• Long QT;
  • Heart rate variability;
  • Heart rate recovery;
  • Sudden cardiac death;
  • QT dispersion


  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Summary and Conclusions
  7. Acknowledgment
  8. References
  9. Supporting Information

Purpose: To determine whether abnormal cardiac repolarization and other electrocardiography (ECG) predictors for cardiac mortality occur in epilepsy patients and whether they are associated with an increased risk for sudden unexpected death in epilepsy (SUDEP).

Methods: In a matched-pair case–control study, recordings of adult patients with pharmacoresistant focal epilepsies who died from SUDEP and who had previously had presurgical video-EEG (electroencephalography) telemetry were reviewed. Living controls were matched for age, gender, and date of admission for video-EEG telemetry. Periictal heart rate (HR), corrected QT interval (QTc), postictal HR recovery, HR variability, and cardiac rhythm were assessed. QT dispersion was analyzed with 12-lead ECG.

Results: A total of 38 patients (19 per group) had 91 recorded seizures. QTc was prolonged above pathologic upper limits in 9 of 89 seizures and 5 of 38 patients. Nine of 34 patients displayed pathologic QT dispersion. Presence of neither pathologic cardiac repolarization nor other ECG features were specifically associated with SUDEP. SUDEP patients were, however, more likely to lack pathologic cerebral magnetic resonance imaging (MRI) findings, less likely to experience antiepileptic drug reduction during telemetry, and had more secondarily generalized tonic–clonic seizures (SGTCS) per year.

Discussion: Our study did not reveal a clear-cut ECG predictor for SUDEP. Pathologic cardiac repolarization is not uncommon in adult patients with pharmacoresistant focal epilepsy and could favor occurrence of fatal tachyarrhythmia as one plausible cause for SUDEP. SGTCS are a risk factor for SUDEP, have, as compared to complex-partial seizures, a greater, unfavorable impact on heart activity, and may thereby additionally compromise cardiac function.

Sudden unexpected death in epilepsy (SUDEP) is the major cause of mortality in younger epilepsy patients, with an incidence of 6.3–9.3 per 1,000 person years in patients entering epilepsy surgery programs (Tomson et al., 2008). Epidemiologic studies have consistently identified generalized tonic–clonic seizures as a risk factor, whereas other factors are controversial (Timmings, 1993; Nashef et al., 1995; Langan et al., 2005). The mechanisms underlying SUDEP are still poorly understood, and preventive strategies, apart from good seizure control, are generally lacking. Possible mechanistic explanations include seizure-related respiratory and cardiac dysfunction. Pathologic cardiac repolarization increases the risk of fatal ventricular tachyarrhythmia and is an established predictor of sudden cardiac death (Schouten et al., 1991; Macfarlane et al., 1998; Chiang, 2004). Therefore, seizure-related prolongation of QT intervals has been postulated to be involved in SUDEP (Tavernor et al., 1996; Langan et al., 2000). Conclusive evidence of pathologic cardiac repolarization in adult epilepsy patients, however, is still lacking. We ascertained if adult pharmacoresistant epilepsy patients display pathologic cardiac repolarization or other electrocardiography (ECG) predictors for cardiac mortality during or after seizures and whether these are associated with an increased risk for SUDEP. This is an important issue as it would have implication for the formulation of preventative measures.


  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Summary and Conclusions
  7. Acknowledgment
  8. References
  9. Supporting Information

Study design

We reviewed medically refractory patients with focal epilepsies who underwent standard presurgical assessment at the National Hospital for Neurology and Neurosurgery between 1992 and 2003 and who later died of SUDEP, and in whom there was no evidence of cardiac pathology postmortem. These patients were matched to living control patients for admission date for video-EEG (electroencephalography) telemetry (on average within 2 months of the admission date of respective SUDEP patients), age, and gender, since cardiovascular parameters are influenced by these two factors. At last follow-up all control patients were still alive. The study was approved by local ethics committee.

Standard presurgical assessment

As standard practice on our video-EEG telemetry unit, recordings were performed using conventional scalp EEG recordings (10–20 system) or intracranial recordings (in two SUDEP patients) at a sampling rate of 200 Hz, with up to 25 channels for scalp and 56 channels for intracranial recordings. ECG was recorded from two channels with a modified lead I (adhesive electrodes placed below the clavicles of either side). All patients had 1.5 Tesla magnetic resonance imaging (MRI) as part of clinical presurgical assessment. Duration of telemetry was similar in both groups (Table 2). Reduction in antiepileptic drugs (AEDs) was instituted more often in control patients (Table 2). Demographic and clinical data were retrieved from patient clinical notes. Surgical outcome was classified according to the International League Against Epilepsy (ILAE) commission report (Wieser et al., 2001).

Table 2.   Demographics and seizure characteristics
 ControlSUDEPp Value95% CIOR
  1. aSignificant finding.

  2. bTotal no. of patients with available information indicated in parentheses if different from 19.

  3. AED, antiepileptic drug; B/C, bilateral or changing laterality; EXT, extratemporal (includes also seizures which arose from areas close to the temporal lobe); L, left; R, right; SGTCS, secondarily generalized tonic-clonic seizure; Temp, temporal; vs, versus.

  4. L left; R right; B/C bilateral or changing laterality; EXT, extratemporal (includes also seizures which arise from areas close to the temporal lobe ).

  5. Mean ± SD of averaged values per patient.

Age at telemetry (years)36.2 ± 9.136.1 ± 9.10.389−0.43 to 0.17 
Age at first seizure (years)9.8 ± 10.510.3 ± 7.10.871−0.06 to 0.07 
Epilepsy duration (years)26.9 ± 10.425.8 ± 10.80.700−0.09 to 0.06 
Positive MRI finding (no. pat.)1790.037a0.01 to 0.880.11
Hippocampal sclerosis (no. pat.)1250.0540.05 to 1.030.22
Febrile seizures (no. pat.)551.0  
Psychiatric comorbidity (no. pat.)6100.2200.60 to 9.022.33
Epilepsy surgery (no. pat.)960.3270.13 to 1.990.5
Total no. of seizures per year per pat.219 ± 269471 ± 8880.336−0.001 to 0.002 
No. of SGTCS per year per pat.9.4 ± 15.352 ± 40 (17)b0.035a1.00 to 1.091.044
No. of seizures during telemetry per pat.2.9 ± 2.45.7 ± 6.20.130−0.05 to 0.38 
No. of SGTCS during telemetry per pat.0.5 ± 1.41.0 ± 1.70.431−0.24 to 0.58 
Seizure clustering during telemetry (no. pat.)9130.2200.60 to 9.022.33
AED reduction during telemetry (no. pat.)16 (17)b100.0156a0 to 0.690.10
Recording time (h per pat.)99 ± 35104 ± 820.823−0.01 to 0.01 
Seizure lateralization11 L, 5 R, 3 B/C7 L, 6 R, 6 B/CB/C vs L 0.237 B/C vs R 0.731B/C vs L 0.07–1.94 B/C vs. R 0.07–6.430.36 0.67
Region of seizure-onset6 EXT, 13 Temp.7 EXT, 12 Temp0.7390.21 to 2.970.8

ECG parameters

Peri- and intraictal one-lead ECG

Where possible, intervals between two consecutive R-waves (RR interval) were analyzed up to 3 min before seizure onset, during the seizure, and up to 10 min after the seizure (Fig. S1). Maximal HR was determined by measuring the shortest RR interval. Postictal HR recovery (HRR) was assessed by subtracting preictal HR from HR 5 min after seizure cessation. HR variability (HRV) reflects the beat-to-beat variability and is modulated by the interplay of the sympathetic and parasympathetic nervous system. HRV was estimated by analysis of sequential RR intervals during a short time period (10 and 30 s epochs) in the 1 min before seizure-onset, during the seizure at the HR “plateau phase” where HR was highest, and 3 min after seizure cessation. This estimation has been established previously and is called ultra short-term HRV (de Bruyne et al., 1999; Schroeder et al., 2004). HRV was expressed as standard deviation (SDNN, in ms) or root-mean-square of successive differences (RMSSD, in ms).

QT intervals were manually measured from the start of the QRS complex to the end of the T wave (defined by the intersection with the isoelectric line). QT and preceding RR intervals were determined from 3–5 successive ECG complexes, and corrected QT intervals (QTc) were calculated using the following four formulas:

  • image(1)
  • image(2)
  • image(3)
  • image(4)

In formulas (1)–(3), QTc, QT, and RR interval values are expressed in seconds. In (4), QTc and QT intervals are in milliseconds and HR is in beats per minute (Aytemir et al., 1999). All of these correction formulas are known to lead to over- or undercorrection of QTc (Aytemir et al., 1999; Batchvarov & Malik, 2002). To reduce a bias error of putatively pathologic QTc intervals (due to higher HR during and after seizures), we used modified normal limits for QTc intervals as determined by Luo et al. (2004) and, conservatively, considered only those QTc intervals to be pathologic on which all four correction formulas agreed.

12-lead resting ECG

12-lead ECG was recorded at a paper speed of 25 mm/s and at 1 mV/cm. QRS intervals were measured from QRS onset to the J-point. QT intervals were determined between QRS onset and the end of the T wave, defined as a return to the isoelectric line. Manual measurements were independently made by two clinicians (RS and PA), blinded to patient outcome with an interobserver variability below 5%. Values for QT intervals were corrected for HR using the correction formulas detailed above. QT dispersion (QTd) is a measure of spatial heterogeneity of ventricular repolarization. QTd was calculated by subtracting the minimal QTc from the maximal QTc interval in individual patients. Only patients who displayed QTd greater than 50 ms after QT correction with all four correction formulas were considered as having pathologic QTd (Macfarlane et al., 1998).


Data (categorical and continuous) consisted of characteristics and values from matched pairs. All electrocardiographic data were averaged per patient (data from 1–6 seizures per patient, mean 2.4 seizures per patient), and resulting data were subjected to conditional logistic regression analysis. We also used a mixed 2- or 3-level linear or logistic regression model adjusted for seizure-clustering in individual patients and matched pairs to compare electrocardiographic or seizure characteristics without averaging available data. Statistical analysis was performed with STATA software (StataCorp LP, TX, U.S.A.). p-Values <0.05 were regarded as statistically significant. Data values are expressed as mean ± standard deviation (SD), if not indicated otherwise. Odds ratios (OR) are given where appropriate. Missing data are indicated in the respective data tables and figures.


  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Summary and Conclusions
  7. Acknowledgment
  8. References
  9. Supporting Information

A total of 38 people with pharmacoresistant epilepsy (19 SUDEP, 19 control patients) with available video-EEG data (total of 51 and 40 seizures in SUDEP and control group) were included. The SUDEP group comprised one probable (Patient 13) and 18 definite SUDEP events (Nashef, 1997). One SUDEP patient (Patient 18) had sarcoidosis as a potentially competing comorbidity for sudden cardiac death, but postmortem did not reveal any structural cardiac abnormalities. Two SUDEP patients (Patients 4 and 11) also had an implanted vagal nerve stimulator (VNS). A contribution of vagal nerve stimulation to sudden death cannot be ruled out in these patients. To date, however, there is no strong evidence to indicate an increased risk of SUDEP after VNS implantation (Annegers et al., 2000). Further clinical characteristics of individual patients are given in Tables 1 and 2.

Table 1.   Clinical details of individual patients
Patient no.SexAge/Epilepsy durationaComorbidityEpilepsy typeMRI findingSeizure lateralizationSurgery/ Outcomeb
  1. aAt telemetry; 0, none; B, bilateral; F, female; FLE, frontal lobe epilepsy; HS, hipppocampal sclerosis; ITP, idiopathic thrombocytic purpura; L, left; LE, lupus erythematosus; M, male; na, not applicable; OSAS, obstructive sleep apnea syndrome; R, right; TLE, temporal lobe epilepsy; VNS, vagal nerve stimulator.

  2. bAccording to ILAE classification. Outcome assessed 24 months after surgery.

1F33/250TLEPostischemic lesion temporoparietal RR0/na
2F32/10DepressionTLEHS LL1/1
3M36/26DepressionTLEGangliocytoma/hamartoma frontal lobe LL1/4
4M28/12Depression Ictal bradycardia Cardiac pacemaker (VVI) VNSCryptogenic TLE0R/B0/na
5M41/40Anxiety disorderTLEHS LR/L0/na
6M28/26AlcoholismCryptogenic partial epilepsy0R/L1/4
7M21/190Cryptogenic FLE0L0/na
8M44/26Exocrine pancreatic deficiencyFLETraumatic lesion frontal lobe RR0/na
9M41/26PsychosisCryptogenic TLE0R/L0/na
10F39/35PsychosisCryptogenic partial epilepsy0R0/na
11F15/7VNSCryptogenic TLE0R/L0/na
12F50/450TLEHS LL0/na
13F40/230Cryptogenic TLE0R0/na
14M42/37HypothyroidismTLEHS RR0/na
15M29/16Mental retardationPartial and generalized epilepsy0L1/4
16F41/150TLESecondary lesions (arachnoid cyst) temporal lobe LL0/na
17M33/280Cryptogenic FLE0B0/na
18F49/40Depression, SarcoidosisCryptogenic FLE0R1/unknown
19M43/34DepressionTLEHS and dysplastic amygdala LL1/4
20F32/11DepressionTLEDiffuse postradiation defect, resection defect (medulloblastoma)R0/na
21F29/22Systemic LE I, TP Antiphospholipid antibodiesFLEUnspecific white matter lesions frontal lobe LL0/na
22M32/290TLEHS LB1/1
23M28/260Cryptogenic FLE0L0/na
24M42/330TLEGlioma temporal lobe LL0/na
25M28/260TLEHS LL1/1
26M25/200TLEHS RR0/na
27M43/260Cryptogenic FLE0L0/na
29F39/36DepressionFLECavernoma RR1/4
30F19/7DepressionPartial TLEAstrocytoma frontotemporal RUndecided1/4
31F53/150TLEHS RL0/na
32F44/430TLEHS LL1/1
33M38/330TLEHS RR1/1
34M35/28DepressionTLEHS LL1/2
35F39/34DepressionTLEHS LL1/1
36M39/38HypertensionTLEHemiatrophy and HS LL0/na
37F46/450TLEDiffuse cerebral atrophy L>R after perinatal hemorrhage and HS LL0/na
38M44/30DepressionBilateral TLEHS BR/L0/na

Using a conditional logistic regression model, three of the analyzed clinical parameters were significantly different in SUDEP compared to control patients (Table 2). First, the presence of pathologic cerebral MRI findings was associated with a lower SUDEP risk [p = 0.037, 95% confidence interval (CI) 0.01–0.88; OR = 0.11]. This association was independent of performed epilepsy surgery, as it was still statistically significant after adjusting for epilepsy surgery. Second, patients in whom AED reduction was instituted during telemetry were less likely to die suddenly later on (p = 0.0156, 95% CI 0–0.69; OR = 0.10). Third, the rate of SGTCS per year per patient was significantly higher in SUDEP than in their counterparts (p = 0.035, per SGTCS per year 95% CI 1.00–1.09), with an OR of 1.044 per tonic–clonic seizure (i.e., each secondarily generalized tonic–clonic seizure increased the risk of SUDEP by 4.4%); this was independent of the maximally observed overall seizure frequency. In contrast, none of the analyzed electrocardiographic predictors for cardiac mortality and sudden cardiac death was significantly different between SUDEP and control patients (conditional logistic regression analysis, Table 3).

Table 3.   Electrocardiographic predictors for cardiac mortality
 ControlaSUDEPap-Valuea95% CIOR
  1. HR, heart rate; HRR, HR recovery; HRV, HR variability; QTc, corrected QT interval; QTd, QT dispersion; SDNN, standard deviation of all RR intervals.

  2. aTotal no. of patients (or pairs) with available information indicated in parentheses if different from 19.

  3. bQT corrected with Fridericia’s formula.

  4. Mean ± SD of averaged values per patient .

Maximal preictal HR (bpm)80 ± 1575 ± 140.2320.17 to 3.35 
Maximal ictal HR (bpm)115 ± 30118 ± 230.688−0.04 to 0.02 
Relative ictal change of HR1.46 ± 0.371.62 ± 0.320.208−0.04 to 0.06 
Maximal postictal HR (bpm)84 ± 1488 ± 170.3850.14 to 7.1 
HRR (bpm)4 ± 1714 ± 130.102−0.08 to 0.02 
Preictal HRV (10 s SDNN, ms)29 ± 1731 ± 19 (17)0.971 (17)−0.02 to 0.03 
Ictal HRV (10 s SDNN, ms)18 ± 15 (18)22 ± 17 (15)0.498 (14)−0.70 to 3.22 
Postictal HRV (10 s SDNN, ms)31 ± 1638 ± 28 (17)0.474 (17)−0.03 to 0.07 
Preictal HRV (30 s SDNN, ms)33 ± 19 (18)35 ± 19 (17)0.93 (16)−0.01 to 0.1 
Postictal HRV (30 s SDNN, ms)34 ± 16 (18)45 ± 30 (17)0.303 (16)−0.04 to 0.04 
Preictal QTc (ms)b414 ± 27420 ± 250.483−0.04 to 0.09 
Ictal QTc (ms)402 ± 31 (18)417 ± 38 (18)0.24 (17)−0.02 to 0.04 
Postictal QTc (ms)415 ± 21420 ± 250.368−0.04 to 0.04 
QRS (12-lead, ms)97 ± 24 (18)93 ± 28 (16)0.418 (15)−0.01 to 0.04 
QTd (12-lead, ms)b45 ± 16 (18)50 ± 19 (16)0.748 (15)0.25 to 8.98 
Pathologic QTd (12-lead, no. pat.)4 (18)5 (16)1.0 (15)  
Pathologic QTc (no. pat.)230.657−0.01 to 0.031.5
Cardiac arrhythmias (no. pat.)540.739−0.02 to 0.050.8
Any ECG abnormality (no. pat.)6 (18)6 (16)0.706 (15)0.21 to 2.980.75

Because a single seizure can result in sudden death, individual outliers are perhaps more relevant than population means. Therefore, we also considered electrocardiographic data at the level of individual patients and seizures. As exemplified in Fig. S1A,B, HR transiently increased in most patients during seizures (35 of 38 patients, see also Fig. 1A). HR was not affected by seizure-activity in only three control patients. HR showed marked ictal instability (Fig. S1B), with initial tachycardia and pronounced transient bradycardia in one SUDEP patient (Patient 4). Brief episodes of asystole were also noted in this patient, and these led to subsequent implantation of a VVI cardiac pacemaker, which was replaced after some years and still present at the time of his death 10 years after telemetry. Ultra short-term HRV decreased during seizures in both control and SUDEP patients to the same extent (Fig. 1B). Postictally, however, HRV increased in SUDEP patients more than in controls, although this difference did not reach statistical significance (Fig. 1B, mixed logistic regression model, HRV expressed as 30 s SDNN, p = 0.262).


Figure 1.   Higher ictal heart rates preferentially occurred during SGTCS. No difference in absolute changes of HRV or QTc during and after seizures between SUDEP and control patients. (A) Left y-axis indicates number of seizures and patients with HR >100 bpm and >120 bpm. Right y-axis refers to number of seizures and patients with HR >150 bpm. (B) Ultra short-term HRV decreased during seizures (expressed as SDNN in a 10 s interval) in both groups (left bars) and returned to baseline levels after seizure cessation in controls, whereas HRV slightly increased in SUDEP group. However, differences in HRV change between controls and SUDEP patients did not reach statistical significance. Number of seizures from left to right: 30, 29, 31, 33, 30, and 33. (C) Left panel displays absolute QTc changes during seizures with respect to preictal QTc using all four correction formulas (Baz, Bazett’s; Fri, Fridericia’s; Fra, Framingham; Hod, Hodges) in controls and SUDEP patients. Difference in QTc according to Bazett’s reached statistical significance (p = 0.036), most likely due to overcorrection for higher ictal HR. Right panel displays absolute postictal QTc changes. 37 seizures in control and 48 seizures in SUDEP group. Data expressed as mean ± SEM. For all panels: blue bars control patients, red bars SUDEP patients.

Download figure to PowerPoint

Seizure-related QTc prolongation >10 ms (according to all four correction formulas, see Supporting Information) occurred in 43 of 89 seizures (∼48%, 25 seizures from SUDEP patients) in 24 patients (∼63%, 13 SUDEP patients). Figure 1C summarizes the absolute changes in QTc according to all four correction formulas during and after seizures. It is important to note that five patients displayed seizure-induced QTc lengthening beyond upper normal limits in nine seizures (see Tables S1 and S2: three SUDEP, two controls; mixed logistic regression model, p = 0.657, 95% CI −0.01 to 0.03, OR = 1.5). Lateralization of seizure onset did not have an effect on occurrence of QTc prolongation (mixed logistic regression model, p = 0.545, 95% CI 0.15–34). In none of the patients did all four correction formulae give a short QTc. QTc values below normal limits were, however, observed in some patients before (three seizures in one control and two SUDEP patients), during (four seizures in three control patients), and 5 min after a seizure (two seizures in two SUDEP patients).

12-Lead ECG recordings were available in 34 patients. Abnormalities were found in 12 patients (six SUDEP, six controls). Incomplete or complete right bundle branch block was found in three patients (two controls), bradycardia in one control, tachycardia in two control patients, and a shifted electrical heart axis in two SUDEP patients (one to either side). Nine patients displayed pathologic QTd beyond 50 ms (with all four formulas): four control (22% of control patients with available 12-lead ECG), and five SUDEP patients (31% of SUDEP patients with available 12-lead ECG, Table S3).

No significant correlation between presence of pathologic QTd and age, epilepsy duration, age of first seizure, gender, and seizure-lateralization was detected (mixed logistic regression model). The presence of both pathologic QTd and QTc prolongation was found in two patients and was not associated with a higher risk of SUDEP (mixed logistic regression model, p = 1.0). One SUDEP patient also exhibited pathologic interictal QTc, as assessed by one-lead ECG recordings and a P mitrale–like configuration (Fig. S2D), without any known cardiac or pulmonary cause. In this patient, QTc prolongation might be related to comedication with citalopram, although there were no signs or indications of overdose.

Cardiac arrhythmias occurred in nine people (five controls), mainly postictally. They were benign and included premature atrial (Fig. S2A,B; two controls, one SUDEP patient) or ventricular (two controls) beats as well as marked postictal sinus arrhythmia (Fig. S2C; three controls, four SUDEP patients).

We next assessed if seizure type influenced intra- and postictal electrocardiographic features (mixed logistic regression model, Table 4). SGTCS lasted longer than nongeneralized seizures (p = 0.018, 95% CI 3.67–39.9), and led to greater absolute ictal HR (p = 0.002, 95% CI 6.23–28.86) and a more impaired HRR (p < 0.001, 95% CI 13.03–31.14). Ictal and postictal HRV were also significantly lower in SGTCS (e.g., postictal HRV expressed as 30 s SDNN: p = 0.002, 95% CI −51.63 to −11.50). Likewise, cardiac arrhythmias tended to occur more frequently after SGTCS, but this difference did not reach statistical significance (mixed logistic regression model, p = 0.091, 95% CI 0.79–21.38, OR = 4.125). Together these findings seem to indicate a significantly greater impact of SGTCS on ECG parameters.

Table 4.   Electrocardiographic predictors for cardiac mortality in SGTCS
 Nongen. seizures (63 seizures/32 pat.)aSGTCS (28 seizures/11 pat.)a p-Valueb 95% CIOR
  1. Values expressed as mean ± SD of averaged data per patient. HR, heart rate; HRR, HR recovery; HRV, HR variability; QTc, corrected QT interval; SDNN, standard deviation of all RR intervals.

  2. aIn parentheses number of patients (columns 2 and 3) or seizures (column 4) with available information if different from total number.

  3. bSignificant finding.

Duration (s)77 ± 3897 ± 310.018a3.67 to 39.9 
No. of seizures arising from sleep14 in 8 pat.18 in 5 pat.0.205 (90)−0.87 to 4.1 
Preictal HR (bpm)77 ± 1576 ± 120.991−9.39 to 9.29 
Ictal HR (bpm)110 ± 23135 ± 270.002 (90)b6.23 to 28.86 
Postictal HR (bpm)81 ± 13107 ± 24<0.001b18.04 to 34.65 
Relative ictal change HR1.46 ± 0.321.82 ± 0.37<0.001 (90)b0.16 to 0.56 
HRR (bpm)4 ± 1431 ± 24<0.001b13.03 to 31.14 
HRV (ms)
 10 s SDNN pre32 ± 1824 ± 14 (7)0.138 (63)−26.46 to 3.65 
 10 s SDNN ictal21 ± 16 (30)9 ± 6 (5)0.024 (58)b−26.93 to −1.89 
 10 s SDNN post37 ± 2420 ± 6 (7)0.014 (63)b−39.26 to −4.44 
 30 s SDNN pre35 ± 18 (31)31 ± 20 (7)0.478 (62)−20.23 to 9.47 
 30 s SDNN post43 ± 27 (31)20 ± 5 (7)0.002 (62)b−51.63 to −11.50 
QTc Fridericia’s (ms)
 Preictal419 ± 29418 ± 260.93−14.42 to 15.83 
 Ictal413 ± 35412 ± 42 (10)0.64 (85)−14.70 to 24.00 
 Postictal418 ± 22410 ± 300.064 (89)−28.58 to −0.93 
No. of seizures with QTc prolongation7 in 4 pat.2 in 2 pat.1.0  
No. of seizures with cardiac arrhythmia9 in 4 pat.11 in 6 pat.0.0910.79 to 21.384.125


  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Summary and Conclusions
  7. Acknowledgment
  8. References
  9. Supporting Information

Our main findings are that a substantial proportion of patients with pharmacoresistant epilepsy display abnormal cardiac repolarization at rest, and during or after seizures, which could facilitate sudden cardiac death. SUDEP patients were more likely to lack pathologic cerebral MRI findings, statistically independent of whether surgery was performed, and less likely to experience drug reduction during video-EEG telemetry. Furthermore, SUDEP patients had more SGTCS per year. In contrast, none of the cardiac abnormalities were more common in those patients dying of SUDEP later on.

Abnormal cardiac repolarization in pharmacoresistant epilepsy patients

We found only small, statistically insignificant differences in ECG predictors for cardiac mortality in SUDEP patients compared to controls. It could be argued that this is due to lack of statistical power. These differences, however, if present, are unlikely to be clinically relevant for SUDEP prediction, given both their small magnitude and the high number of ECG abnormalities that we observed in controls and SUDEP patients alike. We have conclusively shown that adult patients with pharmacoresistant epilepsy display pathologic cardiac repolarization at rest and during or after seizures as assessed by interictal QTd and intra- and postictal QTc. Both are established predictors for cardiac mortality in otherwise healthy people (Schouten et al., 1991; Macfarlane et al., 1998; Chiang, 2004).

Peri-ictal QTc prolongation

Prolongation of QT makes the heart more susceptible to threatening tachyarrhythmia and sudden cardiac death (Chiang, 2004). Overall, the frequency of seizure-related pathophysiologically relevant QTc prolongation may be underestimated in our study for two main reasons. First, we have conservatively chosen only those seizures with QT intervals on which all four corrections formulas agreed (see Supporting Information). Second, we have used modified upper limits for QT intervals to avoid false-positive pathologic QT intervals (Luo et al., 2004). Our observation is supported by a study reporting transient postictal QT prolongations beyond normal limits in children with epilepsy (Kändler et al., 2005). Likewise, a previous study has shown an association of QTc lengthening and interictal epileptiform EEG discharges in 6 of 11 SUDEP patients by 35 ms (mean value), but QTc values did not reach pathologic upper limits (Tavernor et al., 1996). There are at least three possible mechanisms by which cardiac repolarization and thus QT intervals could be altered during seizures. (1) Cardiac repolarization is partially regulated by cerebral mechanisms. Especially the insular cortex is implicated in the regulation of cardiac repolarization; for example, ischemic strokes with insular involvement can lead to QT lengthening (Tatschl et al., 2006). In rat models, QT prolongation could be experimentally induced by stimulation of the left insula or damage of the right hemisphere (Oppenheimer, 2006). Human data concerning cerebral lateralization of QT interval modulation are, however, conflicting (Oppenheimer, 2006; Tatschl et al., 2006), consistent with our data indicating no association between seizure lateralization and QTc prolongation. (2) Alterations in respiratory pattern during seizures lead to ictal hypoxemia and hypercapnia in a substantial number of patients and seizures, and might be, together with cardiac dysfunction, one of the key players in SUDEP (Johnston et al., 1995; Langan et al., 2000; Bateman et al., 2008). It is important to note that hypercapnia and hypoxia both prolong QTc intervals (Kiely et al., 1996; Roche et al., 2003). Thus the observed QTc prolongations are likely to be due, at least partially, to ictal hypoxemia and hypercapnia. Oximetry data were not available in our study population. (3) Increase of catecholamine release during and after seizures can also lead to an increase in QTc (Simon et al., 1984; Lee et al., 2003).

Interictal QT dispersion

An alternative measure of cardiac repolarization is QTd, which reflects spatial distribution of ventricular repolarization. A higher QTd indicates a more heterogeneous ventricular repolarization throughout the ventricular wall and neighboring areas. This is a potential cause for ventricular arrhythmias that could facilitate ventricular microcircuits leading to reentrant tachycardia (Engel et al., 2004). Nine of 34 patients displayed pathologic QTd. The increase in QTd could be a consequence of subtle perivascular and interstitial fibrosis, which have been described in people with epilepsy (including SUDEP patients) (Natelson et al., 1998; Opeskin et al., 2000; P-Codrea Tigaran et al., 2005). Likewise, an increase in sympathetic tone and a nonuniform cardiac sympathetic innervation may contribute to an increased QTd (Schraeder & Lathers, 1983; Hilz et al., 2002). An effect of AEDs on QTd cannot be ruled out, as most patients were on AED treatment during recording of 12-lead ECG and video-EEG telemetry (only in one control and one SUDEP patient AED treatment was completely stopped). There is no unequivocal report on AED-induced QT interval prolongation to date. Rufinamide and primidone, however, were both reported to shorten QT interval (DeSilvey & Moss, 1980; Cheng-Hakimian et al., 2006). In our study, two of the patients in the SUDEP group were on primidone treatment during telemetry. One of them had an available 12-lead ECG recording, which displayed pathological QTd, but QTc intervals were normal. None of the other patients was on AED affecting cardiac repolarization. Thus AEDs that are known to affect QT interval may modify QTd, but are unlikely to have a major effect. This is also in line with the finding that QTd was increased in children with epilepsy regardless of AED treatment (Akalin et al., 2003). Pathologic QTd was also shown to predict cardiac arrhythmias during electroconvulsive therapy (Rasmussen et al., 2007), but we did not observe a significant association between arrhythmias and pathologic QTd (mixed logistic regression model, p = 0.571).

Secondarily generalized tonic–clonic seizures and SUDEP

Our study is consistent with earlier studies that identified generalized tonic–clonic seizures as a strong risk factor for SUDEP (Langan et al., 2005; Tomson et al., 2008). It is, however, surprising that even in this particular population of pharmacoresistant epilepsy patients, generalized tonic–clonic seizures appear to be a risk factor. Likewise, a large proportion of witnessed SUDEP occurred in the aftermath of a GTCS (Langan et al., 2000). In our study the risk of SUDEP increased by 4.4% for each additional SGTCS per year. The association of increased risk of SUDEP with lack of AED reduction during telemetry may also be related to SGTCS, because as part of our telemetry unit policy on drug reduction, we are less likely to reduce medication in those with SGTCS. The association of normal MRI with the SUDEP patients most likely relates to surgical outcome. Those with MRI lesions are most likely to proceed to surgery or to have a good outcome following surgery, and this may protect against SUDEP (Ryvlin et al., 2005). How does the increased SGTCS rate translate to a greater risk of SUDEP? Analysis of cardiac parameters by seizure type revealed that SGTCS have a greater effect on cardiac function than nongeneralized seizures do. In our study, SGTCS led to higher ictal HR than other seizures (Fig. 1A). This finding is consistent with previous reports on epilepsy and SUDEP patients, with the latter exhibiting higher ictal HR (Nei et al., 2000, 2004; Opherk et al., 2002). Previous studies, however, did not adjust for seizure clustering within patients, and respective control patients were not pair-matched for age and gender. In contrast to a previous study (Tigaran et al., 2003), we did not observe ST-segment depression with the increased HR.

Heart rate recovery and heart rate variability

Perhaps of greater relevance is that postictal HRR was reduced in SGTCS. Impaired HRR after exercise is an independent predictor of cardiac mortality in healthy people and probably reflects reduced parasympathetic activity (Cole et al., 2000; Watanabe et al., 2001; Jouven et al., 2005; Goldberger et al., 2006). This in turn leads to dominance of sympathetic tone with potential pro-arrhythmic effects, linking SGTCS to SUDEP. Likewise, there was a tendency in our study for SGTCS to be associated with occurrence of cardiac arrhythmias, as suggested by previous studies (Nei et al., 2000; Opherk et al., 2002; Zijlmans et al., 2002).

Changes in HRV with tonic–clonic seizures did not explain the additional risk posed by SGTCS in our study. We found a decreased HRV during and after SGTCS. This finding might be partially a reflection of the higher HR in SGTCS, as we found a negative correlation between HR and HRV (Fig. S3C). Increases rather than decreases in HRV during exercise and recovery, however, predict increased cardiac mortality (Dewey et al., 2007). There was a tendency for postictal HRV to be higher in SUDEP as compared to patients in the control group (Fig. S1A), but this was a small effect and did not reach statistical significance. In summary, SGTCS seem to have a greater and potentially unfavorable impact on cardiac excitability, most likely due to strong activation of the sympathetic nervous system (Simon et al., 1984), as well as other seizure-related cardiorespiratory alterations. Consequently, patients with a higher SGTCS rate may have a higher risk for seizure-related life-threatening cardiac arrhythmias.

Summary and Conclusions

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Summary and Conclusions
  7. Acknowledgment
  8. References
  9. Supporting Information

This study has not revealed specific ECG predictors for cardiac mortality in patients with pharmacoresistant focal epilepsies that die from SUDEP from those who do not, but has conclusively shown cardiac repolarization defects in adult patients with epilepsy. Similar abnormalities in cardiac repolarization were shown to predict an increased risk of cardiac mortality and sudden cardiac death in other medical conditions or otherwise healthy populations. Therefore, their occurrence in pharmacoresistant epilepsy patients could facilitate fatal ventricular tachyarrhythmia and sudden cardiac death as one plausible cause for SUDEP. Indeed, occurrence of ventricular tachycardia and fibrillation was recently described in a secondarily generalized tonic–clonic seizure of an epilepsy patient without underlying cardiac disease (Espinosa et al., 2009). It is also tempting to hypothesize that sudden cardiac death and epilepsy have common genetic susceptibility genes (Nashef et al., 2007). SUDEP is, however, likely to have a variety of underlying etiologies, and our findings do not exclude other potential cardiorespiratory mechanisms such as ictal bradycardia or asystole and hypoxemia which may be more frequent or worse during generalized tonic–clonic convulsions (Nashef et al., 1996; Bateman et al., 2008). People with SUDEP had a greater number of SGTCS per year, and SGTCS increase the likelihood of potentially detrimental cardiac changes. These findings suggest that SGTCS may favor SUDEP by additionally compromising cardiac function.


  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Summary and Conclusions
  7. Acknowledgment
  8. References
  9. Supporting Information

This work was undertaken at UCLH/UCL, which receives a proportion of funding from the Department of Health’s NIHR Biomedical Research Centres funding scheme. RS was supported by a stipend from the Deutsche Forschungsgemeinschaft. 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.

Disclosure: The authors declare no potential financial conflict of interest.


  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Summary and Conclusions
  7. Acknowledgment
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Summary and Conclusions
  7. Acknowledgment
  8. References
  9. Supporting Information

Supplementary Table 1. Details of patients with seizure-related QTc lengthening on which all 4 correction formulations agreed.

Supplementary Table 2. Number of seizures and patients with seizure-related QTc lengthening.

Supplementary Table 3. Details of patients with pathologic QTd in 12-lead ECG on which all 4 correction formulations agreed

Supplemental Fig 1. Time course of instantaneous HR in individual patients. (A) Time course of instantaneous HR in patient no. 19. (B) Time course of instantaneous HR in patient no. 4. Arrows indicate start and end of seizure.

Supplemental Fig 2. Examples of ECG abnormalities as assessed by one-lead (I) ECG recordings. (A) Postictal premature atrial beat (asterisk) with compensatory pause (lower traces) in a control patient (no. 38). (B) Postictal supraventricular trigeminy (premature supraventricular beats marked with asterisks in lower traces) in a SUDEP patient (no. 7). (C) Postictal sinus arrhythmia (RR interval variability > 0.12 s, lower traces) in a SUDEP patient (no. 4). (D) Interictal prolonged QTc (QT 495 ms, RR interval 1100 ms; QTc Friderici 480 ms) and P mitrale-like configuration (P wave duration 125 ms) in a SUDEP patient (no. 2). Scaling in all panels 1 s.

Supplemental Fig 3. Weak negative correlation between HRV and HR. (A) The two measures of HRV, RMSSD and SDNN, are highly correlated (r2 = 0.81). Therefore, we only presented SDNN data. (B) SDNN values obtained in time intervals of 10 and 30 s in the same seizures are highly correlated (r2 = 0.80). Thus, 10 s intervals may give a good estimate of ultra-short-term HRV. Inset: SDNN values obtained during a 30 s interval are significantly higher than during a 10 s interval (n=60 seizures, paired t-test, p < 0.001). (C) HRV shows a weak negative correlation with HR (r2 = 0.18, slope is significantly different from zero). This may partially explain lower HRV during and after SGTCS.

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

EPI_2330_sm_Supplemental_Material.doc1610KSupporting info item
EPI_2330_sm_Surges-copyright.pdf867KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.