Causal effect of psychiatric disorders on epilepsy: A two‐sample Mendelian randomization study

Abstract Background This study aims to explore the relationship between psychiatric disorders and the risk of epilepsy using Mendelian randomization (MR) analysis. Methods We collected summary statistics of seven psychiatric traits from recent largest genome‐wide association study (GWAS), including major depressive disorder (MDD), anxiety disorder, autism spectrum disorder (ASD), bipolar disorder (BIP), attention deficit hyperactivity disorder (ADHD), schizophrenia (SCZ), and insomnia. Then, MR analysis estimates were performed based on International League Against Epilepsy (ILAE) consortium data (n case = 15,212 and n control = 29,677), the results of which were subsequently validated in FinnGen consortium (n case = 6260 and n control = 176,107). Finally, a meta‐analysis was conducted based on the ILAE and FinnGen data. Results We found significant causal effects of MDD and ADHD on epilepsy in the meta‐analysis of the ILAE and FinnGen, with corresponding odds ratios (OR) of 1.20 (95% CI 1.08–1.34, p = .001) and 1.08 (95% CI 1.01–1.16, p = .020) by the inverse‐variance weighted (IVW) method respectively. MDD increases the risk of focal epilepsy while ADHD has a risk effect on generalized epilepsy. No reliable evidence regarding causal effects of other psychiatric traits on epilepsy was identified. Conclusions This study suggests that major depressive disorder and attention deficit hyperactivity disorder may causally increase the risk of epilepsy.

mood and anxiety disorders, attention deficit hyperactivity disorder (ADHD) and psychosis (Kanner, 2016). The prevalence of psychiatric disorders is higher in patients with epilepsy both before and after the diagnosis of epilepsy (Berg et al., 2017;Dagar & Falcone, 2020). However, it remains challenging to measure the causal relationship between psychiatric disorders and epilepsy independent of possible confounding factors. No randomized controlled trial or large prospective study has elucidated this potential causal effect. If the risk effect of psychiatric disorders on the development of epilepsy is identified, more comprehensive and powerful management may be properly conducted and more mechanisms behind them may be discovered in terms of this spectrum of comorbidities and epilepsy.
Mendelian randomization (MR) is an epidemiological approach that uses genetic variation as a natural experiment to investigate the causal associations between potential risk factors and outcomes in observational data (Emdin et al., 2017). MR, simulating randomized controlled trials, is less likely to be affected by confounding and reverse causality biases than observational studies. Considering the power of the causation evidence, MR sits at the interface of randomized controlled trial and observational studies (Davies & Holmes, 2018). Many studies have been increasingly conducted using this useful method in other fields, while few analyses regarding epilepsy have been reported (Allman et al., 2018).
In this study, we performed a two-sample MR study to evaluate the causal relationship between psychiatric disorders and epilepsy for the first time. Seven psychiatric traits were enrolled from the recent largest genome-wide association study (GWAS), including major depressive disorder (MDD), anxiety disorder, autism spectrum disorder (ASD), bipolar disorder (BIP), ADHD, schizophrenia (SCZ), and insomnia.

Study design
We conducted a two-sample MR analysis to investigate the causal effect of seven psychiatric traits on the risk of epilepsy, following the recommendations of Strengthening the Reporting of Observational Studies in Epidemiology Using Mendelian Randomization (STROBE-MR) (Skrivankova et al., 2021). This MR study relies on three assumptions: (1) the instrumental variable (IV) is associated with the exposure (the relevance assumption); (2) the instrument variable shares no common cause with the outcome (the independence assumption); and (3) the instrument variable only affects the outcome through the exposure (the exclusion restriction assumption), which are presented in Figure 1 ( Davies & Holmes, 2018). All data in this study were published by multiple GWASs; ethics approval and patient consent can be found in the original studies.

Data sources and genetic instruments
GWAS summary statistics of seven psychiatric traits were derived from published studies with large sample sizes of European ancestry. The definitions of the seven psychiatric traits are listed in Table S1. GWASs of ADHD, ASD, MDD, BIP, and SCZ were based on data from the Psychiatric Genomics Consortium (PGC). PGC is the largest international consortium of scientists dedicated to conducting meta-and megaanalyses of genomic-wide genetic data, with a focus on psychiatric disorders (Sullivan et al., 2018). For insomnia and anxiety disorders, we obtained the genetic associations from GWAS based on the UK Biobank data (Table 1) (Rusk, 2018 As shown in Figure 2, SNPs strongly associated with exposure were extracted as candidate IVs at the genome-wide significance threshold (p < 5 × 10 −8 ) for each psychiatric trait, except for anxiety disorder and ASD, the SNPs of which were selected with the threshold of 1 × 10 −5 . Then, dependent SNPs with high linkage disequilibrium (LD) were removed from the candidate IV set based on the following parameter (r2 > 0.001, window size = 10,000 kb). Subsequently, among the candidate SNPs, outcome-related SNPs (SNPs associated with epilepsy) were also removed according to the basic study assump-

Statistical analysis
In this study, the inverse-variance weighted (IVW) method was used to calculate estimates of associations between psychiatric traits and epilepsy. Briefly, the IVW method was performed assuming all SNPs were valid IVs with balanced pleiotropy. To determine whether unbalanced pleiotropy causing bias exists, the intercept from the MR-Egger regression was calculated to test directional pleiotropy (p < .05 infers F I G U R E 1 Conceptual framework for the Mendelian randomization analysis of the causal effect of psychiatric traits on epilepsy. The design is based on the assumption that the genetic variants are associated with psychiatric traits, but not with confounders, and affect epilepsy only through psychiatric traits. SNPs, single-nucleotide polymorphisms. Several sensitivity analyses were performed to validate the robustness of the IVW method, which was robust to pleiotropy. The first method was weighted median regression, which required that at least 50% of the weight for the MR analysis comes from valid instruments.

Psychiatric traits Population
The second method was MR-Egger regression, which can help detect and adjust directional pleiotropy. Moreover, weighted mode and simple mode were also used as supplementary sensitivity analyses.
Finally, we performed a meta-analysis of ILAE and FinnGen data to strengthen the power of the MR analysis. All estimates were reported with p values, and odds ratios with 95% confidence intervals for epilepsy risk were scaled to one SD increase in genetically associated psychiatric traits. All analyses were conducted with R 4.1.3, TwoSampleMR, and MR-PRESSO packages.

Causal association of psychiatric traits with epilepsy in ILAE
The IVW MR analysis showed correlations between certain psychiatric traits and epilepsy based on the ILAE data. Estimates of the In sensitivity analyses, Cochran's Q-derived p was calculated from MR-Egger regression (p = .055, .627, .808, respectively) and showed no evidence of heterogeneity for the instrumental variables of these three exposures. Additionally, no horizontal pleiotropy was found, with an insignificant intercept from the MR-Egger test (p = .132, .507, .910, respectively) and no outliers identified from the MR-PRESSO test. The results of leave-one-out sensitivity analyses suggested that the causal associations between psychiatric traits and epilepsy were not affected by any individual SNP (Figures S1, S3, and S5). More details of the sensitivity test are listed in the Table S12.

Causal association of psychiatric traits with epilepsy in FinnGen and meta-analysis
Further validation was conducted based on data from the FinnGen consortium, and meta-analysis of the ILAE and FinnGen was subsequently performed. Similarly, the IVW method showed a significant causal effect of MDD on epilepsy, with an OR of 1.31 (95% CI   (Tables   S9-S11).
The meta-analysis of the ILAE and FinnGen showed that genetically associated MDD (OR = 1.20, 95% CI 1.08−1.34, p = .001) and ADHD (OR = 1.08, 95% CI 1.01−1.16, p = .020) had a suggestive causal effect on epilepsy and increased the risk of epilepsy as shown in Figure 3.
There was no heterogeneity between the MR analysis enrolled in our meta-analysis (p = .384, .725 for MDD and ADHD, respectively).

Causal association of MDD and ADHD with focal epilepsy and generalized epilepsy
The results showed that MDD had a causal association with focal epilepsy (OR = 1.16, 95% CI 1.01−1.34, p = .039) but no risk effect on generalized epilepsy (OR = 1.19, 95% CI 0.97−1.47, p = .095).

DISCUSSION
In this study, we investigated the causal effect of seven psychiatric traits on epilepsy using MR analysis for the first time. To our knowledge, various comorbidities are up to eight times more prevalent in people with epilepsy than in the general population (Thijs et al., 2019). Psychiatric illness is one of the most common comorbidities, with significant overrepresentation both in adults and in children of patients with epilepsy (Mula et al., 2021). Population-based studies identified a 35% lifetime prevalence of psychiatric comorbidities before and after the diagnosis of epilepsy (Kanner, 2017). Nevertheless, these empirical statistical associations could not clearly clarify the essential relationship between psychiatric comorbidities and epilepsy.
Understanding the causal effect has significant implications for early screening and treatment of corresponding psychiatric disorders, especially in patients with new-onset epilepsy (Keezer et al., 2016). Besides, more undiscovered mechanisms are encouraged to be detected as new targets for effective and low-risk drugs. Shuai et al. conducted an MR study to investigate modifiable risk factors for epilepsy, including depression. However, their study only enrolled one psychiatric trait, and the data on depression were limited by a small sample of 901 cases from the UK biobank (Yuan et al., 2021).

Consistent with previous observational, population-based studies
that showed an increased prevalence of depression in patients with epilepsy, our MR analysis determined a risk effect of MDD on epilepsy.
The prevalence rate in patients with epilepsy was as high as 17%-22%, which was up to 55% in patients with drug-resistant epilepsy (Tellez-Zenteno et al., 2007). Moreover, the odds ratio for the risk of epilepsy in patients with MDD was 2.5 (Adelow et al., 2012). Another study based on the UK General Practice Research Database found that the incidence of depression is significantly higher during the three years preceding the development of epilepsy (Hesdorffer et al., 2012). All these results, to some extent, suggested that depression may increase the risk of the onset of epilepsy. In terms of seizure types, focal epilepsy was reported to be associated with a higher prevalence of depression than generalized epilepsy (Kim et al., 2018;Sanchez-Gistau et al., 2012). Correspondingly, we found a causal effect of MDD only on focal epilepsy.

F I G U R E 3 Associations of three psychiatric traits (MDD, ADHD, and BIP) with epilepsy based on the IVW method in International League
Against Epilepsy (ILAE), in FinnGen, and a meta-analysis of both data sets.

F I G U R E 4
Associations of major depressive disorder and attention deficit hyperactivity disorder with focal and generalized epilepsy based on the IVW method in International League Against Epilepsy (ILAE).
Several animal research studies found that many neurobiological pathogenic mechanisms of primary MDDs may potentially promote the development of seizures either spontaneously or with an insult to the central nervous system, such as endocrine abnormalities, structural and functional abnormalities of cortex, neurotransmitter abnormalities and immunological abnormalities (Singh & Goel, 2021;Kanner, 2011).
First, patients with a primary MDD were found with high blood cortisol concentrations that may cause epileptogenesis. Second, patients with a primary MDD may have decreased cortical thickness, which play a part in worse seizure control. Third, abnormality of neurotransmitters, especially serotonin and norepinephrine, in patients with a primary MDD can increase the risk of epilepsy.
ADHD is common in children with epilepsy, with a prevalence from 12% to 39% in patients with newly diagnosed epilepsy and up to 70% in drug-resistant epilepsy (Rheims & Auvin, 2021). Excluding children, ADHD symptoms also occur in 20−30% of adult patients with epilepsy (Ashjazadeh et al., 2019). The prevalence of ADHD in children with epilepsy is five to ten times higher than that in controls without epilepsy (Cohen et al., 2013;Wagner et al., 2021). A study of 91,605 children (<17 years of age) from the National Survey of Children's Health in America, including 977 children with epilepsy, found that the prevalence of ADHD was much higher than that in children without epilepsy (23% vs. 6%) (Adams & Claussen, 2022). Conversely, epilepsy occurs approximately 4 times more frequently in children with ADHD than in the general population. ADHD was composed of a clear predominance of the combined type (80%), which has been reported to be more common in patients with generalized epilepsy, consistent with the conclusion in our study (Rheims & Auvin, 2021). However, the mechanisms of the relationship between ADHD and epilepsy need to be illustrated in the future.
Anxiety disorder was the second most common comorbidity after depressive disorder in patients with epilepsy. In adult patients with epilepsy, prevalence estimates for anxiety disorder range from 11% to 50% (Hingray et al., 2021). However, our study showed no causal relation between anxiety disorders and epilepsy, the reason for which may be bias from insufficient associated IVs (only 4 SNPs with the threshold of 1 × 10 −5 ) and the objective heterogeneity compared with depressive disorder. Although we found a protective effect of BIP on epilepsy, no statistically significant results were obtained in the confirmation test in FinnGen or the meta-analysis, and we did not find any supportive reports in previous studies (Knott et al., 2015).
The strengths of the present study designed with MR analysis were as follows. First, we used genetic variants allocated randomly to identify the causal effect of exposure (psychiatric traits) on outcome (epilepsy), which could reduce conventional bias and avoid reverse causality because of these three basic assumptions (Davies & Holmes, 2018