The association between psychosis and epilepsy, as suggested by psychotic features during- or after some seizures, has given rise to speculation on common pathophysiologies for both conditions.¹ Seizures may present as nonconvulsive behavioral disturbances ranging from subtle alteration of behavior to psychosis. Prevalence of psychosis among patients with epilepsy has been reported as from 8% to 40%.² Psychosis mimicking schizophrenia has long been recognized as a complication of chronic refractory epilepsy of temporal lobe or orbitofrontal origin. Auditory and visual hallucinations are believed to reflect dysfunction within the temporal lobe, with nearly 20% of adults with refractory temporal lobe epilepsy (TLE) experiencing visual hallucinations as part of the ictus.³ The temporal lobe hypothesis of schizophrenia, first introduced in the 1960s, notes that paranoid psychosis is more prevalent with TLE.⁴

Psychosis of epilepsy is a broad term used to define a group of psychotic disorders associated with seizures. Interictal psychosis occurs in patients with a history of epilepsy, but it may not have a temporal relationship with a seizure episode. It typically consists of brief episodes (days to weeks) of psychosis that occur between seizures and alternate with periods of increased seizure activity.

Classification of psychosis in epilepsy has been based on duration of psychosis, psychopathology, and presence or absence of seizures,⁵ or based on temporal relationship between the two components; that is, ictal, postictal, interictal, and alternative psychoses.⁶ Under this classification scheme, psychotic disorders of epilepsy are considered ictal if they are an expression of ongoing seizure activity, and they are considered postictal if psychosis begins within 7 days of a seizure. Episodes of postictal psychosis typically last less than 1 month⁷ and run a benign course, remitting after several days without neuroleptic treatment; however, it is possible for chronic psychosis to develop, particularly in cases of recurrent postictal psychosis.⁸ A retrospective study in patients with TLE estimates that around 10% of patients with postictal psychosis become chronically psychotic.⁹ Previous case studies have noted that postictal psychosis can be recurrent and can transform into episodic interictal psychosis.¹⁰

However, most of the correlative data about epileptic psychosis are based on studies of patients with recognized temporal or frontal lobe epilepsy. Therefore, it remains unclear to the psychiatrist whether obtaining an electroencephalogram (EEG) should be part of the evaluation of patients presenting with acute psychosis in order to rule out seizure disorder as the underlying etiology. We designed the current study to review the EEG and clinical data collected during in- and outpatient psychiatric evaluation of patients presenting with acute psychotic episode in order to assess the diagnostic value of EEG.

Study Population

We reviewed the EEG and clinical data of 406 adult patients presenting with acute psychosis, including both recurrent and first-time psychoses, from 2001 to 2006. All of these patients were referred by psychiatrists in an academic hospital to rule out ictal or interictal psychosis in patients presenting with an acute psychotic episode. None of these patients had witnessed seizures before or after the onset of psychosis. However, in all patients with new-onset psychosis, a routine EEG was obtained, and, in cases of recurrent psychoses, the EEG was obtained during one of the presentations in order to rule out seizures, in particular, TLE, as a possible underlying etiology for their psychotic episodes. Patients with acute alteration of mental status secondary to medical conditions and/or those referred by nonpsychiatrists were excluded. A total of 240 patients (49% women, age 18–81 years, mean: 41 (SD: 15), and 51% men, age 18–80 years, mean: 37 (SD: 13), including 127 patients (52.92%) with new-onset psychosis and 113 patients (47.08%) with recurrent psychoses were included in the study. Except for 27 outpatients, all of the patients were admitted to the psychiatry unit. Diagnoses of psychosis were made by psychiatrists on the basis of DSM-IV criteria when hallucinations, delusions, or severe behavioral abnormalities were present.

Measurement of Variables

Current and past medical history, including psychosis, paranoid schizophrenia, schizoaffective disorder, bipolar disorder, suicidal ideation, anxiety disorder, depression, attention-deficit hyperactivity disorder (ADHD), substance-induced mood disorder, nonspecific psychotic disorders, and seizure disorder were recorded. Only patients with a primary diagnosis of acute psychotic episode were included in this study, including those who presented with acute psychosis for the first time (“new-onset psychosis”). All nonpsychiatric medical conditions were classified as a separate group for analysis. Medications were categorized as anti-epileptic drugs (AEDs), benzodiazepines, and neuroleptics/antidepressants grouped together as “other medications.”

The EEG studies were obtained within the first week of presentation (mean: 3.72 [SD: 4.13] days), as soon as the patients were cooperative enough for a technically reliable and artifact-free study. All EEGs were recorded for 30 minutes, using standard 10–20 electrode system either in the EEG laboratory or at the bedside in patient’s room on the psychiatry wards. At the time the EEG was obtained, all patients were taking multiple medications for their psychiatric and/or other medical conditions.

The EEG findings were classified either as normal, right- or left- or bilateral temporal or extratemporal spikes, right- or left- or bilateral slowing, diffuse slowing, and posterior basic rhythm slowing. However, in order to achieve enough statistical power, given the relatively small number of cases, the EEG findings were also categorized as normal or abnormal for a separate analysis. Brain-imaging findings, including MRI and CT scan, were classified as normal, or right- or left- or bilateral abnormality. The reported abnormalities included white-matter changes, vascular abnormalities, encephalomalacia, cortical atrophy, nonspecific signal enhancement, and benign cysts. Similar to the EEG findings, the imaging results were also re-categorized for a second analysis as either normal or abnormal, since the more detailed categorization did not provide adequate statistical power.

Statistical Analysis

Wilcoxon rank-sum test and Fisher’s exact test were used to compare the distribution and proportion, respectively. Logistic regression was used to assess the association between potential predictors, including taking AEDs, taking benzodiazepines, taking other medications, time-to-EEG, and EEG findings after adjustment for age and gender. Model discrimination and calibration were assessed by the c-statistic and the Hosmer-Lemeshow statistic, respectively. Because of the small number of abnormal brain MRI and CT cases, each potential risk factor was treated in a separate logistic-regression model, adjusted for age and gender, to test the association between the potential risk factors and status of MRI. The agreement between EEG outcome and MRI outcome was measured by kappa coefficient. Two-tailed p values <0.05 were considered statistically significant. Data were analyzed with SAS Version 9.1 (SAS Institute; Cary, NC).

All patients (N=240) had EEGs; 141 patients had brain MRI scans; and 39 patients had head CT scans; 63 patients (57% women, mean age: 42 [SD: 12]) were on benzodiazepines, and 84 patients (50% women; mean age: 38 [SD: 13]) including 20 patients with a previous history of seizures, were taking AEDs (Table 1); no detailed information regarding their seizures was available.

There was no ictal epileptiform activity recorded in any of the patients; however, 7 patients had intermittent epileptiform discharges (interictal spikes); 10 patients had posterior basic rhythm slowing; 2 patients had right-sided slowing; 9 patients had left-sided slowing; and 19 patients had bilateral slowing.

Among patients with interictal spikes 42.9% were on AEDs, and 28.6% were on benzodiazepines, versus 34.8% and 26.5% among those without spikes, respectively. There was no significant association between the presence of interictal spikes and a series of clinical variables, including age, gender, time interval between the onset of psychosis and EEG, being on AEDs or benzodiazepines, being an inpatient versus outpatient, having new-onset psychosis, and history of seizures (p >0.05).

The association between each variable and abnormal EEG outcome was examined using separate logistic-regression models adjusted for age and gender. Acute psychosis, taking AEDs, and the time interval between reported time of psychosis onset and the EEG were not significant predictors of abnormal EEG (odds ratio [OR]: 0.47 and 95% confidence interval ([CI]: 0.22–1.03); OR: 2.02 (CI: 0.95–4.28), and OR: 2.08 (CI: 0.98–4.40), respectively, with p values very close to 0.05. The same was true for age, gender, taking benzodiazepines, being an inpatient versus outpatient, and history of seizures and taking AEDs before admission (p >0.05). However, taking neuroleptics/antidepressants was a significant predictor of abnormal EEG (OR: 4.45; 95% CI: 1.02–19.37; Table 2).

Of 141 patients who had brain MRI, 29 patients had abnormal imaging. There was no significant association between acute psychosis and MRI imaging abnormalities (OR: 1.02; 95% CI: 0.43–2.41), nor were there any associations between MRI findings and the clinical variables tested. Of 39 patients who had head CT scan, 12 patients had abnormal imaging. Older age and taking AEDs were significant predictors for abnormal CT (p=0.003 and p=0.0375, respectively). There were no significant associations between CT abnormalities and acute psychosis or other clinical variables tested (p >0.05).

Six patients (4.3%) had both abnormal EEG and abnormal MRI; 98 patients (69.5%) had both normal EEG and normal MRI; 14 patients (9.9%) had abnormal EEG, but normal MRI; and 23 patients (16.3%) had normal EEG, but abnormal MRI. The kappa coefficient (95% CI) between EEG and MRI was 0.093 (−0.087 to 0.272), indicating no significant agreement between EEG and MRI findings. One patient (2.6%) had both abnormal EEG and abnormal CT; 21 patients (53.9%) had both normal EEG and normal CT; 6 patients (15.4%) had abnormal EEG, but normal CT; and 11 patients (28.2%) had normal EEG, but abnormal CT. The kappa coefficient between EEG and CT was −0.157 (p=0.403) indicating no significant agreement between EEG and CT findings.

Of all the subgroups of psychosis of epilepsy, interictal psychosis tends to be the most difficult to distinguish from schizophrenia because of its puzzling onset and shared clinical similarities, and therefore, it has been the area for speculation on the mechanisms involved connecting psychosis and epilepsy.^1,9 Despite suggestions that psychosis of epilepsy might be distinguished from schizophrenia on the basis of detailed clinical description of psychotic phenomena, for example, by lack of withdrawn attitude, presence of better premorbid functioning, and a preservation of affect and personality,^3,4,6 without EEG documentation, a definitive distinction remains elusive. In this cross-sectional analysis, we evaluated the association between acute psychosis and temporal lobe epilepsy by examining EEG and brain imaging. Associations between acute psychosis and ictal or interictal epileptiform discharges suggestive of temporal or extratemporal epilepsy, focal abnormal (slow) EEG discharges, or brain MRI/CT scan abnormalities were not significant. Similarly, associations between the EEG and brain-imaging findings and a series of clinical variables (age, gender, time interval between the reported time of onset of psychosis and EEG, taking AEDs or benzodiazepines, being an inpatient versus outpatient, having new-onset psychosis, and history of seizures) were not significant. However, taking neuroleptics/antidepressants was significantly associated with abnormal EEG, and older age and taking AEDs were predictors of abnormal CT scan.

These findings may suggest that, despite the common notion, the incidence of TLE presenting as acute psychosis is negligible. Therefore, routine EEG seems to have a limited role as a screening tool for this purpose in this group of patients. This finding appears to be in contrast to the older reports indicating higher rates. Hill reviewed previous case studies from the 1930s and reported “paroxysmal disturbances similar to those seen in patients with epilepsy” in up to 25%–30% of schizophrenic patients.¹¹ However, considering the EEG technology and lack of unified definitions at the time, old reports indicating “paroxysmal disturbances” may have limited comparative value. Nevertheless, given the limitations of routine EEG in detecting seizure discharges, our findings might only indicate the limited value of a single EEG in diagnosing ictal or interictal psychosis, rather than a low incidence of the syndrome itself. It has been shown¹² that EEG may record epileptiform discharges within 20 minutes in only 37% of patients with epilepsy, whereas a continuous EEG would improve this number to 89% within 24 hours of recording. Therefore, prospective studies using long-term video-EEG monitoring would be needed to provide a more sensitive method of evaluating true incidence of ictal and interictal psychosis. Given the nonspecific nature of focal slow activity or slowing of posterior basic rhythm on EEG, the association between abnormal EEG and taking neuroleptics/antidepressants is likely a nonspecific finding.

The results of this study might have been affected by the fact that the EEG could only be obtained when the patient’s clinical condition was relatively controlled with medications including benzodiazepines and/or AEDs, which might reduce epileptiform discharges on EEG. However, since the chances of recording epileptiform activity on a single EEG, even in patients with documented epilepsy, is only about 50%,^13,14 absence of such EEG discharges may not reliably rule out seizure disorder in about half of patients. Nevertheless, the findings of this study do not provide support for ordering regular EEG as part of evaluation of patients presenting with acute psychosis. We also did not find a significant association between acute psychosis and focal or diffuse slowing on EEG. These discharges are nonspecific findings that may be seen in encephalopathies of different etiologies, structural lesions, or postictal states. Although any of these conditions might present with alteration of behavior or psychosis, regular EEG does not appear to be associated with a particular abnormal EEG pattern.

Another limitation of this study is that our patients included those with new-onset psychosis as well as those with recurrent psychotic episodes. Despite this limitation, our findings did not show any significant association between acute psychosis and ictal or interictal EEG discharges. However, any future studies should include the duration of psychotic illness and the number of psychotic episodes in patients with recurrent psychoses.

Structural brain imaging has been used to differentiate psychosis of epilepsy from schizophrenia. Temporal lobe volume loss may be seen in both schizophrenia and TLE, but hippocampus/amygdala volume loss is one of the most consistent anatomic findings in the latter. Furthermore, it was found that patients with TLE and psychosis differed from those with TLE alone and healthy volunteers in that their total brain volumes were significantly smaller. Although there were no differences in hippocampal volumes among the three groups, there was a significant (16%–18%) bilateral amygdala enlargement in patients with epilepsy and psychosis.⁴

In this study, we did not find any significant association between acute psychosis and brain (structural) imaging abnormalities, whereas older age and taking AEDs were associated with abnormal CT scan. Given the nonspecific nature of CT scan abnormalities and the high ratio of patients on AEDs, this association could reflect age-related or other nonspecific imaging findings. Physiological neuroimaging studies such as single-photon emission computed tomography (SPECT), or positron emission tomography (PET) scans could have complemented the findings. Oshima et.al.¹⁵ have reported a review of 19 SPECT studies from patients with postictal psychosis, including 16 TLE cases, and showed a right-temporal predominance. However, the authors acknowledge that the agitated state of the patients presenting with psychosis remains an impediment to obtaining such studies.

These findings suggest that despite the recognized association between acute psychosis and seizures, in particular TLE, a routine EEG may not provide additional benefits to the clinician as a screening tool in evaluation of patients presenting with acute psychosis. However, further studies are needed to assess the role of long-term video EEG monitoring in this group of patients.