Anomalous Prefrontal-Subcortical Activation in Familial Pediatric BipolarDisorder: A Functional Magnetic Resonance Imaging Investigation

Background  The neurobiological features of pediatric bipolar disorder (BD) are largely unknown. Children and adolescents with BD may be important to study with functional neuroimaging techniques because of their unique status of early-onset BD and high familial loading for the disorder. Neuroimaging studies of adults with BD have implicated the dorsolateral prefrontal cortex (DLPFC) and anterior cingulate cortex (ACC) in the development of this disorder.

Objectives  To study children and adolescents with BD via functional magnetic resonance imaging using cognitive and affective tasks and to examine possible abnormalities in the DLPFC and ACC, as well as selected subcortical areas, in pediatric familial BD.

Design  We evaluated 12 male subjects aged 9 to 18 years with BD who had at least 1 parent with BD as well as 10 age- and IQ-matched healthy male controls. Stimulants were discontinued for at least 24 hours; other medications were continued. Subjects underwent functional magnetic resonance imaging at 3 T while performing a 2-back visuospatial working memory task and an affective task involving the visualization of positively, neutrally, or negatively valenced pictures.

Setting  An academic referral setting, drawing from the Bay Area of San Francisco, Calif.

Results  Compared with controls, for the visuospatial working memory task, subjects with BD had greater activation in several areas including the bilateral ACC, left putamen, left thalamus, left DLPFC, and right inferior frontal gyrus. Controls had greater activation in the cerebellar vermis. In viewing negatively valenced pictures, subjects with BD had greater activation in the bilateral DLPFC, inferior frontal gyrus, and right insula. Controls had greater activation in the right posterior cingulate gyrus. For positively valenced pictures, subjects with BD had greater activation in the bilateral caudate and thalamus, left middle/superior frontal gyrus, and left ACC, whereas controls had no areas of greater activation.

Conclusions  Children and adolescents with BD may have underlying abnormalities in the regulation of prefrontal-subcortical circuits. Further functional magnetic resonance imaging studies of attention and mood with greater sample sizes are needed.

Pediatric bipolar disorder (BD) carries high morbidity and psychosocialdysfunction because of its chronic and frequently rapid-cycling symptoms,high comorbidity with disruptive behavioral disorders, and relative treatmentresistance.^1,2 However, littleis known about the neuropathophysiologic features of pediatric BD. Neuroimagingstudies of children and adolescents with BD may be of particular interestto pursue because these patients often have not had as many years of substanceuse or medication exposure, which may confound similar studies in adults.

Neuroimaging studies of adults with BD and a few in children with BDhave supported the involvement of prefrontal brain regions in this disorder.Positron emission tomographic studies have found that adults in manic states,compared with depressed states may have increased overall brain activity,³ particularly in the inferior frontal areas.⁴ Increased activity in the anterior cingulate cortex(ACC) has also been reported in bipolar manic states vs euthymic states.⁵ Compared with healthy controls, adults with BD reportedlyhave hypometabolism in the dorsolateral prefrontal cortex (DLPFC) accordingto fluorodeoxyglucose F 18–positron emission tomographic studies.⁶ A functional magnetic resonance imaging (fMRI) studyalso reported increased left amygdalar and decreased right DLPFC activationin adults with BD viewing fearful faces.⁷ Spectroscopicstudies have reported decreased DLPFC N-acetylaspartatelevels, a marker of neuronal density, in adults⁸ andchildren⁹ with BD. Additionally, children withBD during a manic episode were reported to have increased myo-inositol levels in the ACC.¹⁰ Inlight of these findings, it is likely that these prefrontal areas are involvedin BD.

A hypothesis implicating dysfunction of the DLPFC and ACC in BD appearsappropriate because both regions are involved in normal mood regulation, assupported by studies of healthy volunteers. Increased activity in the rightACC, bilateral frontal and prefrontal cortices,¹¹ andDLPFC¹² has been observed during transientinduced sadness in healthy volunteers. Other investigators have found reductionsin blood flow of the right dorsal and ventral prefrontal lobes and dorsalACC during more sustained sadness inductions in healthy volunteers.¹³

The DLPFC and ACC also have crucial roles in attention processing, relevantwhen considering that 60% to 94% of children with BD have comorbid attention-deficit/hyperactivitydisorder (ADHD).¹⁴ The DLPFC is activated duringthe implementation of control in cognition, necessary in color-naming Strooptasks¹⁵ and spatial working memory.¹⁶ Abnormalities in the DLPFC, as reflected by decreasedlevels of N-acetylaspartate, have been found in adultswith ADHD.¹⁷ The ACC has been similarly implicatedin the control of attention,^18,19 specificallyin error recognition and overriding a prepotent response bias.²⁰ Thus,Stroop tasks have caused activation in the ACC in healthy subjects^15,21 and lesser activation in subjectswith ADHD.²²

Because of these findings, the prefrontal cortex, including the DLPFCand ACC, is postulated to contain cortical control areas that regulate bothmood and attention. Accordingly, these areas are prime candidates for investigationin childhood BD. We tested the hypothesis that children with BD would showanomalous prefrontal activation compared with healthy controls by using fMRIexperiments that tap brain function related to both attention and emotion.These experiments consisted of a cognitive task involving visuospatial workingmemory and an affective task involving the viewing of emotionally valencedpictures from the International Affective Picture System (IAPS).²³ Becauseof research suggesting sex differences in emotional reactivity in children²⁴ and because of the higher incidence of pediatricBD in boys,²⁵ we limited this initial studyto males only. Furthermore, because we were interested in the involvementof prefrontal-subcortical circuits, we conducted whole-brain analyses of thefMRI data.

Subject families were recruited from the Stanford Adult and PediatricBipolar Disorders Clinics (Stanford, Calif) and from the surrounding community.Written and oral informed consent were obtained from at least 1 parent, andassent was obtained from the subject after explaning possible adverse effectsand alternatives to study participation. The study met all requirements ofthe institutional review board at Stanford University.

Inclusion criteria for bipolar subjects were age between 9 and 18 years,at least 1 biological parent with bipolar I or II disorder, and diagnosisof bipolar I disorder. Exclusion criteria were the presence of a pervasivedevelopmental disorder, a neurological condition (such as a seizure disorder),a substance use disorder, an IQ less than 80, or the presence of metallicimplants or braces.

Parents were diagnosed using the Structured Clinical Interview for DSM-IV Axis I Disorders–Patient Edition (SCID-I/P).²⁶ Family history was obtained using the Family History–ResearchDiagnostic Criteria.²⁷ Children were assessedwith the affective module of the Washington University in St Louis KiddieSchedule for Affective Disorders and Schizophrenia^28,29 andthe Schedule for Affective Disorders and Schizophrenia for School-Age Children–Presentand Lifetime Version.³⁰ Subjects were evaluatedeither by a child psychiatrist (K.C.) or trained masters-level research assistants(K.D. or D.I.S.) who were aware of the parental diagnosis. Current and lifetime DSM-IV diagnoses were ultimately made by a board-certifiedchild psychiatrist (K.C.) based on personal interview, discussion with theresearch assistant, and written notes of interview responses.

Healthy controls did not have a DSM-IV diagnosis,were not taking psychotropic medications, had both parents without any psychiatricdiagnosis according to the SCID-I/P, and did not have a first- or second-degreerelative with BD as determined using the Family History–Research DiagnosticCriteria..

Subjects were all outpatients at the time of scanning. Subjects withBD were administered the Young Mania Rating Scale ^31,32 andcompleted the Childhood Depression Inventory, ³³ withthe help of a parent if they were younger than 12 years, within 3 days offMRI. Stimulants were discontinued for at least 24 hours prior to imaging;other medications were continued. The IQ was assessed with the Wechsler AbbreviatedScale of Intelligence.³⁴

The pool of subjects with BD was the same for both tasks. However, thosewho had movement greater than 3 mm (translation) or greater than 3° (rotation)during imaging were disqualified from further analysis owing to spatial datainaccuracy. Therefore, 11 subjects with BD were analyzed for the visuospatialworking memory task (mean ± SD age, 15.3 ± 2.5 years; range,9.7-18.6 years), and 11 were analyzed for the affective task (mean ±SD age, 14.5 ± 3.0 years; range, 9.2-18.6 years). Ten subjects wereincluded in both groups. Ten healthy controls (mean ± SD age, 14.4± 3.2 years; range, 10.0-17.7 years) completed both the visuospatialworking memory and affective tasks.

The visuospatial working memory task consisted of 6 alternating experimentaland control epochs (Figure 1). Eachexperimental and control epoch consisted of 16 stimuli presented for 500 millisecondseach, with a 1500-millisecond interstimulus interval. The stimulus was theletter O presented in 1 of 9 spatial locations ina 3 × 3 matrix. In the experimental epoch, subjects were instructedto press a button if the stimulus was in the same location as it was 2 trialspreviously. In the control epoch, subjects were instructed to respond if thestimulus was in the center position. Correct response rate, incorrect responserate, and reaction times were recorded. Further details of the task have beendescribed elsewhere.^35,36

The IAPS²³ is a stimulus set that hasbeen used in other functional imaging studies of affective stimulation.^37,38 Specific negative (eg, a mutilateddog) and positive (eg, a hot fudge sundae) picture stimuli were selected thatwere deemed acceptable to a pediatric population. Neutral (eg, a plate) pictureswere selected for the control condition. Valence was determined using previouslypublished ratings of the specific pictures.²³ The4 types of stimuli were organized into blocks, each with 6 stimuli, with eachstimulus presented for 4500 milliseconds with a 500-millisecond interstimulusinterval (Figure 2). Subjects wereasked to indicate how each picture made them feel by pressing 1 of 3 buttonscorresponding to negatively, neutrally, and positively.

The tasks were programmed using Psyscope software (http://psyscope.psy.cmu.edu) on an Apple G3 notebook computer (Cupertino, Calif). Stimuli wereprojected onto a screen using a custom-built magnet-compatible projectionsystem (Sanyo, San Diego, Calif). A custom-built button box was used to measurebehavioral responses.

Images were acquired with a 3-T GE Signa scanner using a standard whole-headcoil (General Electric, Milwaukee, Wis). The following spiral pulse sequencepmeters were used: time to repeat, 2000 milliseconds; echo time, 30 milliseconds;flip angle, 80°; and 1 interleave. To reduce field inhomogeneities, anautomated high-order shimming method based on spiral acquisitions was usedbefore acquiring fMRI data.³⁹ To aid in localizationof the functional data, we used high-resolution, T1-weighted, spoiled gradient-recalledacquisition in the steady state (GRASS) 3-dimensional magnetic resonance imagingsequences with the following pmeters: time to repeat, 35 milliseconds;echo time, 6 milliseconds; flip angle, 45°; field of view, 24 cm; 124slices in the coronal plane; and a 256 × 192 matrix.

Images were reconstructed for each time point using inverse Fouriertransform. The fMRI data were preprocessed using SPM99 software (http://www.fil.ion.ucl.ac.uk/spm). Images were corrected for movement using least squares minimizationwithout higher-order corrections for spin history and were normalized to MontrealNeurological Institute (Montreal, Quebec) coordinates.⁴⁰ Imageswere then resampled every 2 mm using sinc interpolation and smoothed witha 4-mm gaussian kernel to decrease spatial noise. The Montreal NeurologicalInstitute coordinates were transformed into stereotactic Talairach coordinates⁴¹ using nonlinear transformation.⁴²

Statistical analysis was performed for individual and group data usingthe general linear model and the theory of gaussian random fields as implementedin the SPM99 program.⁴⁰ Activation foci weresuperimposed on high-resolution T1-weighted images, and their locations wereinterpreted using the Talairach atlas⁴¹ andknown neuroanatomical landmarks.⁴³

A within-subjects procedure was used to model all effects of interestfor each subject. Individual subject models were identical across subjects(ie, a balanced design was used). Confounding effects of fluctuations in theglobal mean were removed using proportional scaling with the global mean ateach time point. Low-frequency noise was removed with a high-pass filter (0.5Hz) applied to the fMRI time series at each voxel. Group analysis was performedusing a random-effects model that incorporated a 2-stage hierarchical procedure.This model estimates the error variance for each condition of interest acrosssubjects rather than across images⁴⁴ and thereforeprovides a stronger generalization to the population studied. Individual contrastimages were computed for experimental minus control conditions in the visuospatialworking memory task and for negative minus neutral and positive minus neutralconditions in the affective task. These contrast images were analyzed usinga general linear model to determine voxelwise t statistics.Appropriate t tests were then used to determine groupactivation and between-group differences for each contrast of interest. Finally,the t statistics were normalized to z scores, and significant clusters of activation were determined usingthe joint expected probability distribution of height and extent of z scores, with height (z>1.67; P<.05) and extent thresholds (P<.05).⁴⁵

Overall, the pool of subjects with BD (all males) had a mean ±SD age of 14.7 ± 3.0 years, whereas controls had a mean ± SDage of 14.4 ± 3.2 years (Table 1). The mean socioeconomic status, as determined with the Hollingshead2-factor method,⁴⁶ was 3.9 for subjects withBD and 4.7 for controls. Subjects with BD did not significantly differ fromcontrols in age, sex, IQ, handedness, or socioeconomic status (Table 1). Of the parents with BD, 58.3% had bipolar I disorder,41.7% had bipolar II disorder, and 83% were women.

For the subjects with BD, mean duration of illness was 3.1 years. Ofthe patients, 92% had at least 1 comorbid psychiatric diagnosis; 92% had ADHD,58% had oppositional defiant disorder, and 33% had an anxiety disorder. Twosubjects (16.7%) had experienced psychotic symptoms in the past. One subject(8.3%) was not taking medication at the time of fMRI. The mean ± SDnumber of medications at the time of imaging was 4.6 ± 2.0 (Table 1). The mean ± SD Young ManiaRating Scale score was 11.8 ± 7.8, and the mean ± SD ChildhoodDepression Inventory score was 14.1 ± 8.2.

Subjects with BD were slightly less accurate on the visuospatial workingmemory task than controls, although this difference did not reach statisticalsignificance (86% vs 93% correct; P = .08). Reactiontimes were not significantly different between subjects with BD and controls(mean ± SD, 628 ± 138 milliseconds vs 534 ± 141 milliseconds,respectively; P = .12).

For the 2-back task minus control condition contrast, within-group analysesshowed that subjects with BD activated the bilateral DLPFC among other prefrontalareas as well as the left caudate, left inferior parietal lobule, right precuneus,and right thalamus (Table 2).Controls activated the right DLPFC and other prefrontal areas, the right precuneus,and the right superior parietal lobule. Subjects with BD had significantlygreater (P<.05) activation than controls in thefollowing regions: the bilateral anterior cingulate, left putamen, left thalamus,left DLPFC, left middle frontal gyrus, left superior frontal gyrus, left superiortemporal gyrus, and right inferior frontal gyrus (Figure 3). Within the left superior temporal gyrus, greater leftinsular activation was also seen in subjects with BD (Table 2). Controls showed greater activation than subjects withBD in areas within the cerebellum, predominantly the vermis (Figure 3).

Each individual's ratings were averaged across pictures of the samevalence, as classified by the IAPS,²³ to givea subject's mean rating for each valence of the pictures. Across both groups,there was a significant effect (Hunyh-Feldt statistic; P<.001) of valence, indicating significant differences between subjects'ratings for differently valenced IAPS pictures. Follow-up paired t tests revealed that subjects with BD had significantly differentratings for positively and neutrally valenced pictures (P = .001) and for negatively and neutrally valenced pictures (P = .003). Within the control group, ratings for both positivelyvs neutrally valenced pictures and neutrally vs negatively valenced pictureswere significantly different (P<.001). There wasno interaction effect between subjects' ratings of valenced pictures and diagnosis(Hunyh-Feldt statistic; P = .12).

Subjects with BD who were exposed to negative visual stimuli activatedthe bilateral DLPFC, left inferior frontal gyrus, and inferior/middle temporalgyrus, among other areas (Table 3).Control group activation in response to negative stimuli included the bilateralDLPFC, left ACC, and inferior temporal gyrus. Compared with healthy controls,subjects with BD showed significantly greater activation in the bilateralDLPFC, left superior/middle temporal gyrus, left inferior frontal gyrus, andright insula (Figure 4). Controlsshowed greater activation than subjects with BD in response to negative stimuliin the right posterior cingulate gyrus (Figure4).

In response to positive stimuli, subjects with BD activated the bilateralmiddle occipital gyrus, left medial frontal gyrus, left ACC, and right cerebellum(Table 4). Controls activatedthe right cuneus and middle occipital gyrus. Subjects with BD showed significantlymore activation than controls in response to positive stimuli in the bilateralcaudate and thalamus and left middle/superior frontal gyrus, ACC, precentralgyrus, pcentral lobule, and precuneus (Figure 5). Controls did not show greater activation than subjectswith BD in any region when viewing positive stimuli.

Consistent with our hypothesis, children and adolescents with BD demonstratesignificant differences in brain activation patterns in prefrontal areas comparedwith controls when performing both cognitive and affective tasks. The differenceswe detected were mostly increases in cerebral activation in subjects withBD, regardless of task. Brain areas differing in activation patterns includedthe DLPFC and ACC as well as other prefrontal areas and extended to the limbicstructures (insula), striatum (caudate and putamen), and thalamus. These areashave all been implicated in the pathophysiologic mechanisms of BD.⁴⁷

There are several possible explanations for why we detected overallincreased task-related brain activation in subjects with BD. First, pediatricBD may be associated with a hyperreactive brain state, particularly in responseto affective stimulation or any performance demand, even during euthymic periods.Positron emission tomographic studies of patients with mania have shown increasedcerebral blood flow at rest.⁴ Because our subjectswere euthymic, overactivation observed in fMRI experiments may be a traitmarker of this disorder. However, it is also possible that these findingsreflect a developmental stage of BD so that activation patterns begin to decrease,even to lower than normal, after years of sustained illness. For example,it has been shown that the ACC is activated by transient induced sadness¹¹ but deactivated in response to more sustained sadness¹³ in healthy volunteers and patients with depression.⁴⁸ Therefore, with extended duration of an emotionalstate or illness, overall activation patterns may progress from overactivationto underactivation in BD. Our findings should be gauged with these possibledevelopmental considerations in mind.

Visuospatial working memory tasks have been reported to activate theright DLPFC in healthy adults^16,49-51 andchildren,^16,52 and studies usingthe IAPS in healthy adults have also demonstrated DLPFC activation.³⁷ In our study, DLPFC activation was greater for subjectswith BD than for controls, on the left (Brodmann area [BA] 9) in the visuospatialworking memory task and bilaterally (BA 9 and BA 45) in the negative-stimulicondition of the IAPS task.

Previous studies in adults and children support the involvement of theDLPFC in the neuropathophysiologic underpinnings of BD. In an fMRI study,adults with BD watching fearful faces had less activation in the right DLPFCthan healthy controls.⁷ Neuronal and glialDLPFC density may be reduced in adults with BD.^8,53 Wepreviously found decreased N-acetylaspartate levels,signifying decreased neuronal density, in the right DLPFC in pediatric familialBD,⁹ albeit to a lesser extent than was foundin adults with BD.⁸ Therefore, the abnormalitiesin DLPFC activation reported in this article may be related to underlyingDLPFC abnormalities in neuronal density or function.

As hypothesized, we also found differences in ACC activation betweensubjects with BD and controls. In our visuospatial working memory task, subjectswith BD had greater activation in the right (BA 24 and BA 32) and left (BA32) ACC than controls. In the positive-stimuli condition of the IAPS task,subjects with BD demonstrated increased activation in the left ACC (BA 24).Abnormalities in the ACC have previously been reported in children and adultswith BD, including increased ACC blood flow during rest⁵ andwhile performing a decision-making task⁵⁴ andincreased ACC myo-inositol.¹⁰ Ourfindings further support the existence of abnormalities in ACC function inpediatric BD.

Researchers have suggested a functional division of the ACC, with caudalportions associated with cognitive functions and ventral portions respondingto emotional stimuli.^19,21,22,55 Inour study, most of the ACC overactivation in subjects with BD was in the ventralportions, but we did not find differences in activation of the subgenual ACC(a portion of BA 24). Abnormalities of the subgenual cingulate have been reportedin familial BD^56,57 and in unipolardepression in adults⁴⁸ and children.⁵⁸ The IAPS task might have been expected to elicitfunctional differences in this area owing to its affective component; however,it is possible that the task was not sufficient to probe for subgenual ACCactivation or that our subjects simply did not have functional abnormalitiesin this region.

Other prefrontal structures were activated to a greater extent in subjectswith BD, most notably the orbitofrontal cortex (OFC). In the visuospatialworking memory task, subjects with BD had greater activation in the rightinferior OFC (BA 11). Subjects with BD also had greater activation in theleft inferior OFC (BA 47) during the negative-stimuli condition of the IAPStask and greater activation in the left medial OFC (BA 10) during the positive-stimulicondition. The OFC has reciprocal connections with limbic structures, includingthe insula, amygdala, and subgenual cingulate, and OFC lesions may resultin behavioral disinhibition and emotional lability.⁵⁹ Ina positron emission tomographic study, decreased orbitofrontal blood flowwas noted in adults with BD and mania compared with euthymia, both duringrest and during a word generation task.⁶⁰ Ourfinding of increased orbitofrontal activity during the visuospatial workingmemory and IAPS tasks could represent compensatory overactivation to modulateoveractive limbic areas in our subjects with BD.

In subjects with BD, we found increased left thalamic activation duringthe visuospatial working memory task and increased bilateral thalamic activationduring the positive-stimuli condition of the IAPS task. The thalamus, whichhas multiple functions, also has significant connections to the prefrontalcortex and may be a crucial component of limbic circuits, including the DLPFCand OFC circuits. Thalamic abnormalities have been reported in BD, includingincreased^61,62 and decreased⁶³ thalamic volume or density and increased thalamic N-acetylaspartate levels.⁶⁴

Subjects with BD also had greater activation of the bilateral caudateduring the positive-stimuli condition of the IAPS task. Increased caudatevolumes have been reported in men with BD⁶⁵ andin monozygotic twins discordant for BD.⁶⁶ Ina positron emission tomographic study, adults with mania had increased bloodflow in the left caudate while at rest.⁵ Ourfindings further support these previous suggestions of striatal abnormalitiesin BD.

Increased activation of the left insula (BA 21 and BA 22) during thevisuospatial working memory task and the right insula (BA 21) during the negative-stimulicondition of the IAPS task were seen in subjects with BD. Left insular activationhas been noted in positron emission tomographic studies of transient inducedsadness,¹² whereas left insular hypermetabolismin adults with BD may predict the response to carbamazepine.⁶⁷ Therole of the insula in autonomic arousal suggests that future studies couldindirectly assess insular overactivation via psychophysiological measures.

It is notable that we did not find differences in amygdalar activationin either between-groups or within-groups comparisons. Activation of thismesial temporal structure may occur by using strong emotional stimuli.^38,68,69 Increased amygdalaractivation has been reported in adults with BD performing affect related tasks.⁷ However, it is unclear if children activate the amygdalato the same extent in these tasks. Also, amygdalar dysfunction could occurlater in the course of BD, only after sustained disrupted prefrontal modulationof amygdalar input. Alternately, the IAPS task may not be as suited to elicitamygdalar activation as, for example, a task involving facial expressionsof fear or disgust.⁷⁰ Finally, the amygdalamay reportedly habituate after repeated affective stimuli.⁷¹

The only areas in which controls showed greater activation were thecerebellar vermis, in the visuospatial working memory task, and the posteriorcingulate, when viewing negative IAPS pictures. The vermis is a neocerebellarstructure that has multiple higher cognitive functions, including executivefunction and working memory.⁷² It was alsofound to be relatively atrophied in adults with familial BD⁷³ andto have decreased N-acetylaspartate levels in offspringof parents with BD.⁷⁴ Therefore, the cerebellarvermis may represent an area in which patients with BD do not (or are notable to) preferentially activate compared with healthy individuals when performinga visuospatial working memory task. Possible reasons for decreased posteriorcingulate activation in subjects with BD are less clear. This region receivesmajor input from the DLPFC and OFC and may promote the evaluative functionof emotional memory. A relationship has also been found between the rightretrosplenial cortex and unpleasant pictoral stimuli in healthy subjects,⁷⁵ so patients with pediatric BD may not have this associationto the same degree.

Of our subjects with BD, 92% also met the criteria for ADHD. Owing tothe additional role of many of the discussed brain structures in the regulationof attention, it could be argued that increased activation of these areasreflects the underlying pathophysiologic mechanisms of ADHD rather than BD.It is difficult to septe the contributions of these 2 disorders to ourfindings. However, given the high comorbidity of ADHD in pediatric BD,¹⁴ it is likely that pediatric BD is a single underlyingdisorder that adversely affects both mood and attention regulation and thatour fMRI data reflect the underlying disorder of pediatric BD as a whole.

A model of brain circuitry dysfunction in BD consistent with our resultscenters around a prefrontal-subcortical theory of mood regulation.^59,76,77 Subcortical structuressuch as the amygdala and hippocampus have long been thought to interact withcortical areas (eg, the cingulate, OFC, and insula) to create and processemotions.^78-80 Whereassome subcortical structures bypass higher cortical input in circumstancesrequiring a quick reaction, they are also significantly interconnected withthe prefrontal cortex, striatum, and thalamus.⁵⁹ Thus,prefrontal areas such as the DLPFC, ACC, and OFC have been postulated to reciprocallymodulate limbic areas to exert cognitive control of affective responses.⁵⁹ Disruptions in the normal balance of activity inthese 2 broad areas (ventral-limbic and dorsal-cortical) may lead to the disruptionof mood regulation. For example, adults with major depressive disorder mayhave increased ventral activity and decreased dorsal activity during depressedstates,⁸¹ a finding that reverses during remission.^48,81

Although it is not possible to discern temporal sequences of activationwithin the tasks in our blocked design, increased prefrontal activation inour subjects with BD during both cognitive and affective tasks may be in responseto increased activation in the ventral-limbic areas. Children with BD mayrequire increased activation of prefrontal areas during euthymic periods tooppose or cortically control a hyperactive limbic system (Figure 6). Limbic areas may be overactivated by the emotional demandsof a difficult task (visuospatial working memory task) or by direct affectivestimuli (IAPS task). However, our subjects with BD did not display consistentlimbic overactivity across both tasks, perhaps because of their euthymic stateor the success of prefrontal structures in suppressing limbic hyperreactivity.Additionally, patients with BD may have relative deficiencies in more efficientprefrontal-limbic circuits, necessitating compensatory activation of otherprefrontal areas.

According to this theory, patients with acute mania might demonstratesimilar limbic overactivity but without sufficient prefrontal activity toremain euthymic. This manifestation would be consistent with findings of decreasedprefrontal activity and neuronal and glial density in adults with BD; thatis, with a longer duration of illness, prefrontal deficits may enlarge andeventually lead to reduced or less effective activation instead of overactivationin response to relevant tasks. A patient with BD in this latter state wouldtheoretically be more vulnerable to stressors triggering mood episodes withoutsufficient prefrontal input to modulate increased subcortical activity, atheory in line with the kindling hypothesis of illness progression in BD.⁸² Further study of patients with BD across differentmood and developmental states is needed to test these hypotheses.

All subjects with BD in our study were taking concomitant psychotropicmedications, of which the effects on cerebral activation patterns are unknown.Stimulants, however, were withheld for 24 to 48 hours before fMRI to minimizetheir behavioral and cerebral effects. There have been reports of increasedcaudate activity⁸³ and decreased ACC activity⁸⁴ associated with antipsychotic medications as wellas decreased ventral ACC activity associated with antidepressants.^81,85 Although between-group caudate differencesmay have been related to this phenomenon, it is unlikely to explain the findingsof increased ACC activity in subjects with BD. Unfortunately we were not ableto divide the subjects into groups based on medication types that would bemeaningful for statistical analyses owing to the wide variety of medications(Table 1).

Because there was a trend for subjects with BD having less accuracyin the 2-back task, it is possible that these subjects were using more effortto perform the task. Additional effort could result in patterns of increasedactivation. Although the sample size was small, the successful completionof an fMRI study in 12 children with a serious neuropsychiatric disorder suchas BD is an accomplishment given pragmatic issues such as subject compliance.All of our subjects were male, primarily white, so these findings may notbe generalizable to girls or boys of other ethnicities who have BD. Finally,we did not formally assess pubertal status, which may have affected the analysesconsidering the substantial hormonal and potential neurobiological changesassociated with puberty.

The strengths of the study included a relatively homogeneous subjectgroup, with both subjects and parents diagnosed using semistructured interviews.Our controls were also screened with high scrutiny; other studies often havenot considered extended family history. This is essential in studies of childrenbecause they retain the potential to develop affective disorders; historicallythe most common age at onset of BD has ranged from 15 to 19 years.⁸⁶ Furthermore, this study was performed using a high-fieldmagnet (3 T), which provides a significantly higher signal-to-noise ratiocompared with 1.5-T studies, increasing the specificity of the findings anddecreasing type II error.

Finally, our study design is also unique because we used a dual approachof cognitive and affective tasks to probe the brain regions relevant to pediatricBD. Temporal analysis of results from future event-related tasks may helpto prove or disprove theories of disrupted prefrontal-subcortical reciprocalmodulation in BD, as suggested by our results. Longitudinal fMRI studies ofthe brain, with greater numbers of subjects and of patients with early formsof BD, would aid in discerning the role of these brain areas in the pathophysiologicmechanisms of BD and in bridging the gap between studies of children and studiesof adults with this disorder.

Correspondence: Kiki Chang, MD, Stanford University School of Medicine,Division of Child and Adolescent Psychiatry, 401 Quarry Rd, Stanford, CA 94305-5540(kchang88@stanford.edu).

Submitted for publication November 12, 2003; final revision receivedJanuary 30, 2004; accepted February 16, 2004.

This study was supported in part by grants MH01142, MH19908, MH050047,and HD31715 (Dr Reiss) and grant MH64460-01 (Dr Chang) from the National Institutesof Health, Bethesda, Md; a Young Investigators Award from the National Alliancefor Research on Schizophrenia and Depression, Great Neck, NY; and a fellowshipfrom the Klingenstein Third Generation Foundation, New York, NY.

This study was presented in part at the 49th Annual Meeting of the AmericanAcademy of Child and Adolescent Psychiatry; October 22-27, 2002; San Francisco,Calif.

We thank Christine Blasey, PhD, for statistical assistance.