Psychoses and Epilepsy: Are Interictal and Postictal Psychoses Distinct Clinical Entities?


Address correspondence and reprint requests to Dr. N. Adachi at Adachi Mental Clinic, Kitano 7-5-12, Kiyota, Sapporo 004-0867, Japan. E-mail:


Summary:  Purpose: To evaluate further the relevance of designating psychotic episodes as either postictal or interictal, we compared several biologic variables between epilepsy patients with and without psychosis.

Methods: The study subjects comprised 282 patients with psychosis (36 with postictal psychosis, 224 with interictal psychosis, and 22 with both postictal and interictal psychoses, i.e., bimodal psychosis), and 658 epilepsy patients without psychosis. The clinical characteristics of these patients were reviewed retrospectively by experienced neuropsychiatrists. Factors predicting the development of each type of psychosis were determined by serial multivariate logistic regression analyses.

Results: Factors that were comparable between postictal and interictal psychoses were intellectual function, family history of psychosis, epilepsy type, and the presence of complex partial seizures. In contrast, age at the onset of epilepsy and at the onset of psychosis and the presence of generalized tonic–clonic seizures differed for the three types of psychosis. Patients with bimodal psychosis showed characteristics associated with both postictal and interictal psychoses.

Conclusions: This study documented conditions, including both general factors and epilepsy-related factors, common to epilepsy patients with psychosis, regardless of chronologic distinctions. Certain epileptic processes appear to have equal influence on postictal and interictal psychoses. However, some differences between postictal and interictal psychoses suggest that these chronologic descriptors are valid. Our findings confirmed that psychosis associated with epilepsy should not be defined as a single, simple condition but rather as a complex condition with several possible subcategories.

Epilepsy patients with psychosis account for a notable segment of the adult epilepsy population (1–5). Their psychotic symptoms are diverse and explicitly heterogeneous. Despite a long history of clinical research on this matter, many methodologic issues such as the definition of psychosis, the conditions under which studies are conducted, sample size, and the factors examined have prevented full understanding of the nature of these phenomena (6,7).

Pond (10) proposed three types of psychiatric disorders in epilepsy patients: disorders due to brain abnormalities, disorders directly related to the seizures, and disorders without a direct relation to the seizures. Discrimination based on the chronologic relation between the psychotic episodes and seizures has been widely used for the psychoses observed in epilepsy patients since Logsdail and Toone (11) reevaluated postictal psychosis. Postictal psychosis is characterized by antecedent seizures, lucid intervals, short episodes, and mild confusion with subsequent amnesia (11–16). In contrast, interictal psychosis occurs between epileptic episodes, is long lasting, and the psychopathology is usually more distinguishable (i.e., so-called schizophrenia-like psychosis) (17–21). However, there is insufficient biologic evidence to support these chronology-based classifications of psychoses as distinct clinical entities. Some clinical features are common to both postictal and interictal psychoses (11,13). Furthermore, it is sometimes difficult to attribute psychotic symptoms to seizures alone if other conditions (e.g., medication status or electrophysiology) become unstable around the time of the psychotic episode (7,22–27). Over time, in some patients who initially manifested postictal psychosis, chronic interictal psychosis develops (11,16,26–28), and there is no consensus as to how “bimodal” psychosis should be classified.

We have studied clinical characteristics of psychosis in a large population of epilepsy patients with and without psychosis and documented that both general and epilepsy-related variables [i.e., borderline intellectual functioning, family history of psychosis, early onset of seizures, and presence of complex partial seizures (CPSs) or generalized tonic–clonic seizures (GTCs)] are strongly linked to the development of interictal psychosis (29). We also reported no significant difference in the frequency of interictal psychosis and age-related variables (i.e., age at onset of epilepsy, age at onset of psychosis, and the time interval) between each type of focal epilepsy (30). In the present study, we investigated multiple variables to evaluate further the relevance of designating psychotic episodes as either postictal or interictal. This was done with a view toward development of a detailed classification system for the psychoses that occur in patients with epilepsy.


Definition of psychosis and its seizure-related subtypes

Psychosis is defined as the presence of hallucinations, delusions, or a limited number of severe behavioral abnormalities in accordance with the ICD-10 Classification of Mental and Behavioral Disorders (31). A diagnosis of psychosis requires the presence of organic hallucinosis (F06.0), organic catatonic disorder (F06.1), or organic delusional disorder (F06.2). Although the episodic manifestations of these disorders (i.e., acute organic confusional states) can be as short as a few hours or even minutes, the present study focused on patients who had episodes lasting ≥24 h in a state of full consciousness. These were the episodes documented in the outpatient clinics where these patients receive treatment.

Psychoses in our patient population were classified into three types. Psychosis was considered postictal if episodes occurred within 7 days after a decisive seizure or cluster of seizures (11,32). It was considered interictal if episodes occurred when the patient was seizure free or even when the patient was between habitual seizures (19,29). Chronic schizophrenia-like psychotic episodes and brief interictal episodes [in relation to antiepileptic medication (AED) or forced normalizaton (33), for example] were included in this category because no distinct definition of these psychotic episodes is available (19,34,35). Psychosis was considered bimodal if both postictal and interictal episodes were observed at distinct periods in an individual patient (i.e., patients who have recurrent postictal psychoses and in whom eventually a chronic interictal psychosis develops) (11,16,26–28). Designation of bimodal psychosis was limited to novel psychotic episodes that appeared in patients after a previous psychosis had remitted. Psychosis was considered interictal when chronic interictal psychotic episodes became exaggerated after seizures.

Selection of subjects

A total of 282 patients with epilepsy and psychosis was anonymously registered for the study in December 1996. All patients met the criteria for epilepsy as set forth in the International Classification of Epilepsies (36). They had been regular patients at one of five institutions in Tokyo with an adult epilepsy clinic: the National Center Hospital for Mental, Nervous, and Muscular Disorders; Nihon University Hospital; Tokyo Medical University Hospital; Tokyo Medical and Dental University Hospital; and Komagino Hospital. The clinics at these hospitals have maintained a collaborative database especially designed for epilepsy psychosis since August 1996 (29,30). No study patients had an episode of psychosis antedating the development of epilepsy, evidence of senile dementia, a history of substance abuse, or a recent progressive mass lesion.

For controls, 658 epilepsy patients with no history of psychosis were recruited. These were all patients who visited one of the clinics between November 1 and 14, 1996. Because these patients were registered consecutively with no selection criteria, the characteristics of these controls, approximately one fourth of the entire population attending these epilepsy clinics, presumably represent those of the overall clinic population (29,30).

Diagnoses and clinical assessments were made by neuropsychiatrists qualified in both epileptology and psychiatry. The psychiatric diagnosis was based on standard clinical interviews with any differences resolved by consensus; a formal psychiatric rating scale was not used. This study was given approval by the hospitals' ethics committees.

Items of investigation

We investigated the following items, the operative definitions of which have been described elsewhere (29): (a) sex; (b) age at investigation; (c) family history (including at least one first-degree relative) of epilepsy or psychoses in accordance with the Japanese translation of the Family History Research Diagnostic Criteria (37,38) and on the basis of interviews with the patient and at least one family member as well as according to clinical notes; (d) age at the onset of epilepsy; (e) type of epilepsy [partial epilepsy (PE), generalized epilepsy (GE), or unclassifiable epilepsy (UE), in accordance with the International Classification of Epilepsies (36) and determined on the basis of ictal symptoms, EEG, and neuroimaging findings; (f) seizure type, CPS or GTC, whether primary or secondary, in accordance with the International Classification of Seizures (39); (g) lateralization of epileptiform activities (i.e., spikes, sharp waves, or variants) observed on serial interictal EEGs (the definition of lateralization recordings (left; right; bilateral; and no discharges) is provided elsewhere (40)]; (h) intellectual status [mental retardation [full-scale WAIS-R IQ (FIQ) <70; borderline intellectual functioning (FIQ, 70–84); or normal (FIQ ≥85)](41).

In patients with psychosis, additional items were studied: (i) age at onset of psychosis; (j) interval between onset of epilepsy and onset of psychosis. For references, (k) AEDs and (l) excessive EEG changes during psychotic episodes also were evaluated in accordance with the criteria used in previous studies (32,42).


The χ2 (contingency table) test was used to analyze relations in the categoric data. To identify special conditions responsible for significant χ2 values in the contingency table analysis, the expected number (EN) of patients and adjusted standardized residual (ASR) were computed for each cell (43). In accordance with a normal standard deviation, the absolute value of ASR (z score) was used to identify cells that varied markedly from independence (i.e., z > 1.96 for p < 0.05).

One-way analysis of variance (ANOVA) was used for linear data. Because age at the time of evaluation differed significantly between the groups and was correlated with other age-related variables, it was used as a covariate in subsequent analyses. The assumption of homogeneity of regression slopes was tested by fitting a model containing the main effects and each interaction. The estimated marginal means (EMMs) also were calculated with use of age at the time of evaluation (44). The Bonferroni method was used for post hoc multiple comparisons.

Analysis was performed 3 times for evaluating risk factors for each of the three psychosis subtypes in the following combinations: 882 patients representing pure interictal psychosis (n = 224) and no psychosis (n = 658); 694 patients representing pure postictal psychosis (n = 36) and no psychosis (n = 658); and 680 patients representing bimodal psychosis (n = 22) and no psychosis (n = 658). Logistic regression analysis was used to find predictors of each subcategory of psychosis. After the significant likelihood ratio test for the full model, a backward selection method was used to reduce the number of variables. A variable was eliminated if its removal statistic had a probability ≥0.10.

All statistical analyses were performed with the Statistical Package for Social Sciences 9.0 (45).


Distribution of general characteristics

General characteristics of patients are shown in Table 1 according to subcategories. Five variables (epilepsy type, presence of CPSs, lateralization of epileptiform discharges, intellectual function, and family history of psychosis) were significantly correlated with the presence of psychosis. Epilepsy type is related to the presence of CPSs and the lateralization of epileptiform discharges. When only subjects with PE were included in the analysis (n = 686; 230 with psychoses and 456 without), the presence of CPSs was significantly correlated with psychoses (χ2 = 11.5; p = 0.01). Analysis of residuals showed that CPSs were correlated with interictal psychosis (EN, 127.5; ASR, 2.4; p = 0.017) and no psychosis (EN, 330.4; ASR, –3.3; p = 0.001). Laterality of epileptiform discharges showed no significant correlations (χ2 = 4.1; p = 0.662).

Table 1.  Distributions of nominal data
 Epilepsy without
(n = 658)
Pure postictal
(n = 36)
Pure interictal
(n = 224)
(n = 22)
χ2p value
  1. PE, partial epilepsy; EN, expected number; ASR, adjusted standardized residual; CPS, complex partial seizure.

  2. PE was correlated with postictal psychosis (EN = 26.0; ASR = 2.7; p = 0.007), interictal psychosis (EN = 162.2; ASR = 2.5; p = 0.012), bimodal psychosis (EN = 16.4; ASR = 2.3; p = 0.021) and no psychosis (EN = 481.4; ASR = −4.2; p = 0.0001). The presence of CPS was correlated with postictal psychosis (EN = 19.3; ASR = 2.6, p = 0.01), interictal psychosis (EN = 119.9; ASR = 3.2; p = 0.002), bimodal psychosis (EN = 11.8; ASR = 2.7; p = 0.007) and no psychosis (EN = 352.1; ASR = −5.0; p = 0.0001). The lateralization of EEG abnormalities was correlated with interictal psychosis (bilateral; EN = 76.3; ASR = −2.8; p = 0.006) and with no psychosis (left; EN = 158; ASR = −2.1;
    p = 0.036; and bilateral, EN = 218.4, ASR = 3.4; p = 0.001). Normal intellectual function was correlated with no psychosis (EN = 412.3;
    ASR = 3.5; p = 0.001) and interictal psychosis (EN = 140.4; ASR = −2.6; p = 0.01). Borderline function was correlated with postictal psychosis (EN = 4.3; ASR = 2.4; p = 0.017), interictal psychosis (EN = 26.9; ASR = 2.8; p = 0.006), and bimodal psychosis (EN = 2.6; ASR = 2.9;
    p = 0.004) and no psychosis (EN = 79.1; ASR = −4.6; p = 0.0001). Family history of psychosis was correlated with postictal psychosis (EN = 0.7; ASR = 1.7; p = 0.089), interictal psychosis (EN = 4.1; ASR = 4.6; p = 0.0001), and no psychosis (EN = 11.9; ASR = 5.3; p = 0.0001).

Sex (male/female)350/30821/15117/10710/120.980.807
Epilepsy type (partial/generalized)456/19133/2176/4221/122.20.000
Generalized tonic–clonic seizures (present/absent)447/21125/11164/6015/72.20.532
Complex partial seizures (present/absent)317/34127/9141/8318/429.30.000
Laterality of EEG abnormalities (left/right/bilateral)146/158/24012/10/9 65/65/60 6/9/712.60.049
Intelligence (normal/borderline/retarded)436/58/16419/9/8124/39/6110/7/528.70.000
Family history of epilepsy (present/absent) 55/603 1/35 15/209 1/212.30.520
Family history of psychoses (present/absent)  2/656 2/34 12/212 1/2128.00.000

Some patients with psychosis had several precipitating factors. Changes in AEDs appeared to be correlated with psychoses in one patient with postictal psychosis, in nine with interictal psychosis, and in two with bimodal psychosis. Excessive EEG changes from habitual EEG findings (32), such as suspected forced normalization, were observed in one patient with postictal psychosis and in four with interictal psychosis. The real contributions of these variables remain uncertain because of the lack of control data; thus these variables were eliminated from subsequent analyses and interpretation.

Age-related factors

The observed means for age-related variables were as follows. Age at onset of epilepsy was 18.9 years for patients with postictal psychosis, 12.8 years for interictal psychosis, 12.6 years for bimodal psychoses, and 14.6 years for controls. Age at onset of psychosis was 34.8 years for postictal psychosis, 26.8 years for interictal psychosis, and 25.0 years for bimodal psychosis. The interval between onset of epilepsy and that of psychosis was 18.9 years for postictal psychosis, 12.8 years for interictal psychosis, and 12.6 years for bimodal psychoses.

With the age at examination as a covariate, the estimated marginal means and 95% confidence intervals (CIs) are shown in Table 2. There was a significant difference for the age at onset of epilepsy and age at onset of psychosis. Because age at onset of seizure tends to differ by type of epilepsy, the results could depend on the distribution of epilepsy types among subjects. When analysis was restricted to patients with PE, patients with interictal psychosis [n = 176; EMM(SE), 11.7 (0.7) years] showed an earlier onset of epilepsy than did those with postictal psychosis [n = 33; 16.5 (1.6); p = 0.046], and than controls [n = 456;15.9 (0.4); p = 0.000]. Likewise, among PE patients, postictal psychosis [n = 33; 32.5 (1.4) years] showed a later age of onset of psychosis than interictal psychosis [n = 176; 28.0 (0.6); p = 0.012] or the bimodal psychosis [n = 21; 26.1 (1.7); p = 0.015]. Intervals between age at onset of epilepsy and age at onset of psychosis did not differ significantly between subcategories, either in overall psychosis subjects or in those with PE.

Table 2.  Distributions of linear data
 Without psychosis
(n = 658)
Pure postictal
psychosis (n = 36)
Pure interictal
psychosis (n = 224)
Bimodal psychoses
(n = 22)
F valuep value
  • a

     Post hoc Bonferroni test disclosed significant differences between postictal psychosis and no psychosis (p = 0.000), between postictal psychosis and interictal psychosis (p = 0.028), and between interictal psychosis and no psychosis (p = 0.001).

  • b

     With age at examination as covariate (40.43 yr), post hoc Bonferroni test disclosed significant differences between interictal psychosis and no psychosis (p = 0.000).

  • c

     With age at examination as covariate (40.43 yr), post hoc Bonferroni test disclosed significant differences between interictal psychosis and postictal psychosis (p = 0.006) and between postictal psychosis and bimodal psychosis (p = 0.025).

Age at examination; mean (SD) 95% CI35.9 (12.9) 34.9–36.946.3 (13.7) 41.6–50.939.7 (12.7) 38.0–41.438.5 (12.3) 33.1–43.911.1a0.000
Age at onset of epilepsy; estimated marginal mean (SE) 95% CI15.0 (0.4) 14.3–15.715.9 (1.9) 12.9–18.912.0 (0.6) 10.8–13.212.2 (1.9) 8.4–16.06.95b0.000
Age at onset of psychosis; estimated marginal mean (SE) 95% CINot applicable31.7 (1.3) 29.1–34.227.2 (0.5) 26.2–28.326.1 (1.6) 22.8–29.15.41c0.005
Interval between first seizure and first psychosis; estimated marginal mean (SE) 95% CINot applicable14.5 (1.5) 11.5–17.514.2 (0.6) 13.0–15.412.8 (1.9) 9.0–

Risk factors for each type of psychosis

Subsequent logistic regression analysis resulted in a reduced model that included the following variables: family history of psychosis, age at onset of epilepsy, presence of CPSs or GTCs, and intellectual function (Table 3A). When epilepsy type was used as a variable instead of CPSs, similar tendencies were observed (Table 3B). In both combinations of variables (with CPSs and GTCs or with epilepsy type), family history of epilepsy, sex, and lateralization of EEG abnormalities were eliminated because of nonsignificant probabilities. The final models showed that 76.1–96.9% of the patients were correctly allocated.

Table 3A.  Multivariate logistic regression analysis with epilepsy type as a variable
 Postictal psychosisInterictal psychosisBimodal psychosis
  1. p Value and odds ratio (95% confidence interval); BIF, borderline intellectual functioning; MR, mental retardation.

Sex (man/woman)NSNSNS
Family history of psychosis (present/absent)p = 0.0088p = 0.0005p = 0.0134
 58.59 (2.79–1230.97)37.68 (4.86–292.12)89.09 (2.54–3122.52)
Family history of epilepsy (present/absent)NSNSNS
Age at onset of epilepsy (OR/yr)p = 0.0382p = 0.0684NS
 1.03 (1.00–1.06)0.98 (0.97–1.00) 
Epilepsy type (PE/GE)p = 0.0154p = 0.0023p = 0.0377
 6.44 (1.43–29.05)1.85 (1.25–2.74)9.76 (1.14–83.74)
Laterality of EEGNSNSNS
Intellectual functionp = 0.0296p = 0.0046p = 0.0076
 Normal, 1Normal, 1Normal, 1
 BIF, 3.32 (1.37–8.06)BIF, 2.19 (1.37–3.50)BIF, 4.91 (1.78–13.56)
 MR, 1.50 (0.59–3.79)MR, 1.28 (0.87–1.88)MR, 1.36 (0.43–4.26)
Percentage correctly classified95.2%76.1%96.9%
Table 3B.  Multivariate logistic regression analysis with seizure types as variables
 Postictal psychosisInterictal psychosisBimodal psychoses
  1. p value and odds ratio (95% confidence interval); BIF, borderline intellectual functioning; MR, mental retardation; GTC, generalized tonic–clonic seizure; CPS, complex partial seizure; NS, not significant.

Sex (man/woman)NSNSNS
Family history of psychosis (present/absent)p = 0.0097p = 0.0002p = 0.0310
 16.46 (1.97–137.59)18.45 (4.02–84.75)17.52 (1.30–236.35)
Family history of epilepsy (present/absent)NSNSNS
Age at onset of epilepsy (OR/yr)p = 0.0190p = 0.0952NS
 1.03 (1.01–1.06)0.99 (0.97–1.00) 
GTC (present/absent)p = 0.1023p = 0.0062NS
 1.96 (0.87–4.38)1.68 (1.16–2.44) 
CPS (present/absent)3.50 (1.52–8.05)2.19 (1.55–3.09)4.34 (1.42–13.28)
 p = 0.0032p = 0.0000p = 0.0100
Laterality of EEGNSNSNS
Intellectual functionp = 0.0290p = 0.0087p = 0.0118
 Normal, 1Normal, 1Normal, 1
 BIF, 3.21 (1.33–7.74)BIF, 2.08 (1.30–3.32)BIF, 4.72 (1.70–13.14)
 MR, 1.83 (0.74–4.49)MR, 1.26 (0.85–1.85)MR, 1.61 (0.52–4.96)
Percentage correctly classified95.0%76.1%96.8%


In this study, patients diagnosed with postictal, interictal, or bimodal psychosis had features in common but also showed distinct biologic features.

Study limitations

Although the ICD-10 classifications are widely used and well validated, their application to psychotic phenomena observed in patients with epilepsy remains controversial (8,9). Because psychotic symptoms in epilepsy are quite varied, a more detailed classification system would be beneficial (32). Even for interictal psychosis alone, further classifications have been proposed (19,33,34). In this study, to evaluate the significance of seizures antecedent to psychotic states, interictal psychoses included both acute and chronic states. The relation between acute and chronic psychoses may be an important issue for further study.

Limited information on some variables previously proposed as precipitators for psychoses is available when patients were in psychotic states (29). Changes in AEDs and EEG abnormalities sometimes precipitate psychotic phenomena (32,46). However, these factors often fluctuate and interact in the course of epilepsy (23), regardless of accompanying psychotic conditions. Prospective controlled studies are needed to assess the real contributions of these variables. Moreover, neuroimaging findings were only partially applicable to the present collaborative study. Most of our patients have undergone various procedures (i.e., computed tomography, magnetic resonance imaging, single-photon enhanced computed tomography, positron-enhanced tomography) during different periods or conditions. The results were used to exclude patients with progressive mass lesions from study subjects. However, it was difficult to use these results as research variables because no uniform criteria were applied for the different neuroimaging techniques (29). Systematic neuroimaging studies could provide more information on the brain structures associated with psychosis.

Patients with no psychiatric history at the time of our assessment may develop psychosis in the future. Onuma et al. (4) reported that annually in 0.3% of all adults with epilepsy, a novel psychosis develops. Likewise, postictal, or even interictal psychosis, can turn into bimodal psychosis in the future. Few previous studies have had controlled follow-up periods, and we were at least able to correct for age at the time of evaluation in our analysis of age-related variables.

Frequency of each type of psychosis

Among the patients with psychosis, the observed percentage (12.8%) of patients with pure postictal psychosis was a little lower than that previously reported (32,46,47). However, if the numbers of patients with bimodal psychoses and postictal psychoses are combined (n = 58; ∼20%), the frequency is consistent with the frequencies of previous studies (32,46,47). Whereas this study included only patients with psychotic episodes that continued >24 h, detailed evaluations with video-EEG monitoring may show an increased prevalence (16).

Patients who exhibited both interictal and postictal psychoses accounted for 39.2% of the entire population who experienced postictal psychotic episodes and for 8.9% of those with interictal psychotic episodes. Most cross-sectional studies regarded these patients as having a variant of postictal psychoses. However, in some patients with recurrent postictal psychoses, eventually chronic interictal psychosis develop, or vice versa (11,16,26,28). Several studies have demonstrated that in 13.9–33.3% of patients who initially exhibited postictal psychosis, chronic psychosis developed (11,16,28). The retrospective design of this study may account for a comparatively higher frequency of bimodal psychosis. The frequency of bimodal psychosis may depend on the follow-up time.

Similarities between subcategories of psychosis

Between the two major epilepsy types, our patients with PE were at higher risk for psychosis than were those with GE. Whether temporal lobe structures are associated with psychosis is controversial (6,18,30,47). However, according to previous studies in which PE and GE were analyzed, patients with PE are more susceptible to interictal psychosis. Patients with postictal psychosis tend to have PE (11,13–15,48), and there may be a preponderance of this epilepsy type.

Because epilepsy type is tightly linked with seizure type, the question arises whether PE or CPSs are more responsible for the development of psychosis. In our present analysis, the statistical models for PE and those for CPSs showed similar percentages of correct allocation. However, in patients with PE, psychosis was more frequently observed in those with CPSs than in those without it. Some researchers (3,19,49) have reported that the presence of CPSs is significantly associated with interictal psychoses. Schmitz and Wolf (47) suggested that seizures, which cause impaired consciousness, are more important in psychotic phenomena.

Patients with a family history of psychosis showed increased risks for each type of psychosis, 37.7–89.1 times higher compared with those without such history. Several classic studies have shown that some patients with interictal psychosis had a family history of psychosis (50,51). However, genetic aspects have been ignored since Slater et al. (17) reported no correlation. Toone et al. (18) reported a similar frequency of family history of psychosis between psychotic patients with and without epilepsy. Regarding postictal psychosis, Kanner et al. (16) reported that three of 13 patients with postictal psychosis had a family history of psychiatric illness, although there was no significant difference compared with statistics in patients without psychosis. Genetic vulnerabilities are universally observed in patients with psychotic diseases (52,53). Even in patients with organic psychosis, an increased frequency of family history of psychosis has been observed (54,55). Thus genetic factors might play a role in the development of psychosis associated with epilepsy.

Each level of intellectual function showed a similar risk for the development of each psychosis type: borderline intellectual functioning (OR; 2.1–4.9) > mental retardation (1.3–1.8) > normal (1). These findings agree with the general consensus that the prevalence of psychosis in persons with mental retardation (53,56) or with borderline intellectual functioning (57) is higher than that in those with normal intelligence. Several controlled studies showed particular cognitive deficits in patients with chronic interictal psychosis (32,58,59). Other studies, however, have shown no significant IQ difference (21) or a relatively high IQ in patients with interictal psychosis (60,61). Several studies (15,24) showed that patients with postictal psychosis showed low IQ as frequently as did those with interictal psychosis, although one study indicated a higher IQ in those with interictal psychosis (13). Intellectual function appears to be an essential risk variable for the development of psychosis in epilepsy patients.

Differences between subcategories of psychosis

Patients with epilepsy of relatively early onset showed an increased risk for the development of interictal psychosis. Even if the effect of epilepsy type was excluded, PE patients with interictal psychosis also exhibited an earlier onset of epilepsy than did PE patients without psychosis. Recent reports (19,21) have noted this tendency, although several previous studies failed to show it (49,62). Although congenital maldevelopment of the neural systems is proposed as a possible pathogenesis of psychosis (55), this study could not show further evidence on this matter because of lack of systematic neuroimaging examinations. Some patients with neural migration disorder experience early onset of seizures and recurrent schizophrenia-like psychosis in adolescence (63,64). Early brain damage, regardless of etiology, may impede normal neural development (65) and precipitate psychiatric phenomena (29). In contrast, patients with postictal psychoses were not significantly different in age at onset of epilepsy from patients without psychosis, also reported by Devinsky et al. (14). The estimated mean age and 95% confidence interval (CI) of our patients were comparable to those (12–20 years) in previous studies (11–13,15,16). Logistic regression analysis revealed that the tendency toward a later onset of epilepsy in postictal psychosis patients was the opposite of that in interictal psychosis patients.

Patients with postictal psychosis were significantly older at the onset of psychosis than were those with interictal psychosis, consistent with the results of previous controlled studies (15,66). Even in studies without a control group, the mean age (30–35 years) at the first postictal psychotic episode was similar to that in our study (11,12,16). This finding may simply reflect the comparatively later onset of epilepsy in patients with postictal psychosis. However, because age at onset of symptoms is likely to reflect the process of symptom development (36,57,67), these age-related variables may help discriminate between subcategories of psychosis (30). In contrast, the mean interval from the first seizure to the first psychotic episode did not differ significantly between subcategories of psychosis. They were consistent at ∼15 years in interictal (17,19,47,48) and postictal (11,13,15,66) psychoses. During these intervals, the epileptic process and its related complications (i.e., recurrent seizures, epileptiform discharges, and medications) may equally influence the development of psychosis in each subcategory (30).

GTCs were a significant risk factor for interictal psychosis, as some previous controlled studies have noted (19,29,68). In contrast, although 68% of our patients with postictal or bimodal psychoses had GTCs, this frequency was similar to that of control patients. Most controlled studies (13–15) have failed to show a significant association between secondary GTCs and postictal psychosis. In cases of postictal psychosis, CPSs may be a stronger risk factor rather than subsequent GTCs. Furthermore, despite PE preponderance in each subcategory of psychosis, some patients with interictal psychosis had GE. In contrast, few patients with postictal or bimodal psychosis had GE, a finding supported by most controlled studies (11–13,15,16). Postictal psychotic phenomena appear to be strongly associated with focal brain dysfunction.

Classification of patients with postictal and interictal psychotic episodes

Approximately 8% of all patients with psychosis experienced both postictal and interictal psychotic episodes. Although several studies have regarded such bimodal psychosis as a variant of postictal psychosis (11,16), the patients show heterogeneous characteristics of postictal and interictal psychoses. Some risk factors (family history of psychosis, partial epilepsy, CPSs, and lower intelligence) for bimodal psychosis are common to both interictal and postictal psychosis. Ages at the first seizure and at the first psychosis are more similar to those in interictal psychoses than to those in postictal psychoses. That the vast majority of these patients had PE may indicate an association with postictal psychosis. Although some researchers (69) have attributed these phenomena to quasicontinuous subcortical discharges, there is little evidence to support this hypothesis.

Contribution to a classification system for psychosis in epilepsy

Our present results showed that some epilepsy-related variables (epilepsy and seizure types) and general variables (genetic and cognitive predispositions) are common to each subcategory of psychosis. Certain epileptic processes may equally influence the development of postictal or interictal psychotic phenomena, as previously observed (13). The historical expression used for psychosis in epilepsy, “epilepsy psychosis,” may still be useful as a superordinate category when considering these phenomena.

In contrast, differences in the onset of symptoms, whether epileptic or psychotic, and in epilepsy or seizure types suggest the validity of subcategories according to the chronologic relation between psychosis and seizures as well as phenomenologic differences previously reported (11,12,14,15). Umbricht et al. (13) regarded these differences as determinants of whether transient or chronic psychotic symptoms develop rather than as etiologic factors. However, no evidence supports this distinction. Thus these chronologic subcategories may be indispensable in evaluating the direct effect of seizures on psychoses.

Our findings indicated multiple subcategories of psychoses associated with epilepsy. Likewise, previous studies have proposed a multidimensional classification system (8,32,70), although some critical issues remain concerning its validity (i.e., complexity, inclination toward etiological emphasis, and inefficiency for nonspecialists) (9). However, psychosis in epilepsy should not be considered a simple phenomenon, because both epilepsy and psychosis are in themselves complex phenomena. Multimodal classifications will contribute to further elucidation of the mechanisms of psychosis in association with epilepsy.

Acknowledgment: This research was supported in part by the Japan Epilepsy Research Foundation and the Ministry of Health and Welfare of Japan.