Download pdf - ADHD Symptoms in Children With Mild Intellectual Disability

Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.

ADHD Symptoms in Children With MildIntellectual Disability

EMILY SIMONOFF, M.D., ANDREW PICKLES, PH.D., NICKY WOOD, M.SC.,

PAUL GRINGRAS, M.B., B.S., AND OLIVER CHADWICK, PH.D.

ABSTRACT

Objectives: To determine whether the nature and correlates of attention-deficit/hyperactivity disorder (ADHD) symptoms

are different in subjects with mild intellectual disability (ID) compared to subjects with average ability. Method: From a

general population sample of 2,726 12- to 15-year-olds, a stratified subsample was selected to enrich for mild ID. A total of

192 subjects were included in the analyses. ADHD symptoms and other emotional/behavioral problems were measured

with the parent and teacher Strengths and Difficulties Questionnaire and IQ with the WISC-III-UK), and social

communication difficulties were assessed by a short version of the Social Communication Questionnaire and academic

attainments by the Wechsler Quicktest. Results: There was a negative linear relationship between ADHD symptoms and

IQ (" =j.087, p < .001). The relationship could not be explained by inappropriate rater expectations. Neither the profiles of

ADHD symptoms nor the comorbidity with emotional/behavioral problems differed according to the presence of ID. When

IQ was accounted for, the group difference in attainments was nonsignificant. Conclusions: ADHD symptoms are

increased in people with ID. We found no evidence that this increase can by explained by inappropriate expectations or by

confounding associations with other emotional/behavioral or cognitive problems. J. Am. Acad. Child Adolesc. Psychiatry,

2007;46(5):591Y600. Key Words: attention-deficit/hyperactivity disorder, mental retardation, epidemiology, comorbidity,

childhood, intellectual disability.

Hyperactivity syndromes and disorders (DSM-IVattention-deficit/hyperactivity disorder [ADHD] andICD-10 hyperkinetic disorder [HKD]) include symp-toms of overactivity, inattention, and impulsivity. Asneuropsychiatric conditions, it is assumed that ADHDand HKD also affect cognition. It has long been

recognized that children with ADHD have as a groupIQs that are roughly 7 to 12 points below the popu-lation mean (Crosbie and Schachar, 2001; Mariani andBarkley, 1997). However, it is not clear whether themean difference is due to an overall shift in the IQdistribution among those with ADHD or to an increasein ADHD among those with intellectual disability(ID). Within the average intelligence population,ADHD symptoms are correlated with lower ability(Kuntsi et al., 2004), but it is not known whether thisrelationship continues into the mental retardation orID range.Previous epidemiological studies of ADHD and

HKD have suggested these disorders may be increasedin individuals with ID, but there are flaws with eachstudy. Emerson (2003) used the 1999 Survey of theMental Health of Children and Adolescents in GreatBritain (Meltzer et al., 2000) to report rates of HKD inthose with global learning disability of 8.7% comparedto 0.9% to the rest of the general population sample, a10-fold increase in risk. However, the definition of

Accepted December 3, 2006.Dr. Simonoff and Ms. Wood are with the Department of Child and

Adolescent Psychiatry, King_s College London, Institute of Psychiatry, London;Dr. Pickles is with the Biostatistics Group, Division of Epidemiology and HealthScience, University of Manchester, Manchester, UK; Dr. Gringras is with Guy_sand St. Thomas_ Hospitals NHS Trust, London; and Dr. Chadwick is with theDepartment of Psychology, King_s College London, Institute of Psychiatry,London.

This study was supported by project grant 058599 from the Wellcome Trust toDr. Simonoff. The authors thank all of the young people and families whoparticipated in the study and also the researchers who collected the data.

Correspondence to Dr. Emily Simonoff, Department of Child and AdolescentPsychiatry, Box 85, Kings College London, Institute of Psychiatry, De CrespignyPark, London SE5 8AF, UK; e-mail: [email protected].

0890-8567/07/4605-0591�2007 by the American Academy of Childand Adolescent Psychiatry.

DOI: 10.1097/chi.0b013e3180323330

591J . AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:5, MAY 2007


learning disability employed was not based on anintellectual ability measure but rather used parentalaccounts of serious learning concerns and/or attendanceat schools for learning difficulties, provided teacherinformation was not contradictory. A Dutch study usedthe Diagnostic Interview Schedule for Children toassess a large sample of Dutch school-age childrenattending special schools for the educable (IQ ~60Y80)and trainable (IQ <60; Dekker and Koot, 2003).Overall, the rate of DSM-IV ADHD was 14.8%,decreasing to 6.8% when an additional impairmentcriterion was included. The Norwegian study ofStromme and Diseth (2000) ascertained children withID through registers and teacher/clinician nominationsand used semistructured interviews based on ICD-10criteria and observation to identify psychiatric dis-orders. HKD was diagnosed in 16% of the sample.Gillberg et al. (1986) studied Swedish schoolchildrenwith moderate and severe ID in whom semistructuredparental interviews and observation were completed.No cases of HKD were reported among the 73individuals with severe ID (IQ <50), in contrast to arate of 11% in the 91 with mild ID. Because the latterstudy assessed the presence of autism and relatedpervasive developmental disorders, which form anexclusion for diagnosing HKD in ICD-10 (andADHD in DSM-IV ), it may be that subjects withsevere ID who are at higher risk for autism andpervasive developmental disorders were more likely tobe excluded from an HKD diagnosis. The latter threestudies did not use general population comparisons, soit is not possible to say whether the rates of ADHD andHKD are significantly higher in those with ID,although the consistency of the high rates is suggestiveof this.Given the high rates of ADHD and HKD reported

from systematic studies, it is surprising that relativelylittle research or clinical attention has been devoted tothe area. Although research on psychiatric disordersamong people with ID is a generally neglected area, thishas not been the case for autism and aggressive(challenging) behavior. The process of diagnosticovershadowing (Reiss and Szyszko, 1983), wherebypsychiatric disorder is underrecognized in the presenceof ID, is well recognized in adult psychiatry and is likelyalso to apply to child mental disorders. An alternativeexplanation is that ADHD behavior is recognized butconceptualized differently by clinicians when it presents

in children with ID. Clinicians may recognize ADHDsymptoms in such children but decide that the behavioris not out of keeping with their developmental level, arequirement for both DSM-IV and ICD-10 diagnoses.Alternatively, ADHD symptoms may be understood asa secondary phenomenon. Clinicians may decide thatchildren with low IQ who demonstrate overactive,inattentive behavior are behaving in this way becausethe demands placed on them, scholastic and otherwise,are inappropriate for their developmental age. Finally,because children with ID and a psychiatric disorder aremore likely to have comorbid psychiatric disorders(Emerson, 2003), ADHD may be undiagnosed eitherbecause clinicians stop making diagnoses once onedisorder accounting for the primary or presentingsymptoms has been identified or because ADHDsymptoms are viewed as a component of the disorderdiagnosed. Implicit in these alternative accounts is thepossibility that clinicians may obtain reports of ADHDsymptoms but treat these differently from reports onchildren of average ability.We used a general population sample, screened to

include a high proportion of children with mild tomoderate ID, to determine whether ADHD symptomsare more common in children with ID. We have alsolooked for evidence that parent and teacher informationabout ADHD symptoms may be biased or less reliablein children with ID. We have explored whether ADHDsymptoms are qualitatively different in ID, either intheir symptom profiles or their associated problems.

METHOD

Sample

The present study makes use of a general population sample ofstudents ages 12 to 15 years who were ascertained as part of a studyexamining the role of medical and psychosocial risk factors inmental retardation or ID (Simonoff et al., 2006). Secondaryschoolchildren attending local authority schools (public schools inthe United States) in the outer London borough of Croydon wereidentified as the target population. Croydon is broadly representa-tive of the United Kingdom in terms of sociodemographiccharacteristics, with the exception of a higher rate of ethnicminority groups than seen in the United Kingdom as a whole(29.8% versus 9.1% nonwhite; Census, 2001, 2004). Of the 22eligible schools, 15 participated in the screening stage, including theone school for adolescents with moderate learning disability(approximate to mild mental retardation), two special schools forchildren with emotional and behavioral difficulties, and two pupilreferral units for adolescents excluded from mainstream school andbeing evaluated for an alternative placement. A total of 2,726children in years 8 and 9 were screened using an abbreviated version

SIMONOFF ET AL.

592 J . AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:5, MAY 2007


of the Cognitive Abilities Test (CAT; Thorndike and Hagen, 1986).This represented 92.3% of the students on the rolls of theparticipating schools. The CAT is a paper-and-pencil test ofintellectual ability. We used four subtests: verbal classification,sentence completion, figure classification, and figure synthesis. Theitems for the age-appropriate level (level D) were augmented withtwo items from each of the three lower levels (A, B, and C) toimprove scoring at the lower end of the distribution. A total scorebased on all four subtests was regressed against age in months. Thisage-regressed score correlated 0.76 with Full Scale IQ on the WISC-III-UK in the present study. A stratified random sample was drawnfor in-depth evaluation, selecting all of those scoring in the lowest5th centiles of the CAT (high-risk group, n = 133), 17% from thosescoring between the 6th and 25th centiles (moderate-risk group, n =93), and 3% of those scoring above the 25th centile (low-risk group,n = 78). The study received ethical approval from the Institute ofPsychiatry/South London and Maudsley Trust and the Guy_sHospital Research Ethics Committees.

Measures

The in-depth psychometric assessment used nine WISC-III-UKsubtests (Wechsler, 1992): vocabulary, similarities, information,comprehension, and digit span as verbal subtests and picturecompletion, picture arrangement, object assembly, and block designas the nonverbal subtests. Verbal IQ was substituted for Full ScaleIQ for three children unable to complete many of the performancesubtests because of physical disability. The Wechsler Quicktest(Wechsler, 1996) was used to assess attainments. Eight subtests ofthe Children_s Memory Scale (Cohen, 1997) were included to assessboth immediate and delayed verbal and visual recall as well asdelayed recognition: dot locations, stories, faces, and word pairs,immediate and delayed.Children_s emotional and behavioral problems were screened

using the parent-, teacher-, and self-report versions of the Strengthsand Difficulties Questionnaires (SDQ) giving scales of emotional,conduct, hyperactivity and peer relationship problems, andprosocial behavior (Goodman, 1997). Only the parent and teacherresponses were used in the present analyses because the self-reportmeasure has not been validated among those with ID. The SDQ iswidely used as a brief screening instrument for child psychiatricproblems and its psychometric properties have been established inseveral samples, including the United Kingdom (Goodman, 2001)and the United States (Bourdon et al., 2005). The five hyperactivity(ADHD) items were restless, overactive, cannot stay still for long;constantly fidgeting or squirming; easily distracted, concentrationwanders; thinks things out before acting (reverse-scored); and seestasks through to the end, good attention span (reverse-scored). Atotal (parent + teacher) score was generated; where data from onesource but not the other were missing (in 12 parent and 14 teacherSDQs), an imputed score from that informant was generated basedon the score from the other informant. This was possible because ofthe relatively high correlation (0.53) between parent and teacherscores. The mean imputed parent score for hyperactivity was 4.40and was the same for the unimputed mean; for teachers, theimputed mean was 4.37 compared with 4.49 for the unimputedscores. The published algorithms employing parent and teacherSDQ scores (Goodman et al., 2004) was used to generate thetrichotomous classification of unlikely, possible, and probableHKD. This classification was collapsed into a binary grouping ofdubious (including the unlikely and possible cases) and probablecategories. These algorithms have been compared with research

diagnoses from the semistructured Development and Well-beingAssessment; for HKD, the combined parent-teacher algorithmshowed a sensitivity of 85% to 87% (Goodman et al., 2000).Hyperactivity subscales from parent and teacher questionnaires weregenerated: overactivity (restless, fidgets; cannot stay still for long)scores ranged from 0 to 8; inattention (easily distracted; goodattention span) 0 to 8; and impulsivity (thinks things through beforeacting) 0 to 4. The impact scale for the SDQ was calculated fromthe sum of responses to four questions from the stem BDo thedifficulties interfere with the child_s life in the following areas?[Respondents rated this on a 4-point scale (0 = not at all, 1 = only alittle, 2 = quite a lot, 3 = a great deal). Parents were asked to rate thisin relation to home life, friendships, classroom learning, and leisureactivities, whereas teachers rated the last three only. In addition, thequestion BDo the difficulties upset or distress your child?[ (rated onthe same 4-point scale) was added to the impact score. Parentalimpact scores ranged from 0 to 15 and teacher impact scores from 0to 12.Comorbid emotional and behavioral problems were examined

using the SDQ domains of emotional and conduct symptomscombined for parents and teachers as described for hyperactivity.The total score on an abbreviated (20 items) version of the parent-reported Social Communication Questionnaire (Rutter et al., 2003)was used as an index of autism spectrum symptoms. Theabbreviated version was developed jointly with the first author ofthe original version (M.R.) and included items with high loadingsfrom each of the four domains described in the original factoranalysis (Berument et al., 1999). The purpose of including thisabbreviated measure was to gain a dimensional measure of autismspectrum symptoms, which are known to be associated with ID.The inclusion of this scale in the present analyses allowedexploration of the relationship between ADHD and autisticsymptoms, which many clinicians believe is stronger than wouldbe expected by chance. All of the behavioral scales were prorated ife30% items were missing.

Analysis

Data reduction was implemented using SAS version 8.1/8.2.Two stages of adjustment were made. The first stage was to accountfor the nonparticipation of some schools and of some individualswithin participating schools. The second stage was to account forthe CAT score selective two-phase sampling design. In the first stagethe screened sample was adjusted to match data for all Croydon statepublic schools on national standard assessments in school year 9 (age13+), the seven-category Key Stage 3 (KS3) mathematics grades.These were available in aggregate form by sex for both participatingand nonparticipating local authority schools and individually for1,097 subjects from the participating schools. A hot-deckimputation procedure drew CAT scores from a pool of matcheddonors (individuals with a CAT score from the same type of school,special versus other, school year, and mathematics KS3 grade) toimpute CAT scores for all nonparticipants that would be consistentwith the known aggregate and individual KS3 data. These imputedscores were then used to calculate inverse probability weights(Pickles and Dunn, 1988) for completion of a CAT screen. Thesesteps made the weighted screen sample representative of pupilsattending all local authority schools in Croydon. In the second stagea further weighting adjustment method was used to account for thetwo-phase design. The weighting and imputation allowed accountto be taken of differential response rates by school, KS3performance, sex, and school year, and, in addition for in-depth

HYPERACTIVITY IN MILD INTELLECTUAL DISABILITY



measures, by CAT performance. Analyses used the robust estimatorof the parameter covariance matrix that not only took account of theweights but possible correlation resulting from clustering by schooland heteroscedasticity of errors in regression (Binder, 1983). Thislast feature is desirable when examining, as we did in the presentstudy, skewed distributions common to symptom scores. All of theanalyses were undertaken in Stata version 7/8 (StataCorp, 2003).

RESULTS

Of the 304 subjects invited to participate in the in-depth evaluation, 204 (67%) consented to do so.Participation was significantly greater in those at lowrisk (64/78, 82%) than that in the moderate-risk(60/93, 65%, odds ratio [OR] = 2.5, p = .01) and thehigh-risk groups (80/133, 60%, OR = 3.0, p < .001). Ofthose taking part in the in-depth study, 12 onlycommenced education in a school in which Englishwas the first language at age 11 or later (the time oftransfer to secondary school in the United Kingdom).Previous exploratory analyses suggested that their IQresults may be falsely lowered because of difficulties withEnglish, and they were therefore excluded from allsubsequent analyses, reducing the sample size to 192.Themean age for participation in the screening was 13.7years (SD 0.64, range 11.9Y16.0 years) and for the in-depth assessment 14.2 years (SD 0.8, range 12.5Y16.1years). There were 125 males and 67 females in the in-depth study. Full Scale IQ ranged from 40 to 137. IQwas classified in two ways: first, by IQ <70 and Q70 and,second, into four groups of IQ <60, 60 to 69, 70 to 90,and >90. The four-group classification was selected forseveral reasons. In our sample, IQ <60 corresponds tothe lowest 3%, often thought of as the administrativeprevalence of mild mental retardation. In addition, thegroups of 70 to 90 and >90 conceptually reflect lowaverage and average to high ability, which compriseuseful contrasts.We first examined the internal consistency of the

SDQ, both for the entire scale and also for thehyperactivity scale, separately for normal ability and IDsubjects. For the entire parent questionnaire, theCronbach ! was .85 for the normal ability group and.84 for those with ID; for the teacher questionnaire,respective ! coefficients were .90 and .99. Examiningthe parent-rated hyperactivity items alone, Cronbach !coefficients were .77 and .69 for normal ability and IDsubjects, whereas corresponding results for the teacher-rated hyperactivity scale were .90 and .85. For theparent-teacher combined ADHD subscales, the

Cronbach ! coefficients were, according to averageability and ID, respectively, .75 and .71 of the four-itemoveractivity scale, .71 and .70 for the four-item in-attention scales, and .51 and .31 for the two-itemimpulsivity scale.The relationship between Full Scale IQ and the

hyperactivity scale, shown in Figure 1, was examined inseveral ways. First, the mean IQ was compared betweenthose with a probable hyperkinetic diagnosis on SDQ(78.3) and those without (94.0) and was found to besignificantly different (adjusted Wald test F1,13 = 9.64,p = .008). Second, the parent, teacher, and combined(parent + teacher) SDQ hyperactivity scores wereregressed on IQ. There was a significant negativeassociation for each: parent (" = j0.04, p < .001),teacher (" =j0.05, p < .01), and combined ratings (" =j0.09, p < .001). Nonlinearities in the relationshipbetween hyperactivity symptoms and IQ were tested byadding the square of IQ as an additional predictor. Thebest-fit regression line shown in Figure 1 furtherindicates the minimal amount of nonlinearity in theIQYhyperactivity relationship. In all three instances,this was nonsignificant. Sex was added to the bivariateregression of IQ on combined hyperactivity score toexclude it as a confounder. This revealed main effects ofboth male sex (" = j3.50, p < .01) and an increase inthe effect of IQ (to " = j0.10, p < .001). A further testof the interaction between gender and ID wasnonsignificant (" = 0.03, p = .40). The rate of SDQ-defined probable HKD are IQ <60, 0.12 (95%confidence intervals [CIs] 3.6%Y27.9%); IQ 60 to69, 0.06 (95% CIs 0.8%Y10.7%); IQ 70 to 90, 0.03

Fig. 1 The association between combined (parent + teacher) hyperactivityscore and Full Scale IQ is depicted by the best two-degree fractionalpolynomial regression line (Royston and Altman, 1994) and its associated95% confidence envelope.

SIMONOFF ET AL.



(95% CIs 0.3%Y7.1%); and IQ >90, 0.01 (95% CIs0%Y5.1%).

We explored whether ratings of ADHD symptomswere subject to bias among children with lower IQ. Wehypothesized that informants not taking properaccount of children_s lower than average developmentallevel may rate the child as having more ADHDsymptoms than would occur if developmental levelwere accounted for. To examine parental responses abinary variable of parental discrepancy betweenmeasured IQ and parental beliefs regarding abilitywas generated. During the interview parents were askedto rate their child_s academic ability in one of fivecategories: a lot above average, above average, average,below average, or well below average. Expectations weredefined as discrepant where measured IQ was <70 andparents rated their child_s ability as average, aboveaverage, or a lot above average or where IQ was 70 to 90and parents rated their child as well above average. Ofthe 83 children with IQ <70, 45 parents werediscrepant in ratings of their ability; 13 of the 60with IQ 70 to 90 were also classified as discrepant.When the discrepancy classification was added to theregression prediction of parental hyperactivity ratingsfrom Full Scale IQ, it proved significant (" = j1.51,p < .01); IQ remained a significant predictor but witha much reduced coefficient (" = j0.04, p = .03).

Similarly, teacher ratings of ADHD symptoms maybe falsely elevated in children whose ID was not fullyrecognized. If this were the case, then the relationshipbetween IQ and teacher ratings should be mediated bywhether the IDs are recognized. The study did notinclude a direct measure of teachers_ assessment of theirpupils_ intellectual ability. As a substitute, whetherchildren attended a school for learning disabilities wasadded as a covariate to the previous regression of IQ on

teacher hyperactivity ratings and revealed a non-significant effect of school type (" = 0.50, p = .54),with no effect of the significance of the IQ on ratings(" = j0.05, p < .01).We hypothesized that the nature of ADHD

symptoms may be different in children with ID thanthose without. Specifically, attention problems mayappear to be more common in ID if developmentallevel was not accounted for and children were exposedto inappropriately challenging cognitive expectations.This was examined in two ways, as shown in Table 1.We regressed each of the three subscales of overactivity,inattention, and impulsivity (based on both parent andteacher items) on the total score for all otherhyperactivity items to control for the overall hyper-activity score and a binary ID variable of IQ <70 versusIQ Q70. The coefficients associated with the IDclassification were nonsignificant, indicating that therelative endorsement of the three subscales was notdifferent in the ID group. The analysis was repeated,substituting the continuous measure of Full Scale IQfor the ID classification. These analyses were alsononsignificant, except for the regression of Full ScaleIQ on the overactivity score, which showed that lowerIQ was associated with more overactivity, even ac-counting for other hyperactivity symptoms (" = 0.02,p = .04).Comorbidity between ADHD and other emotional/

behavioral symptoms was examined to determinewhether the patterns were different for those with IDfrom those without (Table 2). Three models were fittedto each symptom scale. In model 1 the comorbiditybetween the hyperactivity scale and other symptomdomains was quantified. In the second model the binaryID variable was added as a main effect. In a final modelan interaction term between hyperactivity and ID was

TABLE 1Hyperactivity Subscale Scores and Relationship to Intellectual Disability

Relationship of IQ toHyperactivity Subscale,Covarying for OverallHyperactivity Score

Subjects With Probable Hyperkinetic Disorder

Mean Score Regression Coefficients

"a p IQ <70 IQ Q70 "a p

Overactivity 0.019 .04 6.53 6.22 0.304 .63Inattention 0.000 .97 6.70 7.57 j0.393 .14Impulsivity j0.001 .93 3.15 3.51 j0.354 .31

a Unstandardized regression coefficient.




included. Sex was included as a covariate to ensure thatit did not confound the relationships. Hyperactivitysymptoms showed strong and highly significantassociations with emotional and conduct symptomsfor the entire sample but no significant association withsocial communication problems. Sex was only signifi-cant in relation to emotional symptoms, where femaleswere more likely to be affected. When ID was added asa main effect, it also independently predicted emotionalbut not conduct or social communication symptoms.An interaction between hyperactivity and ID in

predicting emotional symptoms was tested but provednonsignificant (" = 0.04, p = .14). For any level ofhyperactivity symptoms, children with ID had emo-tional symptom scores on average ~1.4 symptomshigher than those without ID.When an interaction termwas added to the social communication symptommodel, no new significant associations were identified.However, adding the interaction term to the conductsymptom model produced a borderline significantinteraction (" = 0.18, p = .06) and the coefficient forID as a main effect became significant (" = j1.52,p = .03). However, these effects were in oppositedirections and a 2-df test for the joint effect of the IDmain effect and its interaction with hyperactivity wasnonsignificant (Wald test F2,12 = 2.84, p = .10). Theparsimonious interpretation is therefore that there areno marked differences in the relationship between con-duct and hyperactivity symptoms resulting from ID.We hypothesized that if ADHD symptoms were

overrated in children with ID, then the effects onpsychosocial functioning may be less. As shown inTable 3 hyperactivity symptoms were strongly andsignificantly related to impact score (model 1) for bothparent and teacher reports. Adding the ID classificationhad minimal effect on the strength or significance ofthese relationships and the ID variable was not itselfassociated with impact. Because the impact ratings onthe SDQ refer to all symptoms that are endorsed, weadded into the second model the informant-specificemotional and conduct disorder scores. Although these

TABLE 3Relationship of Hyperactivity Symptoms to Psychosocial Impact

Parent ImpactScore

Teacher ImpactScore

"a p "a p

Model 1Hyperactivity 0.491 <.001 0.673 <.001Sex j0.151 .83 0.508 .17

Model 2Hyperactivity 0.491 <.001 0.653 <.001Intellectual disability 0.486 .92 0.545 .47Sex j0.154 .90 0.494 .18

Model 3Hyperactivity 0.188 .01 0.198 .07Emotional symptoms 0.373 <.01 0.321 .01Conduct symptoms 0.510 .001 0.722 .001Intellectual disability j0.355 .20 0.561 .38Sex j0.402 .08 j0.028 .94


TABLE 2Comorbidity of Emotional/Behavioral Symptoms With Hyperactivity

Emotional Symptoms Conduct SymptomsSocial Communication

Symptoms

"a p "a p "a p

Model 1Hyperactivity 0.394 <.001 0.389 <.001 0.088 .22Sex 1.446 <.001 0.434 .28 j0.553 .25

Model 2Hyperactivity 0.381 <.001 0.390 <.001 0.084 .24Sex 1.360 <.001 0.0440 .27 j0.58 .23Intellectual disability 1.375 .02 0.072 .87 0.420 .40

Model 3Hyperactivity 0.374 <.001 0.376 .03 0.081 .29Sex 1.379 <.001 0.476 .06 j0.575 .24Intellectual disability 0.344 .53 j1.520 .03 0.072 .92Hyperactivity � intellectual disability 0.095 .14 0.178 .06 0.043 .63


SIMONOFF ET AL.



reduced the strength of the association between thehyperactivity and impact scores, these remained sig-nificant and ID continued to be nonsignificant. Nointeractions between the hyperactivity scale and IDwere significant (model 3).

The relationship of ADHD symptoms to academicattainments in reading, spelling, and mathematics, asmeasured on the Wechsler Quicktest, was explored in asimilar way (Table 4). We focused on attainmentsdiscrepant from intellectual ability or underachieve-ment, by including IQ as a covariate; sex was alsoincluded as a covariate because of its association withboth ADHD symptoms and poorer academic attain-ments. Model 1 shows that the hyperactivity scale wasassociated only with academic underachievement inrelation to mathematical attainment and not reading orspelling attainment discrepancies. ID was then added asa binary covariate to test whether its presence altered therelationship of the hyperactivity scale to discrepantachievement (model 2). For none of the three areas ofattainment discrepancy was ID a significant predictorof the association with the hyperactivity scale.

DISCUSSION

Our findings confirm the view that ADHDsymptoms are increased in children with ID and thatthe mean IQ of children with probable HKD is some15 to 16 points below those without HKD. The size ofthe increased risk of probable HKD was somewhatsmaller that that reported by other groups (Emerson,2003; Hastings et al., 2005), but the CIs on both theearlier and the present estimates are wide. We have

demonstrated that ADHD symptoms, as rated byparents and teachers, are linearly related to lower IQacross the ability range. Our population includesadolescents with IQs ranging from 40 to 137, andthis allows the relationship between ADHD symptomsand cognitive ability to be well delineated. Previousstudies have reported a relationship between lower IQand greater problems with ADHD. In a twin studyfocusing largely on variation in the normal range,Kuntsi et al. (2004) showed that ADHD scores werenegatively correlated with IQ. Our study indicates thatthis relationship continues well below the normal rangeof cognitive ability into that of mild and moderate ID.Our findings do not suggest any change in the strengthof the relationship between IQ and ADHD symptomsacross their range. This indicates that the proportionwith ADHD symptoms and therefore probableADHD/HKD should be greatest in those with lowerIQ. Because our study did not include subjects withsevere/profound ID, it is unknown whether the ratescontinue to increase in this group.Although this finding is largely consistent with the

extant literature, we have nevertheless looked for factorsthat could cause a false or biased relationship betweenIQ and ADHD symptoms. The aim was to ensure asmuch as possible that the increased risk of ADHDsymptoms is a true finding and not caused byconfounders or bias. The present results largely supportthis because the presence of ID has few effects on therelationship of ADHD to a range of associated factors.Diagnostic criteria for both ADHD and HKD stipulatethat adequate adjustment for children_s developmentallevel should be made before endorsing symptoms. The

TABLE 4Scholastic Attainments in Relation to Hyperactivity Symptoms

Reading Spelling Mathematics

"a p "a p "a p

Model 1: Prediction of attainments from hyperactivity having accounted for IQ and sexHyperactivity 0.123 .54 0.076 .73 j0.636 .03IQ 0.569 <.001 0.168 <.01 0.514 <.001Sex 6.670 .01 10.006 .02 j1.819 .59

Model 2: Prediction of attainments from hyperactivity having accounted for IQ, sex, andintellectual disability

Hyperactivity 0.094 .65 0.038 .89 j0.668 .01Intellectual disability j6.665 .08 j9.588 .14 j7.314 .07IQ 0.496 <.001 0.362 .07 0.433 <.005Sex 6.549 .01 9.833 .02 j1.951 .55





purpose is to reflect the normal developmental changesin levels of activity, attention, and impulsivity. TheSDQ gives no such instruction to parents and teachers,nor do many of the other questionnaires for ratingADHD symptoms. Furthermore, the algorithm devel-oped for identifying probable HKD does not vary theSDQ scores according to children_s age. The assump-tion therefore is that in rating an individual child,parents and teachers make an implicit comparison totheir expectations in relation to other children of thesame age. It is not clear how informants modify theirresponses when they are rating children whose devel-opmental age is considerably below their chronologicalage. Our results do not suggest that the higher ADHDratings in children with lower ability are simply theresult of inappropriate expectations. Although aBdiscrepant[ parental view (i.e., one in which parentsappeared not to appreciate the developmental delay ordegree of disability) was related to higher ratings ofADHD symptoms, it failed to explain all of the rela-tionship between ADHD symptoms and ID. Similarly,the relationship between teacher ratings and IQ was notexplained by Brecognition[ of ID, as indexed by atten-dance at a school for children with learning disabilities.We explored whether particular symptom types

(overactivity, inattention, and impulsivity) were differ-entially represented in children with ID. We antici-pated that overactivity may be more common inchildren with ID because of their lower developmentalage and the clinical observation that this is mostfrequently a problem in younger children. We alsothought that inattention may be greater in childrenwith mental retardation because cognitive demandswould be more challenging. Only overactivity symp-toms were marginally associated with ID.We examined patterns of coexisting problems, bothpsychiatric symptoms and academic attainments. Theexpectation was that if ADHD symptoms are under-recognized in pupils with ID when other mental/developmental problems are present, then we wouldexpect the degree of association or comorbidity to beattenuated because of increased measurement error.The only comorbidity affected by the presence of IDwas that of emotional symptoms, in which an asso-ciation remained after accounting for hyperactivitysymptoms and sex. It is interesting that the relationshipbetween ADHD and social communication symptomswas not increased in the presence of ID because social

communication problems are also more common inchildren with lower ability (Fombonne, 1999) andclinicians report difficulty in differentiating the twosyndromes among children with ID. Among childrenwith severe ID, one study reports that ADHD symp-toms were rated as higher in those with a diagnosis ofautism (Hastings et al., 2005) However, the abbreviatedSocial Communication Questionnaire failed to show anassociation with IQ in our sample, although it isstrongly linked to the identification of special educa-tional needs (Simonoff et al., 2006).The finding that ADHD symptoms in children with

ID are associated with equally high psychosocial impactas seen in those of normal ability is a further argumentthat these ratings are not simply the result of in-appropriate expectations. Rather, they suggest thatreports of ADHD symptoms should be considered withequal care in those with ID.Although ADHD symptoms are well recognized as

related to both general poor scholastic achievement(Fergusson and Horwood, 1995; Tannock, 1998) andto specific disorders such as reading retardation/dyslexia(Rucklidge and Tannock, 2002), we are not aware ofprevious studies that have examined whether therelationship between underachievement (i.e., poorerachievement than would be expected on the basis ofintellectual ability) is differentially related to ADHDsymptoms for children with ID. Previous work that hasexamined a genetic link between attainments andADHD has not always accounted for IQ, which maybe a confounder (Stevenson et al., 1993). Children withID showed the same reduction in the level ofattainments predicted by IQ as seen in those of normalability. This finding must be viewed cautiously, how-ever, because all coefficients associated with ID remainnegative and in some cases close to significance in amodel that has already accounted for IQ and thereforehas low power to detect an additional effect of ID.Previous research has highlighted underlying cogni-tive processes that are impaired in subjects withADHD, in particular, different types of inhibitorydysfunction (Tannock, 1998). It is uncertain whethersimilar deficits occur in subjects with ADHD and IDbecause few studies using current methods of examininginhibition have included subjects with ID. The earlyliterature on neuropsychological processes in subjectswith ADHD and ID suggested that different processesmay be operating (Pearson et al., 1997), but new studies

SIMONOFF ET AL.



with up-to-date neuropsychological assessments areneeded.

Strengths

The strengths of the present study include the use of ageneral population sample, screened and weighted toallow appropriate correction for the overrepresentationof subjects of low cognitive ability. Our previousfindings have indicated that with regard to ID, the useof register-based populations detects only a smallproportion of those individuals with IQ <70 and isbiased with respect to behavioral symptoms (Simonoffet al., 2006). Greater confidence can be placed in thepresent high rates of ADHD symptoms because of theascertainment method that was employed. The over-representation in the sample of children with IQ <70means that the group-based comparisons have reason-able power to detect differences where they exist. Hence,we can have reasonable confidence in our failure to findsignificant differences between subjects with and with-out ID. IQ was measured individually on each childusing a full version of the WISC, providing an accuratemeasure of cognitive ability. With respect to assessingbehavioral symptoms, we have ratings from both parentsand teachers. It is well recognized that the agreementbetween informants is usually modest (McConaughyet al., 1992). The use of multiple informants is thereforean advantage in exploring and excluding groupdifferences. The availability of multiple measures hasallowed us to explore from a number of directionswhether theremay be systematic differences in the ratingof ADHD symptoms in those with ID. The range ofanalyses undertaken here lend weight to the view thatratings of ADHD symptoms in those with ID areunlikely to be strongly biased in comparison to childrenof normal ability and also that many of the correlates ofADHD are similar across groups.

Limitations

There are several limitations to the present study.First, the measures of ADHD are limited. The parentand teacher SDQmeasures each include only five items.However, the questionnaire has been used alongsidediagnostic interview measures and its validity previouslyestablished (Goodman et al., 2004). Nevertheless,analyses such as the symptoms subtype comparisonmay be less powerful because of the scale_s brevity. Thepresent study did not use a psychiatric diagnostic

interview or direct observation of behavior. This meansthat the question of whether ADHD ratings in thosewith ID are accurate and unbiased must be approachedindirectly. However, we should not exaggerate thislimitation because it cannot be assumed that eitherdiagnostic interviews (which are also informant based)or direct observation (which are time and situationlimited) are an unequivocal gold standard.

Clinical Implications

Our findings indicate that ADHD problems arelikely to be more common in children with ID.Clinicians need to take such symptoms seriously andshould include systematic questioning about ADHD aspart of their routine assessment of children presentingwith ID and mental health problems. In those withmild to moderate ID, our findings suggest that generalpopulation screening measures such as the SDQ can beused as an indicator of ADHD symptoms.

Disclosure: Dr. Gringas has attended one international academicprofessional conference for which his travel and accommodations werepaid by Janssen-Cilag. He has also lectured to professionals and beenpaid honoraria by Janssen Cilag. Dr. Pickles receives royalties from thedistribution by WPS of autism screening/diagnostic instruments (SCQ,ADOS, ADI) and from publishers from the sales of a number ofacademic texts; he also receives occasional honoraria for methodologicaltraining provided on an ad hoc basis to for-profit organizations. Theother authors have no financial relationships to disclose.

REFERENCES

Berument SK, Rutter M, Lord C, Pickles A, Bailey A (1999), Autismscreening questionnaire: diagnostic validity. Br J Psychiatry 175:445Y451

Binder DA (1983), On the variances of asymptotically normal estimatorsfrom complex surveys. Int Stat Rev 51:279Y292

Bourdon KH, Goodman R, Rae DS, Simpson G, Koretz DS (2005), TheStrengths and Difficulties Questionnaire: U.S. normative data andpsychometric properties. J Am Acad Child Adolesc Psychiatry 44:557Y564

Census 2001 (2004), CD Supplement to the National Report for England andWales and Key Statistics for Local Authorities in England and Wales.London: Office of National Statistics

Cohen MJ (1997), Children_s Memory Scale. London: The PsychologicalCorporation

Crosbie J, Schachar R (2001), Deficient inhibition as a marker for familialADHD. Am J Psychiatry 158:1884Y1890

Dekker MC, Koot HM (2003), DSM-IV disorders in children withborderline to moderate intellectual disability II: child and familypredictors. J Am Acad Child Adolesc Psychiatry 42:923Y931

Emerson E (2003), Prevalence of psychiatric disorders in children andadolescents with and without intellectual disabilities. J Intellect DisabilRes 47:51Y58

Fergusson DM, Horwood LJ (1995), Early disruptive behaviour, IQ, andlater school achievement and delinquent behaviour. J Abnorm ChildPsychol 23:183Y199

Fombonne E (1999), The epidemiology of autism: a review. Psychol Med29:769Y786

Gillberg C, Persson M, Grufman M, Themner U (1986), Psychiatric




disorders in mildly and severely mentally retarded urban children andadolescents: epidemiological aspects. Br J Psychiatry 149:68Y74

Goodman R (1997), The strengths and difficulties questionnaire: a researchnote. J Child Psychol Psychiatry 38:581Y586

Goodman R (2001), Psychometric properties of the strengths and difficultiesquestionnaire. J Am Acad Child Adolesc Psychiatry 40:1337Y1345

Goodman R, Ford T, Corbin T, Meltzer H (2004), Using the Strengths andDifficulties Questionnaire (SDQ) multi-informant algorithm to screenlooked-after children for psychiatric disorders. Eur Child AdolescPsychiatry 13:II25YII31

Goodman R, Ford T, Simmons H, Gatward R, Meltzer H (2000), Using theStrengths and Difficulties Questionnaire (SDQ) to screen for childpsychiatric disorders in a community sample. Br J Psychiatry 177:534Y539

Hastings RP, Beck A, Daley D, Hill C (2005), Symptoms of ADHD andtheir correlates in children with intellectual disabilities. Res Dev Disabil26:456Y468

Kuntsi J, Eley TC, Taylor A et al. (2004), Co-occurrence of ADHD and lowIQ has genetic origins. Am J Med Genet 124B:41Y47

Mariani MA, Barkley RA (1997), Neuropsychological and academicfunctioning in preschool boys with attention deficit hyperactivitydisorder. Dev Neuropsychol 13:111Y129

McConaughy SH, Stanger C, Achenbach TM (1992), Three-year course ofbehavioral/emotional problems in a national sample of 4-to 16-year-olds: I. Agreement among informants. J Am Acad Child Adolesc Psychiatry31:932Y940

Meltzer H, Gatward R, Goodman R, Ford T (2000), Mental Health ofChildren and Adolescents in Great Britain. London: The Stationery Office

Pearson DA, Norton AM, Farwell EC (1997), Attention-deficit/hyper-activity disorder in mental retardation: nature of attention deficits. In:Attention, Development, and Psychopathology, Burack JA, Enns JT, eds.New York: Guilford, pp 205Y221

Pickles A, Dunn G (1988), Prevalence of disease, estimation from screening

data. In: The Encyclopedia of Biostatistics, Vol. 4 , Armitage P, Colton T,eds. New York: Wiley, pp 3484Y3490

Reiss S, Szyszko J (1983), Diagnostic overshadowing and professionalexperience with mentally retarded persons. Am J Mental Defic87:396Y402

Royston P, Altman DG (1994), Regression using fractional polynomials ofcontinuous covariates: parsimonious parametric modelling (with dis-cussion). Appl Stat 43:429Y467

Rucklidge JJ, Tannock R (2002), Neuropsychological profiles of adolescentswith ADHD: effects of reading difficulties and gender. J Child PsycholPsychiatry 43:988Y1003

Rutter M, Bailey A, Lord C (2003), The Social Communication Ques-tionnaire. Los Angeles: Western Psychological Services

Simonoff E, Pickles A, Chadwick O et al. (2006), The Croydon Assessmentof Learning study: prevalence and educational identification of mildmental retardation. J Child Psychol Psychiatry 47:828Y839

StataCorp (2003), Stata Statistical Software Release 8.0. College Station, TX:Stata Corporation

Stevenson J, Pennington BF, Gilger JW, DeFries JC, Gillis JJ (1993),Hyperactivity and spelling disability: testing for a shared geneticaetiology. J Child Psychol Psychiatry 34:1137Y1152

Stromme P, Diseth TH (2000), Prevalence of psychiatric diagnoses inchildren with mental retardation: data from a population-based study.Dev Med Child Neurol 42:266Y270

Tannock R (1998), Attention deficit hyperactivity disorder: advances incognitive, neurobiological, and genetic research. J Child PsycholPsychiatry 39:65Y99

Thorndike RL, Hagen E (1986), Cognitive Abilities Test. Windsor, UK:NFER-Nelson

Wechsler D (1992), The Wechsler Intelligence Scale for Children, ThirdEdition. London: The Psychological Corporation

Wechsler D (1996),Wechsler Quicktest. London: The Psychological Corporation

Grandmother and Parent Influences on Child Self-Esteem Judith S. Brook, EdD, Yuming Ning, PhD, Elinor B. Balka, BA,David W. Brook, MD, Erika H. Lubliner, BA, Gary Rosenberg, PhD

Objective: This study tests a model of intergenerational influences on childhood self-esteem that proposes paths fromgrandmothers’ drug problems to grandchildren’s self-esteem via parents’ drug problems and parental adaptive child rearing andfrom grandmothers’ maternal acceptance to grandchildren’s self-esteem via parents’ unconventionality and adaptive child rearing.Methods: This longitudinal study uses data obtained from interviews with a New York City sample of black and Puerto Ricanchildren (N = 149) and 1 of their parents and from mailed questionnaires or comparable interviews with those parents’ mothers.Structural equation modeling was used to test the proposed model. Results: The LISREL analysis found that, with 3 exceptions, allof the hypothesized paths were significant. The total effects analysis indicated that parents’ adaptive child rearing was the strongestlatent construct, a finding that was consistent with this construct’s proximal position in the model. Conclusions: This studysuggests that mothers’ drug problems are not just near-term risks for their children, but also pose long-term risks for theirchildren’s future functioning as parents and thereby for their grandchildren. The relative strength of parents’ adaptive childrearing in this intergenerational model indicates that this area should be the focus of therapeutic intervention efforts, butaddressing future grandmothers’ drug problems may have positive effects on multiple generations. Pediatrics2007;119:e444Ye451.

SIMONOFF ET AL.
