Association between epilepsy and challenging behaviour in adults with intellectual disabilities: systematic review and meta-analysis

Background Previous systematic reviews showed no significant association between epilepsy and challenging behaviours in adults with intellectual disabilities. Aims To identify whether there is an association between epilepsy and challenging behaviour in adults with intellectual disabilities by carrying out a systematic review of published data. PROSPERO registration number: CRD42020178092. Method We searched five databases and hand-searched six journals. Two authors independently screened titles, abstracts and full articles using a standardised eligibility checklist. Several meta-analyses were carried out. Results The narrative analysis of data from 34 included articles (14 168 adults with intellectual disabilities, 4781 of whom also had epilepsy) showed no significant association between epilepsy and challenging behaviour. Meta-analysis was possible on data from 16 controlled studies. This showed no significant intergroup difference but after sensitivity analysis meta-analysis of 10 studies showed a significantly higher rate of overall challenging behaviour in the epilepsy group (effect size: 0.16) compared with the non-epilepsy group. Aggression and self-injurious behaviour both showed a statistically significant higher rate in the epilepsy group, with very small effect sizes (0.16 and 0.28 respectively). No significant intergroup difference was observed in the rate of stereotypy. Conclusions The findings are contradictory and must be interpreted with caution because of the difficulty in pooling data from varied studies, which is likely to introduce confounding. Where significant differences were found, effect sizes are small and may not be clinically significant, and there are major methodological flaws in the included studies, which should be addressed in future large-scale properly controlled studies.

Epilepsy is common in adults with intellectual disabilities, with an average estimated point prevalence of 25%, compared with <1% in the general population who do not have intellectual disabilities. 1 The rate increases with the severity of disability: 2 around 7-15% in mild to moderate intellectual disability, 67% in severe intellectual disability and 82% in profound intellectual disability. The rate increases if the intellectual disability is associated with other neurological disorders, such as cerebral palsy. 1 Special issues relating to epilepsy in adults with intellectual disabilities Certain genetic syndromes that lead to intellectual disabilities, such as Angelman, Sturge-Weber, fragile-X/ataxia, Rett, Lesch-Nyhan, Rubinstein-Taybi, Lowe and Down syndromes and tuberous sclerosis, are commonly associated with epilepsy. 3 A high proportion of people with autism spectrum disorder (22%) also have epilepsy. 1 Similarly, certain epilepsy syndromes, such as West syndrome (in infants), Lennox-Gastaut syndrome, Landau-Kleffner syndrome and Dravet syndrome, are more commonly associated with intellectual disabilities. 4 Compared with the general population of adults who do not have intellectual disabilities, epilepsy among adults with intellectual disabilities is not only more prevalent, but it also often manifests as multiple seizure types, starts at an early age, is of longer duration and is resistant to anti-epileptic treatment (in over 30% in the general population, compared with over 70% in intellectual disabilities). 5 Diagnosing epilepsy and seizure type can be difficult in this population, and both false-positive (stereotypy, cardiac syncope, non-epileptic attack disorder may all mimic epileptic seizure) and false-negative (difficulty diagnosing absence, focal seizures) diagnoses are possible. 6 Also, these people are more prone to die from sudden unexpected death in epilepsy (SUDEP). 7,8 Relationship between epilepsy and challenging behaviour in adults with intellectual disabilities The relationship between epilepsy and challenging behaviour in adults with intellectual disabilities is complex. 9 Challenging behaviour has been defined as 'socially unacceptable behaviour that causes distress, harm or disadvantage to the persons themselves or to other people, and usually requires some intervention'. 10 Challenging behaviour is prevalent among adults with intellectual disabilities, affecting up to around 62%. [11][12][13][14] More severe forms of challenging behaviour are manifested by a lower proportion (18.7-30%). 13,15 The types of challenging behaviour include aggression, destruction of property, disruptive behaviour, self-injurious behaviour, stereotypy, and sexually inappropriate and harmful behaviours. 14,16 Aggression is reported in 10-20% of adults with intellectual disabilities. 16,17 The aetiology of challenging behaviour is multifactorial, including medical, psychiatric, psychological, social and environmental factors. Therefore, a multiprofessional approach is required to formulate a person-centred treatment plan for this difficult-to-manage problem. 16 Demographic factors such as age, gender, severity of intellectual disability, associated comorbidities (e.g. other neurodevelopmental disorders such as autism spectrum disorder, attention-deficit hyperactivity disorder), medical conditions and psychosocial factors can all affect challenging behaviour in adults with intellectual disabilities. 14,18 Epilepsy is one such factor that may influence a person's behaviour.
Previous systematic reviews on the subject Three systematic reviews looked at the association between challenging behaviour and epilepsy in people with intellectual disabilities: none of these found any association. [19][20][21] We decided to carry out an updated systematic review, as important publications either have appeared since the last reviews or were not included in those reviews. Another reason for carrying out this review is to conduct meta-analyses that were not done in any of the previous reviews.

Method
The aim of the current systematic review and meta-analysis is to identify the rates and types of challenging behaviour in adults with intellectual disabilities who have epilepsy ('the epilepsy group') and compare them with rates and types in adults with intellectual disabilities who do not have epilepsy ('the non-epilepsy group') to determine whether epilepsy is a risk factor for developing challenging behaviour in this population.
We aimed to include studies that compared the overall rate of challenging behaviour as well as different types of challenging behaviour in adults with intellectual disabilities with and without epilepsy within the same cohort. We did not aim to include any study that involved only adults with intellectual disabilities who did not have epilepsy. We included studies that involved participants without epilepsy only where they were part of the control group and participants with epilepsy were also involved in the same study.
We also aimed to include studies that involved adults with intellectual disabilities and epilepsy but no one without epilepsy. These studies are included to assess the rate of different types of challenging behaviour according to different seizure variables.

Search strategy
We decided on our final search strategy after an initial scoping literature search. We followed PROSPERO guidelines 22 and the PRISMA-P checklist to develop our protocol and search strategy. 23 Five electronic databases -Embase, PubMed/MEDLINE, PsycInfo, DARE and ASSIA (ProQuest)were searched for relevant journal articles. Each database was searched between 1 January 1985 and 31 May 2020. In addition, we also cross-referenced pertinent reviews and articles. We hand-searched for relevant articles in the past 20 years' issues, from January 1990 to May 2020, in the following intellectual disabilities journals: Only quantitative studies in English were searched. We excluded non-human studies, studies involving children and conference abstracts.

Search terms
The list of search terms used can be found in the supplementary material available at https://doi.org/10.1192/bjo.2020.96. Search terms were adapted from previous systematic reviews that were carried out to develop a national and an international guide for the use of psychotropic medications in the management of challenging behaviour in adults with intellectual disabilities. 10,24 Criteria for selecting studies We devised a list of eligibility criteria based on PROSPERO 22 and Cochrane review guidelines 25 and adapted from similar systematic reviews on psychotropic medications [26][27][28] (supplementary Appendix 1).

Types of study
There was no restriction on the type of study design included in this review. We included both randomised and non-randomised studies; controlled and non-controlled observational or cross-sectional studies; and controlled studies with both matched and nonmatched control groups.

Types of participant
All participants had intellectual disabilities, were aged 16 years or over and displayed various types of challenging behaviour. We have only included data on challenging behaviour in this review. Data on psychiatric illness without challenging behaviour are not presented in the current paper as they will be included in a separate systematic review article.
No ethical approval was required for this study as no individual patient-related data were collected or analysed.

Sample size
Studies with fewer than 10 participants were excluded. This arbitrary cut-off was used in accordance with our previous systematic reviews. 10

Secondary outcome
The rates of different types of challenging behaviour (verbal aggression and aggression towards people or property or self, stereotypy, overactivity, temper tantrum, etc.) were compared between the epilepsy and the non-epilepsy groups. Data on subgroup comparisons according to different types of seizure, seizure frequency and different pharmacological regimes (e.g. polypharmacy versus monopharmacy of anti-epileptic medications; treatment with carbamazepine versus valproate) were collected to identify the role of different epilepsy-related factors in the development of challenging behaviour.

Selection process
After the definitive search was completed, titles were searched for key terms. Zotero reference management software 29 was used to manage and record references from each database. Duplicates, non-human studies and studies involving children were identified by Zotero and removed manually (by B.A.B.). The remaining abstracts were screened independently using the pre-piloted eligibility criteria (by B.A.B. and B.L.). The two authors were masked to each other's scores. Bibliographies of potential studies were screened to identify articles that required acquisition of the full text. Discrepancies identified were reviewed and discussed between B.A.B. and B.L. to resolve any differences. The full texts were then reviewed and independently assessed for eligibility by B.A.B. and B.L. using the same eligibility checklist that was used for screening the abstracts. The selection process is shown in Fig. 1. It was not necessary for the third review author (S.D.) to arbitrate.

Data extraction
Data from studies meeting eligibility criteria were extracted independently by B.A.B. and B.L. using a standardised data extraction template adapted from Cochrane review guidelines (supplementary Appendix 2). 30 S.D. checked this information for accuracy. Results are reported using a narrative synthesis of information from included studies.

Meta-analysis
The meta-analysis was carried out only on controlled studies. An odds ratio was calculated for the studies that presented the proportion of participants in each group displaying challenging behaviour. A standardised mean difference was calculated for those studies that presented means and standard deviations for scores based on a behaviour rating scale in the two groups. 31 To pool data, we logtransformed all data to standardised mean differences. 31 RevMan 5.3 32 software for Windows 10 was used for random-effects metaanalysis. Heterogeneity was tested using the χ2-test and I 2 -statistic. Where there was substantial heterogeneity (I 2 > 60%), a further sensitivity analysis was carried out. A final meta-analysis was carried out after removing data from the studies that produced a high heterogeneity and strong bias, to bring heterogeneity to an acceptable level (I 2 < 30).

Risk-of-bias assessment and confidence in cumulative estimates
The risk of bias for the 19 controlled studies identified was assessed independently by B.A.B. and B.L. using the Cochrane risk-of-bias tool 33 and the quality of all 32 eligible studies was assessed during the data collection process using the SIGN 50 checklist. 34 Publication bias was assessed using a funnel plot, and the studies included were assessed for consistency and precision. The quality of the overall systematic review was assessed using AMSTAR 2 criteria (see Appendix 3; supplementary material). 35

Included studies
Of the 868 articles screened, 34 papers (from 32 studies) met the eligibility criteria and were included in the systematic review (Fig. 1). A list of excluded articles with reason for exclusion is presented in supplementary Appendix 4. Two studies each published two papers on the same sample but on different outcome measures. Of the 34 papers, 19 included participants with and without epilepsy, as their authors compared the rates of challenging behaviour in these two groups to assess an association between challenging behaviour and epilepsy (Table 1). Of these, nine studies [36][37][38][39][40][41][42][43][44] had equal numbers of participants in both groups and the groups were matched on various demographic variables. The rest (n = 10) 12,45-53 were prevalence studies of challenging behaviour in adults with intellectual disabilities that included a number of participants with epilepsy (around 22% of the cohort). Different variables, such as age, gender and presence/absence of epilepsy, were used to assess whether they are risk factors for developing challenging behaviour. The rates of challenging behaviour were compared on the basis of the presence or absence of epilepsy, but the epilepsy groups were not matched with the non-epilepsy groups. The rest of the studies included participants with epilepsy only, and in these the rates of challenging behaviour were compared between various types of epilepsy (e.g. frequent versus infrequent; generalised Epilepsy and challenging behaviour in adults with intellectual disabilities  Table 2 compares the rates of different types of challenging behaviour (e.g. aggression, self-injurious behaviour) between the epilepsy and the non-epilepsy groups. Table 3 presents data on participants with epilepsy only and compares rates of challenging behaviour according to various epilepsy variables.
Most studies were carried out in the UK (n = 24), some in the USA (n = 4) and one each in Ireland, Sweden, The Netherlands and Spain. In total, data on 14 168 adults with intellectual disabilities are presented (4781 with epilepsy and 9387 without epilepsy).

Diagnosis
Challenging behaviour is defined using a variety of methods in different studies. Five studies 49,54-57 collected data retrospectively from participants' case notes and did not use any validated measures. One study 47 collected information on challenging behaviour from parents' and carers' reports and one study 51 65 Two studies 43,53 used the Aberrant Behaviour Checklist-Community version (ABC-C) total score, 66 and another four 42,46,52,67 used ABC-C subdomain scores. 68 Three studies 37,38,69 used the Psychosocial Behaviour Scale (PBS). 70 The Behaviour Problem Inventory (BPI) 71 was used in two studies 12,72 to rate challenging behaviour. One study each used the Diagnostic Criteria-Learning Disability (DC-LD), [73][74][75] Autism Spectrum Disorders-Comorbidity-Adult version (ASD-CA) 44,76 and Autism Spectrum Disorders-Behaviour Problems for Adults (ASD-BPA) 77,78 respectively.

Statistical methods used
Where a standardised scale was used to measure challenging behaviours, some studies compared means and standard deviations for the two groups, whereas others used an arbitrary cut-off score to compare participants in the two groups. 79 Various statistical methods were used, including χ 2 , regression analysis, univariate and multivariate regression analyses, the Wilcoxon matched-pairs signed-ranks test and Student's t-test.

Outcome (narrative synthesis)
The overall rate of challenging behaviour Of the total 19 controlled studies, 13 12,36,37,39,40,43-45,47,50-53 did not show any significant intergroup difference in the overall rate of challenging behaviour, three 38,42,48 showed a significantly higher rate of challenging behaviour in the epilepsy group and three 41,46,49 showed a higher overall rate of challenging behaviour in the non-epilepsy group (Table 1). Of these three, one 41 was at a significant level and the level of significance for other two 46,49 is not known.
Rates of different types of challenging behaviour Table 2 shows the rates of different types of challenging behaviour in the epilepsy and the non-epilepsy groups. For aggression, nine studies 12,13,37,48,49,53,54,74,77 showed no significant intergroup difference. According to two studies 38,59 the epilepsy group showed a statistically significant higher rate of aggression compared with the non-epilepsy group. In one study 67   Epilepsy and challenging behaviour in adults with intellectual disabilities  37 showed significantly less self-injurious behaviour in the epilepsy group compared with the non-epilepsy group. For aggression to property or destructiveness, three studies 13,37,48 showed no significant intergroup difference and no study showed a higher rate in the epilepsy group compared with the non-epilepsy group. One study 48 showed a significantly higher rate of behaviours such as disturbing others at night, seeking attention and being uncooperative in the epilepsy group compared with the non-epilepsy group. Rates of behaviours such as stereotypy, inappropriate sexual behaviour, irritability, hyperactivity, verbal aggression, antisocial behaviours and lethargy. were reported in a very small number of studies showing equivocal findings.
Association between challenging behaviour and epilepsy-related variables showed a higher rate of challenging behaviour among those who presented with frequent seizures as opposed to those who had less frequent seizures. However, in five studies 37,47,54,55,58 no such significant intergroup difference was found. In one study 61 those who showed generalised epileptiform changes on electroencephalograms (EEGs) showed a significantly higher rate of challenging behaviour compared with those whose EEGs showed focal epileptiform changes.

Association between challenging behaviour and anti-epileptic medication-related variables
Three studies 37,61,69 showed a significantly higher rate of challenging behaviour among those who received multiple anti-epileptic medications (polypharmacy group) compared with those who received single anti-epileptic medication (monopharmacy group). However, no such significant intergroup difference was found in five studies. 37,47,53,54,56 Interestingly, two studies 54,72 reported a higher rate of challenging behaviour in the monopharmacy group than in the polypharmacy group, and in one of these studies this intergroup difference was significant. 72 Only one study reported the rate of challenging behaviour in various monopharmacy groups (phenytoin, sodium valproate, carbamazepine and lamotrigine monopharmacy) but it made no intergroup comparison. 56

Meta-analysis
Of the 19 controlled studies, data were available for meta-analysis from 16. Pooled data from the 16 studies using a random-effects meta-analysis of standardised mean differences showed no significant intergroup difference but the heterogeneity among studies was very high (I 2 = 88%) (Fig. 2). After sensitivity analysis we removed data from studies that produced the highest level of heterogeneity and had a high risk of bias according to the Cochrane riskof-bias tool. The final meta-analysis, using a random-effects model, of pooled data from the remaining 10 studies showed a statistically significant higher rate of overall challenging behaviour in the epilepsy group compared with the non-epilepsy group, with a very small effect size of 0.16. The heterogeneity among studies came down to an acceptable level (I 2 = 18%) (Fig. 3). To study specific types of challenging behaviour, we conducted meta-analyses of pooled data on aggression scores from nine studies (Fig. 4), on self-injurious behaviour from six (Fig. 5) and on stereotypy from three (Fig. 6). The aggression meta-analysis showed a significantly higher rate in the epilepsy group compared with the non-epilepsy Blickwedel    Epilepsy and challenging behaviour in adults with intellectual disabilities group, with a very small effect size of 0.16. The heterogeneity level was low (I 2 = 27%). The self-injurious behaviour meta-analysis also showed a statistically significant higher rate in the epilepsy group compared with the non-epilepsy group, with a very small effect size of 0.28, but the heterogeneity score was high, albeit below 60% (I 2 = 54%). The stereotypy meta-analysis did not show any significant intergroup difference but showed a high heterogeneity value of over 60% (I 2 = 69%).

Quality control
The AMSTAR2 checklist showed a high score, indicating that this systematic review and meta-analysis is of a high standard (supplementary Appendix 3). The SIGN 50 checklist identified only 5 of the 32 studies to be of high quality and the Cochrane risk-of-bias assessment of the 19 controlled studies showed a high risk of bias in most domains for most of the studies (Fig. 7). A summary graph is presented as supplementary Appendix 5. Publication bias could not be identified using a funnel plot for stereotypy, as data from only three studies could be amalgamated into the meta-analysis. Funnel plot data showed no publication bias for aggression, which was further supported by Egger's test of publication bias (P = 0.213). 80 Although the funnel plot showed some publication bias for overall rate of challenging behaviour, this was not significant under Egger's test of publication bias (P = 0.734). There was no publication bias present for self-injurious behaviour (P = 0.307). Eleven articles reported receiving funding from external sources, one did not receive any funding and the rest (n = 22) did not declare the funding source.

Discussion
We included 34 articles in our systematic review that met the eligibility criteria. These included 9 studies that compared the overall rate of challenging behaviour in an epilepsy group with a matched control group and another 10 that compared data with an unmatched control group. Compared with previous systematic reviews this review included data for a much higher number of participants (Table 4). In our review, of the total 19 controlled studies, 13 12,36,37,39,40,[43][44][45]47,[50][51][52][53] did not show any significant intergroup difference in the overall rate of challenging behaviour, 3 38,42,48 showed a significantly higher rate of challenging behaviour in the epilepsy group and 3 41,46,49 showed a higher rate in the non-epilepsy group. Of these three, one 41 was at a significant level and the level of significance for the other two 46,49 is not known.
It was possible to pool data from only 16 of the controlled studies for a meta-analysis. Meta-analysis of pooled data from these 16 studies did not show a significant intergroup difference in the overall rate of challenging behaviour but showed a high heterogeneity. However, after sensitivity analysis meta-analysis of pooled data from 10 studies showed a significantly higher rate of overall challenging behaviour in the epilepsy group compared with the non-epilepsy group. Meta-analysis of pooled data showed a statistically significant higher rate of aggression and self-injurious behaviour in the epilepsy group compared with the non-epilepsy group. However, no such intergroup difference emerged from the meta-analysis of pooled stereotype data.
A funnel plot and Egger's test of publication bias showed no publication bias among the included studies. According to both SIGN 50 and Cochrane risk-of-bias assessments most studies appeared to be of moderate to poor quality.

Interpretation of meta-analyses findings
Our finding (based on meta-analysis of pooled data from 10 studies) of a significant intergroup difference differs from that of other systematic reviews, [19][20][21] which found no such difference. Our finding has to be interpreted with caution. First, when data from all available studies were pooled, no statistically significant intergroup difference emerged, although the heterogeneity among studies was very high. Second, even when the meta-analysis after the sensitivity analysis showed that the epilepsy group had a significantly higher rate of challenging behaviour, the effect size remained very small, so this may not be clinically significant. Third, different studies defined challenging behaviour in different ways. Some did not use any validated tool. Fourth, even when a validated scale was used, the total score was used to define challenging behaviour, which is not always valid. For example, a number of studies used the ABC-C total score, which is not valid. 66 Fifth, many studies used an arbitrary cut-off score on behaviour rating scales to define challenging behaviour, and different studies used different scales and different cut-off scores. Thus, it is difficult to compare data among studies as it is difficult to know when data were pooled and whether all studies are describing the same challenging behaviour.
Deb & Hunter 61 hypothesised that it is possible that underlying brain damage (in adults with severe and profound   Epilepsy and challenging behaviour in adults with intellectual disabilities intellectual disabilities) and psychosocial factors (in those with mild intellectual disabilities) are stronger determinants of challenging behaviour than the presence of epilepsy. Factors that affect behaviour in the presence of epilepsy are related to: (a) underlying brain damage, such as the location and severity of any deformity, tumour or abnormal electrical discharge in the brain; (b) epilepsy-related factors, such as the presence of certain epileptic syndromes and genetic syndromes that are prone to lead to more challenging behaviour; (c) seizure-related factors, such as the severity, type and frequency of seizures; (d) anti-epileptic medication-related factors, such as the adverse effects of certain anti-epileptics and drug-drug interactions; and (e) psychosocial factors, such as loss of occupation, financial problems, lack of support, and locus of control being outside the person so the person does not have any control over the timing of seizures.

Types of challenging behaviour
Many types of challenging behaviour were assessed in the included studies. The most common types that cause most concern and are most difficult to manage are aggression (verbal or physical aggression towards other people or property) and self-injurious behaviour. Pooled data showed a statistically significant higher rate of both aggression and self-injurious behaviour in the epilepsy group compared with the non-epilepsy group. It is worth remembering that, apart from epilepsy, many other medical and psychosocial factors influence these behaviours, including certain genetic syndromes that are known to be associated with aggression and self-injurious behaviour. 81 Many of these syndromes also tend to predispose individuals to epilepsy (examples include Lesch-Nyhan and fragile-X syndromes). 1 Therefore, it is difficult to draw any conclusion about the association between these specific challenging behaviours and epilepsy in isolation without considering all the other predisposing (e.g. genetic syndromes), precipitating (e.g. infection or anxiety) and perpetuating (e.g. inappropriate treatment and/or environment) factors for challenging behaviour in general. Pooled data did not show any significant intergroup difference in the rate of stereotyped behaviour. Only a small number of studies were involved in this meta-analysis and the heterogeneity level was high. Therefore, this finding has to be interpreted with caution.
The rates of other behaviours, such as inappropriate sexual behaviour, irritability, hyperactivity, verbal aggression, antisocial behaviours and lethargy, were reported in a very small number of studies showing equivocal findings. Therefore, it is difficult to draw any definitive conclusion about the association between these specific types of challenging behaviour and epilepsy. Another problem is that none of these specific behavioural types was defined using any standardised criteria (such as the Modified Overt Aggression Scale for rating aggression) 15,82 but were based on single-item scoring.

Total (95% Cl)
Test for overall effect: Z = 0.37 (P = 0.71) Heterogeneity: Tau 2 = 0.05; Chi 2 = 6.46, df = 2 (P = 0.04 ); l 2 = 69% 100.0% 0.06 (-0. 25    The narrative analysis of data from 34 included articles showed no significant association between epilepsy and challenging behaviour. No definite association was found in the rate of challenging behaviour and different epilepsy variables, such as frequency of seizures, generalised v. focal epilepsy, polypharmacy of anti-epileptic drugs (narrative analysis only). Meta-analysis was possible on pooled data from only 16 controlled studies. This showed no significant intergroup difference but, after sensitivity analysis, pooled data from 10 studies showed a significantly higher rate of overall challenging behaviour in the epilepsy group (effect size: 0.16) compared with the non-epilepsy group. Aggression and self-injurious behaviour both showed a statistically significant higher rate in the epilepsy group, with a very small effect size (0.16 and 0.28 respectively). No significant intergroup difference was observed in the rate of stereotypy Epilepsy and challenging behaviour in adults with intellectual disabilities

Association with epilepsy variables
When different subgroups according to various epilepsy variables were compared for the rate of challenging behaviour no clear picture emerged. In three studies 54,61,72 the rate of challenging behaviour was significantly higher in those who had generalised seizures compared with those who had focal seizures, but in one study 67 no significant difference was reported between these two groups. Given the small number of studies involving small numbers of participants in the subgroups, lack of matching of the groups, and different types of challenging behaviour rated in these studies, it is difficult to draw any definitive conclusion about any association between challenging behaviour and seizure type. Furthermore, Deb 62 has shown that a high proportion of adults with intellectual disabilities who had a clinical diagnosis of primary generalised seizure showed focal epileptiform changes in their EEGs, thus raising the possibility that in many cases these generalised seizures are secondarily generalised from focal seizures. Although three studies 38,61,72 showed a significantly higher rate of challenging behaviour among those who had frequent seizures compared with those who had less frequent seizures, five studies 37,47,54,55,58 did not find any significant intergroup difference. Therefore, it is difficult to draw any definitive conclusion about the influence of seizure frequency on the rate of challenging behaviour in this population. One confounder is the way frequency was rated in different studies, which varied widely. One study 61 showed a significantly higher rate of challenging behaviour in those whose EEGs showed generalised epileptiform activities compared with those whose EEGs showed focal epileptiform changes. However, in another three studies 53,62,67 no significant intergroup difference emerged between those with generalised as opposed to focal EEG changes. It is difficult to carry out an EEG for many adults with intellectual disabilities, particularly those who have severe and profound disability and also those who have challenging behaviour. EEG records are available for approximately 50-70% of adults with intellectual disabilities and 70-90% of these recordings are abnormal, although the abnormalities are not necessarily epileptiform in nature; rather, they present mostly as non-specific excess slow background activity. 62 Association with anti-epileptic medication No definite association was found in the rate of challenging behaviour and polypharmacy with anti-epileptic medications. One would expect the participants in the polypharmacy group to have more severe epilepsy and, therefore, possibly more challenging behaviour. However, it is possible that anti-epileptic polypharmacy made these participants more sedated, thus dampening down the expression of challenging behaviour. In some people anti-epileptics improve both epilepsy symptoms and behaviour. However, in others it can have an opposite effect, in that although the epilepsy improves, the behaviour deteriorates. 9 There may be many explanations for this paradoxical response. Old theories, such as forced normalisation 83 or alternative psychosis, 84 may provide some explanation. However, a more practical explanation may be that medication side-effects make the behaviour worse in some, despite improving epilepsy symptoms.
Although certain anti-epileptics, such as sodium valproate (restricted use in women of child-bearing age because of major worry about its teratogenicity), carbamazepine and lamotrigine, are known to improve mental state and are used to treat psychiatric disorders such as bipolar disorder, 85 paradoxically some anti-epileptics are known to precipitate psychopathology, including challenging behaviour. 86 Although the evidence is not strong, the available data suggest that the following anti-epileptic medications are likely to have some association with aggression and other challenging behaviours: phenobarbital, topiramate, vigabatrin, perampanel, zonisamide, levetiracetam, clobazam, clonazepam and tiagabine. Among these perhaps levetiracetam (aggression or agitation in 13% of treated patients), perampanel (in 12% at 8 mg/day and 20% at 12 mg/day) and possibly topiramate (in 2-10%) showed the strongest evidence for precipitating challenging behaviour. 86 However, levetiracetam seems to improve behaviour in some adults with intellectual disabilities and worsen it in others. 87 These anti-epileptic-related adverse effects may be more pronounced among adults with intellectual disabilities. 5 However, it is clear that vast majority of those who receive these medications do not show any challenging behaviour. It is difficult to draw any definite conclusion from this review, as only one study reported the rate of challenging behaviour related to specific anti-epileptic drugs (59% of participants on carbamazepine, 55% on sodium valproate, 53% on phenytoin and 78% on lamotrigine monopharmacy showed challenging behaviour).
Although monopharmacy with anti-epileptic medication is desirable, 88 polypharmacy with anti-epileptics is common in intellectual disability populations. Therefore, anti-epileptic drug-drug interaction are more likely, some of which may lead to challenging behaviour. Given that both antipsychotic and antidepressant medications are commonly prescribed among adults with intellectual disabilities, 89 their interaction with anti-epileptics must be considered in any assessment of challenging behaviour. Also, both antipsychotics and antidepressants are likely to lower seizure threshold (particularly the older generation ones and at a high dose), which may precipitate more seizures and may lead to challenging behaviour.
Subgroup comparisons do not provide adequate power to detect clinically significant difference because of the small numbers involved in each subgroup and the lack of a control group. Also, in the subgroups there is no consistency in the types of challenging behaviour described, as some studies provided the rate of overall challenging behaviour, but others reported the rates of different types of challenging behaviour, such as aggression, self-injurious behaviour and stereotypy, making it difficult to amalgamate data from different studies. It will be necessary to conduct a much larger randomised controlled trial to recruit a reasonable number of participants in each subgroup to provide adequate power to detect clinically significant intergroup differences.

Clinical significance of the findings
Having epilepsy can restrict social activities and wider social integration. A careful risk assessment is necessary to balance independence/quality of life and seizure-related risks (e.g. travelling alone, taking a bath, seizure-related injuries, unpredictability of the timing of seizures, SUDEP). The risk assessment should be part of the person's overall person-centred support plan and should be monitored and reviewed regularly with the person, their family/ caregivers, other relevant professionals and the multidisciplinary team. One has to remember to mitigate against the impact on the family/caregivers. It is also important to remember that, apart from epilepsy, many circumstances, such as medical, psychological, social and environmental factors, affect the behaviour of someone with intellectual disabilities, and a full multidisciplinary personcentred assessment is required to develop an appropriate formulation for the management of challenging behaviour, including psychosocial interventions. 10 Support staff, and the person and their family/carers, need to be informed of the risk factors (including SUDEP) and prognosis. [90][91][92] This also highlights the requirement for regular health checks for all adults with intellectual disabilities, as highlighted in a recent NHS England publication. 93

Strengths
Our review received a high rating on the AMSTAR 2 quality control checklist for systematic reviews, as we complied with all of its requirements (supplementary Appendix 3). We carried out metaanalyses that were not done in any previous systematic reviews. We included a comprehensive Cochrane risk-of-bias table, which was not done by any of the previous systematic reviews. We included a much higher number of articles covering data from a much larger number of participants compared with the previous systematic reviews. We have registered our review with a well-established database, PROSPERO, so that our protocol is available for public scrutiny. This was not done by the previous reviews. We also carried out a very extensive hand-search of journals in the field of intellectual disability and epilepsy, along with stringent cross-referencing.

Weaknesses
We searched for articles in English only. We excluded the grey literature and conference abstracts, as we felt it would be difficult to apply our eligibility criteria and risk-of-bias assessment on the basis of abstracts only. Although we used a stringent method for literature search, it is still possible that we missed some relevant articles. Our analysis showed conflicting evidence, in that the metaanalysis of pooled data from a larger number of studies did not show a significant intergroup difference, whereas pooled data from a smaller number of studies after sensitivity analysis showed a significant difference. Although we amalgamated data where possible to carry out a number of meta-analyses, the heterogeneity among studies remains high. It is difficult to amalgamate data from studies that used such diverse methodologies and defined challenging behaviour in so many different ways. Therefore, our findings must be interpreted with caution, as a lot of confounders could not be controlled for. However, to counteract the problem with study heterogeneity we used sensitivity analyses. Ideally, we should have used raw data for the meta-analysis, but this was not possible. Also, by log-transforming some data we may have lost some power in the meta-analysis.

Research implications
Even though the meta-analysis of pooled data from a smaller number of studies after sensitivity analysis showed a significantly higher rate of challenging behaviour in the epilepsy group, the effect sizes are small, which may not be clinically significant. Also, there are major methodological flaws (highlighted by Cochrane risk-of-bias and SIGN 50 assessments) in the included studies. There is therefore a need for large-scale properly controlled studies. The included studies primarily concentrated on inter-ictal challenging behaviour. However, peri-ictally some people may show aggression, which is not goal directed but inadvertently may injure others. This might also be studied in further research.

Data availability
Data availability is not applicable to this article as no new data were created or analysed in this study.