Reexamination of Interictal Psychoses Based on DSM IV Psychosis Classification and International Epilepsy Classification


Address correspondence and reprint requests to Dr. K. Kanemoto at Utano National Hospital, Ukyo-ku, Narutaki, Ondoyama-cho 8, Kyoto, Japan. E-mail:


Summary: We sought to examine interictal psychoses based on the international epilepsy classification and DSM IV criteria, with special attention paid to epilepsy types as well as to subcategories of psychoses. One hundred thirty-two outpatients were studied, each with definite evidence of both epilepsy and interictal psychosis clearly demarcated from postictal psychosis. We compared them with 2,773 other epilepsy outpatients as a control. Risk factors for psychosis were examined within the temporal lobe epilepsy (TLE) group and the more extended group of symptomatic localization-related epilepsy. Further, nuclear schizophrenia and other nonschizophrenic psychotic disorders were compared. We confirmed a close correlation between TLE and interictal psychoses. Within the TLE group, only early epilepsy onset and a history of prolonged febrile convulsions were revealed to be significantly associated with interictal psychosis. Within the symptomatic localization-related epilepsy group, such parameters as complex partial seizures, autonomic aura, and temporal EEG foci were closely associated with psychoses. There was also a significant difference between groups as to ictal fear and secondary generalization. Whereas patients with early psychosis onset and a low intelligence quotient were overrepresented in the nuclear schizophrenia group, drug-induced psychosis and alternative psychosis were underrepresented. TLE proved to be preferentially associated with interictal psychoses. Within the TLE group, medial TLE in particular was found to be more closely associated with psychosis. Our data support the original postulation of Landolt, stating that alternative or drug-induced psychoses constitute a definite subgroup of interictal psychoses, which are different from chronic epileptic psychoses that simulate schizophrenia.

The association between epilepsy and schizophrenia-like psychosis has attracted extensive attention since the nineteenth century. As long ago as 50 years, most of the basic findings on epileptic psychoses had already been presented. Hill and Pond (1,2) suggested an overrepresentation of psychotic states in patients with temporal lobe epilepsy (TLE) as well as a several-year delay for the beginning of psychotic states after epilepsy onset. In the late fifties, Landolt (3) established the well-known concept of forced normalization and pointed out that there was a reciprocal relationship between abnormal mental states and seizures. Tellenbach (4) emphasized this reciprocal nature and proposed the term alternative psychosis as a purely clinical expression, which does not automatically imply cessation of interictal epileptiform discharge. In the mid-1960s, Slater and Beard (5) published a comprehensive report about a chronic epileptic psychosis that simulated schizophrenia. In the ensuing decades, the fundamental knowledge about epileptic psychoses has been amplified and extended. However, as Sachdev (6) pointed out in his recent review on epileptic psychoses, most of the studies, including classic writings, have lacked control subjects or a precise definition of both psychosis and epilepsy. As a result, controversy remains, even regarding the most fundamental issues such as the association between TLE and interictal psychoses (7). Further, exact demarcations and amplifications of the phenomenology of postictal psychosis in the last decade (8–11) have prompted a reexamination of interictal psychosis, because removing patients with postictal psychosis from a group of study subjects allows the sample subject group to be more homogeneous. Considering such data and conceptual limitations, this study is an attempt to amplify previous results regarding epileptic psychosis on a relatively large scale, based on international seizure and epilepsy classification (12,13) and DSM IV criteria (14).


This purely retrospective study, based on a review of case records, included all outpatients (n = 132) of the Kansai Regional Epilepsy Center who were known to have had epilepsy with interictal psychotic episodes from 1984 to 1999. The clinical data from these subjects, such as sex, age at epilepsy onset, age at psychosis onset, side of the lesion, full-scale IQ, history of prolonged febrile convulsions, prognosis of epilepsy, and types as well as incidence of epilepsy, epileptic seizures, and EEG abnormalities were examined. Any febrile convulsion lasting for >30 min was termed a prolonged febrile convulsion. As a control, we used the data of other outpatients with epilepsy but with no history of interictal psychotic episodes (n = 2,773) who had been treated at our institute.

Epilepsy and seizure classifications were based on the definitions proposed by the International League Against Epilepsy (12,13). Schizophrenia and other psychotic disorders fulfilling DSM IV criteria were the psychoses examined. We then subdivided the psychotic states further, according to the subcategories of schizophrenia and other psychotic disorders as defined by DSM IV criteria (14). Among these subcategories, schizophrenia and schizophreniform disorders were distinguished by the duration of illness. Although an episode of the latter disorder lasts for at least 1 month but <6 months, continuous signs of the disturbance persist for at least 6 months in schizophrenia. The duration of brief psychotic disorder is still shorter (<1 month and >1 day), and the essential symptoms are less than those in schizophrenia and schizophreniform disorders. When nonbizarre real life delusions, such as being followed, poisoned, infected, and loved from a distance, lasted for >1 month, they were categorized as a delusional disorder. We excluded patients with postictal psychosis, which was defined as an episode that occurred within 7 days after the last generalized tonic–clonic seizures or clusters of complex partial seizures. Ictal psychotic episodes directly corresponding to ictal epileptiform discharge, such as nonconvulsive status epilepticus, were also excluded. Episodes of interictal psychoses that occurred with absence or remarkable diminution (>80%) of seizures, at least during the initial phase of the psychosis, were designated alternative psychoses for the current study. We regarded patients who exhibited psychopathologic features of schizophrenia based on the DSM IV subcategories as having nuclear schizophrenia (n = 54) and those who exhibited the other subtypes of schizophrenia and other psychotic disorders as having a nonschizophrenic psychotic disorder (n = 78). Based on the spontaneous complaint descriptions written in the case cards, we operationally judged a symptom to be present only if the symptom in question was rated >4 according to the Brief Psychiatric Rating Scale (15).

Statistical analyses were made with chi-square tests using Yates' modification for small numbers.


Epilepsy classification

TLE exhibited a strikingly high association with interictal psychotic states that was highly significant statistically (chi-square, 133.0; p < 0.001; Table 1). In contrast, interictal psychotic states occurred significantly rarely in patients diagnosed as having idiopathic and symptomatic or cryptogenic generalized epilepsy (chi-square, 7.02; p < 0.01; and chi-square, 7.16; p < 0.01; respectively).

Table 1. Epilepsy classification
Interictal psychotic statesYes
(n = 132)
(n = 2,773)
  • a

     Statistically significant difference (p < 0.01).

Idiopathic localization-
 related epilepsy
Symptomatic localization-
 related epilepsy
Idiopathic generalized
Symptomatic/Cryptogenic generalized epilepsy7369a

Subcategories, symptoms, and recurrence of psychosis

In Table 2, the subtypes of psychotic states are shown. Most of the patients (n = 120) complained about delusions of reference or persecution. Verbal hallucinations were documented in 84 patients. In 34 patients, bizarre delusions such as those of perception or Capgras' syndrome occurred. Thought broadcasting was reported in 17 patients, and in eight, bodily hallucinations were noted. In contrast, visual hallucinations were present as symptoms in only two patients. The number of psychotic episode recurrences are summarized in Table 3.

Table 2. Subcategories of interictal psychotic states (DSM IV)
Schizophreniform psychosis33
Brief psychotic episodes15
Delusional psychosis27
Schizoaffective psychosis1
Psychotic disorder not
 otherwise specified
Table 3. Number of transient psychotic episode recurrences
No recurrence53
Two episodes13
Three episodes6
Four episodes8
More than four episodes or undetermined9

Drug-induced psychoses

In 45 patients, adding or increasing the dose of antiepileptic drugs (AEDs) preceded a beginning of psychotic episodes. As indicated in Table 4, >90% of the patients were either newly given zonisamide (ZSM) or phenytoin (PHT), or the dosage of either drug was increased before development of a frank psychosis. In 22 patients with ZSM-induced psychosis, the total number of additions or increased dose of AEDs was 93. Although a psychotic episode occurred 25 times after adding or increasing ZSM, only four episodes were observed for other AEDs, except for PHT. This correlation between ZSM and psychotic episodes was statistically highly significant (p < 0.001; chi square, 63). Likewise, in 20 patients with PHT-induced psychosis who had AEDs added or increased a total of 92 times, the correlation between occurrence of frank psychosis and augmentation or initiation of PHT therapy was also highly significant (p < 0.001; chi square, 41). Interestingly, whereas ZSM was newly given to 20 of 22 patients, PHT was newly given to only six of 20, and its dosage increased in the other 14. This difference was statistically significant (p < 0.01; chi square, 16.48). Further, whereas 75% of the PHT-induced psychoses (15 of 20 patients) showed apparent alternations between psychosis and seizure, only 64% of the ZSM-induced psychoses (14 of 22 patients) manifested themselves as an alternative psychosis. Whereas only six of 45 patients with episodes of drug-induced psychoses later developed chronic psychotic states, initial transient psychotic episodes became gradually chronic in 18 of 54 patients who experienced only unprovoked psychotic episodes. This difference was also statistically significant (p < 0.05; chi square, 5.35). In addition, nine of the 45 patients experienced unprovoked psychotic episodes as well, none of which preceded a drug-induced one. It should be also noted that at least five patients with drug-induced psychosis achieved a complete suppression of seizures without a recurrence of psychosis ultimately after receiving alternative AEDs.

Table 4. Drug-induced psychosis
  Alternative psychosis
  1. ZNS, zonisamide; PHT, phenytoin; CBZ, carbamazepine; ESM, ethosuximide.


Risk factors for interictal psychotic episodes

When our analysis was limited to TLE, only early epilepsy onset (10 years old or younger) and prolonged febrile convulsions were significantly more often documented in patients with interictal psychotic states (chi square, 4.87; p < 0.05; and chi square, 13.73; p < 0.01, respectively). However, as indicated in Table 5, when the comparison was extended to symptomatic localization epilepsy in general, the difference in early epilepsy onset was obscured. Instead, such parameters as autonomic aura, complex partial seizures, and temporal EEG foci were revealed to have highly significant differences (chi square, 12.87; p < 0.01; and chi square, 16.73; p < 0.01; and chi square, 23.54; p < 0.01, respectively). There was also a significant difference between groups as to ictal fear (chi square, 3.81; p < 0.05) and secondarily generalized seizures (chi square, 4.65; p < 0.05). Although there was a clear bias toward male patients and left-sided lesions in patients with interictal psychoses, this proved to be more apparent than real, as those without psychotic episodes showed a similar tendency as well. Overall, there were no statistically significant differences seen for sex or side of lesions. Further, seizure frequency represented by remission did not show a statistically significant difference.

Table 5. History and clinical data from patients with and without interictal psychoses in symptomatic localization-related epilepsy
Interictal psychotic
(n = 109)
(n = 1322)
  • CPS, complex partial seizure; GTC, generalized tonic–clonic; FC, febrile seizure.

  • a

     Statistically significant difference (p < 0.05).

  • b

     Statistically significant difference (p < 0.01).

Age at epilepsy onset10.6 ± 6.413.8 ± 11.4
 Younger than 10 y32/109138/1,322a
Side of lesion (L/R)32/19179/123
Sex (M/F)66/43741/582
 Ictal fear11/10973/1,322a
 Autonomic aura27/109166/1,322b
 Psychic aura11/109108/1,322
 Temporal foci83/109688/1,322b
 Extratemporal foci18/109364/1,322
 Laterality (L/R)45/43420/427
Prolonged FC21/10982/1,322b

Nuclear schizophrenia versus nonschizophrenic psychotic disorders

Early psychosis onset (20 years old or younger) and low intelligence quotient were more often found in patients with nuclear schizophrenia (chi square, 4.82; p < 0.05; and chi square, 5.5; p < 0.05; Table 6). There was no excess of female subjects who had early psychosis onset in the nuclear schizophrenic group. In contrast, psychotic episodes clearly induced by a change of AEDs as well as alternative psychosis were closely associated with nonschizophrenic psychotic disorders (chi square, 11.16; p < 0.01; and chi square, 11.26; p < 0.01). No marked differences were found in the other clinical parameters as a function of our subcategorization of psychotic disorders. However, whereas the number of patients with left- (n = 6) and right-sided (n = 6) lesions on computed tomography (CT) or magnetic resonance imaging (MRI) was equal in male subjects with nuclear schizophrenia, all female patients (n = 7), except one, had left-sided lesions.

Table 6. Nuclear schizophrenia versus nonschizophrenic psychotic disorders
(n = 54)
(n = 78 (A4)
  • a

     Statistically significant difference (p < 0.05).

  • b

     Statistically significant difference (p < 0.01).

Age at epilepsy onset10.0 ± 5.911.9 ± 6.4
 10 y.o.>26/5426/78
Age at psychosis onset23.9 ± 7.226.4 ± 8.4
 20 y.o.>17/5412/78a
Interval between epilepsy
  and psychosis onset
13.0 ± 8.814.0 ± 10.3
 10 y.o.>35/5444/78
Side of lesion (L/R)13/721/14
Sex (M/F)30/2451/27
Mental retardation  
 Full-scale IQ <8022/5417/78a
Prolonged FC6/5415/78
Alernative psychosis8/5433/78b
 Ictal fear4/5455/78
 Autonomic aura13/5414/78
 Psychic aura3/549/78
 Temporal foci32/5451/78
 Extratemporal foci7/5410/78
 Laterality (L/R)18/1827/25


The prevalence of interictal psychosis in patients with epilepsy in the current study (4.7%) agreed well with that seen in other specialized epilepsy clinics (4–9%) (16–19). Our data also support the suggestion of a majority of previous studies that psychosis in epilepsy might be preferentially associated with the temporal lobe (5,20–23). In the current study, it was also revealed that there were significantly fewer patients with interictal psychotic states in idiopathic as well as symptomatic or cryptogenic generalized epilepsy. Although this superficially contradicts the postulations of Wolf (24), who emphasized the association between idiopathic generalized epilepsy and psychosis, it should be noted that the association in that study depended largely on ethosuximide (ESM) treatment. Because most of the patients with idiopathic generalized epilepsy in our institute were treated successfully with valproate (VPA), ESM was only rarely prescribed.

Our data include an interesting suggestion regarding the relationship between age at epilepsy onset and history of interictal psychosis. Except for the study of Roberts et al. (25), previous investigations have failed to demonstrate the difference between psychotic and nonpsychotic patients as a function of age at epilepsy onset, although psychotic patient groups have consistently revealed a younger age at epilepsy onset, usually in early adolescence (26–28). In the current study as well, comparisons within the symptomatic localization epilepsy group did not reveal a significant difference. However, when we limited the study subjects and controls to only those with TLE, early age at epilepsy onset (10 years old or younger) was significantly more often associated with psychotic episodes. Together with a significantly high incidence of prolonged febrile convulsion in infancy (29), we assume that the close link between history of psychosis and medial TLE might lead to this early age at onset finding. Kristensen and Sindrup (27) provided supportive evidence for the assumption that psychotic patients have a substantial preponderance of temporal mediobasal spike foci, as recorded with sphenoidal electrodes. However, there remain some conflicting data about the histologic nature of medial TLE. Although our previous study with MRI findings (29) confirmed the frequent presence of mesial temporal sclerosis in patients with interictal psychosis, Bruton et al. (30) reported that mesial temporal sclerosis and TLE occurred with equal frequency in psychotic and nonpsychotic groups. Further, Taylor (31) commented that epilepsy patients with schizophrenia-like psychosis were less likely to have mesial temporal sclerosis and more likely to have alien tissue lesions. These discrepancies prompted a further amplification in future studies. Although our data, agreeing with Hermann et al. (32) as well as Kristensen and Sindrup (27), demonstrated an increased incidence of ictal fear and autonomic aura in patients with interictal psychosis, these auras were found to be salient features of TLE, but not medial TLE in particular.

Although we noted only a slight excess of left-sided EEG abnormalities in patients with interictal psychotic states, CT and MRI revealed a clear bias for left-sided lesions. In the nuclear schizophrenia group, the bias was even more extreme, just as Trimble and Perez (33) pointed out. Further, in the current series, left-sided lesions were found in almost all female patients with nuclear schizophrenia, a finding in agreement with Taylor's report (34). However, these differences did not reach a statistically significant level, because patients with epilepsy and without psychotic episodes also revealed a bias for left-sided lesions. Previous findings from EEG (22,27,35–37), neuropathologic (25,30), and neuroimaging studies (including ours) (22,23,29,38) that examined laterality after the suggestion was first made by Flor-Henry (26), remain inconclusive. As Trimble (39) suggested, an investigation procedure, such as positron emission tomography (PET) and single photon emission CT (SPECT) (40,41), which detects more subtle metabolic changes, would be necessary to further elucidate the laterality effect on the genesis of epileptic psychosis.

When we focused on patients with nuclear schizophrenia who demonstrated an undesirable course of psychosis, different risk factors emerged. Perez and Trimble (22) pointed out a significantly shorter interval between onset of seizures and onset of psychosis in epileptic patients with nuclear schizophrenia. Although the interval in the current series was well within the range of 11–15 years suggested by previous authors (17,26,37,42–45), the difference between nuclear schizophrenia and other nonschizophrenic psychotic disorders was not statistically significant. Instead, we noted a higher occurrence of early psychosis onset at an age of 20 years or younger with nuclear schizophrenia as compared with the other subcategories of psychotic disorders. This was found in only a small proportion of our patients, but was significantly greater in the nuclear schizophrenic group. Although Taylor (34) pointed out, by reexamining the data provided by Slater and Beard (5), that female patients have a higher risk for the development of psychosis, and that this risk is passed by the age of 25 years, such dependency on the age at psychosis onset for sex difference was not found in the current series. The data on intellectual performance, when compared with those of previous studies, were also conflicting. There have been reports both in favor of (22,23) and in opposition to (18,33) a preponderance of patients with subnormal intelligence quotient in nuclear schizophrenia subgroups. The 6-month criterion for schizophrenia based on DSM IV used in the current study has been not popular in previous studies; therefore, direct comparison is difficult. It should also be stressed that those parameters indicating involvement of limbic structures, such as complex partial seizures, autonomic aura, temporal EEG foci, and a history of prolonged febrile convulsions, did not reveal significant differences when grouped into subcategories of psychoses. In conclusion, the involvement of limbic structures is a risk factor for development of interictal psychoses in general, but not for chronic nuclear schizophrenia in particular.

Finally, our analysis of subcategories of psychoses supported the original postulation of Landolt (3) that episodes of drug-induced psychoses or alternative psychoses do not lead to chronic schizophrenia in most cases. In other words, a substantial portion of the alternative or drug-induced psychoses constitute a separate subgroup of patients distinct from chronic psychosis simulating schizophrenia, usually resulting in a severe functional decline. Further, our study found that some patients became psychotic after control of seizures with ZSM or PHT but did not develop any sign of psychosis with alternative drugs in spite of the complete disappearance of seizures. This suggests that at least some drug-induced psychotic episodes are not precipitated only by the decreased frequency of the seizures, but rather are drug specific. In a comprehensive book dedicated to forced normalization, Schmitz and Trimble (46) emphasized the renaissance of this classic concept in a new era of potent AEDs. Our data reinforce the value of this puzzling phenomenon as a tool to elucidate schizophrenia or other psychotic experiences. In the current study, although PHT-induced psychoses were dose dependent, those induced by ZSM were not. Potentially, such different modes of induction according to the AEDs given might serve as an ideal model to investigate the biochemical nature of psychoses. Moreover, the prevailing view that the potency of a drug against seizures and not the mode of action is the principal determinant for the emergence of psychotic episodes, remains open to question.