Ding Ding and Wenzhi Wang contributed equally to this paper.
Address correspondence to Zhen Hong, Institute of Neurology, Huashan Hospital, Fudan University, #12 Wulumuqi Zhong Rd., Shanghai 200040, China. E-mail: firstname.lastname@example.org Ley Sander, UCL Institute of Neurology, Box 29, Queen Square, London WC1N 3BG, U.K. E-mail: email@example.com
Purpose: Detailed data on the mortality of epilepsy are still lacking from resource-poor settings. We conducted a long-term follow-up survey in a cohort of people with convulsive epilepsy in rural areas of China. In this longitudinal prospective study we investigated the causes of death and premature mortality risk among people with epilepsy.
Methods: We attempted to trace all 2,455 people who had previously participated in a pragmatic assessment of epilepsy management at the primary health level. Putative causes of death were recorded for those who died, according to the International Classification of Diseases. We estimated proportional mortality ratios (PMRs) for each cause, and standardized mortality ratios (SMRs) for each age-group and cause. Survival analysis was used to detect risk factors associated with increased mortality.
Key Findings: During 6.1 years of follow-up there were 206 reported deaths among the 1,986 people with epilepsy who were located. The highest PMRs were for cerebrovascular disease (15%), drowning (14%), self-inflicted injury (13%), and status epilepticus (6%), with probable sudden unexpected death in epilepsy (SUDEP) in 1%. The risk of premature death was 2.9 times greater in people with epilepsy than in the general population. A much higher risk (SMRs 28–37) was found in young people. Duration of epilepsy and living in a waterside area were independent predictors for drowning.
Significance: Drowning and status epilepticus were important, possibly preventable, causes of death. Predictors of increasing mortality suggest interventions with efficient treatment and education to prevent premature mortality among people with epilepsy in resource-poor settings.
More than 80% of the worldwide burden of epilepsy is found in resource-poor settings (de Boer et al., 2008). Many people with epilepsy worldwide remain untreated, particularly in rural locations. Epilepsy is estimated to affect up to 9,000,000 people in China and a treatment gap of 63% was reported from rural areas (Wang et al., 2003).
In 2008, we conducted a second follow-up survey in one of the original cohorts. This longitudinal prospective study allowed us to explore the pattern of death causality in a larger sample size of people with epilepsy who died and to investigate risk factors associated with premature mortality.
Target cohort of people with epilepsy
Between 2000 and 2004, as part of an epilepsy management program, a pragmatic intervention was undertaken in eight rural counties of six provinces in China (Wang et al., 2006). Counties in Heilongjiang, Henan, and Shanxi provinces were defined as mountainous areas, and counties in Ningxia, Jiangsu, and Shanghai as waterside areas. These sample areas covered a population of 3,185,000 people, from whom 2,455 people with convulsive epilepsy were recruited and treated with phenobarbital monotherapy (Wang et al., 2006).
This cohort was previously followed to 2004, with a median follow-up of 25 months (Ding et al., 2006). We conducted a second survey in 2008. Village physicians at each study site were trained to trace people from the original cohort, and to use the questionnaires. The physicians contacted each person by home visit or telephone to introduce the follow-up survey and to ask if they would participate. Informed consent was obtained from the individuals or their guardians. Village physicians completed the questionnaires with those who agreed to be interviewed. Family members completed the questionnaires for those who were ill, those with learning disability, or those who had died. The questionnaire covered (1) basic demographics, (2) current clinical characteristics, and (3) survival status (alive or dead). For those who died, death certificates were located, where possible, to ascertain the date and cause of death. The local physicians also conducted verbal autopsies (Aspray, 2005).
This study was approved by the Beijing Neurosurgical Institute Medical Ethics Committee, China.
Confirmation and classification of cause of death
All information was consolidated centrally. Putative causes of death were attributed by consensus after discussion and consideration of all the available information, by WW, WJ, DD, and JWS, according to the International Classification of Diseases, 10th Edition (ICD-10). Sudden unexpected death in epilepsy (SUDEP) is defined as sudden unexpected death in an individual with epilepsy with or without evidence of a seizure, and excluding documented status epilepticus, where postmortem examination does not reveal an anatomic or toxicologic cause of death (Nashef, 1997). Because full postmortem and toxicologic data are not available in rural China, we attributed probable SUDEP where there was no other plausible cause of death, and foul play was not suspected (Annegers, 1997).
Continuous variables are expressed as mean and standard deviation (SD), or as median (range, minimum to maximum) if the variable is not normally distributed; categorical variables are given as frequencies (%). Comparisons for continuous variables were evaluated with the Student t-test, or the Mann-Whitney U-test when the data were not normally distributed. Chi-square test was used for comparisons of categorical variables.
Proportional mortality ratios (PMRs) were calculated as the proportion of deaths due to a specific cause in those who died (Logroscino & Hesdorffer, 2005). SMRs, the ratios of the observed number of deaths in the epilepsy population to those expected (Logroscino & Hesdorffer, 2005) were estimated using age- and gender-specific mortality rates in the Chinese population in 2004 to estimate the expected number of deaths. Specific SMRs for gender, age-group, and cause of death were also estimated, and results presented with 95% confidence intervals (CIs) (Vandenbroucke, 1982). Individuals were censored if they were lost to follow-up. Cox proportional hazards regression analysis was used to investigate risk factors associated with mortality. The covariates considered were gender, age of onset of epilepsy, duration, and living environment. Adjusted hazard ratios (HRs) with 95% CIs were estimated.
All p-values and CIs were estimated in a two-tailed fashion. Differences were considered to be statistically significant at p < 0.05. Data were analyzed using SPSS 13.0 (SPSS Inc., Chicago, IL, U.S.A.).
Of 2,455 people with epilepsy, 469 (19%) were lost to follow-up (340 could not be traced, 106 were known to be living out of the study sites at the time of the follow-up assessment and could not be contacted, and 23 refused to participate). Therefore, 1,986 people (81%, 1,129 [57%] male) were traced and surveyed. The average duration of epilepsy was 22.5 (SD 12.9) years. The mean time since baseline interview was 6.1 (SD 1.1) years.
Characteristics of those who died
By 2008, 206 people had died (case fatality rate 10%) with mean age at death 46.0 (SD 18.7) years. There was no statistical difference in gender distribution between those who died and those who survived (p = 0.142). The age of onset of epilepsy and disease duration of those who died were significantly higher than in those who survived (Table 1). Of those who died, 38 (18%) died within 10 years of the onset of epilepsy, 87 (42%) within 10–30 years, and 81 (39%) >30 years after onset.
Table 1. Demographic characteristics of those who survived and those who died
Those who survived
Those who died
aAge at follow-up.
bAge at death.
SD, standard deviation.
Gender, male (%)
Age in years, mean (SD)
Onset age in years, mean (SD)
Duration in years, mean (SD)
Follow-up time in years, mean (SD)
Living in waterside area
Cause-specific PMRs and SMRs
Cause-specific PMRs are shown in Table 2. The main causes of death were cerebrovascular disease (14.6%), drowning (13.6%), and status epilepticus (13.1%).
Table 2. Cause-specific PMRs in people with epilepsy in rural China
Cause of death
Number of cases
Neoplasms outside the central nervous system
Diseases of the digestive system
Ischemic heart disease
Other diseases of the circulatory system
Pneumonia and lung diseases
Toxic effects of carbon monoxide and pesticides
Diseases of the genitourinary system
The overall SMR was 2.9 (95% CI 2.6–3.4). Among all causes of death, the risk of drowning was highest, almost 40 times that in the general population (SMR 39.0, 95% CI 26.4–55.5). Cause-specific SMRs were significantly elevated for toxic effects of carbon monoxide and pesticides (17.0); falls (9.8); suicide (8.2); traffic accidents (6.0); myocardial infarction (3.6); diseases of the digestive system (4.4); and cerebrovascular disease (2.2) (Table 3).
Table 3. Cause-specific SMRs in people with epilepsy in rural China
Overall (95% CI)
Male (95% CI)
Female (95% CI)
aIncluding brain tumors and neoplasms outside the central nervous system.
Toxic effects of carbon monoxide and pesticides
Diseases of the digestive system
People with epilepsy between 10 and 69 years of age had higher mortality rates than the general population (Fig. 1), and SMRs were especially high (between 27.7 and 36.6) in those from 10–29 years of age. The SMRs in people >30 years old decreased with increasing age and were no longer significantly elevated by age 70 and older.
Factors associated with mortality risk
Survival analysis suggests that compared with people whose onset was after 12 years of age, those who developed epilepsy before 12 years of age had a 10.2-fold increased risk of death due to status epilepticus (hazard ratio [HR] 10.2, 95% CI 3.2–32.3, p < 0.001). Duration of epilepsy was also significantly associated with the mortality risk for death due to drowning, status epilepticus, and suicides. The highest risk was found at 10–20 years after onset. For death related to drowning, the risk was greater for those living in a waterside area than for those living in the mountains (HR 3.9, 95% CI 1.7–9.2, p = 0.002). Living in a waterside area appeared to be a protective factor for death from cerebrovascular disease (HR 0.4, 95% CI 0.2–0.9, p = 0.02) (Table 4).
Table 4. Adjusted hazard ratios for factors associated with mortality in subgroups with cerebrovascular disease, drowning, status epilepticus, and suicide (the most frequent causes of death in the cohort)
Status epilepticus (n = 27) (model p-value < 0.001)
Suicide (n = 12) (model p-value = 0.005)
aAdjusted for gender and the other variables in the table.
*p < 0.05; **p < 0.01.
Onset age (reference: >12 years old)
≤12 years old
10.2 (3.2, 32.3)**
2.6 (0.4, 19.3)
Duration (reference: 0–10 years)
>10 and ≤20 years
22.8 (5.1, 95.5)**
38.1 (2.8, 520.8)**
>20 and ≤30 years
7.5 (1.9, 29.1)**
15.5 (1.3, 192.1)*
3.2 (0.9, 11.4)
2.3 (0.1, 39.2)
Living environment (reference: Mountainous area)
1.5 (0.7, 3.2)
0.6 (0.2, 2.0)
In this prospective cohort of people with epilepsy in a resource-poor setting, the overall risk of death during the study period was 2.9 times as great as in the general population, with the highest risk in those 10–29 years of age. Duration of epilepsy of >10 years and living in a waterside area were independent predictors for increasing mortality due to drowning.
Mortality in this cohort has been studied once before. There were 35 deaths among 2,455 people (case fatality rate 1.4%) between 2000 and 2004. The case fatality rate increased to 10% in this longer follow-up. There was a large increase in deaths due to status epilepticus (from 0% to 13%). Putative cause of death was not established for 14% of people, as insufficient information was available.
In the first follow-up period, 91 people were lost to follow-up and 471 people withdrew from the study (Ding et al., 2006). In the second follow-up, we tried to trace all the 2,455 subjects initially recruited, regardless of the treatment condition. The changes in PMRs and SMRs between the two follow-up periods might be due to the greater number of people about whom we now have information. Of the original cohort, 81% were followed up but there were still 340 persons who could not be traced. Any missing data may under- or overestimate the mortality risk, although calculations of SMRs and survival analysis both take follow-up time into account.
In the first follow-up period, 57% of 1,324 people with epilepsy who remained in the study for 2 years were effectively treated with phenobarbital (with seizure frequency reduced by at least 75%) (Wang et al., 2006). At this point, the program with free phenobarbital supply ended, although participants were encouraged to continue taking phenobarbital with an assured supply at very low cost. We believe that the treatment rate of this cohort decreased for several reasons. Some stopped medication as they became seizure free, presumably assuming erroneously that their epilepsy was “cured” (Wang et al., 2006); some switched medications for personal preference; and in some withdrawal was due to socioeconomic factors, such as unavailability of phenobarbital or medical care, or unaffordability of treatment, once the original pragmatic trial was completed. In this cohort, some people died of status epilepticus and seizure-related injuries, and these may be related to poor adherence to antiepileptic treatment. This may suggest people with epilepsy need more education on the rationale of treatment and the importance of continuing treatment through drug titration and beyond.
Few studies have identified factors associated with mortality risk in people with epilepsy using survival analysis. A study in the United Kingdom reported the long-term survival of a cohort with chronic epilepsy that had been followed for 20 years. Multivariate survival analysis showed that older onset age and male sex were significantly associated with reduced survival (Mohanraj et al., 2006). Among a phenobarbital-treated cohort of people with epilepsy in the Sichuan province of China, male sex was reported as a risk factor, and higher phenobarbital dose as a protective factor of mortality risk (Mu et al., 2011). In the current study, we did not find that gender affected the mortality risk. Subgroup analysis of cause of death suggested that, as well as disease duration, living in a waterside area was an independent predictor of increased mortality due to drowning. The Sichuan study estimated that the risk of drowning was 82 times higher in that cohort than in the general population, probably because Sichuan has extensive water resources and people often live close to water (Mu et al., 2011). In the current study, we confirmed the contribution of geographic and living environment to the premature mortality risk.
Studies from developed countries suggest that SUDEP is responsible for up to 17% of deaths in epilepsy (Lhatoo & Sander, 2005). Data concerning SUDEP reported from resource-poor settings is less accurate, since postmortem examinations are the exception. The Sichuan study reported 15 probable SUDEP cases with a PMR of 14.7% during the 2-year follow up (Mu et al., 2011). In the current study with a longer follow-up period, only two probable SUDEP cases (PMR 1%) were identified. We found, however, 30 people whose death was assigned to cerebrovascular diseases (PMR 15%), and 7 from ischemic heart disease (PMR 3%); some of these deaths may well have been due to SUDEP. In addition, SUDEP may have been the cause of death in some of those ascribed to status epilepticus, or from unknown cause. More detailed verbal autopsies including descriptions of the death by witnesses may provide more evidence.
Recent studies in developed regions report PMRs for suicide of between 0% and 20% and SMRs between 1 and 5.8 (Gaitatzis & Sander, 2004), with an overall SMR of 3.3 (Bell et al., 2009). We identified 12 people who died because of self-inflicted injuries (PMR 6%) and a higher SMR of 8. Some causes of death, for example, toxic effects of carbon monoxide and pesticides, traffic accidents, and drowning, could also have been self-inflicted.
Very few people in this cohort were followed from the onset of epilepsy. This prevalence study may overrepresent people with more severe forms of epilepsy, while simultaneously missing some people who may have died previously for whatever reason. It is likely, therefore, that results from incident cohorts provide a more accurate picture (Neligan et al., 2010). Several incidence studies suggest that the risk of premature mortality is significantly higher than that in the general population in the initial 1–2 years following onset, and that this risk decreases thereafter (Hauser et al., 1980; Cockerell et al., 1994; Olafsson et al., 1998; Loiseau et al., 1999; Neligan et al., 2011). If this is the case, the long-term mortality risk of people with newly diagnosed epilepsy may be even higher than that of the current study.
Whether or not the risk of premature mortality is preventable ultimately depends on what is driving this higher mortality rate (Neligan & Sander, 2011). Our results may suggest interventions to prevent premature mortality. For instance, training village physicians in treatment of status epilepticus, rendering people with severe epilepsy seizure-free by treatment, educating people living in waterside areas regarding the risk of drowning, and treating people with epilepsy and depression appropriately may further modify the risk.
We thank all the participants for their cooperation. We also wish to thank the local health workers in the different study sites of China for their help and support. This project was funded by the Chinese Ministry of Health and Stichting Epilepsie Instellingen Nederland. GSB and JWS are based at University College London Hospitals/University College London, which receives a proportion of funding from the Department of Health’s National Institute for Health Research Biomedical Research Centres funding scheme. JWS is supported by the Dr Marvin Weil Epilepsy Research Fund.
ZH serves on the editorial advisory boards of the Chinese Journal of Neurology; has received research support from the Chinese Ministry of Health and the China Association Against Epilepsy; and has been a commissioner of the neurology branch of the Chinese Medical Association. GSB’s husband works for, and has shares in, GlaxoSmithKline. PK serves on the scientific advisory board of GSK, Pfizer, and UCB Pharma; speaker’s bureau of GSK, Sanofi, and UCB Pharma; provides consultancy to Eisai and GSK; and receives research grants from Eisai, Johnson & Johnson, Pfizer, and UCB Pharma. JWS served on scientific advisory boards for GlaxoSmithKline, UCB, and Viropharma; has received funding for travel from UCB and Janssen; serves on the editorial boards of Lancet Neurology and Epileptic Disorders; serves on the speaker’s bureaus of UCB and GlaxoSmithKline; and has received research support from UCB, GSK, the National Institutes of Health (NIH), the European Union Seventh Framework Programme, the Wellcome Trust, World Health Organization (WHO), the National Epilepsy Funds of the Netherlands, and the Epilepsy Society. The other authors have no competing interests. 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.
This work was conceptualized by JWS, WW, JW, and DD. Data collection was done by DD, ZH, GM, XD, BW, TW, and CY. Analysis was undertaken by DD, GSB, and JWS. The first draft was prepared by DD, WW, and JWS, and all authors contributed to and approved the final draft. DD, WW, GSB, PK, and JWS contributed to the interpretation of data. WW, ZH, HB, SL, and JWS organized the funding. ZH and JWS are the guarantors.