Address correspondence to Prof. Ley Sander, Box 29, Department of Clinical & Experimental Epilepsy, UCL Institute of Neurology, Queen Square, London WC1N 3BG, U.K. E-mail: email@example.com
Purpose: Suicide is more common in populations with epilepsy, but estimates vary concerning the magnitude of the risk. We aimed to estimate the risk using meta-analysis.
Methods: A literature search identified 74 articles (76 cohorts of people with epilepsy) in whom the number of deaths by suicide in people with epilepsy and the number of person–years at risk could be estimated. Standardized mortality ratios (SMRs) with 95% confidence intervals (CIs) were calculated for each cohort, for groups of cohorts, and for the total population.
Results: The overall SMR was 3.3 (95% CI 2.8–3.7) based on 190 observed deaths by suicide compared with 58.4 expected. The SMR was significantly increased in people with incident or newly diagnosed epilepsy in the community (SMR 2.1), in populations with mixed prevalence and incidence cases (SMR 3.6), in those with prevalent epilepsy (SMR 4.8), in people in institutions (SMR 4.6), in people seen in tertiary care clinics (SMR 2.28), in people with temporal lobe epilepsy (SMR 6.6), in those following temporal lobe excision (SMR 13.9), and following other forms of epilepsy surgery (SMR 6.4). The SMR was significantly low overall in two community-based studies of people with epilepsy and developmental disability.
Discussion: We confirm that the risk of suicide is increased in most populations of people with epilepsy. Psychiatric comorbidity has been demonstrated to be a risk factor for suicide in the general population and in people with epilepsy, and such comorbidity should thus be identified and treated.
Suicide is often said to be more common in populations with epilepsy. It has been reported that suicide in epilepsy may occur at the same rate as that reported among patients with manic-depressive illness (Blumer et al., 2002). Several attempts to ascertain the rate of suicide have been made, but many have been beset by methodologic problems. It has been said that suicide accounts for more than 10% of all deaths in people with epilepsy (compared with 1.4% in the general population) (Robertson, 1997; Jones et al., 2003), although overall the data presented suggest that less than 5% of deaths in people with epilepsy were due to suicide.
A meta-analysis of suicide in epilepsy included 29 studies (Pompili et al., 2005). The report did not provide a standardized mortality ratio (SMR), but from the data provided this can be estimated as 8.5. Several of these studies, however, may have been inappropriately included, as data from some cohorts were included twice and in others the data are not available to calculate the person–years at risk. A meta-analysis of suicide published in 1997 (Harris & Barraclough, 1997), however, used 12 published studies to provide an overall SMR of suicide in people with epilepsy of 5.1 [95% confidence interval (CI) 3.9–6.6]. Each study used in that meta-analysis was considered suitable for inclusion in the present study, with the exception of two studies that have since been updated.
Suicide appears to be a serious problem, particularly among those with chronic epilepsy who require treatment in specialty clinics (Blumer et al., 2002). One case control study, in adults who had been hospitalized with epilepsy, found a marked increase in relative risk for suicide associated with psychiatric comorbidity, and with the use of antipsychotic drugs (Nilsson et al., 2002). Risk also seemed to increase with high seizure frequency and antiepileptic polytherapy, although the associations were not significant. In contrast to this, however, others have found that suicide may occur in patients with epilepsy with longstanding complex partial seizures and dysphoric disorder shortly after full control of seizures (Blumer et al., 2002). A recent large population-based case control study in Denmark found a 3-fold increased risk of suicide in people with epilepsy [rate ratio 3.17, 95% CI 2.88–3.50) (Christensen et al., 2007). In Iceland, a history of attempted suicide was shown to increase the risk of development of seizures, independent of major depression, suggesting that the mechanisms producing suicidal behavior and depression may both be important in the development of epilepsy (Hesdorffer et al., 2006).
Little is known of the risk of death from suicide in people with prevalent epilepsy in the community. Using hospital cohorts for investigation of death may introduce a selection bias toward people with more severe epilepsy, and those with comorbid disorders that may influence mortality rates (Nilsson et al., 1997). We report the results of a meta-analysis of suicide in people with epilepsy in any setting.
Literature searches were performed for this meta-analysis and for a parallel meta-analysis of drowning in epilepsy (Bell et al., 2008). The search included the terms “epilepsy and drowning,”“epilepsy and mortality and follow-up,”“epilepsy and death,” and “epilepsy and suicide”. In addition, we searched the authors’ collections and the references of review articles, and conducted hand searches of specialist journals from 2000 until February 2008. Studies based on cohorts of people presenting with status epilepticus alone, and those dealing exclusively with rare syndromes were excluded. Publications in which no deaths were reported were also excluded for pragmatic reasons. For a study to be included in this meta-analysis either the number of deaths by suicide needed to be stated, or else it should be possible to conclude that there were no deaths by suicide, as all causes of death were stated and did not include suicide.
Data were extracted independently by two coauthors and, where differences arose, discussions ensued until consensus. Wherever possible, the number of person–years of follow-up (PYFU) was taken directly from the articles. This was not always possible, and in many cases was calculated from the number of patients and the mean duration of follow-up, where supplied. If no mean value was provided, this was estimated as being the same as the median, or as the average of the minimum and maximum follow-up duration. When numbers of patients in the study at certain time points were provided, the number of PYFU was calculated for each time point (as the midpoint between that time point and the previous one, multiplied by the number of patients) and the numbers added. Those cohorts in which it was difficult to establish the PYFU with confidence were excluded from one of the analyses.
From the original search, more than 2,000 publications were found. By February 2008, 273 were considered as possibly holding relevant information for death due to suicide in people with epilepsy. Twenty-three were rejected on the basis of the abstract; the remaining 250 articles were read in full. Of these, 176 (70%) were rejected as not suitable for inclusion: 20 publications had been superseded by further analysis in whole or part of the same cohort; 22 were not follow-up studies (e.g., review articles, editorials, or case reports); 38 lacked data on the cause of death; in 38 it was not possible to calculate the number of person–years at risk (for example, studies starting with deaths in people with epilepsy); and 21 provided information on some causes of death, but not all (and hence the number of deaths due to suicide could not be presumed). The remaining 37 were excluded for a variety of reasons, such as having no information on death at all. For some cohorts the cause of death for all patients who died were listed and did not mention any deaths due to suicide; these cohorts were included (with no deaths due to suicide). In all 74 papers were found in which information on deaths due to suicide was provided or could be calculated.
Two studies; each provided separate data on two distinct cohorts of people with epilepsy: the first study (Mohanraj et al., 2006) provided data on one cohort with newly diagnosed epilepsy and another with chronic active epilepsy, whereas the other (Nilsson et al., 2003) included a surgical cohort and a cohort with presurgical evaluation who did not proceed to surgery, and so each study contributed two cohorts to the analyses. Because the death rate from suicide is known to be different in men and women, in those that provided separate data for each sex the expected number of deaths were calculated for each sex separately and the numbers added.
Population numbers and the number of deaths due to suicide for each country in which studies took place were usually taken from the World Health Organization (WHO) Statistical Information Service (World Health Organization, 2006). Wherever possible, the data were used from the date of the midpoint at which the study was conducted. For many of the earlier studies, however, data were not available, in which case the earliest data on population statistics provided by the WHO were used. An exception to this was studies conducted in the United Kingdom; here the population data for earlier studies was taken from the United Kingdom Office for National Statistics CD “20th Century Mortality” (Office for National Statistics, 2003). Population data for a few studies proved to be difficult to find. One study, of people with intractable epilepsy taking part in a trial of a potential new antiepileptic drug, provided the number of patient–years for patients from America and for patients in the rest of the world (Leestma et al., 1997). The majority of the patients from the rest of the world were from Europe, particularly the United Kingdom, Spain, France, Germany, The Netherlands, and Italy, and so an overall rate for death by suicide was estimated from the rates in these six countries. One study took place in Africa (Kamgno et al., 2003) and one in India (Udani et al., 1993), where few data are available. For the African study the population data and numbers of deaths by suicide were taken from the estimations of the 2000 Global Burden of Disease (Murray et al., 2001). For the study of children with epilepsy in west India (Udani et al., 1993) published data on suicide in children in south India were used (Bose et al., 2006). Few of the reports provided information on age at death, so background population data used were taken from the age group nearest to the ages covered in the study.
SMRs were calculated by the standard person–years method (Vaeth, 2005). Expected deaths due to suicide in each cohort were calculated by applying the relevant population rate to the number of patient–years at risk in the study cohort unless expected numbers of deaths were provided by the publication (in which case these were used). SMRs were estimated as the observed deaths divided by the expected deaths. Total expected deaths were also calculated for all cohorts combined by adding the numbers of expected deaths in the individual cohorts. Total observed deaths were calculated similarly and hence the overall SMR was estimated. This analysis was repeated for groups of cohorts. An SMR was calculated including only cohorts with any suicide deaths for one analysis only. Confidence intervals were calculated based on the Poisson distribution using CIA software (Altman et al., 2000).
Using all the suitable published articles located, 190 deaths by suicide were reported in people with epilepsy over 396,865 person–years (one in 2,089 person–years), compared with 58.4 expected using population rates, giving an SMR of 3.25 (95% CI 2.81–3.75) (Tables 1 and 2; Fig. 1). Excluding the 24 cohorts in which the data were difficult to ascertain accurately, made little difference to the SMR (3.00, 95% CI 2.54–3.53). If cohorts were excluded in which there were no deaths by suicide reported, then the SMR was 3.49 (95% CI 3.01–4.03).
Table 1. Reports including deaths by suicide (45 populations of people with epilepsy)
The SMR for suicide was significantly raised overall in most groups of studies: people with incident or newly diagnosed epilepsy in the community, SMR 2.13, 95% CI 1.35–3.19; community based-studies with mixed prevalence and incidence cases, SMR 3.62, 95% CI 2.81–4.59; in people with prevalent epilepsy in the community, SMR 4.81, 95% CI 3.08–7.16; in people seen in tertiary care clinics, SMR 2.28, 95% CI 1.18–3.98; in people in epilepsy institutions, SMR 4.64, 95% CI 2.87–7.10; in those with temporal lobe epilepsy, SMR 6.57, 95% CI 1.79–16.8; in those following temporal lobe excision, SMR 13.9, 95% CI 8.93–20.74; and in those following other forms of epilepsy surgery, SMR 6.37, 95% CI 3.06–11.72 (Fig. 2). The SMR was significantly low (0.34, 95% CI 0.070–0.99) overall in the two community-based studies of people with epilepsy and developmental disability [although only the one larger study (Day et al., 2005) had a low SMR]. Of the 34 postsurgery patients who committed suicide, in 19 there is no information on seizure control available, 4 were known to be seizure-free, 6 probably had improved seizure control, 2 still had seizures, and in 3 seizures were worse. The SMR was not significantly increased in cohorts in which children contributed most of the person–years (SMR 1.65, 95% CI 0.54–3.86). It was increased, however, in those in which children did not figure significantly (SMR 3.49, 95% CI 2.75–4.37). SMRs from studies grouped by the midpoint of the study are shown in Fig. 3.
The meta-analysis confirms that the rate of suicide is increased in people with epilepsy, with only one study apparently showing a protective effect of epilepsy (Day et al., 2005). The subgroup analysis, although using generalizations to combine the studies, seems to confirm that, although the death rate is increased in community-based incident and new diagnosis studies, it may not be as high as studies in hospital settings have previously suggested. The results confirm the increased SMR in groups with chronic severe epilepsy, and in those following surgery, as found by others (Harris & Barraclough, 1997; Pompili et al., 2006). The SMR for suicide is said to be increased in those with temporal lobe epilepsy (Robertson, 1997), and presumably those undergoing temporal lobe excision all had intractable temporal lobe epilepsy. In addition, in the “various” surgery group many of the patient–years were provided by people with temporal lobe resections. One might expect the SMR for suicide to be reduced (as sometimes noted for other epilepsy-related causes of death) in people with successful surgery for seizures. As previously noted, however, one study found that suicide tended to occur shortly after full control of seizures was achieved (Blumer et al., 2002). There is insufficient information available to speculate on the causes for the raised SMRs in these surgical groups, as the postoperative seizure status in the few in whom it was stated was varied. Surgery, however, provides great benefits to many people with partial onset epilepsy.
It is clear that the overall SMR is affected by the cohorts studied. This meta-analysis included 25 cohorts who had epilepsy surgery and this will have increased the overall SMR. The broad inclusion criteria aimed to include all relevant studies; many, however, did not include “suicide” in their search terms and it is unlikely that we found all relevant studies. Nonetheless, Fig. 2 shows that most groups of cohorts show significantly raised SMRs. The exclusion of studies in which no deaths occurred may have increased the SMRs, but any effect is likely to be small, as these studies usually have few PYFU.
The meta-analysis includes 31 cohorts in which there were no deaths by suicide, but it is important to note that in each of these cohorts the expected number of deaths was less than one. Only one study located has an SMR for suicide in epilepsy of significantly less than one (Day et al., 2005); this is a study of cause of death in people with developmental disability in which the SMR for suicide was low in people with epilepsy, but was even lower in those without epilepsy. The reason for this low SMR is unclear, particularly as people with developmental disabilities have sometimes been found to be at higher risk of suicide (Hiroeh et al., 2001). It is unlikely that it is due to increased supervision of those with developmental disability, as the SMR for drowning in this group was high (Day et al., 2005).
A recent study found that the highest risk of suicide occurred in people with epilepsy and comorbid psychiatric disease, even after adjusting for socioeconomic factors (Christensen et al., 2007). It is thus vital that all clinicians who treat people with epilepsy should identify people with psychiatric comorbidity, so that those at risk of suicide are clearly identified and preventative measures can be taken (Bell & Sander, 2007). It is also important that the risk factors, prognosis, and causes of death in epilepsy are understood fully, to reduce any avoidable morbidity and mortality from epilepsy. In the interim, adequate seizure control and taking of precautions should be established for all patients in order to avoid more deaths.
The initial draft was prepared by G.S.B. and J.W.S., and all authors contributed to the final draft and approved it. J.W.S is the guarantor.
This work was supported by the UK National Society for Epilepsy and by 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. The sponsors were not involved in the design of the study or in the decision to publish. J.W.S. had full access to all the data in the study and had final responsibility for the decision to submit for publication.
We are very grateful to Lu Gao, Matthias Koepp, Louis Lemieux, Anna Piasecka, and Yumi Takayanagi for help with translation.
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 report no conflict of interest in relation to this work.