Regular Article

Psychiatry and Clinical Neurosciences

(2003),

57

, 504–510

Blackwell Science, LtdOxford, UKPCNPsychiatric and Clinical Neurosciences1323-13162003 Blackwell Science Pty Ltd575October 2003

1155Mood stabilization in children with BDP. Davanzo et al.10.1046/j.1323-1316.2003.01155.x

Original Article504510BEES SGML

Correspondence address: Dr Pablo Davanzo, UCLA Neuropsychi-atric Institute, 760 Westwood Plaza, 48-243C, Los Angeles, CA90024-1759, USA. Email: pdavanzo@mednet.ucla.edu

Received 18 November 2002; revised 31 January 2003; accepted16 February 2003.

Regular Article

Mood stabilizers in hospitalized children with bipolar disorder: A retrospective review

PABLO DAVANZO,

MD

,

1

BRENT GUNDERSON,

BS

,

2

THOMAS BELIN,

PhD

,

3

JIM MINTZ,

PhD

,

1

CAROLY PATAKI,

MD

,

1

DEREK OTT,

MD

,

1

CATHERINE EMLEY-AKANNO,

RN

,

1

NAZILA MONTAZERI,

BA

,

4

JAMI OPPENHEIMER,

BA

1

AND MICHAEL STROBER,

PhD

1

1

Department of Psychiatry and Behavioral Sciences, School of Medicine,

3

Department of Biostatistics, University of California, Los Angeles,

2

Department of Psychology, University of Southern California, Los Angeles and

4

Graduate School of Education and Psychology, Pepperdine University, Malibu, California, USA

Abstract

A paucity of naturalistic data supported a rationale for the present retrospective review of clinicalchanges during hospitalization in 44 bipolar pre-adolescents, treated with monotherapy lithium,carbamazepine (CBZ) or divalproex sodium (DVP). Daily staff progress notes and dischargesummaries on each patient were read by four trained clinicians blind to treatment group, and ratedaccording to the Clinical Global Impression Improvement (CGI-I) scale. Consensus rating wasmeasured by kappa reliability. Data were analyzed using a general linear model (

SAS

MIXED

)analysis of variance (

ANOVA

) with repeated measures. The medication groups did not differ inlength of hospitalization, overall severity of illness at the time of admission, or comorbidity. Priortreatment was considered as a covariate. Each group approached serum therapeutic levels at day7 of the medication period. The estimated mean CGI-I scores for CBZ were systematically higher(i.e. worse) than those for lithium and DVP, which overlapped and crossed over time. The differ-ence became increasingly apparent and was statistically significant by week 2 (

P

=

0.036). Thepresent study was limited in that the sample sizes, particularly in the case of CBZ, were small,commensurate with the low prevalence of the disorder. Lack of structured interviews, as anindependent assessment of diagnoses was an intrinsic limitation of the study. Although constrainedby its retrospective nature, our findings suggest that by week 2 of hospitalization both lithium andDVP may be more efficacious than CBZ in bipolar pre-adolescents. Any significant finding mustbe viewed as tentative and subject to confirmation in other studies.

Key words

anticonvulsants, mania, medication pre-adolescents, treatment, valproate.

INTRODUCTION

Although the lifetime prevalence of bipolar disorder(BD) in adolescents has been estimated to be approx-imately 1%

1

there is a paucity of medication studies inthis population.

2

In one prospective controlled study oflithium in adolescents with BD (and substance depen-dency), 46% of the subjects were responders to lithiumcompared to 8.3% who responded to placebo, as mea-

sured by improved screening of substance use.

3

Previ-ously, significant differences in treatment efficacy hadbeen reported in prepubertal-onset bipolar adoles-cents compared to an adolescent-onset BD cohortafter 4 weeks of lithium treatment.

4

By contrast, Kaf-antaris

et al.

showed that the presence of a prepuber-tal-onset psychiatric diagnosis did not moderate thelikelihood of response to lithium in a sample of 48adolescents with BD.

5

The pediatric manic spectrum symptoms appear tobe selectively responsive to mood stabilizers, in con-trast to stimulants, antidepressants, and antipsychot-ics.

6

A comparison of lithium versus divalproex sodium(DVP) and carbamazepine (CBZ) was studied byKowatch

et al.

, who completed an open randomized

Mood stabilization in children with BD 505

trial of mood stabilizers in children with type I and typeII BD.

7

The reported effect size of 1.63 for DVP, 1.06for lithium, and 1.00 for CBZ, suggested better efficacyfor DVP than for lithium and CBZ in this sample(mean age 11.4 years).

7

Contrary to previously reported poor selectiveresponse to antipsychotics for pediatric manic spec-trum symptoms

6

Frazier

et al.

reported that risperidone(mean dose 1.7 mg over 6.1 months) was therapeutic ina group of 28 children with BD, mean age 10.4 years.

8

Likewise, DelBello

et al.

found that the combination ofDVP with quetiapine was more effective than DVPalone after 6 weeks of controlled acceleration pharma-cotherapy for the treatment of mania in adolescents,

9

measured by change in Young Mania Rating Scale(YMRS) scores.

10

In light of these findings, and the paucity of natural-istic data on effectiveness of pharmacotherapy in pre-adolescents diagnosed with BD, we sought to studythe degree of clinical response to lithium, CBZ, andDVP in a group of hospitalized pre-adolescents withBD. A retrospective chart review was employed torate daily changes in the Clinical Global Impression(CGI) scale.

METHODS

The charts of 381 children consecutively admitted tothe University of California, Los Angeles (UCLA)Neuropsychiatric Institute and Hospital betweenJune of 1995 and February of 1998 were screened.One hundred and forty-three (37.5%) were treatedwith mood stabilizers during the index admission.Forty-four children met a priori inclusion criteria forretrospective review. These criteria included age atadmission between 5 and 12 years, a

Diagnostic andStatistical Manual of Mental Disorders

(4th edn;DSM-IV) discharge diagnosis of BD, and a mini-mum of 7 days on monotherapy lithium, CBZ, orDVP. Exclusion criteria included children with a pri-mary diagnosis of mental retardation, obsessive–compulsive disorder, pervasive developmental disor-der, post-traumatic stress disorder, Tourette’s disor-der, schizophrenia, seizure disorder, medical orneurological disease. None of the subjects receivedconcomitant neuroleptics, antidepressant, or othermedication therapy during the full period of hospi-talization. In all subjects the use of neuroleptics wasrestricted to no more than one consecutive day of‘as needed for acute exacerbation of illness’ (PRN)antipsychotic medication during their inpatient hos-pitalization. Categorical use (i.e. typical and atypical1 day PRN neuroleptic use

vs

no use) was analyzedwith

ANOVA

.

Assessment and diagnosis

Diagnoses of BD extracted from the chart review werebased on formal DSM-IV criteria for BD type I, andtype II as specified by the inpatient team at the time ofdischarge. Demographic characteristics (i.e. age, gen-der, ethnicity, family history of psychiatric disorders,history of developmental disorders), as well as initialoverall severity of illness, presence or absence ofpsychosis during hospitalization, comorbid diagnoses,length of stay in the hospital, length of medicationperiod, treatment with lithium, CBZ or DVP prior tothe index hospitalization, and the type of mood-stabi-lizer treatment used during the hospitalization (mono-therapy lithium, DVP or CBZ), were extracted for theanalysis by review of the discharge summaries on eachpatient, daily staff progress notes, and physician orders.Three trained child and adolescent psychiatrists (PD,CP, DO) and one research assistant (BG: bachelor’slevel, minimum 2 years of experience in pediatric psy-chiatric interviewing) extracted the chart data. TheCGI–Improvement (CGI-I) scale was used to monitortherapeutic efficacy by assessment of patient’s behav-ior for each hospital day. Results are here reported onCGI-I assessment. Daily staff progress notes (nursingstaff and child psychiatry fellows) were read by eachof the clinicians blind to treatment group, and rated (asan average for each day in comparison to the day prior)as follows: CGI-I: 1, very much improved; 2, muchimproved; 3, minimally improved; 4, no change; 5, min-imally worse; 6, much worse; 7, very much worse.

Consensus rating among the four raters was arrivedat the end of the evaluation on 10 charts, and measuredby a kappa reliability of 0.71. Kappa reliability wascalculated as follows. The chance agreement that tworaters would consider a subject ‘much improved’ wascalculated by multiplying the proportions each raterwould find as ‘much improved’ from a sample of 10charts (y/10

¥

z/10) and multiplied by the total numberof charts assessed (i.e. 10). Step 2 involved calculationsof the charts that would be deemed as ‘not muchimproved’ (m/10

¥

n/10)

¥

10 (i.e. by multiplying theproportion each rater found ‘not improved’ by the totalnumber of charts). Step 3 involved adding the previousnumbers found in steps 1 and 2 and dividing by thetotal number of observations (charts) to obtain a pro-portion for chance agreement. Step 4 involved calcu-lating the potential agreement beyond chance with thefollowing formula (100%

-

chance agreement/dividedby total)

=

kappa.

Data analysis

The dependent measures were daily ratings of theCGI-I. Data were analyzed using a general linear

506 P. Davanzo

et al

.

model (

SAS

MIXED

) analysis of variance (

ANOVA

)with repeated measures. Data after day 13 (i.e. after2 weeks) were excluded because after that time attri-tion was significant and associated with outcomes (i.e.discharge from the hospital). The within-subject com-ponent of the design was stratified by week of hospi-talization because it was expected that meaningfuldrug effects might not emerge until week 2. Thus, thedesign factors were medication group (A, B, C) andhospitalization week (1–2) with days as a repeatedmeasures factor. The mixed model

ANOVA

yields esti-mated (least-square) means for each group on eachday. The method permits analysis of repeated measureswith some missing data; approximately one-third of thesample was lost by day 13.

Inspection of the results suggested that most of thedifferences observed were due to one drug (CBZ) dif-fering from the others. Accordingly, specific contrastswere performed to compare CBZ to the other twodrugs pooled both across the whole study period, andseparately in weeks 1 and 2. Any a priori differencesbetween medication groups served as covariates in thefinal analysis. Specifically, both the number of days inthe hospital prior to starting medication and the num-ber of patients who had had prior treatment with DVP(compared to lithium and CBZ) served as covariatesin the final analysis.

RESULTS

The demographic variables for the entire sample(

n

=

44) by type of medication treatment were as fol-lows: lithium (

n

=

25), CBZ (

n

=

6) and DVP (

n

=

13).Overall, the 44 children who were treated with mood-stabilizing agents had a mean age of 9.2 years(SD

=

2.1); 84.0% (

n

=

37) were male, and the ethniccomposition was 77.3% Caucasian (

n

=

34), 13.6%African–American (

n

=

6), 6.8% Latino (

n

=

3), and2.3% Asian American (

n

=

1). Twenty-one (52.5%) ofthe sample received at least one comorbid diagnosis;specifically, 16 children (36%) were diagnosed withattention deficit–hyperactivity disorder (ADHD), four

children (9.1%) with conduct disorder (CD), four(9.1%) with intermittent explosive disorder (IED),four (9.1%) with oppositional defiant disorder (ODD),two (4.5%) with impulse control disorder not other-wise specified (NOS), one (2.3%) with psychosis NOSand one (2.3%) with anxiety disorder NOS. Sixteenchildren (36%) received no more than 1 consecutiveday of PRN (i.e. as needed) antipsychotic medicationduring their hospitalization, and the PRN use of neu-roleptics across the three medication groups was notsignificantly different (

P

=

0.8).The average length of stay in the hospital for the

entire sample was 18.9 days (SD

=

9.1, range 9–50 days). The average number of days on medicationwhile in the hospital for the entire sample was 16.4 days(SD

=

7.5, range: 9–42 days). The average number ofdays in the hospital before starting medication for theentire sample was 3.5 days (SD

=

5.0, range 0–21 days).By

ANOVA

there were no treatment group differ-ences in age, gender, ethnicity, family history of psychi-atric disorders, history of developmental disorders,initial overall severity of illness, presence or absence ofpsychosis during hospitalization, number of PRN med-ications, comorbid diagnoses, length of stay in the hos-pital, length of medication period, or treatment withlithium, or CBZ prior to the index hospitalization.

As illustrated in Table 1, no differences were seen bymedication group in the length of stay in the hospital,length of medication period, nor the average numberof days in the hospital prior to starting medication,except for the lithium group, which had significantlylonger number of days in the hospital prior to startingmedication compared to the DVP group (

P

=

0.01).Table 2 shows the average blood serum levels for eachmedication group at days 7 and 13 of the hospitaliza-tion. All three groups achieved serum levels above theminimum standard therapeutic range at week 1 of thehospitalization and were comparatively therapeutic(although not achieving the maximum within the ther-apeutic range) by day 13 of the hospitalization. Table 3shows the medication groups by prior medication trials.A significantly higher number of patients had had prior

Table 1.

Interval characteristics by medication group

Medication Group Hospital days (SD) Medication days (SD)Days in hospital

prior to medication (SD)

Lithium 20.0 (11.1) 14.8 (8.9) 6.1 (5.4)*Carbamazepine 15.0 (4.1) 13.0 (3.5) 3.0 (3.3)Divalproex sodium 18.7 (5.7) 18.2 (5.4) 1.5 (1.4)*Entire sample 18.9 (9.1) 16.4 (7.5) 3.4 (4.9)

*

P

=

0.01 (lithium

vs

divalproex sodium, days prior to starting medications).

Mood stabilization in children with BD 507

treatment with DVP compared to lithium and CBZ.Eight of these patients belonged to the DVP group.One patient in the lithium group, one in the CBZgroup, and two in the DVP group had trials with threemood stabilizers prior to the index hospitalization(

P

=

0.25). The two covariates included in the analysis(i.e. no. days prior to starting lithium and prior treat-ment with DVP) did not have any role in the outcomesor statistical findings.

The data are presented in Fig. 1. The grid at thebottom of Fig. 1 details the number of subjects pergroup, per day of hospitalization, available for theanalysis. The overall group included 27 subjects avail-able on day 13 of the hospitalization. Inspection of thefigure suggests that outcomes were poorer with CBZthan the other two drugs. The estimated mean CGIscores for CBZ were systematically higher (i.e. worse)than those for lithium and DVP, which overlapped andcrossed over time. The overall

ANOVA

main effect forgroup was not significant (

F

=

2.23, d.f.

=

2,41,

P

=

0.12).However, the specific contrast between CBZ and theother drugs pooled was (

F

=

4.12, d.f.

=

1,41,

P

=

0.05).Similarly, although the group–week interaction was notsignificant (

F

=

2.11, d.f.

=

2,41,

P

=

0.13), the contrast

between CBZ and the others was significant only inweek 2 (week 2 contrast t

=

2.26, d.f.

=

41,

P

=

0.03;week 1 contrast t

=

1.50, d.f.

=

41, p

=

0.14).

DISCUSSION

This retrospective review randomly screened 381 chil-dren admitted to the UCLA Neuropsychiatric Instituteand Hospital over 3 years and selected 141 who weretreated with mood stabilizers. From this subgroup, 44pre-adolescents who received the diagnosis of bipolar-ity at the end of their hospitalization, and who weretreated with monotherapy lithium, DVP or CBZ wereanalyzed in detail. Our method of using nurse ratingsbased on direct observation of ward behavior overtime has been noted to be more sensitive than inter-view ratings for detecting characteristics of mixed andmanic episodes.

11

To the best of our knowledge this isthe largest retrospective review conducted so far inhospitalized pre-adolescents diagnosed with juvenile-onset BD.

Because recent controlled studies have suggestedthat olanzapine

12

and risperidone

13

may enhance anti-manic response when added to lithium or DVP,

14

we

Table 2.

Mean blood serum levels at days 7 and 13 of hospitalization

Medication groupMean blood serum

levels and SD at day 7Mean blood serum

levels and SD at day 13Standard therapeutic

levels

Lithium (mEq/L) 0.75

±

0.20 0.85

±

0.19 1.0–1.4Carbamazepine (

m

g/mL) 6.75

±

1.68 7.33

±

1.50 4–14Divalproex sodium (

m

g/mL) 86.38 ± 11.20 85.33 ± 10.10 50–125

Table 3. Medication trials prior to index hospitalization

Lithium (prior) CBZ (prior) DVP (prior)

LithiumCount 9 6 6% within group 36.0% 24.0% 24.0%

CBZCount 4 3 1% within group 66.7% 50.0% 16.7%

DVPCount 4 3 8% within group 30.8% 23.1% 61.5%

TotalCount 17 12 15% within entire sample 38.6% 27.3% 34.1%

Pearson c2 0.30 0.40 0.04*

CBZ, carbamazepine; DVP, divalproex sodium.

508 P. Davanzo et al.

chose to exclude patients who had more than one con-secutive day of ‘as needed’ (i.e. PRN) antipsychoticmedication during their inpatient hospitalization. Con-sequently differences in antipsychotic dosages werenot included for analysis, beyond the ANOVA showingno interactions between clinical response and 1 dayPRN antipsychotic usage.

Our results suggest that CBZ is less effective during2 weeks of hospitalization than lithium or DVP, whichhad a very similar clinical course. The differenceappeared throughout the 2 week course but becameincreasingly apparent and was statistically significantonly by week 2. These findings are consistent with arecent outpatient prospective double blind study ofchildren and adolescents with BD7 in which, on intent-to-treat analysis (n = 32, monotherapy for 5 weeks),34% of youngsters on CBZ were considered respond-ers by mean CGI and YMRS scores, compared to 42%for lithium and 46% for DVP.7

Studies in adults with BD have suggested the efficacyof CBZ in the acute15 and prophylactic treatment ofBD type I and type II.16 Nevertheless, no controlledstudies of the treatment of pediatric mood disorderswith CBZ have been conducted so far.2 A negativecontrolled study of CBZ for the treatment of childrenwith conduct disorder17 suggests a possible concor-dance with our comparatively lower response rate toCBZ in children with BD compared to lithium andDVP.

Although Geller et al. have noted an early adoles-cent BD phenotype consistent of high prevalence ofmixed mania and rapid cycling,18 the retrospectivenature of the present study did not permit the conduct-ing of accurate exploratory cluster analysis of subtypesof bipolarity. However, if we speculate that our pre-adolescent sample was predominantly mixed, the over-

all anticipated acute treatment response to DVP wouldhave been superior to lithium;19 a finding not confirmedby our data.

The pattern of comorbidity found in pediatric maniais unique, especially in its overlap with ADHD anddisruptive behavior disorders.20 Several investigatorshave reported high rates of ADHD symptoms in pre-adolescents,21 adolescents22 and adults with BD type I.23

The reported comorbidity of ADHD in pediatricpatients with BD varies from studies describing 22%,24

29%,5 30%,4 and 59%25 comorbidity in adolescents, to89% in pre-adolescents.26 Most of these studies havereported lifetime comorbidity, in comparison to thepresent study in which we report episode comorbidity.This may explain the discrepancy between our comor-bidity of 36% for ADHD and the higher percentagesdescribed in former studies.

Overall, the sample analyzed does not appear toinclude a subgroup of ‘medication resistant’ childrenbecause the number of children in each cell whoreceived prior trials with three mood stabilizers (sug-gesting medication refractoriness) was not significantlydifferent across groups. The use of DVP monotherapyin the prior medication trials was significantly higherthan the other two groups, and was considered a cova-riate in the final analysis. Blood serum levels for eachmedication group at days 7 and 13 of the medicationperiod suggest that each group was within (low-end)standard therapeutic range at week 1 and at week 2.

Limitations

Our findings are related to pre-adolescents diagnosedwith BD (in a hospital setting) and may not be repre-sentative of adolescents or adults with BD. Twoimportant cautions must be added. The sample sizes,

Figure 1. Mean daily Clinical GlobalImpressions-Global Improvement(CGI-I) by medication group

Mood stabilization in children with BD 509

particularly in the case of CBZ, were small. Thismakes any significant finding even more salient. How-ever, these contrasts reported were post hoc andbased on inspection of the results (the statisticalresults reported are not adjusted). They must accord-ingly be viewed as tentative and subject to confirma-tion in other studies.

Additional limitations of the present retrospectivechart review include the differences in record keepingacross staff and/or physicians, which may have led tosome omissions of useful information, and the assess-ment of clinical course by categorical versus dimen-sional ratings. An intrinsic limitation of the study is theabsence of structured interviews as an independentassessment of diagnoses. Demographically, our samplewas overrepresented by Caucasian children, given theprevalence of Hispanic Americans, African–Ameri-cans, and Asian Americans who live in the sampledarea. This disparity may represent the poor access toinpatient resources of these minority groups.

Clinical implications

Our findings suggest that the first week of inpatienthospitalization does not show robust evidence of clin-ical improvement (as measured by CGI improvement)for any of the three ‘first-line’ mood stabilizers used inpre-adolescents with mood disorders. However, thereis a suggestion that by week 2 of hospitalization bothlithium and DVP may be more efficacious than CBZin the treatment of pre-adolescents with BD. A possi-ble implication is that the titration of mood stabilizers(as tolerated by side-effects) be aggressive, and thatclinicians adopt a comprehensive approach to diagnos-tic evaluation during the first week of hospitalization,a time in which the effect of mood stabilizers may notbe evident.

ACKNOWLEDGMENTS

This project was funded by a K01 NIMH grantMH01601, and a Stanley Medical Research Instituteaward, 2001–02 to PD. We wish to thank Gabrielle A.Carlson, MD, and F. Xavier Castellanos, MD for theirvaluable comments about the manuscript.

REFERENCES

1. Lewinsohn PM, Klein DN, Seeley JR. Bipolar disordersin a community sample of older adolescents: Prevalence,phenomenology, comorbidity, and course. J. Am. Acad.Child Adolesc. Psychiatry 1995; 34: 454–463.

2. Davanzo PA, McCracken JT. Mood stabilizers in thetreatment of juvenile bipolar disorder. Advances and

controversies. Child Adolesc. Psychiatr. Clin. N. Am.2000; 1: 159–182.

3. Geller B, Cooper TB, Sun K et al. Double-blind andplacebo-controlled study of lithium for adolescent bipo-lar disorders with secondary substance dependency. J.Am. Acad. Child Adolesc. Psychiatry 1998; 37: 171–178.

4. Strober M, Morrell W, Burroughs J, Lampert C, Dan-forth H, Freeman R. A family study of bipolar I disorderin adolescence. Early onset of symptoms linked toincreased familial loading and lithium resistance. J.Affect. Disord. 1988; 15: 255–268.

5. Kafantaris V, Coletti DJ, Dicker R, Padula G, Pollack S.Are childhood psychiatric histories of bipolar adoles-cents associated with family history, psychosis, andresponse to lithium treatment? J. Affect. Disord. 1998;51: 153–164.

6. Biederman J, Mick E, Bostic J et al. The naturalisticcourse of pharmacologic treatment of children with man-iclike symptoms: A systematic chart review. J. Clin. Psy-chiatry 1998; 59: 628–637.

7. Kowatch RA, Suppes T, Carmody TJ et al. Effect size oflithium, divalproex sodium, and carbamazepine in chil-dren and adolescents with bipolar disorder. J. Am. Acad.Child Adolesc. Psychiatry 2000; 39: 713–720.

8. Frazier JA, Meyer MC, Biederman J et al. Risperidonetreatment for juvenile bipolar disorder: A retrospectivechart review. J. Am. Acad. Child Adolesc. Psychiatry1999; 38: 960–965.

9. DelBello MP, Schwiers ML, Rosenberg HL, StrakowskiSM. A double blind, randomized, placebo-controlledstudy of quetiapine as adjunctive treatment for adoles-cent mania. J. Am. Acad. Child Adolesc. Psychiatry 2002;41: 1216–1223.

10. Young RC, Biggs JT, Ziegler VE, Meyer DA. A ratingscale for mania: Reliability, validity and sensitivity. Br. J.Psychiatry 1978; 133: 429–435.

11. Swann AC, Bowden CL, Calabrese JR, Dilsaver SC,Morris DD. Pattern of response to divalproex, lithium,or placebo in four naturalistic subtypes of mania. Neu-ropsychopharmacology 2002; 26: 530–536.

12. Tohen M, Jacobs TG, Grundy SL et al. Efficacy of olan-zapine in acute bipolar mania: A double-blind, placebo-controlled study. The Olanzipine HGGW Study Group.Arch. Gen. Psychiatry 2000; 57: 841–849.

13. Segal J, Berk M, Brook S. Risperidone compared withboth lithium and haloperidol in mania: A double-blindrandomized controlled trial. Clin. Neuropharmacol.1998; 21: 176–180.

14. Kupka RW, Nolen WA, Altshuler LL et al. The StanleyFoundation Bipolar Network. 2. Preliminary summaryof demographics, course of illness and response tonovel treatments. Br. J. Psychiatry Suppl. 2001; 41:S177–S183.

15. Brown D, Silverstone T, Cookson J. Carbamazepinecompared to haloperidol in acute mania. Int. Clin. Psy-chopharmacol. 1989; 4: 229–238.

16. Greil W, Kleindienst N. The comparative prophylacticefficacy of lithium and carbamazepine in patients with

510 P. Davanzo et al.

bipolar I disorder. Int. Clin. Psychopharmacol. 1999; 14:277–281.

17. Cueva JE, Overall JE, Small AM, Armenteros JL, PerryR, Campbell M. Carbamazepine in aggressive childrenwith conduct disorder: A double-blind and placebo-con-trolled study. J. Am. Acad. Child Adolesc. Psychiatry1996; 35: 480–490.

18. Geller B, Zimerman B, Williams M et al. DSM-IV maniasymptoms in a prepubertal and early adolescent bipolardisorder phenotype compared to attention-deficit hyper-active and normal controls. J. Child Adolesc. Psychop-harmacol. 2002; 12: 11–25.

19. Keck PE Jr, Nabulsi AA, Taylor JL et al. A pharmaco-economic model of divalproex vs. lithium in the acuteand prophylactic treatment of bipolar I disorder. J. Clin.Psychiatry 1996; 57: 213–222.

20. Spencer TJ, Biederman J, Wozniak J, Faraone SV,Wilens TE, Mick E. Parsing pediatric bipolar disorderfrom its associated comorbidity with the disruptivebehavior disorders. Biol. Psychiatry 2001; 49: 1062–1070.

21. Biederman J, Munir K, Knee D et al. High rate of affec-tive disorders in probands with attention deficit disorder

and in their relatives: A controlled family study. Am. J.Psychiatry 1987; 144: 330–333.

22. West SA, McElroy SL, Strakowski SM, Keck PE Jr,McConville BJ. Attention deficit hyperactivity disorderin adolescent mania. Am. J. Psychiatry 1995; 152: 271–273.

23. Winokur G, Coryell W, Endicott J, Akiskal H. Furtherdistinctions between manic-depressive illness (bipolardisorder) and primary depressive disorder (unipolardepression). Am. J. Psychiatry 1993; 150: 176–181.

24. Kutcher S, Robertson HA, Bird D. Premorbid function-ing in adolescent onset bipolar I disorder: A preliminaryreport from an ongoing study. J. Affect. Disord. 1998; 51:137–144.

25. Faraone SV, Biederman J, Mennin D, Wozniak J, Spen-cer T. Attention-deficit hyperactivity disorder with bipo-lar disorder: A familial subtype? J. Am. Acad. ChildAdolesc. Psychiatry 1997; 36: 1378–1387.

26. Wozniak J, Biederman J, Mundy E, Mennin D, FaraoneSV. A pilot family study of childhood-onset mania. J.Am. Acad. Child Adolesc. Psychiatry 1995; 34: 1577–1583.