Directory UMM :Data Elmu:jurnal:B:Biological Psichatry:Vol49.Issue2.2001:

Effects of Mood and Subtype on Cerebral Glucose
Metabolism in Treatment-Resistant Bipolar Disorder
Terence A. Ketter, Tim A. Kimbrell, Mark S. George, Robert T. Dunn,
Andrew M. Speer, Brenda E. Benson, Mark W. Willis, Aimee Danielson,
Mark A. Frye, Peter Herscovitch, and Robert M. Post
Background: Functional brain imaging studies in unipolar and secondary depression have generally found decreased prefrontal cortical activity, but in bipolar disorders findings have been more variable.
Methods: Forty-three medication-free, treatment-resistant,
predominantly rapid-cycling bipolar disorder patients and
43 age- and gender-matched healthy control subjects had
cerebral glucose metabolism assessed using positron emission tomography and fluorine-18-deoxyglucose.
Results: Depressed bipolar disorder patients compared to
control subjects had decreased global, absolute prefrontal
and anterior paralimbic cortical, and increased normalized subcortical (ventral striatum, thalamus, right amygdala) metabolism. Degree of depression correlated negatively with absolute prefrontal and paralimbic cortical,
and positively with normalized anterior paralimbic subcortical metabolism. Increased normalized cerebello-posterior cortical metabolism was seen in all patient subgroups compared to control subjects, independent of mood
state, disorder subtype, or cycle frequency.
Conclusions: In bipolar depression, we observed a pattern of prefrontal hypometabolism, consistent with observations in primary unipolar and secondary depression,
suggesting this is part of a common neural substrate for
depression independent of etiology. In contrast, the cerebello-posterior cortical normalized hypermetabolism
seen in all bipolar subgroups (including euthymic) suggests a possible congenital or acquired trait abnormality.
The degree to which these findings in treatment-resistant,
predominantly rapid-cycling patients pertain to community samples remains to be established. Biol Psychiatry

2001;49:97–109 © 2001 Society of Biological Psychiatry

From the Department of Psychiatry and Behavioral Science, Stanford University
School of Medicine, Stanford, California (TAKe), Biological Psychiatry
Branch, National Institute of Mental Health, Bethesda, Maryland (TAKe,
TAKi, MSG, RTD, AMS, BEB, MWW, AD, MAF, RMP), Veterans Administration Medical Center, North Little Rock, Arkansas (TAKi), Medical
University of South Carolina, Charleston (MSG), University of California, Los
Angeles (MAF), and National Institutes of Health Clinical Center, Bethesda,
Maryland (PH).
Address reprint requests to Robert M. Post, M.D., Chief, Biological Psychiatry
Branch, National Institute of Mental Health, Bldg. 10, Room 3N212, 10 Center
Drive, MSC 1272, Bethesda MD 20892-1272.
Received January 10, 2000; revised June 23, 2000; accepted June 29, 2000.

© 2001 Society of Biological Psychiatry

Key Words: PET, cerebral metabolism, bipolar disorders,
depression, rapid cycling

Introduction

A

substantial series of investigations using positron
emission tomography (PET) and single photon emission computed tomography (SPECT) to assess regional
cerebral flow (rCBF) and glucose metabolism (CMRglu)
have most consistently found decreased prefrontal cerebral
activity (hypofrontality) in unipolar depressed patients
compared to control subjects (for a review see Ketter et al
1996b).
A small number of studies in bipolar depressed patients
have also found decreased prefrontal CMRglu (Baxter et
al 1989; Buchsbaum et al 1984, 1986; Cohen et al 1989,
1992; Goyer et al 1992; Martinot et al 1990; Schwartz et
al 1987) and CBF (Ebert et al 1993; Ketter et al 1996b)
compared to control subjects; however, some studies
indicated that certain prefrontal and temporal areas had
increased CMRglu (Cohen et al 1992; Goyer et al 1992;
Ketter et al 1994). Global CMRglu in depressed bipolar
disorder patients has commonly been decreased (Baxter et

al 1985, 1989; Cohen et al 1992; Goyer et al 1992;
Martinot et al 1990; Schwartz et al 1987), but has also
been noted to be increased (Buchsbaum et al 1986). This
heterogeneity could be related to methodologic differences, illness subtypes, state-trait differences, or degree of
responsiveness to treatment (Ketter et al 1999).
In a recent study, Drevets and associates (Drevets et al
1997) found decreased rCMRglu and rCBF in the prefrontal cortex ventral to the genu of the corpus callosum
(subgenual prefrontal) in both familial bipolar depressives
and familial unipolar depressives. Decreased activity was
accompanied by corresponding substantial reductions in
gray matter volume in this area of 39% in unipolar and
48% in bipolar patients. Decreased CMRglu in the temporal lobes (Goyer et al 1992; Ketter et al 1996b; Martinot
et al 1990; Post et al 1987) and basal ganglia (Buchsbaum
et al 1986; Cohen et al 1989) has also been reported.
0006-3223/01/$20.00
PII S0006-3223(00)00975-6

—
—
38.1 6 10.4

Control subjects

Dep, depressed; Mild, mildly depressed; Eu, euthymic; BPI, bipolar I; BPII, bipolar II; HRSD, Hamilton Rating Scale for Depression.
a
p ,.05 vs. mild and euthymic.
b
p ,.05 vs. euthymic.
c
p ,.05 vs. nonrapid. All other subgroup comparisons nonsignificant.

—
—
—
—
—
—
—
—
—
23

37.5 6 10.6
All BP

20

10.6 6 5.5
7.5 6 10.4
18.3 6 10.4
18.8 6 9.9
21.5 6 11.0
8
35
29
14
10
16
17
23

37.9 6 10.7
35.5 6 10.9
Rapid
Nonrapid

20

9.5 6 4.5c
14.4 6 7.2
9.1 6 11.0
1.1 6 1.2
18.7 6 10.2
17.0 6 9.2
19.3 6 11.2
17.0 6 6.7
22.0 6 11.3
19.7 6 10.5
—
8
35

—
25
4
10
4
9
1
12
4
14
3
18
5

34.7 6 12.1
38.8 6 9.8
BPI
BPII

17

3

9.6 6 5.3
11.0 6 5.7
10.7 6 15.2
6.1 6 7.3
17.1 6 13.4
18.9 6 8.2
17.1 6 7.9
19.6 6 11.4
19.9 6 13.6
22.2 6 10.1
4
4
10
25
—
29
14
—

3
7
7
9
4
13
9
14

39.4 6 11.8
37.1 6 11.8
34.9 6 5.8
Depressed
Mildly depressed
Euthymic

5
15

11.3 6 3.7

11.6 6 7.1
7.8 6 3.8
3.7 6 3.9
10.0 6 12.9
10.0 6 12.5
20.1 6 11.9
17.6 6 11.5
18.7 6 5.6
30.1 6 7.3a
14.5 6 3.5b
8.2 6 1.5
17
—
—

—
16
—

—

—
10

4
7
3

13
9
7

14
12
9

3
4
1

Onset
(years)
Nonrapid
Rapid
BPII
Age (years)
Subgroup

F

M

Dep

Mild

Eu

BPI

Cycling
Subtype
Mood
Gender

Table 1. Sample description

The study was approved by the National Institute of Mental
Health (NIMH) Institutional Review Board, and written informed consent was obtained from all subjects before participation. We studied 43 inpatients (age range, 21– 65 years) with
DSM-IV bipolar disorders that were resistant to standard treatments in the community and referred to the NIMH (Table 1).
This highly treatment-resistant group of bipolar disorder patients
had medians of 19 years of illness duration, 16 manic/hypomanic
episodes, 24 depressive episodes, and 4 hospitalizations per
patient. Medical and neurologic histories and physical examinations were obtained to rule out medical illnesses in patients and
in age- and gender-matched healthy control subjects. Both
diagnoses in patients and the absence of mental illness in 43 paid
healthy control subjects (age range, 20 – 65 years) were confirmed by the Schedule for Affective Disorders and Schizophrenia (Endicott and Spitzer 1978). Episodes and hospitalizations as
well as the presence of a current rapid cycling pattern (four or
more affective episodes in the year before the PET scan) were
determined by examination of retrospective lifecharts assessed
by the NIMH-Life Chart Methodology (NIMH-LCM) (Leverich
and Post 1996, 1998).
Trained clinical research nurses blind to medication status and
scans obtained twice-daily global Bunney-Hamburg depression
(BH-D) (Bunney and Hamburg 1963), weekly 28-item Hamilton
Rating Scale for Depression (HRSD) (Hamilton 1960), Mood
Analog, and global Bunney-Hamburg mood ratings, and in 30
patients day-of-scan HRSD and Young Mania Rating Scale
(Young et al 1978) ratings. Self-rated Beck Depression Inventory
(Beck et al 1961) scores were also obtained weekly.
The patient sample was subdivided using three clinical phenomenologic parameters: mood state (depressed, mildly depressed, euthymic), bipolar subtype (bipolar I, bipolar II), and
rapid-cycling status (currently rapid cycling, non–rapid cycling).
No patient had a full acute manic syndrome (mean Young Mania
Rating Scale score 3.1 6 3.2, range 0 –13). Patients were
classified as euthymic if they had no or only mild symptoms on
prospective NIMH-LCM and thus no functional impairment,
HRSD scores of 10 or less, and BH-D scores of 3 or less. Those
with HRSD scores of 18 or greater and BH-D scores of 7 or
greater were considered depressed, and those with HRSD scores
ranging from 11 to 17 and BH-D scores ranging from 4 to 6 were
classified as mildly depressed. The four instances of disagreement between HRSD and BH-D were dealt with by referring to

HRSD

Subjects, Medications, and Mood Ratings

9
9
5

Methods and Materials

18.8 6 12.3
19.5 6 9.1
15.6 6 6.5

Duration
(years)

Episodes
(per year)

No. medications
failed

In view of these varying findings in bipolar disorders,
we obtained fluorine-18-deoxyglucose (18FDG) PET scans
in a substantial series of medication-free bipolar disorder
patients compared with age- and gender-matched control
subjects while performing an auditory continuous performance task, to better understand how changes in prefrontal
and anterior paralimbic rCMRglu relate to clinical parameters, such as mood state, degree of depression, disorder
subtype, and cycle frequency.

—

T.A. Ketter et al

BIOL PSYCHIATRY
2001;49:97–109

8
7
5

98

Cerebral Metabolism in Bipolar Disorders

BIOL PSYCHIATRY
2001;49:97–109

99

prospective NIMH-LCM, and HRSD and BH-D ratings for the
periods immediately before and after scans to resolve the mood
state determination. The mood state (depressed vs. mildly depressed vs. euthymic), bipolar subtype (bipolar I vs. bipolar II),
and rapid-cycling status (currently rapid cycling vs. non–rapid
cycling) subgroups did not differ significantly with respect to
age, gender, or subgroup compositions with respect to one
another (Table 1).
Patients and healthy control subjects had taken no medications
for at least 2 weeks before positron emission tomography (PET)
scans. Control subject and patient PET scans were acquired in an
interspersed fashion over the duration of the study.

Scanning Procedure
We used a Scanditronix (Uppsala, Sweden) PC1024-7B tomograph and (18FDG) to assess absolute and normalized CMRglu in
subjects performing an auditory discrimination continuous performance task (CPT) with eyes covered for 30 min as previously
described (Ketter et al 1999). Radial arterial cannulation was
performed on all subjects to allow quantification of absolute
CMRglu. Serial arterial blood samples were obtained so that a
time-activity curve for the determination of absolute rCMRglu
could be constructed. The conversion of image pixel values from
nanocuries per cubic centimeter to milligrams of glucose per
hundred grams of tissue per minute (mg/100 3 g/min) was
performed as described elsewhere (Kumar et al 1992).

Data Analysis
Image processing was performed using Statistical Parametric
Mapping (SPM 95, courtesy of Medical Research Council
Cyclotron Unit, Hammersmith Hospital, London UK) software
(Friston et al 1995). Stereotactic normalization was performed on
the images as previously described (Ketter et al 1999). Wholebrain mean (WBM) activity was calculated for each stereotactically normalized image as the mean of all voxels greater than the
background threshold, which was set at 1/8 of the mean of the
entire image space, to approximately include gray and white
matter only. For use in global analyses, estimated gray matter
mean activity was calculated as the mean of all voxels above
70% of each subject’s WBM. Normalized rCMRglu analyses
used a voxel-based analysis of covariance (Friston et al 1990),
which generated in each voxel the residual of the subject’s
rCMRglu from a regression of rCMRglu on WBM across all
subjects. Statistically significant focal interscan differences in
rCMRglu residuals, assumed to be independent of intersubject
global CMRglu changes, were then calculated. Absolute analyses
were also performed, without any adjustment to normalize global
CMRglu.
Absolute and normalized rCMRglu were compared between
groups of bipolar disorder patients and age- and gender-matched
healthy control subjects, and between subgroups of patients. The
latter analyses were covaried for age and gender to remove
possible effects of demographic differences between patient
subgroups. Correlations between rCMRglu and clinical measures
were also assessed, covarying for age and gender, and, when
appropriate, mood. Results were displayed as transverse statisti-

Figure 1. Global cerebral metabolism (CMRglu) in bipolar
disorder subgroups compared with healthy control subjects. Dep,
depression; BPI, bipolar I; BPII, bipolar II. *p , .05. 6, p , .08.
cal parametric maps (SPMs) with significant voxels color coded
to correspond to decreasing p-values (starting at p 5 .05, not
initially corrected for multiple comparisons). Correction for
multiple comparisons was applied by only displaying voxels
within clusters deemed significant by particle analysis using a
Z-threshold of 1.96 and cluster probability ,0.05 (Friston et al
1994).
Demographic and clinical patient data were analyzed using
unpaired t-tests for between-group comparisons. Categorical data
were analyzed by x2 tests. Means (6 standard deviations) are
reported, and significance thresholds were set at p 5 .05, unless
noted otherwise.

Results
Among subjects with complete CPT data, 29 patients
compared to 34 control subjects had slower mean reaction
time [0.90 6 0.18 vs. 0.75 6 0.21 sec, t(61) 5 2.98, p 5
.004], but did not otherwise differ significantly on CPT
performance parameters.

Mood State Subgroups
The 17 depressed bipolar patients with HRSD scores
greater than 17, compared to 17 age- and gender-matched
control subjects had 7.7% lower global metabolism
[7.15 6 0.89 vs. 7.75 6 0.80, t(32) 5 2.07, p , .05;
Figure 1]. Decreased absolute (but not normalized) regional metabolism was seen in widespread prefrontal and
anterior paralimbic cortical regions, including inferior
(Brodmann areas [BAs] 44, 45, 46), middle (BAs 8, 9, 10,
46), and superior (BAs 9, 10) frontal gyri; superior (BAs
22, 39, 40, 42), middle (BAs 21, 39, 40), and transverse
(BA 41) temporal gyri; as well as in inferior parietal lobule
(BAs 39, 40), and bilateral lateral cerebellum (Figure 2,
top, and Table 2). Decreased absolute (and to a lesser

100

BIOL PSYCHIATRY
2001;49:97–109

T.A. Ketter et al

Cerebral Metabolism in Bipolar Disorders

Figure 2. Regional cerebral metabolism (rCMRglu) in depressed
bipolar disorder patients compared with healthy control subjects. Z
maps of differences in absolute and normalized cerebral metabolism
in 17 depressed bipolar disorder patients compared to 17 healthy
control subjects. Legend indicates two-tailed p values. Numbers in
upper right corners indicate distances from the intercommissural
plane. Absolute prefrontal and anterior paralimbic cortical metabolic
decreases and normalized anterior paralimbic subcortical metabolic
increases evident in these images may be a state marker for
depression in bipolar disorders. Ham-D, Hamilton Rating Scale for
Depression; L, left.
4™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™

degree normalized) metabolism was also seen in fascicular
(BA 19) and lingular (BAs 18, 19) gyri (Figure 2 and
Table 2).
In contrast, increased normalized (but not absolute)
metabolism was noted in subcortical regions, including
right amygdala, bilateral accumbens, and ventral caudate
and putamen, as well as in right anterior superior longitudinal fasciculus, bilateral medioposterior thalamus (dorsomedial nucleus, pulvinar), and bilateral medial cerebellum
(Figure 2, bottom). Thus, in depressed bipolar disorder
patients there were three main components to the pattern
of differences compared to control subjects, namely 1)
widespread prefrontal and anterior paralimbic cortical
absolute decreases; 2) subcortical (ventral striatum, thalamus, right amygdala) normalized increases; and 3) cerebellar and posterior cortical normalized increases.
The 16 mildly depressed patients with HRSD scores
ranging from 11 to 17, compared with 16 control subjects,
did not differ significantly in global metabolism (7.31 6
0.94 vs. 7.37 6 1.11; Figure 1) or in the absolute regional
analysis. Increased normalized (but not absolute) metabolism was noted in bilateral cerebellum (sparsely), lingular
gyrus (BAs 18, 19), cuneus (BAs 18, 31), and hippocampus; left postcentral gyrus (BAs 1, 2, 3); left insula and left
transverse temporal gyrus (BA 41); and left inferior (BAs
44, 45, 47), middle (BAs 9, 10, 46), and superior (BAs 8,
9) frontal gyri. In contrast, decreased normalized (but not
absolute) metabolism was noted in right inferior (BAs 20,
37) and middle (BAs 20, 21) temporal gyri (Table 2).
The 10 euthymic patients, compared with 10 control
subjects, did not differ significantly in global metabolism
(7.73 6 1.21 vs. 7.63 6 1.00; Figure 1) or in the absolute
4™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™
Figure 3. Regional cerebral metabolism (rCMRglu) in euthymic
bipolar disorder patients compared with healthy control subjects.
Differences in absolute and normalized cerebral metabolism in
10 euthymic bipolar disorder patients compared to 10 healthy
control subjects. Normalized cerebello-posterior cortical metabolic increases evident in these images were seen in all bipolar
subgroups compared to control subjects (including in Figure 2)
but not to one another, consistent with being a trait marker.
Ham-D, Hamilton Rating Scale for Depression; L, left.

BIOL PSYCHIATRY
2001;49:97–109

101

regional analysis (Figure 3, top). Increased normalized
(but not absolute) metabolism was noted in bilateral
cerebellum, lingular gyrus (BA 18), cuneus (BAs 18, 19,
31), and precuneus (BA 7); as well as right inferior frontal
gyrus (BA 44), right superior longitudinal fasciculus, right
insula, and right inferior parietal lobule (BA 40; Figure 3,
bottom, and Table 2).
Direct comparison of the 17 depressed versus 10 euthymic patients (covarying for age and gender) revealed no
significant global difference but decreased (lower in depressed) both absolute and normalized metabolism in left
postcentral gyrus (BAs 40, 43), left inferior parietal lobule
(BAs 39, 40), and left supramarginal gyrus (BA 40). Also,
decreased normalized (but not absolute) metabolism was
noted in left precentral (BA 6), left middle frontal (BA 9),
and left inferior frontal (BA 44) gyri; bilateral lingular
gyrus (BA 18), and cuneus (BA 18). In contrast, increased
(higher in depressed) normalized (but not absolute) metabolism was seen in right inferior (BAs 45, 46, 47),
middle (BAs 8, 9, 10, 46), and superior (BA 9) frontal
gyri, bilateral accumbens, and ventral caudate and putamen (Table 3).
Direct comparison of the 17 depressed versus 16 mildly
depressed patients and of the of the 16 mildly depressed
versus 10 euthymic patients (covarying for age and gender) revealed no global or cerebellar differences but some
regional differences (Table 3).
Thus, the metabolic difference patterns obtained from
comparing depressed with both euthymic and mildly
depressed patients appeared to be attenuated forms of the
pattern seen comparing depressed patients with healthy
control subjects, in that they included two main components, namely 1) cortical absolute decreases, and 2)
paralimbic subcortical normalized increases, but they
lacked the cerebellar, posterior cortical, and posterior
thalamic normalized increases seen when comparing depressed patients to healthy control subjects.

Severity of Depression Correlation
On the day of the scan, HRSD (mean 6 SD 21.5 6 11.0;
n 5 30) covarying for age and gender was not significantly correlated with global metabolism but correlated
inversely with absolute (but not normalized) metabolism
in prefrontal and anterior paralimbic cortical regions,
including left inferior (BAs 44, 45) and middle (BA 9)
frontal gyri, left temporal pole (BA 22), and left middle
temporal gyrus (BA 21), as well as left hippocampus
(Figure 4). On the day of the scan, HRSD also correlated
inversely with absolute (and to a lesser extent normalized)
metabolism in left inferior parietal lobule (BAs 39, 40),
bilateral cuneus (BAs 17, 18), fascicular (BAs 19, 20) and
lingular (BAs 18, 19) gyri, and right posteromedial
cerebellum.

102

T.A. Ketter et al

BIOL PSYCHIATRY
2001;49:97–109

Table 2. Summary of rCMRglu in Bipolar Subgroups Compared to Healthy Control Subjects
Lingular
gyrus

Cuneus

Thalamus
(DM, Pulv)

Basal ganglia

Temporal lobe

Frontal
lobe

Other

1;2a
1
1

—
1
1

1
—
—

1NA,Cd,Pu
—
—

1R Am;2 Cxa
1Hi,L TTG;2R Cx
—

2 Cxa
1L Cx
1R IFG

2 IPLa
1L Ins
1R Ins, R IPL

1;1a
6 1L

1
1L

1
1L

—
1

—
—

—
2 R TPa

—
2 R Cxa

—
2 R Insa

35 Rapid
8 Nonrapid

1
1

1
—

1
—

1
—

—
—

—
1R Hi

—
—

—
—

43 All BP

1

1

1

1

—

1L TP, L Cx

1L IFG

—

Subgroup
17 BPdep
16 BPmild
10 BPeu

Cerebellum
1Med; 2 Lata
61
1

14 BPI
29 BPII

All data globally normalized except aabsolute. Note that all patient subgroups compared to control subjects have cerebello-posterior cortical normalized increases. DM,
dorsomedial nucleus; Pulv, pulvinar; Med, medial; Lat, lateral; BPdep, BPmild, BPeu, bipolar depressed, mildly depressed, euthymic; NA, accumbens; Cd, caudate; Pu, putamen;
R, right; L, left; Am, amygdala; Cx, cortex; IPL, inferior parietal lobe; Hi, hippocampus; TTG, transverse temporal gyrus; Ins, insula; IFG, inferior frontal gyrus; TP, temporal pole.

Conversely, degree of depression correlated positively
with normalized metabolism in subcortical anterior paralimbic regions, including basal forebrain and ventral caudate and putamen, as well as in subgenual and pregenual
anterior cingulate (BAs 24, 32), medial (BAs 10, 11) and
right inferior (BAs 45, 46, 47) frontal gyri, and right
middle (BAs 20, 21) and inferior (BA 20) temporal gyri.
The overall pattern thus largely overlapped the prefrontal
and anterior paralimbic cortical absolute decreases and
subcortical normalized increases seen in depressed patients compared to control subjects.

Bipolar I and Bipolar II Subgroups
The 14 bipolar I patients compared with 14 healthy control
subjects did not significantly differ in global metabolism
(7.35 6 0.84 vs. 7.16 6 1.13; Figure 1) and had increased
normalized metabolism in bilateral cerebellum (cerebellum also had increased absolute metabolism), lingular
gyrus (BAs 18, 19), and cuneus (BAs 29, 30, 31; Table 2).
The 29 bipolar II patients compared with 29 healthy
control subjects tended to have 5.8% lower global metabolism [7.34 6 1.07 vs. 7.78 6 0.64, t(56) 5 1.78, p , .08;
Figure 1], and had decreased absolute (but not normalized)
metabolism in right lateral prefrontal cortex, including

superior (BAs 9, 10), middle (BAs 9, 10, 46), and inferior
(BAs 44, 45, 46) frontal gyri, as well as right insula and
right temporal pole. In contrast, increased normalized (but
not absolute) metabolism was noted in left cerebellum
(sparse), left lingular gyrus (BAs 18, 19), left cuneus (BAs
29, 30, 31), and bilateral medioposterior thalamus (dorsomedial nucleus, pulvinar; Table 2). This pattern appeared
to be an attenuated form of the pattern seen comparing
depressed patients to healthy control subjects in that it
included two main components, namely 1) anterior paralimbic and prefrontal cortical absolute decreases, and 2)
cerebellar, posterior cortical, and posterior thalamic normalized increases.
Direct comparison of the 14 bipolar I with the 29
bipolar II patients (covarying for age and gender) lacked
significant differences in global or absolute regional metabolism but showed increased (higher in bipolar I) normalized metabolism in supragenual anterior cingulate (BA
32), right middle frontal gyrus (BA 9), and right inferior
parietal lobule (BA 40) in the bipolar I patients (Table 3).
Thus, comparison of bipolar I and II subgroups with one
another lacked the cerebello-posterior cortical normalized
metabolic increases seen when comparing these patient
subgroups to healthy control subjects.

Table 3. Summary of rCMRglu in Bipolar Subgroups Compared to One Another
Cerebellum

Lingular
gyrus

Cuneus

Thalamus
(DM, Pulv)

Basal
ganglia

Temporal
lobe

Frontal
lobe

17 BPdep vs. 10 BPeu

—

2

2

—

1NA,Cd,Pu

—

2 L IPL;2 L IPLa

17 BPdep vs. 16 BPmild

—

2;2a

2

—

1L Cd, L Pu

2L Hi, L TP

16 Bpmild vs. 10 BPeu

—

—

—

—

—

—

2L MiFG, L IFG
1R Cx
1R IFG,
1 R MeFG, AC
1R MeFG,AC

14 BPI vs. 29 BPII

—

—

—

—

—

—

1R MiFG, AC

1R IPL

35 Rapid vs. 8 nonrapid

—

—

—

—

—

1MeFG, AC

—

Subgroups

Other

2L Ins
—

Normalized data except a absolute. Note that no patient subgroups compared to one another have cerebellar differences. DM, dorsomedial nucleus; BPdep, BPeu, BPmild,
bipolar depressed, euthymic, mildly depressed; NA, accumbens; Cd, caudate; Pu, putamen; L, left; R, right, MiFG, middle frontal gyrus; IFG, inferior frontal gyrus; IPL,
inferior parietal lobule; Cx, cortex; Hi, hippocampus; TP, temporal pole; Ins, insula; MeFG, medial frontal gyrus; IPL, inferior parietal lobule.

Cerebral Metabolism in Bipolar Disorders

BIOL PSYCHIATRY
2001;49:97–109

103

Figure 4. Correlation between degree of depression and regional cerebral metabolism (rCMRglu). Pearson r maps of the relationship
between degree of depression and baseline absolute and normalized regional cerebral metabolism in 30 bipolar disorder patients. Color
bar indicates two-tailed p thresholds. This pattern of absolute cortical negative and normalized subcortical positive correlations overlaps
that of Figure 2, excluding normalized cerebello-posterior cortical increases, which may be a trait marker. HamD, Hamilton Rating
Scale for Depression; L, left.

Rapid-Cycling and Non–Rapid-Cycling Subgroups
The 35 rapid cycling patients, compared with 35 control
subjects, had nonsignificantly (3.6%) lower global metabolism (7.26 6 0.92 vs. 7.53 6 1.03; Figure 1) and no
differences in absolute but increased normalized (but not
absolute) metabolism in bilateral cerebellum, lingular
gyrus (BAs 18, 19), cuneus (BAs 29, 30, 31), and
medioposterior thalamus (dorsomedial nucleus, pulvinar;
Table 2).
The eight non–rapid-cycling patients, compared with
eight control subjects, had similar global metabolism
(7.70 6 1.26 vs. 7.78 6 0.72; Figure 1) and no differences
in absolute but increased normalized (but not absolute)
metabolism in bilateral cerebellum and right hippocampus
(Table 2).
Direct comparison of the 35 rapid cycling and eight
non–rapid-cycling patients (covarying for age and gender)

revealed no significant global difference but increased
(higher in rapid cycling) normalized (but not absolute)
metabolism in right greater than left medial frontal gyrus
(BAs 8, 9, 10), and supragenual and pregenual anterior
cingulate (BAs 24, 32; Table 3). Thus, comparison of
rapid and non–rapid-cycling subgroups with one another
lacked the cerebello-posterior cortical normalized metabolic increases seen when comparing these patient subgroups to healthy control subjects.

All Patients
The heterogeneous group of all 43 bipolar patients, compared with 43 healthy control subjects, had nonsignificantly (3.2%) lower global metabolism (7.34 6 0.99 vs.
7.58 6 0.96; Figure 1) and no significant absolute rCMRglu differences but increased normalized metabolism in
bilateral cerebellum, lingular gyrus (BAs 18, 19), cuneus

104

BIOL PSYCHIATRY
2001;49:97–109

(BAs 29, 30, 31), and medioposterior thalamus (dorsomedial nucleus, pulvinar), as well as left temporal pole (BA
38), left middle (BA 21) and left inferior (BA 20) temporal
gyri, and left inferior frontal gyrus (BAs 45, 46, 47; Table
2). Thus, comparison of all patients to healthy control
subjects yielded cerebello-posterior cortical normalized
metabolic increases seen with all comparisons of patient
subgroups with healthy control subjects, but not with
comparisons of patients subgroups with one another.

Discussion
We found that bipolar disorder patients in the depressed
phase compared to healthy control subjects (and to a lesser
extent compared to mildly depressed or euthymic patients)
had decreased global and prefrontal and anterior paralimbic cortical absolute (but not normalized) metabolism, and
increased normalized metabolism in subcortical structures,
including ventral striatum, thalamus, and right amygdala.
This pattern was also evident in bipolar II patients compared to control subjects (Table 2, right) and in the degree
of depression correlation but not in other patient subgroup
versus control subject (Table 2, right) or patient versus
patient subgroup (Table 3, right) comparisons, suggesting
some specificity of this pattern to the depressed state.
Euthymic patients compared to control subjects had
increased cerebello-posterior cortical (cerebellum, lingular
gyrus, cuneus) normalized metabolism, with similar patterns (in some comparisons also involving thalamic increases) present in all patient subgroup versus control
subject comparisons (Table 2, left) but not in patient
versus patient subgroup comparisons (Table 3, left) or the
degree of depression correlation. This is consistent with
the notion that such cerebello-posterior cortical relative
metabolic increases could represent a trait marker, which
might be congenital, or a “scar” acquired over the course
of many episodes. Unipolar patients may display this pattern
in the depressed but not euthymic state, more consistent with
a state marker (Kimbrell et al, unpublished data).
Our finding that global cerebral metabolism in bipolar
disorder patients is decreased when depressed, but not
when euthymic, is consistent with prior reports. Global
cerebral metabolism was decreased in six reports of
bipolar depression (Baxter et al 1985, 1989; Cohen et al
1992; Goyer et al 1992; Martinot et al 1990; Schwartz et
al 1987), and in four studies (Kishimoto et al 1987; Kumar
et al 1993; Raichle et al 1985; Schlegel et al 1989) of
unipolar depression. One study reported increased global
metabolism in bipolar depressed patients (Buchsbaum et al
1986). Also, global cerebral metabolism in hypomanic or
euthymic bipolar patients did not differ from healthy
control subjects but was increased compared to bipolar
patients in depressed or mixed states (Schwartz et al

T.A. Ketter et al

1987). In rapid-cycling bipolar patients, global cerebral
metabolism (Baxter et al 1985) and blood flow (Speer et al
1997) oscillated as mood state changed. Taken together,
these data are consistent with global cerebral metabolism
being decreased in bipolar depression and varying across
mood states; however, many factors could contribute to
such a phenomenon. For example, plasma thyroid stimulating hormone levels may correlate inversely with global
cerebral blood flow and metabolism in mood disorder
patients (Marangell et al 1997).
The observed decrease in global cerebral metabolism in
bipolar depression could confound findings in the normalized analysis, particularly for the subcortical normalized
increases, which could merely represent areas that failed to
be decreased to the same extent as the global decrease.
This is less of a concern for the finding of increased
normalized cerebello-posterior cerebral metabolic increases, as these were evident in other subgroup comparisons in which there was no significant global difference
between patients and control subjects; however, even in
the absence of global differences, increased normalized
posterior activity in patients could be related to decreased
anterior–posterior gradients.
The metabolic pattern we observed in depressed bipolar
disorder patients compared to control subjects (and to a
lesser extent compared to mildly depressed or euthymic
patients, and in the correlation with degree of depression)
of decreased prefrontal and anterior paralimbic cortical
absolute metabolism, and increased subcortical (ventral
striatum, thalamus, right amygdala) normalized metabolism implicates some of the classic neural substrates
hypothesized to be involved in affective modulation in
general, and bipolar illness in particular. Dysregulated
function of basal ganglia– corticothalamic loops as described by Alexander and colleagues (Alexander et al
1986, 1990) could account for such complementary absolute cortical decreases and relative subcortical increases.
Activation of the right amygdala has been associated with
affective arousal, with increased activity occurring with
both euphoric and dysphoric affects, as opposed to the left
amygdala, in which activity tends to increase with sadness
or fear and decrease with happiness or euphoria (George et
al 1995; Ketter et al 1996a). In bipolar depression, we
have thus observed relative activation of anterior paralimbic subcortical regions classically linked to affective
modulation (Ketter et al 1996b), such as amygdala, accumbens, and ventral striatum.
We also observed relative activation of bilateral medioposterior thalamus. This is consistent with altered thalamic
relay and gating function with respect to communication
between subcortical and cortical regions. We noted increased basal ganglia normalized activity, similar to that
seen in obsessive-compulsive disorder (Baxter et al 1987),

Cerebral Metabolism in Bipolar Disorders

a condition often considered more treatment-resistant than
depressive disorders; however, decreased normalized
basal ganglia and thalamic activity has been reported in
some studies of bipolar disorder patients (Baxter et al
1985; Buchsbaum et al 1986, 1997; Cohen et al 1989).
Variability in findings could be due to methodological
differences, such as normalizing activity to posterior
structures (Buchsbaum et al 1986), partial volume effects
in view of the small size of these structures compared to
cortical regions, the type of task performed during scan
acquisition, or the highly treatment-resistant nature of our
patient sample.
In contrast, depressed bipolar patients had absolute
deactivation of prefrontal and anterior paralimbic cortical
regions, including inferior, middle, superior, and medial
frontal gyri, anterior and midcingulate, and superior,
middle, and transverse temporal gyri. This is consistent
with reports of decreased activity in such regions not only
in bipolar depression (Baxter et al 1989; Buchsbaum et al
1984, 1986, 1997; Cohen et al 1989, 1992; Drevets et al
1997; Ebert et al 1993; Goyer et al 1992; Ketter et al
1996b; Martinot et al 1990; Post et al 1987; Schwartz et al
1987) but also in many studies of unipolar depression
(Austin et al 1992; Baxter et al 1989; Bench et al 1992;
Biver et al 1994; Curran et al 1993; Drevets et al 1997;
Ebert et al 1991; Edmonstone et al 1994; Hagman et al
1990; Hurwitz et al 1990; Kanaya and Yonekawa 1990;
Kimbrell et al, unpublished data; Kishimoto et al 1987;
Kuhl et al 1985; Kumar et al 1993; Lesser et al 1994;
Mayberg et al 1994; O’Connell et al 1989; Philpot et al
1993; Schlegel et al 1989; Upadhyaya et al 1990; Yazici et
al 1992) and secondary depression (Bromfield et al 1992;
Caparros-Lefebvre et al 1996; Mayberg et al 1990, 1991,
1992; Renshaw et al 1992; Ring et al 1994; Volkow et al
1991). This convergence suggests the possibility of a
common pathway to depressive symptoms to some extent
independent of illness etiology (primary vs. secondary)
and subtype (unipolar vs. bipolar). A few studies have
reported increased prefrontal and anterior paralimbic activity, perhaps due to distinctive phenomenologic, pathophysiologic, and treatment response characteristics of
samples. For example, such increases have been noted in
patients with familial pure depressive disease (Drevets et
al 1992) and patients with subsequent antidepressant
responses to sleep deprivation (Ebert et al 1991; Volk et al
1997; Wu et al 1992), fluoxetine (Mayberg et al 1997),
and carbamazepine (Ketter et al 1999).
Depressed bipolar disorder patients also had increased
normalized subcortical paralimbic metabolism, consistent
with a limbic-cortical dysregulation model of depression,
wherein dorsal neocortical hypofunction could lead to
ventral paralimbic overactivity or vice versa (Mayberg
1997).

BIOL PSYCHIATRY
2001;49:97–109

105

Posterior regions, compared to anterior regions, have
been less emphasized in mood disorder models or imaging
studies; however, the consistent increases we observed in
cerebello-posterior cortical normalized metabolism suggest possible relevance to bipolar disorders as a trait
marker, if it can be demonstrated in additional studies that
these normalized posterior increases are not merely secondary to absolute anterior cortical decreases.
The cerebellum, through limbic and brain stem connections, may have a role in mood regulation (Berntson and
Torello 1982; Haines et al 1984; Snider and Maiti 1976).
Case reports and case series have noted affective symptoms in patients with cerebellar pathology (Cutting 1976;
Lauterbach 1996; Schmahmann and Sherman 1998; Starkstein et al 1988; Yadalam et al 1985). Cerebellar electrode
implantation and subsequent chronic cerebellar stimulation for neurologic disorders can yield decreased anxiety
and improved mood (Riklan et al 1977) and has also been
reported helpful in some patients with psychiatric (especially depressive) disorders, although methodologic problems have limited the utility of this approach (Heath et al
1981). Hence, a “cerebellar cognitive affective syndrome”
due to disrupted cerebellar modulation of cerebello-parieto-temporo-limbic-prefrontal circuits has been proposed
(Schmahmann and Sherman 1998).
In healthy volunteers, resting cerebellar (and brainstem)
rCBF correlated with harm-avoidance (George et al 1994),
and lateral cerebellar (and occipito-temporo-parietal) CBF
increases appeared related to emotional responses to exteroceptive sensory (film) stimuli (Reiman et al 1997).
Cerebellar atrophy has been reported in mood disorders.
Thus, in four of seven (DelBello et al 1999, Heath et al
1982, Lippmann et al 1982, and Nasrallah et al 1982 but
not Coffman et al 1990, Dewan et al 1988 or Yates et al
1987) studies of bipolar, two of four (Escalona et al 1993
and Shah et al 1992 but not Pillay et al 1997 or Yates et al
1987) studies of unipolar, and one study (Weinberger et al
1982) of mood disorder patients, vermian or cerebellar
atrophy tended to occur compared to healthy control
subjects. Such changes could confound our functional
cerebellar findings; however, such a confound would be
expected to yield decreased activity due to the decreased
volume, rather than the increased activity which we
observed.
Increased cerebellar vermis (and decreased left anterior
medial prefrontal) rCBF has been seen in depressed
patients with cognitive impairment compared to those
without (Dolan et al 1992), whereas decreased cerebellar
blood volume was reported in mood disorder patients
compared to control subjects (Loeber et al 1999). In
unipolar depression, venlafaxine response was accompanied by decreases in cerebellar (and occipital and anterior
paralimbic) normalized rCBF (Little et al 1997). Manic

106

T.A. Ketter et al

BIOL PSYCHIATRY
2001;49:97–109

compared to euthymic patients had attenuated cognitive
task-related rCBF changes in left cerebellum (and right
orbitofrontal, right brainstem, and right medial thalamus)
(Blumberg et al 1999).
The practice of normalizing regional data to activity in
cerebellum (rather than in hemisphere, slice, or whole
brain) in brain imaging studies of mood disorder patients
(Amsterdam and Mozley 1992; Bonne et al 1996, 1999;
Dube´ et al 1993; Jaracz et al 1996; Kanaya and Yonekawa
1990; Kocmur et al 1998; Mozley et al 1996; Vasile et al
1997) may thus need to be reconsidered. If cerebellar
activity is increased in mood (and particularly in bipolar)
disorder patients compared to healthy control subjects,
then this referencing technique could exaggerate the degree of perceived hypofrontality in patients.
Bipolar II compared to bipolar I patients had somewhat
greater CMRglu deficits. This could be related to multiple
factors, such as bipolar II patients having numerically (but
not statistically) higher proportions of subjects who were
depressed (13/29 [45%] versus 4/14 [29%]) and rapid
cycling (25/29 [86%] vs. 10/14 [71%]), and numerically
(but not statistically) higher HRSD, duration, and number
of medications failed. In addition, comparisons with control subjects involving the 29 bipolar II patients had
greater power than those involving the 14 bipolar I
patients, owing to the larger sample size.
The data presented have a number of important limitations. Our sample is a highly treatment-resistant group of
predominantly rapid-cycling (81%) bipolar disorder patients who were referred to the NIMH, a tertiary center for
clinical research studies, because of resistance to agents
generally utilized in the community, including lithium
and, more recently, carbamazepine and valproate (Frye et
al 2000). As such, the abnormalities outlined could relate
primarily to relatively late-stage, treatment-resistant, rapid-cycling bipolar disorder patients, and determination
whether these same abnormalities are evident earlier in
illness course, in non–treatment-resistant, or in non–rapidcycling patients requires further investigation. In addition,
structural abnormalities in bipolar disorder patients, such
as lateral ventricular enlargement and cerebellar atrophy,
the use of stereotactic and global normalization, and
partial volume effects could have confounded our
findings.
Patients and control subjects were engaged in an auditory continuous performance task, which has the asset of
occupying both patient and healthy control subjects with a
similar set of mental operations, but may distract patients
from negative affects or ruminations that are important
components of the disorder, or introduce spurious metabolic changes in neural substrates not central to affective
modulation or other aspects of cognitive–somatic function
in bipolar illness.

Conversely, the data presented in this article have
certain strengths. We have studied a relatively substantial
number of bipolar disorder patients who have been medication-free for at least 2 weeks and monitored in a highly
controlled environment. In contrast to some other reports,
we assessed not only normalized but also absolute regional
cerebral glucose utilization through the use of arterial
cannulization. Depressed, mildly depressed, and euthymic
patients were included in the analysis, giving preliminary
insights into possible alterations related to both state and
trait phenomena in these treatment-resistant bipolar disorder patients.
In summary, in bipolar depression we found widespread
anterior (prefrontal and temporal) cortical absolute hypometabolism, consistent with previous observations in primary unipolar and secondary depression, suggesting a
relationship to the depressed state, and a common pathway
to depressive symptoms independent of illness etiology
(primary vs. secondary) and subtype (unipolar vs. bipolar).
Moreover, depressed bipolar disorder patients also had
increased normalized subcortical (amygdala, accumbens,
and ventral striatum) metabolism, consistent with a limbic– cortical dysregulation model of depression. In contrast to the findings related to depression, we also observed
normalized cerebello-posterior cortical hypermetabolism
in all (including euthymic) patient subgroups compared to
control subjects, but not in comparisons of patient subgroups with one another, suggesting a possible trait
abnormality for bipolar disorders. In addition, we noted
increased thalamic normalized metabolism in several patient versus control subject but no patient versus patient
subgroup comparisons, suggesting basal ganglia–thalamocortical circuit dysfunction as a possible trait abnormality.
Thus, depressed state and bipolar disorder trait may be the
clinical phenomena with the greatest impact on scan
patterns, appearing more robust than the effects of cycling
pattern or bipolar subtype. Further studies are required to
determine whether these putative state and trait patterns
are replicated in bipolar disorder patients with shorter
duration and less treatment-resistant illness, and using
other behavioral and methodologic paradigms.

Support was received from the Stanley Foundation (TAKe, TAKi,
MWW, AD).

References
Alexander GE, Crutcher MD, DeLong MR (1990): Basal ganglia-thalamocortical circuits: Parallel substrates for motor,
oculomotor, prefrontal and limbic functions. Prog Brain Res
85:119 –146.
Alexander GE, DeLong MR, Strick PL (1986): Parallel organi-

Cerebral Metabolism in Bipolar Disorders

zation of functionally segregated circuits linking basal ganglia and cortex. Annu Rev Neurosci 9:357–381.
Amsterdam JD, Mozley PD (1992): Temporal lobe asymmetry
with iofetamine (IMP) SPECT imaging in patients with major
depression. J Affect Disord 24:43–53.
Austin MP, Dougall N, Ross M, et al (1992): Single photon
emission tomography with 99 mTc-exametazime in major
depression and the pattern of brain activity underlying the
psychotic/neurotic continuum. J Affect Disord 26:31– 43.
Baxter LR Jr, Phelps ME, Mazziotta JC, et al (1985): Cerebral
metabolic rates for glucose in mood disorders. Studies with
positron emission tomography and fluorodeoxyglucose F 18.
Arch Gen Psychiatry 42:441– 447.
Baxter LR Jr, Phelps ME, Mazziotta JC, et al (1987): Local
cerebral glucose metabolic rates in obsessive-compulsive
disorder. A comparison with rates in unipolar depression and
in normal controls. Arch Gen Psychiatry 44:211–218.
Baxter LR Jr, Schwartz JM, Phelps ME, et al (1989): Reduction
of prefrontal cortex glucose metabolism common to three
types of depression. Arch Gen Psychiatry 46:243–250.
Beck AT, Ward CH, Mendelson M, et al (1961): An inventory
for measuring depression. Arch Gen Psychiatry 4:561–571.
Bench CJ, Friston KJ, Brown RG, et al (1992): The anatomy of
melancholia—focal abnormalities of cerebral blood flow in
major depression. Psychol Med 22:607– 615.
Berntson GG, Torello MW (1982): The palleocerebellum and the
integration of behavioral function. Physiol Psychol 10:2–12.
Biver F, Goldman S, Delvenne V, et al (1994): Frontal and
parietal metabolic disturbances in unipolar depression. Biol
Psychiatry 36:381–388.
Blumberg HP, Stern E, Ricketts S, et al (1999): Rostral and
orbital prefrontal cortex dysfunction in the manic state of
bipolar disorder. Am J Psychiatry 156:1986 –1988.
Bonne O, Krausz Y, Aharon Y, et al (1999): Clinical doses of
fluoxetine and cerebral blood flow in healthy volunteers.
Psychopharmacology (Berl) 143:24 –28.
Bonne O, Krausz Y, Gorfine M, et al (1996): Cerebral hypoperfusion in medication resistant, depressed patients assessed by
Tc99 m HMPAO SPECT. J Affect Disord 41:163–171.
Bromfield EB, Altshuler L, Leiderman DB, et al (1992): Cerebral
metabolism and depression in patients with complex partial
seizures. Arch Neurol 49:617– 623.
Buchsbaum MS, Cappelletti J, Ball R, et al (1984): Positron
emission tomographic image measurement in schizophrenia
and affective disorders. Ann Neurol 15:S157–S165.
Buchsbaum MS, Someya T, Wu J, et al (1997): Neuroimaging
bipolar illness with positron emission tomography and magnetic resonance imaging. Psychiatr Ann 27:489 – 495.
Buchsbaum MS, Wu J, DeLisi LE, et al (1986): Frontal cortex
and basal ganglia metabolic rates assessed by positron emission tomography with [18F]2-deoxyglucose in affective illness. J Affect Disord 10:137–152.
Bunney WE, Hamburg DA (1963): Methods for reliable longitudinal observation of behavior. Arch Gen Psychiatry 9:280 –294.
Caparros-Lefebvre D, Girard-Buttaz I, Reboul S, et al (1996):
Cognitive and psychiatric impairment in herpes simplex virus
encephalitis suggest involvement of the amygdalo-frontal
pathways. J Neurol 243:248 –256.

BIOL PSYCHIATRY
2001;49:97–109

107

Coffman JA, Bornstein RA, Olson SC, et al (1990): Cognitive
impairment and cerebral structure by MRI in bipolar disorder.
Biol Psychiatry 27:1188 –1196.
Cohen RM, Gross M, Nordahl TE, et al (1992): Preliminary data
on the metabolic brain pattern of patients with winter seasonal
affective disorder. Arch Gen Psychiatry 49:545–552.
Cohen RM, Semple WE, Gross M, et al (1989): Evidence for
common alterations in cerebral glucose metabolism in major
affective disorders and schizophrenia. Neuropsychopharmacology 2:241–254.
Curran SM, Murray CM, Van Beck M, et al (1993): A single
photon emission computerised tomography study of regional
brain function in elderly patients with major depression and
with Alzheimer-type dementia. Br J Psychiatry 163:155–165.
Cutting JC (1976): Chronic mania in childhood: Case report of a
possible association with a radiological picture of cerebellar
disease. Psychol Med 6:635– 642.
DelBello MP, Strakowski SM, Zimmerman ME, et al (1999):
MRI analysis of the cerebellum in bipolar disorder: A pilot
study. Neuropsychopharmacology 21:63– 68.
Dewan MJ, Haldipur CV, Lane EE, et al (1988): Bipolar
affective disorder. I. Comprehensive quantitative computed
tomography. Acta Psychiatr Scand 77:670 – 676.
Dolan RJ, Bench CJ, Brown RG, et al (1992): Regional cerebral
blood flow abnormalities in depressed patients with cognitive
impairment. J Neurol Neurosurg Psychiatry 55:768 –773.
Drevets WC, Price JL, Simpson JR Jr, et al (1997): Subgenual
prefrontal cortex abnormalities in mood disorders. Nature
386:824 – 827.
Drevets WC, Videen TO