ADHD as a Risk Factor for Incident Unprovoked Seizures and Epilepsyin Children FREE

The relationship between epilepsy and attention-deficit/hyperactivitydisorder (ADHD) has been poorly described. Clinically, there is a perceptionthat ADHD is more common among children with epilepsy, due to the epilepsyor its treatment.^1- 2 In studiesof children with prevalent epilepsy, 28.1% to 39% had "hyperactivity-impulsivity,"^3- 4 42.4% had problems with attention,⁵ and 13.9% had ADHD.⁶ Thesecross-sectional studies cannot examine the temporal relationship between ADHDand epilepsy, and standardized symptom profiles were seldom used to make diagnoses.

The largest studies^3- 4 ofthe association between hyperactivity-impulsivity and prevalent epilepsy arisefrom the 1988 National Health Interview Survey. Hyperactivity assessed bythe Behavior Problem Index was 5.7-fold more prevalent among 121 childrenwith epilepsy (28.1%), aged 5 to 17 years, compared with 3950 control subjects(4.9%).⁴ "Highly impulsive behavior," accordingto 4 questions, occurred among 39% of 118 children with a history of epilepsy,aged 6 to 17 years, compared with 11% of 11 042 children without a historyof epilepsy.³

In 2 prior studies^7- 8 ofchildren with incident unprovoked seizure, behavioral disturbances beforethe onset of first seizure were more frequent than among controls. In thelarger study of 148 children with first unprovoked seizure and 89 seizure-freesibling controls, attention problems as assessed by the Child Behavior Checklistwere 2.4-fold more common before identification of the first seizure (8.1%)than in controls (3.4%).⁷

We undertook a population-based case-control study of newly diagnosedunprovoked seizures among Icelandic children younger than 16 years to addresswhether ADHD is associated with an increased risk for developing unprovokedseizures. An Icelandic translation of the Diagnostic Interview Schedule forChildren⁹ was adapted to ascertain symptomsof ADHD in a standardized fashion and to arrive at a DSM-IV¹⁰ diagnosis.

Our cases were drawn from an ongoing seizure surveillance system atregional health care centers, hospitals, emergency departments, radiologylaboratories, neurophysiology laboratories, and offices of pediatricians,neurologists, and neurosurgeons throughout Iceland. Study nurses maintainedan active surveillance of seizure disorders at these facilities. Once a potentialcase was identified, the medical records were reviewed to verify the occurrenceof an incident seizure disorder. Unprovoked seizure wasdefined as a seizure without an identified proximate precipitant (fever, headtrauma, central nervous system infection, etc).¹¹Epilepsy was defined as at least 2 unprovoked seizuresregardless of seizure type. Definitions of seizure types are discussed inthe "Measures and Assessments" subsection.

The diagnoses of first unprovoked seizure and incident epilepsy weremade only after a complete interview with the parents. Therefore, the groupwith epilepsy includes children initially thought to have had a first unprovokedseizure in whom more detailed interview revealed prior unrecognized unprovokedseizures.

Cases in this analysis were all Icelandic children between 3 and 16years of age with unprovoked seizures or epilepsy first diagnosed betweenDecember 1, 1995, and February 28, 1999. The participation rate was 96.5%.

Age-matched controls were selected from the population registry as thenext 2 same-sex births who were alive, resided in Iceland at the time of theindex seizure, and did not have a history of unprovoked seizure on the dateof the case's incident seizure. Among controls, the participation rate was94.1%.

After obtaining informed consent, a structured telephone interview wasadministered to parents of cases and controls. Within this interview, we usedan Icelandic translation of the lifetime module of the Diagnostic InterviewSchedule for Children to identify symptoms of ADHD among cases and controls3 years and older and to make a DSM-IV diagnosis.Therefore, parents were asked whether their child had ever exhibited certaincharacteristic behaviors before the date of the cases' incident unprovokedseizure or epilepsy. Parents were also asked about the age at which thesebehaviors began; the age at which the behaviors ended; the effect of the behaviorson home, school, and friends; whether medical care had been sought becauseof the behaviors; and whether medication was given to treat the behavioraldisorder. Diagnoses of ADHD, according to DSM-IV,were made: ADHD predominantly inattentive type (ADHD-I), ADHD predominantlyhyperactive-impulsive type (ADHD-H), or ADHD combined type (ADHD-C). We alsoevaluated any inattention (ADHD-I or ADHD-C) and ADHD. We asked about symptomsbeginning at age 3 years and analyzed our data with and without preschoolers,because DSM-IV omits preschoolers from the diagnosisof ADHD. In addition, we considered the number of ADHD symptoms as a continuousvariable.

Three of us (W.A.H., P.L., and E.O.) reviewed all information, includingresults of neuroimaging and electroencephalography, to classify cases by seizuretype and etiology.

Seizure Type. The classification of seizure type was based on the description writtenby the attending medical personnel and information obtained from parents.Seizure type was categorized based on criteria of the International LeagueAgainst Epilepsy¹¹ as generalized from onsetor as partial from onset. Generalized-onset seizures includedcases with generalized tonic, clonic, tonic-clonic, atonic, absence, or myoclonicseizures. Partial seizures included cases with simpleor complex partial seizures with or without secondary generalization. If theclinical characteristics of the seizure could not be determined, seizureswere characterized as "unknown" seizure type.

Etiology. Seizures were categorized as symptomatic or as idiopathic/cryptogenic,according to the recommendations of the International League Against EpilepsyCommission on Epidemiology.¹¹ Seizures wereconsidered symptomatic in the presence of a historyof a central nervous system insult associated with an increased risk of epilepsy(ie, stroke, head trauma, mental retardation, cerebral palsy, and meningitis).The interval between the seizure and the neurological insult had to be morethan 1 week, but was usually months or years. Seizures were also consideredsymptomatic in the presence of nonstatic central nervous system conditions(central nervous system tumors and degenerative neurological diseases). Seizureswere considered idiopathic/cryptogenic in the absenceof an acute precipitating factor or a history of prior neurological insult.

Once the treating physician had given permission to contact parentsor guardians for study purposes, an introductory letter explaining the purposeof the study was sent, followed by a telephone contact, at which time verbalagreement to participate in the study was obtained and a time scheduled toadminister the structured interviews. No cases were lost because of failureof the treating physician to give permission to contact the family.

A letter was sent to parents or guardians of potential controls explainingour study. This was followed by a telephone contact to determine willingnessto participate. If permission was not granted, the next eligible control wasidentified and the same procedure followed until 2 controls agreed to participate.Potential controls were excluded if they had a history of unprovoked seizureor were a sibling of the case.

The study was reviewed and approved by the Icelandic Data ProtectionCommission (Tolvunefnd Rikisins), Ethics Committee of the Chief Medical Officerof Iceland, Medical Ethics Board of the National University Hospital of Iceland(Landspitalinn), Institutional Review Board of the College of Physicians &Surgeons of Columbia University, and the Review Board of the National Institutesof Health, Bethesda, Md.

We used t test to compare continuous variablesand χ² statistic to compare categorical variables. We evaluatedthe prevalence of ADHD by age at seizure diagnosis of the case and the matchedcontrol. Age categories were 3 to 5, 6 to 10, and older than 10 years.

Data were analyzed with SAS,¹² usingconditional logistic regression for matched sets as formulated by Breslowand Day.¹³ Using unprovoked seizure as thedependent variable, models were constructed for the whole study populationand separately by seizure type, etiology, and sex. The following variableswere evaluated as potential confounders of the association between ADHD andunprovoked seizures: a first-degree family history of unprovoked seizure orepilepsy, head injury, prematurity (<266 days' gestational age), and pastfebrile seizures. Confounders were variables that changed the natural logarithmof the odds ratio (OR) for ADHD by at least 10%. Only one case was diagnosedas having a neurodevelopmental abnormality present from birth; therefore,this condition was not examined as a confounder. Adjustment for a historyof febrile seizures changed the natural logarithm of the OR for ADHD-H bymore than 10%, but did not substantially alter the natural logarithm of theOR for ADHD-I or ADHD-C. The confounding effect of a history of febrile seizuresmay have been due to chance given the number of potential confounders tested;therefore, this variable was eliminated from the final model.

One hundred nine cases and 218 controls were 3 years or older. The meanage for cases and controls was 9.2 years (Table 1). Forty-six cases were identified owing to a single incidentunprovoked seizure and 63 were owing to incident epilepsy. Most cases weremale. Partial-onset seizures occurred in 51.4% of cases, and most cases wereidiopathic/cryptogenic. Seizure type at diagnosis was unknown in 1 case. Imagingabnormalities on computed tomography or magnetic resonance imaging were presentin 10.1% of cases, and 64.2% of cases showed electroencephalographic abnormalities.

Two cases with symptoms of ADHD otherwise meeting DSM-IV criteria were classified as unexposed for the analyses becausetheir symptoms of ADHD had not persisted for at least 6 months before theonset of seizures.

Attention-deficit/hyperactivity disorder was associated with an increasedrisk for developing incident unprovoked seizure. This association was significantfor ADHD-I, but not for ADHD-H or for ADHD-C. As shown in Table 2, there was a 2.5-fold increased risk for unprovoked seizurein children with ADHD (95% confidence interval [CI], 1.1-5.5). Having ADHD-Iincreased the risk for unprovoked seizure 3.7-fold (95% CI, 1.1-12.8). Anyinattention (ADHD-I or ADHD-C) increased the risk for unprovoked seizure 3.3-fold(95% CI, 1.1-10.1; data not shown). This pattern was observed among childrenwith partial-onset seizures, generalized-onset seizures, and idiopathic/cryptogenicseizures, but was significant only for the group as a whole. Cells in whichthe controls were unexposed prevented calculation of the OR for remote/progressivesymptomatic seizures (Table 2).Restricting our analysis to children 5 years and older, the risk of unprovokedseizures was increased 2.1-fold for ADHD.

Compared with controls, cases had a greater mean number of symptomsof ADHD-I (2.4 vs 0.9, P<.001), ADHD-H (1.4 vs0.6, P = .01), and ADHD-C (3.7 vs 1.5, P<.001), regardless of DSM-IV diagnosesof these disorders. The OR for unprovoked seizure for each unit increase inthe number of symptoms was 1.17 (95% CI, 1.08-1.26) for ADHD-I, 1.15 (95%CI, 1.04-1.27) for ADHD-H, and 1.10 (95% CI, 1.05-1.16) for ADHD-C.

We examined our data to determine whether unrecognized absence seizuresor delayed diagnosis of complex partial seizures contributed to the diagnosisof ADHD-I. There were 8 children with absence seizures: none met criteriafor ADHD-I, 1 met criteria for ADHD-H, and none met criteria for ADHD-C. Therewere 37 children with complex partial seizures: 3 met criteria for ADHD-I,3 met criteria for ADHD-H, and none met criteria for ADHD-C. Among the 3 childrenwith ADHD-I and complex partial seizures, 1 had epilepsy with 8 days separatingthe first seizure from the second seizure, and 2 had 1 seizure only (one prolongedwith an accompanying Todd paralysis and the other with secondary generalization).Excluding cases with complex partial seizures and their controls from theanalyses did not change the results.

We also examined our data to determine whether the interval betweenthe first unprovoked seizure and diagnosis might explain the results. Resultswere unchanged when children with a single unprovoked seizure at diagnosiswere analyzed separately from children with recurrent unprovoked seizuresat diagnosis (ie, epilepsy) (Table 2).In addition, among children with epilepsy, the median time between first unprovokedseizure and epilepsy diagnosis was 61 days (interquartile range, 12-367 days).

No children in our study had experienced a head injury before the firstunprovoked seizure. Adjusting for a first-degree family history of unprovokedseizure or epilepsy, or for prematurity, did not alter the association betweenADHD and unprovoked seizures. Electroencephalographic abnormalities were notassociated with ADHD (P = .81).

Among cases, epileptiform electroencephalographic abnormalities werepresent in 71.4% with ADHD-I, 50.0% with ADHD-H, and 50.0% with ADHD-C; nochildren with ADHD had rolandic spikes on the electroencephalogram. The ORsdid not differ for the different forms of ADHD when separate analyses wereconducted among cases with epileptiform electroencephalographic abnormalitiesand among cases without epileptiform electroencephalographic abnormalities(data not shown).

A lifetime history of ADHD-I occurred in 6.4% of 109 cases (n = 7) and1.8% of 218 controls (n = 4). A lifetime history of ADHD-H occurred in 5.5%of cases (n = 6) and 3.2% of controls (n = 7). A lifetime history of ADHD-Coccurred in 1.8% of cases (n = 2) and 0.9% of controls (n = 2).

The age distribution of ADHD differed for cases and controls; however,the differences were not statistically significant. Among cases younger than5 years (Figure 1), ADHD-I was morecommon among cases than controls (7.7% vs 0.0%). Cases aged 5 to 10 yearswere more likely than their matched controls to meet criteria for ADHD-I (7.0%vs 0.9% for ADHD-I, 7.0% vs 3.5% for ADHD-H, and 3.5% vs 0.9% for ADHD-C).There was no difference in the frequency of ADHD diagnoses between cases andcontrols older than 10 years. Among cases and controls (Figure 2), there was a nonsignificant increase in the lifetime prevalenceof ADHD-I and ADHD-H for boys compared with girls.

Medication for ADHD was prescribed to 7 children with ADHD: 4 caseswith ADHD and 3 controls with ADHD. Among cases, unspecified medications wereprescribed to 2, methylphenidate hydrochloride to 1, and fluoxetine hydrochlorideand promethazine hydrochloride to 1. Among controls, thioridazine hydrochloridewas prescribed to 1, amitriptyline hydrochloride to 1, and methylphenidate,moclobemide, and clomipramine hydrochloride to 1.

Children with incident unprovoked seizure were 2.5-fold more likelythan age- and sex-matched controls to have a parent-reported history of ADHD,meeting DSM-IV criteria before seizure onset. Theassociation was restricted to ADHD-I. Neither ADHD-H nor ADHD-C was associatedwith an increased risk for unprovoked seizure. Power was limited for analysesof subgroups.

Among controls, we found a lifetime prevalence for ADHD of 6.0% andan excess of boys. This is similar to reports from other population-basedresearch.¹⁴ The lifetime prevalence of ADHDamong cases was twice that of controls, and an earlier onset seems likelyamong cases.

There is no doubt that ADHD exists among preschoolers,^15- 17 butdiagnosis is more difficult in this age group because of overlap with normalage-related behavior. Among children meeting criteria for ADHD as preschoolers,between 50%¹⁵ and 77%¹⁷ havepersistent symptoms later in childhood. When we analyzed our data excludingpreschoolers, we found the same pattern of risk for unprovoked seizures andepilepsy.

In cross-sectional studies,^1- 6 hyperactivityand inattention have been attributed to epilepsy or its treatment. One priorstudy⁷ found that inattention was more commonamong children with new-onset seizures than sibling controls. This study wasnot population-based, it did not diagnose ADHD, and 33.9% of the childrenwith seizures had prior unrecognized seizures.

Attention-deficit/hyperactivity disorder and seizures may be comorbidconditions. That is, the 2 disorders may occur together owing to a causalrelationship between them or owing to an underlying vulnerability to bothdisorders. A history of prior ADHD-I increased the risk for developing unprovokedseizures: 6.4% of our cases had ADHD-I before their seizures. Dunn et al¹⁸ reported that 24% of children with prevalent epilepsyhad ADHD-I and 2.3% had ADHD-C. Children with ADHD-I before their diagnosisof epilepsy were not excluded in the study by Dunn et al. However, the proportionwith ADHD-I was substantially greater than we observed before the occurrenceof seizures. Taken together, these studies may suggest a common underlyingvulnerability to unprovoked seizures and ADHD-I.

Studies^19- 22 examiningthe risk of developing seizures in children with a diagnosis of ADHD havenot been population-based. These studies have reported small numbers of ensuingunprovoked seizures during short follow-up. Nonetheless, the percentage whodevelop unprovoked seizures (0.2%-2%^19- 20,22)is greater than the expected rate, because the average annual incidence ofseizures is approximately 0.05 per year in children aged 5 to 16 years.²³ Therefore, these studies suggest an increased riskof developing seizures in children with ADHD that is far in excess of the3-fold increase we observed in our study.

We found that epileptiform electroencephalographic abnormalities werecommon in children with ADHD and unprovoked seizures, but failed to find adifference in the increased risk for seizures associated with ADHD when analyseswere conducted separately for cases with and without electroencephalographicabnormalities. Others have found epileptiform electroencephalographic abnormalitiesin 14.6% to 30%^19,22 of childrenwith ADHD and rolandic spikes in 5.6%.²¹

Although noradrenergic systems have not generally been implicated inepileptogenesis, the area has not been extensively studied. Evidence for arelationship comes from genetic models of audiogenic seizures.

The mechanism of ADHD-I may share a common underlying antecedent withunprovoked seizure through a hypothesized link to deficits in the centralnorepinephrine system. Among boys with ADHD-I, urinary epinephrine levelswere significantly lower during 3 hours of cognitive testing compared witha non-ADHD group.²⁴ Others have suggested thatthe central norepinephrine system may be dysregulated in ADHD so that it doesnot prime the cortical anterior executive attention system effectively.²⁵ Seizures may also be under the control of the noradrenergicsystem.²⁶ In the genetically epilepsy-pronerat, there is a heritable susceptibility to audiogenic and other induced seizures.These rats have deficits in the noradrenergic system, including reduced norepinephrinecontent in several brain regions.²⁷

Misdiagnosis of ADHD-I, recall bias, and interviewer bias are unlikelyexplanations of our results. Misdiagnosis of ADHD-I was unlikely because thisdiagnosis was independent of absence seizures or complex partial seizures.

Recall bias was unlikely for 2 reasons. First, as expected from population-basedprevalence studies,^28- 29 the lifetimeprevalence of ADHD was greater in boys than girls for cases and controls.Second, the association is specific for ADHD-I and not ADHD-H. Recall biaswould predict that the association would be seen for all subtypes of ADHD.Third, there was no difference in the lifetime prevalence of ADHD betweencases and controls older than 10 years. This is the age group most prone torecall bias, because the onset of ADHD symptoms would be expected in youngerchildren.

Interviewer bias was also unlikely for the following reasons: interviewerswere unaware of the hypotheses; the diagnosis of ADHD does not depend onlyon collected symptoms, but on duration of symptoms, age at first symptoms,and impairment in at least 2 settings, and the interviewers were unaware ofthese diagnostic criteria; the association was specific for ADHD-I, whereasinterviewer bias would predict an association for all subtypes; and interviewerbias would predict a greater endorsement of all factors on the interview,especially those known to be related to epilepsy. However, we did not findan association between head injury and seizures in our data. Finally, of the28 children with ADHD identified by our screening procedure, 11 children hadbeen seen by a developmental pediatrician, whose records confirmed the appropriatediagnosis in all.

Ascertainment of ADHD was retrospective among the older children inthis study. This may account for the decreasing lifetime prevalence of ADHDwith increasing age that we observed among cases. Others have reported a decreasein prevalence with increasing age.³⁰ This hasbeen attributed to underreporting of lifetime history due to remission oreffective use of medication in the older age group. In our study, parentsof older children may have reported on current symptoms rather than currentand past symptoms.

Attention-deficit/hyperactivity disorder was associated with an increasedrisk for incident unprovoked seizure. Therefore, ADHD precedes the developmentof epilepsy, and ADHD or its determinants must be considered risk factorsfor epilepsy. The association was specific for ADHD-I and not ADHD-H. Attention-deficit/hyperactivitydisorder may occur with a greater frequency than expected after the diagnosisof epilepsy, but appropriate population-based studies excluding preexistingADHD have not yet been done to address this question, to our knowledge.

Correspondence: Dale C. Hesdorffer, PhD, Gertrude H. Sergievsky Center,College of Physicians & Surgeons, Columbia University, 630 W 168th St,Physicians & Surgeons Unit 16, New York, NY 10032.

Submitted for publication July 21, 2003; final revision received December27, 2003; accepted February 3, 2004.

This study was supported by grant R01 NS32663 from the National Instituteof Neurological Disorders and Stroke, Bethesda.

We thank Ingigerdur Olafsdottir, RN, Oddny Gunnarsdottir, RN, and IngibjorgEinarsdottir, RN, for their assistance with the study. We also thank DanielPilowsky, MD, for his invaluable comments.