SEARCH

SEARCH BY CITATION

Keywords:

  • bipolar disorder;
  • brain;
  • cortical surface-based analysis;
  • cortical thickness;
  • magnetic resonance imaging

Abstract

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Objective:  This study was conducted to explore differences in cortical thickness between subjects with bipolar disorder and healthy comparison subjects using cortical surface-based analysis.

Methods:  Brain magnetic resonance images were acquired from 25 subjects with bipolar disorder and 21 healthy comparison subjects. Cortical surface-based analysis was conducted using the Freesurfer application. Group differences in cortical thickness, defined by the distance from gray/white boundary to the pial surface, were assessed using statistical difference maps.

Results:  Subjects with bipolar disorder exhibited significantly decreased cortical thickness in left cingulate cortex, left middle frontal cortex, left middle occipital cortex, right medial frontal cortex, right angular cortex, right fusiform cortex and bilateral postcentral cortices, relative to healthy comparison subjects (all p < 0.001). Duration of illness in bipolar subjects was inversely correlated with the cortical thickness of the left middle frontal cortex.

Conclusions:  Cortical thinning was present in multiple prefrontal cortices in bipolar disorder. There was also cortical thinning in sensory and sensory association cortices, which has not been reported in previous studies using region-of-interest or voxel-based morphometry methods. Cortical thinning observed in the current study may be related to impairment of emotional, cognitive, and sensory processing in bipolar disorder but longitudinal studies will be necessary to test this hypothesis.

There have been a number of structural brain imaging studies of bipolar disorder evaluating volumetric changes of the frontal lobes (1–3), and temporal lobe subcortical structures, including the amygdala (4–6), hippocampus (7) and basal ganglia (8, 9). Although findings have been inconclusive, recent studies and meta-analyses of the existing literature suggest that increased prevalence of white matter hyperintensities and enlarged right-sided ventricles are among the more consistent brain structural findings in bipolar disorder (10–12).

Abnormalities of prefrontal and anterior cingulate cortex have been reported in a number of brain imaging studies of bipolar disorder (13–19). These brain regions have been suggested to be important in identifying the emotional significance of stimuli and in producing affective states (20). Therefore, their disruption has been suggested as a neuroanatomical model of bipolar disorder, which is related to affective states and emotional processing.

Region-of-interest (ROI) methodologic approaches have the advantage of identifying specific brain ROI, but do not typically evaluate the entire brain. To supplement this limitation of ROI methods, voxel-based morphometry (VBM) is becoming the standard analytical technique for the systematic assessment of gray and white matter density changes across the whole brain (21). Our previous work (13) and other investigators using VBM (14, 15) have demonstrated decreased gray matter density in orbito-frontal and anterior cingulate cortices in bipolar subjects. Decreased cortical density in medial temporal lobes (14, 15) and left ventral occipitotemporal cortex (15) have also been observed in bipolar subjects using VBM. In addition, our recent study using a shape analysis method in bipolar disorder (22) reported that shape changes of anterior and ventral parts of the striatum are connected with frontal and limbic areas, respectively (23).

The human cerebral cortex has many folds and nearly two-thirds is buried within the cortical sulci (24). This multifaceted folding in the cortex leads to difficulties in using ROIs and VBM to study some structures including cortical thickness (25). As cortical thinning may reflect underlying neuropathological deficits in the cortical laminae (24), the accurate measurement of cortical thickness could provide important additional information for characterizing disease-specific neuroanatomical changes. In this context, measurements of cortical thickness have recently been used to study the neuropathology of schizophrenia, Huntington's disease, multiple sclerosis, and the aging process (25–28). However, despite a number of volumetric studies using ROI or VBM methods to characterize neuroanatomical changes in bipolar disorder, to the best of our knowledge there have been no published studies that measured cortical thickness in bipolar subjects.

Previous manual cortical thickness measurement methods from magnetic resonance imaging (MRI) slice data have almost always overestimated the cortical thickness because the cortical surface is not orthogonal to the viewing plane (29). In addition, the manual measurement of cortical thickness over the entire cortical surface is very time-consuming and subject to rater errors. In this study, we used a recently developed cortical surface-based analysis method, the Freesurfer application, to measure cortical thickness (30, 31). This method uses automated surface reconstruction, transformation and high-resolution inter-subject alignment procedures to accurately and rapidly measure the thickness of the entire cortex (29).

We hypothesized, in accordance with our previous VBM findings (13), that decreased cortical thickness in the subjects with bipolar disorder would be observed primarily in the prefrontal cortices. In addition, we explored whether cortical thickness changes in other brain structures could be identified in subjects with bipolar disorder, which have not been previously identified in studies using ROI or VBM methods. To examine possible relationships between cortical thickness alterations, if present, and clinical manifestations of bipolar disorder, correlation analyses were conducted.

Materials and methods

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Subjects

Inclusion criteria were (i) age: 18–59 years and (ii) DSM-IV bipolar disorder, as determined by the Structured Clinical Interview for DSM-IV (SCID-IV). Exclusion criteria were (i) any other comorbid Axis I disorder within the last 3 months, or Axis II antisocial personality disorder; (ii) concurrent or history of any significant medical or neurologic illnesses; (iii) history of head trauma, seizure, learning disorder, and attention-deficit/hyperactivity disorder; and (iv) contraindications to magnetic resonance (MR) scanning.

Thirty-nine subjects with DSM-IV bipolar disorder were initially recruited through the Psychopharmacology Research Laboratory at McLean Hospital, the Bipolar Disorders Research Unit at the Massachusetts General Hospital or the Center for Anxiety and Depression at the University of Washington. Although data were collected from three sites, there were many more subjects from Boston than Seattle. Therefore, our analysis was confined to only subjects recruited from McLean Hospital and Massachusetts General Hospital, who were scanned at McLean Hospital Brain Imaging Center. We excluded MRIs from 14 subjects with lower-quality MR images, typically due to movement artifacts (less than grade 4 on the Iowa Mental Health Clinic Research Center scan quality rating scale) for the current study, as the measurement of cortical thickness requires a high signal-to-noise ratio. Healthy comparison subjects were recruited through advertisements at local newspapers. Fourteen bipolar I subjects and 18 healthy comparison subjects were study subjects in our prior VBM studies (13).

We compared 25 subjects diagnosed with bipolar disorder, type I or type II, to 21 healthy comparison subjects who were without any DSM-IV Axis I or Axis II disorders and without a family history of major depression or bipolar disorder in first-degree relatives. No subject had a current history of substance abuse and there was no past history of drug or alcohol dependence. Subjects underwent urine toxicology prior to MRI scanning to confirm the absence of drug usage. There were no significant differences of age, gender, duration of illness, and previous drug exposures between bipolar subjects included in the current study and those that were excluded due to MRI quality. Demographic and clinical characteristics for bipolar subjects and healthy comparison subjects are presented in Table 1. The consent process was approved by each hospital institutional review board. All subjects participated after reviewing and signing the consent form in the presence of a study physician.

Table 1.  Demographic and clinical characteristics of subjects with bipolar disorder and healthy comparison subjects
Demographic variables Bipolar subjects (n = 25)Healthy comparison subjects (n = 21)
  1. No significant differences exist between groups in age, gender, education, and handedness. Results are reported as means (SD) or as numbers (%). HDRS = Hamilton Depression Rating Scale; YMRS = Young Mania Rating Scale.

  2. aEducational codes: 1 = completed through grade 6; 2 =completed through grade 7–12; 3 = graduated high school; 4 = some college; 5 = graduated 2-year college; 6 =graduated 4-year college; 7 = some graduate/professional school; 8 = completed graduate/professional school.

  3. bAge of onset and year since the first episode for three bipolar subjects were not determined.

  4. cn = 15.

  5. dn = 17.

Age (years)33.8 (9.6)31.5 (9.7)
Sex (male)10 (40.0%)5 (23.8%)
Educationa5.3 (2.0)6.5 (1.4)
Handedness (right)23 (92%)20 (95.2%)
Age of onsetb17.9 (5.4)
Years since the first episodesb16.5 (11.5)
Number of past manic/ hypomanic/mixed episodesc15.3 (12.0)
Number of past depressive episodesd18.3 (10.8)
Bipolar I disorder18 (72.0%) 
Bipolar II disorder7 (28.0%) 
HDRS scores at the time of scan16.9 (8.8) 
YMRS scores at the time of scan10.4 (7.8)
Current medications
 Lithium6 (24.0%) 
 Depakote®4 (16.0%)
 Others8 (32.0%) 
 None7 (28.0%) 

MR image acquisition

Brain MRI was performed using a 1.5 T GE whole body imaging system (Horizon Echo-Speed; General Electric Medical Systems, Milwaukee, WI, USA) and a custom-made linear birdcage coil with approximately 40% improvement in signal-to-noise ratio and improved homogeneity over standard quadrature head coil (C. Hayes, personal communication, 1996). A three-dimensional spoiled gradient echo pulse sequence was used to produce 124 1.5-mm-thick contiguous coronal images (TE = 5 ms, TR = 35 ms, 256 × 192 matrix; FOV = 24 cm, flip angle = 45°, 1 NEX). Axial proton-density and T2-weighted images (TE = 30/80 ms, TR = 3000 ms, 256 × 192 matrix; FOV = 24 cm, flip angle = 45°, 0.5 NEX, 3-mm-thick slices, no skip) were obtained to screen for brain structural abnormalities.

Surface reconstruction

Brain surface reconstruction and measurements of cortical thickness were conducted by the cortical surface-based analysis developed by Dale and Fiscle using Freesurfer program (30–32). Cortical white matter voxels of the 3D structural scan were segmented for tessellation of gray/white matter boundaries. Automatic correction of topological defects and manual correction of inaccurate segmentation after visual inspection were conducted. Each brain surface tessellation folded by sulci and gyri was inflated, morphed and registered into an average spherical surface template (33). Non-rigid high-resolution surface-based averaging method was used for an optimal alignment of cortical folding patterns (30, 33). This process makes it possible to accurately match anatomically homologous cortical locations of each subject's reconstructed brain while minimizing a metric distortion (30, 33). Smoothing processes of surface tessellation were conducted using an iterative nearest-neighbor averaging procedure in order to remove noise-induced variations (1000 iterations, equivalent to 2-D Gaussian kernel with a full-width half-maximum of approximately 6 cm). Areas with cortical thickness differences were demonstrated in statistical difference maps (Fig. 1).

image

Figure 1. The average statistical map of difference in cortical thickness in subjects with bipolar disorder (n = 25) and healthy comparison subjects (n = 21). Cortical thinning is displayed on lateral (left column), anterior (middle column), and medial (right column) views of inflated standard brains. (A) Left hemisphere: Cortical thinning across subjects was significant in middle frontal cortex, cingulate cortex, postcentral cortex, and middle occipital cortex. (B) Right hemisphere: Cortical thinning across subjects was significant in medial frontal cortex, angular cortex, postcentral cortex, and lateral occipital cortex. Color scale represents the dynamic range of thinning. Full yellow corresponds to a statistical difference in cortical thickness with a p value of 0.001. Blue represents a significant thickening of the cortex.

Download figure to PowerPoint

Measurements of cortical thickness

Shortest distances (i) from the surface to the gray/white boundary and (ii) from the gray/white boundary to the surface were measured for every subject. Cortical thickness was defined as the average of these two shortest distances. Deformable surface algorithm can detect the gray/white boundary and surfaces with sub-millimeter precision (25, 29). Both intensity and continuity information in the reconstruction process was used for areas with ambiguous MR images.

Statistical analysis

Surface regions with significant differences of mean cortical thickness values between two groups (i.e. bipolar subjects versus healthy comparison subjects) were overlaid in statistical difference maps. The statistical difference map was constructed using a random effects model with t-tests for each cortical location. An uncorrected p-value of 0.001 (two-tailed) was considered a significant threshold for statistical difference maps, as this threshold is, when an a priori hypothesis is present, approximately equivalent to a p-value of 0.05 corrected for multiple comparisons (34).

Contiguous regions with significant cortical thinning over an area of 300 mm2 were selected as ROIs. Average thickness values within each ROI for each brain were calculated after the reverse spherical morphing procedure (30). These thickness values were then used to evaluate the effects of age, gender, duration of illness, and previous drug exposure on regional cortical thickness in bipolar subjects and healthy comparison subjects.

Group differences in continuous and categorical variables were computed using independent t-tests and Fisher's exact test for 2 × k table, respectively. Association between continuous variables was calculated using Pearson correlation analysis. Statistical significance was defined at α = 0.05 using two-tailed statistical tests. Statistica 6.0 was used for computations.

Results

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Neuroanatomical loci demonstrating significant differences in cortical thickness between the bipolar group and the healthy comparison group are shown in Fig. 1. In the left hemisphere, there were significant decreases of cortical thickness in the middle frontal cortex [Brodmann area (BA) (35) 46, 2.24 versus 2.38 mm, 6.1% decrease], postcentral cortex (BA 3, 2.53 versus 2.74 mm, 7.6% decrease), pregenual anterior cingulate cortex (BA 32, 2.93 versus 3.11 mm, 5.8% decrease), dorsal anterior cingulate cortex (BA 24, 2.50 versus 2.68, 6.5% decrease), posterior cingulate cortex (BA 23, 2.45 versus 2.57 mm, 4.7% decrease), and middle occipital cortex (BA 18, 1.90 versus 2.06 mm, 7.8% decrease) comparing bipolar subjects with the healthy comparison group (Table 2).

Table 2.  Brain regions with significant cortical thinning (p < 0.001) in subjects with bipolar disorder compared with healthy comparison subjects
Talairach coordinatesBrain regionBACortical thickness of bipolar subjects (mm)Cortical thickness of comparison subjects (mm)Differences between groups (%)
xyz
  1. Results of cortical thickness are reported as means ± SD (mm). Brodmann areas most approximating to each ROI are provided. BA = Brodmann area; DLPFC = dorsolateral prefrontal cortex.

Right hemisphere
 10550Medial frontal cortex (orbitofrontal cortex)102.49 ± 0.182.66 ± 0.166.4
 46−1644Postcentral cortex41.95 ± 0.192.14 ± 0.178.6
 38−6432Parietal, angular cortex392.39 ± 0.162.59 ± 0.167.5
 29−74−11Occipital, fusiform cortex191.90 ± 0.152.05 ± 0.197.2
Left hemisphere
 −412922Middle frontal cortex (DLPFC)462.24 ± 0.182.38 ± 0.126.1
 −55−1024Postcentral cortex32.53 ± 0.242.74 ± 0.247.6
  −6436Anterior cingulate cortex (pregenual regions)322.93 ± 0.203.11 ± 0.305.8
 −101332Anterior cingulate cortex (dorsal regions)242.50 ± 0.242.68 ± 0.266.5
  −6−1034Posterior cingulate cortex232.45 ± 0.172.57 ± 0.184.7
 −26−84−8Occipital, middle cortex181.90 ± 0.162.06 ± 0.177.8

In the right hemisphere, there were also significant decreases in cortical thickness for the medial frontal cortex (BA 10, 2.49 versus 2.66 mm, 6.4% decrease) and postcentral cortex (BA 4, 1.95 versus 2.14 mm, 8.6% decrease), as well as the angular cortex (BA 39, 2.39 versus 2.59 mm, 7.5% decrease), and fusiform cortex (BA 19, 1.90 versus 2.05 mm, 7.2% decrease) comparing the bipolar with the healthy comparison group (Table 2).

There were no regions of significant increases of cortical thickness in bipolar subjects relative to healthy comparison subjects at the level of p < 0.001.

In order to assess the stability of our findings, we have conducted additional analyses using the matched splits of the bipolar and healthy comparison groups. The pattern of cortical thinning in Split 1 bipolar versus healthy comparison groups was similar to that in Split 2 bipolar versus healthy comparison groups. In addition, these patterns were similar to that of cortical thinning in the bipolar group when the full sample was used.

Contiguous regions having significant differences between groups were defined as ROIs. Six ROIs were identified in the left hemisphere: middle frontal cortex, three divisions (dorsal, subgenual, posterior) of cingulate cortex, postcentral cortex, and middle occipital cortex. Four ROIs were identified in the right hemisphere: medial frontal cortex, angular cortex, postcentral cortex, and fusiform cortex.

The influences of potential confounders including age, gender, and educational level were tested in additional analyses. After controlling for age, gender or educational levels using general linear models, between-group differences in cortical thickness remained significant in all ROIs.

Duration of illness in bipolar subjects was inversely correlated with cortical thickness of the left middle frontal cortex (r = −0.58, p = 0.005) (Fig. 2), and the right postcentral cortex (r = −0.45, p = 0.04), but not significantly with other ROIs (all |r|s < 0.29). As there was a high colinearity between age and illness duration in bipolar subjects (r = 0.88, p < 0.0001), we did not conduct correlation analysis controlling for age. Instead, we conducted correlation analyses between cortical thickness and age separately for bipolar and comparison groups.

image

Figure 2. Correlation of the duration of illness with cortical thickness in ‘region of interest’* in left middle frontal cortex in subjects with bipolar disorder. Length of illness duration inversely correlated with cortical thickness of left middle frontal cortex in bipolar subjects (r = −0.58, n = 22, p = 0.005). *Contiguous regions with significant cortical thinning in bipolar subjects compared with healthy comparison subjects (p < 0.001).

Download figure to PowerPoint

Age was inversely correlated with cortical thickness of the left middle frontal cortex in bipolar subjects (r = −0.46, p = 0.02), but not in healthy comparison subjects (r = −0.10, p = 0.66). In contrast, age was inversely correlated with cortical thickness of the left pregenual anterior cingulate cortex in healthy comparison subjects (r = −0.46, p = 0.04), but not in bipolar subjects (r = −0.11, p = 0.58). There were no significant correlations between age and cortical thickness for other ROIs in either diagnostic group (all |r|s < 0.39).

There were no significant differences in cortical thickness for any ROIs between men and women in either diagnostic group. In addition, there were no significant differences in cortical thickness for any ROIs between subjects with bipolar I disorder compared with bipolar II disorder (Table 3), or between drug-naïve and medicated bipolar subjects (Table 4).

Table 3.  Comparison of cortical thickness in brain regions with significant cortical thinning between bipolar I subjects and bipolar II subjects
 Bipolar I subjectsBipolar II subjects tp
  1. Results of cortical thickness are reported as means ± standard deviations (mm). DLPFC = dorsolateral prefrontal cortex.

Right hemisphere
 Medial frontal cortex (orbitofrontal cortex)2.46 ± 0.152.57 ± 0.23−1.440.16
 Postcentral cortex1.99 ± 0.211.87 ± 0.081.380.18
 Parietal, angular cortex2.43 ± 0.182.31 ± 0.071.670.11
 Occipital, fusiform cortex1.92 ± 0.171.87 ± 0.070.780.48
Left hemisphere
 Middle frontal cortex (DLPFC)2.22 ± 0.192.27 ± 0.17−0.520.61
 Postcentral cortex2.54 ± 0.232.53 ± 0.270.050.96
 Anterior cingulate cortex (pregenual regions)2.95 ± 0.222.86 ± 0.161.100.28
 Anterior cingulate cortex (dorsal regions)2.47 ± 0.252.59 ± 0.21−1.110.28
 Posterior cingulate cortex2.44 ± 0.172.47 ± 0.19−0.380.70
 Occipital, middle cortex1.89 ± 0.161.92 ± 0.17−0.540.61
Table 4.  Comparison of cortical thickness in brain regions with significant cortical thinning between drug-exposed bipolar subjects and drug-naïve bipolar subjects
 Drug-exposed bipolar subjectsDrug-naïve bipolar subjects tp
  1. Results of cortical thickness are reported as means ± SD (mm). DLPFC = dorsolateral prefrontal cortex.

Right hemisphere
 Medial frontal cortex (orbitofrontal cortex)2.47 ± 0.182.57 ± 0.15−1.310.20
 Postcentral cortex1.95 ± 0.201.97 ± 0.16−0.230.82
 Parietal, angular cortex2.38 ± 0.162.43 ± 0.16−0.670.51
 Occipital, fusiform cortex1.90 ± 0.161.91 ± 0.12−0.150.88
Left hemisphere
 Middle frontal cortex (DLPFC)2.22 ± 0.202.28 ± 0.14−0.690.49
 Postcentral cortex2.54 ± 0.262.53 ± 0.200.040.97
 Anterior cingulate cortex (pregenual regions)2.89 ± 0.203.02 ± 0.20−1.460.16
 Anterior cingulate cortex (dorsal regions)2.48 ± 0.242.55 ± 0.27−0.620.54
 Posterior cingulate cortex2.44 ± 0.182.49 ± 0.15−0.710.48
 Occipital, middle cortex1.90 ± 0.181.89 ± 0.100.060.95

Discussion

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

To the best of our knowledge, this is the first study to demonstrate regional changes in cortical thickness in subjects with bipolar disorder. We report cortical thinning in prefrontal cortices in bipolar disorder, which is consistent with our prior VBM studies (13). In addition, we also found cortical thinning in sensory and sensory association cortices, which was not observed using VBM methods (13).

In accordance with our primary hypothesis, subjects with bipolar disorder were found to have significantly thinner right orbitofrontal cortex (medial frontal cortex) and left cingulate cortex, relative to the healthy comparison group. These focal cortical thinnings are consistent with previous studies using other brain imaging methods, that have reported decreased prefrontal volume (7), decreased subgenual cortical volume and reduced blood flow to these regions (3). Decreased density of cingulate neurons has also been reported in postmortem studies of bipolar disorder (36), as well as reduced gray-matter volume in the left anterior cingulate in drug-naïve bipolar subjects (37). Our finding of cortical thinning in the cingulate and orbitofrontal regions may be linked to observations of abnormal autonomic responses to emotional stimuli in bipolar disorder (38). Cortical thinning of orbitofrontal and cingulate regions observed in the present study may be related also to our recent findings of shape changes of the ventral striatum (connected limbic structures) (22).

Cortical thinning in orbitofrontal, dorsolateral and cingulate regions was in line with a neuroanatomical model of bipolar disorder, i.e. the disruption of prefrontal and anterior cingulate cortex in bipolar disorder, which has been consistently supported by previous brain imaging studies (13–19). These brain cortices are implicated in emotional processing and emotional experience (39–42). Therefore, our findings of cortical thinning in these brain areas in bipolar disorder might be related to the uncontrolled emotion in bipolar patients.

We also found significant thinning of the left middle frontal cortex (dorsolateral prefrontal cortex) in the bipolar subjects. This observation is consistent with prior studies that found decreased gray matter density (15), decreased N-acetylaspartate (43) and decreased glucose metabolism during the manic phase (18, 44) in left dorsolateral frontal cortex. Findings in bipolar subjects of significant thinning of the left middle frontal cortex may be related to concurrent evidence for altered shape of the anterior striatum that is connected to the prefrontal cortices (22). It is plausible that the thinning we observed in the dorsolateral prefrontal cortex is functionally related to the impaired executive function often observed in bipolar disorder (38).

The association between illness duration and cortical thinning in left middle frontal cortex found in our study may pertain to the progressive decline in a wide range of cognitive functions that is associated with bipolar disorder (45, 46). The association may be due to an aging effect, as age also correlated with the cortical thickness of the left middle frontal cortex. However, as there was no correlation between cortical thickness in the left middle frontal cortex and age within the healthy comparison group, a sole effect of aging was less likely to have contributed to correlations between the cortical thickness and the illness duration. In addition, as multiple comparisons were conducted, readers should exercise caution in interpreting correlation findings of the current study. Future prospective studies are required to establish any putative relationship between the degree of cortical thinning and cognitive decline in bipolar disorder.

It is also intriguing that a different pattern of cortical thinning between right and left hemispheres was observed in the bipolar subjects. Cognitive impairments predominantly in performance IQ and visual spatial dysfunction (47), as well as the occurrence of secondary mania due to right hemispheric lesions (48), have supported theories of right more than left hemispheric dysfunction in bipolar disorder. However, left hemispheric abnormalities have also been reported (38). Our findings for cortical thinning for both right and left hemispheres add to the existing literature implicating both hemispheres in bipolar disorder, although different patterns of distribution were found between hemispheres. In this regard, albeit highly speculative at this time, abnormalities of the dorsolateral prefrontal and cingulate cortex in the left hemisphere could underlie the cognitive decline in bipolar disorder, and abnormalities of orbitofrontal and sensory association cortices in the right hemisphere may underlie deficits of emotional and visual spatial processing in bipolar disorder.

Independent of our primary hypothesis, there were unhypothesized findings of significant cortical thinning in parietal and occipital cortices; in left middle occipital cortex, right angular cortex, right fusiform cortex and bilateral postcentral cortices. Gray matter density decreases of these areas in the bipolar subjects were not detected in our previous VBM study (13). One qualification to note in this regard is that a difference in findings such as this one may be ascribed to the different means of measurement. For example, analogous differential findings were reported in a study of schizophrenia, in which the effect size for detecting parietal cortical thickness alterations was greater than that for detecting changes in parietal gray matter density (49). Our findings in parietal and occipital regions, however, should be interpreted with caution, since an uncorrected p-value of 0.001 may not be conservative enough, considering the lack of a priori hypotheses for these regions.

It is also interesting that the bipolar subjects exhibited cortical thinning in parietal and occipital cortices, which are associated with sensory functions. Lochhead et al. have reported decreased gray matter density in the left ventral occipitotemporal cortex (15). Doris et al. also found decreased gray matter density in the parietal lobe in poor outcome bipolar subjects (50). In a SPECT study, decreased perfusion has been reported in the occipital lobes of bipolar depressive subjects (51). Our findings of cortical thinning in sensory and sensory association cortices may be related to impairments in visual spatial neuropsychologic functions of bipolar subjects (52, 53). In addition, abnormalities of these cortices may be related to sensory neurological soft sign abnormalities that have a high prevalence rate in bipolar disorder (54). Further, functional imaging studies and concurrent assessment of altered sensory processing will be critical in confirming a possible relationship between regional cortical abnormalities and sensory dysfunction in bipolar disorder.

Findings of the current study did not exactly coincide with findings of our previous VBM study (13). This partial discrepancy between our previous VBM study and the current study of cortical thickness may be due to differences in applied methodology, as the current method assessed the thickness of gray matter, rather than probabilistic gray matter density as measured in our previous VBM study (30, 31). In addition, differences in demographic and clinical characteristics may have contributed to those in the findings. Bipolar I and II subjects were included in the current study while only bipolar I subjects were in the previous VBM study.

A number of factors limit the conclusions that can be drawn from the current study. The fact that our bipolar subjects included subjects with prior or current use of psychotropic medications should be considered in interpreting our findings of cortical thinning. We have previously reported changes in brain chemistry that occur after lithium or valproate treatments (55); and such mood stabilizers have been associated with enhanced neurogenesis and increased gray matter density (56). Although there were no differences in cortical thickness in any ROIs between never-medicated and medicated bipolar subjects in the current study, readers should be cautious in interpreting this result as only seven bipolar subjects in our study were drug-naïve. To explore the postulated effects of mood stabilizers on increasing gray matter, future studies measuring cortical thickness before and then during medication treatment will be required.

In summary, our study using cortical surface-based analysis found significant decreases in cortical thickness of the left cingulate cortex, left middle frontal cortex, left middle occipital cortex, right medial frontal cortex, right angular cortex, right fusiform cortex, and bilateral postcentral cortices in bipolar subjects relative to healthy comparison subjects. These findings suggest that abnormalities in multiple cortical areas, which are associated with sensory function as well as emotional and cognitive processing, may play an important role in the pathophysiology of bipolar disorder.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

This study was supported in part by grants from the National Institute of Mental Health (MH58681), the Stanley Medical Research Institute (I.K.L.), the Stanley Foundation Bipolar Disorders Research Grant at McLean Hospital, the Poitras Foundation (P.F.R.), National Alliance for Research on Schizophrenia and Depression Young Investigator Award, and the Harvard–MIT Clinical Investigator Training Program Award (I.K.L.).

References

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References