Trait impulsivity in Juvenile Myoclonic Epilepsy

Abstract Objective Impulsivity is a multidimensional construct that can predispose to psychopathology. Meta‐analysis demonstrates an association between response impulsivity and Juvenile Myoclonic Epilepsy (JME), a common genetic generalized epilepsy. Here, we test the hypotheses that trait impulsivity is (i) elevated in JME compared to controls; (ii) moderated by specific seizure characteristics; and (iii) associated with psychiatric adverse effects of antiepileptic drugs (AEDs). Methods 322 participants with JME and 126 age and gender‐matched controls completed the Barratt’s Impulsiveness Scale (BIS‐brief) alongside information on seizure history and AED use. We compared group BIS‐brief scores and assessed associations of JME BIS‐brief scores with seizure characteristics and AED adverse effects. Results The mean BIS‐brief score in JME was 18.1 ± 4.4 compared with 16.2 ± 4.1 in controls (P = 0.0007). Elevated impulsivity was associated with male gender (P = 0.027), frequent absence seizures (P = 0.0004) and lack of morning predominance of myoclonus (P = 0.008). High impulsivity significantly increased the odds of a psychiatric adverse event on levetiracetam (P = 0.036), but not any other psychiatric or somatic adverse effects. Interpretation Trait impulsivity is elevated in JME and comparable to scores in personality and neurotic disorders. Increased seizure frequency and absence of circadian seizure pattern moderate BIS score, suggesting disruption of both cortico‐striatal and thalamocortical networks as a shared mechanism between seizures and impulsivity in JME. These findings warrant consideration of impulsivity as a distinct target of intervention, and as a stratifying factor for AED treatment in JME, and perhaps other types of epilepsy. The role of impulsivity in treatment adherence and psychosocial outcome requires further investigation.


Introduction
Impulsivity is a notable feature of Juvenile Myoclonic Epilepsy (JME), a common genetic epilepsy with complex inheritance. 1 JME is characterized by adolescent-onset, generalized seizures, occurring shortly after awakening, and often triggered by sleep fragmentation. Frontal lobe anatomy and function, thalamocortical drive and corticostriatal connectivity are altered in JME. [2][3][4][5][6] Early descriptions emphasized emotional instability and unreliability in patients, 7 later linked to Cluster B personality disorders. 8 We have recently shown reduced response inhibition, a dimension of impulsivity, in a meta-analysis of 1,047 JME patients 9 ; however, evidence for impairment in other ª 2020 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association dimensions of impulsivity, such as choice 5,10-14 and trait impulsivity, 10,12,13,15 are scarce or arise from small, uncontrolled studies. The characterization of these discrete dimensions of impulsivity (response, choice and trait) is necessary, as they each contribute to distinct psychiatric phenotypes. 16,17 Furthermore, the clinical utility of antiepileptic drugs (AEDs), such as levetiracetam, in JME is limited by psychiatric adverse effects, which highly impulsive individuals or those with comorbid psychiatric disorders may be predisposed to [18][19][20][21] .
Here, we address three issues with major clinical implications in a large, cross-sectional JME cohort. First, whether trait impulsivity is elevated in JME compared to age and gender-matched controls. If confirmed, this may justify the assessment and management of impulsivity as a distinct therapeutic target, and motivate the investigation of associations with treatment adherence 22,23 and long-term social prognosis. 24,25 Second, an exploration of moderators of impulsivity, where we test the hypothesis that trait impulsivity is positively associated with seizure characteristics. Moderators may indicate mechanisms and potential interventions for impulsivity in epilepsy. Third, we test the hypothesis that psychiatric, but not somatic, adverse effects of AEDs are associated with elevated trait impulsivity. These findings would assist clinical prediction models of adverse effects and motivate the search for genetic mediators in precision medicine. 26,27 Methods

Participants and data collection
We collected cross-sectional data as part of the Biology of Juvenile Myoclonic Epilepsy (BIOJUME) genome-wide association consortium study across 50 sites in 10 countries (appendix). Inclusion criteria were based on Avignon Class II consensus criteria for JME diagnosis 28  Research staff collected clinical data face-to-face in the form of a structured questionnaire, augmented by clinical records, EEG reports, and digital EEGs. The dataset included general demographics and health information, epilepsy history including seizure types, seizure frequency, drug/lifestyle interventions, and BIS-brief (Table 1).
Community control participants with no history of neurological or psychiatric diagnoses from London, UK were also recruited during the same period and completed the BIS-brief. Controls were group-matched to the JME cohort by age and gender.

Clinical data assurance
We uploaded JME clinical data onto a secure central database where a phenotyping panel, comprising seven epilepsy experts (CPB, KH, DKP, MPR, GR, MS, RT) confirmed the diagnosis of JME according to inclusion criteria. Disagreements were resolved through consensus.
Seizure types and frequency JME participants reported whether they had experienced absence or generalized tonic-clonic seizures (GTCS), as well as their pre-diagnostic and current frequency of myoclonic and absence seizures, categorized as at least daily, at least weekly, less than weekly or none. Participants also reported the number of GTCS experienced pre-diagnosis, whether their myoclonic seizures had a morning predominance and approximate ages of seizure onset and JME diagnosis.

Antiepileptic drug therapy
Participants reported their lifetime use of four common AEDs used to treat JME (valproate, lamotrigine, levetiracetam, and topiramate) and their experience of six adverse reactions associated with the same AEDs; (a) weight gain on valproate; (b) a skin rash requiring drug withdrawal on lamotrigine; (c) any psychiatric adverse event to levetiracetam; (d) weight loss on topiramate; (e) any psychiatric adverse event to topiramate; and (f) any cognitive adverse event to topiramate.

Barratt Impulsiveness Scale (BIS)
We used the BIS-brief to measure trait impulsivity in JME and control cohorts. The BIS-brief is a shortened eight-item version of the 30-question BIS-11 29 developed by Steinberg et al. 30 The original BIS-11 is well-  29,31 We chose the BIS-brief due to data available from normative and neuropsychiatric groups for comparison (see Table 2), in addition to ease of administration in a large-scale international study. Also, BIS-brief shows high test-retest reliability across both short and long timescales and concurrent validity with other impulsivity measures. 32 The maximum BIS-brief score (32) indicates high impulsivity and the minimum score (8) indicates low impulsivity. Of the 322 JME individuals who completed the BIS-brief, 92 Norwegian participants also answered the full BIS-11. The validity of the BIS-brief in JME was compared to the BIS-11, using a Spearman's correlation. There was a strong correlation between the two scores ( Fig. 1, r s = 0.85, P < 0.001).

Analysis procedure and statistical methods
We carried out statistical analysis on SPSS statistics (version 25) and produced graphics using GraphPad Prism (version 8.4.2). We first tested all statistical assumptions, and if necessary, transformed data as appropriate or used equivalent non-parametric tests. Individuals with missing/ unknown data for specific variables were excluded from each applicable analysis. We used the following tests: Chisquared test for categorical variables; Spearman's correlation to test associations between two ordinal variables; and Student's t-test or Mann-Whitney test for comparison of continuous variables between two groups. We calculated a mean BIS-brief score (AE standard deviation) for the JME cohort and compared with control and neuropsychiatric samples (described in Table 2) using a pairwise one-way Welch's analysis of variance (ANOVA) with Games-Howell correction (using Graph-Pad Prism). To compare with a broader range of clinical cohorts from the literature, we also calculated a mean BIS-11 score from the subset of 92 individuals who completed the full BIS-11 and observed its rank in context with other cohorts outlined in Table 2.
To test associations of clinical variables with BIS-brief score in JME, we first performed univariate analysis (correlations for continuous data and t-tests or ANOVAs, or non-parametric equivalents, for categorical variables) and calculated effect sizes. Hedge's g was used as sample sizes were unequal for all comparisons. We investigated confounding by gender in each univariate analysis by observing the change in regression coefficient when gender was added to a linear regression of BIS-brief score with each variable. The following variables were used in a multiple linear regression model of BIS-brief score: current frequency of absence seizures (daily, weekly, less than weekly, none); current frequency of myoclonic seizures (daily, weekly, less than weekly, none); gender (male, female); body mass index (BMI); morning predominance of myoclonus (yes, no); topiramate use (yes, no); psychiatric adverse reaction to levetiracetam (yes, no).
To investigate whether highly impulsive participants are at increased risk of a psychiatric adverse event to levetiracetam, we performed a logistic regression of current frequency of absence and myoclonic seizures, gender, BMI, morning predominance of myoclonus, and high impulsivity, against experiencing any psychiatric adverse event to levetiracetam. High impulsivity was defined as BIS-brief ≥ 21, interpreted from the classification of high impulsivity from Stanford et al. 31 where BIS-11 scores ≥ 72 are considered highly impulsive. Using results from a linear regression of BIS-11 and BIS-brief scores (regression line: BIS-11 = 24.2 + 2.3*BIS-brief, Fig. 1), a BIS-brief score of 21 is equivalent to a BIS-11 score of 72.
Ethics and funding BIOJUME is funded by the Canadian Institutes for Health Research (MOP-142405) and received ethical approval from the National Health Service (NHS) Health Research Authority (South Central -Oxford C Research Ethics Committee, reference 16/SC/0266) and the Research Ethics Board of the Hospital for Sick Children, Toronto (REB# 1000033784). Local ethical approvals were also held for all international sites. All procedures complied with appropriate regulatory requirements and ethical principles in line with the Declaration of Helsinki. Informed consent was obtained and documented for all participants. Assent was obtained from minors (under 16), and informed consent was obtained on their behalf by a parent or legally appropriate guardian. All clinical data from participants were deidentified before entry onto the central database.

Results
Demographics and clinical characteristics of JME cohort Eligibility and general demographics 322 JME participants completed the BIS-brief. General demographics and clinical information are presented in Table 1. The female bias, well-known in JME, was

Seizure characteristics
The median age of myoclonic seizure onset was 14.4 years (range 6-25), and the median age of JME diagnosis occurred around one year later at 16.1 years (range 6.9-40.5). Among those who reported other seizure types, the median age of GTCS onset was 15.4 years (range 5.7-37.8) and the median age of absence seizure onset was 6.1 years (range 3.0-11.3) in a subset of individuals who had Childhood Absence Epilepsy (CAE) evolving into JME (n = 22), and 14.2 years (range 6.9-37.1) in those who did not have CAE before a JME diagnosis (n = 131). Frequencies of seizure characteristics are presented in Table 1. We see a significant association between having daily myoclonic seizures and daily absence seizures (P = 5x10 -8 ).

Antiepileptic drug therapy
We found significant gender differences in AED therapy, with lower use of valproate and higher use of lamotrigine and levetiracetam in females (Table 1). This disparity between AED use in males and females is likely due to the revision of prescribing guidelines for valproate in females of childbearing age. 33 42% of individuals experienced at least one severe adverse drug event, with 37% experiencing a psychiatric event to either levetiracetam or topiramate, compared to 32.3% of individuals experiencing a somatic adverse event on valproate, lamotrigine, or topiramate. Experiencing a psychiatric adverse event to levetiracetam (37%) was most common, followed by weight loss on topiramate (36%) and weight gain on valproate (33%). More females reported weight gain on valproate than males (40% female vs. 24% male, P = 0.014).
To test clinical associations of trait impulsivity, we performed exploratory univariate analysis of demographic and clinical variables with BIS-brief scores. Effect sizes and p-values are shown in Table 3, with the largest effect size existing between individuals who are currently not having absence seizures (lower BIS-brief) compared to those having daily absence seizures (higher BIS-brief) (Hedges' g = 1, 95% CI: 0.56-1.44). There was no significant effect of age, gender, duration of JME, GTCS, AED use, adverse reactions to topiramate, or any adverse somatic event (Table 3).
Based on these results, univariately significant variables were entered into a multiple linear regression model. Morning predominance of myoclonus was also included in the model due to an increase in adjusted r 2 when entered (r 2 = 0.17 vs. 0.20, p = 0.008). Therefore, the final model included: gender, current frequency of myoclonic and absence seizures, BMI, topiramate use, morning predominance of myoclonus, and having a psychiatric adverse event to levetiracetam (results presented in Table 4). The model shows an increased frequency of absence seizures is associated with increased BIS-brief score (p = 0.0004), as is male gender (p = 0.027) and a lack of morning predominance of myoclonus (p = 0.008). Having daily absence seizures increases BIS-brief score by a mean of 4.1 points AE 1.1 (standard error). Despite the current frequency of myoclonic seizures being significantly associated with BIS-brief in univariate testing (Hedges' g = 0.72) and showing a clear relationship in Figure 3, it is not a significant predictor in the regression model, likely owing to the high correlation between frequency of myoclonic and absence seizures (P = 1x10 -8 ). However, these variables have low collinearity in the regression model (frequency of absence seizures, tolerance = 0.8, variance inflation factor (VIF) = 1.3; frequency of myoclonic seizures, tolerance = 0.8, VIF = 1.3).
Those who used topiramate showed some evidence of increased impulsivity in the linear regression model, and a small to medium effect size in univariate testing, as do those who experience a psychiatric adverse event to levetiracetam (Table 3 and 5). Figure 3 shows the distribution of BIS-brief scores across variables included in the regression model. Logistic regression (Table 5) showed that having a BIS-brief ≥ 21, i.e. highly impulsive, significantly

DISCUSSION
Trait impulsivity is significantly elevated in JME over matched controls by two points on the BIS-brief and to a score comparable to personality and neurotic disorders in the literature using BIS-11. 34,39,40 The extent of this psychological comorbidity in epilepsy has previously not been well-defined, and adds to known associations of impulsivity with personality disorders and externalizing psychiatric disorders. 9,22,23,30,32,42 Furthermore, our results substantiate evidence of elevated response and choice impulsivity in JME through experimental tasks, 5,10-14 suggesting a robust elevation in broad impulsivity. The association with seizure frequency and absent circadian pattern of seizures, points to possible mechanisms involving frontostriatal and thalamocortical networks, suggesting hypotheses for intervention. The increased odds of a psychiatric adverse event to levetiracetam in those with elevated impulsivity supports previous findings 19 and possibly reflects a genetic predisposition.

Clinically significant elevated impulsivity
Trait impulsivity is significantly higher in individuals with JME compared to matched controls, with a mean 2-point score difference and medium effect size. A BIS-11 score of 72 is a suggested cut-off for severity, 31 therefore, stratifying 24% of the current sample (27% if the BIS-brief equivalent of 21 is used). The clinical severity of this trait is also contextualized by comparison with other samples. BIS-brief scores in JME were higher than in cohorts at risk of psychopathology, 30,32 and BIS-11 scores were higher than PD 37,38 and OCD samples, 39 overlapped with anxiety and personality disorder samples 34,40 and lower than mood and schizophrenia spectrum disorder 35,40,41 and ADHD samples. 35,36,41 This extends findings from smaller, uncontrolled studies of trait impulsivity in JME using the BIS-11 12 and the Temperament and Character Inventory, 10,13 and provides robust evidence for associations across multiple dimensions of impulsivity, including response and choice impulsivity. 5,9,14 We recently reported a moderate and homogeneous effect size (d = 0.50, 95% CI 0.37-0.63) for response inhibition in a meta-analysis of 1047 JME patients. 9 Similarly, with choice impulsivity, JME patients, especially those with persistent seizures, have difficulty learning advantageous decisions in the Iowa Gambling Task (IGT), while seizure-free patients performed equal to controls. 5,14 Although our cross-sectional design does not allow inference about directionality, functional imaging studies support the interpretation that either seizures result in increased trait impulsivity or loss of striato-cortical inhibition influences both seizure susceptibility and impulse control. 3,4,14 A further limitation of the study design is the potential of ascertainment bias in both JME and control cohorts. The influence of an individual's impulsiveness on participating in a study such as this is unknown, however, this would likely influence both JME and control cohorts similarly. Conversely, potential ascertainment biases differing between JME and control cohorts may include educational level, socio-economic background, and ethnicity which were not controlled for. However, we aimed to address this limitation by using a range of control cohorts from previous literature alongside our own age and gender-matched controls, all of which have lower impulsivity than the JME cohort. Impulsivity influences treatment outcome through reduced treatment adherence in substance abuse 22 and gambling disorders. 23 Although we were unable to test the association here, adherence to lifestyle modifications, including limiting alcohol and maintaining a stable sleep routine, in addition to regular AED use, is essential to JME seizure management. 28 The link to social prognosis requires further investigation, but problem behaviors associated with impulsivity in the population, including substance abuse, violence and illegal activities such as shoplifting or reckless driving, 43 closely resembles findings of two population-based interview studies of psychosocial outcome in JME patients. 24,25 In both studies, there was no relationship between seizure and social outcome, which underlines the value of longitudinal investigations of impulsivity and associated psychiatric features as longterm risk factors.

Moderators of impulsivity
Seizure frequency is the strongest predictor of trait impulsivity, with effect sizes ranging from 0.72 to 1.00 for daily versus no current myoclonic or absence seizures, and a clear dose-response relationship. By analogy, the four-point rise in BIS-brief with daily absence seizures is equivalent to the difference between asymptomatic individuals' and psychiatric inpatients' mean scores. 32 However, seizure-free patients also had an elevated mean BIS-brief score compared to controls. Absence seizures demonstrate the strongest effect on impulsivity compared to myoclonic or GTC seizures, possibly reflecting additional disruption to attention and impulse control networks. 44 The effect size, though large, is possibly an underestimate due to notorious under-reporting of absence seizures. A previous study showed that active myoclonic seizures rather than active absence seizures were associated with an increased BIS-11 score in a genetic generalized epilepsy (GGE) cohort, 15 however, no interaction analysis between GGE subtype and the effect of active absence seizures on BIS-11 was performed therefore preventing a direct comparison to results from this study of JME. Other studies demonstrate that seizure frequency also moderates choice impulsivity, 5,14 but there were insufficient data to explore this in response impulsivity. 9 A circadian pattern or morning predominance of seizures is a touchstone of JME diagnosis and enshrined in consensus criteria. 28 Loss of this circadian pattern was associated with a small but significant effect size in trait impulsivity. Loss of morning predominance may be associated with unfavorable prognosis as it is more often found in JME patients with absence seizures or with worse overall seizure control, putatively because of disruption of circadian thalamocortical oscillators. 45 However, the association with trait impulsivity remains even after controlling for absences, suggesting residual confounding (unreported absences), heterogeneity or confounding by sleep. A poor sleep routine affects attention, executive function, and memory in healthy individuals 46 and those with high BIS-11 scores have phase-delayed sleep patterns, decreased total sleep time and efficiency and disrupted circadian function. 47 Since sleep fragmentation triggers seizures in JME, a poor sleep routine may confound this association, causing both increased seizure frequency and increased impulsivity. The extent to which impulsivity could be reduced by improved seizure control remains an open question. JME is usually managed by AEDs with broad efficacy for generalized seizure types, 21 but it is possible that complementing these with an absence seizure-specific AED (e.g. ethosuximide) may reduce impulsivity. However, complete seizure control is not a feasible goal for 20% of patients, 24 therefore warranting consideration of alternative approaches to reducing impulsivity, such as cognitive-behavioral therapy. 48 The complicated relationship between sleep and impulsivity in epilepsy may also yield testable hypotheses for intervention.
Elevated broad impulsivity in JME may partly be attributable to its adolescent-onset and partly to the pathophysiology of JME, which involves disruption to normal development and function of both frontostriatal and thalamocortical structures and respective connections. [2][3][4][5][6] There is convergent evidence for abnormalities in frontostriatal and thalamocortical networks in JME, drawn from neuroimaging and neuropsychological studies. 2,3,6,49 Taken together, there is considerable overlap in prefrontal, striatal, and limbic networks involved in both JME and impulse control. Other epilepsies may also be associated with raised impulsivity but are less studied. 8,44 None < Weekly Weekly Daily

Current myoclonic seizure frequency
None < Weekly Weekly Daily

Yes No
Psychiatric adverse effect on levetiracetam

Treatment outcome
Neuropsychiatric adverse events were reported in over one-third of participants and were more common than somatic adverse events. We observed a two-fold increase in the odds of a psychiatric adverse reaction to levetiracetam in highly impulsive individuals compared to less impulsive individuals. There was no association of impulsivity with any adverse event to topiramate, nor valproate or lamotrigine use, nor any somatic adverse effects to these AEDs. These data alone are not conclusive, partly due to absent baseline impulsivity data before levetiracetam exposure and the inability to exclude potential effects of current AED treatment on BIS scores, since information on lifetime therapy was provided rather than current therapy. However, other studies show similar findings. Helmstaedter et al. 19 showed that individuals with epilepsy and high BIS-11 scores were more likely to suffer from adverse psychiatric effects to levetiracetam. A model used to predict this adverse effect showed that individuals with a history of neuropsychiatric disorders, including depression, anxiety, personality disorder, or recreational drug use (all outcomes associated with impulsivity), were at increased risk. 20 Other studies show that individuals with epilepsy and a history of behavioral or psychiatric conditions were more likely to experience a psychiatric side effect across many AEDs, not only levetiracetam. 18,50 Additionally, we must consider a potential genetic predisposition to psychiatric adverse events to levetiracetam in those individuals who are more impulsive. 26,27 Dopamine receptor alleles are suggested in a candidate gene study, 26 but replication is necessary.
Finally, there is suggestion of increased impulsivity with topiramate use. However, exposure to topiramate, which is a second or third-line AED for JME, 21 may be a proxy for poor seizure control or drug-resistance, rather than representing a sign of chronic topiramate neurotoxicity.

Summary and conclusions
Multiple dimensions of raised impulsivity are associated with JME. The robust association of trait impulsivity with seizure frequency and loss of circadian pattern implies dysfunction in prefrontal, striatal, and limbic brain networks. Improved seizure control, perhaps complemented by psychotropic or cognitive-behavioral interventions, may be considered to mitigate this psychological comorbidity. Whether these findings also apply to other epilepsy syndromes, 8 as well as the implications of impulsivity for AED adherence and psychosocial prognosis in JME remain unexplored. We demonstrate that the BIS-brief, a simple measure of trait impulsivity, may have utility in predicting adverse psychiatric effects to levetiracetam. Genetic mediators of this risk would offer an advance in precision medicine. Overall, these findings demonstrate the importance of looking beyond diagnostic categories toward endophenotypes that better capture the components influencing patient-relevant outcomes and allow the development of better-targeted treatments.

Conflicts of Interests
DA, KKS, RT, and JZ report honoraria from UCB Pharma (manufacturer of levetiracetam) and RT reports honoraria from Sanofi (manufacturer of sodium valproate). All other authors report no conflicts of interest.