Introduction
Recent evidence regarding shared genetic risk factors and familial co-aggregation has raised questions about the validity of a distinction between schizophrenia and bipolar disorder (Owen et al. Reference Owen, Craddock and Jablensky2007; Craddock & Owen, Reference Craddock and Owen2010; Lichtenstein et al. Reference Lichtenstein, Yip, Björk, Pawitan, Cannon, Sullivan and Hultman2009). On the one hand, both disorders share some common genetic vulnerability factors and yet, on the other hand, schizophrenia, unlike bipolar disorder, also seems to share genetic risk factors with other neurodevelopmental disorders (Craddock & Owen, Reference Craddock and Owen2010; Grozeva et al. Reference Grozeva, Kirov, Ivanov, Jones, Jones, Green, St Clair, Young, Ferrier, Farmer, McGuffin, Holmans, Owen, O'Donovan and Craddock2010).
Brain imaging could help to inform the debate by describing similarities and differences between these disorders. Magnetic resonance imaging (MRI) studies of schizophrenia have provided evidence for structural abnormalities in some structures (i.e. hippocampus, superior temporal gyrus (STG), insula, thalamus, medial frontal regions), with most of these studies using a region of interest (ROI) approach (Nelson et al. Reference Nelson, Saykin, Flashman and Riordan1998; Konick & Friedman, Reference Konick and Friedman2001; Steen et al. Reference Steen, Mull, McClure, Hamer and Lieberman2006; Baiano et al. Reference Baiano, David, Versace, Churchill, Balestrieri and Brambilla2007; Fornito et al. Reference Fornito, Yucel, Wood, Adamson, Velakoulis, Saling, McGorry and Pantelis2008b; Takahashi et al. Reference Takahashi, Wood, Soulsby, Kawasaki, McGorry, Suzuki, Velakoulis and Pantelis2009b, Reference Takahashi, Wood, Soulsby, McGorry, Tanino, Suzuki, Velakoulis and Pantelisc). ROI studies have also provided some evidence of structural abnormalities in bipolar disorder (Fornito et al. Reference Fornito, Malhi, Lagopoulos, Ivanovski, Wood, Saling, Pantelis and Yucel2008a; Arnone et al. Reference Arnone, Cavanagh, Gerber, Lawrie, Ebmeier and McIntosh2009; Takahashi et al. Reference Takahashi, Malhi, Wood, Walterfang, Yücel, Lorenzetti, Soulsby, Suzuki, Velakoulis and Pantelis2009a; Hallahan et al. Reference Hallahan, Newell, Soares, Brambilla, Strakowski, Fleck, Kieseppa, Altshuler, Fornito, Malhi, McIntosh, Yurgelun-Todd, Labar, Sharma, Macqueen, Murray and McDonald2011). A recent ROI meta-analysis compared the brain imaging findings of schizophrenia and bipolar disorder and found greater ventricular enlargement and smaller amygdalae in schizophrenia compared to bipolar disorder (Steen et al. Reference Steen, Mull, McClure, Hamer and Lieberman2006).
The possibility that there are differences between schizophrenia and bipolar disorder in the regions outside the frequently examined ROIs has not yet been fully explored (Ashburner & Friston, Reference Ashburner and Friston2000; Good et al. Reference Good, Johnsrude, Ashburner, Henson, Friston and Frackowiak2001). Voxel-based morphometry (VBM) is a popular whole-brain analysis method useful for revealing brain abnormalities in a regionally unbiased manner and this method has been applied to schizophrenia and bipolar disorder in a number of studies (Glahn et al. Reference Glahn, Laird, Ellison-Wright, Thelen, Robinson, Lancaster, Bullmore and Fox2008; Fornito et al. Reference Fornito, Yücel, Patti, Wood and Pantelis2009; Bora et al. Reference Bora, Fornito, Yücel and Pantelis2010a, Reference Bora, Fornito, Radua, Walterfang, Seal, Wood, Yücel, Velakoulis and Pantelis2011; Ellison-Wright & Bullmore, Reference Ellison-Wright and Bullmore2010). Two recent meta-analyses that compared the findings of schizophrenia and bipolar disorder found more extensive abnormalities in schizophrenia (Ellison-Wright & Bullmore, Reference Ellison-Wright and Bullmore2010; Yu et al. Reference Yu, Cheung, Leung, Li, Chua and McAlonan2010).
However, there are important confounders that need to be examined when comparing schizophrenia and bipolar disorder. Most importantly, a markedly higher ratio of male patients are generally enrolled into schizophrenia studies when compared to studies of bipolar disorder. This is an important confound since male gender has been associated with more severe negative symptoms, deficit subtype, poorer prognosis and greater cognitive impairment in schizophrenia (McGlashan & Fenton, Reference McGlashan and Fenton1992; Leung & Chue, Reference Leung and Chue2000; Kirkpatrick et al. Reference Kirkpatrick, Buchanan, Ross and Carpenter2001; Mitelman & Buchsbaum, Reference Mitelman and Buchsbaum2007; Bora et al. Reference Bora, Fornito, Yücel and Pantelis2009). Given the fact that male gender has been associated with greater susceptibility to neurodevelopmental disorders, neurodevelopmental subtypes of schizophrenia would be expected to be more common in male dominant samples of schizophrenia (Fombonne, Reference Fombonne2003; Handen, Reference Handen, Mash and Barkley2007). Therefore, structural brain abnormalities associated with severe cognitive deficits might be more common in male patients with schizophrenia. In line with these notions, our recent meta-analysis of neuropsychological studies suggested that differences between schizophrenia and psychotic mood disorders are driven by studies that included a higher ratio of males in schizophrenia compared to bipolar disorder comparison groups (Bora et al. Reference Bora, Fornito, Yücel and Pantelis2009). However, this issue has not been systematically investigated in brain imaging research in schizophrenia and bipolar disorder.
Our aim in this study was to compare grey matter (GM) abnormalities observed in schizophrenia and bipolar disorder and to examine the effect of gender on the observed differences. The findings will help shed light on the validity of a distinction between schizophrenia and bipolar disorder.
Method
Inclusion of studies
We followed PRISMA 2009 guidelines in conducting this meta-analysis (Moher et al. Reference Moher, Liberati, Tetzlaff and Altman2009). Studies published in English were identified through extensive searches in PubMed, Scopus, PsychINFO and EMBASE databases for the period from January 1995 to January 2011. We used the following keywords in this search: Voxel*; morphometry; schizophrenia; bipolar disorder. We also reviewed the reference lists of published studies. Inclusion criteria were studies that: (1) reported a voxelwise comparison between patients with schizophrenia or bipolar disorder with controls for GM; (2) reported whole-brain results in stereotactic coordinates and used thresholds for significance corrected for multiple comparisons, or uncorrected with spatial extent thresholds. Studies reporting findings based on only small volume correction (SVC) were excluded. When a study reported SVC findings together with whole brain results, only the latter findings were included. Where necessary, we contacted the authors for voxel coordinates and for clarification of whether there was sample overlap between published studies by the same research group. Several multi-site studies were not included in our analyses since these studies varied in methodology between sites and they had sample overlap with already published studies (Meda et al. Reference Meda, Giuliani, Calhoun, Jagannathan, Schretlen, Pulver, Cascella, Keshavan, Kates, Buchanan, Sharma and Pearlson2008; Segall et al. Reference Segall, Turner, van Erp, White, Bockholt, Gollub, Ho, Magnotta, Jung, McCarley, Schulz, Lauriello, Clark, Voyvodic, Diaz and Calhoun2009). Altogether, 72 studies were included in these meta-analyses (Table 1). A flow chart of study selection is shown in Supplementary Fig. 1 (available online). For GM analyses, 52 of these studies compared schizophrenia and controls and 24 of them compared bipolar disorder and controls.
Sch/HC, Schizophrenia/healthy controls; Edu, education; FE, first-episode.
Statistical analysis
For coordinate based meta-analysis, we used SDM (www.sdmproject.com/software/) software to analyse GM abnormalities in schizophrenia and bipolar disorder (Radua & Mataix-Cols, Reference Radua and Mataix-Cols2009). Briefly, a map of GM differences, comprising the reported stereotactic coordinates for each significant group difference, was generated for each study. In SDM, unlike other coordinate-based, meta-analytic methods, both positive and negative differences are reconstructed in the same map (signed map), which prevents a particular voxel appearing to be significant in opposite directions. Importantly, when using SDM, negative studies are also included in the meta-analyses. Then meta-analytic statistical maps were obtained by calculating the corresponding statistics from the study maps, weighted by the square root of the sample size of each study so that studies with large sample sizes contributed more. The statistical significance of each voxel was determined using randomization tests (p<0.001). In our previous meta-analysis of VBM studies in bipolar disorder, this threshold gave more conservative and consistent results compared to activation likelihood activation analysis with false discovery rate (FDR)p<0.05 in the same dataset (Bora et al. Reference Bora, Fornito, Yücel and Pantelis2010a). To compare schizophrenia and bipolar disorder findings, we contrasted the control–schizophrenia and control–bipolar disorder maps using the omnibus test (Q statistic), correcting for age (dichotomized as 0 and 1 based on the mean) (Radua et al. Reference Radua, van den Heuvel, Surguladze and Mataix-Cols2010). Meta-regression analyses and subgroup analyses were conducted to examine the effect of gender on GM abnormalities.
Results
Demographic characteristics
In schizophrenia studies, 52 studies comparing 2090 schizophrenia patients and 2284 healthy controls were included. There was a small but significant gender difference in these studies; the percentage of male patients was higher in the schizophrenia group in comparison with controls [70.0% v. 61.6%, relative risk (RR) 1.10, CI 1.05–1.16, Z=3.70, p=0.0002]. In bipolar disorder studies (734 patients and 914 controls), patient and control samples were well matched for gender (46.0% v. 48.7%). Mean age for schizophrenia and bipolar patients were 33.1 and 35.9 years respectively.
Meta-analysis of VBM studies
Gender effects on GM findings
Meta-regression analysis suggested that a higher percentage of males in schizophrenia studies was associated with reduced GM in a left hemisphere cluster incorporating the inferior frontal cortex, insula, STG and amygdala extending to hippocampus (−32, 6, −12, SDM=0.37, p=0.0002, 202 voxels), and a right hemisphere cluster including insula and claustrum (36, −4, 6, SDM=−0.45, p=0.00001, 212 voxels). In contrast, there was no significant effect of gender on the findings in bipolar disorder.
GM abnormalities in schizophrenia and bipolar disorder compared with controls
Given the effects reported above concerning the influence of gender on GM changes in schizophrenia, and the fact that the percentage of males was higher in the schizophrenia group (70%) compared with the bipolar disorder (46%) group, we divided the schizophrenia sample into two groups based on mean percentage of males: gender balanced (47% males, range=42–57%) and male dominated (75% of males: range=60–100%). As seen in Fig. 1, the male-dominated schizophrenia sample had extensive GM reductions, including bilateral insula, STG, inferior frontal gyrus, dorsomedial frontal cortex, medial frontal cortex/anterior cingulate cortex (ACC) and thalamus extending to red nucleus (Table 2).
SDM, Signed differential mapping; BA, Brodmann area; ACC, anterior cingulate cortex; STG, superior temporal gyrus; R, right; L, left.
However, gender-balanced samples had a less extensive and less distinct pattern of GM deficits characterized by GM decrease in left dorsolateral prefrontal cortex (BA 9) and fronto-insular cortex/STG and right dorsal ACC/dorsomedial frontal (Fig. 2, red colour) compared with controls (Table 2). Bipolar disorder patients had GM deficits in right ACC/medial frontal cortex, bilateral anterior insula/inferior frontal cortex and subgenual ACC/medial frontal cortex (Fig. 2, blue colour). There were no GM increases in bipolar disorder or schizophrenia patients compared with controls.
Comparison of GM between schizophrenia and bipolar disorder
A Q-test showed that the schizophrenia-male dominated samples had statistically significant decreases of GM in regions including left STG/anterior insula, left amygdala, right posterior insula and bilateral thalamus and midbrain including the red nucleus and surrounding structures (Table 3) compared to bipolar disorder. There was no region where GM was more decreased in bipolar disorder compared to schizophrenia.
SDM, Signed differential mapping; STG, superior temporal gyrus; BA, Brodmann area; R, right; L, left.
A Q-test showed that GM differences between gender-balanced schizophrenia and bipolar disorder patients reached statistical significance only in the right dorsomedial frontal cortex and left dorsolateral prefrontal cortex (smaller in schizophrenia; Table 4).
SDM, Signed differential mapping; BA, Brodmann area.
Discussion
This voxelwise meta-analysis compared voxel-based morphometric studies in schizophrenia and bipolar disorder and specifically examined the effect of gender on between-group differences. A comparison of male-dominated schizophrenia samples with bipolar patients suggested more extensive GM reductions in the former. However, when the groups where matched for gender ratio, the GM differences between the two were more circumscribed. Differences between bipolar disorder and gender-balanced samples of schizophrenia were only modest, involving right-sided reductions in dorsomedial frontal cortex and left dorsolateral prefrontal cortex in schizophrenia.
This meta-analysis found that gender had a substantial effect on GM findings in schizophrenia. These findings suggest that samples including a higher ratio of male schizophrenia patients (i.e. samples with greater male:female ratio) have more extensive GM abnormalities than gender-balanced samples, where they localized mainly to dorsal ACC/dorsomedial frontal cortex, inferior frontal, insula, superior temporal gyrus and left dorsolateral prefrontal cortices. Our findings indicate, however, that most of the GM differences between schizophrenia and bipolar disorder were driven by the male-dominated schizophrenia samples. GM differences between gender-balanced samples of schizophrenia and bipolar disorder were restricted to left dorsomedial frontal cortex and dorsolateral prefrontal cortex. These findings are consistent with our meta-analysis, showing a similar pattern of cognitive differences between schizophrenia and affective psychoses (Bora et al. Reference Bora, Fornito, Yücel and Pantelis2009, Reference Bora, Fornito, Yücel and Pantelis2010b).
These findings might be compatible with genetic findings suggesting both overlapping and distinctive genes for schizophrenia and bipolar disorder (Owen et al. Reference Owen, Craddock and Jablensky2007; Craddock & Owen, Reference Craddock and Owen2010; Lichtenstein et al. Reference Lichtenstein, Yip, Björk, Pawitan, Cannon, Sullivan and Hultman2009). Some patients with bipolar disorder and schizophrenia might be associated with similar genetic vulnerability factors, cognitive and structural brain deficits. On the other hand, separate genetic risk factors associated with extensive GM abnormalities and severe cognitive deficits might be associated with poor outcome schizophrenia and a number of neurodevelopmental disorders. It is likely that schizophrenia samples including a higher ratio of male patients have more severe structural brain abnormalities since male schizophrenics are more likely to have neurodevelopmental type of schizophrenia than females since male gender is associated with higher risk of mild mental retardation and autistic spectrum conditions in the general population (DSM-IV TR). However, one can argue alternative explanations such as the gender-specific plastic influence on the illness-related abnormalities.
Patients with bipolar disorder showed reductions in the parts of the ACC and very anterior part of insula and inferior frontal cortex. These regions co-activate in functional imaging studies and are part of a neural network subserving emotional response inhibition and emotional salience (Di Martino et al. Reference Di Martino, Shehzad, Kelly, Roy, Gee, Uddin, Gotimer, Klein, Castellanos and Milham2009; Taylor et al. Reference Taylor, Seminowicz and Davis2009). In gender-balanced samples of schizophrenia, volume reductions in ACC and insula were more dorsal compared to bipolar disorder. Specific abnormality in dorsal ACC and insula in schizophrenia is interesting since these brain regions are components of a general (cognitive) salience network (Sridharan et al. Reference Sridharan, Levitin and Menon2008). It is thought that this network is responsible for switching between internally focused mentation, mediated by the so-called default mode network and externally focused attention supported by frontoparietal networks (Sridharan et al. Reference Sridharan, Levitin and Menon2008; Taylor et al. Reference Taylor, Seminowicz and Davis2009). In concordance with structural MRI studies, a recent functional imaging study found abnormal connectivity in general salience network (White et al. Reference White, Joseph, Francis and Liddle2010). Volumes of dorsomedial and dorsolateral prefrontal cortices, which is known to interact this system, were also reduced in schizophrenia. In male-dominated schizophrenia samples, abnormalities in ACC and insula were more extensive, including both cognitive and emotional salience networks. Thus, abnormalities of the general salience network might be a core characteristic of schizophrenia independent of gender.
Abnormalities in medial frontal lobe and thalamus also suggest an impairment of frontothalamic circuitry in schizophrenia, which may contribute to negative symptoms and cognitive deficits (Pantelis et al. Reference Pantelis, Barnes and Nelson1992, Reference Pantelis, Barnes, Nelson, Tanner, Weatherley, Owen and Robbins1997). Similar abnormalities were not found in bipolar disorder. It is noteworthy that the frontothalamic network is important for emotional information processing despite the absence of abnormality in this network in bipolar disorder (Caligiuri et al. Reference Caligiuri, Brown, Meloy, Eberson, Niculescu and Lohr2006). It is still possible that bipolar disorder (or a subgroup thereof) might be associated with abnormalities in frontothalamic circuits that are masked by the effects of lithium and antipsychotic use (Chakos et al. Reference Chakos, Lieberman, Bilder, Borenstein, Lerner, Bogerts, Wu, Kinon and Ashtari1994; Moore et al. Reference Moore, Cortese, Glitz, Zajac-Benitez, Quiroz, Uhde, Drevets and Manji2009). Also, GM abnormality in midbrain structures in schizophrenia might be important to explain frontothalamo-cerebellar abnormalities observed in functional imaging studies since the red nucleus lies on cerebellar–thalamic pathways (Andreasen et al. Reference Andreasen, Nopoulos, O'Leary, Miller, Wassink and Flaum1999).
Methodological considerations
Meta-analyses of voxel-based imaging studies are generally limited by the reporting practices of VBM studies, which make it difficult to estimate effect sizes. Another issue concerns the methodological differences of VBM studies. These include differences in smoothing kernel size, slice thickness, statistical thresholding and whether Jacobian modulation is used in pre-processing of VBM. Comparing white matter findings in schizophrenia and bipolar disorder might also be interesting, since there is some evidence for a higher degree of similarities of WM abnormalities for both disorders. However, there are insufficient studies examining whole brain white matter volume or anisotropy to conduct a meta-analysis in bipolar disorder. Also, it was not possible to examine the effect of between-group differences in intellectual abilities on the observed findings due to a lack of relevant data in most studies. It is also likely that differences in medication use in schizophrenia and bipolar disorder can influence the group comparisons. For example, lithium can minimize abnormalities in ACC or medial temporal lobe (Bora et al. 2010a; Hallahan et al. Reference Hallahan, Newell, Soares, Brambilla, Strakowski, Fleck, Kieseppa, Altshuler, Fornito, Malhi, McIntosh, Yurgelun-Todd, Labar, Sharma, Macqueen, Murray and McDonald2011) in bipolar disorder and more frequent use (and higher doses) of antipsychotics in the schizophrenia group can contribute to some of schizophrenia–bipolar disorder differences (i.e. cortical regions) and mask the others (i.e. striatum). However, it was not possible to meta-analytically examine the effect of medication on group comparisons due to a lack of data concerning treatment-related variables and the relatively small number of bipolar studies.
Conclusions
In summary, studies that have demonstrated greater GM abnormalities in patients with schizophrenia compared with bipolar disorder have typically examined samples that include more male patients, who typically show poorer prognosis, greater cognitive impairment and more marked neurodevelopmental compromise. When gender is controlled, GM abnormalities in bipolar disorder and schizophrenia are less pronounced and are restricted to brain regions involved in emotional and cognitive aspects of identification of salient stimuli in the environment. Future studies examining patients with schizophrenia and bipolar disorder longitudinally at multiple time points before and after onset of psychosis would be valuable in further understanding the nature of structural abnormalities in the major psychoses.
Note
Supplementary information accompanies this paper on the Journal's website (http://journals.cambridge.org/psm).
Acknowledgements
M.Y. was supported by a National Health and Medical Research Council (NHMRC) clinical career development award (ID: 509345). A.F. was supported by a National Health and Medical Research Council CJ Martin Fellowship (ID: 454797). C.P. was supported by a NHMRC Senior Principal Research Fellowship (ID: 628386) and by NHMRC Program Grant (ID: 566529).
Declaration of Interest
None.