Background
Sudden unexpected death in epilepsy (SUDEP) represents an interesting conundrum of the heart-brain interface that is the focus of intense research. The exact pathophysiologic processes were unknown. Autonomic changes affecting the cardiac rhythm may play a role, although the precise pathways involved are still unclear [
1]. The cardiac effects of epilepsy are widespread and range from subtle changes in heart rate variability (HRV) to ictal sinus arrest.
HRV reflects the beat-to-beat alterations in the HR and is mainly modulated by parasympathetic and sympathetic activity [
2]. HRV can be used as a tool to show information on the functional state of the autonomic nervous system. And HRV is a mirror of neuronal influences on the cardiac pacemaker as one of the important functions of the autonomic nervous system [
3]. It was found to be lower with refractory epilepsy, possibly resulting from parasympathetic or vagal reduction. This can make patients more susceptible to tachycardia and fibrillation and possibly SUDEP [
4,
5]. Although IB in HR during epileptic seizures can occur [
6], the most common abnormality reported is Ictal tachycardia (IT) [
7‐
9], which has been reported following CPS and SPS [
4]. IT has been reported in up to 100% of seizures and can precede, coincide with, or follow ictal discharges. Analysis shows that HR increases significantly in CPS-G compared to non-generalized seizures [
10]. More insight into the mechanisms of generalization of partial seizures may improve treatment, and possibly also contribute to a reduced number of SUDEP, which has been reported previously [
11]. HR increase seems more prominent in temporal lobe seizures (where insular spread is common), compared with extratemporal seizures [
12‐
14]. Whether there is a central hemispheric lateralization of cardiac autonomic control is debatable. A predominance of increase of HR as compared to baseline in right sided temporal seizures can be expected and is confirmed by some [
15,
16], but others describe a more prominent HR change during left-sided seizures [
10]. To clarify these questions, we analyzed the frequency of cardiac rhythm related to seizures and attempted to determine risk factors and associated clinical characteristics.
Apart from SUDEP, mortality and morbidity as a result of seizure-related events (e.g., accidents and drowning) are frequent. As the occurrence of seizures is unpredictable, much research was devoted into prediction or early detection of seizures. Detection of seizures could be very helpful not only in the development of warning systems but also in novel treatment strategies [
17].
Methods
Subjects
We analyzed HR changes on 181 consecutive partial seizures from 81 epileptic patients referred to the Department of Neurosurgery of Liaocheng People’s Hospital, Beijing Tiantan Hospital, and Beijing Neurosurgical Institute at Capital Medical University for presurgical evaluation and scheduled as candidates for surgery. Between January 2011 and June 2013, most of these patients were admitted to our unit to determine their suitability for epilepsy surgery. None of the subjects suffered from cardiovascular disease. Status epileptics have shorter interval time between two seizures, the time period from 60 seconds before EEG seizure onset to 60 seconds after seizure offset is not easy to define. Therefore, status epileptics were excluded from the analysis. For three groups of seizures (SPS, CPS and CPS-G), consecutive events providing satisfactory video – electroencephalograph – electrocardiograph (EEG-ECG) recording were analyzed. The patients with at least one qualifying event were entered consecutively. The number of seizures analyzed per patient ranged from 1 to 4 (average 2.2). The study conformed to the Declaration of Helsinki, was approved by the ethics committee of the Beijing Tiantan Hospital, Capital Medical University.
Data
All the available inpatient video-EEG-ECG data for each subject were reviewed. Digital EEG was recorded with a common reference at a sampling frequency of 256 Hz. The EEG data were obtained by using the international standard 10–20 system electrode placement and sphenoidal electrodes with one lead of ECG monitoring. ECG was recorded (sampling frequency 256 Hz, low-frequency filter 0.3 Hz, high-frequency filter 70 Hz). The non-lateralized patients were recorded with indwelling depth electrodes. HR was assessed by manual analysis. HR were the beats per minute at different time epoch in each seizure out of EEG-ECG which was recorded within 10 seconds in each screen, then count the number of QRS complexs within 10 seconds, multiply the result by 6. The baseline ECG rates were determined for 10 second epochs and expressed as beats per minute, and we measured the HR for the 60 seconds that preceded the seizure onset as the baseline ECG. The maximum and minimum HR during seizures were similarly identified [
18]. HR was calculated at the 60S, 40S, 30S, 20S, 10S before seizure onset, ictal onset, ictal, and 60S after seizure offset. Seizure onset was defined as EEG seizure onset. The first 10s of the ictus (herein called ‘ictal onset’), seizure offset was defined as the end of EEG seizure end. In CPS-G seizures, ECG signals were commonly obscured by muscle and movement artifacts. We found that near the end of the tonic-clonic phase, ECG signals were less susceptible to artifacts interference. Additionally, we can get data by reducing the sensitivity of scalp electrodes, increasing the sensitivity and low-frequency filters of ECG, and debasing high -frequency filters of ECG. Therefore, ictal HR was possible to obtain. The video-EEG data evaluated for each recorded seizure are (a) the location of the ictal onset zone, (b) the seizure duration, (c) the seizure type, and (d) the sleep/wakefulness state during which the seizure arose. All the ECG data were reviewed by a board-certified cardiologist and electrophysiologist. Magnetic resonance imaging (MRI) of the brain for each patient was obtained from a review of the medical records.
Type (IT and IB) of HR changes and curve graph
IT and IB were defined as development of a heart rate more than 100 and less than 60 beats/min respectively after onset of seizure activity on the EEG. We measured the HR for the 10, 20, 30, 40, 60 seconds before the seizure onset, ictal onset, ictal and 60 seconds after seizure offset. We measured the HR for the 60 seconds that preceded the seizure onset as the baseline HR. The average of the difference between the HR at different time and the baseline HR can tell us whether it produce ictal HR changes. The HR changes were evaluated in the stage of epilepsy as time varies. Curve graphs of the seizures could be drawn based on the HR changes and HR.
Statistical analysis
SPSS for Windows version 11.0 (SPSS, Chicago, IL, USA) was used for the statistical analysis. A mean baseline HR (B-HR) and maximum and minimum HR were estimated during seizures. All reported variables were assessed with a non-parametric variable test (One-Sample Kolmogorov–Smirnov), assessing if it was distributed normally. The difference between gender, hemispheric lateralization and sleep/wakefulness on the HR changes, and the difference between ictal and postictal HR during sleep and CPS-G sleep separately were assessed with an independent two–sample t-test. Analysis of variance (ANOVA) was used in comparing the mean HR difference. Post hoc correction for multiple testing has been performed. Spearman’s rank and Pearson's correlation coefficient were used. A multivariate analysis was performed by means multiple linear regressions to assess the influence of the indicators on the HR difference. The process of variable selection was stepwise forward, at the level of 5%, which selected the smallest subgroup of the independent variables that influenced the HR difference. P < 0.05 was considered statistically significant.
Discussion
Addressing the issue of cardiac changes during ictal phase is challenging, the results of previous studies were different. The expected consequence of seizure activity is an increased HR as compared to baseline due to sympathetic discharge, furthermore, study showed that cardiac rhythm abnormalities occur during or after seizure activity in patients with refractory epilepsy. Seizure-related disturbances of cardiac rate, rhythm, and conduction have been identified in patients with epilepsy, with IT being the most frequent [
19]. The consequence of a sustained tachycardia during epileptic seizures could be a ventricular tachycardia and sudden death. This is especially significant for individuals underlying cardiac disease [
20]. These results suggest the possibility of increase SUDEP during seizure period, which is introduced by previous studies [
21,
22].
We noted that seizures are associated with a marked increase in HR in most cases as noted previously [
19]. As shown in Table
1, there were differences in the HR changes between SPS, CPS and CPS-G (
P < 0.01). We observed that CPS-G displayed a significantly higher ictal and postictal HR difference than SPS and CPS, which has been reported previously [
23]. The observed alterations could potentially facilitate sudden cardiac death and might contribute to the association of SUDEP in epilepsy with CPS-G. There is evidence of a hemispheric-specific organization of this response as shown in the depth electrode studies by Oppenheimer et al. With the pressor response lateralized to the right and depressor response to the left hemisphere [
24]. As illustrated in Table
1, the HR in the right focus showed a significant increase over that in the left focus (
P < 0.05).
Another major finding of our study was a central imbalance of autonomic cardiac control in men only. From published reports, there is evidence for gender differences in cardiac sympathovagal regulation. HR changes in women have been shown to be significantly less than in men as a result of a reduced sympathetic influence on the heart [
25]. In our study, the HR changes of 124 male seizures were higher than those of 57 female seizures (
P < 0.001).
The relationship between the circadian clock and epilepsy might provide additional understanding of the mechanisms that induce seizures to occur at certain times [
26]. As shown in Table
1, a total of 72 (39.8%) of the seizures occurred in sleep, and 109 (60.2%) occurred in wakefulness. A previous study analyzed fourteen patients who died during sleep; two patients were awake among 16 SUDEP cases. Greater increases in HR were associated with seizures arising from sleep (78 BPM increase) than from wakefulness (47 BPM;
P < 0.001) in SUDEP cases [
27]. Our study showed significant differences in HR increases which can be observed during seizures arising from sleep more frequently than during seizures arising from wakefulness (the mean increase in HR, 63 BPM in sleep and 47 BPM in wakefulness,
P < 0.001) (Table
1). As illustrated in the Figure
3, we also find that CPS-G seizures tend to arise from sleep. However, a recent study illustrated that seizures arising from sleep which display higher HR changes are somewhat trivial, as baseline and preictal HR are lower during sleep because of higher vagal tone [
28]. To minimize the bias due to sleep-related prominent vagal tone, we have compared the difference during 72 sleep and 17 CPS-G sleep seizures between postictal and ictal HR respectively. Ictal HR were higher than postictal HR in the 72 sleep seizures (P < 0.001). In the CPS-G seizures, ictal HR were higher than postictal HR (P < 0.001).
The circadian distributions of seizures arising from individual regions among seizure localizations have been poorly described in the literature. Our findings confirm that there were differences in the HR changes during wakefulness and sleep in frontal, temporal and occipital lobe seizures (
P < 0.01) (Figure
1).
In our report, we observed that side (left/right), type (SPS, CPS and CPS-G) and state (sleep/wakefulness) were independently related to HR differences by multiple linear regression analysis (Table
6). One limitation of our regression model is the relatively few seizures after the larger dataset was divided into the various seizure types, frequently referred to as the “curse of dimensionality”. It was for this reason that we could look at sleep/wakefulness, left/right, and types of seizures together, but not add location to that analysis. Hence, we were unable to determine if a particular seizure type was more likely to occur in a specific location and whether this location differed depending on the sleep state. Our results add to the available literature by providing additional details on seizure types. Of the 181 seizures with ictal ECGs, 152 seizures (83.98%) were associated with IT. We have documented IT during both left 80 and right 72 onset seizures among 152 seizures. From Table
3, we can add location to that analysis. Whether there is a central hemispheric lateralization of cardiac autonomic control is debatable. a recent intracranial EEG-study comparing HR modulation during left-and right-sided seizures within the same patients has not found a group effect of the side of seizure-onset [
29]. However, our results confirm that the HR of the patients with a right focus showed a significant increase over the HR of the patients with a left focus (
P = 0.01), which has been reported previously [
15,
16]. The mean HR differences of SPS, CPS and CPS-G were (47.5 ± 24.9), (51.2 ± 19.8) and (80.5 ± 21.3) BPM, respectively. The HR changes in CPS-G seizures were the highest of the three types (
P < 0.01) (Table
3), which has been reported previously [
23,
11].
The mean HR differences of sleep and wakefulness were (67.4 ± 22.9) and (53.1 ± 24.2) BPM. The HR changes during sleep are higher than in wakefulness (
P < 0.01) (Table
3), as in previous studies [
30].
Epileptic discharges of inpatients with seizures are thought to propagate to the central autonomic network and change or disturb normal autonomic control of vital cardiac functions. This activation of the central autonomic nervous system is responsible for the periictal autonomic cardiac symptoms observed in patients with epilepsy. As we know, HR changes can precede clinical and encephalographic seizure onset, and early detection of these changes can have an application in seizure detection systems [
17]. Similarly, changes in vagus nerve activity—indicative of the autonomic control of the heart—could be used as seizure predictors.
So far, nobody shows the regularity of HR changes in seizures. Our study confirms that the HR in the stage of epilepsy has a regular evolution varying with time (Figures
4,
5). According to Figures
4 and
5, IT was divided into three types based on the HR at each time in each seizure. We also find that in a total of 152 investigated IT seizures, HR increases rapidly (within 10 s before seizure onset and ictus) and typically slows to normal with seizure offset (Figures
4,
5). In previous reports, ictal HR increases as compared to baseline preceded seizure onset on surface EEG for 8–19 s [
7,
15]. In our study population, the time lag was 10 s, conforming to the time window previously. Our curve graphs can be more intuitive to show HR changes regularity, seizure detection and help to clarify further the basic mechanisms of interactions between heart and brain. Similarly, identification of these early autonomic manifestations in seizures can theoretically contribute in developing new treatment strategies based on seizure detection for patients with refractory seizures. If further investigation validates such techniques, changes in HR might be included in seizure warning systems or even in “on demand” electrical stimulation therapies such as vagus nerve stimulation (VNS). Therefore, the monitoring of cardiac activity could predict on-coming seizures prior to their EEG onset, which could theoretically increase the quality of life and treatment efficiency in at least some patients with refractory epilepsy.
Limitations
There are several limitations to this retrospective study, and much work remains. For future studies, it would be desirable to have more detailed information regarding certain possible risk factors for SUDEP such as EKG abnormalities, seizure etiology, IQ test performance and AED doses and serum levels.
Competing interests
The authors declared no potential conflicts of interest with respect to the research, authorship, finance and/or publication of this article.
Authors’ contributions
Author contributions to the study and manuscript preparation include the following. Conception and design: WC, FGM. Acquisition of data: WC, FGM, CLG, PSZ, JGZ, HQ. Analysis and interpretation of data: CL, PSZ. Drafting the article: WC, FGM. Critically revising the article: all authors. Reviewed final version of the manuscript and approved it for submission: all authors. Statistical analysis: CL, CLG, PSZ. Study supervision: FGM.