversible neuronal damage during a seizure, strategic injury in critical autonomic regions or the cumulative effects of recent poor seizure control [86]. Such cellular injury or neurochemical exhaustion could decompensate normal homeostatic, auto-resuscitative mechanisms following a fatal seizure. Acute eosinophilic neuronal change has been reported in up to half of SUDEP cases [31], although with notable differences between series (Table 2) which may reflect to its varying extent and hence detection (Figure 2). It primarily involves the hippocampus (CA1/subiculum) although widespread involvement including cortical laminar patterns can occur [35]. This non-specific cytopathic change likely reflects hypoxic and/or excitotoxic cellular stresses occurring acutely (rather than immediately) in the last six hours. Hypoxia in epilepsy reflects ictal hypoperfusion [87], impaired central respiration/transient apnoea, external airway compromise during seizure or a synergistic interaction of these. Acute neuronal injury in SDUEP was significantly more frequent when seizure episodes occurred in the 24 h prior to death, the body was found in a prone position, external airways were obstructed, or brain swelling was present [31]. In sudden unexplained death in childhood, hyper-eosinophilic acute ischaemic neurones were identified at similar rates to controls groups with known cause of death [69].

Investigations in SUDEP of markers of acute neuronal death (necrosis, apoptosis) or reversible injury, including cell-stress signals promoting neuronal survival (VEGF, JNK1-3, BNDF) [88], enable more objective data regarding the type, severity and distribution of brain injury, and could aid in localising the epileptogenic zone and explore secondary effects of ictal discharges involving central autonomic networks [48]. Heat Shock Protein (HSP-70)-positive neurones in all hippocampal subfields were more frequent in SUDEP compared to non-epilepsy and other epilepsy sudden deaths; similar differences were not noted for c-JUN neuronal expression [57]. Labelling of insular cortex and brainstem nuclei with HSP-70 and c-JUN was also observed [57]. Hippocampal neuronal hypoxia-inducible factor 1-alpha (HIF1-α) and vascular endothelial growth factor (VEGF) expression, both upregulated in response to hypoxia, occur in SUDEP although did not significantly differ from control groups [89]. A more recent study identified fewer HIF1-α neurones in the parahippocampal gyrus in SUDEP compared to non-epilepsy sudden death controls but no differences in the hippocampus, amygdala or brainstem, where neurones in the pre-Bötzinger complex were labelled strongly [56]. Further investigations correlating evidence of neuronal injury with clinical and genetic variables and circumstantial evidence of death will be important to inform on mechanisms in SUDEP.

Seizure induced neuroinflammation [40] and acute blood brain barrier dysfunction [41] pose obvious candidate mechanisms in SUDEP. Increased, but statistically non-significant, brain weights and brain swelling, likely due to acute cerebral oedema, occur in SUDEP (Table 2) [31]. However, studies as yet do not indicate significant neuroinflammation in SUDEP; less lymphocytic inflammation was noted than controls [80] and quantitative analysis of HLA-DR activated microglia in the hippocampus, amygdala and medulla did not differ between SUDEP, and non-epilepsy sudden deaths [56] (Figure 1g, h). Albumin and IgG, tissue markers of acute blood brain barrier breakdown, showed less labelling in SUDEP in the parahippocampal gyrus, and fewer CD163-positive peripheral macrophages (indicative of increased vascular leakage) were found in the ventrolateral medulla [56]. In summary is unclear if neuroinflammation, or its cumulative effects in the presence of repeated seizures, is related to the pathogenesis of SUDEP.

Cardiopulmonary pathology and investigations in SUDEP

Pulmonary pathology is a common autopsy finding in SUDEP (Table 2) reported in 72% of cases, with the most frequent findings being pulmonary congestion or oedema [90]. Although the incidence of post-ictal pulmonary oedema has been debated [91], radiographic evidence was reported in 29% of patients following GTCS [92]. The pathogenic mechanisms of oedema following GTCS are not understood, and include increased sympathetic drive, mediated by the medulla and hypothalamus, altered hydrostatic pressures and increased pulmonary capillary permeability (‘neurogenic pulmonary oedema’) [91]. In SUDEP, additional exacerbating factors include central respiratory inhibition and cardiogenic and haemodynamic dysfunction.

Cardiac examination in ERD and SUDEP is essential to exclude cardiac causes of death such as undiagnosed myocardial disease, valve disease, coronary artery disease or structural abnormalities of the conduction system. When the clinical diagnosis of epilepsy is less certain and the cause of death is considered likely cardiac, but no lesion is identified, sampling for histology is required with mapped blocks of anterior, lateral and posterior right and left ventricle and septum from a representative mid-ventricular transverse slice and right ventricular outflow tract. In addition blocks from conduction tissue is taken if history of conduction defect is recorded clinically. In some situations, fixation of the heart and referral to a specialist cardiac pathologist is recommended as best practice and the organ can be returned following examination. There should be a low threshold for sending the whole heart to a specialist centre for expert opinion, at the discretion of the pathologist and Coroner or Medical examiner and discussion with the family (for UK guidance see https://www.rcpath.org/profession/clinical-effectiveness/clinical-guidelines/autopsy-guidelines.html). Following status epilepticus or a convulsive seizure, stress-induced cardiac changes or Takotsubo cardiomyopathy can occur [90] (Figure 3). When mild to moderate cardiac pathology is found in a case with circumstances and clinical features suggestive of SUDEP, the classification should be ‘definite SUDEP plus’ if the cardiac condition could be a contributing cause, or ‘possible SUDEP’ if the cardiac condition is a competing cause of death. This nuanced distinction, often challenging to apply in specific cases, was introduced in the more recent SUDEP classification [13]. For example, if a 30-year-old man with treatment resistant nocturnal convulsions is found dead in bed in the morning in the prone position with a tongue laceration, and post-mortem shows a 40% occlusion of the left anterior coronary artery or mild cardiomegaly, the cause of death is likely epilepsy, although the cardiac pathology may have contributed (definite SUDEP Plus). By contrast, a 55-year-old woman with infrequent seizures who dies alone in her flat during the daytime with no evidence of a seizure and who is found to have moderate cardiomegaly with left ventricular hypertrophy could have died from the cardiac disease or epilepsy (possible SUDEP).

Subtle cardiac pathologies reported in SUDEP series include myocyte hypertrophy and mild interstitial fibrosis (Figure 3) (Table 2 and reviewed in [90]). It remains unclear whether these changes represent a sequel of seizures or antiepileptic drugs, or may be similar to baseline rates of cardiac pathology in other non-cardiac sudden deaths; their relevance to the mechanism of SUDEP is unknown. Nevertheless, this again suggests that routine systematic histological examination of the myocardium, even in what appear to be typical SUDEP cases [93], is warranted in to monitor the significance of these unexpected findings. Ictal disturbance in cardiac electrophysiology that induces dysrhythmias has long been considered a candidate mechanism of death in SUDEP [94]. Further, several ion channels have shared expression in brain and heart and cardiac genes associated with long QT syndrome (LQT), bradycardia, and SCD (KCNQ1, KCNH2, SCN5A, RYR2, HCN4) can cause both epilepsy and arrhythmias or increase the risk of seizure-induced arrhythmias and have been linked to SUDEP (reviewed in [5] and see section below) (Table 3).

Genetics of SUDEP and the molecular autopsy

In cases where the scene investigation, autopsy and toxicological studies do not reveal a probable or definite cause of death, post-mortem genetic testing can reveal a variant or mutation that could cause sudden death. These ‘molecular autopsies’ emerged from studies in which sudden cardiac death was suspected but only confirmed when post-mortem DNA analysis revealed genetic findings that established the likely mechanism of death. De novo or known pathogenic mutations in cardiac channelopathy or cardiomyopathy genes were found in 27% of children and young adults with unexplained sudden cardiac death [95]. While molecular autopsy is well established in evaluating cases of suspected sudden cardiac death, its role in defining the mechanism of death in epilepsy patients remains less well defined. Unlike sudden cardiac death, where pathogenic variants in genes that modulate the cardiac rhythm and myocardium are associated with sudden death and have a well-characterized lethal mechanism, defining a SUDEP gene is more challenging. In the largest study to date, 61 SUDEP cases underwent post-mortem whole exome sequencing. The investigators identified mutations in cardiac arrhythmia (all long QT) genes (7%), candidate pathogenic variants in dominant cardiac arrhythmia genes (15%) and mutations or candidate pathogenic variants in dominant epilepsy genes in 25% [96]. The unanswered question is the potential pathogenic relationship and role of these mutations in SUDEP.

Some genetic mutations are associated with an increased risk of SUDEP in humans and in some animal models of SUDEP. However, most of these mutations are established causes of epilepsy that are often treatment-resistant and associated with frequent GTCS (e.g., SCN1A, SCN8A, DEPDC5, dup15q; Table 3). The epileptic encephalopathies are associated with high rates of sudden death and status epilepticus [97]. In Dravet syndrome [98], the mortality rate/1000-person-years was 15.8 and SUDEP rate was 9.3/1000-person-years. These rates are more than 10-fold higher than other paediatric epilepsies but similar to cohorts of adults with refractory epilepsy being evaluated for surgery (6-9/1000-person-years) [99].

The association of these genes with SUDEP could result solely from their role in causing severe epilepsy with frequent convulsive seizures and status epilepticus (e.g., SCN1A, SCN8A, dup15q) or epilepsy of variable severity (e.g., DEPDC5). However, specific genetic mutations may also independently contribute to the mechanism of death by facilitating autonomic instability, cardiac or respiratory dysfunction, or more prolonged postictal depression of arousal following a seizure or interfering with post-ictal autonomic recovery. Experimental models of Dravet syndrome reveal parasympathetic hyperactivity accompanies SUDEP and death can be prevented by parasympathetic blockade [100]. The relevance of this to human Dravet patients is uncertain as we lack autonomic data related to SUDEP, Dravet syndrome, and the broader epilepsy population. Marked sympathetic hyperactivity (elevated electrodermal activity) was documented post-ictally in a 20-year-old SUDEP case (Picard et al 2017 Neurology In press). Sympathetic hyperactivity is well documented after most GTCSs in animals and humans. The mechanism of death in SUDEP may be related to postictal abnormalities of both parasympathetic [101] (e.g., bradycardia, hypotension) and sympathetic systems [5].

The challenges in linking an epilepsy gene to SUDEP are highlighted by DEPDC5. This gene codes for a member of the IML1 protein family involved in G-protein signalling networks. The protein coded by DEPDC5 is a component of the GATOR1 complex that inhibits the mTORC1 pathway and can cause autosomal dominant familial focal epilepsy of variable severity. Pathogenic variants of DEPDC5 were enriched in the largest genetic study of a SUDEP cohort [96]. In a series of 61 SUDEP cases, five cases had DEPDC5 variants; three were nonsense mutations that were very likely pathogenic while two were missense variants, one of which is highly conserved in evolution. Yet DEDPC5 is the most common gene associated with familial focal epilepsy and is a common gene for sporadic focal epilepsy [102]. The challenges of differentiating genetic from other factors in the mechanism of death is exemplified by the two brothers with DEPDC5 mutations who both died from SUDEP but both had histories of AED nonadherence [103]. Since DEPDC5 patients often have frontal lobe foci and frontal lobe deficits, this can potentially impair planning and judgement, and this gene may increase SUDEP risk by causing treatment-resistant epilepsy as well as increasing the risk of life-style factors such as sleep deprivation, medication nonadherence and excess alcohol use.

SUDEP is not limited to treatment resistant epilepsies. SUDEP rarely complicates benign epilepsy of childhood with centrotemporal spikes and a history of convulsions not treated with antiepileptic drugs [104]. Further, one third of young children who die suddenly and have unrevealing comprehensive death investigation, have a history of febrile seizures (ten-fold greater than the general population), suggesting seizure as a mechanism of death [67]. Thus, any patient with seizures is likely at increased risk for SUDEP post-ictally. Larger genetic studies are needed comparing SUDEP cases and control epilepsy populations matched for epilepsy severity, to determine if epilepsy genes increase SUDEP risk beyond the increased risk related to epilepsy severity.

As detailed above SUDEP is also associated with variants and mutations of neuro-cardiac channelopathy genes. The presence of LQT gene mutations in SUDEP suggests that a seizure could provoke a lethal cardiac arrhythmia, or sudden cardiac death could occur independently of a seizure. Both possibilities may occur in some cases, although the available literature on deaths recorded in epilepsy monitoring units do not identify a single case where a seizure led to lethal arrhythmia or where a lethal arrhythmia spontaneously occurred [12]. However, several near-SUDEP cases had arrhythmias and it is possible that patients with LQT gene mutations and epilepsy have treatment-responsive epilepsy and are rarely admitted to epilepsy monitoring units. Further, the extensive literature on cardiac changes during and after seizures recorded in epilepsy monitoring units very rarely identifies potentially lethal seizure-induced cardiac arrhythmias [94, 105, 106]. The most serious cardiac rhythm changes associated with seizures are ictal asystole, atrial fibrillation, and ST-wave depression. These are less often associated with mutations in cardiac genes than torsade des pointes ventricular tachyarrhythmias. However, abnormalities in cardiac repolarization on ECG are more frequent in SUDEP than control epilepsy cases [107], suggesting a potential role of cardiac gene variants in SUDEP. As noted above, pathogenic and rare variants of cardiac channel genes are found in SUDEP cases [108]. Moreover, all channelopathy genes are expressed to some extent in the brain, including the brainstem. Thus, if these cardiac channelopathy genetic variants increase SUDEP risk, they could do so by alterations in the heart or brain, or both organs. The extremely rare occurrence of ventricular tachyarrhythmias in recorded SUDEP cases, or during seizures recorded in epilepsy monitoring units, suggest that if cardiac channel genes contribute to SUDEP pathogenesis, it may be via a different mechanism than their role in sudden cardiac death. Finally, some or many variants in cardiac genes may be unrelated to SUDEP pathogenesis.

The future and biobanking in SUDEP

There is current momentum from professional groups, as well as patient support groups, to move from mere recognition and documentation of SUDEP to identifying its causes and preventing its occurrence [109]. Despite an initial ‘head start’ by pathologists in recognising SUDEP, human tissue and cell based research programmes in ERD have somewhat lagged behind. The reasons are complex and manifold: (i) lack of clinical recognition and importance of the problem to allocate proportionate research funding in this area, (ii) establishing research collections in forensic setting where priorities and human tissue legislation is primarily geared toward establishing cause of death, with limited or restricted capacity to interact with research programmes, (iii) problems obtaining consent from immediately bereaved families, (iv) lack of easy access to a local or referral neuropathology services in some regions [16], (v) problems with post-mortem decomposition of autopsy tissues [80] with many epilepsy research programmes focussing resources on optimally preserved surgical tissues, (vi) lack of sufficient numbers of suitable control cases and tissues, (vii) lack of standardised guidance documents for conducting SUDEP autopsies, the core practice of which can vary between regions and jurisdictions. This last point is pertinent, as despite a flurry of clinical practice documents on SUDEP aimed at neurologists and health care professionals (14 from the first half of 2017 alone [6, 7, 9, 110–120] there are very few directed at the pathologist regarding the elements and scope of the post-mortem examination. There are no consensus international best practice guidelines (in contrast to surgical epilepsy neuropathology practice [32, 33, 37] and informal enquiry indicates that very few countries have any form of autopsy guidelines for investigations in ERD. In the UK guidance documents for the approach to post-mortem examination, tissue sampling protocols (Figure 1 ; Table 4) and classification of cause of death in suspected epilepsy and sudden cardiac cases are issued by the Royal College of Pathologists series (https://www.rcpath.org/profession/clinical-effectiveness/clinical-guidelines/autopsy-guidelines.html) ; these provide guidelines and do not mandate practice and are based on current evidence available in ERD and SUDEP.

In establishing the Centres for SUDEP research [121], the extent of the logistical problems and existing barriers to biobanking in ERD and SUDEP became apparent. These obstacles need to be addressed in the future and invested in to accelerate research into this condition. All biobanking should be conducted in full agreement with Coroners and Medical Examiners and following discussion and informed consent of the families (Figure 4). Local practical and economical strategies to more accurately determine the cause of death in ERD should utilise regional expertise in neuro and cardiac pathology to augment the investigation with timely and detailed reports (Figure 4). Post-mortem imaging, including MRI, CT and angiography is likely to play an increasing role in sudden death examination [122, 123], where it may supplement standard autopsy procedures by revealing lesions difficult to assess by macroscopic examination. There is no published study applying post-mortem imaging in ERD, however, high resolution MRI of whole brains or regions provides more detailed information of structural brain abnormalities enabling correlation with histological findings as ongoing areas of research [124, 125]. Prospective collection of DNA from brain, heart and lungs could also expand the range and scope of investigations. Techniques such as single cell RNAseq, proteomics usually requiring fresh frozen tissue, will provide complementary data on gene expression ; however, with newer technologies these methods are feasible with even formalin fixed, paraffin-embedded (FFPE) [126]. Greater availability of bio-specimens could facilitate numerous studies, including 1) systematic comparison of DNA exomes derived from blood and brain to examine the role of possible somatic mutations, 2) brainstem RNAseq in areas of interest (e.g., medullary raphe and pre-Bötzinger nuclei), 3) detailed examination of the cardiac conduction system. Such systematic tissue biobanking may enable better stratification of cause of death in SUDEP through integrated clinical, imaging, pathological and molecular diagnosis (Figure 4).

In summary, the evidence indicates that SUDEP is a heterogeneous condition in terms of both the underlying epilepsy neuropathology, genetics as well as mechanism of death. SUDEP at present provides a convenient ‘umbrella’ term under which to collect, stratify and further investigate these cases to shed light on any divergent or convergent pathogenetic mechanisms. Hopefully, in the future, epilepsy centres, neuropathologists, cardiac pathologists and medical examiners, coroner offices, and government funding agencies will collaborate in establishing international biorepositories to advance SUDEP research.