Original Article

N-acetylaspartate levels in bipolar offspringwith and at high-risk for bipolar disorder

The advancement of neuroimaging technology hasallowed for the increased study of specific brainregions and neuronal substrates involved in thepathophysiology of bipolar disorder (BD). Protonmagnetic resonance spectroscopy (1H-MRS) is anon-invasive procedure using magnetic resonancetechnology to determine levels of specific neuronalsubstrates such as N-acetylaspartate (NAA), cho-line (Cho), myoinositol (mI), and creatine + phos-

phocreatine (Cr) (1). Relatively low NAA/Cr ratioshave been suggested to indicate decreased neuronalintegrity or functioning (2) and may be a possiblerisk marker or indicator of bipolar illness progres-sion. For example, decreased NAA levels werefound in the cerebellar vermis in children withmood disorders (MDs) and at familial risk for BD(3).

1H-MRS studies of adults and children alreadywith fully developed BD have found decreasedNAA levels in dorsolateral prefrontal cortex(DLPFC) (4, 5). The DLPFC mediates mood andattention regulation (6, 7) and therefore is a logicalbrain region to study for abnormalities in patients

Gallelli KA,WagnerCM,KarchemskiyA,HoweM, SpielmanD,ReissA,Chang KD. N-acetylaspartate levels in bipolar offspring with and athigh-risk for bipolar disorder.Bipolar Disord 2005: 7: 589–597. ª Blackwell Munksgaard, 2005

Objectives: Studies have reported decreased N-acetylaspartate (NAA)in dorsolateral prefrontal cortex (DLPFC) of adults and children withbipolar disorder (BD), suggesting decreased neuronal density in thisarea. However, it is unclear if this finding represents neurodegenerationafter or a trait marker present before BD onset. To address this question,we used proton magnetic resonance spectroscopy (1H-MRS) to compareDLPFC levels of NAA among bipolar offspring with early-onset BD,bipolar offspring with subsyndromal symptoms of BD and healthychildren.

Methods: Participants were 9–18 years old, and included 60 offspringof parents with bipolar I or II disorder (32 with BD and 28 withsubsyndromal symptoms of BD), and 26 healthy controls. 1H-MRS at3 T was used to study 8-cm3 voxels placed in left and right DLPFC.

Results: There were no significant group differences in mean right orleft DLPFC NAA/Cr ratios. Exploratory analyses of additionalmetabolites (myoinositol, choline) also yielded no significant groupdifferences. NAA/Cr ratios were not correlated with age, duration ofillness, or exposure to lithium or valproate.

Conclusions: Our findings suggest that DLPFC NAA/Cr ratios cannotbe used as a trait marker for BD. Although we did not find decreasedDLPFC NAA/Cr ratios in children and adolescents with BD, it is stillpossible that such levels begin to decrease after longer durations of illnessinto adulthood. Longitudinal neuroimaging studies of patients with BDaccounting for developmental and treatment factors are needed tofurther clarify the neurodegenerative aspects of BD.

Kim A Gallellia, Christopher MWagnera, Asya Karchemskiya,Meghan Howea, Daniel Spielmanb,Allan Reissa and Kiki D Changa

Departments of aPsychiatry and Behavioral

Sciences, and bRadiology, Stanford University

School of Medicine, Stanford, CA, USA

Key words: adolescents – bipolar disorder –

children – magnetic resonance spectroscopy –

N-acetylaspartate – offspring – prodromal

Received 3 December 2004, revised and accepted

for publication 2 May 2005

Corresponding author: Kiki D Chang, Division of

Child and Adolescent Psychiatry, Stanford Univer-

sity School of Medicine, 401 Quarry Road, Stanford,

CA 94305-5540, USA. Fax: +1 650 723 5531;

e-mail: kchang88@stanford.edu

The authors of this paper do not have any commercial associations

that might pose a conflict of interest in connection with this manu-

script.

Bipolar Disorders 2005: 7: 589–597Copyright ª Blackwell Munksgaard 2005

BIPOLAR DISORDERS

589

with BD. Postmortem studies of adults withBD when compared with healthy controls havereported reductions in neuronal and glial DLPFCdensity (8, 9). In a functional magnetic resonanceimaging (fMRI) study, adults with BD haddecreased DLPFC activation when viewingfearful faces when compared with healthy controls(10). These findings all suggest that DLPFCabnormalities, especially decreased neuronaldensity, are present in BD.However, it remains to be determined whether

these prefrontal abnormalities exist prior to theonset of BD and could therefore be utilized aspredictive biological markers of the illness. Previ-ous research suggests that decreased NAA levels inBD may be the result of a neurodegenerativeprocess, rather than a static trait marker (acharacteristic present from birth). We previouslyreported decreased NAA/Cr ratios in the rightDLPFC in 15 children and adolescents withfamilial BD when compared with healthy controls.In addition, right DLPFC NAA/Cr tended todecrease as duration of the bipolar illnessincreased, supporting a neurodegenerative process(3). Ideally, in order to further examine neurode-generation in DLPFC in children with BD, exam-ination of NAA levels in an identified group ofchildren at high-risk for developing BD before theonset of the illness is essential.Children of parents with BD, also known as

�bipolar offspring,� are at high-risk for developingthe illness themselves (11, 12). Furthermore, it hasbeen suggested that bipolar offspring who have acombination of mood difficulties and disruptivebehavior disorders may be experiencing a prodro-mal form of BD (11, 13, 14). Specifically, thosewith ADHD and subsyndromal symptoms ofdepression and mania may be at extremely high-risk for full BD development.Through comparing levels of neuronal sub-

strates in children with familial BD to those athigh-risk for developing BD, one can attempt tofurther elucidate the meaning of varying NAA andother substrate levels associated with the illness.Therefore, we wished to expand upon our previousstudy and use 1H-MRS to study DLPFC NAA/Crratios in three different groups of children andadolescents: bipolar offspring in the early course ofBD (bipolar group); bipolar offspring with putativeprodromal symptoms of BD (prodromal group)and children with no psychiatric disorder and nofamily history of BD (control group). We hypoth-esized that decreased NAA may be the result of aneurodegenerative process, rather than being rep-resentative of a neurobiological marker of child-hood BD. Therefore, we anticipated that DLPFC

NAA/Cr ratios would be lower in bipolar offspringwith BD when compared with bipolar offspringwith prodromal symptoms of BD and healthycontrols – with no detectable differences in NAA/Cr ratios between the latter two groups.

Materials and methods

This protocol was approved by the StanfordUniversity Panel of Medical Research in HumanSubjects. Thirty-two children and adolescents withBD, 28 subjects with prodromal symptoms, and26 healthy volunteers were recruited from anongoing study of bipolar offspring and from thecommunity. After obtaining oral and writteninformed consent from parents and oral andwritten assent from their offspring, semi-struc-tured interviews were conducted. All subjects wereevaluated by the affective disorders module of theWashington University in St Louis Kiddie Sched-ule for Affective Disorders and Schizophrenia(WASH-U-KSADS) (15, 16), and the Schedulefor Affective Disorders and schizophrenia forSchool-age Children, present and Lifetime(K-SADS-PL) (17). Subjects were evaluated eitherby a child psychiatrist (KC) or a trained masters-level research assistant, who were both aware ofparental diagnosis. Inter-rater reliability wasestablished at the outset by rating videotapedinterviews, observing trained rater interviews, andperforming interviews with observation by atrained rater, as described by Geller et al. (fourconsecutive patients with 100% agreement ondiagnoses) (18). Diagnostic decisions were ulti-mately made by a child psychiatrist (KC) basedon personal interview, discussion with theresearch assistant, and written notes of parentaland subject responses to individual WASH-U-KSADS questions. Current and lifetime diagnoseswere established according to DSM-IV criteria.Parents were euthymic at the time of their own andtheir child’s interview. Age of onset of BD wasdetermined as the earliest period to the closestmonth that patients met first criteria for a manic orhypomanic episode, as defined by the DSM-IV.For inclusion in the bipolar and prodromal

group, children and adolescents between the agesof 9 and 18 years had at least one parent withbipolar I or II disorder as diagnosed by theStructured Clinical Interview for DSM-IV Axis IDisorders (SCID) (19), administered by a trainedmasters-level clinician and/or board certified psy-chiatrist (KC). In addition to parental diagnosis,for inclusion in the bipolar group, subjects neededa diagnosis of bipolar I or II disorder by theWASH-U-KSADS.

Gallelli et al.

590

For inclusion in the prodromal group, in addi-tion to parental diagnosis of BD, all children metcriteria for ADHD and had at least moderate moodsymptoms, as indicated by a score of >10 on theYoungMania Rating Scale (YMRS) (20) or a scoreof >29 on the Children’s Depressive RatingScale – Revised (CDRS-R) (21). The YMRS is an11-item semi-structured interview that has beenshown to be reliable in detectingmanic symptoms inprepubertal children and adolescents (22, 23). TheCDRS-R is a 17-item semi-structured interview thatreliably measures childhood depressive symptoms.For inclusion in the control group, healthy

volunteers did not have a DSM-IV psychiatricdiagnosis, had both parents without any diagnosisby SCID, and did not have a first- or second-degreerelative with BD as determined by the FamilyHistory Research Diagnostic Criteria (24).Subjects were all outpatients at the time of

scanning. Patients with BD were administered theclinician-rated YMRS (20, 23), and themselvescompleted the Childhood Depression Inventory(CDI) (25), with the help of a parent if they wereless than 12 years old. Patients had any psycho-stimulants discontinued for 24 h before the MRS,primarily due to a concurrent, separate fMRIstudy of attention. The effects of psychostimulantson neurometabolites measured by 1H-MRS areunknown; regardless, as the half-life of the major-ity of psychostimulants available at the time of thisstudy was less than 8 h, 24 h was consideredadequate time for at least cessation of behavioraleffects. They were allowed to continue any othercurrent medications such as mood-stabilizers orantidepressants due to the risk of mood destabili-zation. Subjects� medication history was obtainedand used for exploratory analyses and for covariateanalyses of 1H-MRS findings.Subjects were scanned on a 3-T GE Signa scanner

with Echospeed gradients using a custom-built headcoil with a 50% advantage in signal-to-noise ratiothan the standard GE coil (General Electric MedicalSystems,Milwaukee,WI,USA).Eighteenaxial slices(4 mm thick, 0.5 mm skip) parallel to the anterior–posterior commissure plane covering the wholebrain were obtained with a temporal resolution of3 s using a T2-weighted gradient echo spiral pulsesequence (TR ¼ 3000 ms,TE ¼ 30 ms, flip angle ¼89� and 1 interleave). The field of view was 200 mmand the in-plane spatial resolution 3.125 mm. Toaid in voxel segmentation, high-resolutionT1-weighted spoiled grass gradient recalled (SPGR)3D MRI sequences with the following parame-ters were used: TR ¼ 35 ms; TE ¼ 6 ms; flipangle ¼ 45�; 24 cm field of view; 124 slices incoronal plane; 256 · 192 matrix; acquired resolu-

tion ¼ 1.5 · 0.9 · 1.2 mm3. The images werereconstructed as a 124 · 256 · 256 matrix with a1.5 · 0.9 · 0.9 mm3 spatial resolution.For 1H-MRS, a 2 · 2 · 2 cm voxel was pre-

scribed in the right and then left DLPFC, from thefirst axial slice above the lateral ventricles. As sliceswere 5 mm thick, the voxel was placed anywherefrom 0 to 5 mm above the lateral ventricles,immediately anterior to a line drawn between theanterior aspects of the lateral ventricles, and as farlateral as possible while remaining in the cerebrumand visually maintaining approximately equalparts of gray and white matter (Fig. 1). Aninvestigator blind to diagnosis visually inspectedeach voxel placement to ensure proper placementfully within the brain and that spectra contain nosizable lipid peaks or rolling baselines. MRS datawas acquired using a preselected region of interestfor point-resolved spectroscopy (PRESS) with aTR/TE of 2000/35 ms. MRS scans used 32 aver-ages, 1 kHz spectral bandwidth, 1 k data points,and unsuppressed water was collected for allspectra. The MRS scan was 1 min 44 s in length.We were able to obtain adequate signal-to-noiseratio with this relatively short acquisition time dueto the relatively large field strength of 3 T. Thefully automated PROBE/SV quantification tool(General Electric Medical Systems, Milwaukee,WI, USA) was used to process MRS data. Each ofthe five spectral peaks associated with NAA, Cr,Cho, mI, and H2O was quantified by Levenberg–Marquardt curve fitting over that line region usingthe standard data processing package by GEmentioned above (Fig. 2).

Fig. 1. Position of dorsolateral prefrontal voxels.

Comparison of NAA in bipolar offspring

591

Statistical analysis

Differences in NAA/Cr ratios were consideredprimary outcome measures and were comparedamong groups using analysis of variance (ANOVA).A significance threshold was set at p < 0.025, as aBonferroni-type correction was applied for multi-ple comparisons (left and right hemispheres). Othermetabolite ratios including NAA/Cho, NAA/Cho + Cr, Cho/Cr, and mI/Cr were consideredsecondary and exploratory. Group comparisons onthese additional metabolites were conducted usinga multiple analysis of variance (MANOVA).Demographic information between groups wasanalyzed using either ANOVA or chi-squareanalyses. Spearman rank correlations were usedto test correlations of metabolite ratios withexposure to valproate, exposure to lithium, sex,age, and duration of BD illness (in the bipolargroup only). Because of these multiple compar-isons, conducted in both the right and left hemi-spheres of each subject, a Bonferroni-typecorrection was applied to significance levels. Theadjusted significance threshold for the prodromalgroup was set at p < 0.006; and p < 0.005 for thebipolar group.

Results

Cohort

Subjects in the prodromal group were significantlyyounger than both the bipolar and control groups(12.2 ± 2.6 versus 14.1 ± 3.0 and 14.2 ± 2.8,

respectively; F ¼ 4.75, p ¼ 0.01). There were nodifferences in ethnicity or socioeconomic statusamong the bipolar, prodromal and control groups(Table 1). There was also no significant interactionof gender with NAA/Cr ratios (F ¼ 0.35, df ¼1.84, p ¼ 0.56).There was no significant difference between

YMRS scores in the bipolar (M ¼ 13.8 ± 9.1)and prodromal groups (M ¼ 13.6 ± 5.62; p ¼0.94). In addition, CDI scores did not differbetween the two groups (M ¼ 16.6 ± 9.4 and11.5 ± 9.7, respectively; p ¼ 0.08). However, thebipolar group had significantly higher CDRSscores than the prodromal group (M ¼42.8 ± 16.1 and 33.9 ± 7.28, respectively, p ¼0.05), suggesting the presence of more depressivesymptoms in the bipolar group. Although mostsubjects were not experiencing a manic or depres-sive episode at the time of scan, the scores on themood symptom measures suggest the presence of

Fig. 2. Sample magnetic resonance spectrum.

Table 1. Demographics and comorbid diagnoses of offspring

Bipolar Prodromal Control

n 32 28 26Mean age, years (SD) 14.1 (3.0)* 12.2 (2.6)* 14.2 (2.8)*Gender, (%) female 10 (31) 9 (32) 9 (35)SES (SD) 4.0 (0.91) 4.2 (0.77) 4.5 (0.85)Race (%)

African-American 0 (0) 1 (4) 1 (4)Hispanic 2 (6) 1 (4) 1 (4)Asian 0 (0) 0 (0) 1 (4)Caucasian 28 (88) 22 (79) 20 (77)Other 2 (6) 4 (14) 3 (12)

I.Q. (SD) 108 (12) 108 (13) 115 (9.5)First- and second-degreerelatives with a mooddisorder, % (SD)

51 (21) 47 (13) 0 (0)

Comorbid diagnoses of offspring, n (%)ADHD 29 (91) 28 (100) 0 (0)Anxiety disorders 11 (34) 8 (29) 0 (0)ODD 18 (56) 12 (43) 0 (0)

Past psychotropic medication exposure, n (%)Stimulants 16 (50) 11 (39) 0 (0)TCAs 14 (44) 6 (21) 0 (0)SSRIs 15 (47) 6 (21) 0 (0)Atypical ADs 6 (19) 7 (25) 0 (0)Lithium 8 (25) 3 (11) 0 (0)Valproate 15 (47)* 5 (18)* 0 (0)Antipsychotics 14 (44) 7 (25) 0 (0)Any mood-stabilizer 17 (53)* 6 (21)* 0 (0)

SD ¼ standard deviation; SES ¼ socioeconomic status; Anxietydisorder ¼ separation anxiety disorder, generalized anxietydisorder, social phobia, obsessive–compulsive disorder, orpost-traumatic stress disorder; ADHD ¼ attention-deficit/hyper-activity disorder; ODD ¼ oppositional defiant disorder; TCAs ¼tricyclic antidepressants; SSRIs ¼ selective-serotonin reuptakeinhibitors; ADs ¼ antidepressants. *significant group differenceat p < 0.05.

Gallelli et al.

592

more depressive symptoms, with fewer manicsymptoms at the time of scan. There were nosignificant correlations of YMRS, CDI or CDRSscores with NAA/Cr ratios in right and lefthemispheres.Twenty-eight (88%) subjects in the bipolar group

had previously taken psychotropic medications.Fifty percent of these subjects had significant pastexposure (more than 2 months) to stimulants; 44%to tricyclic antidepressants, 47% to selective-sero-tonin reuptake inhibitors (SSRIs); 19% to atypicalantidepressants; 44% to antipsychotics, and 53%to mood-stabilizers, including 25% with exposureto lithium and 47% with exposure to valproate(Table 1). However, at the time of scan thepercentage of subjects in the bipolar group activelybeing treated with medication were as follows:25% with stimulants (however stimulants werediscontinued for 24 h prior to scan); 6.3% withtricyclic antidepressants; 31.3% with SSRIs; 15.6%with atypical antidepressants; 41% with antipsy-chotics; 50% were being treated with mood-stabilizers, including 15.6% on lithium and 25%on valproate.Twenty (71%) subjects in the prodromal group

had previously taken psychotropic medications.Thirty-nine percent were exposed to stimulants,21% to tricyclic antidepressants, 21% to SSRIs,25% to atypical antidepressants, 25% to antipsy-chotics and 21% tomood-stabilizers, including 11%with exposure to lithium and 18% with exposure tovalproate. At the time of scan the percentage ofsubjects in the prodromal group being activelytreated with medication were as follows: 25% withstimulants; 0% with tricyclic antidepressants;10.7% with SSRIs; 25% with atypical antidepres-sants, 21.4% with antipsychotics and 32.1% were

treated with mood-stabilizers, including 10.7% onlithium and 14.3% on valproate.No subjects in the control group had previous

exposure to psychotropic medications. Comparedwith the prodromal group, significantly moresubjects inthebipolargroupwerepreviouslyexposedto mood-stabilizers in general (Fisher’s exacttest ¼ 0.017). In addition, subjects in the bipolargroup had significantly more lifetime exposure tovalproate (Fisher’s exact test ¼ 0.027). Addition-ally at the time of scan, there were no significantdifferences between groups in current medicationexposure. In the bipolar group, 90% of subjectshad comorbid attention-deficit/hyperactivity disor-der (ADHD); 34% had a comorbid diagnosis ofanxiety disorder; and 56% had comorbid opposi-tional defiant disorder (ODD). By definition, all ofthe subjects in the prodromal group had a diag-nosis of ADHD. Twenty-eight percent of prodro-mal subjects were diagnosed with an anxietydisorder, and 42% with ODD. None of thesubjects in any group had a present or pastsubstance-use disorder.

NAA/Cr

There were no significant group differences in meanright or left DLPFC NAA/Cr ratios (Table 2). Inaddition, exploratory analyses of additional me-tabolites yielded no significant differences in NAA/Cho, NAA/Cho + Cr, Cho/Cr or mI/Cr betweenthe three groups (Table 2). For subjects in thebipolar group, left and right DLPFC NAA/Crratios were not correlated with age, duration ofillness, exposure to lithium or exposure to valpro-ate. Additionally, for subjects in the prodromalgroup, left and right DLPFC NAA/Cr ratios were

Table 2. Neurometabolite levels in dorsolateral prefrontal cortex (DLPFC)

Metabolite ratioBipolar group,mean (SD)

Prodromal group,mean (SD)

Control group,mean (SD) Significance (p) F

n 32 28 26Right NAA/Cr 1.7 (0.13) 1.7 (0.13) 1.7 (0.10) 0.75 0.29Left NAA/Cr 1.6 (0.14) 1.6 (0.13) 1.6 (0.12) 0.99 0.00Right NAA/Cho 2.2 (0.23) 2.2 (0.29) 2.3 (2.3) 0.35 1.1Left NAA/Cho 2.1 (0.21) 2.1 (0.23) 2.2 (0.18) 0.28 1.3Right NAA/Cho + Cr 0.95 (0.07) 0.94 (0.09) 0.97 (0.07) 0.38 0.98Left NAA/Cho + Cr 0.91 (0.06) 0.90 (0.07) 0.92 (0.06) 0.57 0.56Right Cho/Cr 0.77 (0.08) 0.78 (0.08) 0.75 (0.06) 0.46 0.78Left Cho/Cr 0.78 (0.11) 0.79 (0.08) 0.75 (0.06) 0.31 1.2Right mI/Cr 0.48 (0.05) 0.49 (0.06) 0.47 (0.03) 0.19 1.7Left mI/Cr 0.48 (0.05) 0.51 (0.05) 0.50 (0.06) 0.15 2.0

NAA ¼ N-acetylaspartate; Cr ¼ creatine-phosphocreatine; Cho ¼ choline; mI ¼ myoinositol.

Comparison of NAA in bipolar offspring

593

not correlated with exposure to lithium or valpro-ate (p > 0.006).Given that the prodromal group was signifi-

cantly younger than both the bipolar and controlgroup, we also analyzed NAA/Cr ratios in a subsetof age-matched subjects. Mean left and rightDLPFC NAA/Cr ratios were compared between18 control subjects, 25 prodromal subjects and 24subjects with BDs, ranging in age from 9 to16.3 years. The three groups did not differ in age(mean age of 12.82, SD ¼ 21, 12.68, SD ¼ 2.2, and12.95, SD ¼ 2.6, for control, prodromal andbipolar groups, respectively; F ¼ 0.083, df ¼ 2,p ¼ 0.92). Also, with this age-matched subset, nosignificant differences were found between thegroups in either left or right DLPFC NAA/Crratios (F ¼ 0.018, df ¼ 2 p ¼ 0.98; and F ¼ 0.082,df ¼ 2, p ¼ 0.92, respectively).

Voxel composition

We chose a representative sample of subjects fromeach diagnostic group for analysis of MRS voxelcomposition (gray/white matter ratios). Pairwiset-tests (three pairs per each variable) were conduct-ed to test for group differences on a subsample ofsubjects. None of these t-tests showed significantdifferences between groups at p ¼ 0.05, indicatingno differences in percent voxel composition ofgray/white matter (Table 3).

Discussion

This study attempted to replicate and expand uponprevious findings of decreased right DLPFC NAA/Cr ratios in children with familial BD, aiming toclarify the role of this metabolite as one potentialneurobiological marker of pediatric BD. In thepresent study, we did not find any evidence ofdifferences in DLPFC NAA/Cr ratios between thedifferent comparison groups of children and ado-lescents: bipolar offspring with BD, bipolar off-spring with prodromal symptoms of BD, andhealthy children with no family history of BD. Insupport of our neurodegenerative hypothesis, wehad predicted that there would be no significant

difference in NAA/Cr ratios between the prodro-mal and control group. However, based uponfindings from our previous study that founddecreased NAA/Cr levels (by 4.2%) in the rightDLPFC of children with familial BD (5), we hadanticipated significant differences between thebipolar and control groups. To detect NAA/Crdifferences between the groups in this study, at80% power with a significance level of p ¼ 0.05,we were adequately powered to detect a differenceof 4.7%.Similarities in NAA/Cr ratios between the three

groups could not be explained by differences inmedication exposure – specifically the effects ofmood-stabilizers such as lithium and valproate –nor was there any detected relationship betweenduration of illness and metabolite levels in thebipolar group. Finally, exploratory analyses ofother metabolite ratios, such as NAA/Cho, NAA/Cho + Cr, Cho/Cr, and mI/Cr also failed toreveal any significant group differences.Thus while there are some limitations to our

ability to detect true differences in metaboliteconcentrations between these three groups, it islikely that decreases in NAA in DLPFC simplydo not occur before or shortly after the develop-ment of the first manic episode in pediatric-onsetBD. NAA levels in other cortical regions still mightbe used to detect prodromal BD. For example,Cecil et al. conducted an MRS study on ninechildren with a mood disorder and familial risk forBD. In comparison with healthy controls, the MDgroup had an 8% decrease of NAA levels withinthe cerebellar vermis, and a 16% elevation of mIconcentration levels in the frontal cortex (3). Thiscohort was somewhat different from our prodro-mal cohort, as seven subjects (78%) were diag-nosed with BD, and the remaining two childrenwere diagnosed with major depressive disorder.Additionally, only three subjects in the MD group(33%) had co-morbid ADHD when comparedwith the majority of subjects in our bipolar andprodromal groups. So, while direct comparisonsbetween these studies cannot be made, regionsother than DLPFC may still show abnormalmetabolite levels in prodromal BD.

Table 3. Percent voxel composition of subjects

Bipolar group(n ¼ 15), % (SD)

Prodromal group(n ¼ 16), % (SD)

Control group(n ¼ 11), % (SD) Significance (p)

Left white 56.0 (6.6) 52.3 (4.6) 53.9 (5.4) 0.09*, 0.39**, 0.44***Left gray 41.8 (6.5) 45.3 (4.5) 44.0 (5.1) 0.10*, 0.36**, 0.51***Right white 56.5 (8.6) 54.2 (5.1) 55.4 (5.3) 0.36*, 0.80**, 0.57***Right gray 41.3 (8.6) 43.1 (5.8) 42.1 (4.3) 0.51*, 0.76**, 0.62***

*Bipolar–prodromal comparison; **bipolar–control comparison; ***prodromal–control comparison.

Gallelli et al.

594

Although we could not support our previousfindings of decreased DLPFC NAA/Cr in pediatricBD, neurodegenerative processes in BD should notbe ruled out. Because of the relatively short illnessduration in our bipolar group (2.0 years, SD ¼1.92), there was a relatively small range of yearswith which to correlate duration of illness withNAA/Cr change. Therefore, it is possible that aftergreater illness duration we would begin to detectgreater decreases in DLPFC NAA/Cr. Findings ofdecreased NAA/Cr levels in studies of adults withBD (9% decrease in the left DLPFC and a 6.5%decrease in the right DLPFC, when compared withcontrols) appear quite robust and may still suggestneurodegenerative processes (4). Compared withthe short illness duration in our young subjectswith BD, adults in those studies likely had experi-enced their illness for much longer. However, thesestudies did not report illness duration of subjects.Perhaps neurodegeneration, if it is occurring, doesnot appear until significantly more time than2 years has passed after the onset of BD, such asentry into adulthood.Later life prefrontal neurodegeneration is sup-

ported by findings from fMRI studies of childrenand adults with BD. In an fMRI study of adults,subjects with BD had decreased DLPFC activationwhen viewing fearful faces when compared withhealthy controls (10). However, in an fMRI studyof euthymic children and adolescents with BDperforming both cognitive and affective tasks,increased activation in the DLPFC was detected(26). This finding is consistent with these childrenstill having relatively intact neuronal density inDLPFC, allowing for overactivation of this area.These seemingly contradictory findings may beexplained by a prefrontal-subcortical model ofmood regulation (27, 28). Increased prefrontalactivation in euthymic children with BD duringboth cognitive and affective tasks may be incompensatory response to increased activation insubcortical limbic areas, in an effort to regulatemood. With longer duration of illness, it is possiblethat neurodegeneration in DLPFC (as reflected bydecreased NAA and neuronal/glial density inadults with BD) would eventually lead to lessability to overactivate this area, and thus underac-tivity of the DLPFC in response to emotion-relatedtasks (10, 26). Euthymic children and adolescentswith BD, at the early stage of illness, because oftheir relatively intact prefrontal neuronal density,may still able to overactivate prefrontal areas (26).Another developmental confound is the sugges-

tion that in normal development, there are non-linear age-related changes of NAA concentrationsin frontal cortex gray matter. Recent research has

shown that cortical NAA tends to increase throughchildhood, until about age 10 years, and thencontinues to decrease thereafter. It has been arguedthat these changes may be associated with dendriticand synaptic development and regression (29).Thus neurodegeneration may not be detected untillater developmental stages, perhaps early adult-hood, or when NAA concentrations begin tostabilize. Additionally, subjects in the prodromalgroup were significantly younger by approximately2 years when compared with the bipolar andcontrol groups, so it is possible that NAA levelscould still decrease in prodromal subjects to levelsbelow those in control subjects after two moreyears. However, to address this concern we ana-lyzed a subset of subjects that were age-matched tothe younger prodromal subjects and still found nosignificant group differences in NAA/Cr ratios.In addition to the impact of developmental stage

of illness on our findings, there exists anotherpotential confound regarding the exposure tovarious psychotropic medications in most of ourbipolar and prodromal subjects. Although theeffects of psychotropic medications on specificneurometabolites remains to be studied, it hasbeen suggested that pharmacological agents maypotentially compromise the use of Cr as an internalstandard (30), thus influencing our measurement ofprefrontal neuronal density of NAA/Cr. Research-ers have also reported discrepant findingsregarding the effects of medication treatment onbrain NAA levels. MRS and MRI studies havesuggested that lithium may increase brain NAAconcentrations (30) and total brain gray mattervolumes in adults with BD (30, 31). However, a1H-MRS study found that lithium administrationdid not increase NAA in the DLPFC of healthyadult volunteers (32). Past lithium exposureappeared to have had no effect on DLPFC NAA/Cr levels in the children in our sample as well.Thus, it is possible that lithium may preferentiallyaffect gray matter in regions of the brain other thanDLPFC, if at all.It should also be noted that the role of NAA is

unclear and may not only function as a marker ofneuronal density. It is found in both white and graymatter and abnormalities in NAA have also beenfound in diseases of white matter, such as multiplesclerosis (33). Several technological limitations ofthis study also need to be acknowledged. Anormalizing phantom was not used in this studyto control for scanner drift, which could result invariable detection of neurometabolite concentra-tions over time. However, all three groups werescanned in random orders; that is, all groups wouldhave been subject to the same variances in

Comparison of NAA in bipolar offspring

595

technology over time. Also, test/retest checks ofvoxel placement were not done to prevent slightvariations in voxel placements across subjects. Anadditional potential limitation is that at a TE of35 ms, additional resonances will be present in thespectrum, beyond the five that were fitted, includ-ing glutamate, glutamine, aspartate, glucose andother compounds. We decided not to use a longerecho time to more specifically isolate the NAApeak, as decreased signal-to-noise ratiowould occur.Also, we tried to maximize our ability to obtainaccurate data by using a 3-T field strength, whichoffers better peak separation than 1.5 T, which hasbeen used in most previous similar studies.Although there has been some controversy

regarding using metabolite ratios versus absoluteconcentrations, we felt that calculating absoluteconcentrations, which require multiple calcula-tions, would significantly increase the error margin.Also, by using ratios, which remain constant, wecan compensate for potential scanner drift overtime. Finally, our understanding of the relationshipbetween neurochemistry and mood disorders hasbeen clouded by the disparate and sometimescontradictory findings of existing MRS research.Technological confounds, such as the use ofdifferent voxel acquisitions across studies, mayimpact comparative findings. Therefore, MRStechnology may need further refinement beforeminute differences in neurometabolites in ourpediatric subjects can be detected and such studiescan be replicated.Another limitation involves the use of the

prodromal comparison group. We cannot becertain that subjects in the prodromal group wouldall eventually develop BD. In the absence ofdefinitive longitudinal studies of this population,and based on adult retrospective reports andclinical observations of the natural developmentof BD in children (11, 13), we maintain that amajority of these children are truly prodromal forBD. Finally, we should note that our study wasfocused on DLPFC NAA levels only. Otherresearchers have found decreased NAA in cerebel-lar vermis (3) and hippocampus (34) of patientswith BD, suggesting the need for further explora-tion of NAA levels in these and other brain regionsin patients with and at high-risk for BD.Despite these limitations, this is the largest

1H-MRS study performed as yet investigatingNAA/Cr metabolite levels in children with familialBD and in a population at high-risk for developingBD. Our findings suggest that DLPFC NAA/Crlevels should not be used as an early marker forBD. Instead, we suggest refocusing efforts oninvestigating the possibility of prefrontal neurode-

generation after BD development, which may onlybe detected by prospective, longitudinal studiesconducted on subjects of different ages and stagesof development of BD.

Acknowledgements

This work was supported in part by NIH grants MH01142,MH19908, MH050047 and HD31715 (Dr Reiss), grants fromNARSAD, the Klingenstein Third Generation Foundation,the Heinz C. Prechter Fund, and NIH grant MH64460-01 (DrChang). The authors are grateful to Lin Shen, B.A. for histechnical assistance with data processing.

References

1. Miller BL. A review of chemical issues in 1H NMRspectroscopy: N-acetyl-L-aspartate, creatinine and choline.NMR Biomed 1991; 4: 47–52.

2. Ross B, Michaelis T. Clinical applications of magneticresonance spectroscopy. Magn Reson Q 1994; 10: 191–247.

3. Cecil KM, DelBello MP, Sellars MC, Strakowski SM.Proton magnetic resonance spectroscopy of the frontallobe and cerebellar vermis in children with a mood disorderand a familial risk for bipolar disorders. J Child AdolPsychopharm 2003; 4: 545–555.

4. Winsberg ME, Sachs N, Tate DL, Adalsteinsson E,Spielman D, Ketter TA. Decreased dorsolateral prefrontalN-acetyl aspartate in bipolar disorder. Biol Psychiatry2000; 47: 475–481.

5. Chang K, Adleman N, Dienes K, Barnea-Goraly N, ReissA, Ketter T. Decreased N-acetylaspartate in children withfamilial bipolar disorder. Biol Psychiatry 2003; 53: 1059–1065.

6. Pardo JV, Pardo PJ, Raichle ME. Neural correlates ofself-induced dysphoria. Am J Psychiatry 1993; 150: 713–719.

7. MacDonald AW III, Cohen JD, Stenger VA, Carter CS.Dissociating the role of the dorsolateral prefrontal andanterior cingulate cortex in cognitive control. Science 2000;288: 1835–1838.

8. Ongur D, Drevets WC, Price JL. Glial reduction in thesubgenual prefrontal cortex in mood disorders. Proc NatlAcad Sci U S A 1998; 95: 13290–13295.

9. Rajkowska G. Cell pathology in bipolar disorder. BipolarDisord 2002; 836: 253–268.

10. Yurgelon-Todd DA, Gruber SA, Kanayam G, KilgoreWD, Baird AD, Young AD. FMRI during affect discrim-ination in bipolar affective disorder. Bipolar Disord 2000;2: 237–248.

11. Chang K, Steiner H, Dienes K, Adleman N, Ketter T.Bipolar offspring: a window into bipolar disorder evolu-tion. Biol Psychiatry 2003; 53: 945–951.

12. DelBello MP, Geller B. Review of studies of child andadolescent offspring of bipolar parents. Bipolar Disord2001; 3: 325–334.

13. Carlson GA, Wintraub S. Childhood behavior problemsand bipolar disorder-relationship or coincidence? J AffectDisord 1993; 28: 143–153.

14. Wozniak J et al. Mania-like symptoms suggestive ofchildhood-onset bipolar disorder in clinically referredchildren. J Am Acad Child Adolesc Psychiatry 1995; 34:867–876.

15. Geller B, Zimerman B, Williams M et al. Reliability of theWashington University in St. Louis Kiddie Schedule for

Gallelli et al.

596

AffectiveDisorders and Schizophrenia (WASH-U-KSADS)mania and rapid cycling sections. J Am Acad ChildAdolesc Psychiatry 2001; 40: 450–455.

16. Geller BG, Williams M, Zimerman B, Frazier J. WASH-U-KSADS (Washington University in St. Louis KiddieSchedule for Affective Disorders and Schizophrenia).St Louis, MO: Washington University, 1996.

17. Kaufman J, Birmaher B, Brent D et al. Schedule forAffective Disorders and Schizophrenia for School-AgeChildren-Present and Lifetime Version (K-SADS-PL):initial reliability and validity data. J Am Acad ChildAdolesc Psychiatry 1997; 36: 980–988.

18. Geller BG, Warner K, Williams M, Zimerman B. Prepu-bertal and young adolescent bipolarity versus ADHD:assessment and validity using the WASH-U-KSADS,CBCL AND TRF. J Affect Disord 1998; 51: 93–100.

19. First MB, Spitzer RL, Gibbon M, Williams JBW. Struc-tured Clinical Interview for DSM-IV Axis I Disorders –Patient Edition (SCID-I/P, version 2.0). New York, NY:Biometric Research, New York State Psychiatric Institute,1995.

20. Young RC, Biggs JT, Ziegler VE, Meyer DA. A ratingscale for mania: reliability, validity and sensitivity. Br JPsychiatry 1978; 133: 429–435.

21. Poznanski EO, Freeman LN, Mokros HB. Children’sdepressive rating scale-revised. Pyschopharmacol Bull1985; 21: 979–989.

22. Fristad MA, Weller EB, Weller RA. The mania ratingscale: can it be used in children? A preliminary report.J Am Acad Child Adolesc Psychiatry 1992; 31: 252–257.

23. Fristad MA, Weller RA, Weller EB. The Mania RatingScale (MRS): further reliability and validity studies withchildren. Ann Clin Psychiatry 1995; 7: 127–132.

24. Andreasen NC, Endicott J, Spitzer RL, Winokur G. Thefamily history method using criteria. reliability and valid-ity. Arch Gen Psychiatry 1977; 34: 1229–1235.

25. Kovacs M. The Children’s Depression, Inventory (CDI).Psychopharmacol Bull 1985; 21: 995–998.

26. Chang K, Adleman NE, Dienes K, Simeonova DI, MenonV, Reiss A. Anomalous prefrontal-subcortical activation infamilial pediatric bipolar disorder: a functional magneticresonance imaging investigation. Arch Gen Psychiatry2004; 61: 781–792.

27. Mayberg HS. Limbic-cortical dysregulation: a proposedmodel of depression. J Neuropsychiatry Clin Neurosci1997; 9: 471–481.

28. Blumberg HP, Charney DS, Krystal JH. Frontotemporalneural systems in bipolar disorder. Semin Clin Neuropsy-chiatry 2002; 7: 243–254.

29. Horska A, Kaufmann WE, Brant LJ, Naiju S, Harris JC,Parker PB. In vivo quantitative proton MRSI study ofbrain development from childhood to adolescence. J MagnReson Imaging 2002; 15: 137–143.

30. Moore GJ, Bebchuk JM, Hasanat K et al. Lithiumincreases N-acetyl-aspartate in the human brain: in vivoevidence in support of bcl-2�s neurotrophic effects? BiolPsychiatry 2000; 48: 1–8.

31. Sassi RB, Nicoletti M, Brambilla P et al. Increased graymatter volume in lithium-treated bipolar disorder patients.Neurosci Lett 2002; 329: 243–245.

32. Brambilla P, Stanley JA, Sassi RB et al. 1H MRS study ofdorsolateral prefrontal cortex in healthy individuals beforeand after lithium administration. Neuropsychopharmacol-ogy 2004; 29: 1918–1924.

33. Narayana PA, Wolinsky JS, Rao SB, He R, Mehta M;PROMise Trial MRSI Group. Multicentre proton mag-netic resonance spectroscopy imaging of primary progres-sive multiple sclerosis. Mult Scler 2004; 10: S73–S79.

34. Bertolino A, Frye M, Callicott JH et al. Neuronalpathology in the hippocampal area of patients with bipolardisorder: a study of proton magnetic resonance spectro-scopic imaging. Biol Psychiatry 2003; 53: 906–913.

Comparison of NAA in bipolar offspring

597