The presence of comorbid disruptive behavior disorders(oppositional defiant disorder [ODD] or conduct dis-order [CD]) within children with attention-deficit/hyperactivity disorder (ADHD) has been well estab-lished for several decades (e.g., see Bird et al., 1990;Hinshaw, 1987). Only in the past decade has it becomeapparent that internalizing disorders (both anxiety anddepressive disorders) also commonly co-occur withADHD. Thus, both clinical and epidemiological studieshave consistently shown that as many as one third ofchildren with ADHD have co-occurring anxiety dis-

ADHD Comorbidity Findings From the MTA Study:Comparing Comorbid Subgroups

PETER S. JENSEN, M.D., STEPHEN P. HINSHAW, PH.D., HELENA C. KRAEMER, PH.D.,NILANTHA LENORA, B.S., JEFFREY H. NEWCORN, M.D., HOWARD B. ABIKOFF, PH.D.,

JOHN S. MARCH, M.D., L. EUGENE ARNOLD, M.D., DENNIS P. CANTWELL, M.D.,C. KEITH CONNERS, PH.D., GLEN R. ELLIOTT, M.D., LAURENCE L. GREENHILL, M.D.,

LILY HECHTMAN, M.D., BETSY HOZA, PH.D., WILLIAM E. PELHAM, PH.D., JOANNE B. SEVERE, M.S.,JAMES M. SWANSON, PH.D., KAREN C. WELLS, PH.D., TIMOTHY WIGAL, PH.D.,

AND BENEDETTO VITIELLO, M.D.

ABSTRACT

Objectives: Previous research has been inconclusive whether attention-deficit/hyperactivity disorder (ADHD), when comor-

bid with disruptive disorders (oppositional defiant disorder [ODD] or conduct disorder [CD]), with the internalizing disorders

(anxiety and/or depression), or with both, should constitute separate clinical entities. Determination of the clinical significance

of potential ADHD + internalizing disorder or ADHD + ODD/CD syndromes could yield better diagnostic decision-making,

treatment planning, and treatment outcomes. Method: Drawing upon cross-sectional and longitudinal information from 579

children (aged 7–9.9 years) with ADHD participating in the NIMH Collaborative Multisite Multimodal Treatment Study of

Children With Attention-Deficit/Hyperactivity Disorder (MTA), investigators applied validational criteria to compare ADHD sub-

jects with and without comorbid internalizing disorders and ODD/CD.Results: Substantial evidence of main effects of internal-

izing and externalizing comorbid disorders was found. Moderate evidence of interactions of parent-reported anxiety and

ODD/CD status were noted on response to treatment, indicating that children with ADHD and anxiety disorders (but no

ODD/CD) were likely to respond equally well to the MTA behavioral and medication treatments. Children with ADHD-only or

ADHD with ODD/CD (but without anxiety disorders) responded best to MTA medication treatments (with or without behavioral

treatments), while children with multiple comorbid disorders (anxiety and ODD/CD) responded optimally to combined (med-

ication and behavioral) treatments. Conclusions: Findings indicate that three clinical profiles, ADHD co-occurring with inter-

nalizing disorders (principally parent-reported anxiety disorders) absent any concurrent disruptive disorder (ADHD + ANX),

ADHD co-occurring with ODD/CD but no anxiety (ADHD + ODD/CD), and ADHD with both anxiety and ODD/CD (ADHD +

ANX + ODD/CD) may be sufficiently distinct to warrant classification as ADHD subtypes different from “pure” ADHD with

neither comorbidity. Future clinical, etiological, and genetics research should explore the merits of these three ADHD classifi-

cation options. J. Am. Acad. Child Adolesc. Psychiatry, 2001, 40(2):147–158. Key Words: attention-deficit/hyperactivity dis-

order, attention deficit, anxiety, comorbidity, treatment, outcomes, classification, diagnosis.

J . AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 40 :2 , FEBRUARY 2001 147

Accepted September 26, 2000.Dr. Jensen is Professor of Psychiatry, Department of Psychiatry, Columbia

University College of Physicians and Surgeons; Dr. Hinshaw is Professor ofPsychology, Department of Psychology, University of California at Berkeley; Dr.Kraemer is Professor of Biostatistics, Stanford University; Ms. Lenora is a med-ical student, George Washington University School of Medicine. Other authors’affiliations are listed in the MTA Cooperative Group acknowledgment thatappears at the end of the text.

Reprint requests to Dr. Jensen, Center for the Advancement of Children’sMental Health, Columbia University/NYSPI, 1051 Riverside Drive, Unit 78,New York, NY 10032; e-mail: pj131@columbia.edu.

0890-8567/01/4002-0147q2001 by the American Academy of Childand Adolescent Psychiatry.

orders (Biederman et al., 1991; Bird et al., 1990; MTACooperative Group, 1999a; Woolston et al., 1989). Thisphenomenon appears “real” and cannot be attributedsolely to factors such as referral biases or methodologicalartifacts (Angold et al., 1999).

The overlap of ADHD with both the internalizing anddisruptive disorders should be of substantial interest toresearchers and clinicians. For example, if it can be shownthat ADHD when comorbid with either the internalizingor disruptive disorders is unique with respect to its clinicalcorrelates, etiology, course, and outcome (versus “pure”ADHD), it may be useful to consider such comorbidstates as diagnostically meaningful subtypes or even asseparate disorders. Also, if various comorbid forms ofADHD are actually of qualitatively different types from“pure” ADHD, better treatment planning and clinicalprediction, as well as more valid etiological and geneticstudies, should be possible.

Validation of Subtypes

Cantwell (1995) outlined eight domains of clinicalinvestigation that could be used to assess the discriminantvalidity of possible disorder types and subtypes. Thesecriteria, modifications of the original Robins and Guze(1970) criteria, include (1) clinical phenomenology, (2)demographic correlates, (3) psychosocial correlates, (4, 5)family factors (both environmental and genetic), 6) biolog-ical factors, (7) response to treatment, and (8) clinical out-comes. Applying these criteria to ADHD comorbiditystudies, Jensen, Martin, and Cantwell (1997) reported thatseven of the eight validational criteria supported thehypothesis that ADHD comorbid with disruptive behav-ior disorders (ADHD + ODD/CD) constitutes a differentADHD subtype. However, applying these same criteria toADHD comorbid with internalizing disorders (principallyanxiety), they noted somewhat weaker evidence (four ofeight criteria) in support of an ADHD + ANX subtype.

In the past several years since the Jensen et al. (1997)review, two additional pertinent studies have been con-ducted that are of relevance to questions concerning theusefulness and validity of the ADHD + ANX subtype.Beginning with an initial dose titration, Diamond et al.(1999) followed stimulant-treated ADHD children over 4months to determine the impact of comorbid anxiety ontheir response to medication. Unlike the preponderanceof evidence from earlier studies (e.g., Buitelaar et al.,1995; DuPaul et al., 1994; Pliszka, 1989, 1992; Tannocket al., 1995), Diamond and colleagues found that children

with ADHD + ANX and those with ADHD/no ANXdid not show any differential response to methylpheni-date, either in terms of behavioral measures or side effects.

Most recently, findings from the NIMH CollaborativeMultisite Multimodal Treatment Study of Children WithAttention-Deficit/Hyperactivity Disorder (MTA), a multi-site randomized clinical trial testing four treatment groupsin 579 children across six sites, indicated that ADHD chil-dren with comorbid parent-reported anxiety disordershave treatment responses that are qualitatively differentfrom those of children without comorbid anxiety (MTACooperative Group, 1999a,b). Thus, among children withparent-identified anxiety disorders (33.5% of the sample),those children in the behavioral therapy condition (Beh)showed an enhanced response on outcome measures ofparent-reported ADHD and internalizing symptoms, rel-ative to nonanxious children. In addition, among anxioussubjects combined treatments (Comb) yielded greaterimprovements than medication management alone (Med-Mgt) in several outcome domains. Among nonanxiousparticipants, however, MedMgt and Comb both outper-formed Beh and the community comparison group (CC)with regard to ADHD symptoms (MTA CooperativeGroup, 1999b).

Findings from the MTA study provide evidence ofqualitatively different treatment responses in ADHD +ANX versus ADHD/no ANX children and may thussupport the separate ADHD + ANX subtype hypothesis.In contrast, and to the surprise of the MTA investigators,the presence of comorbid ODD/CD did not moderateany differential treatment response to specific treatments.However, these first moderator analyses were restricted toexamination of the main effects of comorbidity and didnot explore possible interactions between ODD/CD andcomorbid anxiety on MTA outcomes.

Remarkably, only one study to date has examined thesefour potentially separate entities. In a study of 138 whitemales with DSM-III–defined ADHD, Livingston andcolleagues (1990) found that boys with an internalizingdisorder and ODD/CD differed along multiple dimen-sions from boys with ADHD + ODD/CD-only andADHD + ANX-only, as well as from boys with ADHD-only. In most areas of functioning, doubly comorbid boysscored worse than ADHD + ODD/CD-only and ADHD+ ANX-only boys, though in at least one instance doublycomorbid boys functioned somewhat better (e.g., teacher-completed aggression ratings). Unfortunately, given thestudy’s small sample size (n = 138) and its restriction to

JENSEN ET AL.

148 J . AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 40 :2 , FEBRUARY 2001

white males from one clinical setting, it could not bedetermined whether the differences between doublycomorbid boys and singly comorbid and ADHD-onlyboys were simply an additive function of the two comor-bid conditions or whether these differences reflectedunique clinical profiles.

Thus, while mounting evidence does support the pos-sibility that individuals with comorbid ADHD + ANXand those with ADHD + ODD/CD may have uniqueclinical profiles compared with children with noncomor-bid ADHD, firm conclusions about these comorbid pat-terns cannot be derived from previous research, given therelatively small sample sizes and previous studies’ failure totease apart the effects of the two comorbid conditionswhen they overlap. Furthermore, no previous investiga-tions have used a comprehensive set of validational crite-ria (either Robins and Guze, 1970, or Cantwell, 1995) toexamine these comorbidity subgroups, all within the samestudy.

The MTA study (MTA Cooperative Group 1999a,b)offers a unique opportunity to examine whether childrenwith ADHD + ANX, ADHD + ODD/CD, and ADHD +ANX + ODD/CD do indeed differ from each other andfrom children with noncomorbid ADHD, and whethersuch findings can be replicated across multiple settings. Inthis report, we examine four different groups of ADHDchildren (ADHD + ANX, ADHD + ODD/CD, ADHD +ANX + ODD/CD, ADHD-only), comparing them usingthe modified Cantwell validational criteria to determinewhether the varying forms of comorbidity convey specialdiagnostic, therapeutic, or outcome significance beyondthe presence of ADHD alone.

METHOD

Recruitment Procedures and Sample Characteristics

MTA recruitment and screening procedures aimed to collect a care-fully diagnosed, impaired sample of ADHD children with a widerange of comorbid conditions and demographic characteristics, repre-sentative of patients seen clinically (MTA Cooperative Group, 1999a).For eligibility, children (of either sex) were between ages 7.0 and 9.9years, in first through fourth grades. All met DSM-IV criteria forADHD, combined type, using the Diagnostic Interview Schedule forChildren, parent report, version 3.0 (DISC-P) (Shaffer et al., 1996).The presence of comorbid disorders such as ODD, CD, internalizingdisorders, or specific learning disabilities were not exclusions, as animportant aim of the MTA study was to examine their interactionswith treatment outcomes.

In a four-group parallel design, 579 children were assigned randomlyto (1) medication management (MedMgt), (2) behavioral treatment(Beh), (3) the combination (Comb), or (4) community comparison(CC) for 14 months. Demographic characteristics of the MTA partic-

ipants were as follows: 80% male and 20% female; 61% white, 20%African American, and 19% Hispanic, racially mixed, or other ethnicorigins. Mean age at study entry was 8.2 years. Extensive informationconcerning subject characteristics, study treatments, and outcomes isreported elsewhere (MTA Cooperative Group, 1999a,b).

Assessments

The major symptom and functioning domains originally used toassess baseline characteristics and 14-month outcomes (see MTACooperative Group, 1999a) were applied to test the Cantwell val-idational criteria on the various comorbidity profiles. Variables selectedfor the Cantwell criteria were as follows: (1) clinical phenomenology—inattentive, hyperactive/impulsive, and aggressive/oppositional symp-toms (assessed via parent- and teacher-completed Swanson, Nolan,and Pelham scale [SNAP] scores) (MTA Cooperative Group, 1999a),Verbal and Performance IQ, impairment (measured via the ColumbiaImpairment Scale) (Bird et al., 1996), internalizing symptoms (assessedvia the Multidimensional Anxiety Scale for Children [March et al.,1997], the Children’s Depression Inventory [Kovacs and Beck, 1977],and parent- and teacher-rated internalizing symptoms scales from theSocial Skills Rating System [SSRS] [Gresham and Elliott, 1989]), andacademic achievement (assessed via the Wechsler Individual Achieve-ment Test [WIAT] [Wechsler, 1992]); (2) demographic factors—gender,age, single- versus two-parent status, income, and family occupationalstatus; (3) psychosocial factors—social skills/peer relations (assessed viaparent and teacher reports using the SSRS) and levels of life stressors;(4) family factors (both environmental and genetic)—family stress (usingthe Parent Stress Index), parent–child relations (using scales from theParent–Child Relationship Questionnaire [Furman and Giverson,1995]), parenting style based on an internally consistent and reliablefactor derived from all MTA-administered parenting assessments(Wells et al., 2000), and maternal depression (using the Beck Depres-sion Inventory); and (5) biological factors (history of pre- and perinataldifficulties). Of note, in the original Cantwell validational schemefamily environment and family genetic factors are defined separately,but here we have combined them into a single criterion, as the MTAdesign did not allow any robust method for separating these twosources of familial effects. These five Cantwell criteria were all based ondata collected at the baseline assessment, allowing us to examine whatKraemer (1995) has termed epidemiologic comorbidity.

To examine the last two Cantwell criteria (outcomes, response totreatment—also termed clinical comorbidity by Kraemer, 1995), we con-structed difference scores by subtracting 14-month endpoint scoresfrom baseline scores of children’s symptoms and functioning ratings ofthose variables from the clinical phenomenology, psychosocial factors,and family factors domains that describe the child’s functioning. Inaddition to these outcome measures, we also used an overall compositeoutcome measure (see Conners et al., 2001) to examine outcomes andresponse to treatment. To avoid the potentially confounding effects ofthe MTA treatments on more “naturalistic” outcomes, analyses of out-comes were limited to CC subjects only (n = 146), while response totreatment analyses included all subjects. By way of caution in interpret-ing these data, our analyses of CC subjects’ “naturalistic” outcomes areshould be viewed with caution, inasmuch as two thirds of these childrenobtained at least some treatment (including stimulants) over the courseof the study.

Comorbidity Subgroups

Diagnoses and comorbidity groups were constructed on the basisof parent-reported DISC diagnoses at study baseline. Diagnoses werebased on the previous 6 months, except for CD, which was based on

ADHD COMORBID SUBGROUPS

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 40 :2 , FEBRUARY 2001 149

the last 12 months according to DSM criteria. While our first reportsconcerning ADHD comorbid conditions (MTA Cooperative Group,1999a,b) did not include simple phobia in the “anxiety disorders” cat-egory, other analyses have indicated that parent-reported simple pho-bias are in fact as likely to be validated by clinicians upon reinterviewas other anxiety disorders (Jensen et al., 1999). Given these findingsand the need to form large enough groups for the analyses for thisreport, we chose here to include subjects with simple phobia in the“anxiety disorders” category, resulting in somewhat higher rates ofcomorbid anxiety disorders (38.7% versus 33.5%).

Each child was grouped into one of four categories: ADHD-only(31.8%, n = 184) ADHD + ANX (14.0%, n = 81), ADHD + ODD/CD (29.5%, n = 171), and ADHD + ODD/CD + ANX (24.7%, n =143). For the 22 cases of mood disorders in the MTA sample, all but 2were comorbid with anxiety disorders. Given recent evidence pre-sented by Angold et al. (1999) that comorbid mood disorders in chil-dren with ADHD occur principally via their association with anxietydisorders, we have named the overall category ADHD + ANX, thoughin fact this group also included two children with mood disorderswithout a full-blown anxiety disorder. All other comorbid conditions(e.g., enuresis, tics, etc.) were ignored for the purposes of forming thefour comorbidity groups.

Analytic StrategyTo determine whether ADHD children with different comorbid-

ity profiles differed meaningfully from “pure” ADHD subjects, ouranalytic strategy proceeded as follows. We first assigned two new vari-ables to all subjects according to the presence or absence of the twocomorbidity forms, i.e., presence (1) or absence (0) of anxiety comor-bidity (Anx) and ODD/CD comorbidity (Cd ). Then, to examine thepotential distinctions between the singly comorbid groups (e.g.,ADHD + ANX) and the doubly comorbid group (ADHD + ANX +ODD/CD), it was necessary to determine whether the effects of eachcomorbidity applied equally across its single and doubly comorbidforms (termed a main effect of that comorbidity), or whether anyinteraction effects (Anx 3 Cd ) existed. These issues were explored forall Cantwell criteria via multivariate analyses, examining the extent towhich Anx, Cd, and Anx 3 Cd effects could be found for specificindicators. If these analyses should demonstrate multiple Anx 3 Cdinteractions, or should indicate that when Anx and Cd effects arefound they usually occur with different variables, such findingswould provide evidence for the need to discriminate among the threepotential comorbid subgroups versus “pure” ADHD-only subjects.

To examine further the response to treatment criterion, we constructedunique codes for each of the MTA treatment group assignments. Then,using these codes in multivariate analyses of the difference scoresdescribed above, along with codes for the four comorbidity subgroups,we examined all interactions between comorbidity subtype and MTAtreatment group status in the final model. Finding interactions betweena given MTA treatment and any comorbidity group would have impor-tant implications for treatment planning and could enhance clinicians’ability to match patients to specific treatments.

Finally, to determine clinical significance of all findings, effect sizeswere computed to examine the impact of comorbidity subtypes onbaseline characteristics, outcomes, and response to treatment.

Given our concern that too stringent methods might result in failureto identify modest to moderate effects within relatively small comor-bidity groups stratified across six sites and four treatment conditions, pvalues are set at a modest p < .05 and no corrections for multiple testsare done here. After statistical consultation, we reasoned that this was adefensible strategy, since these are exploratory analyses and the inter-pretation of results should not rest not on any single statistical analysis

per se. Readers are encouraged to exercise caution in the interpretationof findings. For all analyses, because previous analyses (March et al.,2000; MTA Cooperative Group, 1999b) demonstrated that neithergender nor ethnicity impacted significantly on results, these two vari-ables were not included as covariates. In contrast, because importantdifferences in baseline characteristics have been found across sites, sitestatus was entered as a covariate for all analyses.

RESULTS

Results are presented in the following order: (1) head-to-head comparisons of the two singly comorbid subgroups(to determine whether the specific kind of comorbidity[ANX or ODD/CD] makes a difference); (2) examinationof Cd and Anx main, additive, and/or interaction effects byapplying Cantwell criteria across baseline characteristicsand outcomes; (3) examination of possible differential ben-efits of specific treatments on specific comorbid groups’outcomes; and (4) examination of effect sizes.

Do ADHD + ANX and ADHD + ODD/CD SubjectsDiffer? Table 1 presents the baseline means from each ofthe four groups constructed by comorbidity profiles. Noanalyses are presented for any baseline variables thatsimply reflect how the groups were formed (e.g., par-ent/teacher reports of aggression/oppositionality andinternalizing symptoms). The comparisons of the twononoverlapping ADHD comorbidity profiles (ADHD +ANX, ADHD + ODD/CD) can be seen in the secondand third columns from the left in Table 1 (baseline char-acteristics, Cantwell criteria 1–5). Significant differencesbetween these two groups are indicated by boldface type.Inspection of these columns indicates that these two singlycomorbid groups differed in 11 instances on baselinecharacteristics (Table 1).

Because of space considerations, analyses of the outcomescriterion are not presented here. Nonetheless, findings fromADHD + ANX versus ADHD + ODD/CD outcomecomparisons overlapped substantially with and were consis-tent with the analogous differences noted between thesetwo comorbid groups at baseline for five variables (parent-rated subscale and total scores on the SNAP, overall impair-ment, WIAT math, and Conners composite scores). Inaddition, for three additional variables—all teacher-completed (SNAP inattention, total SNAP scores, andSSRS social skills), significant differences were foundbetween ADHD + ANX and ADHD + ODD/CD sub-jects’ outcomes (tabled data available upon request).

It is important to note that the differences betweenthese two comorbid groups cross many areas of function-ing, including domains where ADHD + ODD/CD sub-

JENSEN ET AL.

150 J . AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 40 :2 , FEBRUARY 2001

TAB

LE 1

Val

idat

ion

Cri

teri

a by

Com

orbi

dity

Sub

type

: Mea

n Sc

ores

and

Eff

ect S

izes

Subt

ype

Mea

n Sc

ores

AD

HD

+Anx

+OD

D/C

D+D

ual

Eff

ect S

izes

Pool

edV

aria

ble

(184

)a(8

1)(1

71)

(143

)A

nxC

dC

d 3

Anx

Stat

isti

csSD

Clin

ical

phe

nom

enol

ogy

(bas

elin

e)In

atte

ntiv

e (T

)2.

272.

282.

212.

16–.

02.0

9.0

9F

= 1.

42, p

< .2

0; 8

,560

0.65

2In

atte

ntiv

e (P

)1.

922.

102.

122.

08–.

29–.

32.3

5F

= 2.

74, p

< .0

06; 8

,563

0.62

7H

yper

acti

ve-i

mpu

lsiv

e (T

)2.

001.

872.

042.

01.1

7–.

05–.

13F

= 1.

82, p

< .0

7; 8

,563

0.76

5H

yper

acti

ve-i

mpu

lsiv

e (P

)1.

721.

872.

042.

05–.

20–.

43**

*.1

9F

= 4.

94, p

< .0

01; 8

,565

0.75

1To

tal S

NA

P (T

)1.

761.

751.

851.

85.0

2–.

18*

.02

F=

2.76

, p<

.005

; 8,5

470.

512

Tota

l SN

AP

(P)

1.46

1.59

1.80

1.86

–.26

*–.

69**

*.1

4F

= 10

.5, p

< .0

01; 8

,551

0.49

4C

DI

(chi

ld)

0.39

0.42

0.41

0.37

–.10

–.06

.23

F=

1.72

, p<

.09;

8,5

670.

311

MA

SC (

child

)2.

522.

572.

462.

52–.

09.1

1–.

02F

= 1.

32, p

< .2

3; 8

,536

0.53

2C

olum

bia

Impa

irm

ent S

cale

(P)

18.0

21.0

23.8

27.5

–.38

***

–.74

***

–.09

F=

17.4

, p<

.001

; 8,4

227.

83C

hild

nee

ds s

ervi

ces

(P)

2.87

3.09

3.33

3.49

–.26

*–.

54**

*.0

7F

= 5.

10, p

< .0

01; 8

,422

0.85

3LD

dia

gnos

is (

yes/

no)

(pro

port

ion)

0.16

0.22

0.13

0.13

–.17

***

.08

.17

F=

1.16

, p<

.33;

8,5

700.

361

Ver

bal I

Q10

110

010

110

1–.

07.0

0–.

07F

= 1.

89, p

< .0

6; 8

,564

14.8

Perf

orm

ance

IQ

102

101

102

100

–.06

.00

.06

F=

2.47

, p<

.012

; 8,5

6415

.6W

IAT

Mat

h99

.294

.599

.196

.7–.

34**

–.01

.17

F=

3.38

, p<

.001

; 8,5

7013

.8W

IAT

Spe

lling

95.8

91.0

94.4

93.1

–.34

*–.

10.2

5F

= 2.

04, p

< .0

4; 8

,570

14.0

WIA

T R

eadi

ng96

.993

.596

.494

.9–.

24*

–.04

.13

F=

1.43

, p<

.19;

8,5

7014

.2O

vera

ll co

mpo

site

sco

re0.

942.

646.

968.

68–.

21**

–.76

**–.

00F

= 17

.1, p

< .0

01; 8

,570

7.94

Dem

ogra

phic

fact

ors

(bas

elin

e)G

ende

r (p

ropo

rtio

n m

ale)

0.80

0.79

0.81

0.81

–.03

.03

.03

F=

1.44

, p<

.18;

8,5

700.

40A

ge (

year

s)7.

87.

57.

87.

9–.

37*

.00

.50*

F=

2.27

, p<

.03;

8,5

700.

808

Min

orit

y (y

es/n

o) (

prop

orti

on)

0.34

0.44

0.40

0.41

–.20

–.12

.18

F=

13.7

, p<

.001

; 8,5

680.

489

Two

pare

nts

in h

ome

(pro

port

ion)

0.75

0.56

0.70

0.59

–.35

**.0

0.0

0F

= 3.

79, p

< .0

01; 8

,570

0.56

7(M

ean

no. o

fpa

rent

s in

hom

e)(1

.7)

(1.5

)(1

.7)

(1.5

)In

com

e (p

ropo

rtio

n on

wel

fare

sta

tus)

0.12

0.21

0.20

0.27

–.23

*–.

20.1

0F

= 7.

07, p

< .0

01; 8

,570

0.39

3Ps

ycho

soci

al fa

ctor

s (b

asel

ine)

Soci

al s

kills

/pee

r re

lati

ons

SSR

S (T

)0.

860.

840.

800.

80–.

07–.

22.0

7F

= 2.

15, p

< .0

3; 8

,531

0.27

8SS

RS

(P)

1.11

1.06

0.98

0.94

–.21

*–.

55**

*.0

4F

= 8.

01, p

< .0

01; 8

,514

0.23

5N

o. o

flif

e st

ress

ors

(all

neg.

eve

nts)

b2.

02.

32.

53.

1–.

12–.

21**

–.12

F=

2.84

, p<

.004

; 8,5

602.

42N

o. o

flif

e st

ress

ors

(con

foun

ded)

1.2

1.4

1.6

2.0

–.11

–.21

**–.

11F

= 3.

37, p

< .0

01; 8

,560

1.90

Fam

ily fa

ctor

s (en

viro

nmen

tal a

nd g

enet

ic, b

asel

ine)

Wel

ls P

aren

ting

Ind

ex (

P)0.

150.

68–0

.48

0.03

.29*

*–.

34**

*–.

01F

= 3.

96, p

< .0

01; 8

,575

1.84

Pare

ntin

g St

ress

Ind

ex (

P)3.

643.

493.

373.

20–.

28**

*–.

50**

*.0

4F

= 8.

78, p

< .0

01; 8

,515

0.53

5Pa

rent

–chi

ld (

pow

er a

sser

tion

) (P

)2.

582.

632.

832.

82–.

09–.

46**

*.0

9F

= 4.

18, p

< .0

01; 8

,552

0.54

2Pa

rent

–chi

ld (

posi

tive

aff

ecti

on)

(P)

3.60

3.64

3.46

3.57

.08

–.28

*.1

4F

= 1.

61, p

< .1

19; 8

,552

0.49

4M

ater

nal d

epre

ssio

n (B

DI)

(P)

0.27

0.35

0.29

0.39

–.28

**–.

07–.

07F

= 3.

22, p

< .0

01; 8

,518

0.29

1

ADHD COMORBID SUBGROUPS

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 40 :2 , FEBRUARY 2001 151

–– C

ontin

ued

jects were more impaired (e.g., baseline hyperactive-impulsive symptoms, overall impairment, parent–childrelations, and the composite index), as well as other, quitedifferent areas of functioning where ADHD + ANX sub-jects were more impaired (e.g., baseline academic perfor-mance scores, likelihood of a learning disability diagnosis,less improvement in teacher-rated inattention by 14months, and less improvement in the composite index).

Are the Two Forms of Comorbidity Simply “Additive”?Turning to the multivariate analyses to explore additiveand interactive effects of the two forms of comorbidity,columns 5 through 7 in Table 1 indicate significant main(Anx, Cd ) and interactive (Anx 3 Cd ) effects. Significanteffects are labeled with asterisks to denote levels of signifi-cance. There were 22 variables in which there were signifi-cant effects of comorbidity on baseline characteristics and7 additional instances of effects of comorbidity on out-comes (not shown here, but available upon request). Ofnote, across these variables, only 7 instances of additiveeffects of comorbidity were found (characterized by signif-icant effects being noted in both columns 5 and 6—parenttotal SNAP scores, overall impairment, need for services,parent-rated social skills, the Parenting Stress Index, theWells parenting factor, and the composite index). Notethat none of the measures, except the parent total SNAP,attempt to measure psychopathology directly; rather, theyare general indices of the overall impact of psychopathol-ogy on impairment, need for services, and families. Thesefindings indicate that while the two different forms ofcomorbidity/psychopathology are not additive per se, theircombined impacts on external factors such as overallimpairment, family burden, and services use may be.

Apart from the 7 variables for which additive effectswere found, findings from Table 1 and our outcome crite-rion analyses revealed that Anx comorbidity exerted maineffects on either baseline or outcome variables in 12instances, while Cd exerted effects on 9 other variables.Thus, the most common main effects of Anx and Cdcomorbidity were noted on different variables, suggestingquite different symptom profiles of subjects, dependingon their particular comorbidity subgroup.

Do Comorbid Subgroups Differ in Response to SpecificTreatments? Table 2 presents findings concerning theresponse to treatment criterion among all three MTA-delivered interventions (versus CC), examined via themultivariate analyses. Turning first to the issue of uniqueeffects of treatments on comorbid subgroups, such pat-terns are revealed by letters B (Beh), M (MedMgt), and C

TAB

LE 1

(Con

tinue

d)

Subt

ype

Mea

n Sc

ores

AD

HD

+Anx

+OD

D/C

D+D

ual

Eff

ect S

izes

Pool

edV

aria

ble

(184

)a(8

1)(1

71)

(143

)A

nxC

dC

d 3

Anx

Stat

isti

csSD

Bio

logi

cal f

acto

rsN

o. o

fpr

enat

al p

robl

ems

1.5

1.4

1.4

1.2

.07

.07

.07

F=

1.25

, p<

.27;

8,5

701.

37N

o. o

fpe

rina

tal p

robl

ems

0.5

0.6

0.5

0.7

–.13

*.0

0–.

13F

= 2.

23, p

< .0

3; 8

,570

0.77

Tota

l pre

-per

inat

al p

robl

ems

1.9

2.0

1.9

1.9

–.06

.00

.06

F=

0.56

, p<

.81;

8,5

701.

72M

at. p

ostp

artu

m d

epre

ssiv

e sy

mpt

oms

2.0

2.2

1.9

2.1

–.18

.09

.00

F=

1.39

, p<

.20;

8,5

301.

11

Not

e: F

or sc

ale

scor

es, h

ighe

r sco

res a

nd p

ositi

ve v

alue

s ind

icat

e im

prov

emen

t/be

tter

func

tioni

ng; n

egat

ive

scor

es in

dica

te d

eter

iora

tion.

AD

HD

, +A

nx, +

OD

D/C

D, a

nd +

Dua

lrep

-re

sent

the

four

com

orbi

dity

gro

ups,

cla

ssifi

ed b

ased

on

pres

ence

or a

bsen

ce o

fea

ch c

omor

bidi

ty. A

nx, C

dre

pres

ent t

he d

umm

y co

des f

or p

rese

nce/

abse

nce

ofea

ch c

omor

bidi

ty c

oded

and

ente

red

in m

ultiv

aria

te a

naly

ses.

Sig

nific

ant m

ain

effe

cts

and

inte

ract

ions

(Anx

3C

d)

for

com

orbi

dity

are

den

oted

in th

e re

spec

tive

colu

mns

by

aste

risk

s to

indi

cate

leve

l of

sta-

tistic

al s

igni

fican

ce. *

p#

.05;

**p

#.0

1; *

**p

#.0

01. S

igni

fican

t di

ffer

ence

s be

twee

n th

e A

DH

D +

AN

X a

nd t

he A

DH

D +

OD

D/C

D s

ubgr

oups

are

indi

cate

d by

bol

dfac

e ty

pe.

AD

HD

= a

tten

tion-

defic

it/hy

pera

ctiv

ity d

isor

der;

AN

X =

anx

iety

dis

orde

rs; O

DD

/CD

= o

ppos

ition

al d

efian

t dis

orde

r or c

ondu

ct d

isor

der;

T =

teac

her;

P =

par

ent;

SNA

P =

Swan

son,

Nol

an, a

nd P

elha

m s

cale

; CD

I =

Chi

ldre

n’s

Dep

ress

ion

Inve

ntor

y; M

ASC

= M

ultid

imen

sion

al A

nxie

ty S

cale

for

Chi

ldre

n; L

D =

lear

ning

dis

abili

ty; W

IAT

= W

echs

ler

Indi

vidu

alA

chie

vem

ent T

est;

SSR

S =

Soci

al S

kills

Rat

ing

Scal

e; B

DI

= B

eck

Dep

ress

ion

Inve

ntor

y.a

n’s.

bSt

ress

sco

res

wer

e co

mpu

ted

first

by

sum

min

g on

ly th

ose

even

ts th

at w

ere

not p

oten

tial

ly a

resu

lt of

subj

ects

’ or

fam

ilies

’ ow

n fu

ncti

onin

g (“

unco

nfou

nded

”), t

hen

by s

umm

ing

only

tho

se e

vent

s th

at m

ay in

fact

res

ult

from

fam

ilies

’ fun

ctio

ning

, e.g

., di

vorc

e or

par

enta

l sep

arat

ion

(“co

nfou

nded

”) (

see

Jens

en e

tal

., 19

91).

JENSEN ET AL.

152 J . AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 40 :2 , FEBRUARY 2001

(Comb) in the columns denoting the comorbidity sub-types. All comorbidity group effects are presented vis-à-viscomparisons to the ADHD-only group. Likewise, treat-ment effects are presented with reference to the CC group.Of note in the table are the commonly found, differen-tially beneficial effects of Beh interventions for theADHD + ANX group. These interactions are indicative oftreatment effects over and above the main effects of all treat-ments for the entire sample (irrespective of comorbidity),beyond Beh treatment benefits for the entire sample, andeven distinct from the impact of Comb and MedMgttreatments on the ADHD + ANX subgroup itself.

The second finding revealed in Table 2 is that duallycomorbid subjects also evidenced a substantial benefitfrom specific treatments, sometimes Comb, sometimes

Beh, while ADHD + ODD/CD subjects rarely showed adifferential response to any of the treatments, comparedwith ADHD-only subjects.

The right-hand column of Table 2 indicates whetherthere were any main effects of comorbidity subgroup ontreatment outcomes. Of note, seven of these eight in-stances were found among subjects with comorbid anxiety(ADHD + ANX or ADHD + ANX + ODD/CD sub-types). Inspection of actual means revealed that these twogroups generally showed the greatest response to treat-ments (not shown here, but tables available from theauthors upon request), regardless of treatment type.

How Large and Clinically Meaningful Are ComorbiditySubgroup Effects? The size and statistical significance ofcomorbidity effects on baseline characteristics (Table 1)

ADHD COMORBID SUBGROUPS

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 40 :2 , FEBRUARY 2001 153

TABLE 2Response to Treatment by Comorbidity Subtypes: Difference Scores (Baseline Minus 14 Months)

Mean Scores Main EffectsADHD +Anx +ODD/CD +Dual of Comorbid(184)a (81) (171) (143) Group on Outcomes? Statistics

Inattentive (T) 0.90 0.78B,M 0.83 0.91 F = 1.47, p < .09; 20,557Inattentive (P) 0.71 0.77B 0.81 0.78B ADHD+Anx F = 3.44, p < .001; 20,555Hyperactive-impulsive (T) 0.98 0.85 0.93 0.99 F = 1.64, p < .04; 20,557Hyperactive-impulsive (P) 0.70 0.79B 0.86 0.81C F = 4.01, p < .001; 20,555Oppositional aggressive (T) 0.49 0.42B 0.54 0.62 F = 1.68, p < .03; 20,557Oppositional aggressive (P) 0.29 0.31 0.55 0.58 ADHD+Anx+ODD/CD F = 3.08, p < .001; 20,555Total SNAP symptoms (T) 0.78 0.64B,M 0.75 0.85 F = 2.22, p < .002; 20,557Total SNAP symptoms (P) 0.56 0.59B 0.70 0.70B ADHD+Anx+ODD/CD F = 4.45, p < .001; 20,555Aggression (P) 0.05 0.05 0.09 0.11 ADHD+ODD/CD F = 1.59, p < .05; 20,558Internalizing symptoms (T) 0.12 0.24M 0.15 0.07 ADHD+Anx F = 1.55, p < .06; 20,524Internalizing symptoms (P) 0.15 0.14B 0.22 0.32B ADHD+Anx+ODD/CD F = 3.34, p < .001; 20,547MASC (child) 0.23 0.25 0.25 0.16 F = 1.17, p < .27; 20,524CDI (child) 0.15 0.14 0.14 0.11 F = 1.40, p < .12; 20,555Columbia Impairment Scale 5.8 5.4B 7.6 9.7 ADHD+Anx+ODD/CD F = 4.23, p < .001; 20,470Needs services (P) 0.76 0.95 0.67 0.92 F = 2.92, p < .001; 20,558Social skills (T) 0.23 0.24B 0.25 0.30 F = 1.68, p < .03; 20,519Social skills (P) 0.14 0.06 0.16 0.17 F = 1.55, p < .05; 20,548Overall composite 7.5 8.5B 9.1 8.6B ADHD+Anx, ADHD+ F = 4.41, p < .001; 20,558

Anx+ODD/CDP-C power assertion 0.22 0.20, 0.32 0.31 F = 1.50, p < .08; 20,544P-C positive affection 0.01 0.07 0.04 0.05 F = 0.70, p < .83; 20,554WIAT Math 2.5 3.7B 1.0B,M 2.5 F = 1.55, p < .06; 20,558WIAT Spelling 0.7 2.7B,C 1.0 1.0 ADHD+Anx F = 0.94, p < .53; 20,558WIAT Reading 1.4 1.2 1.4 1.2 F = 1.48, p < .08; 20,558LD at 14 mo (proportion) 0.10 0.16 0.08 0.16 F = 1.28, p < .19; 20,558ADHD at 14 mo 0.42 0.57 0.50 0.55 F = 3.63, p < .001; 20,558

Note: Boldface values indicate that the MTA treatment type (B = behavioral therapy; M = medication management; C = combined treatment)exerted difference effects on the particular comorbidity groups in that column, compared with the overall effects of that same treatment for allsubjects, irrespective of comorbidity profile. ADHD = attention-deficit/hyperactivity disorder; ANX = anxiety disorders; ODD/CD =oppositional defiant disorder or conduct disorder; T = teacher; P = parent; SNAP = Swanson, Nolan, and Pelham scale; MASC =Multidimensional Anxiety Scale for Children; CDI = Children’s Depression Inventory; WIAT = Wechsler Individual Achievement Test; LD =learning disability.

a n’s.

can be seen in the right-hand column of the table. Reviewof Table 1 suggests that the effects of Cd comorbidity aregenerally more pronounced than Anx effects on baselinecharacteristics and are generally associated with greaterlevels of symptoms (e.g., hyperactive-impulsive symp-toms), lower levels of functioning (teacher-rated socialskills), and overall impairment. Important exceptionswere noted in academic functioning, where ADHD +ANX subjects evidenced sizable effects with lower func-tioning levels. Further review of this table also indicatesthat in almost all instances of additive effects of the twocomorbid conditions (parent total SNAP scores, impair-ment, need for services, teacher-rated social skills, and thecomposite index), effects exerted by Cd were 2-fold ormore larger than those conveyed by Anx.

Analyses of effect sizes of the outcomes criterionrevealed few instances of comorbidity effects on outcomesindependent of MTA treatments. Important exceptionsare the relative lack of improvement over time in teacher-reported inattention and the overall composite indexamong ADHD + ANX and ADHD + ANX + ODD/CDsubjects (table available upon request).

Effect sizes based on comorbidity profiles on the treat-ment response criterion are much more complex, since anygiven variable could have 12 different effect sizes (3 MTAtreatments 3 4 comorbidity profiles). Therefore, effect sizesfor selected variables where comorbidity 3 treatment inter-actions were noted in Table 2 are presented in graphic formin Figure 1. Review of this figure indicates that ADHD +ANX subjects are particularly responsive to Beh, comparedwith subjects in other comorbidity groups, which rarely didas well (i.e., effect size > 0.4) with Beh alone. Figure 1 alsoindicates that ADHD + ANX + ODD/CD subjects werepreferentially responsive to Comb interventions, as indi-cated by a robust response (effect size = 0.92) to Comb andmuch lower effect size for Beh (0.53) and MedMgt (.28) inthe overall composite rating. By and large, the effect sizes ofADHD-only and ADHD + ODD/CD subjects were gen-erally lower than those for ADHD + ANX and ADHD +ANX + ODD/CD subjects. For ADHD-only andADHD + ODD/CD subjects, Beh alone interventionsnever appeared beneficial.

DISCUSSION

Our findings suggest that ADHD children with andwithout ODD/CD and ANX differed on many baselinecharacteristics, outcomes, and response to treatment.

Though ODD/CD comorbidity commonly exerted fairlypowerful main effects on baseline characteristics, it rarelyinteracted with treatment response or outcomes. In con-trast, Anx conditions exerted somewhat less robust effectson baseline characteristics, but frequently interacted withresponse to specific treatments. Concerning our subtypehypothesis, while both Anx and Cd frequently exertedstraightforward main effects, their effects usually did notoperate on the same variables. Thus such effects werefound in only six instances of all variables examined, andthese variables pertained more to the impact of psychopa-thology than psychopathology per se. Taken together, ourfindings—consistent differences between ADHD + ANXand ADHD + ODD/CD groups, rarely found additiveeffects, the impact of Cd and Anx on different variables,and the differential impact of specific treatments on thefour comorbidity types, with some but not all treatmentsyielding effects of sizable magnitude—indicate that theoverall clinical profiles of the comorbidity types are suffi-ciently distinct to support the need to discriminate amongthe comorbid subgroups.

Consistent with our first report of parent-reportedanxiety as a moderating variable on outcomes (MTACooperative Group, 1999b), we found substantial, recur-ring indications of Anx comorbidity (either alone or ininteraction with Cd ) moderating responses to treatment.In addition, however, this analysis, going well beyond ourearlier report, suggests that anxiety may interact withtreatment outcomes across many more variables than firstdescribed, more readily detectable here with Anx 3 Cdinteractions included in the analytic model.

It is of interest to note that Anx exerted relatively mod-est effects on baseline characteristics and outcomes, yetexerted persistent, sometimes sizable effects on response totreatment findings. Reasons for this finding are unclear,but could be due to the fact simply that behavioral inter-ventions are selectively effective for ADHD children withparent-defined anxiety and that these children are qual-itatively different in terms of etiological processes anddevelopmental trajectories. This possibility of qualitativedifferences in these children is further supported by thefindings of WIAT score differences seen in these subjectsat baseline, as well as evidence from other reports thatsuch children may have different physiological reactionsto methylphenidate (Urman et al., 1995).

As a general rule, ANX status appeared to confer cer-tain benefits on ADHD children, regardless of the pres-ence of ODD/CD. Thus, inspection of Table 1 and our

JENSEN ET AL.

154 J . AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 40 :2 , FEBRUARY 2001

outcomes criterion analyses (available upon request) indi-cates that most (40 of 58) Cd 3 Anx interaction effectswere positive (though rarely significant), that is, theyexerted ameliorating effects on concurrent ODD/CD(i.e., ADHD + ANX + ODD/CD versus ADHD +ODD/CD subjects). Furthermore, review of Figure 1indicates that children with Anx tended to be more treat-ment-responsive than ADHD + ODD/CD and evenADHD-only subjects, both in terms of absolute effectsizes and in terms of their relatively robust response to all

treatment modalities. The ADHD + ANX group in par-ticular responded to almost any of the three treatments,with the single exception of academic performance, wherepossible adverse effects of medication management werenoted in the ADHD + ANX group, of some interest givenhistorical reports of adverse effects of medication treat-ments on learning. In contrast, ADHD + ANX + ODD/CD subjects in a number of important instances appearedto derive substantially greater benefits from Comb inter-ventions compared with all other treatments, thus provid-

ADHD COMORBID SUBGROUPS

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 40 :2 , FEBRUARY 2001 155

Fig. 1 Comorbidity type by treatment response: effect sizes across six variables. SNAP = Swanson, Nolan, andPelham scale; ADHD = attention-deficit/hyperactivity disorder; DBD = disruptive behavior disorders; Beh =behavioral therapy; MedMgt = medication management; Comb = combined treatment.

ing some justification of their separate classification fromADHD + ANX subjects.

ADHD-only and ADHD + ODD/CD subjects re-sponded usually only to interventions that included med-ication. While ADHD-only and ADHD + ODD/CDsubjects appeared quite similar in terms of degree of clin-ical change in response to treatment and outcomes, thesizable baseline differences in severity between these twogroups indicates a more problematic prognosis for theADHD/CD group. In fact, additional analyses of out-comes and response to treatment based solely on absolutelevels of 14-month symptoms (rather than differencescores) revealed that these two groups continued to showsubstantial differences at endpoint on a number of criticalvariables, such as parent-reported aggression, need for ser-vices, impairment, social skills, parent–child relations,and the overall composite index (tables available from theauthors upon request). These persistent differences inprognosis may justify the separate classification of thesetwo profiles.

Limitations

In consideration of our findings, a number of caveatsmust be carefully weighed. First, ours was a referred sam-ple, so any inferences that these findings are applicable toADHD comorbidity patterns in the general communityare not warranted. In addition, our analyses of the out-comes criterion “independent of treatment” (i.e., withCC subjects only) must be viewed with caution becausemost CC subjects did indeed obtain some form of treat-ment, and our preliminary findings suggest that thoseseeking treatment benefited slightly more than those whodid not seek care (MTA Cooperative Group, 1999a,b).Another word of caution is warranted concerning thoseanalyses suggesting differential response to treatments asa function of comorbidity, particularly anxiety. As notedelsewhere, very little overlap is found between parent-and child-identified anxiety syndromes (e.g., Jensenet al., 1999). This issue is of consequence because ouradditional analyses (March et al., 2000) of the modera-tor effects of anxiety on MTA outcomes indicate that incontrast to parent-reported anxiety comorbidity, child-reported anxiety syndromes exert little if any moderatingeffects on MTA outcomes. Thus it remains unclearwhether the moderating effects of anxiety reported hereapply to anxiety disorders per se, or to some broader con-struct of parent-identified behavioral and affective dys-regulation. Relatedly, it should be noted that we found

substantially lower rates of affective disorder in the MTAsample than have been reported in other clinical studies(e.g., Biederman et al., 1991), raising concerns about thepossibility of underdiagnosis of these conditions, oralternatively, comorbidity differences between the MTAsample (which drew from a broad range of referralsources) versus samples that have drawn principally frommental health settings.

We note that our decision to combine CD and ODDcases into a single disruptive disorder category is not with-out controversy. However, evidence from clinical studiessuggests that most ADHD children with comorbid CDalso often meet criteria for ODD, which has in fact usuallypreceded CD onset by several years (e.g., Biederman et al.,1996). On the other hand, a number of population-basedstudies have suggested that ODD and CD, while intercor-related 0.70 to 0.76, may nonetheless constitute somewhatseparate dimensions of psychopathology (Quay, 1999).Our decision to combine these two disorders reflects inpart the substantial (but incomplete) overlap between thetwo disorders, as well the lack of power to examine CD asa separate disorder, given our sample sizes.

A final note of caution concerns the use of clinical sam-ples to explore comorbidity. As noted by Angold et al.(1999), clinical samples can be burdened by methodolog-ical artifacts, potentially leading to incorrect estimates ofcomorbidity rates and even mistaken etiological infer-ences. Nonetheless, clinical studies of comorbidity can beuseful to the extent that the conclusions from such studiesare intended to apply only to other clinical samples. Inthis regard, the MTA’s inclusion and exclusion criteria,such as admitting only children with ADHD combinedtype and excluding children with psychosis, with Tou-rette’s disorder, or concurrently being treated with anti-psychotic medications, demarcate some of the possiblelimitations concerning inferences that might be drawn tomore heterogenous clinical samples.

Clinical Implications

Our findings may have considerable implications forfuture research and treatment studies, as well as for the clin-ical process of matching of patients to treatments. As con-structed, our comorbidity groups appear to differ in anumber of areas, including several notable baseline charac-teristics, response to treatment, and outcomes. ADHDchildren with parent-defined anxiety disorders, particularlythose with no accompanying ODD/CD, may have a num-ber of unique characteristics, most notably seen in WIAT

JENSEN ET AL.

156 J . AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 40 :2 , FEBRUARY 2001

performance scores at baseline (also reported by Livingstonet al., 1990) and at 14 months. In addition, this groupevinces a clinical profile that is not a simple function of howit was originally formed (persistent ADHD, teacher-reported inattention at 14 months), and perhaps mostimportantly, in its differential response to treatment. Whilethe discriminating responses to treatments (especially thebehavioral components of Beh and Comb) were frequentlyshared between the ADHD + ANX and ADHD + ANX/CD groups (suggesting a main effect of Anx status), addi-tional interactions with treatment response were found byseparating Anx children into the ADHD + ANX andADHD + ANX/CD groups, compared with our originalreport (MTA Cooperative Group, 1999b).

A simple “rule of thumb” that summarizes these find-ings suggests that if a child presents with an ADHD +ANX profile, all interventions are likely to be effective. Ifa child presents with ADHD-only or ADHD + ODD/CD, treatments with medication appear especially indi-cated, and Beh alone strategies may be contraindicated.Finally, if a child presents with ADHD + ANX + ODD/CD, Comb interventions may offer substantial advan-tages over other treatments, particularly in overall impair-ment and functioning outcomes.

Our findings suggest that more precise matching ofpatients to treatment using patients’ comorbidity pro-files may mitigate initial clinical uncertainty, reduce thenumber of therapeutic trials until a workable treatmentis found, and yield larger treatment gains for specificpatients. Future pathophysiological and genetic studiesmay benefit by these discriminations, to the extent thatthese clinical profiles reflect different etiological anddevelopmental processes.

The NIMH Collaborative Multisite Multimodal Treatment Study ofChildren With Attention-Deficit/Hyperactivity Disorder (MTA) is a coopera-tive treatment study performed by six independent research teams in collab-oration with the staff of the Division of Services and Intervention Research ofthe NIMH, Rockville, MD, and the Office of Special Education Programs(OSEP) of the U.S. Department of Education (DOE). The NIMH PrincipalCollaborators are Peter S. Jensen, M.D., L. Eugene Arnold, M.Ed., M.D.,John E. Richters, Ph.D., Joanne B. Severe, M.S., Donald Vereen, M.D., andBenedetto Vitiello, M.D. Principal Investigators and Coinvestigators from thesix sites are as follows: University of California at Berkeley/San Francisco(UO1 MH50461): Stephen P. Hinshaw, Ph.D., Glen R. Elliott, M.D.,Ph.D.; Duke University (UO1 MH50447): C. Keith Conners, Ph.D., KarenC. Wells, Ph.D., John S. March, M.D., M.P.H.; University of California atIrvine/Los Angeles (UO1 MH50440): James M. Swanson, Ph.D., Dennis P.Cantwell, M.D., Timothy Wigal, Ph.D.; Long Island Jewish MedicalCenter/Montreal Children’s Hospital (UO1 MH50453): Howard B. Abikoff,Ph.D., Lily Hechtman, M.D.; New York State Psychiatric Institute/Columbia

University/Mount Sinai Medical Center (UO1 MH50454): Laurence L.Greenhill, M.D., Jeffrey H. Newcorn, M.D.; University of Pittsburgh (UO1MH50467): William E. Pelham, Ph.D., Betsy Hoza, Ph.D. Helena C.Kraemer, Ph.D. (Stanford University) is statistical and design consultant. TheOSEP/DOE Principal Collaborator is Ellen Schiller, Ph.D.

REFERENCES

Angold A, Costello EJ, Erkanli A (1999), Comorbidity. J Child PsycholPsychiatry 40:57–87

Biederman J, Faraone SV, Milberger S et al. (1996), Is childhood oppositionaldefiant disorder a precursor to adolescent conduct disorder? Findings froma four-year follow-up study of children with ADHD. J Am Acad ChildAdolesc Psychiatry 35:1193–1204

Biederman J, Newcorn J, Sprich S (1991), Comorbidity of attention deficithyperactivity disorder with conduct, depressive, anxiety, and other dis-orders. Am J Psychiatry 148:564–577

Bird H, Gould M, Staghezza-Jaramillo B (1990), The comorbidity of ADHDin a community sample of children aged 6 through 16 years. J Child FamStud 3:365–378

Bird HR, Andrews H, Schwab-Stone M et al. (1996), Global measures ofimpairment for epidemiologic and clinical use with children and adoles-cents. Int J Methods Psychiatr Res 6:295–307

Buitelaar J, Van der Gaag R, Swaab-Barneveld HS, Kuiper M (1995), Predictionof clinical response to methylphenidate in children with attention-deficithyperactivity disorder. J Am Acad Child Adolesc Psychiatry 34:1025–1032

Cantwell DP (1995), Child psychiatry: introduction and overview. In:Comprehensive Textbook of Psychiatry/VI, Kaplan HI, Sadock BJ, eds.Baltimore: Williams & Wilkins, pp 2151–2154

Conners CK, Epstein JN, March JS et al. (2001), Multimodal treatment ofADHD in the MTA: an alternative outcome analysis. J Am Acad ChildAdolesc Psychiatry

Diamond I, Tannock R, Schachar R (1999), Response to methylphenidate inchildren with ADHD and comorbid anxiety. J Am Acad Child AdolescPsychiatry 38:402–409

DuPaul G, Barkley R, McMurray M (1994), Response of children withADHD to methylphenidate: interaction with internalizing symptoms. JAm Acad Child Adolesc Psychiatry 33:894–903

Furman W, Giverson R (1995), Identifying the links between parents andtheir children’s sibling relationships. In: Close Relationships and Socio-Emotional Development, Shulman S, ed. Norwood, NJ: Ablex, pp 95–108

Gresham FM, Elliott SN (1989), Social Skills Rating System: Parent, Teacher,and Child Forms. Circle Pines, MN: American Guidance Systems

Hinshaw SP (1987), On the distinction between attentional problems/hyperactivity and conduct problems/aggression in child psychopathology.Psychol Bull 101:443–463

Jensen P, Martin C, Cantwell D (1997), Comorbidity in ADHD: implicationsfor research, practice, and DSM-IV. J Am Acad Child Adolesc Psychiatry36:1065–1079

Jensen PS, Rubio-Stipec M, Canino G et al. (1999), Parent and child contri-butions to diagnosis of mental disorder: are both informants always nec-essary? J Am Acad Child Adolesc Psychiatry 38:1569–1579

Jensen PS, Ussery T, Bloedau L, Richters J (1991), Stressful life events versusfamily processes: environmental predictors of child dysfunction. J AmAcad Child Adolesc Psychiatry 30:303–309

Kovacs M, Beck AT (1977), An empirical-clinical approach toward a definitionof childhood depression. In: Depression in Childhood, Schulterbrandt JG,Raskin A, eds. New York: Raven, pp 1–25

Kraemer HC (1995), Statistical issues in assessing comorbidity. Stat Med14:721–733

Livingston R, Dykman R, Ackerman P (1990), The frequency and significanceof additional self-reported psychiatric diagnoses in children with ADD. JAbnorm Child Psychol 18:465–478

March JS, Parker JD, Sullivan K, Stallings P, Conners CK (1997), The Multi-dimensional Anxiety Scale for Children (MASC): factor structure, relia-bility, and validity. J Am Acad Child Adolesc Psychiatry 36:554–565

ADHD COMORBID SUBGROUPS

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 40 :2 , FEBRUARY 2001 157

March JS, Swanson JM, Arnold LE et al. (2000), Anxiety as a predictor andoutcome variable in the MTA study: analysis using hierarchical linearmodeling. J Abnorm Child Psychol 28:527–541

MTA Cooperative Group (1999a), 14-Month randomized clinical trial oftreatment strategies for attention deficit hyperactivity disorder. Arch GenPsychiatry 56:1073–1086

MTA Cooperative Group (1999b), Moderators and mediators of treatmentresponse for children with attention-deficit/hyperactivity disorder. ArchGen Psychiatry 56:1088–1096

Pliszka S (1989), Effect of anxiety on cognition, behavior, and stimulantresponse in ADHD. J Am Acad Child Adolesc Psychiatry 28:882–887

Pliska S (1992), Comorbidity of ADHD and overanxious disorder. J Am AcadChild Adolesc Psychiatry 31:197–203

Quay H (1999), Classification of the disruptive behavior disorders. In: Hand-book of Disruptive Behavior Disorders, Quay HC, Hogan AE, eds. NewYork: Kluwer Academic/Plenum Publishers, pp 3–21

Robins E, Guze SB (1970), Establishment of diagnostic validity in psychiatricillness: its application to schizophrenia. Am J Psychiatry 126:983–987

Shaffer D, Fisher P, Dulcan M et al. (1996), The second version of the NIMHDiagnostic Interview Schedule for Children (DISC-2). J Am Acad ChildAdolesc Psychiatry 35:865–877

Tannock R, Ickowicz A, Schachar R (1995), Differential effects of methyl-phenidate on working memory in ADHD children with and withoutcomorbid anxiety. J Am Acad Child Adolesc Psychiatry 34:886–896

Urman R, Ickowicz A, Fulford P, Tannock R (1995), An exaggerated cardio-vascular response to methylphenidate in ADHD children with anxiety. JChild Adolesc Psychopharmacol 5:29–37

Wechsler D (1992), Wechsler Individual Achievement Test: Manual. SanAntonio, TX: Psychological Corporation

Wells CK, Epstein JN, Hinshaw SP et al. (2000), Parenting and family stresstreatment outcomes in attention deficit hyperactivity disorder (ADHD):an analysis in the MTA study. J Abnorm Child Psychol 28:543–553

Woolston JL, Rosenthal SL, Riddle MA, Sparrow SS, Cicchetti D, RosenthalLD (1989), Childhood comorbidity of anxiety/affective disorders andbehavior disorders. J Am Acad Child Adolesc Psychiatry 28:707–713

JENSEN ET AL.

158 J . AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 40 :2 , FEBRUARY 2001