Abstract
Campbell C, McCormack M, Patel S, Stapleton C, Bobbili D, Krause R, Depondt C, Sills GJ, Koeleman BP, Striano P, Zara F, Sander JW, Lerche H, Kunz WS, Stefansson K, Stefansson H, Doherty CP, Heinzen EL, Scheffer IE, Goldstein DB, O’Brien T, Cotter D, Berkovic SF, EpiPGX Consortium; Sisodiya SM, Delanty N, Cavalleri GL. Epilepsia. 2022;63(6):1563-1570. doi:10.1111/epi.17228
Levetiracetam (LEV) is an effective antiseizure medicine, but 10%-20% of people treated with LEV report psychiatric side-effects, and up to 1% may have psychotic episodes. Pharmacogenomic predictors of these adverse drug reactions (ADRs) have yet to be identified. We sought to determine the contribution of both common and rare genetic variation to psychiatric and behavioral ADRs associated with LEV. This case-control study compared cases of LEV-associated behavioral disorder (n = 149) or psychotic reaction (n = 37) to LEV-exposed people with no history of psychiatric ADRs (n = 920). All samples were of European ancestry. We performed genome-wide association study (GWAS) analysis comparing those with LEV ADRs to controls. We estimated the polygenic risk scores (PRS) for schizophrenia and compared cases with LEV-associated psychotic reaction to controls. Rare variant burden analysis was performed using exome sequence data of cases with psychotic reactions (n = 18) and controls (n = 122). Univariate GWAS found no significant associations with either LEV-associated behavioural disorder or LEV-psychotic reaction. PRS analysis showed that cases of LEV-associated psychotic reaction had an increased PRS for schizophrenia relative to controls (p = .0097, estimate = .4886). The rare-variant analysis found no evidence of an increased burden of rare genetic variants in people who had experienced LEV-associated psychotic reaction relative to controls. The polygenic burden for schizophrenia is a risk factor for LEV-associated psychotic reaction. To assess the clinical utility of PRS as a predictor, it should be tested in an independent and ideally prospective cohort. Larger sample sizes are required for the identification of significant univariate common genetic signals or rare genetic signals associated with psychiatric LEV ADRs.Objective:
Methods:
Results:
Significance:
Commentary
Levetiracetam (LEV) is a widely used anti-seizure medication (ASM) in both focal and generalized epilepsies, as monotherapy or as adjuvant therapy. However, almost one-fifth of patients will have some form of neuropsychiatric adverse drug reaction (ADR) to LEV, leading to dose reduction or even discontinuation of this medication. 1 Neuropsychiatric symptoms can vary from irritability, depression, personality changes, and suicidal ideation. Importantly, 1% of patients on LEV will experience psychosis. This number is much higher than the expected number on patients exposed to other ASMs.
It is unclear why some patients develop neuropsychiatric ADR to LEV, while others, sometimes taking much higher doses, do not. One possibility is that the drug itself leads to these reactions. However, this would not explain why 80% of patients do not have a neuropsychyatric ADR. Another possibility is that the patient’s genetic makeup make them more susceptible to neuropsychiatric ADR when exposed to LEV. For years, pharmacogenomic studies have helped us identify the relationships between drugs and genes in several conditions.
One classic example well-known to epileptologists is the severe cutaneous reaction that can happen in people harboring the human leukocyte antigen (HLA) region alleles HLA-B*15:02 or HLA-A*31:01, when they are exposed to aromatic ASM. Here, the presence of a few variants with small to moderate effect in discrete regions of the genome are associated with Stevens-Johnson syndrome and toxic epidermal necrolysis upon exposure to drugs like carbamazepine. These alleles can have different effects in different ethnic groups. For instance, HLA-B*15:02 allele is a strong risk factor for patients of Han Chinese ancestry. 2 For many years, only patients of Asian ancestry were tested for the HLA-B*15:02 allele. However, a recent study showed that people from self-reported other ethnicities may also develop the severe cutaneous phenotypes upon aromatic ASM exposure, making the criteria for genetic screening less clear. 3
Another pharmacogenomic mechanism leading to unusual ADR is the occurrence of rare genetic variants with a large effect. For instance, patients with 22q11.2 microdeletion syndrome are much more likely to develop seizures when exposed to the antipsychotic drug clozapine. In a study comparing patients with and without this microdeletion, clozapine induced seizures in 44% of patients carrying the microdeletion, while less than 1% of those without an identifiable genetic abnormality had seizures when exposed to this drug. 4,5
Finally, some patients develop certain phenotypes because they have a “genetic tendency,” that is, their genetic makeup as a whole is responsible for the phenotype, but across many genes. We call this a polygenic trait. Since 2018 thanks to large number of patients with genetic data and powerful computational tools, what we used to call “genetic tendency” has been further analysed and quantified and can be calculated through polygenic risk scores (PRS). Polygenic risk scores aggregate the small risk conferred by thousands to millions of variants across the genome into a single score, which can stratify affected and healthy individuals. Polygenic risk scores were initially determined for 5 common diseases: coronary artery disease, atrial fibrillation, type 2 diabetes, inflammatory bowel disease, and breast cancer. 6 More recently, PRS have been identified in several other conditions such as epilepsy 7 and schizophrenia. 8 In some cases, the risk of developing a phenotype due to several variants of small effect as demonstrated by a high PRS is equal to the risk of developing the same disease due to a single rare genetic variant of large effect. For example, a high breast cancer PRS confers a similar risk of cancer as a single rare variant in a BRCA gene.
In the study by Campbell and colleagues, 9 the authors set out to study the contributions of common and rare genetic variants to the neuropsychiatric ADR to LEV. Neuropsychiatric ADR were defined as those that occurred a) within 6 months of the initiation of LEV treatment, b) led to withdrawal or dose reduction of LEV, c) were reversed or improved after withdrawal or dose reduction, and d) the ADR could not be attributed to any other cause by the treating or phenotyping clinician.
The authors assembled a cohort of 1106 patients with epilepsy who were treated with LEV. Out of these, 149 patients had some LEV-induced behavioral disorder (agitation, aggression, irritability, confusion, or cognitive decline), 37 patients had LEV-induced psychotic reactions (vivid hallucinations, misidentifications, delusions, etc), and the remaining 920 did not have any neuropsychiatric reaction to LEV. These 920 patients served as the control group.
The authors then employed 3 different methods to evaluate the pharmacogenomics of LEV-induced neuropsychiatric symptoms: To identify individual common genetic risk variants, the authors applied univariate genome-wide association study (GWAS) analysis comparing patients with LEV-induced behavioral and LEV-induced psychotic ADR to controls. To identify if rare variants were associated with neuropsychiatric ADR, rare variant burden analysis was done in a subset of LEV-induced psychosis and controls. To check if a higher polygenic burden for schizophrenia was associated with the development of LEV-induced psychosis, PRS for schizophrenia was calculated in patients with LEV-induced psychosis and compared to the calculated schizophrenia PRS in controls.
The GWAS analysis done through 3.8 million single-nucleotide polymorphisms (SNPs), with 80% power to detect any genetic variants with a relative risk of 3.34 or greater could not identify any common genetic risk loci for LEV-induced behavioral ADR. Similarly, GWAS through 3.8 million SNPs with 80% power to detect variants with a relative risk of 7.22 or greater did not find any risk loci for LEV-induced psychotic ADR.
Rare variant burden analysis was carried out in a subset of patients with psychotic-LEV-induced ADR and controls. Here, 18 668 genes were individually tested for the enrichment of variation. Among the thousands of genes evaluated were genes previously found to harbor rare variants in people with schizophrenia (SLC6A1, SETD1A and BRM12) and genes which are targets of LEV SV2A. No significant enrichment of rare variants was found in any individual gene. Combining the rare-variant burden in the 3 genes associated with schizophrenia as a unit, also did not provide any significant enrichment.
The authors also compared the schizophrenia PRS in LEV-induced psychosis to the PRS in controls. This evaluation finally showed a difference between the groups. The PRS for schizophrenia was significantly higher in the affected group compared with controls (P = .0097). Looking only at the top 10% of the schizophrenia PRS distribution, the LEV-induced psychosis cases makeup 8.3% of the cohort, while only 1% of them are in the bottom 10% of the schizophrenia PRS.
This study has some limitations. For instance, only patients of European ethnicity were included, since PRS translates poorly across different ancestries. 10 Patients with a clear prior history of psychosis were excluded from the study. This is because a history of social deprivation, depression, anxiety, or recreational drug use have been previously linked to a higher chance of neuropsychiatric adverse reactions to LEV. 11 However, it is not clear that milder forms of these mood disorders were ruled out. Finally, the relative small number of patients with psychosis might have precluded the identification of rare variants with large effect.
This work is relevant as it shows evidence that a high PRS for schizophrenia is a risk factor for psychosis in patients exposed to LEV. This initial work needs to be replicated in larger, prospective and non-European populations. That will help define the utility of PRS in clinical practice. Polygenic risk scores (for epilepsy, schizophrenia, or other conditions) is not routinely used in clinical scenarios yet. However, as GWASs get steadily larger in size, the predictive ability of PRS gets better and soon they will become clinically valuable. As such, epileptologists should understand PRSs role and be ready to use it on a clinical basis in the near future. Just as we check HLA to avoid Steven-Johnson’s syndrome in certain patients, we might be using schizophrenia PRS to avoid psychotic reactions to LEV.
