Shared genetic influences on ADHD symptomsand very low-frequency EEG activity: a twin

study

Charlotte Tye, Fruhling Rijsdijk, Corina U. Greven, Jonna Kuntsi,Philip Asherson and Grainne McLoughlin

MRC Social, Genetic and Developmental Psychiatry Centre, Institute of Psychiatry, King’s College London, UK

Background: Attention deficit hyperactivity disorder (ADHD) is a common and highly heritableneurodevelopmental disorder with a complex aetiology. The identification of candidate intermediatephenotypes that are both heritable and genetically linked to ADHD may facilitate the detection of sus-ceptibility genes and elucidate aetiological pathways. Very low-frequency (VLF; <0.5 Hz) electroenceph-alographic (EEG) activity represents a promising indicator of risk for ADHD, but it currently remainsunclear as to whether it is heritable or genetically linked to the disorder. Methods: Direct-current (DC)-EEGwas recordedduring a cognitive activation condition in30monozygotic anddizygotic adolescent twinpairs concordant or discordant for high ADHD symptom scores, and 37 monozygotic and dizygotic mat-ched-control twin pairs with low ADHD symptom scores. Structural equation modelling was usedto quantify the genetic and environmental contributions to the phenotypic covariance between ADHDand VLF activity. Results: Attention deficit hyperactivity disorder was significantly associated withreduced VLF power during cognitive activation, which suggests reduced synchronization of widespreadneuronal activity. Very low-frequency power demonstrated modest heritability (0.31), and the geneticcorrelation ()0.80) indicated a substantial degree of overlap in genetic influences on ADHD and VLFactivity. Conclusions: Altered VLF activity is a potential candidate intermediate phenotype of ADHD,which warrants further investigation of underlying neurobiological and genetic mechanisms.Keywords: ADHD, EEG, very low-frequency activity, endophenotype, genetics, heritability.

IntroductionAttention deficit hyperactivity disorder (ADHD) is oneof the most prevalent neurodevelopmental disorders,and is characterized by impairing levels of inatten-tive, impulsive and hyperactive symptoms. ADHDpersists beyond childhood in around 65% of cases,and is associated with high levels of clinical, psy-chosocial and economic burden (Faraone, Bieder-man, & Mick, 2006; Kendall, Taylor, Perez, & Taylor,2008). Family and twin studies suggest that ADHD isunder substantial genetic influence, with an averageheritability estimate of 0.76 (Faraone et al., 2005).Candidate gene association studies support the roleof genetic factors; however, these studies haveshown inconsistency and nonreplication of findings,which suggests a complex genetic inheritance with asmall risk conferred by individual genetic variants(Faraone et al., 2005).

One strategy to facilitate the detection of suscep-tibility genes and elucidate aetiological pathways isthe identification of neurobiological processes thatunderlie the disorder and potentially mediatebetween genes and behaviour (Gottesman & Gould,2003). To be useful for genetic research, theseintermediate phenotypes or endophenotypes, mustbe associated with the disorder, be heritable andshare genetic effects with the disorder (Gottesman &

Gould, 2003). Several candidate endophenotypeshave been reported for ADHD (Kuntsi, McLoughlin,& Asherson, 2006; McLoughlin, Kuntsi, Brandeis, &Banaschewski, 2005). Electrophysiological abnor-malities as measured using electroencephalography(EEG), in particular, are among the most promisingindicators of increased genetic risk for ADHD withconsistent associations and moderate-to-high heri-tability (McLoughlin et al., 2005; Tye, McLoughlin,Kuntsi, & Asherson, 2011).

Investigations of brain function in ADHD haverecently extended to include very low-frequencyactivity (VLF; <.05 Hz) that can be measured usingdirect-current (DC)-coupled EEG recordings. Spon-taneous VLF fluctuations synchronize activity acrossfunctionally specific, but widespread distributedneural networks, demonstrating a high degree ofcoherence within these circuits (Balduzzi, Riedner, &Tononi, 2008; Buszaki & Draguhn, 2004; Fransson,2005; Vanhatalo et al., 2004). It has been suggestedthat VLF activity represents a reflection of the brain’sdefault-mode network (DMN) that is typically char-acterized by a low-frequency BOLD signal (Fox et al.,2005; Sonuga-Barke & Castellanos, 2007). Specifi-cally, based on findings from fMRI studies, VLFfluctuations are posited to reflect toggling betweentwo anti-correlated brain networks: a task-negativenetwork that is active during wakeful resting statesand characterized by these slow oscillations, and theConflict of interest statement: No conflicts declared.

Journal of Child Psychology and Psychiatry *:* (2011), pp **–** doi:10.1111/j.1469-7610.2011.02501.x

� 2011 The Authors. Journal of Child Psychology and Psychiatry � 2011 Association for Child and Adolescent Mental Health.Published by Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main St, Malden, MA 02148, USA

task-positive network that activates during goal-oriented activity (Fox et al., 2005; Sonuga-Barke &Castellanos, 2007). In addition, EEG studies havereported that VLF activity modulates the activity ofhigher frequency bands, suggesting regulation ofgross cortical excitability (Monto, Palva, Voipio, &Palva, 2008; Vanhatalo et al., 2004). During anactive task condition, slow cortical activity can alsobe generated in response to stimuli, and as suchreflect event-related rather than spontaneous activ-ity. These time-locked slow cortical potentials havebeen posited as an index of conscious perception (He& Raichle, 2009).

Abnormalities in VLF activity are associated withseveral neuropathological disorders (Broyd et al.,2009). Adults with attention problems demonstratedreduced spontaneous VLF power and reduced rest-to-task VLF attenuation (Helps, Broyd, James, Karl,& Sonuga-Barke, 2009; Helps, James, Debener, Karl,& Sonuga-Barke, 2008), which was replicated in asample of adolescentswithADHD (Helps et al., 2010).In addition, reduced rest-to-taskVLF attenuation hasbeen associated with poor task performance (Helpset al., 2010) and similar slow fluctuations in taskperformance (Helps et al., 2009; Monto et al., 2008),which may underlie the deficits in task performanceexhibited in ADHD (Castellanos et al., 2005). This issupportive of the default-mode interference hypoth-esis, which proposes that VLF activity, usuallyexhibited at rest, persists during cognitive activationin ADHD producing periodic lapses of attention (Foxet al., 2005; Sonuga-Barke & Castellanos, 2007).Measurement of VLF activity during a cognitive taskallows investigation of this brain-behaviour associa-tion and its persistence during activation states thatare altered in ADHD (Sergeant, 2000).

Twin research conducted on frequency bandsabove 1 Hz suggests that EEG is highly heritable(Smit, Posthuma, Boomsma, & Geus, 2005). A pre-liminary study of affected sibling pairs with ADHDindicated high sibling correlations of 0.53–0.76 dur-ing a cognitive activation condition (Loo & Smalley,2008) although smaller estimates (0.22–0.61) havebeen reported in a larger study of multiplex familieswithADHD (Loo et al., 2010). These studieshaveusedfamily designs that are unable to discriminatebetween genetic and environmental influences. Thetwin design, however, allows separation of theseeffects by utilizing the different levels of geneticrelatedness between monozygotic (MZ; 100%) anddizygotic (DZ; 50%) twinpairs (Neale&Cardon, 1992).

No twin study on VLF activity or its genetic overlapwith ADHD has been conducted to date. This studyaims to evaluate VLF activity as a potential inter-mediate phenotype of ADHD during a key develop-mental period by (a) estimating the heritability ofVLF activity, (b) quantifying the strength of thephenotypic relationship of VLF activity with ADHDsymptoms, and (c) examining the genetic andenvironmental overlap with ADHD symptoms. We

measured VLF activity during a cognition activationcondition [cued continuous performance test (CPT-OX)]; (Doehnert, Brandeis, Straub, Steinhausen, &Drechsler, 2008; McLoughlin et al., 2010; Valkoet al., 2010) in adolescent MZ and DZ twin pairs (12–15 years old) concordant and discordant for highand low ADHD symptom scores. Structural equationmodelling was applied, enabling separation of thephenotypic covariance between these two parame-ters (i.e. ADHD and VLF activity) into genetic andenvironmental components (Toulopoulou et al.,2007). Significant heritability and genetic overlapbetween ADHD and VLF activity would support thismeasure as a candidate endophenotype for the dis-order, reflecting mediating processes on ADHD(Gottesman & Gould, 2003) or pleiotropic effects ofgenes (Kovas & Plomin, 2006).

MethodsSample

The sample was selected from the Twins’ Early Devel-opment Study (TEDS), a birth cohort study of all twinsborn in England and Wales between 1994 and 1996(Trouton, Spinath, & Plomin, 2002). Zygosity wasdetermined using a zygosity questionnaire that hasbeen shown to have 95% accuracy (Price et al., 2000);For cases where zygosity remains unclear from thisquestionnaire, DNA testing was conducted. The TEDSsample is representative of the general population interms of parental education, ethnicity and employmentstatus (Kovas, Haworth, Dale, & Plomin, 2007).

The Neurophysiological Study of Activity and Atten-tion in Twins (NEAAT) subset used in this study con-sisted of 67 male twin pairs in groups of 22 pairsconcordant for high levels of ADHD symptoms (MZ: 11;DZ: 11), eight pairs discordant for ADHD symptoms(MZ: 2; DZ: 6) and 37 control pairs concordant for lowlevels of ADHD symptoms (MZ: 21; DZ: 16). Twin pairswere selected based on an analysis of symptom devel-opment over time using the program MPLUS (Data S1).This identified sub-groups of individuals who have beenstably high or stably low at ages 8, 12 and 14, using the18 DSM-IV ADHD items from the Long Version of theConners’ Parent Rating Scale (Conners, Sitarenios,Parker, & Epstein, 1998a). This ensured that theselected twin pairs were consistently concordant ordiscordant for high levels of ADHD symptoms (corre-sponding to a clinical diagnosis) or unaffected controlswho had consistently low ADHD symptoms. Demo-graphic characteristics are given in Table 1. Partici-pating families gave their written informed consent, andthe study was approved by King’s College LondonPsychiatry, Nursing and Midwifery Research EthicsSub-Committee (PNM/08/09-89).

Task and stimuli

The cued-CPT (flanker version; (Doehnert et al., 2008;McLoughlin et al., 2010, 2011; Valko et al., 2010)consists of a black letter array formed of a centre letterflanked on each side by distractor letters, presented in

2 Charlotte Tye et al. J Child Psychol Psychiatry 2011; *(*): **–**

� 2011 The Authors. Journal of Child Psychology and Psychiatry � 2011 Association for Child and Adolescent Mental Health.

four identical blocks of 100 letter arrays each. Patientswere instructed to ignore the distractor letters and at-tend only to the centre letter. There were 11 differentcentre letters (O, X, H, B, C, D, E, F, G, J and L) sub-tending approximately 0.5 degrees. Centre letters ‘X’and ‘O’ were flanked by the incompatible letter ‘O’ or ‘X’,

and distractor letters were flanked by either ‘X’ or ‘O’.The letter arrays were presented briefly (150 ms) every1.65 s in a pseudo-random sequence at the centre of acomputer monitor, at the viewing distance of 120 cm.Duration of the task was 11 min. Subjects were seatedon a height-adjustable chair in a video-monitored test-ing cubicle. They were instructed to respond only tocue-target sequences (i.e. XOX-OXO) by pressing abutton as quickly as possible with the index finger oftheir preferred hand. The task was practised and com-prehension ascertained prior to task performance. Ifnecessary, participants were told to minimize eyemovements or blinks. Participants completed 6 min ofresting EEG prior to completing the cued CPT-OX. Thetask was followed by other tasks not reported here.

Scoring overt performance

Performance measures in the cued-CPT task includedtarget reaction time (MRT, mean latency of respondingin ms after target onset), within-subject variability inreaction times (SD-RT) and the coefficient of reactiontime variability (CV, SD-RT/MRT). MRT and SD-RTwere calculated across correctly answered target trials.Hits were characterized by target-OXOs that were de-tected between 200 and 1500 ms after stimulus onset.False alarms were responses to letters other than tar-get-OXO. Errors were broken down into subcategories(omission errors, total commission errors, and O-not-Xcommission errors).

Other measures

General cognitive ability was assessed at age 14 as partof ongoing TEDS web-based data collection. The twinswere tested on the WISC-III-PI vocabulary multiplechoice subtests (Wechsler, 1992) and Raven’s standardand advanced progressive matrices (Raven, Court, &Raven, 1996). Missing scores (Table 1) were imputedfrommultiple IQ subtest scores across ages 7, 12 and 14using the ICE command in Stata version 10 statisticalsoftware (Stata Corp, College Station, TX) for each out-come variable. A g score was created with equal weightsfor the two tests by summing their standardized scoreswithin the NEAAT sample. Further information about gas measured in TEDS can be found elsewhere (e.g.(Haworth et al., 2010). Measures of g and IQ correlatehighly and both provide an index of general intelligence(Jensen, 1998), and as such, IQ can be calculated fromstandardized g using the formula IQ = (g · 15) + 100.

EEG recording and analysis

Electroencephalography was recorded using 62 chan-nel DC-coupled recording system (extended 10–20montage). Electrode impedance was kept below 5 kW.The reference electrode was positioned at FCz. Verticaland horizontal electro-oculograms (EOGs) were simul-taneously recorded from electrodes above and below theleft eye and at the outer canthi. The signal was digitizedat 500 Hz sampling rate, stored and analysed offline.

Data were analysed in Brain Vision Analyzer (2.0)(Brain Products, Munich, Germany). The signal wasre-referenced offline to the average reference, and

Table 1 Raw scores and mean comparisons of demographiccharacteristics adjusted for genetic relatedness

Mean SD

All ADHDversus allcontrols

t p

Age (years)MZ Control (n = 46) 14.45 1.05 )2.72 <.001DZ Control (n = 35) 14.56 0.69MZ ADHD (n = 22) 14.01 0.65DZ ADHD (n = 31) 14.03 0.75

IQa (n = Raven’s matrices, WISC-III-PI multiple choicevocabulary subtests)MZ Control (n = 27, 28) 103.28 13.82 )2.29 .04DZ Control (n = 21, 23) 103.51 12.93MZ ADHD (n = 6, 6) 91.88 16.15DZ ADHD (n = 7, 8) 96.80 15.13

Parent Connersb

DSM-IV symptom subscaleInattentionMZ Control (n = 46) 43.00 3.42 12.60 <.001DZ Control (n = 36) 42.17 2.49MZ ADHD (n = 23) 57.39 8.49DZ ADHD (n = 28) 54.18 9.10

Hyperactivity/impulsivityMZ Control (n = 46) 44.65 2.69 10.74 <.001DZ Control (n = 36) 43.92 1.46MZ ADHD (n = 23) 62.43 11.81DZ ADHD (n = 28) 59.50 12.84

Teacher Connersc

DSM-IV symptom subscaleInattentionMZ Control (n = 27) 46.00 6.65 5.57 <.001DZ Control (n = 21) 46.57 6.75MZ ADHD (n = 11) 64.55 10.97DZ ADHD (n = 15) 51.87 11.35

Hyperactivity/impulsivityMZ Control (n = 27) 48.00 8.29 1.86 .07DZ Control (n = 21) 46.33 5.29MZ ADHD (n = 11) 57.36 9.11DZ ADHD (n = 15) 49.27 7.67

ADHD, attention deficit hyperactivity disorder.aGeneral cognitive ability was assessed at age 14 as part ofongoing TEDS web-based data collection. The twins weretested on the WISC-III-PI vocabulary multiple choice subtests(Wechsler, 1992) and Raven’s standard and advanced pro-gressive matrices (Raven et al., 1996). Missing scores wereimputed from multiple IQ subtest scores across ages 7, 12 and14 using the ICE command in Stata for each outcome variable.A g score was created with equal weights for the two tests bysumming their standardized scores within the NEAAT sample.Further information about g as measured in TEDS can befound elsewhere (e.g. (Haworth et al., 2010). Measures of g andIQ correlate highly and both provide an index of generalintelligence (Jensen, 1998), and as such IQ can be calculatedfrom standardized g using the formula IQ = (g · 15) + 100.bLong Version of the Parent Conners’ Rating Scale T scores(Conners et al., 1998a) collected on the day of testing.cLong Version of the Teacher Conners’ Rating Scale T scores(Conners, Sitarenios, Parker, & Epstein, 1998b). Teacherswere contacted following completion of the testing session.

doi:10.1111/j.1469-7610.2011.02501.x Very low-frequency EEG intermediate phenotypes of ADHD 3

� 2011 The Authors. Journal of Child Psychology and Psychiatry � 2011 Association for Child and Adolescent Mental Health.

downsampled to 256 Hz. We applied 0.02–4 Hz (12dB/Oct) Butterworth filters. Ocular artefacts were removedfrom the data using biased infomax Independent Com-ponent Analysis (ICA (Jung et al., 2000) The extractedindependent components were manually inspected,and ocular artefacts were removed by back-projection ofall but those components. Continuous EEG was seg-mented into 50 s epochs. Segments with artefactsexceeding 120 lV peak-to-peak in any channel wererejected. A DC-Detrend command was executed to re-move linear drifts from the data. Fast-Fourier Trans-form analysis was performed on the data from each ofthe electrodes for each participant. Fifty second Han-ning windows were used, and power was calculated.The current study focused on the frequency band 0.02–0.2 Hz that has been implicated previously in ADHD(Helps et al., 2010). To allow comparison with otherstudies investigating familial effects on higher fre-quency bands in ADHD (Loo et al., 2010), VLF powerwas compared across frontal (Fz, F3, F4), central (Cz,C3, C4), parietal (Pz, P3, P4) and, in addition, occipital(Oz, O1, O2) scalp electrode locations.

Statistical analysis

Two participants were excluded from analysis due toexcessive artefact (DZ control) and extreme commissionerrors (n = 37) indicative of insufficient task engage-ment (MZ ADHD).

For the analysis of performance data, the effects ofage and general intelligence were regressed out of thedata using Stata. All measures had pronounced heter-ogeneity of variance and skewed distributions, and werelog-transformed (optimized minimal skew through thelnskew0 command in Stata).

Preparation of EEG data prior to model fitting. Theeffects of age and general intelligence were regressedout of the data using Stata due to significant associa-tions with both ADHD (Table 1) and VLF activity (ager = )0.12 to )0.19; g r = )0.26 to )0.33). Althoughgeneral cognitive ability is confounded by the presenceof psychiatric disorder, given that age and IQ are highlypredictive of cognitive performance, regressing outthese effects before twin modelling ensures straight-forward interpretation; any reported associationsbetween VLF activity and ADHD are independent ofgeneral cognitive ability and neurodevelopmentalchanges. Simultaneous analyses of dichotomous andcontinuous data could not be performed in the Mxprogram, and so both ADHD, which was scored as adichotomous attribute, and the VLF measure, whichwas scored as a continuous variable, were modelled asthreshold traits. Accordingly, each VLF measure wasordinalized into five equal classes in terms of propor-tions, which should capture most of the information inthe continuous data. To provide a normal distributionfor a more successful ordinal cut and to fulfil twinmodel assumptions, outliers were removed separatelyfor each VLF parameter (±3 SD from the mean; finalnumbers shown in Table 3) and log-transformed (opti-mized minimal skew through the lnskew0 command inStata). The Mx software for structural equation model-ling (Neale, Boker, Xie, & Maes, 1999) was then used toestimate polychoric correlations and genetic model

parameters using maximum likelihood statistics, whilecorrecting for the selected nature of the sample.

Comparison of means. The comparisons of meanvalues were analysed by means of a regression com-mand in Stata that allows for nonindependent obser-vations (e.g. twin pairs) using a robust clustercommand to estimate standard errors. The associationbetween the VLF activity and ADHD symptom scores,and VLF activity and performance measures, wasinvestigated using Pearson’s product moment correla-tion coefficient on transformed residuals. Performancemeasures for correlation analysis were selected on thebasis of trends towards case-control differences.

Twin correlations. Twin correlations between VLFactivityandADHDwereestimatedbyfittingaconstrainedcorrelational model to the observed MZ and DZ data toproduce (a) one overall within-twin across-trait correla-tionregardlessofzygosity, i.e.betweenADHDandVLF,(b)the MZ and (c) the DZ cross-twin within-trait correlationfor VLF, and (d) the MZ and (v) DZ cross-twin cross-traitcorrelation, by comparing one twin’sADHDscore and theco-twin’s VLF score. The MZ and DZ cross-twin correla-tionswerefixedaccordingtothepointestimatesforADHDderived from the heritability estimate of a meta-analysis(rMZ = 0.76, rDZ = 0.38; (Faraone et al., 2005), due tothe uncertain ascertainment process for twins, concor-dant and discordant for ADHD symptoms [see belowAccounting for Selection; (Toulopoulou et al., 2007)].

Genetic model fitting. A more sophisticated approachto the analysis of twin data is using structural equationmodels that model correlations between variableswithin individuals and across twins using relationshipsbetween observed and latent variables (Rijsdijk & Sham2002). Liability-threshold models were used for bothADHD and VLF variables (Falconer & Mackay, 1996).The liability-threshold models for the dichotomizedADHD phenotype (affected versus nonaffected) assumethat risk is normally distributed on a continuum andthat the disorder occurs only when a certain thresholdis exceeded (Neale & Kendler, 1995). Both affected andunaffected individuals were assumed to be part of thesame distribution of liability to the disorder, with eachindividual being either below or above the threshold.

Through examination of the differences in correla-tions between MZ and DZ twin pairs, the variance ofVLF activity can be decomposed into additive genetic(A), shared environmental (C) and unique environmen-tal (E) components to estimate (a) the heritability of VLFactivity, and (b) genetic and environmental correlationsof ADHD with VLF activity. The applied ACE bivariatecorrelated factors liability-threshold model is illustratedin Figure 1. The parameter estimates from the modelcan be used to estimate the correlation between thegenetic factors for ADHD and VLF activity (rg), which isan index of shared genetic effects between theseparameters, and similarly for correlations of uniqueenvironmental factors (re). As the rg and re correlationsdo not take into account the heritability of either trait, itis possible for a large genetic correlation to actuallyexplain a very small portion of the observed covariationbetween these two traits. Combining the informationfrom the rg and re with the heritabilities of each trait, we

4 Charlotte Tye et al. J Child Psychol Psychiatry 2011; *(*): **–**

� 2011 The Authors. Journal of Child Psychology and Psychiatry � 2011 Association for Child and Adolescent Mental Health.

can establish the genetic (rph-a) and unique environ-mental (rph-e) contributions to the total phenotypiccorrelation (rph) between ADHD and VLF activity(Toulopoulou et al., 2007).

Accounting for selection. As the data are from twinsselected for high or low ADHD symptoms scores, ratherthan a random sample, the heritability of ADHD cannotbeestimated.Selectedsamplesaremoreefficientandcanbe more powerful when studying low prevalence disor-ders (Neale, Eaves, & Kendler, 1994), but model fittinganalyses will usually require an ascertainment correc-tion. As selection is based on ADHD and blind to VLFvalues, the required ascertainment correction willdepend only on the model for ADHD, and therefore themodel parameters for ADHD were fixed to constant val-ues supported by ameta-analysis of 20 studies of ADHD(model 1: h2 = 0.76, c2 = 0, e2 = 0.24; (Faraone et al.,2005); univariate twinmodelling ofDSM-IV basedADHDscoreson theConners’ scoresatage8 in theTEDSsample(model 2: h2 = 0.89, c2 = 0, e2 = 0.11 (Ronald, Simonoff,Kuntsi, Asherson, & Plomin, 2008) and at age 12 in theTEDS sample (model 3: h2 = 0.73, c2 = 0.13, e2 = 0.14;C. Greven, unpublished data). In these models, the var-iance components for VLF activity, as well as their rela-tionship with ADHD, are free parameters to be estimatedfrom the data (Figure 1). The different parameter esti-mates for ADHD from the different models had no effecton the variance components for VLF activity, and thus,we only report those of the first model. In addition, wefixed the ADHD prevalence rate to a lifetime risk of 5%(Polanczyk, de Lima, Horta, Biederman, &Rohde, 2007).

ResultsResults from the regression and correlation analyses

Regression analyses indicated no significant differ-ences between ADHD and control twins for cognitiveperformance measures (Table 2). VLF power was

highest in central regions and was significantlyreduced for central and parietal locations in theADHD group, with the strongest association withcentral locations (Table 3). Therefore, to reduce thetotal number of variables and subsequent multipletesting bias VLF power at central regions was used incorrelations between symptom scores and perfor-mance measures, and in twin modelling analyses tominimize heteroscedascity. Significant associationsbetween VLF power and symptom scores were foundin the control sample only (Table 4), suggesting thatincreased symptoms of inattention and hyperactiv-ity/impulsivity are associated with increased VLFpower in typically developing adolescents. ReducedVLF activity was significantly associated with in-creased response variability in the ADHD group only(Table 5). All other associations were nonsignificant.

Results from the twin modelling analyses

The MZ cross-twin within-trait correlation for VLFpower (r = 0.37; 95% CI, 0.002 to 0.64) was greater

A1 A2 C2

E1 E2

ADHD VLF activity

a1

e1 e2

a2 c2

rg

re

√.76

√.24

Figure 1 Constrained correlated factors bivariate model forADHD and VLF activity. Circles represent latent additive genetic(A), shared environmental (C) and nonshared environmental (E)factors. a1, a2, e1, e2, c2 represent genetic and environmentalpath estimates. r g and r e represent genetic and environmentalcorrelations between ADHD and VLF activity. Parameters forADHD (heritability and unique environmental estimates) arefixed values according to meta-analysis (Faraone et al., 2005)

Table 2 Summary statistics of raw scores and mean compar-isons adjusted for genetic relatedness for task performancemeasures based on age- and IQ-regressed scores

Mean SD

All ADHDversus allcontrols

t p

MRTMZ Control (n = 46) 409.63 59.26 1.27 .21DZ Control (n = 36) 379.61 48.29MZ ADHD (n = 23) 441.30 75.18DZ ADHD (n = 28) 407.93 56.01

SD-RTMZ Control (n = 46) 93.33 42.58 1.88 .06DZ Control (n = 36) 78.56 30.18MZ ADHD (n = 23) 128.09 64.38DZ ADHD (n = 28) 102.00 38.91

CVMZ Control (n = 46) 0.22 .08 1.95 .06DZ Control (n = 36) 0.20 .07MZ ADHD (n = 23) 0.28 0.12DZ ADHD (n = 28) 0.25 .09

Commission errors (n)MZ Control (n = 46) 2.59 2.60 1.21 .22DZ Control (n = 36) 3.92 3.49MZ ADHD (n = 23) 4.78 4.35DZ ADHD (n = 28) 3.79 4.15

O-not-X commission errors (n)MZ Control (n = 46) 1.89 2.58 )0.69 .49DZ Control (n = 36) 2.36 2.24MZ ADHD (n = 24) 2.74 3.40DZ ADHD (n = 28) 2.39 3.02

Omission errors (n)MZ Control (n = 46) 0.67 0.84 1.43 .16DZ Control (n = 36) 0.94 1.35MZ ADHD (n = 22) 1.87 1.98DZ ADHD (n = 28) 1.82 2.58

ADHD, attention deficit hyperactivity disorder; DZ, dizygotic;CV, coefficient of variation (SD-RT/MRT); MRT, mean reactiontime in milliseconds; MZ, monozygotic; SD-RT, within-subjectvariability in RTs in milliseconds.

doi:10.1111/j.1469-7610.2011.02501.x Very low-frequency EEG intermediate phenotypes of ADHD 5

� 2011 The Authors. Journal of Child Psychology and Psychiatry � 2011 Association for Child and Adolescent Mental Health.

than the DZ cross-twin within-trait correlation forVLF power (r = 0.21; 95% CI, )0.21 to 0.56) whichsuggests that genetic effects contribute to VLF pow-er. As the MZ correlation for VLF power deviated from1, this suggests that unique environmental influ-

ences contribute substantially to VLF power. Fur-thermore, the DZ cross-twin within-trait correlationfor VLF power was slightly more than half the MZcorrelations, suggesting the presence of sharedenvironmental effects. The MZ cross-twin cross-traitcorrelation between ADHD and VLF power(r = )0.42; 95% CI, )0.65 to )0.09) was greater thanthe DZ cross-twin cross-trait correlation (r = )0.14;95% CI, )0.39 to 0.13), suggesting that genetic ef-fects contribute to the association between reducedVLF power and ADHD. The DZ cross-trait cross-twincorrelation is less than half the MZ correlation,suggesting that nonadditive genetic dominanceeffects may contribute to this overlap.

Structural equation modelling indicated thatgenetic factors accounted modestly for the totalvariation in VLF power (h2 = 0.31; 95% CI, 0.01–0.64). Shared environment did not significantly ex-plain individual differences in VLF power (c2 = 0.06;95% CI, 0–0.44), whereas unique environmental ef-fects (incorporating also measurement error) ac-counted for a moderate part of the variance in VLFpower (e2 = 0.63; 95% CI, 0.38–0.92).

The extent to which ADHD and VLF activity sharethe same genetic and unique environmental effects isgiven by the correlations rg and re, respectively.Significant genetic correlations (rg = )0.80; 95% CI,)1.00 to )0.15) indicated substantial genetic overlapbetween ADHD and reduced VLF power. Uniqueenvironmental effects did not have a significantoverlap with ADHD (re = 0.41; 95% CI, )0.61 to0.96). The negative phenotypic correlation (rph)suggested that increased liability to ADHD wasassociated with reduced VLF power ()0.23; 95% CI,)0.44 to )0.01). Due to the moderate to high herit-abilities and moderate genetic correlation, the phe-notypic correlation between ADHD and VLF powerappears to be largely attributable to genetics (rph-a= )0.39; 95% CI, )0.60 to )0.08). Unique environ-mental contributions to phenotypic variancewere nonsignificant (rph-e = 0.16; 95% CI, )0.24 to0.42).

DiscussionThis study aimed to evaluate very low-frequency(VLF) neuronal activity as an intermediate phenotypeof ADHD in a sample of adolescent monozygotic anddizygotic twin pairs concordant and discordant forADHD symptoms. Genetic analyses showed that VLFactivity demonstrates modest heritability, with noevidence of significant shared environmental effects.Structural equation modelling revealed a significantphenotypic association between high ADHD symp-tom scores during adolescence and reduced VLFactivity during cognitive activation. Genetic factorswere the main source of this association, and uniqueenvironmental factors were not significant. This isthe first study to support VLF activity as an electro-physiological marker of genetic risk in ADHD.

Table 3 Summary statistics and mean comparisons adjustedfor genetic relatedness for VLF power based on transformedage- and IQ-regressed scores with extreme outliers removed

Mean SD

All ADHDversus allcontrols

t p

VLF frontalMZ Control (n = 45) 4.36 0.36 )1.88 .07DZ Control (n = 35) 4.36 0.40MZ ADHD (n = 21) 4.10 0.45DZ ADHD (n = 27) 4.28 0.48

VLF centralMZ Control (n = 46) 5.29 0.18 )3.03 .003DZ Control (n = 35) 5.28 0.19MZ ADHD (n = 20) 5.17 0.14DZ ADHD (n = 26) 5.18 0.17

VLF parietalMZ Control (n = 46) 4.92 0.27 )2.25 .03DZ Control (n = 35) 4.87 0.23MZ ADHD (n = 21) 4.79 0.33DZ ADHD (n = 27) 4.74 .27

VLF occipitalMZ Control (n = 46) 4.34 0.46 1.99 .05DZ Control (n = 36) 4.28 0.37MZ ADHD (n = 21) 4.20 0.51DZ ADHD (n = 26) 4.10 0.38

ADHD, attention deficit hyperactivity disorder; DZ, dizygotic;MZ, monozygotic; VLF, very low-frequency power.

Table 4 Correlations between VLF power at central scalplocations and symptoms of ADHD (Pearson’s product momentcorrelation on transformed age and IQ-regressed scores)

Control MZ DZ ADHD MZ DZ

Inattentiona 0.41** 0.19 0.70** )0.01 0.04 )0.05Hyperactivity/impulsivitya

0.44** 0.34* 0.71** )0.09 )0.22 0.02

*p < .05, **p < .01ADHD, attention deficit hyperactivity disorder; DZ, dizygotic;MZ, monozygotic.aADHD symptom scores based on the Long Version of theParent Conners’ Rating Scale (Conners et al., 1998a) collectedon the day of testing.

Table 5 Correlations between VLF power at central locationsand performance measures (Pearson’s product moment cor-relation on transformed age and IQ-regressed scores)

Control MZ DZ ADHD MZ DZ

SD-RT 0.07 0.02 0.11 )0.21 )0.46* 0.20CV 0.08 0.06 0.09 )0.28* )0.49* 0.04

*p < .05.ADHD, attention deficit hyperactivity disorder; DZ, dizygotic;CV, coefficient of variation (SD-RT/MRT); MRT, mean reactiontime in milliseconds; MZ, monozygotic; SD-RT, within-subjectvariability in RTs in milliseconds.

6 Charlotte Tye et al. J Child Psychol Psychiatry 2011; *(*): **–**

� 2011 The Authors. Journal of Child Psychology and Psychiatry � 2011 Association for Child and Adolescent Mental Health.

Heritability of VLF activity during a cognitiveactivation condition is consistent with reports of highheritability (Smit et al., 2005) and high sibling sim-ilarity in an ADHD sample (Loo et al., 2010) in higherfrequency bands. This suggests that VLF activity hasa genetic basis combined with a moderate contribu-tion from unique environmental effects. The sub-stantial genetic correlation between ADHD and VLFactivity indicates that they are substantially influ-enced by the same genes, supporting VLF activity asa putative intermediate phenotype of the disorder.This finding is an important step in understandingthe neurobiological pathways involved in the disor-der, and potentially to facilitate in the detection ofsusceptibility genes. For example, findings suggestthat a catecholaminergic deficiency underliesabnormal VLF oscillations, which is also widely re-ported in ADHD (Castellanos et al., 2005). Indeeddopamine reuptake inhibitors (e.g. methylpheni-date), which are used as treatment for ADHD, havebeen found to modulate slow oscillations in sub-cortical structures (Ruskin et al., 2001). Causal testsof mediation are necessary to identify if these elec-trophysiological markers mediate aetiological effectson ADHD, rather than pleiotropic (or epiphenome-nal) processes (Walters & Owen, 2008).

The phenotypic association between ADHD andVLF activity is consistent with previous studiesreporting reduced VLF power at rest in ADHD (Helpset al., 2008, 2010). In addition, previous studies re-port an association between increased rest-to-taskVLF attenuation and a higher number of inattentionsymptoms in typical adults (Helps et al., 2009) and aclinical ADHD group (Helps et al., 2010). In thepresent study, higher VLF activity was associatedwith increased levels of inattention and hyperactiv-ity/impulsivity in the control group. Such discrep-ancies between control and ADHD participants mayreflect neuropathological differences betweengroups, and also the robustness of the longitudinalmethod employed for group selection compared withinvestigating symptom scores at the single time-point of data collection.

Overall, the findings for both groups supportspontaneous VLF activity, typically associated withresting periods as exhibited by slow fluctuations ofthe BOLD signal, as present during cognitive acti-vation, suggesting that it represents a continuousprocess that is present during both resting and goal-oriented periods (Fransson, 2006). VLF activity areproposed to be an index of information integration orfunctional connectivity through synchronizationover widely distributed neuronal networks (Biswal,Yetkin, Haughton, & Hyde, 2005; He & Raichle,2009). The association between ADHD and reducedVLF power in this study may therefore reflectreduced synchronization of these widespread cir-cuits. As VLF activity is proposed to relate to theDMN, one of several widely distributed networks asdefined by fMRI (Sonuga-Barke & Castellanos,

2007), these findings suggest that individuals withADHD demonstrate reduced DMN synchronizationduring cognitive activation or abnormal togglingbetween the anti-correlated introspective task-negative and extroceptive task-positive networks,producing impairment to the functions they eachserve (Fox et al., 2005; Fransson, 2005; Sonuga-Barke & Castellanos, 2007).

The measurement of VLF activity during this cog-nitive task may include both spontaneous VLFactivity and event-related VLF slow cortical poten-tials time-locked to the cue stimulus (the contingentnegative variation). Although in this particular studythere is limited evidence for a relationship betweenspontaneous VLF activity and event-related VLFactivity (see Data S2), it is still important that futurework considers this relationship, particularlybecause without correction for multiple testing,there is a small yet significant correlation betweenthese measures. Along with the significant correla-tion between VLF activity and delta activity, suchphenotypic overlap may suggest that the geneticcontributions reported may be shared across fre-quency bands. In accordance, VLF phase has beenshown to correlate with the magnitude of higherfrequency bands suggesting a modulation of grosscortical excitability (Monto et al., 2008; Vanhataloet al., 2004). Further work investigating phaselocking with higher frequency bands as a measure ofsynchronization will increase our understanding ofthe mechanisms underlying this overlap, particu-larly when used in genetically sensitive designs.

Advances in EEG source localization techniquesare likely to provide further insight into the rela-tionship between the BOLD signal and EEG; recentwork using sLORETA identified differential localiza-tion of rest-task VLF attenuation dependent onADHD status, in the absence of case-control differ-ences in rest-task attenuation itself (Broyd, Helps, &Sonuga-Barke, 2011). This highlights potentialtopographical and source differences between con-trol and clinical populations. In accordance withthis, in the current study, we identified group dif-ferences in VLF activity at central and parietal scalplocations only, and genetic effects emerged primarilyfor central scalp locations. Localization differencesmay reflect the measure of VLF activity used (i.e. acomparison of VLF activity between rest and task, asopposed to VLF activity during the task itself) andtask-specific demands that particular brain pro-cesses/regions subserve. Examination of changesover time and space in VLF activity in well-validated,and robust EEG/ERP paradigms is required.

For measures of task performance, phenotypicassociations with ADHD were nonsignificant. This iscontrary to findings reporting slower and morevariable reaction times in ADHD (Klein, Wendling,Huettner, Ruder, & Peper, 2006; Kuntsi et al., 2010)and increased omission errors (Willcutt, Doyle, Nigg,Faraone, & Pennington, 2005). The lack of a group

doi:10.1111/j.1469-7610.2011.02501.x Very low-frequency EEG intermediate phenotypes of ADHD 7

� 2011 The Authors. Journal of Child Psychology and Psychiatry � 2011 Association for Child and Adolescent Mental Health.

difference in the performance data is likely to be dueto the use of the CPT-OX with flankers that was notspecifically designed to optimally measure cognitiveperformance. Another possible reason for our lack ofcognitive performance findings is our use of a popu-lation sample, rather than a clinical sample, albeitone selected for low and high ADHD symptom scores.One study reported increased RT variability in apopulation sample using ADHD symptom scores(similar to the study design here), but the data werecollected at a younger age, in a larger sample andusing task conditions that maximise RT variability,such as a slow event rate and lack of rewards (Kuntsi,Wood, Van derMeere, & Asherson, 2009). The greaterassociation between ADHD and neurophysiologicalmarkers compared with task performance suggestsEEG as a more sensitive index of genetic loading forADHD than the cognitive performance measures, atleast as derived from the task used in this study.

According to the default-mode interferencehypothesis, a failure to fully attenuate VLF activity incognitive activation conditions contributes to atten-tion lapses associated with ADHD (Sonuga-Barke &Castellanos, 2007). VLF activity was not associatedwith task performance measures in this study acrossthe whole sample, but when collapsing the sampleby ADHD status and zygosity, significant associa-tions were found between reduced VLF power andincreased response variability in the ADHD grouponly. This suggests that reduced synchronizationduring a task is associated with poor performance inADHD, and is in accordance with previous studiesthat report associations between rest-task attenua-tion and increased errors and variability (Helpset al., 2010). Further work incorporating analysis oftrial-to-trial changes in the synchronicity betweentask performance and VLF activity is required(Castellanos et al., 2005; Helps et al., 2009), andlinking this to optimal measures of RT variability willhelp to clarify this association.

Certain limitations must be taken into consider-ation. First, the sample size is relatively small for atwin study, such that although the estimates aremodest to large, the confidence intervals aroundthese estimates are also large introducing an overlapin the confidence intervals for MZ and DZ twins. Thismight suggest that the contribution from genetics isnot significant; nevertheless, combining the signifi-cant MZ correlations with the significant mean cor-relation differences suggests a familial nature that islikely to be of a genetic basis based on the fixedparameters used. This limited statistical power alsorestricted the estimation of shared environmental (assuggested by cross-twin within-trait correlations forVLF power) and nonadditive genetic influences (assuggested by cross-twin cross-trait correlationsbetween ADHD and VLF power). We acknowledgethat such influences may have contributed to theestimates of heritability and genetic overlap betweenADHD and VLF activity in our study, although the

genetic contribution presented is not overestimated.The substantial contribution of unique environmen-tal influences on VLF activity may reflect error orinstability in the measure. Reliability sets an upperlimit on the estimates of heritability; any deviationsfrom perfect reliability will increase measurementerror and therefore unique environmental influences(Kuntsi et al., 2006). Studies of test–retest reliabilityare required to investigate the stability of both phe-notypic associations and genetic influences on VLFactivity (de Geus, 2002). In addition, the imputationof a relatively substantial proportion of general cog-nitive ability as a potential confounding influencemay introduce further error arising from predictedscores, although the use of a robust multiple-impu-tation method is likely to have reduced this errorvariance. As we do not know whether the underlyingcauses of altered VLF activity are shared or disorder-specific, particularly as abnormalities are reported inseveral other neuropathological disorders such asautism and schizophrenia (Broyd et al., 2009), it isimportant that future work directly compares VLFactivity profiles across disorders.

In conclusion, applying genetic model fitting forthe first time to VLF neuronal activity in ADHD, wehave demonstrated modest heritability of VLF activ-ity and substantial genetic overlap between ADHDand VLF activity. This supports the relevance of thisnovel measure of physiological arousal to ourunderstanding of ADHD, and indicates that lowerVLF activity during cognitive activation is a potentialintermediate phenotype for ADHD. This provides abasis for identification of specific genes that influ-ence both ADHD and VLF activity, and warrants theinvestigation of the underlying mechanisms of VLFactivity to provide insight into the pathophysiologicalunderpinnings of ADHD.

Supporting informationAdditional Supporting Information may be found in theonline version of this article:

Data S1. Growth mixture modelling on longitudinalTEDS ADHD data

Data S2. Associations between VLF power and higherfrequency bands

Please note: Wiley-Blackwell are not responsible forthe content or functionality of any supporting materialssupplied by the authors. Any queries (other thanmissing material) should be directed to the corre-sponding author for the article.

AcknowledgementsThe authors gratefully acknowledge the participatingfamilies and all staff involved in this study, inparticular, Chloe Booth, Sarah Lewis, StuartNewman, the TEDS research team and the Director ofTEDS, Robert Plomin. The PI (GM) was supported bya fellowship from the National Institute for HealthResearch.

8 Charlotte Tye et al. J Child Psychol Psychiatry 2011; *(*): **–**

� 2011 The Authors. Journal of Child Psychology and Psychiatry � 2011 Association for Child and Adolescent Mental Health.

Correspondence toCharlotte Tye, MRC Social, Genetic and DevelopmentalPsychiatry Centre, Institute of Psychiatry, De Crespigny

Park, London SE5 8AF, UK.; Tel: +44 (0) 20 7848 0748;Fax: +44 (0) 20 7848 0866; E-mail: charlotte.tye@kcl.ac.uk

Key points

• The identification of neurobiological processes that mediate between genes and behaviour may elucidate theaetiological pathways underlying complex psychiatric disorders such as ADHD.

• The current study investigated genetic effects on very low-frequency (VLF) neuronal activity in adolescent twinpairs concordant or discordant for high or low ADHD symptom scores.

• Subjects with high ADHD symptom scores demonstrated reduced VLF power that was associated with poorperformance.

• VLF activity demonstrated moderate heritability and shared genetic influences with ADHD symptoms, sug-gestive of an ideal intermediate phenotype of the disorder.

• Investigating these processes provides insight into the pathophysiological underpinnings of ADHD, and mayhelp in understanding their impact on social and academic functioning.

ReferencesBalduzzi, D., Riedner, B.A., & Tononi, G. (2008). A BOLD

window into brain waves. Proceedings of the National Acad-emy of Sciences, 105, 15641–15642.

Biswal, B., Yetkin, F.Z., Haughton, V.M., & Hyde, J.S. (2005).Functional connectivity in the motor cortex of resting humanbrain using echo-planar MRI. Magnetic Resonance in Med-icine, 34, 537–541.

Broyd, S.J., Demanuele, C., Debener, S., Helps, S.K., James,C.J., & Sonuga-Barke, E.J.S. (2009). Default-mode braindysfunction in mental disorders: A systematic review. Neu-roscience & Biobehavioral Reviews, 33, 279–296.

Broyd, S.J., Helps, S.K., & Sonuga-Barke, E.J.S. (2011).Attention-induced deactivations in very low frequency EEGoscillations: Differential localisation according to ADHDsymptom status. PLoS ONE, 6, e17325.

Buszaki, G., & Draguhn, A. (2004). Neuronal oscillations incortical networks. Science, 304, 1926–1929.

Castellanos, F.X., Sonuga-Barke, E.J.S., Scheres, A., Di Mar-tino, A., Hyde, C., & Walters, J.R. (2005). Varieties ofattention-deficit/hyperactivity disorder-related intra-indi-vidual variability. Biological Psychiatry, 57, 1416–1423.

Conners, C., Sitarenios, G., Parker, J., & Epstein, J. (1998a).The Revised Conners’ Parent Rating Scale (CPRS-R): Factorstructure, reliability, and criterion validity. Journal of Abnor-mal Child Psychology, 26, 257–268.

Conners, C., Sitarenios, G., Parker, J., & Epstein, J. (1998b).Revision and restandardization of the Conners’ TeacherRating Scale (CTRS-R): Factor structure, reliability, andcriterion validity. Journal of Abnormal Child Psychology, 26,279–291.

Doehnert, M., Brandeis, D., Straub, M., Steinhausen, H.-C., &Drechsler, R. (2008). Slow cortical potential neurofeedbackin attention deficit hyperactivity disorder: Is there neuro-physiological evidence for specific effects? Journal of NeuralTransmission, 115, 1445–1456.

Falconer, D.S., & Mackay, T.F.C. (1996). Introduction toquantitative genetics. Harlow: Addison Wesley Longman Ltd.

Faraone, S.V., Biederman, J., & Mick, E. (2006). The age-dependent decline of attention deficit hyperactivity disorder:A meta-analysis of follow-up studies. Psychological Medi-cine, 36, 159–165.

Faraone, S.V., Perlis, R.H., Doyle, A.E., Smoller, J.W., Goral-nick, J.J., Holmgren, M.A., & Sklar, P. (2005). Moleculargenetics of attention-deficit/hyperactivity disorder. Biologi-cal Psychiatry, 57, 1313–1323.

Fox, M.D., Snyder, A.Z., Vincent, J.L., Corbetta, M., VanEssen, D.C., & Raichle, M.E. (2005). The human brainis intrinsically organized into dynamic, anticorrelated

functional networks. Proceedings of the National Academyof Sciences of the United States of America, 102, 9673–9678.

Fransson, P. (2005). Spontaneous low-frequency BOLD signalfluctuations: An fMRI investigation of the resting-statedefault mode of brain function hypothesis. Human BrainMapping, 26, 15–29.

Fransson, P. (2006). How default is the default mode of brainfunction?: Further evidence from intrinsic BOLD signalfluctuations. Neuropsychologia, 44, 2836–2845.

de Geus, E.J.C. (2002). Introducing genetic psychophysiology.Biological Psychology, 61, 1–10.

Gottesman, I.I., & Gould, T.D. (2003). The endophenotypeconcept in psychiatry: Etymology and strategic intentions.American Journal of Psychiatry, 160, 636–645.

Haworth, C.M.A., Wright, M.J., Luciano, M., Martin, N.G., DeGeus, E., Van Beijsterveldt, C.E., … & Plomin, R. (2010). Theheritability of general cognitive ability increases linearlyfrom childhood to young adulthood. Molecular Psychiatry,15, 1112–1120.

He, B.J., & Raichle, M.E. (2009). The fMRI signal, slow corticalpotential and consciousness. Trends in Cognitive Sciences,13, 302–309.

Helps, S.K., Broyd, S.J., James, C.J., Karl, A., Chen, W., &Sonuga-Barke, E.J.S. (2010). Altered spontaneous lowfrequency brain activity in attention deficit/hyperactivitydisorder. Brain Research, 1322, 134–143.

Helps, S.K., Broyd, S.J., James, C.J., Karl, A., & Sonuga-Barke, E.J.S. (2009). The attenuation of very low frequencybrain oscillations in transitions from a rest state to activeattention. Journal of Psychophysiology, 23, 191–198.

Helps, S.K., James, C.J., Debener, S., Karl, A., & Sonuga-Barke, E.J.S. (2008). Very low frequency EEG oscillationsand the resting brain in young adults: A preliminarystudy of localisation, stability and association with symp-toms of inattention. Journal of Neural Transmission, 115,279–285.

Jensen, A.R. (1998). The g factor: The science of mental ability.Wesport: Praeger.

Jung, T.-P., Makeig, S., Humphries, C., Lee, T.-W., McKeown,M.J., Iragui, V., & Sejnowski, T.J. (2000). Removing electro-encephalographic artifacts by blind source separation. Psy-chophysiology, 37, 163–178.

Kendall, T., Taylor, E., Perez, A., & Taylor, C., & GuidelineDevelopment Group. (2008). Diagnosis and management ofattention-deficit/hyperactivity disorder in children, youngpeople, and adults: Summary of NICE guidance. BritishMedical Journal, 337, a1239.

doi:10.1111/j.1469-7610.2011.02501.x Very low-frequency EEG intermediate phenotypes of ADHD 9

� 2011 The Authors. Journal of Child Psychology and Psychiatry � 2011 Association for Child and Adolescent Mental Health.

Klein, C., Wendling, K., Huettner, P., Ruder, H., & Peper,M. (2006). Intra-subject variability in attention-deficithyperactivity disorder. Biological Psychiatry, 60, 1088–1097.

Kovas, Y., Haworth, C.M.A., Dale, P.S., & Plomin, R. (2007).The genetic and environmental origins of learning abilitiesand disabilities in the early school years. Monographs of theSociety for Research in Child Development, 72, 1–144.

Kovas, Y., & Plomin, R. (2006). Generalist genes: Implicationsfor the cognitive sciences. Trends in Cognitive Sciences, 10,198–203.

Kuntsi, J., McLoughlin, G., & Asherson, P. (2006a). Attentiondeficit hyperactivity disorder. Neuromolecular Medicine, 8,461–484.

Kuntsi, J., Rogers, H., Swinard, G., Borger, N.A., Meere, J.v.d.,Rijsdijk, F.V., & Asherson, P. (2006b). Reaction time, inhi-bition, working memory and performance: Genetic influ-ences and their interpretation. Psychological Medicine, 36,1613–1624.

Kuntsi, J., Wood, A., Rijsdijk, F., Johnson, K., Andreou, P.,Albrecht, B., … & Asherson, P. (2010). Separation ofcognitive impairments in attention deficit hyperactivity dis-order into two familial factors. Archives of General Psychia-try, 67, 1159–1166.

Kuntsi, J., Wood, A.C., Van der Meere, J., & Asherson, P.(2009). Why cognitive performance in ADHD may not revealtrue potential: Findings from a large population-basedsample. Journal of the International NeuropsychologicalSociety, 15, 570–579.

Loo, S.K., Hale, T., Hanada, G., Macion, J., Shrestha, A.,McGough, J.J., … & Smalley, S.L. (2010). Familial clusteringand DRD4 effects on electroencephalogram measures inmultiplex families with attention deficit/hyperactivity disor-der. Journal of the American Academy of Child & AdolescentPsychiatry, 49, 368–377.

Loo, S.K., & Smalley, S.L. (2008). Preliminary report of familialclustering of EEG measures in ADHD. American Journal ofMedical Genetics Part B: Neuropsychiatric Genetics, 147B,107–109.

McLoughlin, G., Albrecht, B., Banaschewski, T., Rothenberger,A., Brandeis, D., Asherson, P., & Kuntsi, J. (2010). Electro-physiological evidence for abnormal preparatory states andinhibitory processing in adult ADHD. Behavioral and BrainFunctions, 6, 66.

McLoughlin, G., Asherson, P., Albrecht, B., Banaschewski, T.,Rothenberger, A., Brandeis, D., & Kuntsi, J. (2011). Cogni-tive-electrophysiological indices of attentional and inhibitoryprocessing in adults with ADHD: Familial effects. Behavioraland Brain Functions, 7, 26.

McLoughlin, G., Kuntsi, J., Brandeis, D., & Banaschewski, T.(2005). Electrophysiological parameters in psychiatric re-search: ADHD. Psychiatry, 4, 14–18.

Monto, S., Palva, S., Voipio, J., & Palva, J.M. (2008). Very slowEEG fluctuations predict the dynamics of stimulus detectionand oscillation amplitudes in humans. Journal of Neurosci-ence, 28, 8268–8272.

Neale, M.C., Boker, S.M., Xie, G., & Maes, H. (1999). Mx:Statistical modeling, 5th edn. Richmond, VA 23298: Depart-ment of Psychiatry, Box 126 MCV.

Neale, M.C., & Cardon, L.R. (1992). Methodology for geneticstudies of twins and families. Dordrecht: Kluwer.

Neale, M.C., Eaves, L.J., & Kendler, K.S. (1994). The power ofthe classical twin study to resolve variation in thresholdtraits. Behavior Genetics, 24, 239–258.

Neale, M.C., & Kendler, K.S. (1995). Models of comorbidity formultifactorial disorders. American Journal of Human Genet-ics, 57, 935–953.

Polanczyk, G., de Lima, M.S., Horta, B.L., Biederman, J., &Rohde, L.A. (2007). The worldwide prevalence of ADHD: A

systematic review and metaregression analysis. AmericanJournal of Psychiatry, 164, 942–948.

Price, T.S., Freeman, B., Craig, I., Petrill, S.A., Ebersole, L., &Plomin, R. (2000). Infant zygosity can be assigned byparental report questionnaire data. Twin Research andHuman Genetics, 3, 129–133.

Raven, J.C., Court, J.H., & Raven, J. (1996). Manual forRaven’s progressive matrices and vocabulary scales. Oxford:Oxford University Press.

Rijsdijk, F.V., & Sham, P.C. (2002). Analytic approaches totwin data using structural equation models. Briefings inBioinformatics, 3, 119–133.

Ronald, A., Simonoff, E., Kuntsi, J., Asherson, P., & Plomin, R.(2008). Evidence for overlapping genetic influences onautistic and ADHD behaviours in a community twinsample. Journal of Child Psychology and Psychiatry, 49,535–542.

Ruskin, D.N., Bergstrom, D.A., Shenker, A., Freeman, L.E.,Baek, D., & Walters, J.R. (2001). Drugs used in thetreatment of attention-deficit/hyperactivity disorder affectpostsynaptic firing rate and oscillation without preferentialdopamine autoreceptor action. Biological Psychiatry, 49,340–350.

Sergeant, J. (2000). The cognitive-energetic model: An empir-ical approach to attention-deficit hyperactivity disorder.Neuroscience & Biobehavioral Reviews, 24, 7–12.

Smit, D., Posthuma, D., Boomsma, D.I., & Geus, E.J.C. (2005).Heritability of background EEG across the power spectrum.Psychophysiology, 42, 691–697.

Sonuga-Barke, E.J.S., & Castellanos, F.X. (2007). Spontane-ous attentional fluctuations in impaired states and patho-logical conditions: A neurobiological hypothesis.Neuroscience and Biobehavioral Reviews, 31, 977–986.

Toulopoulou, T., Picchioni, M., Rijsdijk, F., Hua-Hall, M.,Ettinger, U., Sham, P., & Murray, R. (2007). Substantialgenetic overlap between neurocognition and schizophrenia:Genetic modeling in twin samples. Archives of GeneralPsychiatry, 64, 1348–1355.

Trouton, A., Spinath, F.M., & Plomin, R. (2002). Twins earlydevelopment study (TEDS): A multivariate, longitudinalgenetic investigation of language, cognition and behaviorproblems in childhood. Twin Research and Human Genetics,5, 444–448.

Tye, C., McLoughlin, G., Kuntsi, J., & Asherson, P. (2011).Electrophysiological markers of genetic risk for attentiondeficit hyperactivity disorder. Expert Reviews in MolecularMedicine, 13, e9.

Valko, L., Doehnert, M., Muller, U.C., Schneider, G., Albrecht,B., Drechsler, R., … & Brandeis, D. (2010). Differences inneurophysiological markers of inhibitory and temporal pro-cessing deficits in children and adults with ADHD. Journal ofPsychophysiology, 23, 235–246.

Vanhatalo, S., Palva, J.M., Holmes, M.D., Miller, J.W., Voipio,J., & Kaila, K. (2004). Infraslow oscillations modulateexcitability and interictal epileptic activity in the humancortex during sleep. Proceedings of the National Academy ofSciences of the United States of America, 101, 5053–5057.

Walters, J.T.R., & Owen, M.J. (2008). Endophenotypes inpsychiatric genetics. Molecular Psychiatry, 12, 886–890.

Wechsler, D. (1992). Wechsler intelligence scale for children –Third edition UK (WISC-IIIUK) manual. London: The Psycho-logical Corporation.

Willcutt, E.G., Doyle, A.E., Nigg, J.T., Faraone, S.V., &Pennington, B.F. (2005). Validity of the executive functiontheory of attention-deficit/hyperactivity disorder: A meta-analytic review. Biological Psychiatry, 57, 1336–1346.

Accepted for publication: 18 October 2011

10 Charlotte Tye et al. J Child Psychol Psychiatry 2011; *(*): **–**

� 2011 The Authors. Journal of Child Psychology and Psychiatry � 2011 Association for Child and Adolescent Mental Health.