The incidence of SUDEP

DISCUSSION

In this nationwide population-based study of SUDEP among all ages, we applied 2 different SUDEP classifications and noted that the overall incidence of SUDEP is very similar between the two. Including definite, probable, and possible SUDEP according to the Annegers criteria, or definite, probable, definite SUDEP Plus, and possible as defined by Nashef et al., the incidence was 1.74 per 1,000 person-years, and 1.20 and 1.24, respectively, with possible excluded. In accordance with previous reports,^4,–,6 the incidence was higher among male patients. The sex difference was most pronounced in the young age group, with 7 definite/probable cases among boys and none among girls. In contrast to most previous studies,^3,7 we found similar incidence rates of definite/probable SUDEP in all age groups, including in those younger than 16 years as well as those over 50. Most remarkable is the high incidence (2.05 per 1,000 person-years) among boys under 16 years, all of whom had treatment-resistant epilepsy (table e-3). The observed high proportion of possible SUDEP in patients older than 50 is hardly surprising considering the increasing occurrence of competing causes of death as well as lower autopsy rates among older persons. This highlights the difficulties in obtaining a reliable assessment of the SUDEP risk in older people, and our estimates based on definite/probable SUDEP should probably be considered to be minimum numbers and suggest a higher incidence than previously thought. Psychiatric comorbidity has been shown to be associated with premature mortality from external causes in people with epilepsy, although SUDEP was not analyzed in that study.¹⁴ We found a more than 2-fold higher SUDEP incidence in those with psychiatric comorbidity compared to those without, and an almost 5-fold difference among female patients (table 3). The numbers did not allow us to determine differences across subgroups of neuropsychiatric comorbidities, but the highest incidence was noted with organic mental disorders (table e-4). Whether the higher incidence among those with psychiatric comorbidity is related to other SUDEP risk factors such as poor seizure control, concurrent medication, or poorer adherence to the prescribed medication is unclear. Finally, our study confirms the difficulties in using death certificates to identify SUDEP cases as the indicated immediate cause of death was SUDEP in only one case, and SUDEP, seizures, or epilepsy together accounted for no more than 27/99 cases.

The main strength of our study lies in it being nationwide and population-based. Considering that the total Swedish population in 2002 was 8.925 million, the 78,524 individuals identified in the SNPR during 1998–2005 most likely represent the vast majority of inhabitants with epilepsy. A further unique strength is the review of all deaths during 1 year, with meticulous assessment of all relevant documentation for all potential SUDEP cases.

Limitations include that we relied on the ICD codes for epilepsy in the SNPR to define the study population. Our review of medical charts of 660 potential cases, however, suggests a high validity, as more than 90% of reviewed individuals had epilepsy. Assuming that the same degree of misdiagnosis applies for the entire study population (which remains to be shown), our estimates of the SUDEP incidence may thus underestimate the incidence by approximately 10%. Another limitation is that our detailed review of all deaths covered 2008, whereas the study population as well as the occurrence of psychiatric comorbidity was defined by a diagnosis in the SNPR anytime between 1998 and 2005. Fatalities from time of inclusion up to December 31, 2007, were thus not included in our analysis, which was restricted to those 57,775 persons alive on January 1, 2008. Likewise, persons with epilepsy onset after December 31, 2005, were not included in our analysis. Hence, we determined the incidence of SUDEP during 1 year in a prevalent epilepsy population and did not follow newly diagnosed patients from epilepsy onset.

Due to convincing evidence of non-SUDEP causes of death on the death certificate, we restricted our comprehensive review of all documents to 660 individuals. We cannot exclude the possibility of missing a few SUDEP cases in this procedure. But should this be the case, they would most likely have been possible SUDEPs due to competing causes of death and thus not have altered our estimates of the incidence of definite and probable SUDEP. A further limitation of our study is the rather small number of cases, hampering subgroup analysis.

In terms of study design, a nationwide Danish study of SUDEP in the young is the one most similar to ours, using Danish health registers to define the epilepsy study population, and review of death certificates to identify SUDEP cases.¹ There are, however, important differences: The Danish study was restricted to ages 1–35 years. Furthermore, the occurrence of SUDEP was determined based on review of death certificates and autopsy records, but not including medical charts, as in our study. Our results, with SUDEP, seizures, or epilepsy listed as the immediate cause of death in only 27% of SUDEP cases, and epilepsy mentioned in any position on the death certificate in 63.6%, clearly show the risk of underestimation of SUDEP incidence with this approach, possibly contributing to the lower incidence of definite and probable SUDEP, 0.41 (0.32–0.55) per 1,000 person-years, in the Danish study.

In previous population-based studies including all age groups, the estimated incidence of definite and probable SUDEP ranged from 0.33 to 1.35 per 1,000 person-years.^{2,4,–,6,8,10,–,12,22} The wide range could depend on methodologic differences in terms of study populations, SUDEP criteria, and not least for identification of SUDEP cases, in some based on coroner or medical examiners,^4,–,6 some based mostly on death certificates.¹ Despite some differences in methodology, our observed incidence of 1.20 (0.93–1.52) per 1,000 person-years is similar to the rate of 1.16 (0.95–1.36) reported in a pooled analysis³ of 3 high-quality population-based studies.^4,–,6 However, our findings of similar incidence rates across age groups including children and people above 50 years of age differ markedly from what has been reported in the past.^3,23,–,25 Our incidence rate of 1.11 (0.45–2.29) <16 years of age is considerably higher than 0.20 and 0.43 per 1,000 reported in 2 community-based studies of children with epilepsy,^24,–,26 and the 0.33 per 1,000 person-years reported in a pooled analysis of 4 cohorts of childhood onset epilepsy.²⁶ While our data in children should be interpreted with caution considering the small numbers, our comprehensive approach where we review all possible sources of information to identify people with epilepsy, and the cause of their deaths, provides more complete data. The fact that all SUDEP cases in the young age group were in boys is intriguing. We have no explanation but this observation should prompt further studies. All SUDEP children had refractory focal epilepsy, most with frequent generalized tonic-clonic seizures (table e-3), which highlights the importance of early consideration of epilepsy surgery for children to possibly prevent this outcome. Despite the refractory nature of all the cases and some having an epileptic encephalopathy, there was no Dravet case among them. Importantly, in children with epilepsy, the relative contribution of SUDEP to the overall mortality was no less than 36%, compared to 21% in ages 16–50 and only 3% among epilepsy patients above 50 years of age. The lower relative contribution in higher age groups is evidently due to mortality from other causes becoming more frequent with increasing age. Nevertheless, SUDEP risk in the older age group is higher than previously thought. Our results thus corroborate the opinion of Thurman et al.³ and Devinsky et al.²⁷ that the undercount of SUDEP might be great in older age groups. In fact, it is possible that our rates based on definite and probable SUDEP underestimate the risk given the substantially greater number of possible SUDEP cases in the higher age group (figure e-1). Our current experience is that there is less information and documentation to be found regarding the circumstances of death in the older age groups compared to the younger, which likely contributes to underascertainment.

Being nationwide and population-based, our results are truly representative for the incidence of SUDEP in the Swedish prevalent epilepsy population. We can only assume that they are generalizable to other countries with similar living conditions and health care. Our novel findings of significant SUDEP risk for children, in particular boys, for older people, and for women with psychiatric comorbidities need to be confirmed. The findings should, however, be taken into account in the important individualized patient counseling about epilepsy and its associated risks. More research into risk factors for SUDEP is urgently needed. We are currently extending the review of deaths in our study population to cover more years. This will increase the number of SUDEP cases and allow for a case-control study where we will compare characteristics of SUDEP cases to living epilepsy controls in detail.