Introduction

Schizophrenia (SCZ) and major depressive disorder (MDD) are the two most common psychiatric disorders, accounting for a considerable portion of healthcare expenditure, and are major public health concerns.[1,2] Historically, SCZ and MDD have been regarded as separate disorders based on their distinct clinical presentation and course. An obvious clinical argument for this is that they are treated with pharmacologically distinct therapeutics, i.e., dopamine antagonists for SCZ[3] and enhancers of serotonin and/or noradrenaline activity for MDD.[4] However, symptoms such as psychotic or emotional symptoms, apathy, and cognitive impairment are common in both diseases.[5] Identifying commonalities and differences in pathophysiology across these disorders is important for more precise and objective diagnosis and the development of new treatments.

A growing number of studies on the relationship between SCZ and MDD have found overlap in genetic risk variants,[6,7] neuropsychological defects,[8] endophenotypic characteristics,[9] and biological foundation of susceptibility.[10] Clinical analyses have found similarities in some psychotic symptoms and high comorbidity between the diagnoses of SCZ and MDD,[11] which is clinically difficult to distinguish. These similarities raise the possibility that the neural correlation of two syndromes may exist and highlight the need to elucidate the shared and different brain abnormalities.

There is increasing interest in the trans-diagnosis of major psychiatric disorders.[5,12] Recent progresses in psychoradiology, an evolving subspecialty of radiology, have led to a better understanding of the complex brain abnormalities in neuropsychiatric disorders.[13] Although a few studies have attempted to identify the structural abnormalities of SCZ and MDD,[14,15] it is difficult to characterize the histopathological changes in vivo that underlie them. Imaging techniques that are sensitive to relatively subtle macromolecular changes can help to overcome above difficulties. Magnetization transfer imaging (MTI) has been increasingly used to evaluate macromolecular alterations in healthy brains[16] and neuropsychiatric disorders.[17,18] By utilizing the magnetization exchange between the spins of free water and water bounded to macromolecules, MTI creates a contrast between tissues, and expresses as a magnetization transfer ratio (MTR) map. MTR measures the efficiency of these exchange phenomena, which depend on the biophysics of macromolecular protein pools and their local microenvironment.[19] Previous studies have reported abnormalities in similar brain areas in the two disorders,[20,21] which could partially explain the overlap of signs and symptoms observed in clinical practice, as well as the co-occurrence of the two disorders. However, without a direct comparison, it remains unclear whether there is a true systematic overlap in the brain structural abnormalities between SCZ and MDD and how these two disorders relate to each other.

This study aims to conduct a quantitative, voxel-based meta-analytic comparison of all published whole-brain studies of MTR abnormalities in patients with SCZ and MDD to identify shared or disorder-specific MTR abnormalities between the two disorders, which could aid in understanding the pathophysiological basis of the clinical continuum of psychosis. We integrated all eligible studies reporting whole-brain MTR alterations in SCZ and/or MDD using the meta-analytic technique of anisotropic effect size-signed differential mapping (AES-SDM), which allows comparison of effect size between two disorders. In addition, we performed exploratory meta-regression analyses to find the potential influences of demographic and clinical variables on identified patterns of abnormalities.

Methods Search strategy

In accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement,[22] studies included in this meta-analysis were collected from the PubMed, EMBASE, International Scientific Index (ISI) Web of Science, and MEDLINE databases up to March 2022. The search strategies were as follows: (1) (“schizophrenia” OR “SCZ”) AND (“magnetization transfer imaging” OR “magnetization transfer ratio” OR “MTI” OR “MTR”) and (2) (“major depression” OR “depression” OR “unipolar depression” OR “depressive disorder” OR “major depressive disorder” OR “depressed”) AND (“magnetization transfer imaging” OR “magnetization transfer ratio” OR “MTI” OR “MTR”). The reference lists of retrieved articles and review articles were checked to obtain additional publications. Two independent researchers conducted the literature search. The results from these two researchers were compared, any inconsistencies were discussed until a consensus was reached.

Study selection

A study was included if (1) it was published in English with a peer review; (2) it compared groups of SCZ/MDD subjects with healthy control (HC) groups; (3) it acquired images on an MR scanner with a minimum field strength of 1.5 T and used MTR to investigate alterations in whole-brain structure; and (4) it used thresholds that were either corrected for multiple comparisons or uncorrected with spatial extent thresholds for judging statistic difference. The following types of studies were excluded: (1) studies used region-of-interest approaches; (2) case reports, letters to the editor, meta-analysis, or systematic review; (3) studies using correlation analyses between brain MTR measures and clinical variables; (4) studies did not report the results of MTR changes in Talairach/Montreal Neurological Institute (MNI) stereotactic space; and (5) duplicated studies. In cases with a sample overlap, the study with the largest sample size was selected. For studies that used multiple independent patient samples and separately compared them with the same HC sample, the results were regarded as separate datasets. If adequate information was not available in the original manuscript, we contacted the corresponding authors by e-mail to obtain additional details.

Study quality assessment

The quality (completeness) of each study was assessed using a 12-point checklist,[23] which classes items into three categories: participants (items 1–4), methods for image acquisition and analysis (items 5–10), and results and conclusions (items 11 and 12). Each item received a score of 1, 0.5, or 0 if the criteria were fully, partially, or not met, respectively [Supplementary File, https://links.lww.com/CM9/B442].

Meta-analysis

Voxel-wise meta-analyses of regional brain differences were conducted using the AES-SDM software package (http://www.sdmproject.com, version 5.15),[24,25] which has been applied in several neuropsychiatric disorders.[26] As described in detail by Carlisi et al[27] and Radua et al[28], separate analyses were performed to examine regional MTR abnormalities in each patient group relative to HC. Following this, a quantitative comparison was performed for MTR abnormalities (relative to HC) between SCZ and MDD patients by calculating the between-group difference in each voxel and using standard randomization tests to establish statistical significance. Then, a conjunction analysis examined shared abnormalities in both patient groups relative to HC; for regions based on the location of significant peak effects showing alterations in one or both of the groups, we computed the union of the P values for each group in each voxel accounting for noise.[24] A statistical threshold of P < 0.005 was used with a cluster extent of 50 voxels. In addition, we performed a subgroup meta-analysis in studies with statistical threshold correction separately in SCZ and MDD groups. The same subgroup analysis could not be performed for other methodological variables (e.g., slice thickness) due to the limited number of studies.

Jackknife sensitivity, heterogeneity, and publication bias analysis

The robustness of the results was assessed by systematic whole-brain voxel-based jackknife sensitivity analysis. The analysis was repeated over and over by discarding a different study each time, and a result was considered replicable if it is similar to the main findings of the pooled meta-analysis.[29]

The studies heterogeneity of individual clusters (the mean of the brain regions reported in the original papers) was examined using a random-effect model with Q statistics and tested with a permutation approach (P < 0.005, peak height z = 1, cluster extent = 50 voxels). The possibility of publication bias for regions showing shared alterations of MTR was examined using Egger's test.[30]

Meta-regression analysis

We used meta-regression analysis to identify the potential effects of demographic and clinical variables such as age, gender (percentage of female patients), duration of illness, medication status, Hamilton rating scale for depression (HRSD), and positive and negative syndrome scale (PANSS) on MTR abnormalities. To minimize the detection of spurious relationships, we decreased the probability threshold to 0.0005, and ignored the findings in regions other than those detected in the main analysis[28]; we also inspected the regression plots to avoid effects driven by too few studies.

Results Study selection and characteristics

Figure 1 shows the identification processes and exclusion criteria of the studies. Finally, 17 datasets from 15 studies were included in the meta-analysis: nine datasets from nine studies on SCZ[21,31–38] and eight datasets from six studies on MDD.[20,39–43] The final sample comprised 365 SCZ patients (mean age range: 24.0–37.3 years), 224 MDD patients (mean age range: 32.5–69.9 years), and 550 HC (392 from SCZ vs. HC studies, mean age range: 24.0–35.1 years; 158 from MDD vs. HC studies, mean age range: 32.0–71.8 years). Two SCZ studies[31,35] and one MDD study[43] contributed no coordinates, reporting no significant between-group differences. Two MDD studies recruited two subgroups of patients: those with and without a history of a suicide attempt[39] and those with or without medication-resistant depression.[41]Table 1 summarizes the demographic characteristics of the study participants, methodological details, and study quality scores. Supplementary Table 1, https://links.lww.com/CM9/B442 summarizes the main results of the included studies, and Supplementary Table 2, https://links.lww.com/CM9/B442 shows the details of the quality scores for each included study.

Figure 1:

Flowchart of literature search and selection criteria. MDD: Major depressive disorder; ROI: Region of interest; SCZ: Schizophrenia.

Table 1 - Demographics and clinical characteristics of studies that were included in the meta-analysis. Sample size (female) Age (years), mean (SD) PANSS Methodological aspects Study Disorder Patients Controls Patients Controls Disease duration (years) Medication status HRSD PS NS MRI scanner (Tesla) Slice thickness (mm) Multiple comparison corrections Quality scores Foong et al[33] SCZ 25 (6) 30 (8) 37.3 (6.7) 35.1 (7.2) 14.3 M NA 12.4 (4.6) 18.6 (6.2) 1.5 5 No, uncorrected P < 0.001 10.5 Bagary et al[21] SCZ 30 (11) 30 (12) 27.3 (7.4) 28.9 (5.5) 5.3 M NA 12.6 (4.4) 11.9 (6.6) 1.5 1.5 No, uncorrected P < 0.001 11.0 Antosik-Biernacka et al[31] SCZ 20 (7) 23 (10) 35.5 (6.3) 33.9 (6.8) 10.7 M NA 14.3 (5.7) 15.1 (5.8) 3 2 Yes, GRF P < 0.05 11.5 Price et al[35] SCZ 16 (4) 12 (8) 26.3 (NA) 31.6 (NA) NA M NA NA NA 1.5 1.5 Yes, FDR P < 0.05 12.0 Price et al[36] SCZ 48 (15) 47 (20) 26.2 (NA) 24.8 (NA) NA M NA 30.8 (NA) 16.3 (NA) 1.5 1.2 Yes, FWE P < 0.05 12.0 Bachmann et al[32] SCZ 32 (12) 32 (12) 33.1 (8.6) 33.5 (10.1) NA M NA NA NA 3 0.5 Yes, corrected P < 0.05 11.5 Palaniyappan et al[34] SCZ 17 (5) 20 (5) 33 (10.0) 32 (8.2) 7 M NA NA NA 7 1 Yes, FWE P < 0.05 12.0 Raghava et al[37] SCZ 34 (15) 51 (21) 25.1 (5.3) 25.3 (5.2) NA N NA 20.7 (3.9) 20.6 (7.5) 3 2 No, P < 0.05 11.0 Lei et al[38] SCZ 143 (73) 147 (74) 24.0 (8.0) 24.0 (7.0) 0.9 N NA 25.0 (6.2) 19.7 (7.5) 3 3 Yes, FWE P < 0.05 12.0 Gunning-Dixon et al[20] MDD 55 (32) 24 (19) 69.9 (6.1) 71.8 (6.1) NA NA 23.7 – – 1.5 1.25 Yes, corrected P < 0.05 11.5 Zhang et al[42] MDD 15 (5) 15 (5) 33.5 (10.2) 33.4 (10.2) 10.3 M 21.1 – – 3 3 No, uncorrected P < 0.01 11.0 Dalby et al[43] MDD 22 (15) 22 (15) 57.4 (4.6) 59.2 (7.3) NA NA NA – – 3 1.2 Yes, GRF P < 0.05 12.0 Chen et al[39]∗ MDD 19 (7) 28 (17) 38.0 (13.2) 32.0 (10.4) 9.3 F 22.6 – – 3 3 Yes, AlphaSim P < 0.05 12.0 17 (12) 28 (17) 36 (12.8) 32.0 (10.4) 3.2 F 25 – – 3 3 Yes, AlphaSim P < 0.05 12.0 Chen et al[40] MDD 27 (14) 28 (14) 33.0 (10.8) 33.0 (11.7) 6.6 F 22 – – 3 3 Yes, AlphaSim P < 0.05 12.0 Jia et al[41]∗ MDD 30 (10) 41 (23) 36.7 (12.7) 33.9 (11.9) 9.9 M 22.7 – – 3 3 Yes, AlphaSim P < 0.05 12.0 39 (24) 41 (23) 32.5 (12.6) 33.9 (11.9) 3.3 M 24.1 – – 3 3 Yes, AlphaSim P < 0.05 12.0

∗Two data sets were included. FDR: False discovery rat; FEW: Family-wise error rate; GRF: Gaussian random fields; HRSD: Hamilton rating scale for depression; MDD: Major depressive disorder; Medication status: Medication-cessation (F), on medication (M), and medication-naïve (N); MRI: Magnetic resonance image; NA: Not available; PANSS: Positive and negative syndrome scale, including positive score (PS) and negative score (NS); SCZ: Schizophrenia; SD: Standard deviation; –: Not applicable.

Pooled meta-analysis of MTR alterations SCZ vs. HC

As shown in Figure 2 and Supplementary Table 3, https://links.lww.com/CM9/B442, SCZ patients showed higher MTR than HC in the left cerebellum (hemispheric lobule VI), left thalamus, left fusiform gyrus, left paracentral lobule, left precuneus/cuneus and posterior cingulate gyrus (PCG), and lower MTR than HC in the anterior cingulate/paracingulate gyrus (ACG), right superior temporal gyrus (STG), and right superior frontal gyrus (SFG, medial part).

Figure 2:

Pooled meta-analysis of MTR alterations in SCZ and MDD, respectively. Regions of higher (red) and lower (light blue) MTR in patients with SCZ and MDD compared with HC. ACG: Anterior cingulate/paracingulate gyrus; HC: Healthy control; L: Left; MDD: Major depressive disorder; MOR: Middle occipital region; MTR: Magnetization transfer ratio; PCG: Posterior cingulate gyrus; R: Right; SCZ: Schizophrenia; SFG: Superior frontal gyrus; STG: Superior temporal gyrus.

MDD vs. HC

MDD patients showed higher MTR than HC in the left cerebellum (hemispheric lobule VIII), left fusiform gyrus, and left middle occipital region (MOR); and no areas showed a lower MTR in MDD patients than HC [Figure 2 and Supplementary Table 3, https://links.lww.com/CM9/B442]. We noted that only one study involved geriatric MDD,[20] we repeated the analysis after excluding it: the results were unchanged [Supplementary Table 4, https://links.lww.com/CM9/B442].

Subgroup analysis

In addition, the results for the subgroup meta-analysis of SCZ/MDD studies with threshold correction remained largely unchanged [Supplementary Table 5, https://links.lww.com/CM9/B442].

Conjunction analysis

As shown in Table 2 and Supplementary Figure 1, https://links.lww.com/CM9/B442, both patient groups showed significantly higher MTR in the left cerebellum (MNI coordinates: −24, −60, −40; voxels: 185) and left fusiform gyrus (MNI coordinates: −46, −56, 2; voxels: 154). Because the independent samples t tests revealed that SCZ and MDD patient groups differed in age distribution (t = 2.40; P = 0.040) and the proportion of female patients (t = 2.72; P = 0.020), we repeated the conjunction analysis with age and sex ratio as covariates and the results were maintained [Supplementary Table 6, https://links.lww.com/CM9/B442].

Table 2 - MTR alteration results of conjunction analysis of SCZ and MDD. Brain region MNI coordinates SDM P value No. of voxels∗ Cluster breakdown (No. of voxels)† X Y Z Z score Higher MTR in both SCZ and MDD than HC  L. cerebellum −24 −60 −40 1.805 0.0006 185 L. cerebellum, hemispheric lobule VI (93) L. cerebellum, hemispheric lobule VIIB (37) L. cerebellum, hemispheric lobule VIII (31)  L. fusiform gyrus −46 −56 2 1.654 0.0004 154 L. fusiform gyrus (82) L. inferior network, inferior longitudinal fasciculus (34) L. middle temporal gyrus, BA 37 (27)

∗Cluster extent threshold: 50 voxels.

†Regions with <10 voxels are not reported in the cluster breakdown. BA: Brodmann area; HC: Healthy control; L.: Left; MDD: Major depressive disorder; MNI: Montreal Neurological Institute; MTR: Magnetization transfer ratio; SCZ: Schizophrenia; SDM: Signed differential mapping.

Jackknife sensitivity, heterogeneity, and publication bias analysis

The results of the meta-analyses showed high replicability. For SCZ [Supplementary Table 7, https://links.lww.com/CM9/B442], the higher MTR in the left cerebellum and lower MTR in the ACG were preserved in all but one combination; the lower MTR in the right STG was preserved in all but two combinations; and the results in other brain areas were unchanged in all combinations. For MDD [Supplementary Table 8, https://links.lww.com/CM9/B442], the higher MTR in the left cerebellum and fusiform gyrus were preserved in all combinations, and the higher MTR was preserved in the left MOR in all but one combination. No regions with altered MTR showed significant between-study heterogeneity in either group. Egger's test did not identify any evidence of publication bias in any cluster (P > 0.05).

Meta-regression analysis

No significant association was found, at least not linear, between MTR changes and age, proportion of female patients, duration of illness, medication status, HRSD, or PANSS in either the SCZ or MDD group.

Discussion

To our knowledge, this voxel-based whole-brain structural imaging meta-analysis identified shared and disorder-specific MTR abnormalities in patients with SCZ and MDD compared with HC. Both disorders share a higher MTR in the left cerebellum and left fusiform gyrus compared with HC. SCZ patients additionally showed a higher MTR in the left thalamus, left paracentral lobule, left precuneus/cuneus and PCG, as well as a lower MTR in the ACG, right STG, and right SFG compared with HC. In MDD patients, the prominent disorder-specific findings were higher MTR in the left MOR compared with HC.

MTR mechanisms

The pathophysiology of MTR changes is not fully understood. In one aspect, it is a semiquantitative metric of macromolecule structure and properties.[44] In white matter, it is thought to reflect mainly myelin and axonal density, with lower MTR being associated with demyelination and lower axonal density; in gray matter, it reflects properties of neuronal cell membranes, with lower MTR being linked to neuronal and synaptic loss, abnormal cell membrane proteins and phospholipids.[21] However, most of our findings are increased MTR. About these findings, one interpretation is that a higher MTR reflects increased amounts of macromolecules (including myelin) that are important in axonal function.[45] It is tempting to suggest that the high MTR implies enhanced signal transfer reflecting either primary neocortical hyperexcitability or a compensation mechanism for the well-established reductions in intra- and inter-hemispheric connectivity in SCZ/MDD.[46,47] Of course, a higher communication speed does not necessarily mean a better communication efficiency.[48] Another aspect of MTR is its dependence on the T1 of tissue water[49]: more tissue free water, higher MTR. Although this could result from several pathologies (e.g., cellular atrophy, cell membrane breakdown, and reduction of dendrite numbers or cell density), some studies have linked increased extracellular water (and consequent increased MTR) to neuroinflammation.[50,51] Studies with additional modalities giving complementary information will be needed to resolve this issue.

Main findings

The conjunction analysis found that the MTR in the cerebellum was robustly higher in both diseases than that in HC. In addition to the classical role in balance and motor control, the cerebellum plays a major role in emotion regulation and cognition and is a key structure in mood regulation.[52–54] Structural and functional abnormalities of the cerebellum[55–57] and disrupted connections between the cerebellum and frontal lobe[58,59] have been described in both SCZ and MDD; and cerebellar gamma-aminobutyric acid (GABA) receptor protein expression is increased in both disorders.[60] This may underlie the overlap between the negative symptoms of SCZ and partial clinical symptoms of MDD.[11] Thus, a picture, which is related to cerebellar involvement in the pathophysiology of both conditions and involved in abnormal mood regulation processes, such as blunting of affect, withdrawal, highly emotional lability and inappropriate behavior, is building.[61,62]

Both patient groups showed higher MTR in the left fusiform gyrus, which has been implicated in high-level visual processing, particularly the processing of information related to faces and stimuli characterized by high spatial frequencies.[63] Early studies linked altered responsiveness to facial emotional stimuli to the fusiform gyrus in both SCZ and MDD.[64,65] These results suggest that neuroanatomic abnormalities of fusiform gyrus underlie at least some of the deficits associated with facial recognition in both disorders.

For SCZ, our study confirmed and extended earlier reports,[18,66] finding widespread MTR alterations, including the left precuneus, PCG, right SFG, and right temporal lobes. These are the key structures of the default mode network (DMN),[67] which has been proved implicated with SCZ in neuroimaging,[68–70] neuropathology,[71] and genomics[72] studies. The DMN is related to stimulus-independent thought and self-reflection, and greater suppression of the DMN is associated with better performance on attention-demanding tasks.[73] In SCZ, the DMN is often hyperactivated and hyperconnected, which may relate to overly intensive self-reference and impairments in attention and working memory.[74,75] Notably, we also found altered MTR in the thalamus, prefrontal regions (including the ACG and paracentral lobule) and cuneus in SCZ patients, which were consistent with the findings of our recent review that structural deficits in SCZ are most robust in thalamo-cortical networks.[76] Such abnormalities may be linked to the fundamental cognitive deficits in SCZ, which may be a useful neural target for pro-cognitive interventions.[77]

Unlike the extensive MTR abnormalities in SCZ, MDD presented a relatively limited pattern of higher MTR in the left MOR in addition to the cerebellum and fusiform gyrus. This finding was in line with previous research reporting higher perfusion in the left occipital lobe and higher occipital activation in MDD patients than in HC.[78,79] Moreover, a recent magnetic resonance spectroscopy study found that GABA concentration was altered in the occipital cortex of patients with MDD, thus pointing out the importance of the occipital cortex in MDD, including its role as a candidate biomarker.[80] The MOR is functionally similar to the fusiform gyrus and is involved in the perception of facial emotion.[81] Studies have shown that deficits in facial emotion processing may be associated with certain emotional and social symptoms in MDD patients.[82,83] Thus, the increased MTR in the MOR may be one of the neuropathological bases of MDD. However, our results failed to identify a lower MTR in the right caudate nucleus in the MDD group, in contrast to previous reports using the region of interest approach.[17,84] This discrepancy may be due to the heterogeneity of the patient population and methodological differences. Interestingly, the results of MDD were all located in the left hemisphere, which might imply that left hemisphere damage is more associated with depressive symptoms.[41] However, a recent large-scale study found that altered brain macroanatomical asymmetry is in most cases of little relevance to MDD.