This protocol was approved by the Stanford University Panel of Medical Research in Human Subjects. Twenty-two patients and twenty healthy volunteers were recruited from an ongoing study of BD offspring and from the community. Patients were included consecutively if they met inclusion criteria. Inclusion criteria for high-risk subjects were age 9 – 18 years, a biological parent with bipolar I or II disorder, and a diagnosis of “high-risk” for BD, as defined below. Exclusion criteria were presence of a pervasive development disorder (such as Autism or Aspergers Disorder), a neurological condition (such as a seizure disorder), a substance use disorder, IQ less than 80, or presence of metallic implants or orthodontic braces, which would make the MRI scan not feasible.
Oral and written consent from the parents as well as oral and written assent from the youth were obtained, and both the parents and the offspring were interviewed. At least one parent had BD I or II diagnosed by the Structured Clinical Interview for DSM-IV Axis I disorders (SCID) (First MB 1995
), administered by a trained master's degree level clinician and/or a psychiatrist board certified in general psychiatry. Both parents were interviewed, and on occasion, the non-bipolar parent had an Axis I diagnosis within the high-risk group. For inclusion in the high-risk BD group, in addition to parental diagnosis of BD, all children met criteria for ADHD and had at least moderate mood symptoms, as indicated by a score of >10 on the Young Mania Rating Scale (YMRS, (Fristad et al., 1995
)) or a score of >30 on the Children's Depressive Rating Scale – Revised (CDRS-R, (Poznanski et al., 1985
)). Subjects could have depression or dysthymia, but not cyclothymia, bipolar I, or bipolar II disorder. Subjects were allowed to meet bipolar disorder, not otherwise specified (BD-NOS) criteria, defined as having either one less criterion B symptom than necessary for a (hypo)-manic episode, or enough criterion B symptoms but only a 2–3 day duration of the episode. All subjects, patients and healthy volunteers, were evaluated by the affective disorders module of the Washington University in St. Louis Kiddie Schedule for Affective Disorders and Schizophrenia (WASH-U-KSADS) (Geller et al., 1996
; Geller et al., 2001
) and the Schedule for Affective Disorders and Schizophrenia for School-Age Children, Present and Lifetime (K-SADS-PL) (Kaufman et al., 1997
). Researchers with at least a Masters degree and two years of clinical experience administered the KSADS-PL and WASH-U-KSADS. Diagnostic decisions were ultimately made by a board-certified child psychiatrist who reviewed the interview, and also performed a clinical interview on the child to confirm the diagnoses. . Inter-rater reliability was established at the outset by rating videotaped interviews, observing trained rater interviews, and performing interviews with observation by a trained rater, as described by Geller et al. (1998)
(four consecutive patients with 100% agreement on diagnoses). The inter-rater reliability for diagnoses was a kappa of > 0.9. Current and lifetime diagnoses were established according to DSM-IV criteria.
Bipolar offspring had psychostimulants discontinued for at least 24 hours before the scan, primarily due to a concurrent functional MRI study of attention. They were allowed to continue any other current medications such as mood stabilizers or antidepressants due to the risk of mood destabilization. Medication history was obtained from interviews with subjects and parents and review of medical records when available. Past exposure to lithium or valproate was recorded if the subject had at least 6 months treatment of either agent at standard doses or serum levels.
For inclusion in the control group, healthy volunteers could not have a current or lifetime DSM-IV psychiatric diagnosis, neither parent had a lifetime or current psychiatric diagnosis by SCID, and the participant did not have a first or second degree relative with BD as determined by the Family History Research Diagnostic Criteria (Andreasen et al., 1977
). None of the healthy control group parents had any Axis I diagnoses.
All the study participants and healthy volunteers were scanned on a GE 3Tesla scanner. Coronal 3D volumetric spoiled gradient echo (SPGR) series were obtained with the following parameters: TR = 35, TE = 6, flip angle = 45, slice thickness = 1.5 or 1.6 mm, and matrix = 256×192 for 124 slices. The number of scans using a 1.6 mm rather than a 1.5 mm slice thickness was not different between the at-risk and control groups (at-risk=5 of 22 scans; control=3 of 22 scans; chi-square(1)=.611, p=.43.)
The volumetric analysis was performed using BrainImageJava software v. 0.13.4 ((Reiss, 2002
), Center for Interdisciplinary Brain Sciences Research, CIBSR; http://cibsr.stanford.edu
) for semi-automated image processing and quantification.
Image processing included removal of non-brain tissue, correction of non-uniformity, and positional normalization to anterior and posterior commissures in a stereotactic space (Talairach and Tournoux, 1988
). Each brain was divided into lobes with a semi-automated stereotactic-based parcellation method (Kates et al., 1999
), based on the raters' identification of the anterior commissure, the posterior commissure, and a midsagittal point above the axis created by the first two points. Raters who conducted morphometric analyses were blind to the diagnosis of each subject. Voxels comprising brain tissue were then segmented into gray matter, white matter, and cerebrospinal fluid (CSF) using a semi-automated fuzzy tissue segmentation algorithm (Reiss et al., 1998
). The total brain volume (TBV) was calculated as the sum of all brain regions. Total cerebral volume was calculated by adding cerebral total tissue with cortical and ventricular CSF. Total brain tissue was calculated by adding cerebral total tissue, cerebellar tissue, and brainstem tissue.
Subcortical regions were outlined manually by a rater who had demonstrated reliability with gold standard tracings established with an independent rater and dataset (single measure intraclass correlation coefficient) > 0.9). Regions were drawn on positionally normalized brain image stacks in the coronal orientation. Hippocampi were traced starting at the slice where a clear distinction between amygdala and hippocampus was first visible and outlined proceeding posteriorly until the structure disappeared. The superior white matter tract extending from the temporal lobe was used as an inferior border of the hippocampus, medial border was defined by CSF and by the pons, where present, and the lateral border was marked by CSF or white matter tracts on the lateral edge of the hippocampus.
Thalami were traced starting on the slice where the structure was first visible and followed until thalamic gray matter disappeared; the border between the gray matter of the thalamus and the surrounding white matter was used to outline the thalamus.
Amygdalae were traced starting on the slice demonstrating the thickest extent of the anterior commissure and following the structure towards the posterior end of the brain. The most superior white matter tract extending from the temporal lobe marked the inferior border, CSF marked the medial border, endorhinal sulcus marked the superior border, and a thick, central white matter tract of the temporal lobe was used as the lateral border of amygdala ().
Figure 1 Outline of the left and right amygdalae on the positionally normalized brain stack in coronal orientation. The most superior white matter tract extending from the temporal lobe marked the inferior border, CSF marked the medial border, endorhinal sulcus (more ...)
2.1 Statistical Analysis
Independent t-tests were used to compare demographic measures and TBV in BD offspring and healthy controls. Brain volume data distributions were first examined for normality to confirm the assumptions of parametric statistics. One-way analyses of covariance (ANCOVAs) were used to compare brain structure volumes, using age and TBV as covariates. A p value of 0.05 (two-tailed) was chosen as the significance threshold. No corrections for multiple comparisons were made for these exploratory analyses. We calculated the effect size (Cohen's d) as the difference between the means (at-risk group mean volume minus control group mean volume for each region) divided by the pooled (average) standard deviation for that region.