The probability of sudden unexpected death in epilepsy given postictal prone position

Abstract

Sudden Unexpected Death in Epilepsy (SUDEP) is strongly linked to prone position in the immediate aftermath of a generalized tonic-clonic seizure (GTCS). The risk of SUDEP after a GTCS resulting in prone position has not yet been estimated. We calculated the probability of SUDEP given prone position after a GTCS using Bayesian analysis with inputs obtained from known risk of SUDEP and probability of prone position after a GTCS. The risk for SUDEP given the prone position is estimated to be 0.41% (95% Credible Interval 0.13–0.69%). The relative risk of SUDEP in prone vs. non-prone position following a GTCS is estimated to be 63 (95% Credible Interval 30–96). Sudden Unexpected Death in Epilepsy might be prevented by repositioning the patient after a seizure to avoid the prone position.

1. Introduction

Sudden Unexpected Death in Epilepsy (SUDEP) is the leading cause of death directly related to epilepsy, and preferentially affects young people [1]. Despite its devastating consequences, the underlying mechanism of SUDEP remains unknown, impeding prevention. Although we do not completely understand the underlying mechanisms of SUDEP, population-based studies, case-control studies, and inpatient video-EEG monitoring information have greatly advanced understanding of the circumstances surrounding SUDEP. The most consistent risk factor for SUDEP is frequent generalized tonic–clonic seizures (GTCS) [2]. Most deaths due to SUDEP occur in sleep [3]. Postictal prone position has been consistently reported in association with SUDEP [3,4]. However, the risk of SUDEP after a GTCS resulting in prone position has not yet been estimated. Although the exact relationship between prone position and SUDEP is unclear, prone position likely contributes to SUDEP as a result of fatal airway obstruction [5]. To further evaluate the risk of SUDEP associated with prone position, we used a Bayesian analysis to estimate the conditional probability of SUDEP after a GTCS resulting in prone position.

2. Methods

2.1. Application of Bayesian analysis

Our primary interest is the probability of SUDEP resulting from a GTCS where the patient is found in prone position; this cannot be observed directly, but can be estimated using a Bayesian analysis. In our case, the events we consider are GTCS, and we define the following subsets of these events:

• D: events that result in SUDEP,

• D⁻: events that do not result in SUDEP,

• E: events for which the patient is found in prone position,

• E⁻: events for which the patient is found in a non-prone position.

Using standard notation for conditional probabilities, we write the probability of SUDEP resulting from a GTCS where the patient is found in prone position as P(D|E). Bayes’ theorem then gives

$$P(D\mid E)=\frac{P(E\mid D)*P(D)}{P(E)}.$$

Using the law of total probability, this can be written in terms of known quantities

$$P(D\mid E)=\frac{P(E\mid D)*P(D)}{\left[P(E\mid D)*P(D)+P\left(E\mid D^{-}\right)*P\left(D^{-}\right)\right]}.$$

Similarly, for P(D|E⁻) we can write

$$P\left(D\mid E^{-}\right)=\frac{P\left(E^{-}\mid D\right)*P(D)}{\left[P\left(E^{-}\mid D\right)*P(D)+P\left(E^{-}\mid D^{-}\right)*P\left(D^{-}\right)\right]},$$

with

• P(D⁻): probability of GTCS not resulting in SUDEP,

• P(D): probability of GTCS resulting in SUDEP,

• P(E|D): probability being found prone in the setting of SUDEP,

• P(E|D⁻): probability of being prone after GTCS not resulting in SUDEP,

• P(D|E): probability of SUDEP when the patient is prone at the end of a GTCS, our main value of interest.

2.2. Credible intervals

We used a Monte Carlo simulation to determine 95% credible intervals for the quantities P(D|E) and P(D|E⁻), and the relative risk of SUDEP given prone position vs. non-prone position. The proportions in the simulation were modeled as scaled binomial distributions.

3. Results and application of prior information

3.1. P(D) and P(D⁻)

The incidence of SUDEP based on population-based studies is approximately 1.2 (95% CI 0.64–2.32) per 1000 patients per year [2]. Patients with drug-resistant epilepsy who have frequent GTCS, especially during sleep, have the highest risk for SUDEP [2]; however, SUDEP can occur in patients with well-controlled epilepsy or those who have relatively few GTCS. Moreover, considering the relatively large size of the population of patients with milder epilepsy, more patients in this group would suffer SUDEP compared to the smaller population of patients with more frequent GTCS. In a recent population-based study, the average annual seizure frequency was 8.10 (95% CI 4.68–11.52) per year in probable SUDEP group compared to 6.07 (95% CI 4.77–7.37) seizures per year in the control group of patients with epilepsy [6]. For P(D), SUDEP incidence rate of 1.2 (95% CI 0.64–2.32) per 1000 patients per year, an average of 4 GTCS per patient per year gives an estimate of P(D) of 0.0003 and P(D⁻) of (1–0.0003) = 0.999.

3.2. P(E|D)

Although, the exact role that prone position plays in pathophysiology of SUDEP has yet to be determined, there is a significant association between prone position and SUDEP. In the retrospective study of mortality in epilepsy monitoring units (MORTEMUS), the body position was documented in 13 of the 16 patients suffering SUDEP, and all were exclusively found in prone position [7]. In several recent population-based studies, approximately 69–74% of patients who suffered SUDEP were found in prone position [8–10]. Studies using meta-analysis of SUDEP literature have also shown similar results: 73.3% (95% CI 65.7%, 80.9%) of patients who died from SUDEP were found in prone position. The prone position at the time of SUDEP was observed in 85.7% (95% CI 74.6%, 93.3%) of patients who were younger than 40 years old and was up to 87.6% (95% CI 81.1%, 94.2%) when SUDEP occurred during sleep [3,4]. Thus, we will estimate P(E|D) to be between 68.6% (95% CI 61.6%, 75.3%) (lower limit) to 87.6% (95% CI 81.1%, 94.2%), particularly because night-time SUDEP is a feared entity as there is less likelihood of nearby assistance.

3.3. P(E|D⁻)

Despite the high percentage of SUDEP patients found in prone position, the incidence of prone position in the postictal state of GTCS is low: a retrospective study of the video-EEG data of 193 patients with 308 generalized tonic-clonic seizures in the epilepsy monitoring units showed 13 patients were in the peri-ictal prone position, though all but 4 either spontaneously or with nursing intervention turned away from the prone position. Patients were prone in 4 of 96 seizures without any ictal interventions, of which one resulted in SUDEP, thus yielding 3 of 95 seizures who were prone and alive (3.16%; 95% CI 0.66–8.95%) [11]. A smaller video-EEG study of 180 GTCS in 90 patients revealed patients were in a prone position in 13 GTCS (7.22%; 95% CI 3.90–12.03%) at the time of nursing intervention [12]. As such, we estimate P(E|D⁻) to be 0.0316–0.0722.

3.4. P(D|E)

We calculated the probability of SUDEP given prone position in the postictal state of a GTCS using Bayes’ theorem and applying the prior information stated above (Table 1). We obtained probability of SUDEP in the setting of a prone position P(D|E) = 0.0041 (95% Credible Interval 0.0013–0.0069), with a relative risk of SUDEP in the prone vs. non-prone position of 63 (95% Credible Interval 30–96).

Table 1.

Summary of prior information used to determine the probability of sudden unexpected death in epilepsy given postictal prone position. SUDEP: sudden unexpected death in epilepsy; GTCS: generalized tonic-clonic seizure.

Prior information	Value	References
SUDEP incidence rate	1.2 (95% CI 0.64–2.32) per 1000 patients per year	2
Average number of GTCS per patient per year	6.07 (95% CI 4.77–7.37)	6
Percentage ofpatients withprone position at the time of SUDEP	68.6% (95% CI 61.6–75.3%)–87.6% (95% CI 81.1–94.2%)	4, 10
Percentage ofpatients withprone position after a survived GTCS	3.16% (95% CI 0.66–8.95%)–7.22% (95% CI 3.90–12.3%)	11,12

4. Discussion

The exact role of prone position in SUDEP remains unclear; however, SUDEP commonly occurs in the prone position. Utilizing parameters from prior literature, we report that the risk for SUDEP in the setting of the post-convulsion prone position is estimated to be between 0.1% and 0.7%, with a relative risk of SUDEP in the prone vs non-prone position of 30–96. Although it is possible that prone position may be associative rather than mechanistic, due to version to a prone position after a severe GTCS, it is more likely that prone position contributes to SUDEP through airway obstruction; in vulnerable infants, increase in central apnea and cardiorespiratory instability have been attributed to the prone position [10,13,14]. As such, SUDEP might be prevented by repositioning the patient after a seizure to avoid the prone position. Having night-time supervision or an alarming system for the caregivers can decrease the risk of SUDEP, suggesting that simple interventions such as turning and stimulating the patient can decrease the risk [13]. Given the fact that both SUDEP and postictal prone position overall occur rarely, the significant association between the two warrants further study to reveal the nature of this association.

The probability calculated here is dependent on the assumptions and parameters driven from prior literature and should be interpreted in the proper context. An exception to this model is patients with severe epilepsy and frequent GTCS who have a high risk of SUDEP. However, many deaths due to SUDEP occur in patients with milder types of epilepsy who have about or less than 4 GTCS per year. The risk of prone position for SUDEP may be underestimated in this study due to the lack of available data in the community setting of the probability of prone position after a survived GTCS. These data are only available in the inpatient monitoring setting, where postictal nursing intervention may be a contributing factor for the patients who survived the prone position after a GTCS. In contrast, this study uses data available from population-based studies in the community setting to obtain the probability of prone position after SUDEP.