Increased risk of sudden unexpected death in epilepsy in females using lamotrigine: A nested, case-control study

Authors

  • Dag Aurlien,

    1. Department of Neurology, Stavanger University Hospital, Stavanger, Norway
    2. The Norwegian Centre for Movement Disorders, Stavanger University Hospital, Stavanger, Norway
    Search for more papers by this author
  • Jan Petter Larsen,

    1. The Norwegian Centre for Movement Disorders, Stavanger University Hospital, Stavanger, Norway
    Search for more papers by this author
  • Leif Gjerstad,

    1. Department of Neurology, Oslo University Hospital – Rikshospitalet, Oslo, Norway
    2. University of Oslo, Faculty of Medicine, Oslo, Norway
    Search for more papers by this author
  • Erik Taubøll

    1. Department of Neurology, Oslo University Hospital – Rikshospitalet, Oslo, Norway
    2. University of Oslo, Faculty of Medicine, Oslo, Norway
    Search for more papers by this author

Address correspondence to Dag Aurlien, Department of Neurology, Stavanger University Hospital, P.O.Box 8100, 4068 Stavanger, Norway. E-mail: auda@sus.no

Summary

Purpose:  To estimate the incidence of sudden unexpected death in epilepsy (SUDEP) in Rogaland County, Norway, in the period August 1 1995–July 31 2005, and to investigate whether use of lamotrigine (LTG) was associated with increased risk in female patients or other subgroups.

Methods:  SUDEP victims were identified from autopsy reports and data from the Norwegian Cause of Death Registry. In all cases where SUDEP was considered as a possible cause of death, the hospital records were also reviewed. For each deceased, at least three living patients with epilepsy were randomly selected as controls. The market share in defined daily doses was collected for each year to estimate the number of patient-years at risk on each antiepileptic drug.

Key Findings:  We identified 26 cases of SUDEP: 16 definite, 3 probable, and 7 possible; 15 patients were female and 11 were male. Of these, 10 patients (38.5%) were treated with LTG: 9 of these patients were female. The incidence of SUDEP was estimated as 1.0 per 1,000 patient-years when all cases were included, and 0.7 per 1,000 patient-years for definite and probable SUDEP. Seven of 12 (58.3%) of female patients with definite and probable SUDEP and 10 of 41 (24.4%) of controls matched on age and gender were on LTG (p = 0.038). The incidence of definite and probable SUDEP in women on LTG, was estimated as 2.5 per 1,000 patient-years and 0.5 per 1,000 patient-years in female who were not taking LTG (p = 0.007).

Significance:  The incidence of SUDEP was significantly higher among female patients with epilepsy who were being treated with LTG than among female patients with epilepsy who were not taking LTG, and a significantly higher proportion of female SUDEP cases than controls were taking LTG. Our findings may have implications for treatment of epilepsy in female patients.

Sudden unexpected death in epilepsy (SUDEP) is the most common, directly epilepsy-related cause of death, and it has been estimated that sudden death occurs up to 40 times more frequently in patients with epilepsy than in the general population (Annegers & Coan, 1999). In community-based studies, reported rates have ranged from 0.09 per 1,000 person-years in an incidence cohort of newly diagnosed epilepsy (Lhatoo et al., 2001) to 0.9–2.7 per 1,000 person-years in a prospective prevalence cohort of epilepsy (Leestma et al., 1989). The majority of observed cases have occurred in relation to a generalized tonic–clonic seizure (GTCS) (Tomson et al., 2005), and high seizure frequency appears to be the most important risk factor for SUDEP (Nilsson et al., 1999; Walczak et al., 2001; Tomson et al., 2005). In addition, polytherapy and long duration of epilepsy have been identified as the most consistent risk factors (Surges et al., 2009). The highest rates have been found in patients between 20 and 40 years of age (Tomson et al., 2005). Cardiac arrhythmia, central or obstructive apnoea, and electrocerebral shutdown have been proposed as possible pathophysiologic mechanisms (Tomson et al., 2008). However, a clear mechanism that is common to all cases has not been established.

We have previously reported a case series from our outpatient clinic of four consecutive SUDEP cases that were all in young women with idiopathic epilepsy being treated with lamotrigine (LTG) in monotherapy (Aurlien et al., 2007). In that article we suggested certain mechanisms that might explain an association among SUDEP, LTG treatment, and female gender. However, we concluded that a systematic study was needed to reveal whether LTG really may increase the risk of SUDEP in subgroups of patients. Recently, an increased risk of SUDEP in patients with idiopathic generalized epilepsy (IGE) being treated with LTG was reported (Hesdorffer et al., 2011).

Herein we present the results of a study that aimed to estimate the incidence of SUDEP in our county and to investigate whether use of LTG was associated with increased risk of SUDEP in female patients with epilepsy or other epilepsy patient subgroups.

Materials and Methods

Patients

We aimed at identifying all individuals with SUDEP in Rogaland County, Norway, in the 10-year period from August 1, 1995 to July 31, 2005. A prerequisite for inclusion was a permanent address in Rogaland County.

The SUDEP cases in this study were found by review of autopsy reports and data from the Norwegian Cause of Death Registry, in which all cases in our county with epilepsy as an underlying or contributory cause of death were listed (N = 136). To evaluate the quality of our methods in identifying cases, we also obtained, from the Norwegian Cause of Death Registry, the causes of death of all deceased patients who were registered with a diagnosis of epilepsy at Stavanger University Hospital in the same period (N = 268). The hospital records were reviewed for all cases for which SUDEP was considered as a possible cause of death.

Diagnosis of SUDEP

The inclusion of cases was based on the following definition of SUDEP: “Sudden, unexpected, witnessed or unwitnessed, nontraumatic and nondrowning death in patients with epilepsy, with or without evidence for a seizure and excluding documented status epilepticus, in which postmortem examination does not reveal a toxicologic or anatomic cause for death” (Nashef, 1997). Cases were classified as definite, probable, and possible (Tomson et al., 2008).

Two cases, both classified as possible SUDEP, were found by review of the causes of death of the deceased epilepsy patients at Stavanger University Hospital during the 10-year period. Four of the SUDEP victims included in this study are those that have been previously described in our publication concerned with cases from our outpatient clinic (Aurlien et al., 2007).

Estimates of patient-years at risk

The epilepsy population at risk was estimated by multiplying the number of residents in the county each year (Statistics Norway: http://www.ssb.no) by 0.7%, which is the estimated prevalence of epilepsy in Western countries (Forsgren, 2004).

The market share in defined daily doses (DDDs) in Rogaland County (Farmastat AS: http://www.farmastat.no) was collected for each year to estimate the number of patient-years at risk for each antiepileptic drug (AED) (Table 1).

Table 1.   Market share of AEDs in DDDs and estimated patient-years at risk on each AEDa
 199519961997199819992000200120022003200420051995–2005 (sum)
  1. AEDs, antiepileptic drugs; CBZ, carbamazepine; DDDs, defined daily doses; F, females; LTG, lamotrigine; M, males; MS, market share; Pt., patient; VPA, valproate.

  2. aThe complete table (Table S2) including corresponding information about the remaining AEDs is available in the online version of the article.

Residents in the county354,447357,027360,403364,341369,059373,210375,225381,375385,020388,848393,104 
Patients with epilepsy2,4812,4992,5232,5502,5832,6122,6272,6702,6952,7222,752 
Patient-years at risk1,034
(August–December)
2,4992,5232,5502,5832,6122,6272,6702,6952,7221,605
(January–July)
26,120
CBZ MS (%) 41.240.540.239.638.536.834.231.926.8824.0720.1 
Pt.-years at risk on CBZ425.91012.11014.21009.8994.5961.2898.4851.7724.4655.2322.68,870
Patients with CBZ prescribed         1,965 (927 F/1,038 M)1,876 (12 months) (873 F/1,003 M) 
VPA MS (%) 12.813.213.915.415.81717.91918.8618.7717.7 
Pt.-years at risk on VPA132.3329.9350.7392.7408.1444470.2507.3508.3510.9284.14338.5
Patients with VPA prescribed         1,069 (528 F/541 M)1,109 (12 months) (547 F/562 M) 
LTG MS (%)  1.72.94.14.76.06.87.78.711.7614.4215.9 
Pt.-years at risk on LTG 17.672.5103.4119.9155177.6202.3232.3316.9392.5255.22045.2
Patients with LTG prescribed         868 (501 F/367 M)1,055 (12 months) (614 F/441 M) 

Control group

For each individual with SUDEP, we randomly selected at least three living controls, who had been registered in the database of Stavanger University Hospital with a diagnosis of epilepsy in the same year as the SUDEP case occurred. To investigate whether age and gender could be risk factors, an unmatched primary control group was selected. There were 89 unmatched controls for definite and probable SUDEP and 120 for definite, probable and possible SUDEP. For the remaining analyses, a new secondary control group was selected and matched with the deceased SUDEP cases by age and gender. There were 63 matched controls for definite and probable SUDEP and 86 for definite, probable, and possible SUDEP; three to four controls for each of the deceased. Because the prescription rates of AEDs changed during the period of the study (Table 1), the hospital record data from each control were extracted from the year in which the death of the corresponding SUDEP case occurred. The epilepsy syndromes of the cases and controls were classified according to the International League Against Epilepsy (ILAE) 1989 definition (ILAE, 1989).

For the SUDEP cases and controls, clinical data were extracted from the hospital records, including information about possible risk factors (Table 2). In addition, the serum concentrations of AEDs and electrolytes were registered.

Table 2.   Clinical characteristics of definite and probable SUDEP cases compared to matched controlsa
 Cases (%)Controlsp-valueCORCI
  1. Bilat., bilateral; CI, 95% confidence interval; COR, conditional odds ratio; CPS, history of complex partial seizures; excl., excluding; epil., epileptogenic; fam., family; F, females; gen., generalized; freq., frequency; GTCS, history of generalized tonic–clonic seizures; IGE, idiopathic generalized epilepsy; local., localization; LTG, lamotrigine; rel., related; SPS, history of simple partial seizures; unclass., unclassifiable.

  2. aThe complete version of the table (Table S3) including the possible SUDEP cases is available in the online version of the article.

Duration >10 years12/19 (63.2)38/62 (61.3)0.9781.00.3–3.0
F on LTG; duration >10 years3/7 (42.9)13/24 (54.2)   
Onset before age 168/19 (42.1)32/62 (51.6)0.3960.60.2–1.9
F on LTG; onset before age 161/7 (14.3)11/24 (45.8)   
Confirmed fam. history of epilepsy3/19 (15.8)7/63 (11.1)   
GTCS18/19 (94.7)51/63 (81)0.1794.20.5–34.2
CPS11/19 (57.9)29/63 (46)0.4181.50.5–4.4
SPS5/19 (26.3)10/63 (15.9)0.3391.80.5–6.2
Absences1/19 (5.3)9/63 (14.3)   
Myoclonic seizures3/19 (15.8)10/63 (15.9)   
Atonic seizures0/190/86   
Gen. tonic seizures0/191/63 (1.6)   
Unclass. seizures2/19 (10.5)0/63   
Gen. epil. EEG6/19 (31.6)15/58 (25.9)0.5801.40.4–4.4
Bilat. epil. EEG4/19 (21.1)15/58 (25.9)   
Gen. or bilat. epil. EEG10/19 (52.3)30/58 (51.7)0.9051.10.3–3.4
LTG – gen. or bil. epil. EEG7/8 (87.5)12/25 (48)   
Focal epil. EEG6/19 (31.6)12/58 (20.7)0.3091.90.6–6.1
Bilat. pathology in CT/MRI2/18 (11.1)6/59 (10.2)   
Left pathology in CT/MRI5/18 (27.8)6/59 (10.2)0.0684.00.9–17.6
Right pathology in CT/MRI0/1811/59 (18.6)   
IGE5/19 (26.3)15/63 (23.8)0.7311.20.4–4.1
Local. rel. symptomatic6/19 (31.6)17/63 (27.0)0.6991.30.4–4.0
Local. rel. cryptogenic or idiopathic6/19 (31.6)28/63 (44.4)0.2350.50.2–1.6
Unclass.epil. syndrome2/19 (10.5)3/63 (4.8)   
Seizure freq. >1/week7/18 (38.9)11/63 (17.5)0.0573.00.97–9.2
F on LTG; seizure freq. >1/week3/7 (42.9)4/24 (16.7)   
Comorbidity10/19 (52.6)25/63 (39.7)0.2671.80.6–5.4
F on LTG; comorbidity3/7 (42.9)12/24 (50)   
Cardiovascular disease3/19 (10.5)4/63 (6.3)   
Brain tumor (excl. pineal gland cyst)0/183/59 (5.1)   
Hippocampal sclerosis0/183/59 (5.1)   
Learning disability2/19 (10.5)11/63 (17.5)   
Concurrent medication10/19 (52.6)11/63 (17.5)0.0068.91.9–42.1
F on LTG; concurrent medication4/7 (57.1)3/24 (12.5)   
Psychotropic drugs5/19 (26.3)1/63 (1.6)   
F on LTG; psychotropic drugs1/7 (14.3)0/24   
Cardiovascular drugs2/19 (10.5)4/63 (6.3)   

Approval

This study was approved by the Regional Committee of Medical Research Ethics.

Statistical analysis

Before conducting the study, a power analysis was performed to evaluate whether our material would be sufficient to detect a possible increase in risk of SUDEP associated with treatment with LTG. Based on the following: (1) a base population of about 400,000, (2) a prevalence of epilepsy of 0.7% (Forsgren, 2004), (3) a mean market share in DDDs of LTG of 6.7% in the 10-year period (Aurlien et al., 2007), and (4) an incidence of SUDEP of 1.4 per 1,000 patient-years (Nilsson et al., 1999), the number of patient-years at risk on LTG was estimated as 1,900, and, in those that were not on LTG, was estimated as 26,000 person-years. An anticipated relative risk (RR) of three could be detected with a power of 80% at a significance level of 0.05. Because of the statistical uncertainty connected with small numbers, statistical analysis was performed only when there were at least five observations. The rate ratios of SUDEP incidences between patients treated with an AED compared with patients not on a specific drug were estimated with 95% confidence intervals (Cis) using the web-based software Open Epi (http://www.openepi.com). For comparison of frequencies between cases and matched controls, conditional odds ratios (Ors) for matched age and sex case–control design were estimated using Epi Info version 3.5.1 (Centers for Disease Control, Atlanta, GA, U.S.A.). Comparison of mean duration of epilepsy between cases and matched controls were performed in SPSS (version 18; IBM Corporation, Armonk, NY, U.S.A.) using a linear mixed model with a random intercept. Analyses comparing cases with unmatched controls were performed in SPSS using Pearson chi-square tests for frequencies and Mann-Whitney tests for means.

Because the primary outcome of the study comparing cases with controls was the proportion treated with LTG, correction for multiple comparisons was not performed.

Results

We identified 26 cases of SUDEP: 16 definite, 3 probable, and 7 possible; 15 patients were female and 11 were male (Table S1).

Age and gender

Age and gender distribution did not differ significantly between cases and unmatched controls.

However, a significantly higher proportion of SUDEP cases than unmatched controls were between 20 and 40 years of age (Table 3).

Table 3.   Age and gender—definite and probable SUDEPa
 CasesUnmatched controlsMatched controls
  1. F, females; M, males.

  2. aThe complete version of the table (Table S4) including all SUDEP cases is available in the online version of the article.

GenderF/M = 12/7 (63% F)F/M = 40/49 (45% F) (p = 0.149)F/M = 41/22 (65% F)
Age (years)Mean 34.0, median 33.0, range 16–67Mean 29, median 26, range 0.75–75 (p = 0.140)Mean 33.9, median 33.0, range 15–69
Age 20–40 years13/19 (68%)30/89 (34%) (p = 0.005) 

AEDs

Of the 26 SUDEP cases, 10 (38.5%) were being treated with LTG; five of these patients were on monotherapy and 9 of 10 were female. Of 19 cases of definite or probable SUDEP, 8 (42%) were on LTG; 5 of these were on monotherapy and 7 of 8 were female. Of the 26 cases, 8 (4 female) were being treated with valproate (VPA), 7 (two female) with carbamazepine (CBZ), three were being treated with oxcarbazepine (OXC), three with phenytoin (PHT), three with vigabatrin (VGB), two with topiramate (TPM), one with phenobarbital (PB), and one was not being treated with an antiepileptic drug (AED).

Incidence of SUDEP

The total incidence of SUDEP was estimated as 1.0 per 1,000 patient-years when all cases were included, and 0.7 per 1,000 patient-years when only definite and probable SUDEP were included.

The incidence of SUDEP in patients on CBZ or VPA was not significantly different from the incidence among those not being treated with these drugs (Table 4).

Table 4.   Estimated incidence (cases per 1,000 patient-years) of definite and probable SUDEP in patients on LTG, VPA, and CBZ. For definite, probable, and possible SUDEP, the results are given in parentheses
 With the AEDWithout the AEDIncidence rate ratio/CIp-value
  1. AED, antiepileptic drug; CBZ, carbamazepine; CI, 95% confidence interval; DDD, defined daily doses; F, females; LTG, lamotrigine; VPA, valproate.

  2. Corr*: Patient-years at risk corrected through the 10-year period for the ratio between the exact number and the originally estimated number at risk in 2005 (based on the market share in DDDs). For CBZ this correction was impossible because the corrected number at risk was higher than the total epilepsy population.

  3. Corr**: Patient-years at risk corrected through the 10-year period for the ratio between the exact number and the originally estimated number at risk in 2005 (based on the market share in DDDs) and assuming the same share of females among patients on LTG through the whole 10-years period as in 2005 (58%).

LTG3.9 (4.9)0.5 (0.7)8.6/3.3–21.5 (7.4/3.2 –16.2)<0.001 (<0.001)
F on LTG6.8 (8.8)0.4 (0.5)16.5/5.1–56.8 (17.7/6.2–53.3)<0.001 (<0.001)
Corr.* LTG1.6 (2.0)0.5 (0.8)3.1/1.2–7.9 (2.7/1.2–5.9)0.021 (0.020)
Corr.** F on LTG2.5 (3.2)0.5 (0.6)5.0/1.6–17.3 (5.4/1.9–16.2)0.007 (0.002)
VPA1.2 (1.8)0.6 (0.8)1.8/0.6–4.8 (2.2/0.9–5.1)0.278 (0.075)
F on VPA1.4 (1.8)0.8 (1.0)  
Corr.* VPA0.5 (0.8)0.9 (1.1)0.6/0.2–1.6 (0.7/0.3–1.7)0.312 (0.469)
CBZ0.6 (0.8)0.8 (1.1)0.7/0.2–1.9 (0.7/0.3–1.7)0.504 (0.465)
F on CBZ0.2 (0.5)1.3 (1.5)  
Corr.* CBZ    

Estimated incidence of SUDEP in patients treated with LTG

Based on the market share in DDDs, the incidence of definite and probable SUDEP in patients on LTG was 3.9 per 1,000 patient-years, and 0.5 per 1,000 patient-years in those that were not on LTG (p < 0.001, incidence rate ratio 8.6, 95% CI 3.3–21.5) (Table 4).

Because the exact number of individuals with a prescription for each AED was available for the years 2004 and 2005 (The Norwegian Prescription Database: http://www.reseptregisteret.no) (Table 1), we corrected the number of patient-years at risk through the 10-year period for the ratio between the exact number and the originally estimated number at risk in 2005. The corrected incidence of definite and probable SUDEP in patients using LTG was then estimated to be 1.6 per 1,000 patient-years and 0.5 per 1,000 patient-years in those that were not on LTG (p = 0.021, incidence rate ratio 3.1, 95% CI 1.2–7.9).

In 2005, 58% of the patients on LTG in Rogaland County were female (The Norwegian Prescription Database: http://www.reseptregisteret.no). When this (58% females) was also taken into account, assuming the same proportion of female users during the whole 10-year period, the incidence of definite and probable SUDEP in female patients who were taking LTG was estimated to be 2.5 per 1,000 patient-years and in female patients that were not taking LTG the incidence of SUDEP was estimated to be 0.5 per 1,000 patient-years (p = 0.007, incidence rate ratio 5.0, 95% CI 1.6–17.3).

Four of the deaths in patients treated with LTG occurred during the first 5 years of the study period and 6 during the last 5 years.

AEDs in SUDEP victims and controls

Of the 12 female patients with definite and probable SUDEP, 7 (58.3%) were taking LTG, and of the 41 controls, 10 (24.4%) were taking LTG (p = 0.038). However, at the group level (male and female), the difference in the proportion of LTG users between cases and controls did not reach statistical significance. Mean dose and treatment time on LTG did not differ significantly between cases and controls (Table 5).

Table 5.   Antiepileptic drugs in definite and probable SUDEP cases and controlsa
 CasesControlsbp-value
MeanMedianRangeMeanMedianRange
  1. AEDs, antiepileptic drugs; CBZ, carbamazepine; CI, 95% confidence interval; COR, conditional odds ratio; LTG, lamotrigine; VPA, valproate.

  2. aThe complete table (Table S5) including corresponding information about the remaining AEDs and possible SUDEP cases is available in the online version of the article.

  3. bSUDEP victims who were taking LTG compared with controls taking LTG.

LTG dose (mg)197150100–350158100175–3000.228
Months on LTG26.917.50.35–9429.3112–1070.771
 CasesMatched controlsp-valueCORCI 
  % %    
Patients on LTG8/1942.115/6323.80.1112.60.8–8.4
Females on LTG7/1258.310/4124.40.0385.61.1–28.2
Patients on CBZ5/1926.327/6342.90.1500.40.1–1.4
Females on CBZ1/128.314/4134.10.1190.20.02–1.53
Patients on VPA5/1926.314/6322.20.5391.50.4–5.6
Females on VPA3/12259/41220.6571.50.3–8.1
Untreated1/195.37/6311.1   
>2 AEDs2/1910.55/637.9    

There was no statistically significant difference in the proportion of SUDEP victims and the proportion of controls treated with VPA and CBZ . For OXC, PHT, VGB, TPM, and PB, statistical analysis was not performed because of the small number of cases (Table S5). Only 2 of 26 cases and 5 of 86 controls were taking more than two AEDs.

Clinical characteristics

Mean duration of epilepsy did not differ significantly between definite and probable SUDEP victims who were taking LTG and controls (13.25 and 13.59 years, respectively). Likewise, there was no statistically significant difference in mean duration of epilepsy between cases (16.25 years) and controls (15.15 years), when all cases with definite and probable SUDEP were included. In addition, the proportion of cases with a duration of epilepsy >10 years and the share with onset of epilepsy before 16 years of age did not differ significantly between SUDEP victims and controls (Table 2). The proportion of patients with a total seizure count of more than one seizure per week was higher among cases than controls, but the difference was not statistically significant for definite and probable SUDEP. A logistic regression analysis, including both high seizure frequency (more than one per week) and treatment with LTG as possible risk factors, identified LTG as an independent risk factor.

The proportion of patients with generalized or bilateral epileptogenic activity did not differ significantly between SUDEP cases and controls, but among the cases on LTG, 9 of 10 had generalized or bilateral epileptogenic activity compared to 16 of 32 controls. A significantly higher proportion of cases than controls was on concurrent medication (Table 2). Among the SUDEP victims on LTG, 2 of 10 were also being treated with a psychotropic drug and one of 10 with a cardiovascular drug. Only one of 26 had a subtherapeutic serum concentration of an AED, although analyses were missing for at least one AED in 8 of the cases. In comparison, 17 of 86 controls had a subtherapeutic serum concentration (analyses missing for 28 of 86, of whom 9 were untreated). Five of 7 cases on CBZ had plasma concentrations within range and under 40 μm, whereas the plasma concentration was not measured in 2 of 7. In comparison, among the 35 controls on CBZ, 2 of 24 with a measured plasma concentration of CBZ had levels that exceeded 40 μm. At postmortem examination, 5 of 26 cases had subtherapeutic levels of at least one AED and 3 had no levels of their AED. When all cases of SUDEP were included, only one of 18 had an electrolyte level out of range compared with 3 of 49 controls, and all of these abnormalities were considered to be clinically unimportant.

Discussion

Our study indicates an increased risk of SUDEP in female patients taking LTG. To our knowledge this has previously not been found in a clinical study, and, in our opinion, our findings may have implications for the treatment of women with epilepsy.

The total incidence of SUDEP in our county was similar to that which has been reported from previous, population-based studies (Leestma et al., 1989; Tomson et al., 2008). However, the incidence in patients treated with LTG was significantly increased compared with the incidence in patients not treated with this drug. The difference was statistically significant both for male and female patients together and for females, even with a very conservative estimate of the numbers of epilepsy patients at risk (Table 4). LTG was approved in Norway for the treatment of bipolar disorder in early 2004 and an unknown part of the increasing market share of the drug during the last years of the study period reflects prescription for this indication. If this had been accounted for in estimating the number of epilepsy patients at risk, the calculated incidence of SUDEP in LTG users would have been even higher. We also found a statistically significantly higher proportion of LTG users among the female SUDEP victims than among the controls matched for age and gender. However, for male patients and for male and female patients together, the difference in use of LTG between SUDEP victims and controls was not statistically significant.

The preponderance of female SUDEP victims who were taking LTG is strikingly different from the findings in the analysis of the SUDEP cases in the clinical development program of LTG (Leestma et al., 1997), which included 4,700 patients of whom 56% were female. In that study, only 6 (30%) of the 20 SUDEP victims (possible cases included) that were being treated with LTG at death were female. A possible explanation may be that the clinical development program included only patients with localization-related epilepsy, whereas our population-based study also included idiopathic epilepsy. The findings from a recently published combined analysis (Hesdorffer et al., 2011) of two previous studies from England and Scotland, where LTG was associated with a statistically significantly increased risk of SUDEP in patients with IGE, support this explanation. It is possible that a genetic predisposition to cardiac arrhythmia may play a role in SUDEP (Nashef et al., 2007), and we have previously suggested that this may be of particular importance in idiopathic epilepsy (Aurlien et al., 2007, 2009). A cell-based experimental study (Danielsson et al., 2005) has shown that LTG inhibits the cardiac potassium current Ikr, and drugs with this ability have been associated with the long QT syndrome (LQTS), torsade de pointes arrhythmia, syncope, and sudden death (Witchell & Hancox, 2000; Redfern et al., 2003). A possible explanation for the increased risk of SUDEP with LTG in IGE (Hesdorffer et al., 2011) may be a genetically determined increased risk of cardiac arrhythmia, in combination with a drug that, under certain circumstances, could have a proarrhythmic effect. In addition to a possibly increased risk from LTG in the presence of this genetic factor, other risk factors may be high serum concentration of LTG or comedication with other Ikr-blocking drugs (Danielsson et al., 2005). On the other hand, in a study of healthy subjects on therapeutic doses of LTG, the QTc was not prolonged (Dixon et al., 2008). However, this does not exclude the possibility that LTG may be associated with prolonged QTc in epilepsy or, in particular, in idiopathic epilepsy. The QTc may increase during epileptic seizures (Brotherstone et al., 2010), and the risk of arrhythmia may be further increased during GTCS with seizure-induced acidosis (Vereecke & Carmeliet, 2000; Danielsson et al., 2005). If the increased rate of SUDEP associated with LTG in our study is due to a proarrhythmic effect of the drug, this could also explain the preponderance of women, since the risk for drug-induced torsade de pointes arrhythmia is up to three times greater in female than in male patients (Makkar et al., 1993; Lehmann et al.,1999).

We found about the same proportion of IGE in SUDEP victims and controls. Likewise the proportion with IGE of female SUDEP victims who were taking LTG did not differ significantly from that of controls. Interestingly, however, 9 of the 10 SUDEP victims who were taking LTG had generalized or bilateral epileptogenic activity on their electroencephalography. In case 1, a mutation in the LQTS-associated gene, SCN5A, was found postmortem, and we have previously suggested the possibility that this mutation may have caused both the epilepsy and a predisposition to cardiac arrhythmia (Aurlien et al., 2009). Both case 1 and case 11 (primarily considered to be unclassifiable) had partial seizures and generalized epileptogenic activity on their EEG studies, and no significant pathology on cerebral magnetic resonance imaging (MRI), and therefore case 11 could, in our opinion, possibly also have had idiopathic epilepsy. Classifying epilepsy syndromes based on medical history, interictal EEG, and radiologic findings may be associated with uncertainty, but if the assumption about an idiopathic epilepsy in both case 1 and case 11 is correct, then 5 of 7 female definite and probable SUDEP victims who were taking LTG had idiopathic epilepsy compared with 5 of 24 of the controls; this was a statistically significant difference. Although there is still scanty evidence of an association between a genetically determined predisposition to cardiac arrhythmia and idiopathic epilepsy, the existence of this association was recently strongly suggested in a case report of an epilepsy patient with a mutation in the SCN5A gene and a rare variant in the KCNE2 gene, which is also associated with the LQTS (Heron et al., 2010).

Although there was no statistically significant difference between the calculated incidence of SUDEP in patients VPA and CBZ and the incidence in those not treated with these drugs, we are aware that the estimated numbers with epilepsy on these drugs is uncertain, since AEDs are also prescribed for indications other than epilepsy (Oteri et al., 2010). In line with LTG, VPA and CBZ are also used to treat psychiatric disorders, and CBZ is also used in the treatment of neuropathic pain and alcohol abstinence. However, in comparing the proportions of cases with controls who were on these drugs, there was no statistically significant difference, and consequently our study could not confirm the findings of previous studies that have suggested an increased risk of SUDEP with CBZ (Timmings, 1998; Langan et al., 2005). An increased risk has also been associated with plasma levels of CBZ exceeding 40 μm (Nilsson et al., 2001; Tomson et al., 2005), but no indication of this association was found in our study.

The proportion of patients between 20 and 40 years of age was significantly higher among cases than unmatched controls, and this is in line with findings in previous studies (Tomson et al., 2005). There were no statistically significant differences between cases and controls concerning the proportion with epilepsy duration of >10 years, or onset before 16 years of age, and for female patients who were taking LTG the proportion was lower in cases than in controls for both of these parameters. Polytherapy, another potential risk factor for SUDEP (Surges et al., 2009), did not appear to be a likely risk factor in this study. The proportion of patients with concurrent psychotropic medication, which may be associated with an increased risk of cardiac arrhythmia (Alvarez & Pahissa, 2010), was higher in cases than in controls, but among the nine female patients (cases) who were taking LTG, only one was on psychotropic medication (zuclopenthixol). There were three other female patients who were taking LTG with a concurrent medication, but neither these drugs nor zuclopenthixol are, to our knowledge, associated with an increased risk of cardiac arrhythmia.

In the cases with an absent or subtherapeutic level of an AED postmortem, the possibility of noncompliance cannot be ruled out. However, these findings do not permit the conclusion that the drug levels at the time of death were similarly low, since postmortem concentrations of AEDs are generally considered unreliable (Tomson et al.,1998, 2008).

One of the female patients who was taking LTG (case 16) developed a moderate exanthema on the day before SUDEP, and that was believed to represent an allergic reaction to LTG. The drug was, therefore, withdrawn and the patient treated with the antihistamine promethazine. Treatment with LTG had been initiated (50 mg/day) 12 days antemortem, and the dose was increased (to 100 mg/day) 6 days before death. Because of the long half-life of LTG (Rambeck & Wolf, 1993), this case was included as an LTG-associated SUDEP. However, even if this case had been excluded as an LTG-associated SUDEP victim, and with the most conservative estimate of the numbers at risk on LTG (Table 4), then the incidence of definite and probable SUDEP in female patients who were taking LTG was still significantly higher than in women who were not taking this drug. Likewise, the difference between the proportion of female cases on LTG compared with the proportion of controls was still statistically significant when all cases were included, but did not reach statistical significance for definite and probable SUDEP.

One strength of our study is that a statistically significant increased risk of SUDEP in women who were taking LTG has been found both using a conservative estimate of the incidence, and through a significantly higher proportion of cases than controls taking the drug. However, one may argue that correction for multiple comparisons could have been performed. Doing so, the difference would be statistically insignificant probably due to the limited number of cases included in this study. Furthermore, because the primary outcome of the study was the proportion of cases and controls who were taking LTG, in our opinion performing this correction may be disputable. Likewise, a weakness of our study may be that we cannot exclude the possibility that a higher frequency of GTCSs may be a part of the explanation, since the relative frequency of GTCSs in the female cases who were taking LTG compared to their controls was unknown. Nevertheless, in our opinion, our findings indicate that a genuine adverse effect of the drug may be present, and it causes a higher risk of SUDEP in female patients.

LTG has been a preferred AED in the treatment of epilepsy in women of childbearing age (Legros et al., 2009). Our findings of an increased risk of SUDEP in female patients who were taking this drug should be explored in future studies before firm conclusions can be made. Whether a gender difference can be identified in other populations should be investigated, and also whether the increased risk occurs only in idiopathic epilepsy or also in localization-related epilepsy.

Acknowledgments

This study was financed by Stavanger University Hospital and was also supported by the Norwegian Chapter of the International League Against Epilepsy. We would like to thank Inge Morild and Sverre Undheim for reviewing the autopsy reports in the search for possible SUDEP cases, as well as Are Hugo Pripp, Morten Aarflot, and Odd Bjarte Nilsen for statistical assistance; and Hilde Alvseike Dahle for her help with collecting the informed consent from controls. 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

Dag Aurlien has received sponsorship for travel from UCB Pharma, Eisai, and GSK; payment for talks at meetings arranged by GSK; and payment as site investigator in studies conducted by GlaxoWellcome, Janssen-Cilag, and UCB Pharma. Jan Petter Larsen has participated in Advisory Boards for Lundbeck and GSK and has received payment for talks at meetings arranged by Orion Pharma. Leif Gjerstad has received speaker’s honoraria and financial support for conference attendance from GSK, Lundbeck, and UCB Pharma. Erik Taubøll has received speaker’s honoraria and financial support for conference attendance from UCB Pharma and GSK, and an unrestricted research grant from GSK (2007).

Ancillary