Psychoses of Epilepsy: Unravelling the Phenotypic and Genotypic Features

Abstract

Objectives

We analyzed the genotypic and phenotypic features of patients with psychosis of epilepsy (POE).

Methods

Patients with POE recruited to an epilepsy genetics research program underwent phenotyping and genetic analysis. The latter included screening for rare pathogenic variants in epilepsy genes, and polygenic risk score (PRS) calculation for common risk variants associated with schizophrenia.

Results

One hundred twenty‐two individuals with POE were identified. Eighty‐six of 122 of the individuals (70%) had interictal psychosis, with schizophrenia the most common interictal phenotype (36/86, 42%). Twenty‐eight of 122 of the individuals (23%) had postictal psychosis (PIP), 2 of 122 of the individuals (2%) had antiseizure medication‐induced psychosis, and 6 of 122 of the individuals (5%) had substance‐induced psychosis. Focal epilepsies were more frequently associated with PIP (24/28, 86%) compared to interictal psychosis (39/86, 45%; p < 0.05). Twenty‐nine percent of the patients with POE with genetic data had a rare pathogenic variant: 19 in an epilepsy gene (PCDH19, SCN1A, DEPDC5, KCNT1, CHD2, SLC2A1, NPRL3, CLN3, NPRL3, ATP1A3, and CACNA1A) and 4 had a chromosomal anomaly. Fifty‐seven percent of the patients with a rare pathogenic variant had interictal schizophrenia/schizophreniform disorder rather than PIP (9%; p < 0.05). PRSs showed that schizophrenia‐related common risk variants were enriched in patients with POE compared to population controls (p = 0.0007), however, among the POE phenotypes, a raised PRS was only observed in interictal schizophrenia (p = 0.015) and not in those with PIP or other interictal POEs.

Interpretation

Interictal POE is threefold more common than PIP, and more likely to be associated with both rare pathogenic variants for epilepsy and common risk variants for schizophrenia. Distinguishing between different POE phenotypes enhances clinical practice and our understanding of etiology, paving the way for precision medicine. ANN NEUROL 2025;98:35–47

Psychosis is an altered mental state characterized by abnormal perceptions or interpretations of the world. ¹ Although the lifetime prevalence of psychosis in the general population is < 1%, ² it affects 6% to 10% of people with epilepsy. ³ , ⁴ Psychoses of epilepsy (POE) are a mental health emergency associated with excess mortality. ⁵

The POE are classified by the temporal relationship between the onset of psychotic features and seizure occurrence. Ictal psychosis forms part of the seizure and resolves within days. ⁶ Postictal psychosis (PIP) comprises time‐limited psychotic symptoms emerging proximal to a seizure(s), that is, within 7 days, and lasting up to 2 weeks. ⁷ Evidence suggests that PIP largely occurs in the context of focal epilepsy, ⁸ and is somewhat associated with male sex and right‐sided epileptiform discharges. ⁷ , ⁹ Interictal POE occurs independently of seizures, with active psychosis for at least a month. ¹⁰ It is associated with a younger age of seizure and psychosis onset than PIP, ⁸ and is said to present with higher rates of disorganized and negative psychotic symptoms. ⁴ The current Diagnostic and Statistical Manual for Mental Disorders (DSM‐5‐TR) cautions against assuming interictal POEs are synonymous with schizophrenia in symptom profile or severity. ¹

The increased frequency of psychosis in people with epilepsy suggests a shared etiology, with genetic mechanisms a likely contributor. ¹¹ Familial aggregation of epilepsy and psychosis suggests that the 2 conditions share an underlying genetic cause. A large population‐based family study (N = 9,653 families) found that people with epilepsy had 5.5 times higher risk of a psychotic disorder and 8.5 times the risk of schizophrenia. ¹² Moreover, offspring of people with a psychotic disorder were at 2.7 times higher risk of developing generalized epilepsy, whereas offspring of people with epilepsy had a 2‐fold increase in the risk of developing psychosis. A small number of rare genetic variants of major effect are also associated with both epilepsy and schizophrenia in a minority of patients, such as recurrent microdeletions (eg, 15q13.3 and 16p13.11), ¹³ or specific genetic epilepsies (eg, PCDH19‐clustering epilepsy). ¹⁴

A genetic contribution to POE from common risk variants linked to primary psychosis has also been identified. A recent genome‐wide association study (GWAS; N > 80,000) sub‐analysis estimated that 97% of the common risk variants implicated in generalized epilepsy were shared with schizophrenia. ¹⁵ , ¹⁶ At a subgroup level, polygenic risk score (PRS) analysis recently showed 58 people with PIP had a higher PRS for common risk variants associated with schizophrenia compared with 1,366 control patients with epilepsy with no psychosis history; in fact, patients with PIP had a similar PRS to 945 individuals with primary schizophrenia, meaning the 2 groups had the same genetic propensity to psychosis. ⁷ The genetics related to other phenotypes like interictal POE, however, remain unexplored.

Phenotypic and genetic heterogeneity of POE have likely hindered understanding of the underlying mechanisms. ³ We aimed to evaluate the genotypic and phenotypic features of POE in a large cohort of individuals with epilepsy, and explore both rare and common variant contribution to the genetic basis of POE.

Methods

Cohort and Phenotyping

A systematic search for patients with a history of POE in the Epilepsy Genetics Research Program database at The University of Melbourne was performed. Each file was searched for free‐text keywords: “psych*,” “psychosis,” “psychotic,” “hallucinations,” “delusions,” and “negative symptoms,” as well as those noted to have a diagnosis of psychosis. Additional cases were recruited when clinicians informed us about patients with a history of POE.

Files were screened by 2 clinical neuropsychologists with expertise in epilepsy and psychiatric comorbidities (authors G.R. and E.H.) to establish that compelling psychotic symptoms or disorders were documented in a person with epilepsy. This was based on available medical records from neurologists and psychiatrists, neuropsychology reports, video‐electroencephalogram (EEG) monitoring, neuroimaging, family pedigrees, and genetic testing. Genetic testing comprised candidate gene screening for rare pathogenic variant detection, and clinical microarray analysis for copy‐number variant detection; all genetic findings met American College of Medical Genetics and Genomics criteria for pathogenic/likely pathogenic variants. ¹⁷

Epilepsy variables included age of seizure onset, seizure types, epilepsy syndrome, magnetic resonance imaging (MRI)‐brain abnormalities, seizure frequency at time of psychosis, seizure clustering, anti‐seizure medication (ASM), epilepsy surgery, and family history of seizures. Epilepsy diagnosis was based on the International League Against Epilepsy and seizure classifications. ¹⁸ , ¹⁹ Patients were classified as having focal epilepsy, idiopathic generalized epilepsy (IGE), genetic generalized epilepsy (GGE), epileptic encephalopathy (EE), or developmental and epileptic encephalopathy (DEE). IGEs comprise 4 epilepsy syndromes with therapeutic and prognostic implications: Childhood Absence Epilepsy, Juvenile Absence Epilepsy, Juvenile Myoclonic Epilepsy, and Epilepsy with Generalized Tonic–Clonic Seizures Alone. ²⁰ Patients that did not fulfill criteria for 1 of these 4 IGE syndromes but had generalized epilepsy were classified as having GGE. ²⁰ EEs and DEEs were differentiated by whether the patient developed normally prior to showing developmental slowing or regression (EE) compared with those with abnormal development from birth (DEE). Other data included age, sex, handedness, use of anti‐psychotic, anti‐depressant, or anxiolytic medication, comorbidities, and cognitive functioning.

The study was approved by the Austin Health Human Research Ethics Committee and the Walter and Eliza Hall Institute of Medical Research (HREC approval G20/01). Written informed consent was obtained from each participant, or else their legal guardian in the case of minors or those with intellectual disabilities.

Phenotyping of Psychosis

Psychiatric data included categorical diagnosis of psychotic disorder by a psychiatrist or psychologist, age of onset of psychotic symptoms, description of typical psychotic episodes, number of psychotic episodes, co‐morbid Axis I and II psychiatric diagnoses, and family history of psychiatric disorders. All psychotic symptoms across episodes were recorded for each case and stratified according to DSM‐5‐TR descriptors: delusions, hallucinations, disorganized thinking (speech), grossly disorganized or abnormal motor behavior (including catatonia), and negative symptoms.

Case files for patients diagnosed with a psychosis phenotype by a psychiatrist or psychologist were verified by both authors Genevieve Rayner and Eliza Honybun. For patients where a psychosis phenotype was not recorded by a psychiatrist or psychologist, a consensus diagnosis was reached by both authors Genevieve Rayner and Eliza Honybun who reviewed all available clinical data with reference to the DSM‐5 criteria. Raters were not able to be blinded to genetic diagnosis, but typically were not aware of the patients’ genetic diagnoses at the time of reviewing the psychosis diagnosis.

Patients were classified into those whose psychosis was temporally associated with seizures (PIP), epilepsy treatment (ASM‐induced psychosis), or substance use; versus those who experienced psychosis interictally.

Psychosis Related to Seizures, Treatment, or Substance Use

Postictal psychosis was defined according to the International League Against Epilepsy Commission on Psychobiology of Epilepsy criteria as a psychotic episode within 7 days of a seizure or cluster of seizures, in many cases after a lucid interval of no more than 48 hours, and lasting no longer than 2 weeks. ⁶ ASM‐Induced Psychotic Disorder was defined as psychotic symptoms that correlated with taking a specific ASM. No patients had “forced normalization” or “alternate psychosis” where POE emerges in the context of EEG normalization or seizure freedom (eg, post‐surgery), respectively. ²¹ Substance‐induced POE was diagnosed according to DSM‐5 criteria for a substance‐induced psychotic disorder, where psychotic symptoms were a consequence of a drug and ceased after cessation of the agent. ¹

Interictal Psychosis

Interictal POE was diagnosed when the psychotic episode(s) occurred without a temporal relationship with seizures, defined as 7 days after a seizure or seizure cluster. Psychoses were classified according to DSM‐5 criteria for schizophrenia, delusional disorder, brief psychotic disorder, schizophreniform disorder, or schizoaffective disorder. ¹ Differential diagnoses were based on the length of the psychotic symptom phase, number or type of psychotic symptoms, functional impact of the psychosis, and presence or absence of comorbid mood episodes (depression or mania). Patients were characterized as “other” when they did not meet criteria for a specific disorder. Schizotypal personality disorders were considered but not identified in our cohort.

Statistical Analyses

Group‐level statistical analyses comparing different phenotypes across clinicodemographic and rare genetic variant findings (such as pathogenic variants of major effect including copy number variants) were undertaken using Jamovi. Linear relationships were assessed using 2 sample t tests or analysis of variance (ANOVA), with categorical data analyzed with Chi‐square tests using a continuity correction where applicable. Statistical significance was set at p < 0.05; effect sizes are reported for all analyses.

Schizophrenia Polygenic Risk Score Analysis

Schizophrenia PRSs were calculated for patients with POE with DNA available (n = 113/122) for single‐nucleotide polymorphism (SNP) genotyping. We excluded patients with POE with ASM/substance‐induced psychosis due to their small sample size (n = 8). Population controls were obtained from the QSkin Sun and Health Study, a prospective cohort study randomly sampled from the Australian state of Queensland. ²²

SNP Genotyping and Quality Control

The SNP genotyping for patients and controls was performed on the Ilumina Global Screening Array. We excluded samples with call rates <95%, high (>0.2) or low (< −0.2) rates of heterozygosity or where there was a missing, ambiguous, or sex mismatch between genotype and reported sex. SNPs with low minor allele frequency (MAF < 0.5%) or failing Hardy–Weinberg equilibrium (p < 10–6) were excluded.

Imputation to the HRC r1.1 2016 (GRCh37/hg19) reference panel was performed using Minimac4 as implemented on the Michigan Imputation Server 23 with pre‐imputation phasing using Eagle version 2.4 (Broad Institute of Harvard and MIT, Cambridge, Massachusetts, USA). ²³ Post‐imputation we selected high‐quality imputed and genotyped SNPs (imputation quality scores R² > 0.9) and repeated the above standard quality control measures. Principal components analysis for ancestry was performed on our final cohort merged with 1,000 Genomes data using GCTA‐PCA ²⁴ ; non‐European clustering samples were excluded.

PRS Analyses

The PRSs were calculated using PRSice‐2 ²⁵ for POE cases and population controls using SNP effect sizes from the 2022 Schizophrenia GWAS. ²⁶ We used a significance threshold of p < 0.05 for SNP inclusion (n = 40,233 SNPs). All PRS values were normalized to a standard normal distribution with a mean of zero and standard deviation (SD) of one. Statistical analyses were performed using linear regression, accounting for sex, and the first 5 principal components of ancestry in R version 4.2.0 (A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria).

Results

Our cohort comprised 155 patients with a history of psychosis, including 146 identified in our database search and 9 identified by clinicians. Figure 1 summarizes the case ascertainment. Thirty‐three patients were excluded: 5 were given an incorrect diagnosis of epilepsy, 14 had no records pertaining to psychosis, and 12 had insufficient information. The final sample therefore comprised 122 cases of POE.

FIGURE 1.

Case ascertainment methodology. Dx = diagnosis; POE = psychosis of epilepsy. [Color figure can be viewed at www.annalsofneurology.org]

The average age of this POE cohort was 44 years (SD = 15.6, range = 15–95 years), and 53% were female subjects (65/122). Seventy‐six of 122 (62%) patients had a formal diagnosis of POE with a documented DSM‐5‐TR phenotype, ¹ verified on file review. The remaining 46 individuals had psychotic symptoms on file review but had not been given a formal DSM‐5 phenotype.

Phenotypic Features of POE

The predominant psychosis was interictal POE, occurring in 86 of 122 (70%) patients (see Table 1). Seven of these people with interictal POE had a mixed presentation, with interictal POE being the predominant phenotype at file review but with a medical history of PIP too (Table 2). In terms of the “mixed” interictal phenotype, most had a schizophrenia phenotype (4/7), focal seizures (5/7), and abnormal MRI findings (4/7). Three had a genetic cause for their epilepsy (NPRL3, KCNT1, and 2p13.1 dup). Average age at onset of seizures was 9.7 years, which was closer in age at onset to the interictal group (M = 11.8 years, SD = 14.6) than that of the pure PIP group (M = 20.2 years, SD = 13.8). Average age at the onset of psychosis for this mixed presentation was 21.3 years, which again was closer in age at onset of psychosis to the interictal POE group (M = 23.0 years, SD = 9.38) compared to the pure PIP group (M = 31.6 years, SD = 9.48). No other interictal POE cases had a history of PIP.

TABLE 1.

Summary of the Phenotype–Genotype Findings

Phenotype	N	Age Onset Psychosis Symptoms	Age Onset Epilepsy	Sex, M:F	% With a Genetic Finding a	Monogenic Findings b
Interictal POE
Schizophrenia	36 a	25.1 ± 10.0 12–50	14.8 ± 13.0 < 1–46	19:17	23%	PCDH19 (2); SCN1A; CHD2; KCNT1; 2p13.1 duplication; ring chromosome 20; 22q11.2q11.2 deletion
Schizophreniform	16	24.5 ± 8.3 11–42	6.2 ± 5.0 < 1–17	8:8	29%	CACNA1A, CHD2, SCL2A1, 11q23.3q25 dup (mosaic), ATP1A3
Schizoaffective	12	17.5 ± 6.3 c 2–25	10.0 ± 6.4 c 2–20	4:8	8% c	DEPDC5
Other	10	18.7 ± 9.5 3–34	3.8 ± 7.0 < 1–8	2:8	17%	PCDH19; SCN1A
Brief psychotic disorder	7	21.8 ± 7.3 15–32	9.5 ± 6.8 < 1–17	6:1	29%	DEPDC5, SCN1A (2)
Delusional disorder	5	31.8 ± 9.6 18–40	18.8 ± 13.9 2–33	3:2	20%	NPRL3
Seizure‐, substance‐, and treatment‐related POE
PIP	28	31.6 ± 9.5 18–50	20.2 ± 13.8 < 1–52	11:17	4%	CLN3, NPRL3
Substance‐induced	6	18.7 ± 3.5 15–22	13 ± 8.8 5–27	3:3	20%
ASM‐induced	2	15.0 ± NAc 15‐NA c	1.3 ± 0.4 1–1.5	1:1	50%	PCDH19

^a Note that findings where the group sample size is < 20 the percentage is likely to be exaggerated.

^b All genetic findings meet American College of Medical Genetics and Genomics criteria for pathogenic/likely pathogenic.

^c One case of missing data.

ASM = anti‐seizure medication; NA = not applicable; PIP = post‐ictal psychosis; POE = psychosis of epilepsy.

Table 2.

Epilepsy‐Related Characteristics of Interictal POE Versus PIP

Epilepsy Variables	Interictal POE (n = 86)	PIP (n = 28)	p
Age at seizure onset, yr
M (SD)	10.58 (10.30) d	20.21 (13.50) a	< 0.001***
Epilepsy duration, yr
M (SD)	30.42 (13.2) d	30.75 (19.03) a	0.921
Epilepsy type, N (%)
Focal epilepsy	39 of 86 (45%)	24 of 28 (86%)	0.003**
Temporal	15 of 39 (38%)	17 of 24 (71%)
Frontal	10 of 39 (26%)	5 of 24 (21%)
Parietal	1 of 39 (2.5%)	1 of 24 (4%)
Occipital	1 of 39 (2.5%)	0
Hypothalamic hamartoma	3 of 39 (8%)	0
Unclear	9 of 39 (23%)	1 of 24 (4%)
Idiopathic generalized epilepsy	29 of 86 (34%)	2 of 28 (7%)
GGE	1 of 86 (1%)	1 of 28 (4%)
EE	8 of 86 (9%)	1 of 28 (4%)
DEE	9 of 86 (10%)	0
History of PIP, N (%)
Yes	7 (8%)	100%	–
No	79 (92%)	N/A
Frequency of PIP
One episode	0	8 of 24 (33%) e	0.655
Recurrent	7 of 7 (100%)	16 of 24 (66%) e
Focal lesion on MRI, N (%)
Yes	20 of 86 (23%)	15 of 28 (54%) a	0.003**
Hippocampal sclerosis	6 of 20 (30%)	5 of 15 (33%)
Cavernoma	0	2 of 15 (13%)
Focal cortical dysplasia	5 of 20 (25%)	2 of 15 (13%)
Hypothalamic hamartoma	3 of 20 (15%)	0
Periventricular nodular heterotopia	2 of 20 (10%)	0
Other	4 of 20 (20%) f	5 of 15 (33%) b
No	66 of 86 (77%)	13 of 28 (46%)
Exposure to levetiracetam, N (%)
Never	61 of 86 (71%)	16 of 28 (57%)	0.150
Past c	16 of 86 (19%)	6 of 28 (21%)
Current c	9 of 86 (10%)	9 of 28 (32%)

^** p < .01.

^*** p < .001.

^a One case of missing data.

^b One case each of tuberous sclerosis, encephalocele, dysembryoplastic neuroepithelial tumor, ganglioglioma, and gliosis.

^c Categories not mutually exclusive.

^d Eleven cases of missing data.

^e Four cases of missing data.

^f One case each of ganglioglioma, traumatic encephalomalacia, hemiatrophy, and polymicrogyria.

DEE = developmental and epileptic encephalopathy; EE = epileptic encephalopathy; GGE = genetic generalized epilepsy; MRI = magnetic resonance imaging; N/A = not applicable; PIP = postictal psychosis; POE = psychosis of epilepsy.

Of the remaining cases of POE, 28 of 122 (23%) people had PIP, 2 (2%) had iatrogenic ASM‐induced psychosis associated with levetiracetam, and 6 (5%) had a substance‐induced psychosis (see Table 1). Of the 6 cases with substance‐induced psychosis, 2 were triggered by methamphetamine use, 1 by amphetamines, 1 by cannabis, 1 by steroids, and there was 1 case where the drug was not specified. Of note, the steroids were not initiated as an ASM but were prescribed for the treatment of comorbid systemic lupus erythematosus. Given the small number of patients with ASM‐ and substance‐induced psychosis, these individuals were not included in subsequent analyses.

Univariate analysis found that patients with interictal POE had a younger age of seizure onset (mean = 11.8 years, SD = 14.6) compared with those with PIP (mean = 20.2 years, SD = 13.8 and t ₉₈ = 2.6, p < 0.05, Cohen's d = 0.59, 95% confidence interval [CI] = 0.13–1.04, ie, medium‐large effect size). That is, people with interictal POE had seizure onset around 8.4 years earlier than people with PIP. Psychosis also began at a younger age in individuals with interictal POE (mean = 23.0 years, SD = 9.38) compared with those with PIP (mean = 31.6 years, SD = 9.48 and t ₉₀ = 3.6, p < 0.05, Cohen's d = 0.92, 95% CI = 0.40–1.44], ie, large effect size); that is, people with interictal POE experienced the onset of psychosis around 8.6 years earlier than people with PIP. Sex did not differ between interictal POE and PIP (X ² _1,114 = 0.8, p = 0.38, Cramér's V = 0.08, ie, small effect size).

Among the patients with interictal POE, 36 of 86 (42%) had schizophrenia, 16 of 86 (19%) had schizophreniform disorder, 12 of 86 (14%) had schizoaffective disorder, 7 of 86 (8%) had a brief psychotic disorder, 5 of 86 (6%) had a delusional disorder, and 10 of 86 were classified as “other” (12%). In the “other” group, 3 patients had persistent auditory hallucinations only, 2 had major depressive disorder with mood (in)congruent psychotic features, 2 experienced the onset of psychosis associated with the terminal phase of their EE, and 2 individuals had definite psychosis without further detail available. The remaining patient had strikingly early onset of recurrent psychosis by age 7 years with regular vivid and malevolent hallucinations (visual and auditory). The diagnosis was made by an experienced child psychiatrist who treated her with risperidone to an excellent effect.

Psychotic Symptoms in POE

We examined negative and positive psychosis symptoms across the cohort. Negative symptoms associated with psychosis, such as blunted affect, alogia, avolition, asociality, and anhedonia, only occurred in 19 of 122 (15%) of the patients. They were largely seen in interictal POE phenotypes, including schizophrenia (11/19, 58%), schizophreniform disorder (4/19, 21%), and schizoaffective disorder (3/19, 16%; Fig 2). Disorganized behavior occurred in 3 of 122 (2%) cases; all had interictal schizophreniform disorder.

FIGURE 2.

Relative prevalence of psychotic symptoms across different phenotypes of POE. The consistently high frequency of positive symptoms (hallucinations and delusions) across phenotypes is in contrast to infrequent negative symptoms and disorganized behavior. PIP = postictal psychosis; POE = psychosis of epilepsy. [Color figure can be viewed at www.annalsofneurology.org]

In contrast, positive psychotic symptoms, such as delusions and hallucinations occurred across all phenotypes, with delusions documented in 84 of 122 (69%) cases, and hallucinations in 76 of 122 cases (62%). Chi‐square tests showed that hallucinations and delusions were more common in POE than negative and disorganized symptoms (X ² _8,28 = 394.4, p < 0.001, Cramér's V = 0.27, ie, moderate effect size).

Hallucinations were auditory in nature for 60 of 76 (79%) of the individuals with POE who experienced them, followed by visual hallucinations in 32 of 76 (42%; Supplementary Table S1), noting that 32 individuals had more than one type of hallucination. Other types of hallucinations were experienced by less than 25% of cases. Delusions in POE were dominated by paranoid and persecutory themes (69/84, 82%). Ideas of reference, grandiose, and religious themes were present in < 20% of POE delusions (see Supplementary Table S1).

Epilepsy Types Associated with POE Phenotypes

In terms of epilepsy diagnoses, 52% of individuals with POE overall had focal epilepsy (64/122); 28% had IGE (34/122), 3% had GGE (4/122), 8% had developmental and DEEs (10/122), and 8% had EEs (10/122). ¹⁸ PIP was predominantly associated with focal epilepsies (24/28, 86%), compared with low rates of IGE (2/28, 7%), GGE (1/28, 4%), or EE (1/28, 4%; Fig 3). Most PIP cases with focal epilepsy had temporal lobe epilepsy (TLE) 17 of 24 (71%), in line with the relative frequency of TLE among patients with focal epilepsies referred to tertiary epilepsy centers like ours. ²⁷ Interictal POE showed a more even distribution of epilepsy types: 39 of 86 (45%) had focal epilepsy, versus 29 of 86 (34%) with IGE, 9 of 86 (10%) with DEE, 8 of 86 (9%) with EE, and 1 of 86 (1%) with GGE. Focal epilepsies thus occur more frequently in PIP than interictal POE (X ² _4,114 = 16.2, p < 0.05, Cramér's V = 0.38, ie, moderate effect size) and are more likely to have a focal lesion on MRI (see Table 2; X ² _1,113 = 7.54 continuity corrected, p < 0.05, Cramér's V = 0.28, ie, moderate effect size). Of the interictal POE cases with focal epilepsy, only 15 of 39 (38%) had TLE; therefore, there was a higher proportion of people with TLE in the PIP group compared to the interictal POE group (z = −2.50, p = 0.013).

FIGURE 3.

Epilepsy types associated with different POE phenotypes. PIP is associated with a high proportion of focal epilepsies, whereas Interictal POEs show a more even distribution of focal, generalized, and encephalopathic epilepsy types. ASM = antiseizure medication (levetiracetam); DEE = developmental and epileptic encephalopathy; EE = epileptic encephalopathy; Focal = focal epilepsy; GGE = genetic generalized epilepsy, IGE = idiopathic generalized epilepsy; Other = interictal psychosis not otherwise specified; PIP = postictal psychosis; POE = psychosis of epilepsy. [Color figure can be viewed at www.annalsofneurology.org]

Interictal POE was especially common in patients with epileptic encephalopathies. Specifically, 9 of 10 (90%) of individuals with DEE had interictal POE, encompassing schizophrenia (3/10), schizophreniform disorder (3/10), schizoaffective disorder (1/10), brief psychotic disorder (1/10), and persistent auditory hallucinations in the absence of other psychotic features (1/10); there was also 1 patient with DEE who had an ASM‐induced psychosis. Similarly, 8 of 10 (80%) of the individuals with EE had interictal POE; 4 of 10 had “other” forms of interictal POE, 1 had a brief psychotic disorder, and only 3 of 10 had classical schizophrenia or schizoaffective disorders (there was also 1 patient with EE each with PIP and steroid‐induced POE).

POE Phenotypes in People with Intellectual Disability

Intellectual disability (ID) was present in 38 of 122 (31%) patients (Fig 4), which is higher than the estimated prevalence of ID in the adult epilepsy population of approximately 20%. ²⁸ , ²⁹ Among the 38 people with POE and ID, interictal POE occurred in 35 of 38 (95%) patients, compared with PIP in 2 of 38 patients (5%; X ² _1,110 = 11.0, p < 0.05, Cramér's V = 0.32, ie, moderate effect size).

FIGURE 4.

Differential rates of ID across the POE phenotypes, showing that ID is significantly more common in interictal schizophrenia and schizophreniform phenotypes than in PIP. ASM = antiseizure medication; ID = intellectual disability; other = interictal psychosis not otherwise specified; PIP = postictal psychosis; POE = psychosis of epilepsy. [Color figure can be viewed at www.annalsofneurology.org]

Family History

A family history of seizures was more common in individuals with interictal POE (42/85, 49%, 1 case with missing data) than PIP (7/28, 25%; X ² _1,113 = 4.165 continuity corrected, p < 0.05, Cramér's V = 0.21, ie, moderate effect size). Although a family history of psychosis was documented more frequently in interictal POE (11/86, 13%) than PIP (2/28, 7%), this was not significant (X ² _1,114 = 0.225 continuity corrected, p = 0.635).

Genetic Findings

Rare Pathogenic Variants

Genetic testing was performed in 79 of 122 (65%) individuals, of whom 23 of 79 (29%) patients had a pathogenic variant of major effect detected (see Table 1). This included pathogenic variants in epilepsy genes in 19 of 23 (83%) patients: PCDH19 (4 cases), SCN1A (4 cases), CHD2 (2 cases), DEPDC5 (2 cases), NPRL3 (2 cases), KCNT1 (1 case), SLC2A1 (1 case), CLN3 (1 case), CACNA1A (1 case), and ATP1A3 (1 case), and chromosomal abnormalities in 4 of 23 (17%) individuals, including 11q23.3q25 duplication, 2p13.1 duplication, 22q11.2 deletion, and ring chromosome 20.

These pathogenic variants were more likely to be associated with interictal POE, specifically, 13 of 23 (57%) patients had interictal schizophrenia/schizophreniform disorder, compared with PIP (2/23, 9%, X ² _1,79 = 3.96, p < 0.05, Cramér's V = 0.22, ie, moderate effect size).

In terms of epileptology, rare pathogenic variants of major effect were seen in 9 of 10 DEE cases with POE. This is unsurprising as 90% of the currently known approximately 1,000 monogenic epilepsy genes are associated with DEEs. ³⁰

Common Polygenic Risk Variants

Schizophrenia PRSs were calculated for 91 of 122 (75%) patients. Reasons for exclusion from this analysis included lack of DNA available (n = 8), non‐European ancestry (n = 14), failed SNP quality control measures (n = 4), and phenotypes with small numbers (ie, ASM‐ and substance‐induced psychosis).

Common risk variants contributing to the schizophrenia PRSs were significantly enriched in patients with POE compared with controls (p = 0.0007; Fig 5A). Stratification of the POE cases demonstrated a positive trend for the 3 largest phenotypic groups (PIP, schizophrenia, and all other interictal POE), however, only the schizophrenia group reached significance (p = 0.015; Fig 5B).

FIGURE 5.

Schizophrenia common polygenic risk is elevated in patients with POE (A) and in the three largest POE phenotypic groups (B). Shown are the means of the normalized schizophrenia PRS values with 95% CIs for European POE cases and population controls. The uncorrected p values for mean differences between cases and population controls are given as numbers and derive from regression models that included sex and the first 5 ancestry principal components as covariates. CI = confidence interval; PIP = postictal psychosis; POE = psychosis of epilepsy; PRS = polygenic risk score. [Color figure can be viewed at www.annalsofneurology.org]

Discussion

Psychosis has long been known to occur in individuals with epilepsy, either postictally or interictally. Although PIP is easily recognized, interictal psychosis has emerged as a heterogeneous entity that ranges from short‐lived psychotic symptoms to chronic schizophrenia. In our cohort of 122 patients with epilepsy, we showed that interictal POEs are 3 times more common than PIP, and ASM‐ or substance‐induced psychosis is comparatively rare. The most frequent interictal POEs were severe phenotypes like schizophrenia and schizophreniform disorder. Interictal POEs were associated with a younger age at onset of both seizures and psychosis, higher rates of ID, and a fairly even distribution of epilepsy type compared to PIP in which focal epilepsies, such as TLE, predominated. More than a quarter (29%) of individuals with POE had a pathogenic gene variant of major effect, which was more likely to be associated with interictal POEs; specifically, schizophrenia and schizophreniform disorder. Interictal schizophrenia was also more enriched for psychosis‐related common risk variants of minor effect based on PRSs, compared with other POE phenotypes. Together, these phenotypic features and etiological findings highlight that distilling POE phenotypes will inform clinical practice and underpin development of precision medicine.

Interictal Psychosis is the most Common Form of POE

Interictal psychosis accounted for 70% of our POE cohort, with 23% due to PIP. This contradicts a portion of the epilepsy literature purporting that peri‐ictal psychoses is more frequent than interictal psychosis. ⁴ , ³¹ This is likely due to several reasons: first, our cohort is drawn from a large epilepsy database that typically follows patients over many years, and therefore captured 7 individuals with PIP evolving to interictal POE—a progression estimated to occur in approximately 10% of people with PIP. ³² Second, interictal POE is under‐researched relative to PIP, likely because the strong association of psychosis following a seizure is the province of the epileptologists, whereas interictal psychosis is often regarded as a second, unrelated disorder. Our findings should reframe concepts of POE to consistently incorporate interictal POEs, as well as the progression from PIP to interictal POE over time. ³²

We use the DSM‐5 diagnostic criteria for psychosis to classify our large, diverse POE cohort into phenotypes, highlighting its utility in understanding phenotypic associations in epilepsy. ¹¹ The most common presentations of interictal POE were chronic, severe phenotypes of psychosis where patients exhibit 2 or more different major psychotic symptoms for at least a month: namely, schizophrenia (41% of interictal POE), schizophreniform disorder (16%), and schizoaffective disorder (14%). Milder interictal phenotypes like delusional disorder and brief psychotic disorder were rare, accounting for just approximately 5% of interictal POE cases. Neuroimaging suggests that interictal POEs are associated with more widespread neurocognitive network abnormalities than PIP. ³³ In our study, markers of widespread network dysfunction in interictal POE included higher rates of ID, earlier seizure onset, and more generalized epilepsy, echoing previous work. ⁴ , ⁸ , ³⁴ In contrast, PIP was strongly associated with focal epilepsies and normal intellect, again commensurate with previous studies. ⁷ , ⁸

Interictal POE More Frequently has a Genetic Etiology Identified than PIP

Patients with chronic interictal POE had more evidence of a genetic basis for their psychosis. Interictal POE was more commonly linked to a family history of seizures than PIP. Chronic and severe interictal POEs, such as schizophrenia and schizophreniform disorder, were more likely to be associated with rare pathogenic variants, in contrast to PIP and other interictal POEs. PCDH19 and SCN1A were the most commonly implicated genes. Psychosis is a well‐known feature of PCDH19‐clustering epilepsy, occurring in nearly a quarter of affected female patients. ¹⁴ Psychosis has only rarely been reported in SCN1A‐Dravet syndrome, and may warrant closer examination. ³⁵ Moreover, the interictal schizophrenia phenotype had a higher schizophrenia PRS compared to PIP, other interictal phenotypes, and population controls. This highlights that both rare and common genetic risks vary by POE phenotype, and underpin vulnerability to this serious comorbidity in patients with epilepsy.

These results expand Kousuke Kanemoto's gene–environment interaction model of POE, which suggests that a genetic predisposition to psychosis in some people with epilepsy is activated by seizure‐related factors (environment) to varying degrees (Fig 6). ²¹ Seizures might influence the expression of psychosis‐related genes to give rise to positive symptoms in particular, given that hallucinations and delusions were the predominant psychotic symptoms in our cohort. Positive psychotic symptoms are associated with overactivity of the mesolimbic dopamine system, and it has been speculated that the abnormal dopamine signaling that accompanies seizures could be involved in the expression of psychosis‐related genes, ²¹ , ³⁶ This could also account for why negative symptoms were uncommon in our cohort of POE and others, ⁴ supporting the persistent notion that POE is qualitatively different to phenotypes of primary schizophrenia characterized by chronic negative symptoms, cognitive impairment, and poor treatment response. ³⁷

FIGURE 6.

An evidence‐based gene–environment interaction model of POE relative to primary schizophrenia, showing that different phenotypes of POE have varying genetic liabilities, which are strongest for interictal schizophrenia and schizophreniform POEs. Seizures most clearly activate a genetic vulnerability to POE in the case of postictal psychoses. POE = psychosis of epilepsy. [Color figure can be viewed at www.annalsofneurology.org]

In addition to the current findings, PRS analyses using schizophrenia GWAS have also shown that individuals with PIP have an elevated number of common schizophrenia‐related genetic variants that place them at higher risk of psychosis ⁷ ; the same has been shown for levetiracetam‐induced psychosis in people with epilepsy. ³⁸ Therefore, common genetic variation contributes to PIP and ASM‐induced psychosis, as well as interictal POE, albeit accounting for only a small part of the overall variance. ⁷ , ¹⁵ Future research will integrate genetic variants of major, intermediate, and minor effect in unravelling the genetic architecture of POE in larger cohorts. ³⁹ Understanding the complex genetic architecture, together with the impact of environmental factors, will inform risk prediction for POE and enable earlier diagnosis and targeted management. Moving forward, patients with epilepsy at high‐risk of psychosis could benefit from increased surveillance and lifestyle modification to minimize the likelihood of psychosis.

Symptomatology of POE

Delusions in POE were common and dominated by paranoid and persecutory themes (>80% of cases), ⁴⁰ in contrast with older reports that grandiose and religious delusions predominate in epilepsy (only 20% of our cohort). ⁴¹ , ⁴² Given that the content of delusions reflects the patient's cultural milieu, ¹ our findings are consistent with broader trends in Western psychiatry showing reductions in religious delusions mirroring declines in societal religiosity. ⁴³ Paranoid and persecutory delusions are underpinned by several psychological factors that are elevated in people with epilepsy, ⁴⁴ , ⁴⁵ including worrying, heightened interpersonal sensitivity, anomalous internal experiences, and insomnia. ⁴⁶ These factors, amenable to cognitive therapies, ⁴⁷ provide a valuable treatment pathway for people with POE.

Whereas up to 60% of patients with primary schizophrenia have prominent negative symptoms that require treatment, ⁴⁸ this was seen in only a minority of our POE cases (< 30% of POE‐schizophrenia and < 5% of POE–PIP). ¹¹ The paucity of negative symptoms, such as alogia, amotivation, and asociality, particularly in PIP compared with interictal POE, aligns with a previous empirical study of symptomatology in POE. ⁴ It showed occasional negative and disorganized symptoms in people with interictal POE (17%) compared to none in those with PIP (0%), with no difference in the high rates of positive symptoms between phenotypes. Future studies should ensure that negative symptoms of POE are not under‐reported, however, especially in the absence of systematic psychiatric evaluation, by ascertaining that they are not misattributed to common cognitive comorbidities of epilepsy such as bradyphrenia, dysnomia, and poor social cognition. ⁴⁹ This underscores the value of skillful mental health and cognitive evaluation in individuals with complex morbidities (eg, epilepsy, ID, and psychiatric disorder), to ensure appropriate diagnosis and management. ⁵⁰

Limitations and Future Directions

Our study offers important empirical insights into the diverse phenotypes of POE, highlighting distinct clinical features and genetic factors. Its retrospective file review design has limitations that can be overcome by prospective research that incorporates structured psychiatric evaluation. Subjectivities in detecting and documenting psychiatric features can lead to underdiagnosis of cases and symptoms, particularly negative symptoms. The cohort was recruited from our epilepsy genetics research database, where probands may be referred if they have a family history of seizures or are thought to have a genetic etiology, such as a DEE. Replication of the current findings in larger sample sizes and non‐European populations will be important.

Conclusions

Psychosis of epilepsy is not a single disorder, but a complex entity with diverse phenotypes based on symptom presentation, duration, chronicity, and outcomes. Postictal and interictal psychoses are each underscored by distinct neurobiological mechanisms that are both epilepsy‐ and etiologically related. Recognizing the heterogeneity of POE informs understanding of the genetic basis and, in the future, will guide diagnosis and the development of targeted therapies for individuals with epilepsy who present with psychosis.

Author Contributions

G.R. and I.E.S. contributed to the conception and design of the study; G.R., E.H., K.O., and I.E.S. contributed to the acquisition and analysis of data; G.R., E.H., K.O., M.B., and I.E.S. contributed to drafting the text and preparing the figures.

Potential Conflicts of Interest

I.E.S. has served on scientific advisory boards, has been an investigator, and received speaker honoraria, travel, and research funding from UCB, and served on scientific advisory boards, received speaker honoraria, and been an investigator for Chiesi. None of the remaining authors has any conflicts of interest to disclose.

Supporting information

Table S1. Relative frequency of different hallucination and delusion types (N = 122).

Data Availability

Data supporting the current findings are available from the corresponding author upon reasonable request.