Four-Year Prospective Outcome and Natural History of Mania in ChildrenWith a Prepubertal and Early Adolescent Bipolar Disorder Phenotype FREE

In the approximately 50 years since lithium salts were introduced, therehave been several 4- to 5-year prospective follow-up studies^1- 6 ofolder adolescent and adult-onset bipolar disorder (BP). In contrast, to ourknowledge, this is the first prospective 4-year follow-up study of a samplesystematically ascertained for a prepubertal and early adolescent BP (PEA-BP)phenotype. Because the area of child mania has been highly contentious, aninvestigation of longitudinal diagnostic validation was clearly warranted.^7- 9 Although once thoughtrare, in the National Institute of Mental Health (NIMH)–funded "Phenomenologyand Course of Pediatric Bipolar Disorders," 6.9% of consecutive new cases,ascertained from multiple pediatric and psychiatric facilities, fit PEA-BPby stringent research assessment.¹⁰

Because this was the first NIMH-funded study of the phenomenology andlongitudinal course of child mania, conservative inclusion criteria were selectedfor credibility in a highly contentious field.⁹ Therefore,analogous to the requirement of depressed mood and/or anhedonia for DSM-IV major depressive disorder (MDD), the PEA-BP phenotypewas defined by elated mood and/or grandiosity as one inclusion criterion.^10- 14 Thisdefinition avoided diagnosing mania by criteria (hyperactivity and distractibility)that overlapped with those for attention-deficit/hyperactivity disorder (ADHD)and ensured that subjects had at least 1 of the 2 cardinal features of mania(elated mood and grandiosity). This PEA-BP cardinal symptom phenotype is notuncommon, as evidenced by the first 60 subjects (of the same age as in thePEA-BP sample) enrolled in the ongoing NIMH-funded multi-site Treatment ofEarly Age Mania study. In the Treatment of Early Age Mania study, elated moodor grandiosity was not required to diagnose bipolar I disorder (BP-I). Nevertheless,98.3% of the subjects had elation, 96.7% had grandiosity, 95.0% had both,and 100.0% had either (B.G., unpublished data, 2004).

In this regard, Biederman et al¹⁵ reportedthat mania in children is characterized by irritable, rather than elated,mood. These investigators, however, did not interview children younger than12 years, which has recently been shown to be essential in evaluating childmania.¹⁶ In addition, Biederman et al usedlay raters, did not use a scale with prepubertal age-specific items, did nothave a severity cutoff, and ascertained for ADHD. These methodological differencesbetween the Biederman et al and the PEA-BP and Treatment of Early Age Maniasamples may in part account for the differences in prevalence of mania symptoms.Moreover, the nonspecificity of irritability in child psychiatric disordershas also been demonstrated by recent randomized trials for aggression/irritabilityin children with autism¹⁷ and a low IQ¹⁸ and by the report¹⁹ that20% to 60% of young adults with various diagnoses had aggression/irritabilitydiagnoses as children. Thus, irritability is a highly sensitive symptom ofchild and adolescent mania, but is nonspecific because it occurs in many otherdiagnoses.

In earlier reports^11- 13 onthe outcome of the PEA-BP sample, 6-month, 1-year, and 2-year diagnostic stabilityof mania was found. But, there was a need for 4-year data for the followingreasons. Baseline data showed that subjects had been continuously ill withthe intake episode of mania for a mean ± SD of 3.5 ± 2.5 years.To prospectively validate this history of long-episode duration required prospectivedata for at least 4 years. Furthermore, to our knowledge, there are no 4-yearfollow-up data to guide researchers developing intervention and preventionstrategies or to aid physicians answering parental inquiries regarding naturalhistory and prognosis.

Contrary to commonly held beliefs that most adults with BP-I had episodicillnesses, recent weekly examination of mood symptoms during a 20-year periodsupported a complex pattern of varying affective states.²⁰ Prospectivefollow-up of the PEA-BP sample provides the opportunity to compare the weeklycourse in children with that reported for adults.

Study participants were 86 subjects with PEA-BP and were obtained throughconsecutive new case ascertainment from multiple child psychiatric and pediatricsites between September 25, 1995, and December 15, 1998, using methods detailedelsewhere.¹⁰ During this naturalistic study,treatments were provided by participants' own clinicians in the communityand not in any way by the research nurses who conducted the assessments. Therefore,subjects received treatments exactly as if they were not participating ina longitudinal research study.

The Washington University in St Louis Kiddie Schedule for AffectiveDisorders and Schizophrenia (WASH-U-KSADS)^21- 22 isa semistructured interview with excellent reliability for mania symptoms,mood diagnoses, rapid cycling patterns, and time frames (κ, 0.82-1.00),and is the most widely used instrument in NIMH-funded studies of child mania.⁷ It was administered by highly experienced researchnurses, blind at baseline, to mothers about their children and separatelyto children about themselves. Different raters were used for the mother andchild within each family to avoid bias from knowing what the other informanthad reported.¹⁶ The WASH-U-KSADS was developedfrom the KSADS²³ by adding the following: (1)an expanded mania section that included items developed specifically to assessprepubertal mania; (2) a section to assess patterns of ultrarapid and ultradiancycling, calibrated as cycles per day that last 4 hours or more per day; (3)items to assess both lifetime and current episodes; (4) items for specifictiming of onsets and offsets for all symptoms and syndromes, calibrated byweeks; and (5) sections for ADHD and multiple other DSM-IV diagnoses. Skip-outs were minimized to enhance collection of phenomenologydata. Establishing time frames for children's ratings was done by using birthdays,holidays, start of school, end of school, and whether present in earlier grades(eg, if the subject is in fourth grade, was it there in third grade?) as anchorpoints. The WASH-U-KSADS narrative next to each rating is part of this assessmenttool (eg, part of the narrative next to a suicidal ideation item read "cuther wrists four times with a kitchen knife and wanted to die to escape hersad feelings"). The data collection guideline is that the narrative must justifythe rating with respect to onset, offset, frequency, duration, intensity,and specific examples. Severity ratings for items were as follows: 1, no pathology;2, doubtful pathology; 3, mild, with no impairment (eg, a child with ticswho is not teased or ashamed); and 4 or higher, clinically significant pathology(eg, a child with tics who refuses to go to school). Items needed to be rated4 or higher to count toward a diagnosis of mania, MDD, or minor depression,and 3 or higher to count toward a diagnosis of hypomania or dysthymia. Examplesof the phenomenology of mania criteria in children are published.²⁴ To score the WASH-U-KSADS items, mother and childresponses were combined by using either, in accordance with the methods describedby Bird et al.²⁵ Templates to the WASH-U-KSADSto assess DSM-IV substance use disorders (SUDs) inchildhood were also given.^26- 27

The Children's Global Assessment Scale (CGAS)^28- 29 isa global measure of severity based on psychiatric symptoms and adaptive impairmentin family, social, school, and work areas. Ratings were obtained by researchnurses who administered the WASH-U-KSADS. On this scale, 0 is the worst, 100is the best, and 60 or less is definite clinical impairment.²⁵

The Psychosocial Schedule for School Age Children–Revised³⁰ was used to obtain comprehensive measurements ofchild interaction with parents, siblings, peers, and teachers and of maritalrelationships, and was administered to mothers about their children and separatelyto children about themselves by the research nurses who administered the WASH-U-KSADS.The Psychosocial Schedule for School Age Children–Revised has good psychometricproperties.³¹ Psychosocial impairments reportedby either informant (mother or child) were used in the analyses by methodspreviously detailed.³²

Socioeconomic status was established by the Hollingshead Four-FactorIndex of Social Status.³³ The Duke PubertalStatus Questionnaire³⁴ was completed, to obtainTanner stage, by subjects 10 years or older at baseline.

For symptoms that occurred in more than 1 diagnosis, the investigatorsdid not know a way of discerning which diagnosis to count the item toward.Therefore, symptoms that were concurrent to more than 1 diagnosis were countedtoward each, but each diagnosis was only given if there were sufficient symptomsto fit DSM-IV criteria. To avoid overdiagnosing ADHDas BP, due to symptoms that occur in both (eg, hyperactivity, distractibility),children only received mania or hypomania diagnoses if elated mood and/orgrandiosity was one criterion.

Consensus conferences were held after each rating at all time pointsto establish DSM-IV consensus diagnoses. At theseconferences, all assessment instruments, school reports, agency records, andpediatrician records were reviewed. Furthermore, at baseline, videotapes ofthe WASH-U-KSADS interviews from parent informants and, separately, from childinformants were reviewed and discussed.

Because of the paucity of data on hypomania, minor depression, and dysthymiain children, mixed mania was conservatively defined as mania or hypomaniaand MDD. This definition was also used by Solomon et al³⁵ intheir article on long-term follow-up of adult unipolar mania. For heuristicpurposes, data were also analyzed by using the adult definitions.²⁰

Psychosis required a pathologic delusion or hallucination that did notonly occur hypnagogically or hypnopompically, and was assessed with the psychosissection of the WASH-U-KSADS.²¹ In addition,ratings of 6 on the grandiosity, hopelessness, hypochondriasis, and guiltitems signify delusions in these areas.²¹

As recently reported by Tillman and Geller,³⁶ thereis a need to have definitions of cycling and of episodes that are relevantacross the age span. Tillman and Geller have proposed that episodes referto the entire length of illness. In this schema, subjects with 4 episodesper year, who were previously labeled as "rapid cycling," are designated ashaving 4 episodes per year. Cycling refers to mood changes during an episode.Mood changes that occur every few days are ultrarapid cycling, and mood changesthat occur at least once daily are ultradian cycling, using definitions modifiedfrom Kramlinger and Post.³⁷ For example, an8-year-old boy had mania lasting 2 years during which he had daily cycling.Thus, this child had 1 episode of mania with a duration of 2 years characterizedby ultradian cycling (daily cycling for 2 years during 1 episode). Specifically,ultrarapid cycling was defined as 5 to 364 cycles per year, and ultradian(daily) rapid cycling was defined as 365 cycles or more per year.

Research nurses were trained to interrater reliability and were recalibratedannually.²² Raters had virtual 100% agreementon diagnostic categories and symptom severity ratings 5 times in a row asboth interviewer and observer. In addition, the study site has been a majortraining site for investigative groups who use the WASH-U-KSADS in other NIMH-fundedstudies. Training of other investigative teams as of January 9, 2004, included66 research clinicians from 23 sites, who rated as observers during the follow-upstudy to obtain interrater reliability with our group.

The same instruments were used. At baseline, the WASH-U-KSADS and CGASinstruments were obtained for lifetime and current episodes. At follow-upvisits, the WASH-U-KSADS and CGAS instruments were obtained for the periodsince the prior assessment (eg, at the 6-month follow-up, the time frame wasbaseline to 6 months). Similar to baseline, the specific dates of onsets andoffsets of each symptom and the severity of each symptom for each time framewere obtained.

The number of weeks for every occurrence of each symptom (and the severityat each occurrence) were used to calculate the number of weeks for each diagnosis,for the natural history data analyses.

At this time point, we did not analyze a category equivalent to the"subthreshold but some symptoms" reported in samples of adults²⁰ becausenumerous investigators^9,38 reporta high prevalence of ADHD in child BP. Based on this, it seemed problematicto designate subjects with interepisode mania symptoms that were similar tothose for ADHD (eg, hyperactivity, distractibility) as having subthresholdmania. In addition, the difficulty (for developmental reasons) of differentiatingthe nonimpairing happiness and expansiveness of normal children from pathologicallyimpairing manic euphoria and grandiosity has been reported.²⁴

The PEA-BP phenotype was defined as current DSM-IV BP-I (manic or mixed phase) for at least 2 weeks, with elation and/orgrandiosity as 1 criterion.

Exclusion criteria were as follows: adopted, IQ of less than 70, pervasivedevelopmental disorders, schizophrenia, epilepsy or other major medical orneurological disorder, baseline SUDs or pregnancy, or mania only with medicationsthat may produce manic symptoms (eg, antidepressants).

The rationales for these inclusion and exclusion criteria were as follows:the duration criteria for PEA-BP were similar to conservative durations inmultiple nosological schemata and were selected to increase the likelihoodof caseness.⁷ Baseline (index) episodes ofmania were required because this was a phenomenological study, but data werecollected on all lifetime episodes. The rationale for including elation and/orgrandiosity as one criterion is as previously noted. A younger age of 7 yearswas chosen because of the credibility of interview assessments, and an olderage of 16 years was selected so subjects would still be teenagers at the 2-yearfollow-up assessment.¹³ Scores on the CGASwere selected to ensure definite caseness.²⁹ Atbaseline only, SUDs and pregnancy were exclusion criteria to avoid confoundingthe diagnosis of PEA-BP with mental status effects of substance use or gestationalstate, but due to the prepubertal age of the subjects, this did not affectenrollment into the study. Subjects continued in the follow-up phase of thestudy if they developed SUDs or became pregnant after baseline. Adoption wasan exclusion criterion due to concurrent family and genetic studies.^39- 40

Characteristics of the PEA-BP sample are provided in Table 1.

After complete description of the study was provided to parents andchildren, written informed consent was obtained from the parents and writtenassent was obtained from the children.

Analyses included data from the 6-, 12-, 18-, 24-, 36-, and 48-monthfollow-up points.

Definitions of recovery and relapse during follow-up were adapted fromFrank et al.⁴¹ Recovery was defined as 8 consecutiveweeks without meeting DSM-IV criteria for mania orhypomania. Remission was defined as 2 to 7 weeks without meeting DSM-IV criteria for mania or hypomania. Relapse after recovery wasdefined as 2 consecutive weeks of meeting DSM-IV criteriafor mania or hypomania with clinically significant impairment, evidenced bya CGAS score of 60 or less; and partial relapse was similar to full relapse,but lasted only 1 week.

The number of weeks ill with mania, hypomania, major depression, minordepression, dysthymia, or mixed mania was calculated from the severity scoresand from the onset and offset dates for each symptom on the WASH-U-KSADS.Diagnostic categories were constituted from a sufficient number of symptomsto fit DSM-IV diagnoses.

To examine baseline differences between those subjects who did not experiencea recovery from mania during 4 years of follow-up and those subjects who recoveredwithin the first 6 months and remained mania free during follow-up, χ² and t tests of baseline characteristic differencesbetween the 2 groups were performed.

For predictor analyses, a priori predictors were selected based on theliterature.^{1- 7,20,42- 45} Theseincluded age, sex, puberty status, CGAS score, psychosis, mixed status (maniaplus MDD), rapid cycling, and maternal warmth, which is akin to expressedemotion.^42- 44 Detailson the construction of categories for maternal warmth and other categoriesfrom the Psychosocial Schedule for School Age Children–Revised havebeen reported.¹³ Exploratory analyses wereconducted on the following variables: age of onset of first episode, durationof baseline episode, all categories of mixed mania, ascertainment site (pediatricvs psychiatric), comorbid diagnoses (Table1), maternal tension/hostility, paternal warmth and tension/hostility,and living situation.

The cumulative probability of recovery and relapse was estimated usingthe Kaplan-Meier (K-M) method.⁴⁶ During follow-up,durations of BP diagnoses were calculated as the number of weeks between onsetand offset of full DSM-IV syndromal criteria. Recoveryand relapse were modeled using Cox proportional hazards modeling.⁴⁷ Covariates significant in univariate analyses wereselected for K-M analyses, controlling for sex, puberty, and mixed mania.

A mixed model using proportion of weeks with BP diagnoses during the4 years as the outcome measure was constructed with a priori predictors. Covariatessignificant in the model were examined post hoc, controlling for sex, puberty,and mixed mania. Following the methods of Wolfinger⁴⁸ andLittell et al,⁴⁹ various underlying covariancestructures were tested to select the model with the best fit, which was theunstructured covariance model.

Bonferroni correction was used to limit the possibility of type I errors,resulting in a significance level of P<.006 toassess significance for all models. SAS statistical software V8.2 was usedfor all statistical analyses.⁴⁹ Data are givenas mean ± SD unless otherwise indicated.

Of the baseline 93 subjects, 86 were interviewed at all follow-up points.Thus, the retention after 4 years of follow-up was 92.5%. The 7 subjects whodiscontinued the study during the 4 years of follow-up did not differ significantlyfrom those who continued on baseline age, sex, puberty status, age of onsetof index episode, duration of index episode, or CGAS score (data not shown).At baseline, 100% of both parent and child informants were separately interviewed.During the 516 follow-up interviews, only 2.3% (n = 12) were conducted ona single informant. The socioeconomic status was 4.0 ± 0.9, which isthe second highest of 5 classes.³³

The baseline intake episode of mania was also the first episode of maniafor 70 (81.4%) of the 86 subjects. Age of onset for the entire sample (N =86) was 6.9 ± 3.5 years.

Prospective episode duration of manic diagnoses, using onset of maniaas baseline date, was 79.2 ± 66.7 consecutive weeks.

The time to recovery was 60.2 ± 47.5 weeks (Figure 1). One subject was a remitter. Data analyses with or withoutthis subject were similar, so this subject was included in the analyses. TheK-M estimate of the rate of recovery was 87.2% (95% confidence interval, 80.2%-94.3%).The time to relapse after recovery was 40.4 ± 33.4 weeks (Figure 1). The K-M estimate of the rate ofrelapse was 70.2% (95% confidence interval, 57.4%-83.0%). Only 1 subject experienceda partial relapse, so a separate analysis was not conducted for partial relapse.

There were no significant differences in baseline characteristics betweenthose subjects who did not experience a recovery from mania during 4 yearsof follow-up (n = 11) and those subjects who recovered within the first 6months and remained free of mania or hypomania during follow-up (n = 7) (datanot shown). At baseline, 10 subjects had hypomania, but 8 of these developedmania during follow-up and were, therefore, included in the analyses. Dataanalyses with or without these 2 subjects with hypomania were similar, sothese 2 subjects were also included in the analyses.

These data are presented in Table2. Any BP diagnosis occurred during 67.1% ± 28.5% of totalweeks, during the 209.4 ± 3.3 weeks of follow-up. Subjects spent 56.9%± 28.8% of total weeks with mania or hypomania (unipolar or mixed),and 38.7% ± 28.8% of these weeks were with mania. Major or minor depressionand dysthymia (unipolar or mixed) occurred during 47.1% ± 30.4% oftotal weeks.

Polarity switches from mania or hypomania to MDD, dysthymia, or minordepression occurred 1.1 ± 0.7 times per year.

The overall mixed model for proportion of weeks ill with mania or hypomaniawas significant (χ²₇₇ = 1554.8, P<.0001), with a significant interaction of time × baselinepsychosis (F_11,77 = 2.1, P = .028), indicatingthat subjects with baseline psychosis spent more weeks ill with mania or hypomania(Figure 2). A post hoc analysisindicated that a quadratic model best fit the time × baseline psychosiseffect (F_1,80 = 12.2, P = .0008).

Estimates using K-M analyses showed that low maternal warmth predictedearlier relapse to mania or hypomania (Figure3). No other covariates predicted the outcome of BP diagnoses.

Findings from the 4-year prospective follow-up of PEA-BP validate theexistence of child mania. Moreover, these data address the controversy overwhether PEA-BP can be differentiated from ADHD by demonstrating that thesesubjects with PEA-BP had persistent mania and did not become diagnosed withonly ADHD during follow-up. Whether ADHD in children with pediatric BP isa separate diathesis, a reflection of mania plus normal childhood developmentalbehaviors (eg, children are naturally more active than adults), shared geneticvulnerabilities, and/or the result of overlapping criteria for both diagnosesis not yet known.^50- 51

Validation of long durations of episodes of child mania or mixed maniawas provided by the long time to recovery from mania or mixed mania. Becausesubjects were obtained by consecutive new case ascertainment, it is likelythat this picture is representative of BP children who present clinically.This study does not address, however, the question of whether BP childrenwith shorter episodes are not brought to treatment (ie, are there childrenwith briefer episodes tolerated by families for a few weeks until the episodespontaneously resolves?).

Chronicity of PEA-BP was validated by the percentage of weeks duringthe 4 years that subjects spent in BP episodes. The chronicity and severityof the clinical picture in child BP-I warrant comment. Subjects with PEA-BPpresented with long-episode durations, chronicity, and severity (ultradianrapid cycling, psychosis, and mixed mania). These chronic, severe symptomshave been reported in about 20% of adults,^20,52 butwere seen in most PEA-BP subjects (Table1). In addition to longitudinal validation of PEA-BP, data froma parallel ongoing study⁵³ of psychopathologyin first-degree relatives of PEA-BP probands and 2 control groups (subjectswith ADHD and normal controls) evidence high familial aggregation. The clinicalpicture of marked severity, high familial aggregation, and early onset fitsthe paradigm reported by Childs and Scriver⁵⁴ formany medical illnesses. Future studies of genetic and environmental factorswill be needed to elucidate the mechanisms of this association of early onset,loaded family psychopathology, and malignant course.

Low maternal warmth as a predictor of relapse, also found at the 2-yearfollow-up,¹³ is consistent with findings insamples of adults with BP that show the predictive effect of impaired expressedemotion, a concept akin to maternal warmth.^42- 44 Thisrobust finding can be useful for planning intervention and prevention studiesof children with mania and perhaps for those who are high-risk offspring.⁵⁵

Continuity between child- and adult-onset BP is supported by the similarityof mania symptom distribution between PEA-BP and adult-onset cases,^10,14,52 the occurrence ofchild- and adult-onset BP within the same families,⁵³ andthe occurrence of maternal warmth and psychosis as predictors of outcome ofPEA-BP and adult-onset BP.^13,20,42,45

The long time to recovery and the high number of weeks ill during follow-upemphasize the need for early recognition and for the development of preventionand intervention strategies.

Tohen et al⁴⁵ reported recovery ratesof 85.5%, 91.6%, and 97.5% at 6 months, 1 year, and 2 years, respectively,in adult subjects with first-episode mania. By comparison, the recovery ratesat 6 months, 1 year, and 2 years for the PEA-BP sample were 14.0%, 36.0%,and 65.1% (Figure 1).^11- 13 Thesecomparisons support the chronicity of child mania, even in light of methodologicaldifferences in the samples (eg, the subjects in the study by Tohen et al wereinpatients).

Recently, Judd et al²⁰ reported 20-yearfollow-up results of subjects with baseline mood disorders, most of whom hadintake mania. Data were not broken down by 4 years, but during the 2 decades,depressive episodes predominated in the adult sample. In contrast to thesedata from adult mania, there was a higher percentage of weeks with mania orhypomania in the child PEA-BP sample. Speculations on the reasons for thisdifference include developmental trajectories (ie, mania or hypomania maypredominate during the prepubertal and early adolescent age range, and depressivestates may become more prevalent with age). Another possibility is that childand adult BP are discontinuous, but this is unlikely because child- and adult-onsetBP occur within the same families.⁵³ A thirdpossibility is that the difference may be methodological, due to differencesbetween the Judd et al study and the PEA-BP sample on credentials of the ratersand on use of multiple interviewed informants. The Judd et al study used trainedraters compared with the highly experienced research nurses used in the PEA-BPstudy, which raises the possibility that professionally credentialed clinicallyexperienced raters may be more likely to elicit and/or recognize manic symptoms.In this regard, some epidemiological studies⁵⁶ havereported that lay rater–administered interviews identify chronic severemania, but not less severe cases. With respect to multiple informants, inthe Judd et al study, subjects were the only interviewed informants comparedwith 2 interviewed informants (separate research interviews of mothers abouttheir children and of children about themselves) in the PEA-BP investigation.Child mania symptoms are more likely to be identified if multiple informantsare used, as reported for the PEA-BP sample and as reported in studies^16,57- 64 ofchild depression. It is possible that dual informants for adult BP subjectsmay have elicited more symptoms of mania.

Lack of mixed/cycling as a predictor of child outcome is different fromfindings in adult BP,^3,20,45 likelydue to the higher prevalence of cycling in the PEA-BP sample.

The fewer polarity switches per year in the PEA-BP study compared withthe Judd et al²⁰ study are consistent withthe long duration of mania or hypomania episodes in the child BP sample.

During the time this sample was obtained through consecutive new caseascertainment (1995-1998), there was no facility or private practice in StLouis available to obtain consecutive subjects from lower socioeconomic statusbackgrounds. Therefore, these findings may not generalize to lower socioeconomicstatus settings.

In addition, due to the paucity of inpatient facilities in St Louisduring this study, there were no inpatients in the PEA-BP sample. This isdissimilar to studies^45,65 ofBP adults, in which subjects with mania are often hospitalized. By contrast,many severely ill prepubertal and early adolescent children, including thosewith psychosis, are not inpatients. This is evidenced by the entirely outpatientstatus of the NIMH-funded PEA-BP and Treatment of Early Age Mania samples,in which pathologic psychosis occurred in 59.3% (Table 1) and 83.3%, respectively.

Since this study was the first phenomenology and longitudinal studyof PEA-BP, to our knowledge, we elected to use a conservative phenotype thatrequired DSM-IV BP (manic or mixed phase) with elationand/or grandiosity as one inclusion criterion to ensure differentiation fromADHD.⁶⁶ Thus, these findings may not generalizeto other phenotypes.⁷

Several of the predictors of outcome in studies^1,20,67- 68 ofadults with BP were not examined at the 4-year point because they were notpresent at baseline (SUD and panic disorder). This is likely due to the youngage of the subjects, because panic disorder and SUD usually begin at a laterage.^67- 68 Future follow-up ofthe PEA-BP sample will include SUDs and panic disorders as covariates, sincethese conditions are likely to occur as the PEA-BP sample ages.⁶⁸ Separatepublications will report the predictive value of community physician–administeredtreatments. The direct interview family study of these PEA-BP probands and2 control groups is in the data collection phase and, thus, data are not availablefor use as predictors of outcome or to determine the relationship of maternalwarmth to maternal psychopathology. Later data analyses will examine theseissues. More important, later follow-up of the PEA-BP sample will providedata on whether the predominantly manic or hypomanic, long-episode duration,severe, ultradian cycling picture present during childhood continues intolater adolescence and adulthood.

Corresponding author: Barbara Geller, MD, Department of Psychiatry,Washington University in St Louis, 660 S Euclid Ave, St Louis, MO 63110 (e-mail: gellerb@medicine.wustl.edu)

Submitted for publication June 27, 2003; final revision received December2, 2003; accepted December 16, 2003.

This study was supported by grant R01 MH-53063 from the NIMH, Rockville,Md.

We thank Betsy Zimerman, BSN, MA, and Marlene Williams, RN, for contributingto administration and data collection for this study; and Jeanne Frazier,BSN, and Linda Beringer, RN, for contributing to data collection for thisstudy.