In this study we could not observe any significant differences of grey or white matter volumes between euthymic patients with bipolar I disorder and healthy control subjects. Firstly we looked at regions with an a priori hypothesis as laid out in the introduction (prefrontal cortex including anterior cingulate cortex, amygdala, ventral striatum and thalamus) with a significance level of P
= 0.001 and a threshold of 200 contiguous voxels. Secondly we conducted analyses with small volume correction around the Talairach coordinates of all originally reported volume abnormalities in previous studies [1
]. Neither the first nor the second analysis revealed significant volume abnormalities of grey or white matter in our samples.
Our results are in accordance with three studies [4
] apart from the finding of decreased grey matter volume in the prefrontal lobe in the latter. Our results are in contrast to three studies using the optimized VBM protocol [1
] and two using traditional VBM [5
]. In these studies decreased grey matter density or decreased and increased grey matter volume in multiple cortical regions were reported. For the most part these differences were observed in frontal cortical regions involving prefrontal and anterior cingulate cortices. But no study could replicate prior findings in exactly the same localization according to the Talairach coordinates. Moreover, some reported abnormalities are contradictory, for example increased and decreased volumes of the anterior cingulate gyrus. Taken together the reported grey matter volume or density abnormalities in bipolar disorder are highly inconsistent. We hypothesize that this inconsistency might be caused by differences in the investigated samples with regard to sample size, clinical outcome, type of bipolar disorder, age range and number of previous manic or depressive episodes.
Three studies investigated small samples with 16 [9
] or 11 patients [5
]. Significant results in such small samples might be false positive findings. In the present study the sample consisted of 33 patients with bipolar disorder and 30 healthy control subjects. Since we used a rather conservative level of significance as compared to previous VBM studies [1
] this might have led to a false negative finding. To resolve this problem, we lowered the significance level as suggested by several authors [1
] but again we could not find any abnormalities. Nonetheless our sample might be too small to detect hypothesized abnormalities, but to our knowledge there is no study that investigates many more patients.
Previous VBM studies in bipolar disorder also differed from each other with respect to the clinical characteristics of the samples. One study [5
] investigated patients with poor outcome whereas the other studies did not mention the outcome of the patients. In the present study we analysed only patients with good outcome that are euthymic and on stable medication.
The majority of studies assessed patients with bipolar I disorder [1
] but three included also patients with bipolar II disorder into analysis [4
]. Lochhead et al. [10
] investigated seven patients with bipolar I and four patients with bipolar II disorder, Bruno et al. [4
] included 28 patients with bipolar I and 11 patients with bipolar II disorder and Nugent et al. [19
] investigated seven bipolar I and 29 bipolar II patients. In our study we investigated only patients with bipolar I disorder. Clinical differences between patients like recurrent episodes of hypomania or mania might be caused by different structural brain abnormalities or might lead to different structural changes. Therefore, investigations of patient samples that are heterogeneous with respect to clinical criteria might produce inconsistent results.
The age range of the subjects in most studies was between 18 and 65 years. In most studies matching was done with regard to mean age and not pair-wise to single subjects. Because the effect of age on brain structure may be non-linear and may also vary between individuals, matching by mean values alone might not be able to fully control for age effects. We analysed a large sample in which patients with bipolar disorder were significantly older than the controls. When using age as a covariate, we could not observe any changes of grey or white matter volumes.
A meta-analysis of MRI studies revealed that hippocampal volume is reduced in patients with repeated periods of major depressive disorder [27
]. Patients with bipolar disorder did not show a reduction of hippocampal volume in that study. Nevertheless, this finding suggests a possible influence of number of affective episodes on brain structure. Therefore, this criterion should be taken into account in discussions of study results of brain volume abnormalities. The majority of previous VBM studies in bipolar disorder did not report number of episodes [4
]. Two studies investigated patients with multiple depressive and manic episodes [10
]. One study so far [1
] and the present study investigated patients with only few previous illness episodes (Table ). Due to these clear differences regarding number of affective episodes the comparability of the studies is limited. We performed additional multiple regression analyses concerning number of manic or depressive episodes. In the present study we did not find any significant correlation between number of affective episodes and grey or white matter volume.
There is some evidence that psychiatric medication could influence brain structure. Some studies reported increased grey matter volume due to treatment with lithium in patients with bipolar disorder [18
]. So far no data is available on the effect of other mood-stabilizers or antidepressive drugs on grey matter volume. The effect of first and second generation antipsychotics on brain structure in patients with schizophrenia was recently reviewed [23
]. First generation antipsychotics seem to increase the volume of nucleus caudatus and to reduce the overall grey matter volume. In contrast to this, second generation antipsychotics seem to affect neither the cortical grey matter nor the basal ganglia volumes except the thalamus whose volume has been reported to be increased. Most patients in the VBM studies in bipolar disorder were treated with mood stabilizers or antipsychotic drugs. However, no study investigated their possible influence on the reported abnormalities of brain structure. In this study, we did not differentiate between patients treated with lithium or other mood stabilizers. We performed however multiple regression analyses regarding treatment with lithium or antipsychotics. This analysis did not reveal any significant correlation of grey or white matter volume with medication.
VBM studies on patients with schizophrenia reported grey matter reductions in frontal and temporal regions [8
]. These differences—in particular in the dorsolateral prefrontal cortex and the superior temporal gyrus—were already found at the time of the first psychotic episode [16
]. One recent study observed additionally an increased CSF volume in the frontal and temporal lobe of patients with schizophrenia [20
]. A recent study on patients with bipolar disorder revealed a ventricular enlargement in patients with psychotic symptoms in contrast to patients without psychotic symptoms [25
]. This result suggests that patients with bipolar disorder with psychotic symptoms could be considered a clinically distinguishable sub-group with possibly distinctive neuropathological features. We performed, therefore, an additional multiple regression analysis concerning the existence of prior psychotic symptoms. However, we could not detect a correlation between psychotic symptoms and grey or white matter volume in our sample.
Investigation with VBM allows structural analyses free from hypotheses and user-bias [2
]. Despite these advantages, its general application in structural MRI studies might be restricted by a low sensitivity in detecting marginal abnormalities. A recent study concluded that manual tracing of the corpus callosum was more sensitive in detecting discrete structural changes than VBM [26
]. Negative findings in VBM-studies, therefore, could only exclude major structural changes but not minor abnormalities.
The lack of structural abnormality in this study may speak against a structural vulnerability hypothesis in bipolar disorder. Nevertheless, studies using region of interest analyses in patients with bipolar disorder revealed for example consistently a decreased amygdala volume in children and adolescents and more heterogeneous an increased amygdala volume in adult patients [24
]. We conclude, therefore, that we need both methods, VBM and region-of-interest analyses to detect major and discrete structural changes in patients with bipolar disorder.
In summary, using the optimized VBM-procedure the present study could not provide any evidence for differences of grey and white matter volume in a sample of patients with bipolar I disorder. This is in contrast to some prior findings. Our negative finding does not eliminate possible grey and white matter volume changes in the longitudinal course of the illness. Perhaps, structural changes emerge only in distinct sub-groups of bipolar patients. Larger well matched samples in consideration of clinical characteristics are needed to examine brain volume abnormalities in longitudinal studies. A combination of structural and functional imaging might reveal more robust data on patho-physiological relevant brain regions in bipolar disorder.