The study protocol was approved by the Institutional Review Board at UCLA and the VA Greater Los Angeles Healthcare System. Each participant gave written informed consent. Participants with bipolar I disorder were recruited through the UCLA Mood Disorders Clinic, the bipolar Disorders Clinic of the Veterans Affairs Greater Los Angeles Health Care System, and inpatient units of both hospitals. Participants enrolled in other research projects of the UCLA Mood Disorders Research Program were also invited to participate. Patients were evaluated using the Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders-IV to confirm an accurate diagnosis. Illness duration and medication information were obtained by patient self-report and from medical records when available. Exclusion criteria for all participants included other Axis I comorbidities, left handedness, hypertension, neurologic illness, metal implants and a history of skull fracture or head trauma with loss of consciousness for more than 5 min. Participants were designated as lithium-treated if the participant was treated with lithium at the time of scanning, and lithium-free if the participant was not taking lithium at the time of scanning.
A total of 49 scans were obtained (37 lithium-free, 12 lithium-treated). Lithium-free participants (37.5±10.7 years; 57% women) had been ill for 17.8±12.8 years, had a prior history of 5.2±4.7 manic episodes and 7.6±12.5 depressive episodes. Forty-six percent of these participants were euthymic, 32% were depressed and 22% were manic at the time of scanning and were taking one or more of the following medications: antipsychotics (n=4), selective serotonin reuptake inhibitors (n=12), benzodiazepines (n=4), anticonvulsants (n=25). Lithium-treated participants (42.0±9.1 years; 8% women) were ill for 15.0±11.9 years, and had a prior history of 8.3±8.8 manic episodes and 6.3±4.5 depressive episodes. Fifty-eight percent of these participants were euthymic, 17% were depressed, and 25% were manic at the time of scanning and were taking one or more of the following medications: benzodiazepines (n=1), anticonvulsants (n=4). The two groups did not differ significantly in terms of age (P=0.19), illness duration (P=0.51), prior number of manias (P=0.18) and prior number of depressions (P=0.77).
Structural neuroimaging was performed on a 3-T MRI scanner (General Electric, Waukesha, Wisconsin, USA). A 3-plane gradient echo scan was acquired for alignment and localization, followed by a shimming procedure to improve magnetic field homogeneity. A T-1 weighted three-dimensional volume scan (SPGR, T1 500/TE 3.7/FOV 20/1 NEX) was obtained for each participant that spanned the entire brain, with a slice thickness of 1.2 mm. Precautions were taken to minimize participant motion during scanning.
Automated extraction of brain tissue was performed using the Brainsuite software package [11
], following which manual corrections were made by an image analyst blind to participant characteristics. All scans were then corrected for intensity inhomogeneity [12
] and spatially smoothed using a 1.5-mm full-width at half-maximum Gaussian kernel. Each brain was then linearly transformed to the International Consortium for Brain Mapping-27 brain template using a 9-parameter transformation matrix to maximize the mutual information between the participant’s image and the standard template.
TBM was used to measure differences in amygdalar and hippocampal volumes between groups. This technique derives regional brain volumes from the amount of warping needed to match each patient’s subcortical structures to that of a common template [13
]. This template was the average of 27 T1-weighted MRI acquisitions from a single healthy participant, aligned within the stereotaxic space of the International Consortium for Brain Mapping [14
]. Rather than using a multiparticipant average as a template in this deformation, we preferred registration to a single participant’s image given the improved contrast and better spatial resolution. Template optimization for TBM, however, is the subject of further on-going study [15
]. For each individual, this warping was encoded in a deformation map with expansion or contraction encoded at every voxel (sometimes called the ‘Jacobian determinant’). Values less than 1 indicated volume decreases in the patient relative to the template, whereas values greater than 1 indicated volume increases. Overall scaling of these deformation maps was removed and retained for use as a covariate in subsequent statistical analyses to control for total brain volume.
Using the deformation maps, group differences in the volume of the amygdala and hippocampus were obtained by first averaging across point values contained within the amygdala and hippocampus, manually defined on the template brain by a trained neuroanatomist. Regional averages were then entered into a second-level statistical analysis in SAS (PROC mixed SAS, Version 8.1, SAS Institute Inc., Cary, North Carolina, USA) to predict volume difference as a function of lithium status, controlling for age, sex, mood state and total brain scale. As nine patients were scanned in two different mood states, within-subject variability was accounted for using a mixed-effects model, with subject modeled as a random effect, nested within treatment group. Treatment was modeled as a between-subjects factor.
Spearman rank correlation was used to examine the effects of specific clinical variables on the subcortical volumes across patients within lithium-treated and lithium-free bipolar subgroups.