The prevalence of psychosis in epilepsy; a systematic review and meta-analysis

We conducted a systematic review of all published literature in English on the prevalence of psychosis in patients with epilepsy. We were guided by PRISMA criteria[22]. A search was undertaken of the electronic databases PUBMED, OVID MEDLINE and EMBASE from their inception to September 2010 with the following terms: prevalence, incidence, rate, rates, psychosis, schizophrenia, schizophreniform illness, epilepsy, seizures and temporal lobe epilepsy. In addition to this search procedure, we used the reference lists of the identified publications to find further relevant articles.

Pooled estimates of the prevalence of psychosis in epilepsy patients were calculated using random-effects meta-analysis. This allows a more robust and true estimate of effect size and one that is weighted by the sample size of individual studies. A random effects model weights studies more equally and is considered more appropriate for meta-analyses with substantial heterogeneity [24, 25] and this meta-analytic approach has been used with data containing similar levels to those seen here [25]. The between-study variance or heterogeneity in estimates was modelled using Cochran Q and the I² statistic. The Q statistic is reported with χ² and p-values and the I² statistic is reported as a percentage with increasing values indicating greater heterogeneity between estimates of individual studies (I² <25% indicates low heterogeneity; 30-70% = moderate heterogeneity and >75% indicates high heterogeneity [25]. Meta-regression was used to estimate the extent to which measured covariates (study design, study setting, assessment instrument used, types of epilepsy, year of study, family history of psychiatric illness, International League Against Epilepsy (ILAE) classification used, exclusion criteria applied, country in which study took place) could explain the observed heterogeneity in prevalence estimates across studies. The regression coefficients (β) reported indicate the average difference in prevalence proportion for one category compared to the other (eg assessment by clinical interview versus case notes). Effects of individual covariates were examined first in univariate models and then in a multivariate model constructed in a step-wise fashion. All analyses were carried out using STATA statistical software package, version 11.0.

Our preliminary search identified a total of 431 papers on EMBASE, 773 papers on PUBMED and 999 papers on MEDLINE via OVID. There was a substantial degree of overlap between the 3 databases.

After these titles were screened, 215 papers were examined in detail. After applying inclusion and exclusion criteria, 58 papers (27%) were deemed to have data relevant to the systematic review and meta-analysis. See Figure 1.

Thirteen of the studies were from the UK, 11 from the US, 11 from Japan, 5 from Brazil, 2 from Denmark, 2 from the Netherlands, 2 from Canada, 2 from India,2 from Iceland, 2 from Australia and 6 from other countries.

Ten papers specified the gender of participants in the study. The number of patients with psychosis in the studies varied from 1 [16, 26‐28] to 795 [29].

The number of participants in studies varied from 50 [30] to 34,494 [29]. With regard to the prevalence rates, forty-eight studies were cross sectional studies, 9 were cohort and 1 was case control. The time periods over which the prevalence rates were calculated were as follows: 27 studies were for an unknown time period, 7 studies were lifetime prevalence studies, 21 studies were for more than 1 year and 3 studies were less than 1 year. Only 6 of the studies had control groups [31‐36].

Seven studies were in a community setting, 3 were in a mixed hospital/community setting, 1 was unknown, 3 were from population databases and the remaining 44 were from hospital settings.

In 39 of the studies, patients were interviewed, in 19 studies the patients were not interviewed or it was not stated in the study.

The prevalence rates of psychosis varied from 0.02% [37] to 27% [17]. Forty of the studies used ICD criteria or DSM criteria for diagnosis. Of these studies, 17 used DSM IV criteria, 5 used DSM IIIR criteria, 1 used DSM III criteria, 1 used DSM II criteria, 6 used ICD 10 criteria, 4 used ICD 9 criteria and 1 used ICD 8 criteria. 5 studies used a mixture of these criteria.

Seven studies used a Structured Clinical Interview for DSM IV on patients in their studies [26, 31, 38‐42]. Nine studies were conducted before 1990.

Seventeen of the 58 studies gave the prevalence of schizophrenia. Ten studies had prevalence rates for interictal psychosis. Thirteen studies had prevalence rates for postictal psychosis. Three studies used a sample group of patients with children/adolescents only [14, 27, 43]. Four studies studied patients with Learning Disability only [32, 44‐46]. Seventeen studies included patients with TLE only. Seven of the studies mentioned a family psychiatric history, only 3 mentioned a family history of psychosis. 17 studies had TLE patients only, in 26 studies, the authors did not differentiate whether patients had TLE or not and in 15 studies, there was a mixture of patients with TLE and other forms of epilepsy but unfortunately it was not possible to extract the non TLE or TLE sample from each other. Therefore for the TLE prevalence sample, we only used studies where all the patients had TLE. We did not contact authors personally to access data as this was not possible in many of the studies as the contact details were not current and 22 of the studies were conducted before 2000.

Overall, the pooled prevalence rate for psychosis in epilepsy patients was 5.6% (95% CI =4.8-6.4). (see Figure 2). The pooled prevalence rate when postictal psychosis was excluded was 5.4% (95% CI = 4.5-6.2).

The pooled prevalence rate for psychosis in patients with temporal lobe epilepsy was 7.0% (CI 4.9-9.1). (See Figure 3). The pooled prevalence rate of interictal psychosis was 5.2% (95% CI 3.2-7.2) and the pooled prevalence rate for post ictal psychosis was 2.0% (95% CI 1.2-2.8).

Four papers studied an intellectual disability sample only [32, 44‐46]. The prevalence of psychosis in patients with epilepsy and comorbid learning disability was 7.4% (95% CI 2.6-12.2).

Three studies used child and adolescent subjects only [14, 27, 43]. The prevalence of psychosis in children and adolescents with epilepsy was 5.4%, (95% CI 0.6-10.2).

Three papers mentioned a family history of psychosis. The prevalence of psychosis in patients with epilepsy where a family history of psychosis was present was 5.4% (95% CI = 1-9.8).

Six of the studies had control groups. However, only 4 studies [31, 34‐36] were used for analysis because two of them [32, 33] had no psychosis outcome in the control group so an odds ratio could not be calculated. The pooled odds ratio for risk of psychosis among people with epilepsy was 7.83. (See Figure 4).

There was substantial heterogeneity between the prevalence estimates of psychosis in epilepsy patients across individual studies (I² > 70%). We examined the effect of nine factors on the prevalence of psychotic symptoms in epilepsy patients found across studies in a meta-regression analysis: (1) study design (2) study setting (3) instrument used to assess psychotic symptoms (4) types of epilepsy (5) year of study (6) family history of psychosis noted (7) ILAE classification used (8) exclusion criteria applied (9) country in which the study took place.

None of the following factors significantly explained any of the variance in the estimates on an individual basis. (1) Study Design: (β = -1.1,SE(β = 1.9, p =0.58). (2) Study setting: (whether recruiting patients from the community or from hospitals or from tertiary centres) = (β -0.11, SE = (β) = 0.9, p =0.90). (3) Instrument used to assess psychosis: (employing a clinical interview or using case notes) (β = 0.31, SE(β) = 1.7, p =0.85). 4) Types of epilepsy reported: (all types of epilepsy patients or temporal lobe epilepsy patients only or other types of epilepsy) = (β = 0.74, SE(β) = 0.8, p =0.39). 5) Year of study: (whether studies carried out before 1990 or after 1990) (β = -2.8, SE (β) = 2.1, p =0.17). 6) Family history of psychosis: family of history of psychosis assessed or not assessed = (β = 3.7, SE(β) = 2.1, p =0.09).

7) ILAE classification used: studies reporting use of ILAE classification or not reporting this classification = (β = 0.3 , SE (β) = 1.7, p =0.85). 8) Exclusion criteria applied: studies reporting the use of exclusion criteria or those not reporting exclusion criteria (β = 1.1, SE(β) = 1.4, p =0.45). 9) Country in which study took place: studies from countries classified as ‘developed’ or those classified as ‘not developed’ (β = 2.0, SE(β) = 1.9, p =0.31).

A multivariate model constructed in step-wise fashion showed that the specific combination of 5 factors explained the largest amount of heterogeneity in the prevalence estimates: Study setting; Assessment instrument used; Year of study, Family history of psychosis; Type of epilepsy. These 5 variables together explained 10% of the variance in estimates of the prevalence of psychotic symptoms in epilepsy patients across studies.

Most of the studies included in this review were cross sectional in design (77.6%) and based on samples from tertiary referral centres (71%) and these factors may limit the interpretation of the results with respect to the general population. Only 10% of the studies were population-based. Prevalence measures derived from such unrepresentative samples may therefore overestimate psychiatric morbidity among epilepsy patients [54].

We were unable to draw conclusions about the influence of a genetic vulnerability to psychosis from this study as data was too limited. Family history of psychosis was only mentioned in 3 studies [29, 34, 39]. Further family studies are needed to elucidate this association. Heterogeneity may also play a role.

It was not possible to identify a group of patients which did not have temporal lobe epilepsy to compare with the TLE group as some studies did not state whether the patients had TLE or not and also some studies has TLE and non TLE patients mixed together in their study samples, thus it was impossible to differentiate these 2 disparate groups out.

The question of whether epilepsy is a risk factor for psychosis and/or whether psychosis is a risk factor for epilepsy has repeatedly arisen in the literature, however, very few studies had the data necessary to adequately provide answers [55]. We have not addressed the temporality of the association in this study because the majority of studies included do not have temporal information. However, Clarke et al. have recently found in a population based family study that patients with epilepsy have a 5.5 fold increase in the risk of having a broadly defined psychotic disorder and an 8.5 fold increase in the risk of having schizophrenia [56]. Individuals with a parental history of epilepsy had a 2 fold increase in the risk of developing psychosis, compared to individuals without a parental history of epilepsy. Individuals with a parental history of psychosis had reciprocally a 2.7 fold increase in the risk of having a diagnosis of generalised epilepsy compared to individuals without a parental history of psychosis. Post hoc analyses showed that these analyses were not driven by the comorbidity of epilepsy and psychosis in the parents.

A consensus on the classification of psychotic symptoms associated with epilepsy is lacking [57]. Neither ICD 10 nor DSM V classify seizure-related psychosis separately. It could be debated whether postictal psychosis for example should be classified as a brief psychotic disorder/psychotic disorder due to a general medical condition or a psychotic disorder not otherwise specified in DSM. Furthermore, the psychopathology of patients with epilepsy can be atypical and does not readily conform to these diagnostic manuals [58]. A dedicated sub-commission of the International League Against Epilepsy commission on neuropsychiatric aspects has developed a proposal on the classification of neuropsychiatric disorders in epilepsy [59]. Recent genetic studies reinforce the view that more attention should be given to the relationship between the functional psychoses and neurodevelopmental disorders such as autism [60].

It has been suggested that the neurotoxic effect of epilepsy explains the association between epilepsy and psychosis [55]. Various mechanisms through which this effect might be brought about have been proposed. These include firstly, a ‘kindling’ process whereby acute seizure discharges may cause changes in brain function perhaps through receptor-based changes and changes in cerebral blood flow [61]; secondly, a ‘forced normalisation’ process whereby an inverse relationship exists between seizure control and psychotic symptoms [57, 62] and thirdly, on-going subictal activity in the limbic system that is undetectable on EEG but which leads to brain changes that result in psychosis [55, 63].

Antiepileptic medication may also play a role in the development of psychosis especially among patients with other risk factors such as family history or past psychiatric history. Psychoses have been noted as a potential adverse effect in many different antiepileptic medications suggesting that the phenomenon is not medication specific. Antiepileptic medications implicated include ethosuxamide, topiramate, vigabatrin, zonisamide and leviteracetam [64]. One study found a prevalence of psychosis of 3.7% in patients after they were commenced on topiramate [65]. High starting doses of medication and a rapid titration schedule in more vulnerable patients with past psychiatric history and with more severe epilepsy with high seizure frequency were associated with greater risk.

However, it is also possible that epilepsy and psychotic illness may represent different outcomes of a common aetiological process. Neuropathological, neuroimaging and genetics findings show that similar structural brain abnormalities and genetic abnormalities are present in patients with schizophrenia and patients with epilepsy [55, 66‐71]. For instance, enlarged ventricles have been found to be common to first episode psychosis and temporal lobe epilepsy patients without psychosis [67]. Neuronal migration defects have been proposed as a mechanism related to enlarged ventricles and this defect could be common both to schizophrenia and epilepsy. From a neurobiological perspective, significant grey and white matter deficits occur in temporal lobe epilepsy with psychosis. Some of these deficits overlap with those found in schizophrenia. These include the medial temporal structures but also extend to lateral temporal and extratemporal regions [70].

Recent genetic work shows that a rare genetic mutation can lead to either epilepsy or schizophrenia. A micro-deletion in the genomic area 15q 13–14 containing the nicotine receptor was linked to development of either schizophrenia or juvenile epilepsy [68, 69, 71]. The gene leucine-rich glioma-inactivated 1 gene (LGI1) in autosomal dominant partial epilepsy with auditory features, [72, 73] may also play a role in regulating glutaminergic synaptic transmission, a process that is involved in the pathophysiology of schizophrenia [74]. Genes encoding ion channels may also be a source of interest. Ion channelopathies are known to underlie some epilepsies and it has been shown that variation within the gene CACNA1C (encoding a subunit of the L-type voltage dependent calcium channel) is associated with schizophrenia as well as depression and bipolar affective disorder [75]. This study provides support for the possibility that some people might experience both psychosis and epilepsy at least in part because of an underlying vulnerability to both.