Address correspondence to Roland D. Thijs, SEIN-Epilepsy Institute in the Netherlands Foundation, PO Box 540, 2130 AM Hoofddorp, The Netherlands. E-mail: email@example.com
Purpose: Most people with epilepsy who die suddenly and whose death is attributed to sudden unexpected death in epilepsy (SUDEP) are found in or by the bed for unknown reasons. We assessed whether those with sleep-related SUDEP were more likely to have nocturnal seizures, and whether seizure patterns (diurnal vs. nocturnal) differed from people dying suddenly and living controls with epilepsy.
Methods: Seizure patterns in a cohort of 154 people with epilepsy who died suddenly and after autopsy conformed to the definition of SUDEP and 616 controls living with epilepsy were classified as having “exclusively diurnal” or “nocturnal seizures.” Comparisons were made between the groups. SUDEP was classified as sleep-related or non–sleep-related based on eyewitness accounts and the circumstances surrounding death.
Key Findings: SUDEP was primarily a sleep-related (58%) and unwitnessed (86%) event. If sleep-related, SUDEP was more likely to be unwitnessed [odds ratio (OR) 4.4, 95% confidence interval (CI) 1.6–12]. Those with sleep-related SUDEP were more likely to have a history of nocturnal seizures than those who had non–sleep-related SUDEP (OR 3.6, 95% CI 1.4–9.4). Those who died were more likely to have a history of nocturnal seizures than living controls (OR 3.9, 95% CI 2.5–6.0). After correction for previously established SUDEP risk factors (Langan et al., 2005), the presence of nocturnal seizures remained significant (OR 2.6, 95% CI 1.3–5.0).
Significance: Nocturnal seizures seem to be an independent risk factor for SUDEP. These findings underscore the importance of preventive measures, which may include night supervision.
SUDEP is usually not witnessed or monitored and therefore little is known about its pathophysiology (Langan et al., 2000). Despite all efforts to identify risk factors from descriptive epidemiologic studies, predicting and also preventing SUDEP is thus still challenging, particularly on an individual basis.
By contrast, in many cardiovascular conditions (e.g., sudden cardiac death [SCD] and in sudden infant death syndrome [SIDS]), a peak incidence in the early morning has been found (Elliot, 2001; Daltveit et al., 2003).
We assessed whether people with sleep-related SUDEP were more likely to have nocturnal seizures. We ascertained whether seizure patterns (diurnal vs. nocturnal) differed between those with SUDEP and controls living with epilepsy.
Selection of cases and controls
The cohort in this study has been reported previously (Langan et al., 2005). There were 154 autopsy confirmed SUDEP cases and 616 age- and sex-matched controls living with epilepsy. In short, cases were identified by coroners, neurologists, and self-referred family members. Each case had four controls matched for age (±5 years) and geographic location. For cases, historical data were obtained from general practitioners, hospital consultants, and through semistructured interviews with family members. For controls, data were extracted from medical records. For both cases and controls, 10% of entries were randomly chosen and the accuracy of data retrieved from notes and data entry into the database was checked. The error rate was <5%.
Information on seizure patterns was recorded using questionnaires that were sent to the treating physician of each participant. Seizure patterns were classified as “diurnal seizures” (seizures occur exclusively during the day) and “nocturnal seizures” (seizures also occur during the night).
Circumstances of SUDEP
The circumstances surrounding death were analyzed to determine if SUDEP had been sleep-related or not. In witnessed SUDEP cases, eyewitness accounts were used to this end. In unwitnessed cases, SUDEP was assumed to be sleep-related if patients were found in bed. If they were found by the bed additional evidence of a recent seizure was required: lateral tongue bite, urinary incontinence, or blood near the nose or mouth. In unwitnessed cases found other than in a bedroom, SUDEP was assumed to be non–sleep-related. If information was lacking regarding the location or circumstances surrounding death, cases were labeled as “unknown” and not included in the analysis.
Time of death
Six 4-h periods were defined starting from midnight to study the circadian pattern of SUDEP. In unwitnessed cases the time of death was estimated by dividing by 2 the interval between the moment the person who died was last seen alive and when he or she was found dead. If this interval exceeded 4 h, the individual was excluded from this analysis.
Chi-square statistics were used to compare the frequency of nocturnal seizure patterns in people dying with sleep-related versus non–sleep-related SUDEP and the frequency of nocturnal seizure patterns in those who died versus controls living with epilepsy. Odds ratios (ORs) with 95% confidence intervals (CIs) were calculated. For the analysis of the circadian pattern of SUDEP, a chi-square goodness-of-fit test was used. In addition, logistic regression was used to correct for six independent predictors for SUDEP found in the previous study: history of convulsive seizures (CS), number of CS in the last 3 months, total number of antiepileptic drugs (AEDs) ever taken, current use of carbamazepine, night-time supervision, and asthma (Langan et al., 2005).
State of wakefulness in SUDEP
Most SUDEP cases were unwitnessed (133 of 154) and sleep-related (90 of 154). In almost one third (48 people) deaths were non–sleep-related, and in 16 people the state of wakefulness at time of death could not be determined. If sleep-related, SUDEP was more likely to be unwitnessed: = 9.4, p < 0.01 (OR 4.4, 95% CI 1.6–12 [Fig. 1]).
Circadian seizure patterns in sleep-related versus non–sleep-related SUDEP
In cases, seizure patterns were nocturnal in 36% (55 of 154) and diurnal in 32% (49 of 154). In controls, 17% (104 of 616) had a nocturnal pattern and 58% (357 of 616) had a diurnal seizure pattern.
Data on seizure patterns were unknown in 32% (50 of 154) of cases and 25% (155 of 616) of controls. Those who died during sleep were more likely to have a history of nocturnal seizures than those who died while awake: = 7.3, p < 0.01 (OR 3.6, 95% CI 1.4–9.4).
Circadian seizure patterns in those who died versus controls living with epilepsy
Excluding missing values, most people dying suddenly had a nocturnal seizure pattern (53%; 55 of 104), whereas most controls had an exclusive diurnal seizure pattern (77%; 357 of 461). Those who died suddenly were more likely to have a history of nocturnal seizures than controls living with epilepsy: = 38.6, p < 0.01 (OR 3.9, 95% CI 2.5–6.0). When adding the variable “seizure patterns” to the original logistic regression model, it was an independent risk factor for SUDEP. The risk of dying suddenly was higher for people with nocturnal seizures than for those with a strictly diurnal seizure pattern (OR 2.6, 95% CI 1.3–5.0).
The variable “current use of carbamazepine” became nonsignificant and was removed from the model after step 1.
Diurnal variation of SUDEP
In 40 of 154 of people who died suddenly, it was possible to narrow down time of death to a period of 4 h or less (Fig. 2). The diurnal variation of SUDEP differed significantly from an expected uniform distribution over time: = 11.3, p < 0.05.
Sleep-related SUDEP occurred most frequently in the early morning from 04:00 to 08:00 hours, whereas non–sleep-related death occurred most frequently in the periods from 08:00 to 12:00 and from 16:00 to 20:00 hours (Fig. 2).
This study provides some evidence that nocturnal seizures are an independent risk factor for SUDEP: people dying suddenly were more likely to have a history of nocturnal seizures than controls living with epilepsy. Those with sleep-related SUDEP were also more likely to have a history of nocturnal seizures than those with non–sleep-related SUDEP.
This study’s major strength is that a population-based cohort was used, so these results can be extrapolated to the general population with epilepsy. A limitation is that seizure patterns could not be accurately determined in more than one fourth of people. It could be argued that this may have affected results, since information on seizure patterns was not only obtained from medical records, but also from interviews with bereaved relatives in 17% of all SUDEP cases. No interviews took place with controls.
The proportion of missing values for this variable was, however, similar in both groups: 32% versus 25%. We, do not believe, thus, that this will have seriously biased our results. The relation between nocturnal seizures and SUDEP was confirmed in two different ways. First, a link between sleep-related SUDEP and a nocturnal seizure history was found. Furthermore, there was an association between SUDEP risk and a nocturnal seizure history. This relation remained significant after correcting for other risk factors. Therefore, we believe that the ultimate conclusions of this study are still valid. It could be argued that nocturnal seizures are a marker for seizure severity and may therefore act as a confounder: People with epilepsy with a high seizure burden are more likely to have both diurnal and nocturnal seizures. The fact that nocturnal seizures remains a significant risk factor for SUDEP after correcting for other variables including frequency of CS, however, renders this unlikely.
Impaired heart rate variability (HRV), an independent predictor of sudden cardiac death in a healthy population (Tsuji et al., 1996), was most markedly decreased during sleep in patients with epilepsy surgery (Ronkainen et al., 2005; Persson et al., 2007). To our knowledge, no studies have yet documented different effects of diurnal versus nocturnal seizures on respiratory control. It is not known whether postictal generalized electroencephalography (EEG) suppression (PGES), recently suggested as a risk factor for SUDEP, occurs preferentially during sleep (Lhatoo et al., 2010). An insufficient arousal response to autonomic changes has not yet been detailed in people dying with SUDEP, but several deficiencies in the medullary 5-hydroxytryptamine (5-HT) serotonergic arousal system have been found in children who died of SIDS (Richerson & Buchanan, 2011). Similar findings have been reported in animal models of epilepsy, but this still needs confirmation in humans (Tupal & Faingold, 2006; Uteshev et al., 2010).
This study suggests a possible link between nocturnal seizures and risk of SUDEP, which underscores the importance of further studies to determine the effectiveness of preventive measures, including night supervision. Such measures should ideally be directed toward individuals who have seizures during sleep. Future studies should focus on periictal autonomic changes in both diurnal and nocturnal seizures, as this could produce valuable new insights into SUDEP pathophysiology.
This work was partly undertaken at UCLH/UCL, which receives a proportion of funding from the Department of Health’s National Institute for Health Research Biomedical Research Centres funding scheme. RJL is supported by the Nationaal Epilepsie Fonds Nederland (project number 10-07). RDT is supported by the Christelijke Vereniging voor de Verpleging van Lijders aan Epilepsie (Nederland). JWS receives research support from the Epilepsy Society, Dr. Marvin Weil Epilepsy Research Fund, Eisai, UCB, the Wellcome Trust, the World Health Organization, National Health and Medical Research Council (Australia), the National Institutes of Health (NIH), Tuberous Sclerosis Association, and the Brain Research Trust. The authors are grateful to Dr. GS Bell for critically reviewing the manuscript.
RDT has received fees for lectures from Medtronic. JWS has been consulted by and received fees for lectures from Eisai, GSK, Medtronic, and UCB. The remaining authors have no conflicts of interest. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.