Address correspondence and reprint requests to Dr. Samuel Wiebe, Division of Neurology, Foothills Medical Centre, 1403–29 St. NW, Calgary, Alberta, Canada, T2 N 2T9. E-mail: firstname.lastname@example.org
Purpose: The estimated prevalence of mental health disorders in those with epilepsy in the general population varies owing to differences in study methods and heterogeneity of epilepsy syndromes. We assessed the population-based prevalence of various psychiatric conditions associated with epilepsy using a large Canadian national population health survey.
Methods: The Canadian Community Health Survey (CCHS 1.2) was used to explore numerous aspects of mental health in persons with epilepsy in the community compared with those without epilepsy. The CCHS includes administration of the World Mental Health Composite International Diagnostic Interview to a sample of 36,984 subjects. Age-specific prevalence of mental health conditions in epilepsy was assessed using logistic regression.
Results: The prevalence of epilepsy was 0.6%. Individuals with epilepsy were more likely than individuals without epilepsy to report lifetime anxiety disorders or suicidal thoughts with odds ratio of 2.4 (95% CI = 1.5–3.8) and 2.2 (1.4–3.3), respectively. In the crude analysis, the odds of lifetime major depression or panic disorder/agoraphobia were not greater in those with epilepsy than those without epilepsy, but the association with lifetime major depression became significant after adjustment for covariates.
Conclusions: In the community, epilepsy is associated with an increased prevalence of mental health disorders compared with the general population. Epilepsy is also associated with a higher prevalence of suicidal ideation. Understanding the psychiatric correlates of epilepsy is important to adequately manage this patient population.
Originally coined by Feinstein, the term comorbidity is used to refer to the greater than coincidental association of two conditions in the same individual (Feinstein, 1970). The psychiatric comorbidities in people with epilepsy have important clinical and therapeutic implications. Prevalence studies on the association between epilepsy and psychiatric disorders have found that epilepsy can precede, co-occur with or follow the diagnosis of a psychiatric disorder (Gaitatzis et al., 2004a). The most frequent psychiatric diagnoses reported in people with epilepsy include psychoses, neuroses, mood disorders (DSM-IV axis I disorders), personality disorders (DSM-IV axis II disorders), and behavioral problems (Gaitatzis et al., 2004a). Psychiatric symptoms can be classified according to their temporal relationship with seizure occurrence. They can be divided into peri-ictal symptoms (related to the seizure itself) or interictal symptoms (independent of seizures). Peri-ictal symptoms can precede the seizure (pre-ictal), occur during a seizure itself (ictal), or follow the seizure (postictal) (Swinkels et al., 2005).
There are few population-based studies evaluating the prevalence of psychiatric conditions in people with epilepsy. Most studies involve selected groups of patients from tertiary centers or specialized clinics. Psychiatric psychopathology may be overrepresented in selected populations such as patients with temporal lobe epilepsy or those with refractory seizures. Forsgren (1992) performed a survey in 713 adults with active epilepsy in northern Sweden. In that study, 5.9% of patients reported a psychiatric disorder: 0.8% had schizophrenia, 0.6% had psychoses, and 1.7% had personality disorders. Another group (Jalava and Sillanpaa, 1996) followed a cohort of 94 persons with epilepsy and 199 without epilepsy from the general population. The lifetime prevalence of psychiatric disorders was 0.7% in the general population and 24% in people with epilepsy. Another study using a general practitioner database (Mensah et al., 2006) evaluated 499 patients with epilepsy and found that 11.2% had depression. Gudmundsson (1966) studied all individuals with epilepsy in Iceland (n = 987) and found that 52% had some nonpsychotic mental condition and 7% had lifetime psychosis. Some studies have also been performed in children. Havlova (1990) reported mental abnormalities in 6.7% of 225 patients with childhood-onset epilepsy, Rutter et al. (Rutter et al., 1970) reported a prevalence of 27% of psychiatric conditions in children with epilepsy and 7% in children from the general population and finally Hacket et al. (1998) reported a 23% prevalence of psychiatric disorders compared with 8.1% of children from the general population.
Broad-ranging general population studies using validated psychiatric diagnoses are essential to understand the mental health problems of the entire spectrum of people suffering from epilepsy. Also, the use of standard measures and definitions, as employed in the broader literature of psychiatric epidemiological studies, is important for interpretation of population-based estimates. Using data from the 1.2 cycle of the Canadian Community Health Survey (CCHS), we assessed the prevalence of psychiatric conditions in individuals with epilepsy compared with those without epilepsy. This study is one of the few available studies in the literature exploring psychiatric comorbidities in people with epilepsy from the general population using a structured interview based on the DSM-IV criteria.
Canadian community health survey—mental health and well-being (CCHS-MHWB)
The Canadian community health survey (Cycle 1.2)
The CCHS cycle 1.2 is a cross-sectional survey that collects a wide range of information relating to health status, health care utilization, and determinants of health for the Canadian population. Information was collected between May 2002 and December 2002 from a representative sample of persons 15 years or older, living in private dwellings in the ten provinces and the three territories (Gravel and Beland, 2005). Individuals living on Indian Reserves or Crown lands, clientele of institutions, full-time members of the Canadian Armed Forces and residents of certain remote regions were excluded from the sampling frame. One person aged 15 years or older was randomly selected from sampled households. A significant effort was made to interview respondents in person at their place of residence (86% of cases). Interviews were conducted in English, French, Chinese, or Punjabi (as required or requested by the interviewee). From the initially selected 48,047 households, there was an 86.5% household-level response rate, and among responding household, there was an 89.0% person-level response rate. The overall response rate was thus 77.0%, resulting in a total sample size of 36,984 respondents.
Ascertainment of psychiatric conditions
The CCHS 1.2 mental health interview was based on the WMH-CIDI (World Mental Health (WMH) Survey Initiative Version of the World Health Organization (WHO) Composite International Diagnostic Interview (CIDI) (Kessler and Ustun, 2004). A copy of the Canadian adaptation is available on the Statistics Canada website (Statistics Canada, 2003). Trained lay interviewers using computer-assisted interviewing procedures administered the survey. The following disorders were evaluated: major depressive disorder, mood disorder, anxiety disorder, bipolar disorder, social phobia, agoraphobia, panic disorder, substance dependence and suicidal ideation. Diagnostic algorithms followed DSM-IV criteria, with the exception of the duration requirement for a manic episode. The CCHS asked only whether manic symptoms had lasted “several days or longer” whereas duration of seven days is required by the DSM-IV when there has not been a need for hospitalization. In this analysis we differentiated between major depressive disorder and bipolar disorder by identifying subjects with one or more lifetime manic episodes according to the WMH-CIDI. The Canadian adaptation of the WMH-CIDI evaluated subjects for illicit drug dependence, and alcohol dependence was assessed using the CIDI Short Form (Kessler et al., 1998). In this analysis, substance dependence was defined as having either WMH-CIDI drug dependence, CIDI Short Form alcohol dependence, or both. Two disorders, dysthymia and schizophrenia were not assessed using a CIDI module, but instead were assessed by eliciting self reported professional diagnoses using items similar to those used to assess epilepsy, see below.
Ascertainment of epilepsy
The presence of epilepsy was probed by an interviewer asking directly “Do you have epilepsy diagnosed by a health professional?” Subjects responding “yes” were considered as having epilepsy. This method of ascertainment creates opportunity for under and overreporting, but it is the only practical approach to explore population-based, nation-wide morbidities. We have previously discussed issues of validity of this method, and concluded that analyses of prevalence of epilepsy using these health surveys yield results that are compellingly similar to those of classic epidemiological studies (Hauser and Annergers, 1990) and which are also consistent across several health surveys explored in the Canadian population (Wiebe et al., 1999; Tellez-Zenteno et al., 2004, 2005).
The CCHS 1.2 used a multistage, stratified cluster design to select eligible households. To correct for the potential bias resulting from this complex survey design, Statistics Canada recommends a bootstrap procedure using a set of replicate weights that they supply. All results presented here were produced with this approach and are therefore representative of the targeted population. The standard error associated with specific estimates, p-values and confidence intervals (95% CI) were adjusted for survey design effects by the bootstrap procedure. All analyses were conducted at the Prairie Regional Data Centre of the University of Calgary campus, using SAS software. The same methodology and analysis has been validated in previous analysis using this survey (Patten et al., 2006a, 2006b).
Initially, the prevalence of the various mental health disorders was estimated, and 95% confidence intervals for these estimates were calculated. Stratified analysis and logistic regression analysis were carried out to explore the impact of demographic variables on prevalence. In the logistic regression models, age was added as a continuous variable, so that the models described the log odds of having a disorder as a function of epilepsy status, age (in years), and sex. Interaction terms were also explored, and their statistical significance was evaluated using Wald tests adjusted for design effects as described above. For ease of presentation, the fitted log odds from the logistic regression models were exponentiated to produce fitted odds and then converted to proportions by dividing the fitted odds by the fitted odds plus one.
Prevalence of epilepsy
Of the 36,984 CCHS cycle 1.2 participants, 253 reported having epilepsy (weighted prevalence of 6 per 1,000 people (95% CI = 5–7).
Prevalence of psychiatric conditions in those without epilepsy
The prevalence of mental health disorders in people without epilepsy was (95% CI): major depressive disorder (12 month) 3.9% (3.6–4.2), major depressive disorder (lifetime) 10.7% (10.2–11.2), dysthymia (12 months) 0.3% (0.3–0.4), bipolar disorder (12 month) 1.0% (0.8–1.1), panic disorder (12 month) 1.5% (1.3–1.7), agoraphobia (12 month) 0.5% (0.4–0.7), social phobia (12 month) 3.0% (2.7–3.2), schizophrenia (12 months) 0.2% (0.2–0.3), drug dependence (12 month) 0.8% (0.6–0.9) and alcohol dependence (12 month) 2.6% (2.4–2.8).
The prevalence of any evaluated mood disorder (12 month) was 5.2% (4.9–5.5), any evaluated anxiety disorder (12 months) 4.6% (4.3–4.9), anxiety disorder (lifetime) 11.2 (10.8–11.7), mood or anxiety disorder (12 month) 8.0% (7.6–8.5), mood disorder lifetime (includes dysthymia 12 month) 13.2% (12.7–13.7), mood disorder/anxiety disorder/dysthymia (lifetime) 19.6% (19.0–20.2), panic disorder/agoraphobia (12 month) 2.0% (1.8–2.2), panic disorder/agoraphobia (lifetime) 3.6 (3.3–3.9) and substance dependence (12 month) 3.1% (2.8–3.3). The prevalence of any mental health disorder (12 month) was 10.9% (10.4–11.3) and for lifetime was 20.7 (19.5–20.7).
Prevalence of psychiatric conditions in people with epilepsy
The prevalence of mental health disorders in people with epilepsy was as follows (95% CI): major depressive disorder (lifetime) 17.4% (10.0–24.9), mood disorders (12 month) 14.1% (7.0–21.1), mood disorder (lifetime) 24.4 (16.0–32.8), anxiety disorders (12 month) 12.8% (6.0–19.7), anxiety disorder (lifetime) 22.8 (14.8–30.9), mood/anxiety disorder (12 month) 19.9% (12.3–27.4), mood disorder including dysthymia (lifetime) 34.2% (25.0–43.3), panic disorder/agoraphobia (12 month) 5.6% (1.9–9.2) and panic disorder/agoraphobia (lifetime) 6.6 (2.9–10.3). The prevalence of any mental health disorder (12 month) was 23.5 (15.8–31.2) and for lifetime was 35.5 (25.9–44.0). This list of estimates of disorder prevalence for respondents with epilepsy is not identical to that presented above for people without epilepsy because in some cases the epilepsy estimates were too imprecise to be presented: major depressive disorder (12 month), dysthymia (12 month), bipolar disorder (12 month), panic disorder (12 month), agoraphobia (12 month), social phobia (12 month), schizophrenia (12 month), drug and alcohol dependence (12 month), and substance dependence (12 month).
The lifetime prevalence of suicidal ideation in people with epilepsy was 25% (17.4–32.5) compared with 13.3% (12.8–13.8) in those without epilepsy.
Odd ratios for lifetime mood, anxiety disorders, and suicidal thoughts
Individuals with epilepsy were more likely than individuals without epilepsy to report lifetime anxiety disorders or suicidal thoughts with odds ratio of 2.4 (95% CI = 1.5–3.8) and 2.2 (1.4–3.3), respectively. The odds of lifetime major depression or panic disorder/agoraphobia were elevated in those with epilepsy compared with those without epilepsy to a similar extent as for any anxiety disorder or suicidal thoughts, although the confidence intervals for the odds ratios in each case reached the null value of 1.0. The odds ratio for lifetime major depression was 1.8 (1.0–3.1) and for panic disorder/agoraphobia was 1.9 (1.0–3.7), respectively.
In multivariate analyses, several interactions were identified, providing a more detailed description of the epidemiological pattern. Major depressive disorder (12-month) had a statistically significant age by sex interaction (Wald χ2 = 9.5, p = 0.002), such that the sex difference (women > men) was seen to decline with age. However, the difference in the effect of sex over age did not differ in people with or without epilepsy (Wald χ2 = 1.36, p = 0.24), nor was evidence of an age by epilepsy (Wald χ2 = 0.96, p = 0.33) and epilepsy by sex interaction (Wald χ2 = 1.21, p = 0.27). For this reason, it was possible to identify a single age- and sex-adjusted odds ratio describing the association of epilepsy with major depressive disorder (12 month): 2.3 (95% CI = 0.99–5.23), which was not statistically significant (Wald χ2 = 3.7, p = 0.05).
Because 12-month prevalence is predictably lower than lifetime prevalence, a comparison of major depressive disorder lifetime prevalence is expected to be less vulnerable to Type II error than the 12-month analysis, and to achieve greater precision of estimation. The logistic regression model for lifetime major depression prevalence followed the same pattern as that of 12-month prevalence (an age by sex interaction, but no interactions involving epilepsy). Here, the age- and sex-adjusted odds ratio was 1.8 (95% CI = 1.1–3.2), a statistically significant elevation (Wald χ2 = 4.9, p = 0.03).
A model for any 12-month mood disorder produced a similar pattern. However, with the inclusion of other mood disorders (bipolar disorder and dysthymia), the OR for epilepsy became higher (OR = 3.2, 95% CI = 1.7–5.9).
Panic disorder with agoraphobia (12 month) presented a more complex epidemiological pattern. There was no age by sex interaction (Wald χ2 = 2.5, p = 0.11) but a statistically significant epilepsy by disorder interaction was identified (Wald χ2 = 5.8, p = 0.02). Because of this interaction, a statistical test for the overall effect of epilepsy is not presented. However, the epilepsy parameter from the logistic regression model (β= 0.043) indicated no association at the baseline age (15 years), OR = 0.6 (95% CI = 0.1–7.4), and an increasing effect with age. In contrast, age itself was associated with an odds ratio below 1, indicating a tendency for panic disorder with agoraphobia to decline in prevalence with age in people without epilepsy.
The rest of the explored variables are shown in Table 1. Figs. 1A and B provide a graphic representation of the logistic regression models predicting the prevalence of mental health disorders (on the y-axis) based on age (on the x-axis.) and gender. The major depression model displays a consistent relative effect of epilepsy across the age range in men and women. In the panic disorder model, the age by epilepsy interaction indicates that the prevalence of panic disorder in people with epilepsy increases with age, in distinction to the pattern seen in people without epilepsy where a decline with age is seen.
Table 1. Psychiatric comorbidity in people with epilepsy and the general population
We explored the prevalence of different psychiatric conditions in people with epilepsy using a large, validated, general population-based health survey in Canadians. Few studies have explored the psychiatric comorbidities of epilepsy using population-based data sources (Table 2) (Gaitatzis et al., 2004a). A recent nonsystematic review, found a prevalence of psychiatric disorders in people with epilepsy of 6% using population-based data and 20% using more selective populations such as temporal lobe epilepsy patients (Gaitatzis et al., 2004a). In that review, the prevalence of various psychiatric disorders in persons with epilepsy was: mood disorders 24-74%, depression 30%, anxiety disorders 10–25%, psychoses 2–7% and personality disorders 1–2% (Gaitatzis et al., 2004a).
Table 2. Nonselected populations
Ascertainment method of psychiatric conditions
Type of population
Use of controls
Prevalence of psychiatric conditions (%)
Prevalence of depression (%)
Prevalence of anxiety disorder (%)
Prevalence of schizophrenia (%)
Prevalence of psychosis (%)
Prevalence of personality disorders (%)
Prevalence of alcohol dependence (%)
aPoint prevalence, bprevalence during a follow-up of 35 years, clifetime prevalence, d12 month prevalence. eCombined anxiety and depression symptoms.
PE, patients with epilepsy; PWE, patients without epilepsy; CES-D, Center for Epidemiology Studies-Depression Scale; CIDI, Composite International Diagnostic Interview; CIS, Clinical Interview Schedule; HADS, Hospital Anxiety and Depression Scale; ICD, International Classification of Diseases; GP, general practionners; PBS, population-based study; NCOE, noncontrolled, only epileptics; NS, not stated; NE, not examined.
Several studies have explored psychiatric comorbidities in nonselected populations of people with epilepsy (Table 2). Studies that used survey methods similar to ours include those by Forsgren (1992), Strine et al. (2005), Hacket et al. (1998), Ettinger et al. (2004), Gudmundsson (1966), Davies et al. (2003) and Pond and Bidwell (1960). There is extensive variation in the ascertainment methods of psychiatric comorbidities, the prevalence rates obtained and the population sources. The only studies using standardized interviews such as ours are those by Davies et al. (2003) and Pond and Bidwell (1960). Their findings are similar to ours, with general prevalence rates of mental health conditions of 37% (Davis et al.), 29% (Pond and Bidwell), and 23.5% (our study). Furthermore, all of these studies found a higher prevalence of mental health disorders in people with epilepsy than in the general population. The higher rate found in Davies et al.'s study (2003) could be related to the small sample of subjects with epilepsy in that study (n = 67). Psychiatric epidemiological studies may include coverage of different sets of mental health disorders, which can also affect “any disorder” prevalence rates. Some disorders, such as specific phobia, are common and were not included in the CCHS 1.2.
Examples of studies with methodology different from ours include the studies of Strine et al. (2005) and Kobau et al. (2006), which only evaluated self-reported depressive and anxiety symptoms, and found higher rates than with other methods of ascertainment. This is not surprising, since not all psychiatric symptoms necessarily represent disorders. The concept of a disorder requires symptoms to be present, but also imposes additional requirements such as persistence, associated distress and associated dysfunction. Gudmundsson (1966) used nonstandardized interviews, and found a high prevalence of psychiatric diagnoses of 54%. It is conceivable that these studies may overestimate the prevalence of psychiatric conditions, in part due to the lack of standardized interviews. Other common methods of ascertainment include the use of the World Health Organization International Classification of Diseases (ICD) codes from administrative data (Jalava and Sillanpaa, 1996; Hackett et al., 1998; Stefansson et al., 1998; Gaitatzis et al., 2004b). Significant variations in the prevalence of psychiatric conditions exist among these studies. Yet, their reported prevalence rates are similar to ours. Along the same line, there is always the possibility that some of the epilepsy subjects in our study may have non epileptic events. However, our prevalence rates of epilepsy are comparable to that reported in other studies, making it less likely that it would represent a significant limitation.
Depression is the most common comorbid psychiatric disorder in patients with epilepsy (Kanner, 2005). We found a lower prevalence of depression than studies using scales of symptoms or self reported prevalence of depression (Strine et al., 2005; Kobau et al., 2006). Ours is one of the few population-based studies using a well-validated instrument for psychiatric diagnoses (Kessler et al., 1998; Kessler and Ustun, 2004). We obtained prevalence rates similar to those studies using questionnaires to screen for depression (Edeh and Toone, 1987; Mensah et al., 2006) and also to studies using ICD coding and administrative data (Gaitatzis et al., 2004b). Ettinger et al. (2004) used survey methodology and a validated questionnaire to screen for depression and found a higher rate of depression compared with ours (Table 2). Our lower prevalence rates could be explained by the use of a validated instrument to diagnose rather than screen for psychiatric conditions. In summary, the prevalence of depression in nonselected populations such as in our study is lower compared with the prevalence found in selected populations such as patients with temporal lobe epilepsy, or in those with refractory epilepsy where lifetime prevalence ranges between 8% and 48% (Koch-Weser et al., 1988; Victoroff et al., 1994; Umbricht et al., 1995; Altshuler et al., 1999), with a mean estimate of 30% across studies (Hermann et al., 2000). On the other hand, our study and all the studies performed in nonselected populations showed consistently higher rates of depression in people with epilepsy compared with the general population (Table 1).
The rate of suicidal ideation in our study was greater than that in the general population, in keeping with the existing literature (Jones et al., 2003; Pompili et al., 2005). This likely explains in part why mortality overall is greater in those with epilepsy compared with the general population. This emphasizes the importance of preventive strategies (Blumer et al., 2002).
The logistic regression models of prevalence by age and gender provide important new information. The lifetime prevalence of major depressive disorders declined with age in women with and without epilepsy, and remains stable in men, but is overall higher in people with epilepsy (Fig. 1A). Panic disorder and agoraphobia increased with age only in people with epilepsy (Fig. 1B). These data demonstrate that different psychiatric problems vary in prominence with age in a pattern that is different from the general population. This can help raise clinicians' alertness for psychiatric conditions accordingly.
The main strength of our study is the ascertainment method. The CIDI has been extensively validated to ascertain psychiatric conditions in large populations (Kessler et al., 1998; Kessler and Ustun, 2004) and allow us to provide realistic estimates of the prevalence of various psychiatric conditions associated with epilepsy. Unfortunately, the prevalence rates of some psychiatric comorbidities such as bipolar disorder, schizophrenia, personality disorders, psychosis, and substance dependence could not be calculated because of the small sample size. The current study emphasizes the important contribution of medical morbidity to the burden of mental health disorders in modern society. These findings are important in planning health services and provision of adequate medical therapy in individuals with epilepsy and comorbid mental health conditions.
This study was supported by an operating grant from the MSI Foundation received by Dr. Samuel Wiebe.