Biological

Psychiatry Archival Report

Cross-Disorder Analysis of Brain Structural

Abnormalities in Six Major Psychiatric Disorders:
A Secondary Analysis of Mega- and Meta-
analytical Findings From the ENIGMA Consortium
Nils Opel, Janik Goltermann, Marco Hermesdorf, Klaus Berger, Bernhard T. Baune, and
Udo Dannlowski

ABSTRACT
BACKGROUND: Neuroimaging studies have consistently reported similar brain structural abnormalities across
different psychiatric disorders. Yet, the extent and regional distribution of shared morphometric abnormalities be-
tween disorders remains unknown.
METHODS: Here, we conducted a cross-disorder analysis of brain structural abnormalities in 6 psychiatric disorders
based on effect size estimates for cortical thickness and subcortical volume differences between healthy control
subjects and psychiatric patients from 11 mega- and meta-analyses from the ENIGMA (Enhancing Neuro Imaging
Genetics Through Meta Analysis) consortium. Correlational and exploratory factor analyses were used to quantify
the relative overlap in brain structural effect sizes between disorders and to identify brain regions with disorder-
specific abnormalities.
RESULTS: Brain structural abnormalities in major depressive disorder, bipolar disorder, schizophrenia, and
obsessive-compulsive disorder were highly correlated (r = .443 to r = .782), and one shared latent underlying factor
explained between 42.3% and 88.7% of the brain structural variance of each disorder. The observed shared
morphometric signature of these disorders showed little similarity with brain structural patterns related to
physiological aging. In contrast, patterns of brain structural abnormalities independent of all other disorders were
observed in both attention-deficit/hyperactivity disorder and autism spectrum disorder. Brain regions showing high
proportions of independent variance were identified for each disorder to locate disorder-specific morphometric
abnormalities.
CONCLUSIONS: Taken together, these results offer novel insights into transdiagnostic as well as disorder-specific
brain structural abnormalities across 6 major psychiatric disorders. Limitations comprise the uncertain contribution
of risk factors, comorbidities, and medication effects to the observed pattern of results that should be clarified by
future research.
Keywords: Cross-disorder, ENIGMA, Neuroimaging, Psychiatric disorders, Structural MRI, Transdiagnostic
https://doi.org/10.1016/j.biopsych.2020.04.027

Psychiatric disorders represent one of the worldwide leading In contrast, various independent structural neuroimaging
causes of disability, accounting for 7.4% of overall global studies have reported abnormalities in identical brain regions
disability-adjusted life years in 2010 (1). Over the last decades, such as the hippocampus across different major psychiatric
neurobiological research has aimed to provide a better un- disorders, which seems to indicate cross-disorder common-
derstanding of the etiological mechanisms of psychiatric dis- alities (6,7). These findings have raised the question of to what
orders, with the goal of enhancing preventive and therapeutic extent brain structural abnormalities might represent neurobi-
efforts (2). To this end, numerous studies have provided evi- ological traits of mental illness that are disorder specific or
dence for brain structural abnormalities in psychiatric disorders transdiagnostic.
(3). Neuroimaging studies reporting disorder-specific A previous meta-analysis by Goodkind et al. (8) directly
morphometric characteristics in major depressive disorder demonstrated overlapping gray matter reductions in the
(MDD), bipolar disorder (BD), and schizophrenia (SCZ) have cingulate cortex and the insula in MDD, BD, SCZ, obsessive-
raised hopes regarding the potential utility of structural neu- compulsive disorder (OCD), anxiety disorders, and addiction,
roimaging for future clinical applications (4,5). and suggested the existence of transdiagnostic morphometric

SEE COMMENTARY ON PAGE e41

678 ª 2020 Society of Biological Psychiatry.

Biological Psychiatry November 1, 2020; 88:678–686 www.sobp.org/journal ISSN: 0006-3223
 Biological
Cross-Disorder Analysis of Structural MRI Findings Psychiatry

abnormalities across major psychiatric disorders. Similarly, a and adults, we used the effect size estimates obtained from
meta-analysis by Wise et al. (9) confirmed shared gray matter the analysis on the adult sample in order to achieve highest
reductions in the bilateral insula, cingulate cortex, and pre- possible comparability between the study results.
frontal cortex in MDD and BD, while a single-center cross- Inclusion criteria were met for 11 studies on the following 6
disorder study reported enlarged putamen volume as a psychiatric disorders: MDD (13,14), BD (15,16), SCZ (17,18),
transdiagnostic trait in MDD, SCZ, OCD, and posttraumatic OCD (19,20), attention-deficit/hyperactivity disorder (ADHD)
stress disorder (10). (21,22), and autism spectrum disorder (ASD) (23) (Table 1). The
Yet, little is still known about the precise extent and regional published effect size estimates in these studies were Cohen’s
distribution of overlap and specificity in the morphometric d values reflecting the mean difference in each cortical or
profiles of psychiatric disorders. Addressing this question ap- subcortical region of interest after adjustment for age, sex, site
pears relevant for several reasons. First, it could help to shift (scanner), and, in the case of subcortical volume, additional
the focus of psychiatric research on brain regions that are adjustment for total intracranial volume. For all disorders,
central to disorder-specific biological processes and hence negative effect sizes represent a reduction in cortical thickness
might facilitate the discovery of etiological mechanisms. Sec- or subcortical volume in cases compared with controls.
ond, identification of shared and disorder-specific brain Because not all studies provided lateralized effect sizes,
structural signatures might enhance the increasing efforts in average effect sizes across the left and right hemispheres for
developing biomarkers for differential diagnosis or treatment each region of interest were used for the present study. In
response in psychiatric disorders. addition, because surface-area data were not available for all
Addressing this question would require comprehensive in- studies, cortical thickness and subcortical effect sizes were
vestigations of shared and unique morphometric abnormalities employed for the main analyses of the present study. This yiel-
across major psychiatric disorders throughout the entire brain, ded complete sets of k = 41 data points, including regional effect
which have been missing up to now. Recent worldwide sizes for 34 thickness measures and 7 subcortical measures for
collaborating neuroimaging efforts now provide for the first each of the 6 psychiatric disorders under investigation.
time a sufficient database for a systematic investigation of
shared and unique associations between psychiatric disor-
ders. With the present study, we therefore aimed to conduct a
cross-disorder analysis of brain structural abnormalities Data Analysis
throughout the brain, based on regional effect sizes derived SPSS version 25 (IBM Corp., Armonk, NY) was used for all
from published mega- and meta-analyses from the ENIGMA statistical analyses. Figures were created by using the Brain-
(Enhancing Neuro Imaging Genetics Through Meta Analysis) Painter visualization software (https://brainpainter.csail.mit.
consortium. We introduce a factor-analytic approach to edu) (24).
aggregate mega- and meta-analysis data, thus obtaining a First, to investigate if the 6 psychiatric disorders would
quantitative analysis of the communalities and differences of overlap in the extent and distribution of brain structural alter-
effect sizes across psychiatric disorders. In contrast to a ations across the brain, correlational analyses were conducted
qualitative review or a conventional meta-analysis, this by simultaneously correlating all k = 41 regional effect sizes of
approach investigates the potential underlying structure or cortical thickness and subcortical volumes of each disorder
pattern of regional brain structural abnormalities to quantify the with effect sizes of all respective other disorders.
extent of shared and unique brain structural abnormalities Second, to further investigate patterns of correlating brain
across major psychiatric disorders instead of focusing on structural abnormalities across the 6 disorders, an exploratory
dichotomous significance thresholds or the absolute extent of factor analysis was conducted based on the identical dataset,
effect sizes. The quantification and localization of relative employing principal component analysis with oblique rotation
conformity versus deviation from the identified patterns (see also Supplemental Methods).
furthermore allows relevant insights on the disorder specificity Third, we aimed to assess if the observed factor solution
of regional abnormalities in brain structure. was substantially confounded by systematic differences in the
age distributions between the psychiatric samples on which
METHODS AND MATERIALS effect sizes were based. Therefore, we investigated the overlap
between shared brain structural patterns in psychiatric disor-
Study Selection ders and brain structural correlates of physiological aging. To
For the present study, regional effect sizes of published this end, we computed additional regional brain structural ef-
structural neuroimaging mega- and meta-analyses of the fect sizes for age based on 2 cohorts of healthy subjects
ENIGMA consortium were analyzed. Inclusion criteria were derived from the Human Connectome Project (25,26)
defined as ENIGMA studies that had carried out structural (N = 1113; mean age, 28.80 years [range, 22–37 years]) and the
neuroimaging investigations of group comparisons between BiDirect Study (27,28) (N = 431; mean age, 52.15 years [range,
healthy control subjects and patients with a psychiatric dis- 35–65 years]) (Supplemental Methods). Both samples are
order. We only included data from psychiatric disorders for completely independent from all other samples on which effect
which effect size estimates for all 34 cortical brain regions sizes of the psychiatric disorders were based. In order to
based on the Desikan-Killiany atlas (11) and for all 7 subcortical investigate the similarity of brain structural patterns of physi-
regions included in the standardized FreeSurfer imaging ological aging with the observed latent structure of brain
pipeline (12) used by the ENIGMA consortium were available. If structural abnormalities in psychiatric disorders in our analysis,
studies reported separate analyses for children/adolescents we added effect sizes for age to our factor analysis (separately

Biological Psychiatry November 1, 2020; 88:678–686 www.sobp.org/journal 679

Biological
Psychiatry Cross-Disorder Analysis of Structural MRI Findings

Table 1. Overview of Included Studies With Details on the Availability of Subcortical and Cortical Results, Number of
Participants, and the Respective Analytic Design (Meta- or Mega-analysis)
Study Disorder Subcortical Cortical Cases Controls Meta- vs. Mega-analysis
Schmaal et al. (13) MDD x 1728 7199 Meta
Schmaal et al. (14) MDD x 1902 7658 Meta
Hibar et al. (16) BD x 1710 2594 Meta
Hibar et al. (15) BD x 1837 2582 Mega
van Erp et al. (18) SCZ x 2028 2540 Meta
van Erp et al. (17) SCZ x 4474 5098 Meta
Boedhoe et al. (20) OCD x 1495 1472 Both
Boedhoe et al. (19) OCD x 1498 1436 Both
Hoogman et al. (22) ADHD x 515 422 Mega
Hoogman et al. (21) ADHD x 733 539 Mega
van Rooij et al. (23) ASD x x 1658 1606 Both
ADHD, attention-deficit/hyperactivity disorder; ASD, autism spectrum disorder; BD, bipolar disorder; MDD, major depressive disorder; OCD,
obsessive-compulsive disorder; SCZ, schizophrenia.

for age effect sizes derived from the Human Connectome positive correlation between ADHD and MDD, and a weak
Project and BiDirect study samples). negative correlation between ASD, BD, and OCD, emerged
Fourth, we further aimed to assess the regional distribu- (Figure 1, Table S1).
tion of shared and independent variance in brain structural
abnormalities. To address this question, we assessed to Exploratory Factor Analysis of Brain Structural Data
what extent regional morphometric effect sizes of a disorder The factor analysis yielded 3 latent factors (eigenvalues: F1 =
would deviate from the latent factors. This was done using 2.897, F2 = 1.232, F3 = 0.979) (Figure 2A, C) and confirmed the
linear regression analyses in which the true regional effect pattern of results observed in the correlation analysis. While
size of a disorder was predicted by the respective regional MDD, BD, SCZ, and OCD all loaded onto one shared latent
factor score. The regional residuals of these regression an- factor that explained 48.29% of variance of brain structural
alyses represent the difference between the true effect size effect sizes in all disorders, ADHD and ASD loaded onto
and the predicted regional effect size based on the shared separate unique factors (Figure 2A, D). The 3 extracted latent
factor score. Higher absolute residuals indicate less optimal factors were virtually not correlated with each other and
representation of the true effect size based on the shared explained a total variance of 85.14% of the regional brain
latent factor, and thus indicate a lower degree of shared structural effect sizes (Figure 2A). Next, we calculated regional
variance between the common factor and the respective factor scores for each of the 3 factors to identify the contri-
disorder in a specific region (Supplemental Methods). Cor- bution of each brain structure to the observed factor solution.
relation analyses of absolute residuals between disorders The strongest contributors were the hippocampus and the
were conducted to systematically assess if overall regional fusiform gyrus for F1, the rostral anterior cingulate cortex
deviation from the observed factor solution was driven by (ACC) and the amygdala for F2, and the entorhinal cortex and
similar brain regions across disorders. Then, correlation an- the fusiform gyrus for F3 (Figure 2B, Table S2). Additional
alyses of raw residuals were conducted to assess the di- analyses investigating the relationship between the factor
rection of deviation across brain regions between the scores and the respective original effect sizes for each region
disorders. of interest indicated that while F1 was driven mostly by strong
Fifth, additional confirmatory analyses were conducted negative effect sizes and near-to-null effect sizes, F2 and F3
based on differing subsets of imaging data entities, and with were driven by the largest positive as well as largest negative
effect sizes for medicated and nonmedicated patients sepa- effect sizes (Table S3, Figure S1).
rately, in order to account for potential bias resulting from the
specific choice of analyzed morphometric parameters or from Overlap Between Brain Structural Patterns Related
the presence of medication. to Psychiatric Disorders and to Physiological Aging
Adding the calculated effect sizes for age (Table S4) to the 6
RESULTS psychiatric disorders in a factor analysis again yielded 3
prominent latent factors (Table 2, Table S5). The previous
Comparison of Brain Structural Abnormalities factor structure was confirmed in this analysis, with MDD, BD,
Across Psychiatric Disorders SCZ, and OCD loading on F1, while ASD loaded mainly on F2
We observed medium-to-strong positive correlations (r = and ADHD loaded strongly on F3. Age loaded moderately on
.443 to r = .782) of regional effect size estimates between a F1 and showed strong negative loadings with F2. Factor so-
set of 4 (MDD, BD, SCZ, and OCD) out of all 6 psychiatric lutions did not differ depending on the two samples on which
disorders (MDD, BD, SCZ, and OCD), with the most pro- age effect sizes were based.
nounced correlations between BD and SCZ (r = .782) and A similar factor solution was observed in a factor analysis
between OCD and SCZ (r = .745). Furthermore, a weak that only included effect sizes of disorders that loaded on F1

680 Biological Psychiatry November 1, 2020; 88:678–686 www.sobp.org/journal

 Biological
Cross-Disorder Analysis of Structural MRI Findings Psychiatry

Figure 1. Graphical depiction of correlations

A B

ADHD
MDD

OCD

ASD
SCZ
BD
1 between effect sizes of brain structural abnor-
malities of 6 psychiatric disorders. Effect sizes
MDD

1
0.8
comprise regional cortical thickness and
0.6
subcortical volume alterations averaged across
hemispheres. Correlations are depicted (A) as a
BD

0.4 scatter matrix and (B) using the CorrPlot func-

tion in R software, with the color bar displaying
SCZ

0.2
3
the correlation coefficient (r). ADHD, attention-
0 deficit/hyperactivity disorder; ASD, autism
spectrum disorder; BD, bipolar disorder; MDD,
OCD

4
0.2
major depressive disorder; OCD, obsessive-
0.4
compulsive disorder; SCZ, schizophrenia.
ADHD

0.6
ASD

0.8
6

1
MDD BD SCZ OCD ADHD ASD

(MDD, BD, SCZ, OCD) and effect sizes for age: age loaded SCZ, and OCD) showing high proportions of overall shared
weakly on the first and highly on the second latent factor. variance by analyzing residuals as a measure of regional
Effect sizes of the psychiatric disorders loaded either nega- deviation of original effect sizes from the shared latent factor
tively or weakly positively (BD) on the second factor, and the F1 (Figure 3). The full results of this analysis step are pro-
two factors were weakly correlated (r = .133) (Table 2). vided in Table S6.
Correlation analyses of the resulting residuals revealed
that MDD and SCZ exhibited a pronounced degree of
Identification of Disorder-Specific Morphometric overlap in the regional deviation from the common latent
Abnormalities factor, as represented by a medium-sized positive corre-
We aimed to assess the regional distribution of shared and lation of absolute residuals (r = .380). Yet, the directions of
disorder-specific variance for the 4 disorders (MDD, BD, deviation in MDD and SCZ were opposed to one another,

A B

C D

Figure 2. Results of the exploratory factor analysis with (A) a diagram displaying the structure of the factor solution, with factor loadings displayed in
between factors 1 and 3 (F1–F3) and between the respective disorder and extracted variance (ExtrV), and correlation among the 3 factors; (B) regional factor
scores mapped on the cortex and on subcortical structures; (C) a scree plot displaying number of components and respective eigenvalues; and (D) a bar graph
displaying the factor loadings on F1–F3 of each disorder. ADHD, attention-deficit/hyperactivity disorder; ASD, autism spectrum disorder; BD, bipolar disorder;
MDD, major depressive disorder; OCD, obsessive-compulsive disorder; SCZ, schizophrenia.

Biological Psychiatry November 1, 2020; 88:678–686 www.sobp.org/journal 681

Biological
Psychiatry Cross-Disorder Analysis of Structural MRI Findings

Table 2. Factor Structure of the FA, Including Regional Effect Sizes of Psychiatric Disorders and Age
FA of All Disorder and Age Effect Sizes FA of MDD, BD, SCZ, OCD, and Age Effect Sizes
Component Loading Component Loading
F1 F2 F3 Communalities F1 F2 Communalities
Age Effect Sizes Based on the HCP Sample
MDD 0.649a 0.278 0.526 0.775 0.686a 20.256 0.536
BD 0.915 a
20.073 20.111 0.855 0.904a 0.264 0.887
SCZ 0.855a 0.303 20.200 0.863 0.889a 20.117 0.804
OCD 0.841a 0.255 20.034 0.774 0.856a 20.238 0.790
ADHD 20.054 20.058 0.945a 0.900 – – –
ASD 20.389 0.762a 0.030 0.733 – – –
Age 0.354 20.831a 0.065 0.821 0.258 0.944a 0.957
Eigenvalue 2.977 1.515 1.228 – 2.877 1.096 –
% of Variance 42.526 21.643 17.549 – 57.532 21.923 –
Age Effect Sizes Based on the BiDirect Study Sample
MDD 0.691a 0.426 0.421 0.836 0.719a 20.438 0.709
BD 0.886a 20.185 20.067 0.824 0.868a 0.259 0.819
SCZ 0.886a 0.031 20.218 0.834 0.895a 0.148 0.824
OCD 0.872 a
20.041 20.007 0.762 0.867a 0.129 0.768
ADHD 20.073 0.153 0.941a 0.915 – – –
ASD 20.282 0.714a 20.208 0.633 – – –
Age 20.106 20.848a 0.213 0.775 20.164 0.933a 0.898
Eigenvalue 2.904 1.470 1.204 – 2.850 1.168 –
% of Variance 41.489 21.005 17.204 – 57.005 23.358 –
ADHD, attention-deficit/hyperactivity disorder; ASD, autism spectrum disorder; BD, bipolar disorder; F, factor; FA, factor analysis; HCP, Human
Connectome Project; MDD, major depressive disorder; OCD, obsessive-compulsive disorder; SCZ, schizophrenia.
a
Component loading ..60 (58).

as represented by a strong negative correlation between Confirmatory Analyses

raw residuals (r = 2.564). A similar pattern of results was The reported factor structure could be replicated in analyses
found for correlation of residuals between BD and OCD including only cortical thickness data and only subcortical
and between SCZ and OCD, indicating that deviations volume data, respectively (Supplemental Results, Tables S8
from the common factor solution between these disorders and S9) as well as in analyses including laterality and surface
are present in similar brain regions but in opposing di- information that were only available for MDD, BD, SCZ, and
rections (Table S7). OCD (Tables S10 and S11).
For MDD, the largest absolute residuals above the 95th Additional analyses based on effect size estimates from
percentile were found in the rostral ACC (residual analyses accounting for medication intake in a subset of
[res] = 20.081, standardized residual [std-res] = 22.147) and studies including MDD, SCZ, and OCD indicated a similar
the medial orbitofrontal cortex (OFC) (res = 20.067, std- factor structure, with one single extracted latent factor
res = 21.786), where original effect sizes were more pro- explaining over 50% of total variance (Table S12).
nounced than predicted based on the shared factor score. In
contrast, for BD, regions with the largest absolute residuals
were in the parahippocampal gyrus (res = 0.101, std-res = DISCUSSION
2.631), for which predicted effect sizes were overestimated, The present study provides a comprehensive cross-disorder
and the pallidum (res = 20.078, std-res = 22.038), for which analysis of brain structural abnormalities throughout subcor-
a positive effect size was predicted based on the latent tical and cortical regions in 6 major psychiatric disorders based
shared factor, but a negative effect size was observed. For on the largest imaging database currently available. We found
SCZ, the largest residuals were observed in the superior that brain structural abnormalities of MDD, BD, SCZ, and OCD
temporal gyrus (res = 20.138, std-res = 22.325), for which can be traced back to one single latent factor that explained
effect sizes were underestimated, and the medial OFC (res = the predominant share of variance of brain structural alter-
0.136, std-res = 22.301), for which the predicted effect size ations in these disorders. Our findings thus suggest the pres-
was overestimated. The largest residuals for OCD were found ence of a shared morphometric signature in these 4 major
in the pallidum (res = 0.067, std-res = 2.508) and the superior psychiatric disorders. By contrast, ADHD and ASD appear to
temporal gyrus (res = 0.056, std-res = 2.095). be characterized by brain structural correlates that were found

682 Biological Psychiatry November 1, 2020; 88:678–686 www.sobp.org/journal

 Biological
Cross-Disorder Analysis of Structural MRI Findings Psychiatry

as a potential contributing factor to gray matter reductions

in psychiatric disorders might contribute to the current
findings (29).
While we are not aware of any other study based on a
similar approach compared with ours, it appears relevant to
consider findings from a previous voxel-based meta-analysis
of structural neuroimaging studies across MDD, BD, SCZ,
OCD, anxiety, and addiction that reported shared gray matter
reductions in the bilateral insula and the dorsal ACC (8). In
analogy, we found strong negative factor scores for the insula
for the latent factor that seems to commonly represent brain
structural abnormalities of MDD, BD, SCZ, and OCD. Yet, it
appears important to note that converging negative effect
sizes for hippocampal volume and thickness of the fusiform
gyrus across MDD, BC, SCZ, and OCD were the most prom-
inent drivers of shared brain structural abnormalities in the
present study. Again, it appears relevant to highlight that in
contrast to a conventional meta-analysis, the factor-analytic
approach presented here investigates the underlying structure
of regional effect sizes across disorders, rather than their ab-
solute extent, which might explain the aforementioned
differences.
Regions such as the hippocampus and the fusiform gyrus
that show high conformity with the shared morphometric
signature of MDD, BD, SCZ, and OCD through converging
pronounced negative effect sizes across all disorders might be
of particular interest for future investigation of transdiagnostic
processes.
The observed pattern of results raises questions regarding
Figure 3. Regional residuals from the linear regression analysis of factor potential transdiagnostic causes and mechanisms underlying
scores F1 on original (true) effect sizes for each disorder are mapped on the our findings. Of note, converging evidence of a shared genetic
cortex and on subcortical structures. The absolute size of residuals repre-
pattern across mood and psychotic disorders (MDD, BD, SCZ)
sents the degree of representation through the shared latent factor, factor 1.
Regions with high absolute residuals are thus not well represented by the that differs from genetic correlates found in early develop-
shared latent factor. Lower (negative) residuals represent underestimation mental disorders (ADHD, ASD) appears to correspond to the
and higher (positive) residuals represent overestimation based on the factor brain structural findings in the present article (30,31). Com-
score. BD, bipolar disorder; MDD, major depressive disorder; OCD, munalities and differences in genetic variants across psychi-
obsessive-compulsive disorder; SCZ, schizophrenia. atric disorders might thus contribute to the pattern of shared
and distinct morphometric abnormalities that we observed, a
notion that is further supported by evidence for central nervous
to be largely independent from abnormalities in all other psy- system enrichment of shared genetic variants for the most
chiatric disorders under investigation. Moreover, we identified highly correlated disorders (BD and SCZ) (30) as well as by
disorder-specific morphometric alterations for MDD, BD, SCZ, robust evidence from multigenerational register studies
and OCD that substantially deviated from the identified shared (32,33). Moreover, recent large-scale evidence highlights that
latent factor. 26% of variation in cortical thickness is explained by common
While previous research has suggested overlapping brain genetic variants (34). In the same study, variation in cortical
structural correlates of MDD, BD, SCZ, and OCD, the extent of thickness was linked to neural processes such as myelination,
shared variance observed in the present study, with one branching, and pruning, which might thus represent candidate
shared latent factor explaining between 42.3% and 88.7% mechanisms potentially underlying overlapping thickness al-
variation in brain structural abnormalities of each disorder, is terations in psychiatric disorders (34). Yet, this notion remains
surprising. speculative at the moment and needs to be addressed by
Importantly, the aforementioned shared pattern of brain future research.
structural abnormalities in psychiatric disorders showed only Besides shared genetic factors, it appears highly relevant to
little overlap with brain structural alterations related to physi- consider shared environmental factors as potential contribu-
ological aging. Hence, it appears unlikely that our finding of a tors to the observed pattern of correlated brain structural ab-
shared morphometric signature in MDD, BC, SCZ, and OCD normalities. Stressful life events and, especially, early-life
can be explained by mere effects of physiological aging. adversity have been shown to act as risk factors for affective
However, it is noteworthy that the presented analytic approach and psychotic disorders, with little specificity for a single dis-
does not allow the inference on potential effects of patholog- order (35–38). Corresponding to the latter notion, regional brain
ical or accelerated aging. It thus remains unclear if structural alterations have been shown to mediate the pro-
abnormal brain aging that has recently gained much attention spective association between early-life trauma and

Biological Psychiatry November 1, 2020; 88:678–686 www.sobp.org/journal 683

Biological
Psychiatry Cross-Disorder Analysis of Structural MRI Findings

psychopathology across a range of psychiatric disorders difficulties (55). The specific developmental impact may ac-
including depression, substance use, antisocial behavior, and count for patterns of brain structural alterations distinct from
trait anxiety (39–43). Moreover, low-grade inflammation and the other disorders. Although there are many examples in the
hypothalamic-pituitary-adrenal axis deviation have been literature pointing toward an overlap between ADHD and ASD
shown to be present across diagnostic groups and therefore in occurrence, clinical characteristics, and etiological factors
represent further candidate mechanisms that might mediate (55), it seems that these similarities are not mirrored by over-
the development of common brain structural volume decline in lapping patterns of brain structural alterations. In this regard,
psychiatric disorders (44–46). our results are consistent with previous findings of generally
Besides these mechanistic considerations, it appears rele- distinct patterns of structural alterations between ADHD and
vant to consider that shared comorbidities and psychopha- ASD (56).
rmacological treatment have been shown to impact brain Limitations might stem from the remaining heterogeneity of
structure (47–49) and might thus constitute further explanatory studies. First, studies differ in certain methodological aspects,
factors leading to the observed result. Specifically, while an- with some studies using a meta-analytical design and others a
tidepressant treatment has been linked to hippocampal gray mega-analytical design. Yet, an ENIGMA study has directly
matter increase in patients with MDD (48), treatment with an- compared both methods and concluded that effect size esti-
tipsychotics has been associated with gray matter volume loss mates between both approaches are largely comparable (57).
in psychosis (50,51); therefore, systematic differences in the Furthermore, while a similar factor structure was confirmed in
distribution of past and present pharmacological treatment additional analyses accounting for present intake of psycho-
between the investigated disorders might contribute to the tropic medication in MDD, SCZ, and OCD, the lack of avail-
observed results. ability of similar data in BD, ADHD, and ASD limits results from
Despite the overall large proportion of shared brain struc- this additional analysis. Moreover, presence of psychiatric
tural correlates between the studied disorders, we identified comorbidities might have introduced further bias to our results.
regional differences in the extent to which abnormalities ENIGMA studies reporting on sensitivity analyses with sub-
overlapped across psychiatric disorders. Interestingly, brain samples without comorbidities do not point to an overall
regions frequently implicated in previous psychiatric research strong influence by comorbidities (13,14,22,23). Yet, it remains
such as the medial OFC and the rostral ACC strongly deviated to be investigated if the overlap in brain structural alterations
from the identified shared latent factor driven by underrepre- across disorders that we observed may partly reflect similar-
sentation in MDD, for which the extent of thickness decrease in ities between or co-occurrences of clinical phenotypes.
these regions was more than 50% higher than predicted by the Another potential limitation may stem from partial overlap
shared factor. Our results could thus point to an MDD-specific between healthy control samples that were used across
relevance of the medial OFC and the rostral ACC, both of studies. Even though the factor structure reported in this work
which have been implicated in MDD-relevant cognitive pro- does not appear to ultimately reflect the extent of sample
cesses such as in self-referential processing of emotional overlap, as, for example, MDD has very small overlap with the
stimuli (52). Interestingly, Howard et al. (53) reported significant control samples of BD, SCZ, and OCD while being highly
enrichment of MDD-related genetic variants in the ACC and represented by the common factor that we report, we
the frontal cortex; hence, it could be speculated that enrich- acknowledge that partial overlap in healthy control samples
ment of disorder-specific genetic variants might contribute to could have biased the present findings to a limited extent.
the formation of these disorder-specific brain structural ab- Future large-scale mega- and meta-analyses should be
normalities in MDD. encouraged to carefully consider and to report overlap in
Regarding SCZ, a potentially disorder-specific region of healthy control populations with previous studies in a trans-
importance appears to be the superior temporal gyrus, for parent manner to ease comparability.
which effect sizes were much larger in SCZ than implicated by Furthermore, it is important to note that generalizability of
the common factor. Again, this notion appears plausible, our findings to psychiatric disorders not included in the present
considering the important role of the superior temporal gyrus in study cannot be claimed. Generalizability may be particularly
clinical core features of SCZ, including hallucination (54). questionable with regard to so-called externalizing disorders
Importantly, these results might be used to inform future that are underrepresented in our analyses. In addition, we state
research aiming at the development of disorder-specific bio- that future voxel-based morphometry studies applying a
markers based on brain structural data. Brain regions showing similar factor-analytic approach are warranted to delineate
the least overlap with the detected shared morphometric communalities and differences between psychiatric disorders
signature might thus represent optimal candidates for neuro- with higher spatial resolution.
biologically guided differential diagnosis of psychiatric To conclude, the present study provides evidence for
disorders. shared and unique patterns of regional morphometric char-
Last, it appears important to acknowledge that brain acteristics among 6 major psychiatric disorders. While MDD,
structural abnormalities in ASD and ADHD were to a great BD, SCZ, and OCD exhibited a surprisingly high level of shared
extent independent of the common factor aggregating brain variance in brain structural abnormalities, the morphometric
structural abnormalities in MDD, BD, SCZ, and OCD. Impor- properties of ADHD and ASD largely differed from all other
tantly, as opposed to all other disorders under investigation, disorders. The reported regional distribution of shared and
ADHD and ASD are considered to be developmental disorders. unique variance in brain structure across disorders differenti-
These are characterized by neurodevelopmental deviations ates transdiagnostic from disorder-specific morphometric ab-
early in life that are accompanied by behavioral and social normalities and could thus be of great value for future studies

684 Biological Psychiatry November 1, 2020; 88:678–686 www.sobp.org/journal

 Biological
Cross-Disorder Analysis of Structural MRI Findings Psychiatry

focusing on biomarker development based on structural im- pathophysiology of schizophrenia: A selective review and hypothesis
aging data. for early detection and intervention. Mol Psychiatry 23:1764–1772.
8. Goodkind M, Eickhoff SB, Oathes DJ, Jiang Y, Chang A, Jones-
Hagata LB, et al. (2015): Identification of a common neurobiological
ACKNOWLEDGMENTS AND DISCLOSURES substrate for mental illness. JAMA Psychiatry 72:305–315.
9. Wise T, Radua J, Via E, Cardoner N, Abe O, Adams TM, et al. (2017):
The BiDirect study is funded by German Federal Ministry of Education and
Common and distinct patterns of grey-matter volume alteration in
Research Grant Nos. 01ER0816, 01ER1506, and 01ER1205. Data were
major depression and bipolar disorder: Evidence from voxel-based
provided in part by the Human Connectome Project, WU-Minn Consortium
meta-analysis. Mol Psychiatry 22:1455–1463.
(Grant No. 1U54MH091657; principal investigators, David Van Essen and
10. Gong Q, Scarpazza C, Dai J, He M, Xu X, Shi Y, et al. (2019):
Kamil Ugurbil) funded by the 16 National Institutes of Health Institutes and
A transdiagnostic neuroanatomical signature of psychiatric illness.
Centers that support the National Institutes of Health Blueprint for Neuro-
Neuropsychopharmacology 44:869–875.
science Research; and by the McDonnell Center for Systems Neuroscience
11. Desikan RS, Ségonne F, Fischl B, Quinn BT, Dickerson BC, Blacker D,
at Washington University in St. Louis. NO was supported by the Interdis-
et al. (2006): An automated labeling system for subdividing the human
ciplinary Centre for Clinical Research (IZKF) of the medical faculty of
cerebral cortex on MRI scans into gyral based regions of interest.
Münster (Grant No. SEED 11/18). These funders had no role in designing the
NeuroImage 31:968–980.
study; collection, management, analysis, and interpretation of data; writing
12. Fischl B, Salat DH, Busa E, Albert M, Dieterich M, Haselgrove C, et al.
of the report; nor the decision to submit the report for publication.
(2002): Whole brain segmentation: Automated labeling of neuroana-
NO, JG, BTB, and UD were responsible for study concept and design.
tomical structures in the human brain. Neuron 33:341–355.
NO, JG, MH, KB, BTB, and UD were responsible for acquisition, analysis, or
13. Schmaal L, Veltman DJ, van Erp TGM, Sämann PG, Frodl T,
interpretation of data. NO and JG were responsible for drafting of the
Jahanshad N, et al. (2016): Subcortical brain alterations in major
manuscript. BTB, UD, MH, and KB were responsible for critical revision of
depressive disorder: Findings from the ENIGMA Major Depressive
the manuscript for important intellectual content. NO and JG were respon-
Disorder working group. Mol Psychiatry 21:806–812.
sible for statistical analysis. KO obtained funding. NO and UD were
14. Schmaal L, Hibar DP, Sämann PG, Hall GB, Baune BT, Jahanshad N,
responsible for study supervision.
et al. (2017): Cortical abnormalities in adults and adolescents with
The authors report no biomedical financial interests or potential conflicts
major depression based on brain scans from 20 cohorts worldwide in
of interest.
the ENIGMA Major Depressive Disorder Working Group. Mol Psychi-
atry 22:900–909.
ARTICLE INFORMATION 15. Hibar DP, Westlye LT, Doan NT, Jahanshad N, Cheung JW,
Ching CRK, et al. (2018): Cortical abnormalities in bipolar disorder: An
From the Department of Psychiatry (NO, JG, BTB, UD), Institute of Epide- MRI analysis of 6503 individuals from the ENIGMA Bipolar Disorder
miology and Social Medicine (MH, KB), and Interdisciplinary Centre for Working Group. Mol Psychiatry 23:932–942.
Clinical Research (NO), University of Münster, Münster, Germany; and the 16. Hibar DP, Westlye LT, van Erp TGM, Rasmussen J, Leonardo CD,
Department of Psychiatry (BTB), Melbourne Medical School, University of Faskowitz J, et al. (2016): Subcortical volumetric abnormalities in bi-
Melbourne; and the Florey Institute of Neuroscience and Mental Health polar disorder. Mol Psychiatry 21:1710–1716.
(BTB), University of Melbourne, Parkville, Victoria, Australia. 17. van Erp TGM, Walton E, Hibar DP, Schmaal L, Jiang W, Glahn DC,
NO and JG contributed equally to this work as joint first authors. et al. (2018): Cortical brain abnormalities in 4474 individuals with
BTB and UD contributed equally to this work as joint senior authors. schizophrenia and 5098 control subjects via the Enhancing Neuro
Address correspondence to Nils Opel, M.D., at n_opel01@uni- Imaging Genetics Through Meta Analysis (ENIGMA) Consortium. Biol
muenster.de. Psychiatry 84:644–654.
Received Jan 21, 2020; revised and accepted Apr 30, 2020. 18. van Erp TGM, Hibar DP, Rasmussen JM, Glahn DC, Pearlson GD,
Supplementary material cited in this article is available online at https:// Andreassen OA, et al. (2016): Subcortical brain volume abnormalities
doi.org/10.1016/j.biopsych.2020.04.027. in 2028 individuals with schizophrenia and 2540 healthy controls via
the ENIGMA consortium. Mol Psychiatry 21:547–553.
19. Boedhoe PSW, Schmaal L, Abe Y, Alonso P, Ameis SH, Anticevic A,
REFERENCES et al. (2018): Cortical abnormalities associated with pediatric and adult
1. Whiteford HA, Degenhardt L, Rehm J, Baxter AJ, Ferrari AJ, obsessive-compulsive disorder: Findings from the ENIGMA Obsessive-
Erskine HE, et al. (2013): Global burden of disease attributable to Compulsive Disorder Working Group. Am J Psychiatry 175:453–462.
mental and substance use disorders: Findings from the Global Burden 20. Boedhoe PSW, Schmaal L, Abe Y, Ameis SH, Arnold PD,
of Disease Study 2010. Lancet 382:1575–1586. Batistuzzo MC, et al. (2017): Distinct subcortical volume alterations in
2. Linden DEJ (2012): The challenges and promise of neuroimaging in pediatric and adult OCD: A worldwide meta- and mega-analysis. Am J
psychiatry. Neuron 73:8–22. Psychiatry 174:60–69.
3. Falkai P, Schmitt A, Andreasen N (2018): Forty years of structural brain 21. Hoogman M, Muetzel R, Guimaraes JP, Shumskaya E, Mennes M,
imaging in mental disorders: Is it clinically useful or not? Dialogues Clin Zwiers MP, et al. (2019): Brain imaging of the cortex in ADHD: A co-
Neurosci 20:179–186. ordinated analysis of large-scale clinical and population-based sam-
4. Redlich R, Almeida JJR, Grotegerd D, Opel N, Kugel H, Heindel W, ples. Am J Psychiatry 176:531–542.
et al. (2014): Brain morphometric biomarkers distinguishing unipolar 22. Hoogman M, Bralten J, Hibar DP, Mennes M, Zwiers MP, Schweren LSJ,
and bipolar depression: A voxel-based morphometry-pattern classifi- et al. (2017): Subcortical brain volume differences in participants with
cation approach. JAMA Psychiatry 71:1222–1230. attention-deficit/hyperactivity disorder in children and adults: A cross-
5. Koutsouleris N, Meisenzahl EM, Borgwardt S, Riecher-Rössler A, sectional mega-analysis. The lancet Psychiatry 4:310–319.
Frodl T, Kambeitz J, et al. (2015): Individualized differential diagnosis of 23. van Rooij D, Anagnostou E, Arango C, Auzias G, Behrmann M,
schizophrenia and mood disorders using neuroanatomical biomarkers. Busatto GF, et al. (2018): Cortical and subcortical brain morphometry
Brain 138:2059–2073. differences between patients with autism spectrum disorder and
6. Arnone D, McIntosh AM, Ebmeier KP, Munafò MR, Anderson IM healthy individuals across the lifespan: Results from the ENIGMA ASD
(2012): Magnetic resonance imaging studies in unipolar depression: Working Group. Am J Psychiatry 175:359–369.
Systematic review and meta-regression analyses. Eur Neuro- 24. Marinescu RV, Eshaghi A, Alexander DC, Golland P (2019): Brain-
psychopharmacol 22:1–16. Painter: A software for the visualisation of brain structures, biomarkers
7. Lieberman JA, Girgis RR, Brucato G, Moore H, Provenzano F, and associated pathological processes. arXiv. https://arxiv.org/abs/
Kegeles L, et al. (2018): Hippocampal dysfunction in the 1905.08627.

Biological Psychiatry November 1, 2020; 88:678–686 www.sobp.org/journal 685

Biological
Psychiatry Cross-Disorder Analysis of Structural MRI Findings

25. Van Essen DC, Smith SM, Barch DM, Behrens TEJ, Yacoub E, 43. Gorka AX, Hanson JL, Radtke SR, Hariri AR (2014): Reduced hippo-
Ugurbil K (2013): The WU-Minn Human Connectome Project: An campal and medial prefrontal gray matter mediate the association
overview. Neuroimage 80:62–79. between reported childhood maltreatment and trait anxiety in adult-
26. Van Essen DC, Ugurbil K, Auerbach E, Barch D, Behrens TEJ, hood and predict sensitivity to future life stress. Biol Mood Anxiety
Bucholz R, et al. (2012): The Human Connectome Project: A data Disord 4:12.
acquisition perspective. Neuroimage 62:2222–2231. 44. Pinto JV, Moulin TC, Amaral OB (2017): On the transdiagnostic nature
27. Teismann H, Wersching H, Nagel M, Arolt V, Heindel W, of peripheral biomarkers in major psychiatric disorders: A systematic
Baune BT, et al. (2014): Establishing the bidirectional relationship review. Neurosci Biobehav Rev 83:97–108.
between depression and subclinical arteriosclerosis–rationale, 45. Frodl T, O’Keane V (2013): How does the brain deal with cumulative
design, and characteristics of the BiDirect Study. BMC Psychiatry stress? A review with focus on developmental stress, HPA axis
14:174. function and hippocampal structure in humans. Neurobiol Dis 52:24–
28. Teuber A, Sundermann B, Kugel H, Schwindt W, Heindel W, 37.
Minnerup J, et al. (2016): MR imaging of the brain in large cohort 46. Wei Y, Zhu Y, Womer FY, Duan J, Yin Z, Zhang R, et al. (2019): In-
studies: Feasibility report of the population- and patient-based BiDir- flammatory cytokines mediate brain dysconnectivity and cognitive
ect study. Eur Radiol 27:231–238. deficits across major psychiatric disorders. Available at SSRN:
29. Cole JH, Marioni RE, Harris SE, Deary IJ (2019): Brain age and other https://ssrn.com/abstract=3436249 or http://dx.doi.org/10.2139/
bodily ‘ages’: Implications for neuropsychiatry. Mol Psychiatry 24:266– ssrn.3436249. Accessed August 12, 2019.
281. 47. Dannlowski U, Stuhrmann A, Beutelmann V, Zwanzger P, Lenzen T,
30. Cross-Disorder Group of the Psychiatric Genomics Consortium, Grotegerd D, et al. (2012): Limbic scars: Long-term consequences of
Lee SH, Ripke S, Neale BM, Faraone SV, Purcell SM, et al. (2013): childhood maltreatment revealed by functional and structural magnetic
Genetic relationship between five psychiatric disorders estimated from resonance imaging. Biol Psychiatry 71:286–293.
genome-wide SNPs. Nat Genet 45:984–994. 48. Arnone D, McKie S, Elliott R, Juhasz G, Thomas EJ, Downey D, et al.
31. Lee PH, Anttila V, Won H, Feng YCA, Rosenthal J, Zhu Z, et al. (2019): (2012): State-dependent changes in hippocampal grey matter in
Genomic relationships, novel loci, and pleiotropic mechanisms across depression. Mol Psychiatry 18:1265–1272.
eight psychiatric disorders. Cell 179:1469–1482.e11. 49. Opel N, Redlich R, Grotegerd D, Dohm K, Heindel W, Kugel H,
32. Lichtenstein P, Yip BH, Björk C, Pawitan Y, Cannon TD, Sullivan PF, et al. (2015): Obesity and major depression: Body-mass index
Hultman CM (2009): Common genetic determinants of schizophrenia (BMI) is associated with a severe course of disease and specific
and bipolar disorder in Swedish families: A population-based study. neurostructural alterations. Psychoneuroendocrinology 51:219–
Lancet 373:234–239. 226.
33. Kendler KS, Ohlsson H, Sundquist J, Sundquist K (2020): An extended 50. Vita A, De Peri L, Deste G, Barlati S, Sacchetti E (2015): The effect of
Swedish national adoption study of bipolar disorder illness and cross- antipsychotic treatment on cortical gray matter changes in schizo-
generational familial association with schizophrenia and major phrenia: Does the class matter? A meta-analysis and meta-regression
depression. JAMA Psychiatry 77(8):1–9. of longitudinal magnetic resonance imaging studies. Biol Psychiatry
34. Grasby KL, Jahanshad N, Painter JN, Colodro-Conde L, Bralten J, 78:403–412.
Hibar DP, et al. (2020): The genetic architecture of the human cerebral 51. Torres US, Portela-Oliveira E, Borgwardt S, Busatto GF (2013): Struc-
cortex. Science 367:eaay6690. tural brain changes associated with antipsychotic treatment in schizo-
35. Copeland WE, Shanahan L, Hinesley J, Chan RF, Aberg KA, phrenia as revealed by voxel-based morphometric MRI: An activation
Fairbank JA, et al. (2018): Association of childhood trauma exposure likelihood estimation meta-analysis. BMC Psychiatry 13:342.
with adult psychiatric disorders and functional outcomes. JAMA Netw 52. Yoshimura S, Okamoto Y, Onoda K, Matsunaga M, Okada G,
Open 1:e184493. Kunisato Y, et al. (2014): Cognitive behavioral therapy for depression
36. Scott KM, McLaughlin KA, Smith DA, Ellis PM (2012): Childhood changes medial prefrontal and ventral anterior cingulate cortex activity
maltreatment and DSM-IV adult mental disorders: Comparison of associated with self-referential processing. Soc Cogn Affect Neurosci
prospective and retrospective findings. Br J Psychiatry 200:469–475. 9:487–493.
37. Kessler RC, McLaughlin KA, Green JG, Gruber MJ, Sampson NA, 53. Howard DM, Adams MJ, Clarke T-K, Hafferty JD, Gibson J, Shirali M,
Zaslavsky AM, et al. (2010): Childhood adversities and adult psycho- et al. (2019): Genome-wide meta-analysis of depression identifies 102
pathology in the WHO World Mental Health Surveys. Br J Psychiatry independent variants and highlights the importance of the prefrontal
197:378–385. brain regions. Nat Neurosci 22:343–352.
38. Arseneault L, Cannon M, Fisher HL, Polanczyk G, Moffitt TE, Caspi A 54. van Tol M-J, van der Meer L, Bruggeman R, Modinos G, Knegtering H,
(2011): Childhood trauma and children’s emerging psychotic symp- Aleman A (2014): Voxel-based gray and white matter morphometry
toms: A genetically sensitive longitudinal cohort study. Am J Psychi- correlates of hallucinations in schizophrenia: The superior temporal
atry 168:65–72. gyrus does not stand alone. NeuroImage Clin 4:249–257.
39. Rao U, Chen L-A, Bidesi AS, Shad MU, Thomas MA, Hammen CL 55. Visser JC, Rommelse NNJ, Greven CU, Buitelaar JK (2016):
(2010): Hippocampal changes associated with early-life adversity and Autism spectrum disorder and attention-deficit/hyperactivity dis-
vulnerability to depression. Biol Psychiatry 67:357–364. order in early childhood: A review of unique and shared charac-
40. Opel N, Redlich R, Dohm K, Grotegerd D, Zaremba D, Goltermann J, teristics and developmental antecedents. Neurosci Biobehav Rev
et al. (2019): Mediation of the influence of childhood maltreatment on 65:229–263.
depression relapse by cortical structure: A 2-year longitudinal obser- 56. Dougherty CC, Evans DW, Myers SM, Moore GJ, Michael AM (2016):
vational study. Lancet Psychiatry 6:318–326. A comparison of structural brain imaging findings in autism spectrum
41. Busso DS, McLaughlin KA, Brueck S, Peverill M, Gold AL, disorder and attention-deficit hyperactivity disorder. Neuropsychol
Sheridan MA (2017): Child abuse, neural structure, and adolescent Rev 26:25–43.
psychopathology: A longitudinal study. J Am Acad Child Adolesc 57. Boedhoe PSW, Heymans MW, Schmaal L, Abe Y, Alonso P,
Psychiatry 56:321–328.e1. Ameis SH, et al. (2019): An empirical comparison of meta- and mega-
42. Van Dam NT, Rando K, Potenza MN, Tuit K, Sinha R (2014): Childhood analysis with data from the ENIGMA Obsessive-Compulsive Disorder
maltreatment, altered limbic neurobiology, and substance use relapse Working Group. Front Neuroinform 12:102.
severity via trauma-specific reductions in limbic gray matter volume. 58. Stevens JP (2002): Applied Multivariate Statistics for the Social Sci-
JAMA Psychiatry 71:917–925. ences, 4th ed. Hillsdale, NJ: Erlbaum.

686 Biological Psychiatry November 1, 2020; 88:678–686 www.sobp.org/journal