Although bipolar disorder (BD) is traditionally considered as a discrete diagnostic entity, its phenotypic variability suggests the need to redefine this illness based on pathophysiology in the framework of research domain criteria (RDoC) (Insel et al. Reference Insel, Cuthbert, Garvey, Heinssen, Pine, Quinn, Sanislow and Wang2010). A recent focus has been placed on the dimension of psychotic symptoms, which is relevant for the clinical outcome of BD and may be neurobiologically determined. Psychotic BD (PBD) and non-psychotic BD (NPBD) may indeed represent two distinct biological subtypes of the illness, with the former clinically and genetically closer to schizophrenia (Mamah et al. Reference Mamah, Alpert, Barch, Csernansky and Wang2016). The non-consideration of this factor may have also contributed to the heterogeneous findings of neuroimaging studies on BD (Ellison-Wright & Bullmore, Reference Ellison-Wright and Bullmore2010).

Here, in the attempt to clarify the neuroanatomical differences within BD, we summarise the results of the few magnetic resonance imaging (MRI) studies that investigated grey matter volume (GMV) in PBD and NPBD. We conducted a bibliographic research on PubMed using ‘bipolar psychosis MRI’ and ‘psychotic non-psychotic bipolar’ as keywords. From the resulting lists of works (321 and 147, respectively), we selected the original articles that compared PBD and NPBD patients between each other and possibly to healthy controls and/or schizophrenia patients in terms of GMV. Only studies on adult patients (mean age >18 years) were considered.

The methods and results of the eight papers that met the inclusion criteria are summarised in Table 1. In these studies, the neuroanatomy was investigated at the voxel-level and/or in regions of interest. Half of them used voxel-based morphometry to locally compare GMV in the whole brain (Chen et al. Reference Chen, Wen, Malhi, Ivanovski and Sachdev2007; Keramatian et al. Reference Keramatian, Dhanoa, McGirr, Lang, Honer, Lam and Yatham2016; Neves Mde et al. Reference Neves Mde, Duarte, Albuquerque, Nicolato, Neves, Souza-Duran, Busatto and Correa2016; Ekman et al. Reference Ekman, Petrovic, Johansson, Sellgren, Ingvar and Landen2017), whereas the other half compared GMV in a set of pre-defined subcortical regions (Strasser et al. Reference Strasser, Lilyestrom, Ashby, Honeycutt, Schretlen, Pulver, Hopkins, Depaulo, Potash, Schweizer, Yates, Kurian, Barta and Pearlson2005; Womer et al. Reference Womer, Wang, Alpert, Smith, Csernansky, Barch and Mamah2014; Mamah et al. Reference Mamah, Alpert, Barch, Csernansky and Wang2016) or in the cerebellum (Laidi et al. Reference Laidi, d'Albis, Wessa, Linke, Phillips, Delavest, Bellivier, Versace, Almeida, Sarrazin, Poupon, Le Dudal, Daban, Hamdani, Leboyer and Houenou2015). Except from Neves et al. (Reference Neves Mde, Duarte, Albuquerque, Nicolato, Neves, Souza-Duran, Busatto and Correa2016), who merely compared PBD with NPBD, all the works included a group of healthy subjects, and three of them also patients with schizophrenia.

Table 1. Results of GMV comparisons among psychotic and non-psychotic bipolar patients as well as with healthy controls or schizophrenia patients. Only the results concerning bipolar patients and GMV or total intracranial volume are reported. Correlations among structural and clinical variables are listed only if relative to the bipolar subgroups

GMV, grey matter volume; TIV, total intracranial volume; TBV, total brain volume; BD, bipolar disorder; PBD, psychotic bipolar disorder; NPBD, non-psychotic bipolar disorder; MCPBD, mood congruent psychotic bipolar disorder; MIPBD, mood incongruent psychotic bipolar disorder; HC, healthy controls; SCZ, schizophrenia; SCZ-S, schizophrenia spectrum; ST, schizotypal personality disorder; ADHD, attention deficit hyperactivity disorder; AP, antipsychotic; VBM, voxel-based morphometry; ROI, region of interest; LDDMM, Large Deformation Diffeomorphic Metric Mapping; ANOVA, analysis of variance; ANCOVA, analysis of covariance; MANOVA, multivariate analysis of variance; MANCOVA, multivariate analysis of covariance; FWE, family wise error; SVC, small volume correction; LSD, least significant difference; FDR, false discovery rate; DLPFC, dorsolateral prefrontal cortex.

Concerning global brain measures, all the studies that investigated total GMV agreed on the absence of significant differences among PBD, NPBD and healthy subjects (Chen et al. Reference Chen, Wen, Malhi, Ivanovski and Sachdev2007; Keramatian et al. Reference Keramatian, Dhanoa, McGirr, Lang, Honer, Lam and Yatham2016; Mamah et al. Reference Mamah, Alpert, Barch, Csernansky and Wang2016). However, the latter study found higher total GMV in both PBD and NPBD compared with schizophrenia (Mamah et al. Reference Mamah, Alpert, Barch, Csernansky and Wang2016). As per the total intracranial volume, the findings are less consistent. Keramatian et al. (Reference Keramatian, Dhanoa, McGirr, Lang, Honer, Lam and Yatham2016), who distinguished BD patients with mood congruent and mood incongruent psychosis, found lower total intracranial volume in mood congruent PBD compared with NPBD and healthy controls, whereas Mamah et al. (Reference Mamah, Alpert, Barch, Csernansky and Wang2016) reported significant differences only between PBD and schizophrenia, with higher intracranial volume in the former group.

Focusing on the cerebral cortex, the results of the voxel-based studies suggest the presence of specific features in PBD and NPBD. Except from Chen et al. (Reference Chen, Wen, Malhi, Ivanovski and Sachdev2007), who found two clusters of higher GMV in PBD than in controls, in right precentral and middle frontal gyri, the other studies did not report regions of increased GMV in the BD groups compared with the control group (Keramatian et al. Reference Keramatian, Dhanoa, McGirr, Lang, Honer, Lam and Yatham2016; Ekman et al. Reference Ekman, Petrovic, Johansson, Sellgren, Ingvar and Landen2017). On the contrary, GMV deficits in PBD compared with HC were found in the middle temporal (Chen et al. Reference Chen, Wen, Malhi, Ivanovski and Sachdev2007) and fusiform (Ekman et al. Reference Ekman, Petrovic, Johansson, Sellgren, Ingvar and Landen2017) gyri of the left hemisphere, as well as bilaterally in regions of the prefrontal (Keramatian et al. Reference Keramatian, Dhanoa, McGirr, Lang, Honer, Lam and Yatham2016; Ekman et al. Reference Ekman, Petrovic, Johansson, Sellgren, Ingvar and Landen2017) and anterior cingulate cortices (Keramatian et al. Reference Keramatian, Dhanoa, McGirr, Lang, Honer, Lam and Yatham2016). Only one study reported GMV deficits in NPBD compared with healthy subjects, in temporal and occipital clusters (Chen et al. Reference Chen, Wen, Malhi, Ivanovski and Sachdev2007).

The direct comparisons between PBD and NPBD in terms of cortical GMV led to results that are only partially consistent. Only in the work from Chen et al. (Reference Chen, Wen, Malhi, Ivanovski and Sachdev2007) PBD showed increased GMV compared with NPBD in a set of clusters located in bilateral frontal cortex and right cuneus. Opposite findings emerged from Neves et al. (Reference Neves Mde, Duarte, Albuquerque, Nicolato, Neves, Souza-Duran, Busatto and Correa2016) and Ekman et al. (Reference Ekman, Petrovic, Johansson, Sellgren, Ingvar and Landen2017), reporting in PBD compared with NPBD GMV deficits in left fusiform and inferior frontal gyri and in right prefrontal, insular and parieto-occipital regions. Interestingly, Keramatian et al. (Reference Keramatian, Dhanoa, McGirr, Lang, Honer, Lam and Yatham2016) found GMV deficits in mood incongruent PBD compared with both mood congruent PBD and NPBD in a set of cortical regions, as well as in the cerebellum. However, Laidi et al. (Reference Laidi, d'Albis, Wessa, Linke, Phillips, Delavest, Bellivier, Versace, Almeida, Sarrazin, Poupon, Le Dudal, Daban, Hamdani, Leboyer and Houenou2015) did not find differences in cerebellar GMV between PBD and NPBD.

The studies that explored the subcortical structures reported no differences in hippocampal, amygdala and thalamic volumes among patients with PBD, NPBD and schizophrenia and healthy subjects (Strasser et al. Reference Strasser, Lilyestrom, Ashby, Honeycutt, Schretlen, Pulver, Hopkins, Depaulo, Potash, Schweizer, Yates, Kurian, Barta and Pearlson2005; Womer et al. Reference Womer, Wang, Alpert, Smith, Csernansky, Barch and Mamah2014; Mamah et al. Reference Mamah, Alpert, Barch, Csernansky and Wang2016). Interestingly, Womer et al. (Reference Womer, Wang, Alpert, Smith, Csernansky, Barch and Mamah2014) and Mamah et al. (Reference Mamah, Alpert, Barch, Csernansky and Wang2016) suggested GMV deficits in NPBD, specifically in the caudate when compared with controls, and in the globus pallidus when compared with PBD. Reduced GMV in the right caudate body of NPBD compared with PBD was also described (Chen et al. Reference Chen, Wen, Malhi, Ivanovski and Sachdev2007). Moreover, GMV deficits in the caudate, globus pallidus and putamen emerged in BD compared with schizophrenia (Womer et al. Reference Womer, Wang, Alpert, Smith, Csernansky, Barch and Mamah2014; Mamah et al. Reference Mamah, Alpert, Barch, Csernansky and Wang2016), raising questions on the relationship among basal ganglia and psychotic symptomatology.

Taken together, these findings suggest the presence of GMV differences between PBD and NPBD, with the two groups showing specific abnormalities when compared with healthy controls. Indeed, the presence of psychosis in BD seems to be associated with reduced cortical GMV, spanning from prefrontal to temporo-occipital cortices, and increased subcortical GMV, mainly in the basal ganglia. Overall, this evidence confirms the importance to classify BD based on the psychotic dimension, which may have a specific neurobiological architecture that partially overlaps across PBD and schizophrenia. Future larger and longitudinal studies are needed to further explore fronto-striatal and fronto-limbic dysconnectivity in first episode psychotic patients and in subjects at risk to develop schizophrenia or BDs to have a better perspective on the neural basis of psychosis spectrum.