Transdiagnostic structural neuroimaging features in depression and psychosis: A systematic review

Background: Previous research suggests that there may be similarities in structural brain changes seen in patients with depression and psychosis compared to healthy controls. However, there is yet no systematic review collating studies comparing structural brain changes in depression and psychosis. Establishing shared and specific neuroanatomical features could aid the investigation of underlying biological processes. Aims: To identify structural neuroimaging similarities and differences between patients with depression and psychosis. Method: We searched PubMed, PsychInfo, Embase, NICE Evidence, Medline and the Cochrane Library were searched from inception to 30/06/2021 using relevant subject headings (controlled vocabularies) and search syntax. Papers were assessed for quality using the Newcastle-Ottawa Scale. Results: Five-hundred and twenty papers were retrieved, seven met inclusion criteria. In narrative collation of results, grey matter volume ( GMV) reductions were found in the medial frontal gyrus (MFG), hippocampus and left-sided posterior subgenual prefrontal cortex in both psychosis and depression. GMV reductions affected more brain regions in psychosis, including in the insula and thalamus. White matter volume (WMV) decline was found in both depression and psychosis. Reduced fractional anisotropy (FA) was more commonly seen in depression. Conclusions: Our results suggest potential transdiagnostic patterns of GMV and WMV reductions in areas including the MFG, hippocampus, and left-sided posterior subgenual prefrontal cortex. These could be investigated as a future biomarker of transdiagnostic signature across mental illnesses. However, due to the limited number and poor quality of studies future research in large samples and harmonised imaging data is first needed.


Introduction
Depression and psychotic disorders constitute the major challenges of mental ill health to the world's population (Charlson et al., 2018;Vos et al., 2015;WHO | Depression and Other Common Mental Disorders, 2017). Major Depressive Disorder (MDD) has a lifetime prevalence rate of approximately 20% and is one of the most common and disabling psychiatric diseases globally (WHO | Depression and Other Common Mental Disorders, 2017). Psychotic disorders, including schizophrenia affect up to 3% of the world population (Charlson et al., 2018;World Health Organization, 1992) and are severely debilitating.
Despite depression and psychosis being two different diagnostic structures, they may share some aetiologies and maintenance processes. Depression in schizophrenia has long been a taxonomic challenge, with various approaches to address it including that core, deficit schizophrenia is "non-affective" and early more heterogeneous stages of psychosis have higher affective load (Upthegrove et al., 2017). Craddock and Owen (2010) argue genetic studies reinforce evidence that suggests a clear dichotomy between mood and psychotic disorders is absent.
Symptoms such as social withdrawal, blunted affect, anhedonia, avolition, and alogia are observed both in psychosis and depression (Krynicki et al., 2018). Social withdrawal, blunted affect, and alogia are associated with GMV reduction in the cerebellum, while anhedonia and avolition are negatively correlated with left anterior limb of internal capsule white-matter volume (WMV) and positively correlated with left superior longitudinal fasciculus WMV (Chuang et al., 2014;Meisenzahl et al., 2010). Structural magnetic resonance imaging (sMRI) has shown that patients with psychiatric disorders may have distinct brain changes compared to healthy controls. Depression and psychosis may share certain neuroanatomical characteristics; e.g. hippocampal grey matter volume (GMV) reductions are observed in both disorders (Chuang et al., 2014;Meisenzahl et al., 2010). Overlap in neuroanatomical characteristics between the two disorders may be suggestive of common aetiological mechanisms, with depression potentially playing an intrinsic role in the pathogenesis of schizophrenia (Birchwood et al., 2005;Sönmez et al., 2013). However, to our knowledge there has been no previous systematic review of structural neuroimaging studies comparing depression to psychosis, and here we aim to address this gap. It should however be noted that there is an explicit lack of studies that directly compare the two (See Table 1).

Methods
The protocol was pre-registered on PROSPERO ID CRD42020190200 and is reported in line with Preferred Reporting lines for Systematic review and Meta-Analyses (PRISMA) guideline (Moher, 2009). Please see Fig. 1 for a PRISMA flow diagram documenting the study selection process.

Eligibility criteria
Eligibility criteria were established following PICOS format -patients/population, intervention, comparisons and outcomes (Liberati et al., 2009): a) only human participants; b) ages of 14-70 years; c) include participants with a diagnosis of major depressive disorder or recent onset depression AND patients with a diagnosis of first episode psychosis, or recent onset psychosis, or schizophrenia spectrum disorders, or schizophrenia, or schizophreniform disorders; d) studies reporting primary cross-sectional or longitudinal or follow up data; e) studies reporting GMV, or cortical thickness, or gyrification, or surface area, or white matter anisotropy or diffusion tensor imaging.
Studies including only bipolar disorder were excluded however, if these also included psychosis and depression, they were included as it may be possible to extract data from the relevant groups of interest. Animal studies, non-English language studies and pharmacological interventions were excluded.
Two authors (H.A. and C.W) screened titles and abstracts and then assessed eligible studies for potential inclusion with a thorough full text search. Two separate authors (A.M. and H.S.) checked the final list and any conflicts between authors were discussed and consensus achieved.

Information sources and search strategy
PubMed, PsychInfo, Embase, NICE Evidence, Medline and the Cochrane Library were searched from inception to 30/06/2021 using relevant subject headings (controlled vocabularies) and search syntax. See Supplementary Material and PROSPERO registration for full search terms.
A systematic search combining terms describing psychosis with AND to terms describing depression and with OR to terms describing different structural neuroimaging modalities was conducted on the following databases: PubMed, PsychInfo, Embase, NICE Evidence, Medline and the Cochrane Library from inception to 30/06/2020. The search terms were adjusted for different databases in subject headings and proximity operators. Please see supplement 1.1 for full details on the search strategy.
Reference lists from included studies were also hand searched for studies which met eligibility criteria that may not have been found from the databases.

Results
The systematic search identified 516 unique abstracts. Full text assessments was performed on 44 studies, with 37 of these excluded. Seven studies reporting on seven unique samples with a total of 1,452 participants fulfilled eligibility criteria were included in the review. The mean age of depression patients in the sample was 36.0 (12.1). The mean age of psychosis patients was 31.9 (8.8). Five studies investigated gray matter and three studies explored white matter. For details please see Table 2.

Grey matter
Five studies (Coryell et al., 2005;Jiang et al., 2019;Keller et al., 2008;Meisenzahl et al., 2010;Ota et al., 2013) identified: GMV reduction in the bilateral middle frontal gyrus was reported in both depression and psychosis; hippocampal GMV reductions both in depression and psychosis; bilateral amygdala GMV reduction in psychotic depression compared to non-psychotic depression, and reduced GMV in the insula and the thalamus in psychosis compared to depression.

White matter
Three studies (Cui et al., 2020;Kochunov et al., 2013;Ota et al., 2013) reported reduced WMV volume fractional anisotropy (FA) in the corpus callosum and bilateral anterior and posterior insula in depression compared to psychosis.

Risk of bias
The quality check was completed using the Newcastle-Ottawa Scale (NOS), (Wells et al., 2000). Four out of the seven studies were rated 4 stars which is considered low quality. See Supplementary Table S1 for full details. Jiang et al. (2019), conducted voxel-based-morphometric analysis of the total brain volume between patients with schizophrenia and major depressive disorder (MDD). Results found four clusters with significant group effects. In detail, lower GMV in the bilateral MFG, the medial prefrontal cortex (MPFC), and the rectus gyrus was found in patients with schizophrenia compared to patients with MDD. The authors investigated the correlation and causal effects between GMV alterations and clinical variables for each patient group. Specifically, in the schizophrenia group reduction in GMV was negatively correlated with PANSS negative scores. Using Granger Causality Analysis (GCA) they found that the causal link between the right cerebellum and the medial prefrontal cortex had a positive correlation with PANSS negative scores. Finally, authors found a significant negative correlation between GMV in the MPFC and the causal link between the right cerebellum and the MPFC. However, this was not the same for the depression group where no significant correlations or causal effects between GMV alterations and clinical features were observed. Coryell et al. (2005) looked at GMV in the anterior and posterior Subgenual Prefrontal Cortex (SGPFC) in patients with psychotic depression compared to patients with schizophrenia and healthy controls. Results found that the SGPFC GMV did not significantly differ between patients with psychotic depression and schizophrenia. However, patients with psychotic depression reported the smallest volumes in the left-sided posterior SGPFC. The differences found in the left-sided posterior SGPFC volumes were compared with other structure measures within this region including cortical depth and surface area. Surface area did not differ significantly between groups; however, the depth of the surface area was significantly less in the psychotic depression group compared to the schizophrenia group. On follow up, increases in the posterior SGPFC volume were found in the psychotic depression group only. Keller et al. (2008) performed a region of interest (ROI) analysis to investigate volumetric differences in amygdala and hippocampus in patients with psychotic depression and nonpsychotic depression. Patients with psychotic depression had marginally smaller bilateral amygdala volumes than patients with non-psychotic depression. The study found no significant differences in bilateral hippocampal volume between groups. However, researchers report that their sample of participants did not differ on measures of recurrent chronic depression,    which might be the reason for no differences in hippocampal volume. Meisenzahl et al. (2010) looked at differences in hippocampal volume between patients with depression and schizophrenia and found a significant effect of diagnosis on reduced bilateral hippocampal GMV and hippocampal WMV. Furthemore, when studying differences between patients with first episode of depression and recurrently ill patients with depression the authors reported found no significant differences concerning hippocampal volume measurements. Ota et al. (2013) found that patients with schizophrenia showed reduced GMV in the bilateral anterior and right posterior insula and the bilateral anterolateral, anteromedial, central lateral, central medial, and posterior regions of the thalamus.

White matter volume and integrity
Cui et al. (2020) examined white matter integrity and found that the depression group exhibited more extensive FA reductions compared to the schizophrenia group. Compared to the schizophrenia group the depression group also had significantly reduced FA measurements in the bilateral anterior section of the anterior corona radiata (CR). Results also showed significantly reduced FA in the anterior and posterior portions of the anterior CR in depression patients compared to patients with schizophrenia. In contrast, the similarities between depression and schizophrenia can be seen in the overlapping regions with decreased FA which were primarily in the posterior section of the anterior CR. Depression patients who were on medication had significantly lower FA values in the left anterior CR in comparison to depression patients who were not on medication. Finally, the effect of symptom severity and duration of illness on these results did not survive correction for multiple comparisons in voxel-wise and ROI analysis. Kochunov et al. (2013) showed that a quadratic diagnosis by age interaction was significant in five white matter tracts with the impact of the diagnosis of schizophrenia by age factor indicating that the rate of age-related rate decline was two times as high than in the healthy control group. Also results in the schizophrenia cohort showed that higher age-related decline in FA was present in all twelve tracts analysed. Specifically, the age-of-intercept for average FA was 32.4 years but regionally it was from 27.7 years for the genu of the CC to 49.9 years for the thalamic radiation. The age-of-intercept was negatively correlated with normative age of FA peak suggesting earlier onset of accelerated aging in FA in schizophrenia in tracts that are normally matured slowly. On the other hand, in the depression group, patients demonstrated higher age-related decline in FA for seven tracts compared to HC of which none were statistically significant (genu of CC, body of CC, internal capsule, external capsule, Cingulum, SLF and inferior frontal occipital). Therefore, this suggests no evidence of tract-specific   Ota et al. (2013) found that patients with depression exhibited reduced FA in the CC, bilateral anterior and posterior insula, anterolateral and posterior regions of the thalamus, and left rostral and right subcallosal regions of the thalamus.

Discussion
This is the first systematic review to summarise the results from structural brain imaging studies directly comparing patients with psychosis and depression. Our findings suggest that patients show GMV and WMV reductions in a number of similar areas, with some studies suggesting that GMV loss in the insula and thalamus may have some specificity to psychosis. Studies report WMV loss in depression and psychosis, with some potential for specificity in CR and CC FA reductions to depression.
Five studies reported on GMV in depression and psychosis. Overall, they suggest that patients with psychosis and depression show reductions in GMV in the left-sided posterior subgenual prefrontal cortex, the MFG, and the hippocampus. In schizophrenia, the greatest reductions are found in the MFG. In one study, an increase in the posterior subgenual prefrontal cortex volume was found only in the depression group upon follow-up. Some of these studies reported no significant differences in GMV between the two groups: including no significant differences in the female brain, subgenual prefrontal cortex as well as one study reporting no hippocampal GMV volume differences between both groups. Four studies reported on WMV and a transdiagnostic pattern of decreased WMV in both depression and psychosis was identified. This included FA reductions and WMV decline. Two studies found significantly reduced FA specifically in depression. This suggests that both psychosis and depression patients display significantly reduced patterns of WM integrity, with some studies indicating this is more common in depression patients.
Existing literature has demonstrated that depression and psychosis share common etiologies and symptoms which indicates that both disorders may be presenting transdiagnostic mechanisms (Correll & Schooler, 2020;Upthegrove et al., 2017). The literature for amygdala and hippocampal volume in depression and psychosis is conflicting, similar to our results. Part of Keller et al. (2008) results align with a systematic review which also found no significant differences between the two disorders when examining the hippocampus (O'Connor & Agius, 2015). Keller et al. (2008) found volume reductions in the right and left amygdala whereas O' Connor and Agius (2015) found no distinctions. They suggest that reductions in amygdala volume in patients with psychosis contribute to adverse implications such as severe anxiety and fear (Keller et al., 2008). Reduced hippocampal volume has also been closely linked to psychosis in the literature (Haukvik et al., 2018), causing implications such as intensified symptom severity and impaired learning and memory functioning (Arnold et al., 2014). This supports Meisenzahl et al's (2010) findings where patients with psychosis displayed a greater reduction in hippocampal volume compared to patients with depression.
Research comparing depression and psychosis directly is limited, especially in regards to brain structure, suggesting further evidence is needed to disentangle transdiagnostic signatures and specificity of each of the two disorders. Although comorbidities of depression and psychosis are common (Correll and Schooler, 2020;Kambeitz et al, 2015;Lalousis et al., 2021;Lalousis et al., 2022), the existing literature is sparse when investigating brain structure and its impact on diagnosis, symptom severity, and prognosis. Our results highlight that patients with depression and psychosis exhibit reductions in GMV and WMV with GMV loss being more prominent in psychosis. Coryell et al. (2005), however, found that patients with psychotic depression exhibited more GMV loss than patients with schizophreniaa result going against the GMV loss hierarchy suggested by our results.
Our findings are in line with previous results showing WMV and FA disparities in both schizophrenia (Kelly et al., 2018) and depression (Jiang et al., 2017;Wise et al., 2016). Furthermore, literature has also suggested that GMV reductions have been found in both psychosis and depression, such as (Arnone et al., 2012) systematic review on depression, which found both WMV and GMV reductions in patients with depression as well as a significant amount of WMV lesions which have been associated with depression and cognitive decline. Moreover, (Steen et al., 2006) meta-analysis documented that patients with first episode psychosis experienced GMV loss, with continuous decline found over time. Subsequent systematic reviews confirmed this, with reduced GMV being found in both patients with schizoaffective disorder and patients with schizophrenia (Madre et al., 2016). A novelty in our findings is that GMV reductions were suggested to be more prevalent in psychosis, with WMV reductions being more prevalent in depression. This finding should however be approached with caution, as some studies did not follow this pattern such as Coryell et al. (2005) who found GMV reductions in both depression and schizophrenia compared to healthy controls and volume loss in the left-sided posterior subgenual prefrontal cortex being significantly lower in the depression group.
The relevance of WMV reductions in the aetiology of depression still remains in question. WMV in patients with depression has been show to increase after serotonin and norepinephrine reuptake inhibitor (SNRI) or selective serotonin reuptake inhibitor (SSRI) medication compared to recovered patients and healthy controls (Zeng et al., 2012). This implies that SSRIs and SNRIs could possibly reverse WMV reductions, suggesting that WMV may play a role in the aetiology of depression (Zeng et al., 2012).
On the contrary, medication for schizophrenia has shown to cause further impairment in brain structure Zhang et al., 2016). Studies have demonstrated how antipsychotic drug treatment can further reduce GMV as it can cause neuronal and cell complications (Zhang et al., 2016). However, it can be argued that GMV reductions may be a result of the disorder's severity and need for higher doses of medication (Zhang et al., 2016). Yet, studies show that antipsychotic medication for schizophrenia increases GMV in the thalamus, which modulates perception, thoughts, and feelings (Yue et al., 2016), resulting in improvements in positive symptoms. Structural brain imaging findings can differ between individuals with recent onset versus chronic psychiatric disorders due to a variety of factors. Age-related changes in brain development and plasticity may influence the degree and extent of structural abnormalities seen on imaging. Neurodegenerative processes may also play a role, with chronic disorders potentially leading to progressive changes in brain structure. Additionally, medication effects may differ between those with recent onset versus chronic disorders, potentially affecting brain structure differently. Als, the cumulative illness burden associated with chronic disorders may result in greater structural abnormalities compared to those with recent onset disorders. All of these factors may differentially affect brain structure depending on the duration and severity of the disorder, and on the age of the individual and should be taken into account when interpreting the findings of this systematic review.
To date, this is the first systematic review which has provided an overview of structural neuroimaging brain comparisons between depression and psychosis. Key strengths of this systematic review include that it reduced reviewer bias as its approach of validly assessing individual studies is objective and reproducible (Bero, 2006;Collins and Fauser, 2005), it used a standardised search method in multiple databases, it was pre-registered, authors were consistently cross-referencing, and efficient quality checks were performed. However, limitations included the lack of coordinate information which prevented the conduction of a meta-analysis, the lack of individual studies directly comparing depression and psychosis, lack of controls for medication exposure, small sample sizes, and heterogeneity of symptoms. Four out of the 7 studies (Coryell et al., 2005;Kochunov et al., 2013;Ota et al, 2013;Cui et al., 2020) scored 4 on the NOS guidelines, indicating low quality. Furthermore, there is a large body of literature that examines structural abnormatlities in depression and psychosis separately that could not be incorporated in the introduction. Finally, our inclusion criteria selected for studies from 2005 and 2020 in order to have as many studies as possible included in this systematic review. The advances in the technology of neuroimaging is quite fast moving and such a large time range introduces problems in the intepreptation of results. Our interepretation of the results should be viewed under these limitations.

Conclusions
We identified a pattern of GMV reductions and WMV reductions in areas such as the MFG, hippocampus, and left-sided posterior subgenual prefrontal cortex in both depression and psychosis. GMV loss was seen specifically in the insula and thalamus in psychosis. However, the lack of literature directly comparing brain structure in depression and psychosis is further heightened by the poor quality of the studies and future highquality studies, including those utilising large data sets is needed.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
No data was used for the research described in the article.

Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi. org/10.1016/j.nicl.2023.103388.