- Top of page
Summary: Purpose: Suicide is considered to be one of the most important causes of death contributing to the increased mortality of persons with epilepsy. We investigated the association between the risk of suicide in persons with epilepsy and clinical factors that might increase or have been suggested to increase the risk of suicide.
Methods: A case–control study was nested within a cohort of 6,880 patients registered in the Stockholm County In-Patient Register with a diagnosis of epilepsy. The study population was followed up through the National Cause of Death Register. Twenty-six cases of suicide, 23 cases of suspected but not proven suicide, and 171 controls, living epilepsy patients, were selected from the cohort. Clinical data were collected through medical record review.
Results: There was a ninefold increase in risk of suicide with mental illness and a 10-fold increase in relative risk (RR) with the use of antipsychotic drugs. The estimated RR of suicide was 16.0 [95% confidence interval (CI), 4.4–58.3] for onset of epilepsy at younger than 18 years, compared with onset after 29 years. The risk of suicide seemed to increase with high seizure frequency and antiepileptic drug (AED) polytherapy, although the estimates were imprecise and the associations not statistically significant. Insufficient data on seizure frequency and changes in AED dosage due to incomplete case records were associated with high RRs. We found no association between risk of suicide and any particular AED, with type of epilepsy, or localization or lateralization of epileptogenic focus on EEG [RR = 0.3 (95% CI, 0.1–1.7)].
Conclusions: The profile of the epilepsy patient who commits suicide that emerges from our study is a patient with early onset (particularly onset during adolescence) but not necessarily severe epilepsy, psychiatric illness, and perhaps inadequate neurologic follow-up. Previous reports of an association with temporal lobe epilepsy could not be confirmed.
Suicide is considered to be an important contributor to the increased mortality of persons with epilepsy. Many studies have investigated the percentage of deaths due to suicide in different epilepsy populations. Although the rates vary as much as between zero and 25%(1–11), most investigators find 5–7% of deaths due to suicide, which can be compared with 1.4% suicide deaths in the general population in the United States. The majority of these studies comprise small and selected populations. Although a few large population studies have been conducted, conclusions are difficult to draw from the results. Zielinski (5), who based his study on all persons with epilepsy in Warsaw, found 7.3% deaths in suicide; however, Hauser et al. (7) and Cockerel et al. (8) showed no increase in suicide rate. Very few investigations provide data that allow calculation of the incidence of suicide or of standardized mortality ratios (SMRs). Two studies based on large cohorts from the Chalfont Center for Epilepsy in the United Kingdom (2) and from the Stockholm County In-Patient Register (9) showed SMRs of 5.4 and 3.5, respectively.
Several hypotheses have been put forward to explain an increased incidence of suicide among persons with epilepsy (11). Retrospective studies indicate that the absolute majority (81–100%) of suicides occur in subjects with a psychiatric illness, depression being the most common diagnosis (12). Depression occurs in as many as 62% in certain epilepsy populations (13). The prevalence of psychosis has been reported to be 2–9%(5,14–16) in epilepsy, higher than in the general population. Psychosocial consequences associated with the burden of a chronic illness and the unique stress of epilepsy and seizures (17,18), as well as iatrogenic factors, such as the effects on mood of some antiepileptic drugs (AEDs) (19,20) may contribute to increased suicide rates. Clinical characteristics directly related to the epilepsy, such as type of epilepsy, localization and lateralization of the epileptic focus (13,14,21–23), age at onset, and the duration of epilepsy (24,25) have been associated with psychiatric comorbidity and also with attempted suicide. However, these factors must not necessarily be regarded as risk factors also for completed suicide. Studies are lacking in which persons with epilepsy who have committed suicide are compared with relevant controls to identify risk factors for suicide.
In a study of cause-specific mortality in a large cohort of adults with a diagnosis of epilepsy in Stockholm (9), we found 53 cases of suicide, 3.5 times more than expected. We report a case–control study of suicide, based on an extended follow-up of the Stockholm study population. The objective was to investigate the association between suicide and clinical factors that might increase, or have been suggested to increase, the risk of suicide among people with epilepsy.
- Top of page
Twenty-six suicides and 23 deaths in which the cause of death was possible suicide were included in the study. Psychiatric characteristics for the two groups are presented in Table 2. We found a higher frequency of persons with alcohol or substance abuse and, probably as a consequence of this, more often intoxication as the immediate cause of death in the undetermined suicide group. Previous suicide attempts were more frequent among the suicide cases.
Table 2. Characteristics of cases
|Characteristics||All cases (n = 49) (%)||Suicide (n = 26) (%)||Undetermined suicide (n = 23) (%)|
|Earlier attempt of suicide|| || || |
| Yes||15 (30.6)||12 (46.2)||3 (13.0)|
| No||34 (69.4)||4 (53.8)||20 (87.0)|
|Method of suicide|| || || |
| Intoxication||22 (44.9)||7 (26.9)||15 (65.2)|
| Hanging, cutting weapons, guns, and explosives||5 (10.1)||4 (15.3)||1 (4.3)|
| Drowning||5 (10.2)||2 (7.7)||3 (13.0)|
| Jump from height||7 (14.3)||4 (15.4)||3 (13.0)|
| Other||4 (8.2)||3 (11.5)||1 (4.3)|
| Unknown||6 (12.2)||6 (25.1)||0|
|Intoxication with|| || || |
| Antiepileptic drugs||9 (33.3)||2 (7.6)||7 (30.4)|
| Anxiolytic/sedative drugs||3 (11.1)||0||3 (13.0)|
| Alcohol/narcotics||3 (11.1)||1 (3.8)||2 (8.7)|
| Other||12 (44.4)||4 (15.4)||8 (34.8)|
|Psychiatric illness|| || || |
| At present, within 1 yr from death||4 (8.2)||3 (11.5)||1 (4.3)|
| Earlier only||5 (10.2)||3 (11.5)||2 (8.7)|
| At present and earlier||30 (61.2)||14 (53.8)||16 (69.6)|
| No||7 (14.3)||4 (15.4)||3 (13.0)|
| Unknown||3 (6.1)||2 (7.7)||1 (4.3)|
|Psychiatric diagnosis|| || || |
| Depression||10 (25.0)||7 (26.9)||3 (16.7)|
| Psychosis||5 (12.5)||2 (7.7)||3 (16.7)|
| Alcohol or drug abuse||13 (32.5)||3 (11.5)||10 (55.5)|
| Depression + psychosis||3 (7.5)||3 (11.5)||0|
| Depression + abuse||6 (15.0)||5 (19.2)||1 (5.5)|
| Psychosis + abuse||1 (2.5)||0||1 (5.5)|
| Unknown||2 (5.0)||2 (7.7)||0|
|Psychiatric (in- or outpatient) care during last year|| || || |
| Yes||26 (53.1)||15 (57.7)||11 (47.8)|
| No||10 (20.4)||6 (23.1)||4 (17.4)|
| Unknown||13 (26.5)||5 (19.2)||8 (34.7)|
For the case–control comparison, the primary analyses were confined to the 26 cases with definitive suicide. Subsequently the same analyses were made for the undetermined suicide cases and for the two groups taken together, and the results did not differ in any major respect from those presented.
No increase in RR associated with neurodeficit from birth could be demonstrated (Table 3) nor with the occurrence of any other medical condition included in our analysis (not shown in Table 3).
Table 3. Relative risk (RR) of suicide associated with other conditions
|Condition|| ||Number of||RRa (95% CI)|
|Neurodeficit from birth||No||25||152||1.0 (reference)|
| ||Yes||1||19||0.4 (0.1–3.1)|
|Psychiatric illness||No||11||149||1.0 (reference)|
| ||Yes||15||22||9.3 (3.7–23.4)|
| ||Yes||11||50||1.9 (0.8–4.7)|
We found a ninefold increase in RR of suicide associated with psychiatric illness (Table 3) and a >10-fold increase in RR of suicide with the intake of antipsychotic drugs (Table 4). After adjustment for psychiatric illness and substance abuse, a RR of almost 4 still remained (Table 4). Onset of epilepsy at an early age was the factor most strongly associated with risk of suicide in this sample (Table 5). RR of suicide was 16.0 (95% CI, 4.4–58.3) for onset of epilepsy at younger than 18 years, compared with epilepsy onset after age 29 years (not shown), the risk being highest with epilepsy onset in adolescence (Table 5). In relation to seizure control, it was found that having any seizures compared with being seizure free during the last year gave an RR of 2.2 (0.6–8.0) (not shown). The RR seemed to increase with high seizure frequency (Table 5) and with AED polytherapy compared with monotherapy (Table 5), although the estimates were imprecise. A substantial proportion of cases were categorized as “unknown” as regards seizure frequency and changes of AED dosages, representing patients for whom medical records were not sufficiently detailed to allow an accurate quantification. These categories showed high RRs compared with seizure freedom and unchanged AED dosage (Table 5). Adjustment for the number of health care (in- or outpatient) contacts during the last 2 years for each patient did not alter the RR level (not shown). There was no increase in risk of suicide associated with any of the drugs PHT, CBZ, or VPA as monotherapy, nor with any of the particular AED combinations used (not shown). We found no association between RR of suicide and any particular type of epilepsy or with localization or lateralization of the epileptogenic focus on EEG (Table 6). All analyses were also made with adjustments for mental illness and for alcohol abuse, but this did not alter the results in any major way (not shown).
Table 4. Relative risk (RR) of suicide associated with psychopharmacologic drugs adjusted for psychiatric illness and alcohol abuse
|Psychopharamcologic drugs||Number of||RRa (95% CI)||RRb (95% CI)||RRa,b (95% CI)|
|No||13||141||1.0 (reference)||1.0 (reference)||1.0 (reference)|| || || || || || || || |
|Yes||13||30||4.5 (1.8–10.9)||1.8 (0.6–5.2)||1.8 (0.7–5.2)|| || || || || || || || |
| Antipsychotics||11||13||10.3 (3.6–29.0)||3.5 (1.0–12.7)||3.8 (1.0–13.7)|| || || || || || || || |
| Antidepressants||2||3||6.9 (0.9–55.1)||3.5 (0.4–32.9)||4.2 (0.4–39.6)|| || || || || || || || |
| Anxiolytics||4||12||3.6 (1.0–13.2)||1.2 (0.3–5.7)||1.3 (0.3–6.4)|| || || || || || || || |
Table 5. Relative risk (RR) of suicide associated with various clinical characteristics
|Variable||Number of||RRa (95% CI)|
|Age at onset of epilepsy (yr)|| || || |
| ≥30||5||86||1.0 (reference)|
| 18–29||4||46||2.8 (0.7–11.9)|
| 10–17||9||17||19.6 (4.8–80.5)|
| 0–9||5||18||12.0 (2.6–55.7)|
|Type of epilepsy|| || || |
| Generalized idiopathic||2||12||1.0 (reference)|
| Localization-related symptomatic||10||92||0.4 (0.1–2.4)|
| Localization-related cryptogenic||6||45||0.5 (0.1–3.2)|
| Undetermined||8||22||1.5 (0.3–8.9)|
|Seizure frequency during last year|| || || |
| 0||3||54||1.0 (reference)|
| 1–2||4||33||1.9 (0.4–9.4)|
| 3–12||3||33||1.6 (0.3–8.3)|
| 13–50||3||26||2.2 (0.4–11.9)|
| >50||3||13||4.3 (0.8–24.9)|
| Unknown||10||12||16.4 (3.8–70.9)|
|Number of AEDs|| || || |
| 1||12||115||1.0 (reference)|
| 2||10||49||2.0 (0.8–5.2)|
| 3||2||7||3.1 (0.6–17.5)|
|Changes of AED dosages per year|| || || |
| 0||8||88||1.0 (reference)|
| 1–5||8||75||1.2 (0.4–3.4)|
| Unknown||8||8||13.6 (3.8–49.2)|
Table 6. Relative risk (RR) of suicide associated with localization of epileptic focus on EEG
|Variable||Number of||RRa (95% CI)|
|EEG: epileptic focus|| || || |
| Not focal||13||74||1.0 (reference)|
| Focal, extratemporal||1||7||0.8 (0.1–7.0)|
| Focal, temporal||7||60||0.6 (0.2–1.7)|
| Unknown||5||30||0.8 (0.3–2.6)|
|EEG: lateralization of focus|| || || |
| Right||5||23||1.0 (reference)|
| Left||3||39||0.3 (0.1–1.7)|
|EEG: lateralization of temporal focus|| || || |
| Right||5||22||1.0 (reference)|
| Left||2||36||0.3 (0.1–1.7)|
- Top of page
Incidence rates of suicide in the general population usually also comprise so-called undetermined suicides: cases in which uncertainty remains, even after autopsy, whether death was caused on purpose or by accident. The percentage of undetermined suicides varies internationally, in Sweden being ∼25%, which is relatively high. In the present study, 45% of all cases were undetermined suicides. We do not know whether this discrepancy is an effect of the method of identification and selection of our cases or if epilepsy itself increases the rate of undetermined suicides in some way. The incidence of suicide in the general population is higher for men than for women, although this difference decreases in urban areas like Stockholm. The male/female ratio in Sweden is 2.5:1, in Stockholm 1.7:1 (28), and in our study covering the Stockholm area, this ratio was 1.9:1. In the present study, distribution of age at death in suicide did not differ in any major aspect from that of the general population in Stockholm (28). The methods of suicide used by our cases were more often intoxication and less often violent like hanging and gun shooting compared with suicides in the general population of Sweden. This is probably an effect of the larger proportion of undetermined suicides in our study.
In the majority of suicide cases in the general population, the underlying condition is a psychiatric illness or psychiatric symptoms (12). Depression is most common, followed by psychosis and certain personality disorders. Alcohol or drug abuse, alone or in combination with other diagnoses, occurs in more than half of the cases in certain suicide populations. However, suicide also occurs without preceding psychiatric illness, after reactions to crises, or as a well-considered decision in the presence of a serious somatic illness. Certain factors are of importance to start the suicide process in these conditions, such as feelings of hopelessness, meaninglessness, despair, and catastrophic anxiety, but also certain personality traits, negative life events (especially personal losses), and social disintegration. Several of these factors and conditions are associated with epilepsy. Higher unemployment, lower marriage rates, and the experience of social stigmatization of persons with epilepsy are well known (18,21,29). Alcohol abuse can be an etiology of epilepsy, as well as of depression. In the present study, 41% of male suicide cases were alcohol or drug abusers, compared with 60% of men committing suicide in the general population in Stockholm.
An abundance of investigations of epilepsy and psychiatric comorbidity have, despite methodologic difficulties, established associations with depression and psychosis in epilepsy. Depression has been found in higher frequency in epilepsy than in other neurologic disorders (22). Feelings of intense depression, as an ictal manifestation, can cause risk of suicide (30). Postictal and interictal depressions are more common, 20–60% in some highly selected epilepsy populations (13,23,31). Psychotic symptoms in connection with an epileptic seizure have been shown to occur at certain conditions in ≤6% of patients with intractable seizures (32). Although varying, in most reports, the prevalence of interictal psychosis is considered higher than the prevalence of psychosis in the general population. Kanemoto et al. (33) showed suicide attempts to be more closely associated with postictal psychosis than with interictal psychosis.
In the present study, psychiatric illness was associated with a ninefold increase in RR of suicide, naturally a risk factor for suicide also in the general population. The use of antipsychotic drugs, which can be seen as a marker of psychiatric illness, also increased the RR 10 times. However, after adjustment for psychiatric illness and alcohol abuse, a RR of almost 4 associated with the intake of antipsychotic drugs still remained. We have no clear explanation for this.
Several methodologic problems exist when studying psychiatric comorbidity as well as suicide in epilepsy. Many investigations comprise small and difficult-to-treat, highly selected epilepsy populations, decreasing the possibility of generalizing from the results. Several findings discussed in the literature are in fact the results of reviews and meta-analyses based on small samples from different populations and methods (34,35). Our study population is based on a large epilepsy cohort covering >40,500 person-years at risk. Although we estimate that the majority of the adult epilepsy population in Stockholm is included, a certain selection bias toward more severe epilepsy may occur, because all patients had been hospitalized at least once (9). Furthermore, in the process of selecting cases as well as controls, a number of patients were excluded because of erroneous diagnosis of epilepsy or lack of retrievable medical records. None of these facts, however, invalidates the case–control comparison, although the number of patients in some analyses became rather small, and the results should be interpreted with caution.
Investigations that use adequate control patients to discover the risk factors are lacking. In some studies, persons with epilepsy and psychiatric illness or suicide attempts are compared with persons not having epilepsy, and yet conclusions about epilepsy-related risk factors are drawn, which in fact is not possible (21,25,36). In the present study, cases and controls came from the same epilepsy cohort. Our selection criteria differed in one aspect between cases and controls: all controls had received AED treatment for ≥1 year, whereas there were a few suicide cases that had a shorter treatment period or no AED treatment. Whether this affects the results in any way is open to speculation, but it is at least reasonable to assume that the cases did not have more severe epilepsy than the controls. Owing to the large number of analyses, one might ask whether some of the results might have arisen by chance alone. However, all analyses were based on a hypothesis, most often emerging from previous observations by other researchers, and were therefore deemed relevant to interpret.
Onset of epilepsy at an early age was the only directly epilepsy-related factor strongly associated with an increased relative risk of suicide. The relative risk was 12 with onset of epilepsy at younger than 10 years, but even higher, almost 20, with onset in the age group including puberty compared with onset after age 29 years. Our result is in keeping with Zielinski's study of persons with epilepsy in Warsaw (5), in which most patients who committed suicide had early epilepsy onset. Early onset of epilepsy also has been associated with depression, marital status, and perceived stigma (18). No association between duration of epilepsy and relative risk of suicide was found in the present study, in keeping with a study by Batzel and Dodrill (38). Mendez et al. (16) compared 62 patients with epilepsy and psychosis with 62 mentally well patients with epilepsy and found the mean age at epilepsy onset among psychotic patients to be 13 years, compared with 9 years in the control group. Taylor (37) compared epilepsy surgery candidates with a large series of patients with epilepsy and psychosis published by Slater in 1963, and found an association between psychosis and epilepsy onset around the puberty growth spurt. These findings suggest that onset of epilepsy in a period of biologic and psychosocial maturation is of greater importance than a long-standing illness, although the precise mechanisms leading to the increased suicide risk are not known and might well be multifactorial. The results, however, might have implications for the provision of epilepsy care for adolescents.
Although we observed the point estimates of the RR to be 3.1 for three AEDs compared with monotherapy and 4.3 for >50 seizures per year compared with seizure freedom, we could not ascertain any statistically significant association between the risk of suicide and the severity of epilepsy as expressed by seizure frequency, AED polytherapy, or changes of AED dosages during the last year. Earlier studies are ambiguous concerning the impact of epilepsy severity. Controlled studies of suicide or suicide attempts describing data on epilepsy severity are lacking. However, Schmitz (24) compared 25 patients with epilepsy and psychosis, 25 patients with epilepsy and depression, and 50 mentally well patients with epilepsy and found associations between psychosis (but not depression) and severity of seizures, multiple seizure types, a history of status epilepticus, and AED polytherapy. In a study by Mendez (22), depressed patients with epilepsy had lower frequency of generalized tonic–clonic seizures, but more often AED polytherapy than nondepressed epilepsy patients. We interpret the categories “unknown” seizure frequency and “unknown” number of dose changes rendering high RRs of suicide in the present study as markers of a deficient health care contact; see further comments later.
Certain AEDs have been shown to affect mood and mental status, such as phenobarbital (PB)-induced depression (19), and vigabatrin (VGB) has been shown to be associated with psychosis (20). Schmitz (24) found the use of PHT associated with psychosis and treatment with VPA inversely associated with depression (24). The present study was unable to find any association between the RR of suicide and any particular AED or combination of AEDs, although only a small number of patients in our study were treated with PB, VPA, or any of the newer AEDs such as VGB.
We did not find any association between RR of suicide and any particular type of epilepsy, or with localization or lateralization of epileptogenic focus on EEG. Localization-related epilepsy, especially with foci in temporal regions, has previously been associated with psychiatric morbidity (14,24). In 1969 Flor-Henry (39) described an association between depression and epileptogenic focus in the right temporal lobe. Later studies have not confirmed this, but instead demonstrated depression to be overrepresented in epilepsy with a left temporal focus (21,23) or not associated with lateralization (40). None of these studies used patients with epilepsy as controls. More recent studies explored these associations, which probably are of a more complex nature, with other more sophisticated methods. The risk of depression seems to increase with focus in the temporal lobes in the presence of frontal dysfunction demonstrated by neuropsychological methods (41) and in frontal or temporal hypometabolism demonstrated by positron emission tomography (13,31).
It was possible only in few of our cases to discriminate between postictal and interictal psychiatric disorders by our review of medical records. The main reason for this seemed to be the fact that different specialists treated different symptoms apparently without coordination, and that simultaneous occurrence of epilepsy and psychiatric symptoms were regarded as a coincidence in almost all cases without further discussion. The strong overrepresentation of cases with incomplete records concerning epilepsy severity and AED treatment, rendering 13- to 16-fold increased RRs of suicide, might be another aspect of this problem. Adjustment for number of health care contacts during the last year did not change this association, suggesting that some other variable than not visiting the doctor was the reason for incomplete data. One may question if the psychiatric symptoms prevented the patient—or the doctor—from bringing about an adequate follow-up of the patient's epilepsy.
We present one of very few existing case–control studies of epilepsy and suicide. The objective has been to compare clinical characteristics of persons with epilepsy who commit suicide with those of living persons with epilepsy and thus to identify epilepsy-related risk factors for suicide. In many aspects, the profile of our suicide cases was not different from that of suicides in the general population. The strong association between risk of suicide and mental illness is a common risk factor for suicide in the general population, but a contribution of epilepsy itself to these conditions cannot be excluded. We found no significant associations between the risk of suicide and epilepsy severity expressed as seizure frequency, AED polytherapy, or frequent changes of AED dosages, nor was there any association between the risk of suicide and type of epilepsy, lateralization, or localization of epileptogenic focus. However, we found a strong association between risk of suicide and onset of epilepsy at an early age, particularly with onset during adolescence.