Schmaal et al. Translational Psychiatry (2020)10:172 https://doi.org/10.1038/s41398-020-0842-6 Translational Psychiatry REVIEW ARTICLE Open Access ENIGMA MDD: seven years of global neuroimaging studies of major depression through worldwide data sharing Lianne Schmaal 1,2, Elena Pozzi1,2, Tiffany C. Ho3,4,5, Laura S. van Velzen1,2, Ilya M. Veer6, Nils Opel7, Eus J. W. Van Someren8,9,10, Laura K. M. Han 10, Lybomir Aftanas11,12, André Aleman13, Bernhard T. Baune7,14,15, Klaus Berger16, Tessa F. Blanken8,9, Liliana Capitão 17,18, Baptiste Couvy-Duchesne19, Kathryn R. Cullen20, Udo Dannlowski7, Christopher Davey 14, Tracy Erwin-Grabner21, Jennifer Evans22, Thomas Frodl23, Cynthia H. Y. Fu24,25, Beata Godlewska17, Ian H. Gotlib3, Roberto Goya-Maldonado21, Hans J. Grabe 26,27, Nynke A. Groenewold28, Dominik Grotegerd7, Oliver Gruber29, Boris A. Gutman30, Geoffrey B. Hall31, Ben J. Harrison32, Sean N. Hatton 33, Marco Hermesdorf16, Ian B. Hickie33, Eva Hilland34,35,36, Benson Irungu37, Rune Jonassen38, Sinead Kelly39, Tilo Kircher40, Bonnie Klimes-Dougan20, Axel Krug 40, Nils Inge Landrø34,35, Jim Lagopoulos41, Jeanne Leerssen8,9, Meng Li23, David E. J. Linden 42,43,44, Frank P. MacMaster45, Andrew M. McIntosh 46, David M. A. Mehler7,43,44, Igor Nenadić40,47, Brenda W. J. H. Penninx10, Maria J. Portella48,49,50, Liesbeth Reneman51, Miguel E. Rentería 52, Matthew D. Sacchet53, Philipp G. Sämann54, Anouk Schrantee51, Kang Sim 55,56, Jair C. Soares37, Dan J. Stein 57, Leonardo Tozzi 4, Nic J. A. van Der Wee58,59, Marie-José van Tol 13, Robert Vermeiren60, Yolanda Vives-Gilabert61, Henrik Walter 6, Martin Walter62,63, Heather C. Whalley 46, Katharina Wittfeld 26,27, 1234567890():,; 1234567890():,; 1234567890():,; 1234567890():,; Sarah Whittle32, Margaret J. Wright 64,65, Tony T. Yang5, Carlos Zarate Jr66, Sophia I. Thomopoulos67, Neda Jahanshad 67, Paul M. Thompson67 and Dick J. Veltman10 Abstract A key objective in the field of translational psychiatry over the past few decades has been to identify the brain correlates of major depressive disorder (MDD). Identifying measurable indicators of brain processes associated with MDD could facilitate the detection of individuals at risk, and the development of novel treatments, the monitoring of treatment effects, and predicting who might benefit most from treatments that target specific brain mechanisms. However, despite intensive neuroimaging research towards this effort, underpowered studies and a lack of reproducible findings have hindered progress. Here, we discuss the work of the ENIGMA Major Depressive Disorder (MDD) Consortium, which was established to address issues of poor replication, unreliable results, and overestimation of effect sizes in previous studies. The ENIGMA MDD Consortium currently includes data from 45 MDD study cohorts from 14 countries across six continents. The primary aim of ENIGMA MDD is to identify structural and functional brain alterations associated with MDD that can be reliably detected and replicated across cohorts worldwide. A secondary goal is to investigate how demographic, genetic, clinical, psychological, and environmental factors affect these associations. In this review, we summarize findings of the ENIGMA MDD disease working group to date and discuss future directions. We also highlight the challenges and benefits of large- scale data sharing for mental health research. Correspondence: Lianne Schmaal (lianne.schmaal@unimelb.edu.au) 1 Orygen, The National Centre of Excellence in Youth Mental Health, Parkville, VIC, Australia 2 Centre for Youth Mental Health, The University of Melbourne, Parkville, VIC, Australia Full list of author information is available at the end of the article © The Author(s) 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/. Schmaal et al. Translational Psychiatry (2020)10:172 Page 2 of 19 Introduction Worldwide pooling of existing neuroimaging data offer Major depressive disorder (MDD) is the largest con- a highly effective alternative to larger non-generalizable tributor to the disease burden caused by poor mental studies and retrospective meta-analyses, as it (1) makes health worldwide owing to its high prevalence, high optimal use of valuable and costly existing data sets from recurrence rates, chronicity, and comorbidity with phy- individual studies; (2) collates large data sets at a relatively sical illness1. Thus, effective and early treatment is crucial. low cost; (3) allows coordinated analysis using standar- Unfortunately, current strategies for treating MDD, which dized protocols for data processing and analysis; and (4) do not take neurobiological markers into consideration, combines expertize of hundreds of professionals in the have not been particularly effective2–6. fields of neuroimaging, psychiatry, statistics, and mathe- Over the past several decades, technical advances in matics. Here, we discuss the work of the worldwide neuroimaging have provided the impetus for identifying Enhancing NeuroImaging Genetics through Meta- measurable indicators of brain processes associated with Analysis (ENIGMA) Major Depressive Disorder (MDD) MDD in order to detect individuals at risk for the dis- consortium. order, to facilitate the development of novel interven- tions, and to evaluate treatment effects. Although The ENIGMA MDD consortium several neuroimaging markers have been found that The MDD Working Group was founded in 2012 as part differentiate patients with MDD from healthy controls of the ENIGMA consortium. ENIGMA was initiated in (e.g., Mulders et al.7; Kempton et al.8), progress has still 2009 to boost statistical power in genome-wide associa- been limited in part by underpowered studies and a lack tion studies (GWAS) that aimed to identify common of reproducible findings (e.g., Kapur et al.9). Many genetic variants that affect brain structure19–21. Because research studies in the field have been restricted by most major mental illnesses have a high-dimensional small samples, resulting in a lower probability of finding genetic architecture with polygenic influences, epistasis “true” effects (low-powered studies tend to produce and gene by environment interactions, by focusing on more false negatives than do high-powered studies) and intermediate phenotypes—or endophenotypes—at the an inflated estimate of the effect size when a true effect level of MRI-derived brain measures, it was thought that is discovered10–12. This is particularly problematic when researchers would be better able to identify the neuro- the true effect size is modest, which is often the case for biological underpinnings of psychiatric disorders22–24. differences in brain imaging measures between patients Therefore, the ENIGMA consortium was launched to and controls. Underpowered studies also make it diffi- combine existing genomic and neuroimaging data around cult to reproduce significant findings, leading to the world to conduct well-powered GWAS analyses. inconsistent and poorly replicated neuroimaging find- ENIGMA has since published the largest genetic studies ings in depression13,14. of the brain, in partnership with other consortia25–30, Larger samples and meta-analytic approaches represent mapping genome-wide effects of over a million genetic good strategies to overcome issues associated with small loci in over 30,000–50,000 brain MRI scans (for a recent sample sizes. Large-scale data collection initiatives with review of the ENIGMA imaging genetics findings, see harmonized assessments (including neuroimaging), such Thompson et al.21). as the population-based UK Biobank study (N = Building on ENIGMA’s initial successes in imaging 500,000)15, are yielding key insights into brain mechan- genetics, disease working groups were formed to study isms involved in MDD (e.g., Howard et al.16; Shen et al.17; patterns of brain abnormalities in major psychiatric, Harris et al.18). However, recruiting large samples is not neurodevelopmental, neurological, and neurogenetic dis- always feasible because of limited access to patient orders. ENIGMA MDD was established with the initial populations at any one site or limited availability of aim to (1) identify structural and functional brain altera- scanning facilities and the financial costs of scanning tions associated with MDD that can be reliably detected hundreds or even thousands of participants. Moreover, and replicated across many different samples worldwide; large-scale population-based samples also typically focus and (2) identify demographic, genetic, clinical, psycholo- on individuals from a single geographic region or country gical, and environmental factors that affect these within a restricted age range, thereby limiting the gen- associations. eralizability of findings across countries, cultures, and Since it was established, ENIGMA MDD has grown to developmental stages. Issues with retrospective meta- 35 participating research institutions (including 45 study analyses include the potential over-representation of cohorts) from 14 different countries across six continents positive findings in the published literature (publication in September 2019 (Fig. 1). For an up to date overview of bias) and a lack of harmonization of data processing and all participating research institutions, see: http://enigma. statistical analysis methods across the different studies ini.usc.edu/wp-content/uploads/2019/10/List_members_ included in the meta-analysis. oct2019-2.pdf. To date, participating researchers have Schmaal et al. Translational Psychiatry (2020)10:172 Page 3 of 19 Fig. 1 World map of cohorts participating in ENIGMA MDD. Locations of cohorts included in the ENIGMA MDD consortium in September 2019. reproducibility. Brain measures derived from the ENIGMA protocols have shown good reliability31–34. Because ENIGMA is dedicated to “open science”, all ENIGMA protocols are publicly available on the ENIGMA website. Here, we discuss findings of structural brain alterations associated with MDD and depression-related phenotypes that were obtained through this worldwide data-sharing initiative. We also discuss clinical implications, future directions for the ENIGMA MDD consortium and iden- tify challenges of large-scale data sharing. Fig. 2 Connections between ENIGMA MDD and other ENIGMA working groups. Note: not all ENIGMA working groups are displayed ENIGMA MDD findings to date in this figure. In September 2019, ENIGMA includes 50 working groups, The initial studies conducted with data available of which 26 working groups focus on mental and neurological through the ENIGMA MDD consortium focused on disorders. dMRI diffusion magnetic resonance imaging (MRI), rsfMRI identifying associations between MDD and structural resting state functional MRI, EEG electroencephalogram, tbfMRI task- brain measures that could reliably be detected across based functional MRI, MDD major depressive disorder, PTSD post- traumatic stress disorder, AD anxiety disorder. many samples worldwide. Because inconsistent findings across previous retrospective meta-analyses may be owing in part to differences in data processing and statistical shared demographic, clinical and neuroimaging data from analyses among the primary studies, we conducted indi- 9788 healthy individuals and 4372 individuals with MDD. vidual participant data (IPD)-based (or prospective) meta- ENIGMA MDD and other disease working groups are analyses to identify differences in subcortical volume, supported by the ENIGMA Methods working groups (Fig. subcortical shape, cortical thickness, cortical surface area, 2), which are dedicated to developing standardized pro- and white matter integrity between patients with MDD cessing, quality assurance, and statistical analysis proto- and healthy controls, and to examine the effects of cols, to reducing statistical heterogeneity and researcher demographic and clinical characteristics35–38. Harmo- degrees of freedom, and to ensuring or evaluating nized imaging processing (e.g., Freesurfer39), quality Schmaal et al. Translational Psychiatry (2020)10:172 Page 4 of 19 assurance (http://enigma.ini.usc.edu/protocols/imaging- effect size was modest (Cohen’s d = −0.14). The hippo- protocols/) and statistical analysis protocols were run campal volume deficit was greater in MDD patients with locally on data from participating cohorts. In addition, the recurrent episodes (N = 1119, Cohen’s d = −0.17), com- scope of retrospective meta-analyses in terms of examined pared with healthy controls, whereas no hippocampal brain regions is limited to those reported in the original volume alterations were observed in first-episode patients studies, as many published studies adopted a hypothesis- (N = 583). Our findings may suggest that depression- driven approach and focused on specific regions of related reductions in hippocampal volume are a result of interest (ROIs). The large sample size available in longer illness duration or greater number of episodes, ENIGMA MDD ensures the statistical power needed to instead of a premorbid vulnerability factor. This is con- investigate whole-brain structural brain alterations. For sistent with prior longitudinal studies showing greater the IPD-based meta-analyses, summary statistics of each hippocampal atrophy in individuals with persistent, site were shared to be included in a random effects meta- recurring, or worsening of depressive symptoms over analysis to examine differences in structural brain mea- time46–49. Nonetheless, it is unclear whether hippocampal sures between MDD patients and controls. atrophy associated with prolonged illness duration or A few ENIGMA MDD studies have started to adopt a recurrence represents a state marker instead of a per- mega-analytic approach, where individual-level measures manent scar. Fortunately, hippocampal alterations may derived from the harmonized imaging processing proto- normalize as hippocampal enlargement has been observed cols are pooled across sites and regression analyses are following remission or treatment of MDD48,50,51. We also conducted on this pooled dataset while correcting for found smaller hippocampal volumes in patients with an confounding site effects (e.g., linear mixed models with a adolescent onset of MDD (⩽21 years; N = 541; Cohen’s random intercept for site). Key advantages of a meta- d = −0.20), compared with controls, whereas no differ- analytic versus mega-analytic approach include: (1) ences were observed in those with an adult onset of MDD allowing the analysis of individual studies to account for (N = 997) (Fig. 3a). This is in line with previous studies local population substructure; (2) allowing analysis of showing smaller hippocampal volumes in adolescents and study-specific covariates that may be better dealt with even children with depression52–55, whereas other studies within each study; and (3) allowing analyses to be con- found lower hippocampal volumes only in adults with an ducted within each participating site and results to be age of onset >3056 or no differences in hippocampal shared through a central site without requiring individual- volume between adults with an adolescent versus adult level data to be shared40. However, the advantages of a age of onset of MDD57. Because only about half (57%) of mega-analytic versus meta-analytic approach include (1) the adolescent-onset patients had a recurrent episode of greater flexibility in the control of confounders at the level MDD, adolescent disease onset may, in part, have an of individual patients and studies; and (2) not having to independent association with hippocampal volumes. assume within-study normality and known within-study Smaller hippocampal volumes may precede disease onset, variances40, as these assumptions can be especially pro- especially in this early-onset group, perhaps as a result of blematic with smaller samples41. Moreover, pooling all factors commonly associated with early-onset MDD data in a single statistical model may boost statistical including childhood adversity58,59 and genetic influ- power to detect certain effects42, such as higher-level ences60. Longitudinal studies designed to track hippo- interactions, when the phenotype of interest is rare (e.g., campal volume changes prior to disease onset and over suicide attempt, number of medication naive patients) or the disease course are required to elucidate whether when the range of continuous variables is limited (e.g., hippocampal abnormalities result from a prolonged age, symptom severity, childhood trauma scores) in duration of chronic stress associated with depressive individual studies. All ENIGMA MDD studies published episodes, represent a vulnerability factor for MDD, or to date are summarized in Table 1. both. Interestingly, we did not detect significant differences Structural brain alterations in MDD for any of the other subcortical volumes, including the Subcortical brain regions amygdala, nucleus accumbens, caudate, putamen, thala- The first ENIGMA MDD project focused on differences mus, and pallidum, or the lateral ventricles and intracra- in subcortical volume between MDD patients (N = 1728) nial volume (ICV). Previous reports have varied regarding and healthy controls (N = 7199)35. Consistent with prior volume abnormalities in subcortical regions other than studies and retrospective meta-analyses8,43–45, in this the hippocampus, such as the amygdala8,61,62. None- IPD-based meta-analysis we found significantly lower theless, associations with MDD may still be present for hippocampal volumes in individuals with MDD compared functionally distinct subregions within these broader with healthy controls. This effect was also consistently subcortical regions. Patterns of depression-related observed across individual cohorts, although the overall alterations in subregions of subcortical surfaces have Table 1 Overview of ENIGMA MDD studies published to date. Study Modality Meta- or mega- Sample size No. of cohorts Main findings analysis Schmaal et al. 2016 in Molecular sMRI (FreeSurfer subcortical volumes) Meta HC: 7199 MDD: 1728 15 Mean hippocampal volume was significantly lower in MDD Psychiatry compared with HC. This effect was driven primarily by recurrence of MDD (i.e., >1 episode). MDD with an early age of onset (⩽21 years) showed significantly lower mean hippocampal volumes than HC. Schmaal et al. 2017 in Molecular sMRI (FreeSurfer cortical thickness and Meta Adults: HC: 7658 MDD: 1902 20 Compared with adult HC, adults with MDD showed lower cortical Psychiatry surface area) Adolescents: HC: 294 thickness in the bilateral medial OFC, cingulate cortex, insula and MDD: 213 temporal lobes, but no surface area alterations. Subgroup analysis revealed lower cortical thickness in adults with MDD with adult onset, but not adolescent onset, age of onset, relative to HC. Compared with adolescent HC, adolescents with MDD had lower total surface area (but no differences in cortical thickness), with most pronounced effects in medial OFC, superior frontal gyrus, and primary and higher order visual, somatosensory and motor areas. These effects were driven by adolescents with recurrent depression. Schmaal et al. Translational Psychiatry (2020)10:172 Renteria et al. 2017 in sMRI (FreeSurfer subcortical volumes) Meta HC: 1996 MDD: 1101 20 No significant differences were found between MDD with suicidal Translational Psychiatry ideation and HC. MDD with suicidal behavior (reported suicidal attempts or plans) showed a trend toward significant smaller ICV, compared with HC. No significant differences were found between MDD with and without suicidal ideation and/or behavior. Frodl et al. 2017 in Journal of sMRI (FreeSurfer subcortical volumes) Mega HC: 2078 MDD: 958 9 Severity of childhood maltreatment (CM) was associated with lower Psychiatry Research caudate volumes in females, but no significant effects were found in males. The effect was associated with all subcategories of CM, but most pronounced for childhood emotional and physical neglect. The effect was independent of MDD diagnosis. Tozzi et al. 2019 in Psychological sMRI (FreeSurfer cortical thickness and Mega HC: 2588 MDD: 1284 12 Regardless of MDD, overall severity of childhood maltreatment (CM) Medicine surface area) was associated with lower thickness in the supramarginal gyrus, banks of the superior temporal sulcus and lower surface area in the middle temporal lobe. Compared with no CM, the combination of childhood abuse and neglect showed lower cortical thickness in the same areas, in addition to the inferior parietal lobe, middle temporal lobe, and precuneus, whereas no effects were found for abuse or neglect alone. Males—but not females—with MDD and a history of CM showed greater surface area in the rostral ACC compared with the no CM group. The negative association between CM severity and thickness of various prefrontal, cingulate, and temporal regions was more pronounced with increasing age. No significant interaction effect between MDD diagnosis and CM. de Kovel et al. 2019 in American sMRI (FreeSurfer subcortical volumes, Mega Cortical regions: HC: 3504 Cortical regions: 31 No differences in the laterality of cortical regions thickness and Journal of Psychiatry cortical thickness and surface area) MDD: 2256 Subcortical surface area or subcortical volumes were found between MDD Subcortical regions: HC: regions: 32 and HC. 4230 MDD: 2540 Ho et al. 2020 in Human Brain sMRI (FreeSurfer subcortical shapes) Meta HC: 2953 MDD: 1781 10 Compared with HC, MDD had lower thickness and surface area in Mapping the subiculum and CA2/3 areas of the hippocampus and basolateral amygdala. These effects were primarily driven by MDD with an adolescent age of onset (⩽21 years). Recurrence of MDD was associated with lower surface area and thickness in the basolateral amygdala and in the CA1 region of the hippocampus. Page 5 of 19 Schmaal et al. Translational Psychiatry (2020)10:172 Page 6 of 19 of onset (>21 years) and antidepressant-free patients at the time of examined, relative to HC. These effects appeared to be global, with scanning. Higher RD in adults with MDD was also observed across the corona radiate and the corpus callosum contributing most. The effects were mainly driven by recurrent MDD, MDD with adult age DTI Diffusion tensor imaging, FA fractional anisotropy, HC healthy controls, ICV intracranial volume, MDD major depressive disorder, CM childhood maltreatment, OFC orbitofrontal cortex, ACC anterior cingulate cortex, RD predicted age difference of 1.08 years). Strongest effects were found there were no significant differences between the MDD subgroups. in MDD using antidepressants at time of scanning, patients in an individuals with MDD were estimated to be older than expected been difficult to identify as there are few identifiable active episode and patients in remission compared with HC, but Adults with MDD showed lower FA in 16 of the 25 WM tracts Brain-PAD was positive in all MDD subgroups, indicating that Compared with HC, MDD showed higher brain-PAD (brain- surface landmarks, in contrast with the landmarks con- different ROIs. No differences were found between healthy sistently found in cortical surfaces (e.g., deep sulcal pat- radial diffusivity, MD mean diffusivity, AD axial diffusivity, ROIs regions of Interest, Brain-PAD brain-predicted age differencefractional anisotropy, sMRI structuralmagnetic resonance imaging, WM white matter. terns). In this context, subcortical shape analysis may be a more sensitive method to identify more localized effects in subdivisions in subcortical regions that were not cap- tured by the volumetric analysis of subcortical regions in adolescents and adolescents with MDD. the first paper from the ENIGMA MDD consortium35. To address this, we conducted an additional multi-site meta- based on the brain age model. analytic investigation to test whether MDD patients, and specific subgroups of MDD based on important clinical characteristics, differ from healthy controls in subcortical shape. Specifically, we applied meta-analytic models on Main findings effect sizes generated from 1781 patients with MDD and 2953 healthy controls across 10 study cohorts. Consistent with the findings from our first meta-analysis, we found that relative to healthy controls (N = 2879), patients with an adolescent onset of MDD (N = 476) had lower thick- ness (Cohen’s d = −0.17) and smaller surface area No. of cohorts (Cohen’s d = −0.18) in the hippocampus, with most pronounced effects in the subiculum and cornu ammonis (CA) subfields two and three of the hippocampus37 19 20 (Fig. 3a). Extending our prior findings, we also observed Adults: HC: 1265 MDD: 921 lower thickness (Cohen’s d = −0.16) and smaller surface area (Cohen’s d = −0.17) in the amygdala in adolescent- Adolescents: HC: 290 HC: 4314 MDD: 2675 onset patients, specifically within the basolateral subdivi- sion of the amygdala37 (Fig. 3a). These subregions are rich Sample size in glucocorticoid receptors, emphasizing that disturbed MDD: 372 glucocorticoid signaling during stress response promotes the development of MDD63. Importantly, shape analyses of subcortical structures clarify results in the extant lit- Meta- or mega- erature of smaller hippocampal volumes and ambiguous effects in the amygdala in patients with MDD; delineating analysis nuanced effects in depression-related subregions of sub- Mega Meta cortical structures may help to identify more precise intervention targets or more sensitive biomarkers of sMRI (FreeSurfer subcortical volumes, cortical thickness and surface area) treatment response. Noteworthy, additional ENIGMA DTI (FA, RD, MD and AD for atlas- MDD analyses regarding associations between MDD and defined white matter tracts of FreeSurfer-derived hippocampal subfields are currently underway. Cortical thickness and surface area Following these studies of subcortical morphology, we Modality interest) examined cortical thickness and surface area in relation to MDD and clinical characteristics in a meta-analysis of data from 20 participating ENIGMA MDD cohorts36. Because more research groups joined ENIGMA MDD Han et al. 2020 in Molecular after the publication of our subcortical volume meta- Van Velzen et al. 2019 in Table 1 continued analysis study, we were able to conduct separate analyses Molecular Psychiatry in young people (⩽21 years) and adults (>21 years). Most published studies to date have focused on regional cortical volume, which is a function of cortical thickness and Psychiatry surface area. Advances in neuroimaging data processing Study have made it possible to separate cortical surface area and Schmaal et al. Translational Psychiatry (2020)10:172 Page 7 of 19 Fig. 3 Converging findings across ENIGMA MDD studies. Specific characteristics of brain structure are differentially affected by MDD (or vice versa) at different stages of life. a Alterations in hippocampal and amygdala volumes and shapes are observed in adolescent-onset MDD and lower cortical surface area in adolescents with MDD. b Cortical thickness alterations and white matter abnormalities are specifically associated with adult- onset MDD and older age in individuals with MDD and/or childhood maltreatment. *This association was independent of MDD diagnosis. MDD major depressive disorder, FA fractional anisotropy, RD radial diffusivity. cortical thickness, which is important to do in the context the hippocampal volume finding that were driven by adult of understanding brain correlates of MDD, given that patients with an adolescent onset of their first depressive these neural characteristics are genetically and phenoty- episode, the lower cortical thickness findings were driven pically distinct27,64,65. In adults, we observed subtle cor- mostly by adult patients with an adult onset of MDD tical thickness alterations in 13 of 68 cortical regions in (N = 1214; Cohen’s d −0.11 to −0.18) relative to controls. patients with MDD (N = 1902) compared with healthy No cortical surface area differences were found among the controls (N = 7658) (Cohen’s d’s between −0.10 and adult groups. We speculated that the more pronounced −0.14), including lower thickness of the bilateral medial effects in adult patients with an adult age of onset may be orbitofrontal cortex (OFC), fusiform gyrus, insula, rostral driven in part by their older age compared with adult anterior (ACC) and posterior cingulate cortex (PCC) and patients with an adolescent age of onset of their first unilaterally in the left middle temporal gyrus, right depressive episode, which was confirmed by a post hoc inferior temporal gyrus and right caudal ACC (Fig. 3b). moderator analysis with mean age of patients in each Our findings in adults with MDD were consistent with sample. This suggests that mental illness has a greater prior meta-analyses showing depression-related structural impact on cortical thickness in the context of aging. alterations in the medial PFC and ACC8,66–69; however, Indeed, cortical thickness has been shown to be a more they extended previous findings by demonstrating struc- sensitive indicator of aging than is surface area or tural abnormalities in the temporal regions (middle and volume70–72. Because our meta-analytic approach did not inferior temporal gyri and fusiform gyrus), posterior cin- allow us to pool all data across samples, we were not able gulate cortex and insula. These cortico-limbic thickness to investigate age-by-diagnosis effects across the entire alterations may contribute to the broad spectrum of age range (individual samples had restricted age ranges). emotional, cognitive, and behavioral disturbances Therefore, future mega-analyses could further elucidate in MDD. these dynamic relations with development and aging. The largest effect size was observed in the medial OFC, Surprisingly, in contrast to adults with MDD, adoles- which—in contrast to lower hippocampal volume—was cents with MDD showed no cortical thickness alterations, already detectable in first-episode patients. In contrast to but rather, alterations in global cortical surface area Schmaal et al. Translational Psychiatry (2020)10:172 Page 8 of 19 (Fig. 3a); the effect sizes of these results were larger right)/(left + right)) for local and global cortical and (Cohen’s d −0.31 to −0.41 in local regions) than the subcortical brain regions in individuals with MDD (N = cortical thickness alterations observed in adults. Cortical 2256) compared with healthy controls (N = 3504)87. The surface area has been understudied in the context of results showed no significant differences in brain struc- MDD. Nonetheless, a recent longitudinal study showed tural asymmetry between individuals with MDD and that lower surface area was specifically observed in young controls for any of the structural brain measures, nor any people experiencing depressive symptoms in early ado- associations with clinical characteristics. These findings lescence but not in those developing depressive symptoms suggest that altered brain structural asymmetry is of little later in adolescence, and that lower surface area was relevance to the pathophysiology of MDD, although already observable in young people with subclinical functional asymmetries may still play a role. depressive symptoms, not all of whom will develop a full- threshold MDD diagnosis36. Thus, cortical surface area Brain aging in MDD reductions may represent an early developing subtype of MDD is associated with an increased risk of aging- depressive disorder, shaped by genetic factors or early life related medical illnesses such as cardiovascular disease adversity (prenatal73,74 or perinatal or during child- and cancer88,89. Although aging is associated with loss of hood75–77), and potentially precede the onset of MDD. gray matter, depression may accelerate age-related brain This hypothesis is consistent with the observation that, atrophy90. Therefore, we examined deviations from nor- compared with cortical thickness, cortical surface area has mative brain aging in adults with MDD and associated a higher genetic heritability27,64,78, has a genetic correla- clinical heterogeneity by pooling data from >6900 healthy tion with MDD and depressive symptoms (this genetic controls and individuals with MDD from 19 different association is absent for cortical thickness27), is deter- scanners participating in the ENIGMA MDD con- mined earlier in development, and is less strongly affected sortium91. Normative brain aging was estimated by pre- by later environmental influences71,79. dicting chronological age (18–75 years) from 7 subcortical Importantly and paradoxically, no differences in surface volumes, 34 cortical thickness and 34 surface area, lateral area were observed in adult patients with adolescent- ventricles and ICV measures using Ridge Regression, onset MDD. This might be explained by (1) normalization separately in 952 male and 1236 female controls. We of cortical surface area when transitioning into adulthood; showed that our brain age prediction model generalized (2) cortical surface area alterations being only present in a to unseen hold-out samples (927 male controls and 986 specific subgroup (subtype) of adult patients with males with MDD, and 1199 female controls and 1689 adolescent-onset MDD, which we were unable to detect; females with MDD; correlations r between predicted and or (3) those with cortical surface area alterations in early actual age ranged from 0.77–0.85, mean absolute errors adolescence may be at higher risk for transitioning from (MAE) ranged from 6.32 to 7.18 years), as well as to MDD to other mental disorders over time. This latter completely independent samples from different scanning possibility is consistent with reports of lower cortical sites (N = 1330 from 23 different scanners; r = 0.71 and surface area in adolescents and adults with psychosis or MAE = 7.49 for male controls, r = 0.72 and MAE = 7.26 schizophrenia80,81, and in individuals at high risk for and/ for female controls)91. or transitioning to psychosis82,83. Longitudinal studies are Brain-predicted age difference (brain-PAD) was com- required to test the hypothesis that cortical surface area puted from the difference between predicted “brain age” alteration is a pre-existing risk factor for the development and chronological age92. We found that, at the group of MDD, and to investigate the subsequent clinical course level, MDD patients had a + 1.08 years (Cohen’s d = 0.14, of these depressed young people with global surface area p < 0.0001) greater discrepancy between their predicted reductions. and actual age compared with control participants. In other words, individuals with MDD were estimated to be Brain asymmetry ~ 1 year older than expected based on the brain age Altered brain asymmetry may have a role in MDD, model. The brain age model relied mostly on cortical especially at a functional level84, but only a few studies thickness measures (compared with subcortical volumes, have investigated structural asymmetry85,86. In our stu- cortical surface area and ICV) in order to make good age dies, the majority of cortical thickness or surface area predictions. Brain-PAD differences were observed in all alterations were observed across both hemispheres, with a subgroups of patients compared with controls, with no few regions showing only unilateral differences36. How- significant differences in brain-PAD between the patient ever, we did not explicitly test whether the effect sizes of groups. our findings differed significantly by hemisphere. There- As many of the MDD patients did not show advanced fore, in a separate study, we investigated structural brain aging compared with controls, the clinical sig- asymmetries by investigating asymmetry indices ((left − nificance of the observed higher brain-PAD in MDD Schmaal et al. Translational Psychiatry (2020)10:172 Page 9 of 19 patients may be limited. However, there may still be a radiata is part of the limbic-thalamo-cortical circuitry and is subgroup of MDD patients with more extreme patterns of also implicated in mood regulation101. brain aging, which would be important to identify as The effect sizes for case–control differences in adults accelerated brain aging may be reversed with targeted were small, but very similar to the effect sizes reported in treatment. For example, one study showed that brain- the meta-analysis of subcortical volume and cortical PAD was temporarily reduced by 1.1 years in healthy morphology35,36. Also, in line with the previous cortical controls owing to the acute anti-inflammatory effects of and subcortical meta-analysis findings, the widespread ibuprofen93. Inflammatory biomarkers are commonly alterations in FA in adult patients were driven by MDD dysregulated in MDD and negative relationships between patients with recurrent episodes (N = 645), as there were levels of inflammatory cytokine (e.g., interleukin-6) and no significant differences between first-episode patients cortical thickness have been found in medication-free, (N = 169) and controls (N = 816). This again suggests first-episode MDD patients94, suggesting that inflamma- that these alterations may reflect the cumulative effect of tion may be a common biological mechanism between stress on brain morphology, rather than a vulnerability MDD and brain aging. Notably, brain-PAD has been factor for MDD, although the reduced statistical power shown to be a general predictor of psychiatric and neu- for the first-episode patients comparison may also explain rological disorders, with low clinical disease specificity95. this difference. In line with the cortical meta-analysis findings, but in contrast with findings on subcortical White matter microstructure volume, we observed lower FA in patients with an adult We also examined white matter microstructure in 1305 age of onset (N = 399) compared with controls MDD patients, and 1602 healthy controls from 20 sam- (N = 869) (Fig. 3b), but no differences when comparing ples participating in ENIGMA MDD worldwide, again patients with an adolescent age of onset (N = 251) com- using a meta-analytic approach38. The ENIGMA protocol pared with controls (N = 853). We hypothesized that for diffusion tensor imaging (DTI)33 calculates fractional MDD may interact with the normal aging process of white anisotropy (FA) for 25 atlas-defined white matter tracts of matter, which was in line with findings from our diagnosis- interest. FA is a commonly used measure in DTI analysis, by-age interaction analysis, in which we observed that and higher values indicate directionally constrained dif- MDD was associated with an accelerated decline in overall fusion of water molecules within the white matter, which FA with increasing age compared to controls. is mostly interpreted as higher degree of myelin integrity. We could not replicate the case–control differences in In addition to FA, the ENIGMA DTI protocols also yield white matter microstructure in adults in a sample from the following diffusivity metrics: axial diffusivity (AD), the UK Biobank (N = 2096 patients and 3275 healthy which is thought to represent the number, caliber, and controls), which may be related to lower severity of MDD organization of axons, radial diffusivity (RD), which may symptoms in the UK Biobank sample or subtle differences be a measure of myelination, and mean diffusivity (MD), in image processing. In addition, there were no significant which is often considered a measure of membrane den- differences in FA or diffusivity measures between ado- sity96. As maturation of white matter tracts continues lescent patients and controls in the ENIGMA MDD through adolescence and young adulthood, adolescent sample after correction for multiple testing, with smaller (age ≤21 years) and adult (age >21 years) patients and effects in adolescents potentially related to lower disease controls were analyzed separately. The meta-analysis duration and number of episodes in adolescent patients showed subtle but widespread changes in FA, with compared with adult patients. However, we cannot rule lower FA in adult MDD patients observed in 16 out of 25 out that our sample of adolescent participants may still ROIs (Cohen’s d between 0.12 and 0.26) (Fig. 3b). Adult have been too small to detect subtle effects (N = 372 MDD patients also showed higher RD in multiple tracts patients and 290 healthy controls). (Cohen’s d between 0.12 and 0.18), potentially reflecting changes in the morphology of glial cells or myelina- Sex differences in depression-related structural brain tion97,98. These alterations in FA and RD appeared to be alterations global effects, as after correction for average FA and RD Major depression is more than twice as prevalent and the across the white matter skeleton respectively, these effects disease burden of MDD is 50% higher in females than in were no longer significant. Nevertheless, and in accordance males1, which may suggest different etiological pathways to with previous studies, the strongest regional changes in FA developing MDD in males and females. However, across were observed in the genu and body of the corpus callosum our ENIGMA MDD studies examining subcortical, cortical, and the corona radiata99,100. The corpus callosum connects and white matter integrity differences in MDD35,37,38,80, we brain regions in both hemispheres, including regions found no diagnosis-by-sex interaction effects in adult MDD involved in mood regulation such as the anterior cingulate patients, indicating that structural brain alterations likely do cortices and orbitofrontal cortices, whereas the corona not contribute to these sex differences in MDD. In addition, Schmaal et al. Translational Psychiatry (2020)10:172 Page 10 of 19 even though the model fits of the brain aging models cortex (PCC), insula, precuneus, and frontal and temporal improved when trained separately in males and females, the lobe regions. This regional pattern is consistent with the (subtle) advanced brain aging that we observed in adults cortical meta-analysis study36, where we found that adult with MDD was not different for male versus female patients (>21 years old), and especially those with an adult patients91. Nonetheless, sex differences in structural brain age of onset of MDD, had lower cortical thickness in the alterations may be present during specific sensitive periods bilateral OFC, ACC, PCC, insula, frontal, and temporal of brain development, such as adolescence or more speci- lobe regions. It is thus possible, then, that in adults these fically, during puberty102. We did indeed found higher RD depression-related cortical regions are explained by the only in adolescent males with MDD, but not females, severity of childhood maltreatment. compared with adolescent controls38. However, we did not The effects of childhood maltreatment on subcortical observe sex differences in the cortical alterations in the structures in MDD and healthy controls, however, were adolescent MDD group36. This could perhaps be explained distinct from the effects of MDD on subcortical struc- by the observation that sex differences in white matter tures. Notably, in females only—although the same pat- volumes increase from birth, through adolescence, until tern showed a trend towards significance in males— males and females reach adulthood, whereas sex differences greater maltreatment severity was associated with smaller in gray matter remain relatively stable across develop- caudate volumes110. This result stands in contrast to the ment103. Future research would also benefit from a first paper from the ENIGMA MDD consortium, where separation between gender and sex analyses. we found that MDD was associated with smaller hippo- campal volumes, but not smaller caudate volumes, Childhood maltreatment regardless of sex. As part of the dorsal striatum, the Of central relevance to understanding the role of caudate is involved in motor planning, procedural learn- environmental factors contributing to neurophenotypes ing, and reward-based reinforcement learning111,112. of MDD is research focused on childhood adversity and Specifically, the caudate codes representations of expec- maltreatment. Indeed, childhood maltreatment is rela- tation violation and reward prediction errors that underlie tively common in the general population104,105 and is approach and avoidance behaviors and reward-based associated with an increased risk of a multitude of psy- learning, all of which are significantly altered in indivi- chiatric illnesses, including MDD104,106. In addition to duals exposed to childhood adversity and maltreat- epidemiological and clinical evidence linking childhood ment113,114. While keeping in mind the heterogeneity of maltreatment and MDD, recent neuroimaging studies MDD and the limitations of retrospective reports of show that brain structures affected by childhood mal- childhood maltreatment115, in the context of under- treatment are also implicated in the etiology and expres- standing neurophenotypes of MDD thus far, our results sion of MDD symptoms107–109. may suggest that smaller hippocampal volumes result Two studies from the ENIGMA MDD consortium from (stress-related) mechanisms directly associated with examined the effects of childhood maltreatment on brain MDD, whereas smaller caudate volumes may result from structure in depressed and non-depressed individuals exposure to stress during sensitive periods of develop- using a mega-analytic approach75,110. In the two largest ment (i.e., childhood). Of note, MDD was no longer studies to date examining associations between childhood associated with smaller hippocampal volumes when cor- maltreatment and brain structure, Frodl et al. and Tozzi rected for childhood maltreatment110, although the et al. examined the association between severity of overall sample included in the subcortical meta-analysis35 childhood maltreatment—including emotional, physical was larger than the sample included in this childhood and sexual abuse, or emotional and physical neglect as maltreatment mega-analysis and there was also a differ- assessed with the childhood trauma questionnaire (CTQ) ence in the composition of the cohorts evaluated. Future —and brain morphometry in a total of 3036 and 3872 longitudinal studies in youth exposed to childhood mal- individuals with and without MDD, respectively. Across treatment are needed to disentangle primary con- all individuals, and correcting for MDD diagnosis, greater sequences of childhood maltreatment on brain integrity, exposure to childhood maltreatment was associated with from secondary associations caused by prolonged stress lower cortical thickness of the banks of the superior experiences and/or the development of maltreatment- temporal sulcus (STS) and supramarginal gyrus (SMG), associated psychiatric diseases such as MDD. and with lower surface area across the whole brain and in the middle temporal gyrus (Fig. 3b). Childhood mal- Suicidal thoughts and behaviors treatment severity interacted with age such that greater Many individuals with MDD experience suicidal severity and older age were associated with lower cortical thoughts, and major depressive episodes account for at thickness in banks of the STS, SMG, rostral anterior least half of suicide deaths116. The lifetime probability of cingulate cortex (rACC), OFC, ACC, posterior cingulate suicide attempts is 20–25% among people with major Schmaal et al. Translational Psychiatry (2020)10:172 Page 11 of 19 mood disorders117. Prior studies had identified structural Impact of antidepressant medication brain alterations in individuals with MDD and a history of With regard to antidepressant medication use at the suicidal thoughts and behaviors, with most consistent time of scanning, patients taking antidepressants tended evidence for structural deficits in the ventromedial and to show greater structural alterations than antidepressant- ventrolateral PFC, dorsomedial and dorsolateral PFC, free patients, both in the subcortical and cortical ACC, insula and posterior structures including PCC, ENIGMA MDD meta-analysis studies35,36. These findings temporal regions, and the cerebellum (for recent reviews, are counterintuitive as antidepressant treatment has been see118–123). Findings of structural alterations in sub- associated with reduced hippocampal atrophy by puta- cortical regions, including the amygdala, hippocampus tively enhancing synaptic plasticity and neurogenesis125. and striatal regions have been less-consistent across stu- However, as the majority of the ENIGMA MDD cohorts dies. However, published studies have been primarily did not collect detailed information on the history of conducted using small sample sizes (typically N < 50 per antidepressant use, the duration of use, time since last group). Therefore, we performed an IPD-based meta- antidepressant treatment and dose of the antidepressant, analysis of subcortical volumes, lateral ventricle volumes, and given the cross-sectional nature of the studies, these and total ICV using data from N = 1101 people with findings cannot be interpreted as direct effects of anti- MDD (451 of whom exhibited suicidal ideation or beha- depressant medication use. MDD patients taking anti- vior) and N = 1996 healthy controls from seven research depressants at the time of scanning were likely the most cohorts participating in the ENIGMA MDD con- severe/chronic or recurrent patients in the sample, so the sortium124. Groups were identified based on the presence results are likely to be confounded by the severity or of suicidal ideation, defined as thinking about suicide or course of the disorder. Potential neuroprotective effects of taking one’s life, but without making any specific plan or antidepressant medication are more consistent with our acting upon those thoughts; suicidal planning, or the cortical surface area findings in adolescents with MDD, systematic formulation of a program of action that has showing lower cortical surface area in several regions in the potential to lead to a suicide attempt; and a suicide antidepressant-free adolescent patients compared with attempt, defined as any self-initiated action aimed at healthy controls but no differences between adolescent terminating one’s life, regardless of the method or degree patients taking antidepressants and healthy adolescents36. of its consequences. Because the number of suicide Confounding effects of recurrent or chronic illness were attempters (N = 14) was too small to allow any cross- also minimized in this group given their earlier stage of group comparison, the MDD individuals with suicide illness. attempts were grouped with those MDD individuals with In contrast to associations between antidepressant use suicide planning into a single category (suicidal behavior). and more pronounced cortical thickness and hippocampal No significant association of suicidal thoughts and volume abnormalities in adults with MDD, the meta- behavior with any of the subcortical volumes was analysis of white matter microstructure revealed no dif- found124. MDD patients reporting suicidal plans or ferences between adult patients who were taking anti- attempts did show a 2.87% smaller ICV (Cohen’s d = depressants at the time of scanning and healthy −0.284) than controls, but no significant differences were individuals38. Differences in white matter microstructure found when compared with the MDD patients with only were, however, present in adults with MDD who were suicidal ideation without a plan or those without suicidal antidepressant-free at the time of scanning compared ideation and behavior. These null findings with regard to with controls38. This finding was unexpected as the meta- subcortical volumes could perhaps be explained if addi- analysis of white matter microstructure and the meta- tional involvement of subcortical regions, and especially analysis of cortical thickness were both performed in a the hippocampus35, in suicidal thoughts and behaviors partly overlapping sample of adults with MDD with a beyond their role in MDD is subtle and only apparent in similar prevalence of recurrent episode patients (79% studies with very large sample sizes. Alternatively, given versus 71%). Therefore, if the greater and more wide- the highly heterogeneous nature of both MDD and sui- spread cortical thickness alterations were driven by a cidality, subcortical structural alterations may only more severe course of the disorder, a similar effect would become apparent in specific subgroups of people with have been expected with regard to white matter micro- suicidal thoughts and behaviors. Finally, it may be that structure. These findings raise the question of whether cortical structural alterations play a greater role than antidepressant medication may have differential effects on subcortical alterations in suicidal thoughts and behaviors. different characteristics of the brain (e.g., cortical thick- Efforts to identify cortical structural alterations associated ness versus surface area and white matter microstructure) with suicidal thoughts and behaviors are currently ongo- within the same individuals. Effects of antidepressant ing within ENIGMA MDD in a sample with a higher medication use on measures of gray matter and white prevalence of suicide attempts. matter microstructure require further investigation in a Schmaal et al. Translational Psychiatry (2020)10:172 Page 12 of 19 sample with more detailed and comprehensive informa- preceding the onset of MDD. If confirmed in future tion on antidepressant treatment (e.g., information on longitudinal studies, this could provide important infor- history, type of antidepressant, and duration of use) and mation for development of novel prevention and early simultaneous use of other medications (e.g., atypical intervention strategies for depression. antipsychotics). In addition, although many human and Many of the structural brain alterations identified in the animal studies have examined the effects of short-term ENIGMA MDD studies have smaller effect sizes than had antidepressant medication use on brain structure, there is been assumed based on previously published studies, even limited information on the effects of long-term anti- in more homogeneous subgroups of MDD patients. depressant medication use on brain structure. In this However, many of the larger effect sizes observed in prior respect, to disentangle indirectly associated phenomena studies may have been owing to small sample sizes and from causal effects of antidepressants, longitudinal studies publication bias. Effect sizes of neuroimaging measures are needed, preferably with a focus on long-term exposure have been shown to have noticeable instability up to as and at different stages of brain development and aging. many as 1000–2000 subjects (e.g., see Figure S1 in Miller et al.126). Indeed, large-scale studies, including studies that Scientific and clinical relevance pool existing data such as ENIGMA as well as large Our findings were computed from many data sets population-representative samples126,127, are beginning to combined, which has provided a more reliable estimate of show that variability in structural and functional brain effect sizes of structural brain alterations associated with imaging accounts for only a small percentage of the MDD than have individual small sample studies. This explained variance of clinical phenotypes. Thus, similar to inclusion of large-scale, diverse samples also enabled us to genetics literature, it appears that individual measures of calculate and report how reproducible these structural structural brain alterations account for limited variance in brain alterations are across data sets and how well find- complex phenotypes such as depression. ings generalize to cohorts with different ages of onset, These findings have important implications for our duration of illness, and with different geographic origins. theoretical understanding of MDD; small effect sizes Critically, most of our findings were based on a meta- make it unlikely that MDD can be explained by a generic analytic approach, which increases rigor. Moreover, we disease process, which is perhaps not surprising given the also extend retrospective meta-analyses of published multi-causal nature of this highly complex disorder. In studies by including data that have not been previously addition, from a clinical perspective, these small effect published owing to publication bias and by using har- sizes may make it unlikely for individual structural brain monized data processing and statistical analysis protocols measures to provide diagnostic biomarkers. Effect sizes across all data sets. between a Cohen’s d of 1.5 and 3 are likely to be required Our work has identified subtle structural brain altera- for a biomarker to be clinically useful, depending on the tions that are associated with specific demographic and nature of the application128. Nonetheless, given that data clinical characteristics of MDD. In particular, specific available in ENIGMA MDD are cross-sectional, it remains features of brain structure were differentially associated to be elucidated whether any of these structural brain with MDD at different stages of life and different stages of measures could serve as predictive or prognostic bio- illness. Specifically, the associations with hippocampal and markers, or as indices of treatment response that are amygdala volumes/shapes and cortical surface area were related to long-term mental and physical health out- documented in adults with an adolescent-onset MDD and comes. Furthermore, multiple factors with small effect in adolescents with MDD, respectively. In contrast, cor- sizes can be combined to create a large effect. Therefore, tical thickness reductions and white matter abnormalities the findings to date motivate future ENIGMA MDD were associated specifically with adult-onset MDD and studies to investigate whether the combination of differ- with older age in individuals with MDD and childhood ent neuroimaging modalities as well as combining neu- maltreatment (Fig. 3). Moreover, the subcortical and roimaging with clinical, psychosocial, and other biological white matter alterations found in patients with recurrent data modalities (e.g., using machine-learning methods) episodes compared with healthy controls, where absent in could explain more variance in the depressive phenotype, first-episode MDD patients, compared with healthy con- with the ultimate goal of developing clinically useful trols. These findings have generated novel hypotheses diagnostic or predictive tools. regarding different features of brain structure being involved in the onset and progression of depression at Future directions different stages of brain development and provide The ENIGMA MDD consortium is a dynamically important directions for future research. For example, evolving consortium, in which new research groups con- reductions in cortical surface may represent an early tinue to join and new projects are continually being developing subtype of depressive disorder, potentially initiated. Our first studies have mainly focused on Schmaal et al. Translational Psychiatry (2020)10:172 Page 13 of 19 case–control differences in structural brain measures that atypical depression) and the presence or absence of phe- can reliably be identified and replicated across many notypes closely related to MDD (e.g., obesity). samples worldwide. An important next step within In addition, it is also important to investigate associa- ENIGMA MDD is investigating higher dimensional tions with individual symptoms, as individual symptoms structural brain measures (e.g., using vertex-wise or voxel- differ in their impact on impairment of functioning, their wise analysis), which may be better able to detect subtle response to specific life events, their risk factors130, as well regional structural brain alterations in MDD, with as their response to treatment (e.g., Chekroud et al131.). potentially larger effect sizes. Future work to identify Therefore, several ongoing projects in ENIGMA MDD potential histological, genetic, and environmental have taken a dimensional approach to identify neural mechanisms underlying these structural brain alterations correlates of between-subject differences in the severity of is also underway. Furthermore, as can be seen in Fig. 1, individual symptoms (e.g., insomnia, suicidality). Asian research institutions are under-represented in Moreover, the pathophysiology of MDD is also likely ENIGMA MDD. Many research institutions in China highly heterogeneous. Different pathophysiological have shared neuroimaging data from individuals with mechanisms can result in similar symptoms for different depression with the REST-meta-MDD consortium, which individuals (equifinality) and the same underlying biolo- has recently published the first large-scale mega-analysis gical risk factors may result in a different expression of a on resting state functional MRI data of 1300 depressed certain disorder depending on an interaction with the patients and 1128 healthy controls from 25 research environment and genetic vulnerability (multifinality). In groups in China129. Future collaborations between the line with recent studies using brain imaging markers to ENIGMA MDD and REST-meta-MDD consortia will be identify subtypes of MDD defined by different profiles of important for identifying potential cultural differences in biological markers, so-called “biotypes” (e.g. Drysdale brain alterations associated with MDD. et al.132, but also see Dinga et al.133 for limitations asso- Future plans of the ENIGMA MDD consortium also ciated with this approach), we aim to investigate potential include (but are not limited to): (1) parsing the hetero- biotypes and their replicability across multiple cohorts in geneity of MDD, (2) moving beyond structural brain ENIGMA MDD. measures to include functional brain alterations, and (3) elucidating whether the identified neuroimaging markers Functional neuroimaging are unique to MDD or shared across mental disorders, An important next frontier in ENIGMA MDD will be to which are further discussed below. characterize brain functional deficits in MDD. Although the past two decades have witnessed a surge in studies on Addressing the heterogeneity of MDD resting state fMRI, the vast heterogeneity in analysis By pooling data across many samples worldwide, the methods, choice of seeds, templates, and parcellation ENIGMA MDD consortium performs studies encom- atlases has yielded a patchwork landscape of results in the passing a range of depressive phenotypes—from very mild literature. So far, only a few resting state meta- or mega- to severe and a broad range of previous treatments analyses exist for major depression129,134, either focusing received. This broad spectrum of depressive phenotypes on one specific network or combining results from dif- combined with a very large sample size provides the ferent analysis strategies across studies. We therefore plan opportunity to study the phenotypic and neurobiological to conduct a large resting state analysis by pooling data heterogeneity of MDD. Analyses within subtypes, rather from sites across the world, assessing a range of resting than across a heterogeneous diagnosis-based sample, state features using harmonized processing and a standard could reveal more pronounced changes in brain structure set of seeds, templates, and atlases. and function. For example, large samples allow the stra- In addition to rsfMRI, future endeavors will also include tification or clustering patients into different subgroups task fMRI. In the spirit of the Research Domain Criteria while preserving sufficient statistical power within each framework135, we will hone future analyses to task para- subgroup. digms that tap into functional domains relevant to In addition, heterogeneity could be addressed by depressive symptomatology, in particular, spanning the examining individual differences, for which large samples negative valence and cognitive domains. So far, meta- are required to capture the full range of variation in the analyses in MDD mainly have been based on coordinates phenotype. Such approaches may reveal clues for the of peak statistical difference reported in single published development of treatments tailored to subtypes or indi- studies14, using tools such as activation likelihood esti- vidual differences. At present, there are various ongoing mation (ALE)136. As potential case–control differences ENIGMA MDD projects that aim to address this het- with small effect sizes (typically not reported if not sig- erogeneity by examining associations between brain nificant in single studies) may be omitted in such meta- alterations and depressive symptom subtypes (e.g., analyses and previous negative findings may have not Schmaal et al. Translational Psychiatry (2020)10:172 Page 14 of 19 been published, it will be important to expand this work Given that all ENIGMA psychiatric disease working through applying IPD-based meta- and mega-analyses. groups use the same standardized preprocessing pipelines Therefore, we will conduct voxel- (or vertex-)wise meta- and analysis protocols, ENIGMA is well positioned to analyses, which have been shown to be superior to perform cross-disorder comparison studies. Several cross- coordinate-based meta-analyses137,138. A caveat is whe- disorder initiatives are ongoing, including comparisons of ther different versions of task paradigms—or even dif- brain morphology across SCZ, BD, and MDD as well as ferent tasks probing the same functional domain—can be across neurodevelopmental disorders (autism spectrum combined meaningfully in a meta- or mega-analysis. This disorder, attention deficit hyperactivity disorder, OCD, needs to be further confirmed empirically, although pre- and Tourette’s syndrome)146. The presence of suicidal liminary results from the ENIGMA task-based fMRI thoughts and behaviors and childhood maltreatment are workgroup are promising139. also relevant, as these constitute transdiagnostic Variance introduced by the different scanner types and constructs. acquisition parameters at each of the contributing sites cannot be avoided, but we can overcome the apparent Challenges of large-scale data-sharing initiatives heterogeneity in preprocessing, feature extraction, and sta- Worldwide data-sharing initiatives such as ENIGMA tistical testing of fMRI data by harmonizing software are not without their challenges. Some of these challenges packages, preprocessing settings, task contrasts, seed masks, encompass ethical and computational issues with regard parcellation atlases, network templates, and statistical to data sharing, as well as science and data sharing policies models across participating sites. Similarly, the same rig- that vary from one research institute to another, from orous quality assessment (QA) procedures should be country to country or even from continent to continent employed across participating sites, judging data quality and may change over time. This may restrict some against centrally defined criteria. Following the example of researchers from sharing raw neuroimaging data, using standardized analytical tools and QA procedures for although sharing de-identified, individual-level data may structural analyses in ENIGMA, such procedures and tools still be feasible. In addition, several challenges need to be are currently developed within the ENIGMA task-based addressed in translating the small to moderate effect sizes and ENIGMA resting state fMRI methods working groups. observed throughout ENIGMA MDD to the individua- lized and generalizable prediction of MDD-related phe- Identifying shared and unique brain alterations across notypes. The first is the need for rigorous testing and mental disorders validation of predictive models in multiple independent Until recently, all ENIGMA disease working groups samples. A challenge in this respect is the relative una- have focused on comparisons of a single disorder with vailability of deeply characterized phenotypes and long- healthy individuals using neuroimaging data. Results from itudinal data. To date, ENIGMA MDD has largely relied primary projects in ENIGMA have indicated that schi- on existing data, which implies a degree of heterogeneity zophrenia (SCZ)80, bipolar disorder (BD)140, and MDD36 with respect to phenotyping including clinical assess- patients are all characterized by lower prefrontal and ments, limiting the analysis of sources of clinical hetero- temporal cortical thickness relative to healthy control geneity. In addition, the current focus of ENIGMA MDD subjects. However, effect sizes differed between disorders, is on cross-sectional studies. Consequently, our findings the largest (up to Cohen’s d 0.5) having been observed in require further investigation in longitudinal studies to SCZ, followed by BD (Cohen’s d 0.3) and MDD (Cohen’s elucidate, for example, influences of brain development d 0.15). A similar gradient was observed for hippocampal and aging, medication effects, and the clinical relevance of volume across these disorders35,141,142, suggesting an the observed structural brain alterations in MDD. Com- “affective-psychotic severity continuum” (Fig. 4). More- bining longitudinal samples is not without its challenges, over, in ENIGMA, the obsessive–compulsive disorder but has already been successfully done for healthy indi- (OCD) consortium likewise found lower hippocampal viduals by the ENIGMA Plasticity working group147. volume in OCD patients versus healthy controls, but this Another limitation is that neuroimaging data were effect was at least partly driven by patients with comorbid collected using different MRI scanners, different sequen- MDD143, again suggestive of a shared mechanism. Nota- ces, different brain coverage and, for functional analyses, bly, hippocampal volume loss was also observed in post- different paradigm versions or acquisition lengths, which traumatic stress disorder even after accounting for may all introduce noise and further complicate the search childhood trauma144. Finally, the findings of structural for robust biological markers of MDD. Efforts to develop brain alterations in people with substance use disorders post-processing methods to reduce noise associated with from the ENIGMA Addiction consortium also overlap differences in scanning and other characteristics between with our findings in MDD, showing similar effect sizes for cohorts will be important, especially in the context of the hippocampus, insula, and medial OFC145. machine-learning analysis. Schmaal et al. Translational Psychiatry (2020)10:172 Page 15 of 19 Fig. 4 Subcortical volume and cortical thickness alterations in schizophrenia, bipolar disorder, and MDD. Results from the ENIGMA major depressive disorder (MDD), schizophrenia (SCZ), and bipolar disorder (BD) working groups suggest that there is considerable overlap in subcortical volume a and cortical thickness b alterations across these diagnostic groups. Most widespread effects and highest effect sizes were observed in SCZ (up to Cohen’s d 0.5), followed by BD (Cohen’s d 0.3), and with more local effects and lower effect sizes in MDD (Cohen’s d 0.15). Importantly, results displayed are based on case–control comparisons within each disorder separately and are not derived from direct comparisons between patient groups. Data were analyzed with the same harmonized methods across the disorders. Finally, it could be the case that certain findings Acknowledgements regarding the neurobiology of MDD will not be obtained ENIGMA MDD work is supported by NIH grants U54 EB020403 (Thompson), R01 MH116147 (Thompson), and R01 MH117601 (Jahanshad & Schmaal). LS by ever larger meta-analyses of existing samples; we may was supported by an NHMRC Career Development Fellowship (1140764). need alternative methods of data collection or new data AFFDIS cohort: this study was funded by the University Medical Center types that are sensitive to effects that are undetected Goettingen (UMG Startfoerderung) and the research team is supported by German Federal Ministry of Education and Research (Bundesministerium fuer today. We stress the need for a many-pronged approach Bildung und Forschung, BMBF: 01 ZX 1507, “PreNeSt - e:Med”). Barcelona using novel data collection and the coordinated analysis of cohort: MJP is funded by the Ministerio de Ciencia e Innovación of the Spanish the data already available, as well as the development of Government and by the Instituto de Salud Carlos III through a ‘Miguel Servet’ research contract (CP16–0020); National Research Plan (Plan Estatal de I + D + I new approaches. 2016–2019); and co-financed by the European Regional Development Fund (ERDF). BRC DeCC cohort: CHYF is supported by NIHR BRC. Calgary cohort: Conclusion supported by Canadian Institutes for Health Research, Branch Out Neurological Foundation. Cardiff cohort: supported by the Medical Research Council (grant Over the past 7 years since its initiation, ENIGMA G 1100629) and the National Center for Mental Health (NCMH), funded by MDD has brought together research groups across the Health Research Wales (HS/14/20). CLING cohort: this study was partially world with broad expertize to work together to gain a supported by the Deutsche Forschungsgemeinschaft (DFG) via grants to OG (GR1950/5–1 and GR1950/10–1). CODE cohort: Henrik Walter is supported by a better understanding of brain abnormalities associated grant of the Deutsche Forschungsgemeinschaft (WA 1539/4–1). The CODE with MDD. By addressing issues of underpowered stu- cohort was collected from studies funded by Lundbeck and the German dies, our work has provided more reliable estimates of Research Foundation (WA 1539/4–1, SCHN 1205/3–1, SCHR443/11–1). DIP- Groningen cohort: this study was supported by the Gratama Foundation, the the extent of structural brain abnormalities in depres- Netherlands (2012/35 to NG). Edinburgh cohort: The research leading to these sion, showing that variability in structural brain altera- results was supported by IMAGEMEND, which received funding from the tions may only account for a small percentage of the European Community’s Seventh Framework Programme (FP7/2007–2013) under grant agreement no. 602450. This paper reflects only the author’s views depression phenotype. Future work is underway that and the European Union is not liable for any use that may be made of the aims to address the heterogeneity of depression and to information contained therein. This work was also supported by a Wellcome integrate across data modalities to better understand the Trust Strategic Award 104036/Z/14/Z. FOR2107-Marburg cohort: funded by the German Research Foundation (DFG, grant FOR2107 KR 3822/7–2 to AK; multi-causal nature of depression, with the ultimate goal FOR2107 KI 588/14–2 to TK and FOR2107 JA 1890/7–2 to AJ). Houston cohorts: to help develop or select more effective treatments supported in part by NIMH grant R01 085667 and the Dunn Research for MDD. Foundation. JCS is supported by the Pat Rutherford, Jr. Endowed Chair in Schmaal et al. Translational Psychiatry (2020)10:172 Page 16 of 19 Psychiatry. IMH Study cohort: supported by funding from NHG (SIG/15012) and Molecular Bioscience, the University of Queensland, Brisbane, QLD, Australia. 20 NMRC CISSP (2018). Melbourne cohort: funded by National Health and Medical Department of Psychology, University of Minnesota, Minneapolis, MN, USA. 21 Research Council of Australia (NHMRC) Project Grants 1064643 (Principal Laboratory of Systems Neuroscience and Imaging in Psychiatry (SNIP-Lab), Investigator BJH) and 1024570 (Principal Investigator CGD). Minnesota cohort: University Medical Center Göttingen, Göttingen, Germany. 22Experimental the study was funded by the National Institute of Mental Health Therapeutics Branch, NIMH, NIH, Bethesda, MD, USA. 23Department of Psychiatry (K23MH090421; Dr. Cullen) and Biotechnology Research Center (P41 RR008079; and Psychotherapy, Jena University Hospital, Jena, Germany. 24School of Psychology, Center for Magnetic Resonance Research), the National Alliance for Research University of East London, London, UK. 25Centre for Affective Disorders, Institute of on Schizophrenia and Depression, the University of Minnesota Graduate Psychiatry, Psychology and Neuroscience, King’s College London, London, UK. 26 School, and the Minnesota Medical Foundation. This work was carried out in Department of Psychiatry and Psychotherapy, University Medicine Greifswald, part using computing resources at the University of Minnesota Greifswald, Germany. 27German Center for Neurodegenerative Diseases (DZNE), Site Supercomputing Institute. Münster cohort: funded by the German Research Rostock/Greifswald, Germany. 28Department of Psychiatry & Mental Health, Foundation (DFG, grant FOR2107 DA1151/5–1 and DA1151/5–2 to UD; SFB- University of Cape Town, Cape Town, South Africa. 29Section for Experimental TRR58, Projects C09 and Z02 to UD) and the Interdisciplinary Center for Clinical Psychopathology and Neuroimaging, Department of General Psychiatry, Heidelberg Research (IZKF) of the medical faculty of Münster (grant Dan3/012/17 to UD). University Hospital, Heidelberg, Germany. 30Illinois Institute of Technology, Chicago, NESDA cohort: The infrastructure for the NESDA study (www.nesda.nl) is IL, USA. 31Department of Psychology, Neuroscience & Behaviour, McMaster funded through the Geestkracht program of the Netherlands Organisation for University, Hamilton, ON, Canada. 32Melbourne Neuropsychiatry Centre, Department Health Research and Development (Zon-Mw, grant number 10–000–1002) and of Psychiatry, The University of Melbourne & Melbourne Health, Melbourne, VIC, is supported by participating universities (VU University Medical Center, GGZ Australia. 33Brain and Mind Centre, University of Sydney, Camperdown, NSW, inGeest, Arkin, Leiden University Medical Center, GGZ Rivierduinen, University Australia. 34Clinical Neuroscience Research Group, Department of Psychology, Medical Center Groningen) and mental health care organizations, see www. University of Oslo, Oslo, Norway. 35Department of Psychiatry, Diakonhjemmet nesda.nl. Pharmo cohort: supported by ERA-NET PRIOMEDCHILD FP 6 (EU) Hospital, Oslo, Norway. 36Norwegian Centre for Mental Disorders Research grant 11.32050.26. PSYABM-NORMENT: supported by the Research Council of (NORMENT), Institute of Clinical Medicine, University of Oslo, Oslo, Norway. 37 Norway (project number 229135). The South East Norway Health Authority Department of Psychiatry & Behavioral Sciences, The University of Texas Health Research Funding (project number 2015052). The Department of Psychology, Science Center at Houston, Houston, TX, USA. 38Faculty of Health Sciences, Oslo University of Oslo, Norway. San Francisco cohort: supported by NIH/NCCIH Metropolitan University, Oslo, Norway. 39Beth Israel Deaconess Medical Centre, 1R61AT009864–01A1. NIMH R01MH085734. SHIP and SHIP-trend cohorts: SHIP Harvard Medical School, Boston, MA, USA. 40Department of Psychiatry and is part of the Community Medicine Research net of the University of Psychotherapy, University of Marburg, Marburg, Germany. 41Sunshine Coast Mind Greifswald, Germany, which is funded by the Federal Ministry of Education and and Neuroscience Thompson Institute, University of the Sunshine Coast, Birtinya, Research (grants no. 01ZZ9603, 01ZZ0103, and 01ZZ0403), the Ministry of QLD, Australia. 42Division of Psychological Medicine and Clinical Neurosciences, Cultural Affairs and the Social Ministry of the Federal State of Mecklenburg- Cardiff University, Cardiff, UK. 43MRC Center for Neuropsychiatric Genetics and West Pomerania. MRI scans in SHIP and SHIP-TREND have been supported by a Genomics, Cardiff University, Cardiff, UK. 44Cardiff University Brain Research Imaging joint grant from Siemens Healthineers, Erlangen, Germany and the Federal Center, Cardiff University, Cardiff, UK. 45Psychiatry and Pediatrics, University of State of Mecklenburg-West Pomerania. Stanford cohorts: this work was Calgary, Addictions and Mental Health Strategic Clinical Network, Calgary, AB, supported by NIH grant R37 MH101495. The BiDirect Study was supported by Canada. 46Centre for Clinical Brain Science, University of Edinburgh, Edinburgh, UK. 47 grants from the German Federal Ministry of Education and Research (BMBF; Marburg University Hospital UKGM, Marburg, Germany. 48Institut d’Investigació grants FKZ-01ER0816 and FKZ-01ER1506). MDS is partially supported by an Biomèdica-Sant Pau, Barcelona, Spain. 49CIBERSAM, Madrid, Spain. 50Universitat award funded by the Phyllis and Jerome Lyle Rappaport Foundation. TCH is Autònoma de Barcelona, Barcelona, Spain. 51Department of Radiology and Nuclear supported by NIMH grant 5K01MH117442. EJWVS, JL, and TFB are supported Medicine, location AMC, Amsterdam UMC, Amsterdam, The Netherlands. 52 by European Research Council grant no. ERC-ADG-2014–671084 INSOMNIA. Department of Genetics & Computational Biology, QIMR Berghofer Medical TFB is supported by a VU University Amsterdam University Research Fellowship Research Institute, Brisbane, QLD, Australia. 53Center for Depression, Anxiety, and 2016–2017. JL is supported by a VU University Amsterdam University Research Stress Research, McLean Hospital, Harvard Medical School, Belmont, MA, USA. 54Max Fellowship 2017–2018. Planck Institute of Psychiatry, Munich, Germany. 55West Region/Institute of Mental Health, Singapore, Singapore. 56Yong Loo Lin School of Medicine/National University of Singapore, Singapore, Singapore. 57SA MRC Research Unit on Risk & Author details Resilience in Mental Disorders, Department of Psychiatry & Neuroscience Institute, 1 Orygen, The National Centre of Excellence in Youth Mental Health, Parkville, University of Cape Town, Cape Town, South Africa. 58Department of Psychiatry, VIC, Australia. 2Centre for Youth Mental Health, The University of Melbourne, Leiden University Medical Center, Leiden, The Netherlands. 59Leiden Institute for Parkville, VIC, Australia. 3Department of Psychology, Stanford University, Stanford, CA, Brain and Cognition, Leiden University Medical Center, Leiden, The Netherlands. USA. 4Department of Psychiatry & Behavioral Sciences, Stanford University, Stanford, 60 Curium-LUMC, Leiden University Medical Center, Leiden, The Netherlands. CA, USA. 5Department of Psychiatry & Weill Institute for Neurosciences, University of 61 Instituto ITACA, Universitat Politècnica de València, Valencia, Spain. 62Department California, San Francisco, CA, USA. 6Division of Mind and Brain Research, Department of Psychiatry and Psychotherapy, Jena, Germany. 63Clinical Affective Neuroimaging of Psychiatry and Psychotherapy CCM, Charité - Universitätsmedizin Berlin, corporate Laboratory, Leibniz Institute for Neurobiology, Magdeburg, Germany. 64Queensland member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Brain Institute, The University of Queensland, Brisbane, QLD, Australia. 65Centre for Institute of Health, Berlin, Germany. 7Department of Psychiatry, University of Advanced Imaging, The University of Queensland, Brisbane, QLD, Australia. 66Section Münster, Münster, Germany. 8Department of Sleep and Cognition, Netherlands on the Neurobiology and Treatment of Mood Disorders, National Institute of Mental Institute for Neuroscience (NIN), an institute of the Royal Netherlands Academy of Health, Bethesda, MD, USA. 67Imaging Genetics Center, Mark and Mary Stevens Arts and Sciences, Amsterdam, Netherlands. 9Department of Integrative Neuroimaging and Informatics Institute, Keck School of Medicine, University of Neurophysiology, Center for Neurogenomics and Cognitive Research (CNCR), Southern California, Marina del Rey, CA, USA Amsterdam Neuroscience, VU University Amsterdam, Amsterdam, The Netherlands. 10 Department of Psychiatry, Amsterdam UMC, Vrije Universiteit Amsterdam, Amsterdam Neuroscience, Amsterdam Public Health Research Institute, Amsterdam, The Netherlands. 11FSSBI Scientific Research Institute of Physiology & Basic Medicine, Laboratory of Affective, Cognitive & Translational Neuroscience, Novosibirsk, Russia. Conflict of interest 12 Department of Neuroscience, Novosibirsk State University, Novosibirsk, Russia. HJG has received travel grants and speakers honoraria from Fresenius Medical 13 Department of Biomedical Sciences of Cells and Systems, University Medical Care, Neuraxpharm, Servier and Janssen Cilag as well as research funding from Center Groningen, University of Groningen, Groningen, The Netherlands. Fresenius Medical Care. DJS has received grants and/or honoraria from Lundbeck 14 Department of Psychiatry, University of Melbourne, Melbourne, VIC, Australia. and Sun. MDS has received consulting fees from Vorso Corporation. Dr. Zarate is a 15 The Florey Institute of Neuroscience and Mental Health, University of Melbourne, full-time US government employee. He is listed as a co-inventor on a patent for Melbourne, VIC, Australia. 16Institute of Epidemiology and Social Medicine, University the use of ketamine and its metabolites in major depression and suicidal ideation. of Münster, Münster, Germany. 17Department of Psychiatry, Oxford University, Dr. Zarate has assigned his patent rights to the US government but will share a Oxford, UK. 18Oxford Health NHS Foundation Trust, Oxford, UK. 19Institute for percentage of any royalties that may be received by the government. Schmaal et al. Translational Psychiatry (2020)10:172 Page 17 of 19 Publisher’s note 23. Holland, D. et al. Beyond SNP heritability: polygenicity and discoverability Springer Nature remains neutral with regard to jurisdictional claims in estimated for multiple phenotypes with a univariate gaussian mixture model. published maps and institutional affiliations. PLoS Genet 16, e1008612 (2020). 24. Flint, J. & Kendler, K. S. The genetics of major depression. Neuron 81, 484–503 Received: 11 October 2019 Revised: 9 April 2020 Accepted: 7 May 2020 (2014). 25. Satizabal, C. L. et al. Genetic architecture of subcortical brain structures in over 40,000 individuals worldwide. Nat Genet 51, 1624–1636 (2019). 26. Hibar, D. P. P. et al. Common genetic variants influence human subcortical brain structures. Nature 520, 224–229 (2015). References 27. Grasby, K. L. et al. The genetic architecture of the human cerebral cortex. 1. World Health Organization. Depression and other common mental dis- Science 367, eaay6690 (2020). orders: global health estimates. 1–24 (2017). 28. Adams HHHHH et al. Novel genetic loci underlying human intracranial 2. Rush, A. J. et al. Acute and longer-term outcomes in depressed outpatients volume identified through genome-wide association. Nat. Neurosci. 19, requiring one or several treatment steps: a STAR*D report. Am. J. Psychiatry 1569–1582 (2016). 163, 1905–1917 (2006). 29. Hibar, D. P. P. et al. Novel genetic loci associated with hippocampal volume. 3. Machado, M., Iskedjian, M., Ruiz, I. & Einarson, T. R. Remission, dropouts, and Nat. Commun. 8, 13624 (2017). adverse drug reaction rates in major depressive disorder: a meta-analysis of 30. Stein, J. L. et al. Identification of common variants associated with human head-to-head trials. Curr. Med. Res. Opin. 22, 1825–1837 (2006). hippocampal and intracranial volumes. Nat. Genet 44, 552–561 (2012). 4. Thase, M. et al. Remission rates following antidepressant therapy with 31. Acheson, A. et al. Reproducibility of tract-based white matter microstructural bupropion or selective serotonin reuptake inhibitors: a meta-analysis of ori- measures using the ENIGMA-DTI protocol. Brain Behav. 7, 1–10 (2017). ginal data from 7 randomized controlled trials. J. Clin. Psychiatry 66, 974–981 32. Adhikari, B. M. et al. Comparison of heritability estimates on resting state fMRI (2005). connectivity phenotypes using the ENIGMA analysis pipeline. Hum. Brain 5. Trivedi, M. H. et al. Evaluation of outcomes with citalopram for depression Mapp. 39, 4893–4902 (2018). using measurement-based care in STAR*D: implications for clinical practice. 33. Jahanshad, N. et al. Multi-site genetic analysis of diffusion images and vox- Am. J. Psychiatry 163, 28–40 (2006). elwise heritability analysis: a pilot project of the ENIGMA-DTI working group. 6. Cuijpers, P. et al. The effects of psychotherapies for major depression in Neuroimage 81, 455–469 (2013). adults on remission, recovery and improvement: a meta-analysis. J. Affect. 34. Kochunov, P. et al. Heritability of fractional anisotropy in human white matter: Disord. 159, 118–126 (2014). a comparison of Human Connectome Project and ENIGMA-DTI data. Neu- 7. Mulders, P. C., Eijndhoven, P. F., Van Schene, A. H., Beckmann, C. F. & Ten- roimage 111, 300–301. (2015). dolkar, I. Neuroscience and biobehavioral reviews resting-state functional 35. Schmaal, L. et al. Subcortical brain alterations in major depressive disorder: connectivity in major depressive disorder: a review. Neurosci. Biobehav. Rev. findings from the ENIGMA Major Depressive Disorder working group. Mol. 56, 330–344 (2015). Psychiatry 21, 806–812 (2016). 8. Kempton, M. J. Structural neuroimaging studies in major depressive disorder. 36. Schmaal, L. et al. Cortical abnormalities in adults and adolescents with major Arch. Gen. Psychiatry 68, 675–690 (2011). depression based on brain scans from 20 cohorts worldwide in the ENIGMA 9. Kapur, S., Phillips, A. G. & Insel, T. R. Why has it taken so long for biological major depressive disorder working group. Mol. Psychiatry 22, 900–909 (2017). psychiatry to develop clinical tests and what to do about it. Mol. Psychiatry 37. Ho T. C. et al. Subcortical shape alterations in major depressive disorder: 17, 1174–1179 (2012). findings from the ENIGMA Major Depressive Disorder Working Group. Hum. 10. Button, K. S. et al. Power failure: why small sample size undermines the Brain Mapp. 2020. https://doi.org/10.1002/hbm.24988. [Epub ahead of print]. reliability of neuroscience Katherine. Nat. Rev. Neurosci. 14, 365–376 (2013). 38. van Velzen L. S. et al. White matter disturbances in major depressive disorder: 11. Ioannidis, J. P. A. Why most discovered true associations are inflated. Epide- a coordinated analysis across 20 international cohorts in the ENIGMA MDD miology 19, 640–648 (2008). working group. Mol. Psychiatry 2019. https://doi.org/10.1038/s41380-019- 12. Algermissen, J. & Mehler, D. M. A. May the power be with you: are there 0477-2. [Epub ahead of print]. highly powered studies in neuroscience, and how can we get more of 39. Fischl, B. et al. Whole brain segmentation: automated labeling of neuroa- them? J. Neurophysiol. 119, 2114–2117 (2018). natomical structures in the human brain. Neuron 33, 341–355 (2002). 13. Miller, J. What is the probability of replicating a statistically significant effect? 40. Burke, D. L., Ensor, J. & Riley, R. D. Meta-analysis using individual participant Psychon. Bull. Rev. 16, 617–640 (2009). data: one-stage and two-stage approaches, and why they may differ. Stat. 14. Müller, V. I. et al. Altered brain activity in unipolar depression revisited: meta- Med. 36, 855–875 (2017). analyses of neuroimaging studies. JAMA Psychiatry 74, 47–55 (2017). 41. Debray, T. P. A., Moons, K. G. M., Abo-Zaid, G. M. A., Koffijberg, H. & Da Riley, R. 15. Sudlow, C. et al. UK Biobank: an openaccess resource for identifying the Individual participant data meta-analysis for a binary outcome: one-stage or causes of a wide range of complex diseasesof middle and old age. PLoS Med. two-stage? PLoS ONE 8, 1–10. (2013). 12, 1–10. (2015). 42. Boedhoe, P. S. W. et al. An empirical comparison of meta- and mega-analysis 16. Howard, D. M. et al. Genome-wide association study of depression pheno- with data from the ENIGMA obsessive-compulsive disorder Working Group. types in UK Biobank identifies variants in excitatory synaptic pathways. Nat. Front. Neuroinform. 12, 1–8 (2019). Commun. 9, 1–10 (2018). 43. McKinnon, M. C., Yucel, K., Nazarov, A. & MacQueen, G. M. A meta-analysis 17. Shen, X. et al. Subcortical volume and white matter integrity abnormalities in examining clinical predictors of hippocampal volume in patients with major major depressive disorder: Findings from UK Biobank imaging data. Sci. Rep. depressive disorder. J. Psychiatry Neurosci. 34, 41–54 (2009). 7, 1–10. (2017). 44. Campbell, S., Marriott, M., Nahmias, C. & Macqueen, G. M. Lower hippo- 18. Harris M. et al. Stratifying major depressive disorder by polygenic risk for campal volume in patients suffering from depression: a meta-analysis. Am. J. schizophrenia in relation to structural brain measures. Psychol Med 2019. Psychiatry 161, 598–607 (2004). https://doi.org/10.1017/S003329171900165X. [Epub ahead of print] 45. Stratmann, M. et al. Insular and hippocampal gray matter volume reductions 19. Thompson, P. M. et al. The ENIGMA Consortium: large-scale collaborative in patients with major depressive disorder. PLoS ONE 9, e102692 (2014). analyses of neuroimaging and genetic data. Brain Imaging Behav. 8, 153–182 46. Elbejjani, M. et al. Depression, depressive symptoms, and rate of hippo- (2014). campal atrophy in a longitudinal cohort of older men and women. Psychol. 20. Thompson, P. M. et al. ENIGMA and the individual: predicting factors Med. 45, 1931–1944 (2015). that affect the brain in 35 countries worldwide. Neuroimage 145, 47. Phillips, J. L., Batten, L. A., Tremblay, P., Aldosary, F. & Blier, P. A prospective, 389–408 (2017). longitudinal study of the effect of remission on cortical thickness and hip- 21. Thompson, P. M. et al. ENIGMA and global neuroscience: a decade of large- pocampal volume in patients with treatment-resistant depression. Int J. scale studies of the brain in health and disease across more than 40 Neuropsychopharmacol. 18, pyv037 (2015). countries. Transl. Psychiatry 10, 100 (2020). 48. Frodl, T. et al. Effect of hippocampal and amygdala volumes on clinical 22. Gottesman, I. I. & Gould, T. D. The endophenotype concept in psychiatry: outcomes in major depression: a 3-year prospective magnetic resonance etymology and strategic intentions. Am. J. Psychiatry 160, 636–645 (2003). imaging study. J. Psychiatry Neurosci. 33, 423–430 (2008). Schmaal et al. Translational Psychiatry (2020)10:172 Page 18 of 19 49. Taylor, W. D. et al. Hippocampus atrophy and the longitudinal course of late- 77. Opel, N. et al. Mediation of the influence of childhood maltreatment on life depression. Am. J. Geriatr. Psychiatry 22, 1504–1512 (2014). depression relapse by cortical structure: a 2-year longitudinal observational 50. Oltedal, L. et al. Volume of the human hippocampus and clinical response study. Lancet Psychiatry 6, 318–326 (2019). following electroconvulsive therapy. Biol. Psychiatry 84, 574–581 (2018). 78. Eyler, L. T. et al. A comparison of heritability maps of cortical surface area and 51. Arnone, D. et al. State-dependent changes in hippocampal grey matter in thickness and the influence of adjustment for whole brain measures: a depression. Mol. Psychiatry 18, 1265–1272 (2013). magnetic resonance imaging twin study. Twin Res. Hum. Genet. 15, 304–314 52. Barch, D. M. et al. Hippocampal volume and depression among young (2012). children. Psychiatry Res. Neuroimaging 288, 21–28 (2019). 79. Fjell, A. M. et al. Continuity and discontinuity in human cortical development 53. MacMaster, F. P. et al. Amygdala and hippocampal volumes in familial early and change from embryonic stages to old age. Cereb. Cortex 29, 3879–3890 onset major depressive disorder. Biol. Psychiatry 63, 385–390 (2008). (2018). 54. MacMaster, F. P. & Kusumakar, V. Hippocampal volume in early onset 80. van Erp, T. G. M. et al. Cortical brain abnormalities in 4474 individuals with depression. BMC Med. 2, 2 (2004). schizophrenia and 5098 control subjects via the enhancing neuro imaging 55. Kim, J., Suh, S., Lee, H., Lee, J. & Lee, M. Cortical and subcortical gray matter genetics through meta analysis (ENIGMA) consortium. Biol. Psychiatry 84, alterations in first-episode drug-naïve adolescents with major depressive 644–654 (2018). disorder. Neuroreport 30, 1172–1178 (2019). 81. Jalbrzikowski, M. et al. Structural brain alterations in youth with psychosis and 56. Shen, Z. et al. Changes of grey matter volume in first-episode drug-naive bipolar spectrum symptoms. J. Am. Acad. Child Adolesc. Psychiatry 58, adult major depressive disorder patients with different age-onset. Neuro- 1079–1091 (2019). Image Clin. 12, 492–498 (2016). 82. Chung, Y. et al. Cortical abnormalities in youth at clinical high-risk for psy- 57. Jaworska, N. et al. Influence of age of onset on limbic and paralimbic chosis: findings from the NAPLS2 cohort. NeuroImage Clin. 23, 101862 (2019). structures in depression. Psychiatry Clin. Neurosci. 68, 812–820 (2014). 83. de Zwarte, S. M. C. et al. Running in the family? Structural brain abnormalities 58. Hill, J., Pickles, A., Rollinson, L., Davies, R. & Byatt, M. Juvenile- versus adult- and IQ in offspring, siblings, parents, and co-twins of patients with schizo- onset depression: multiple differences imply different pathways. Psychol. Med. phrenia. Schizophr. Bull. 45, 1209–1217 (2018). 34, 1483–1493 (2004). 84. van der Vinne, N., Vollebregt, M. A., van Putten, M. J. A. M. & Arns, M. Frontal 59. Jaffee, S. R. et al. Differences in early childhood rsk factors for juvenile-onset alpha asymmetry as a diagnostic marker in depression: fact or fiction? A and adult-onset depression. Arch. Gen. Psychiatry 59, 215–222 (2002). meta-analysis. NeuroImage Clin. 16, 79–87 (2017). 60. Kendler, K. S., Fiske, A., Gardner, C. O. & Gatz, M. Delineation of two genetic 85. Kumar, A., Bilker, W., Lavretsky, H. & Gottlieb, G. Volumetric asymmetries in pathways to major depression. Biol. Psychiatry 65, 808–811 (2009). late-onset mood disorders: an attenuation of frontal asymmetry with 61. Du, M. et al. Brain grey matter volume alterations in late-life depression. J. depression severity. Psychiatry Res. Neuroimaging 100, 41–47 (2000). Psychiatry Neurosci. 39, 397–406 (2014). 86. Liu, W. et al. Structural asymmetry of dorsolateral prefrontal cortex correlates 62. Hamilton, J. P., Siemer, M. & Gotlib, I. H. Amygdala volume in major with depressive symptoms: evidence from healthy individuals and patients depressive disorder: a meta-analysis of magnetic resonance imaging studies. with major depressive disorder. Neurosci. Bull. 32, 217–226 (2016). Mol. Psychiatry 13, 993–1000 (2008). 87. de Kovel, C. G. F. et al. No alterations of brain structural asymmetry in major 63. Hammen, C. Stress and depression. Annu Rev. Clin. Psychol. 1, 293–319 (2005). depressive disorder: an ENIGMA consortium analysis. Am. J. Psychiatry 176, 64. Winkler, A. M. et al. Cortical thickness or grey matter volume? The importance 1039–1049 (2019). of selecting the phenotype for imaging genetics studies. Neuroimage 53, 88. Penninx, B. W. J. H. Depression and cardiovascular disease: epidemiological 1135–1146 (2010). evidence on their linking mechanisms. Neurosci. Biobehav Rev. 74, 277–286 65. Panizzon, M. S. et al. Distinct genetic influences on cortical surface area and (2017). cortical thickness. Cereb. Cortex 19, 2728–2735 (2009). 89. Spiegel, D. & Giese-Davis, J. Depression and cancer: mechanisms and disease 66. Bora, E., Fornito, A., Pantelis, C. & Yücel, M. Gray matter abnormalities in major progression. Biol. Psychiatry 54, 269–282 (2003). depressive disorder: a meta-analysis of voxel based morphometry studies. J. 90. Koutsouleris, N. et al. Accelerated brain aging in schizophrenia and beyond: a Affect Disord. 138, 9–18 (2012). neuroanatomical marker of psychiatric disorders. Schizophr. Bull. 40, 67. Lai, C. H. Gray matter volume in major depressive disorder: a meta-analysis of 1140–1153 (2014). voxel-based morphometry studies. Psychiatry Res. 211, 37–46 (2013). 91. Han L. K. M. et al. Brain aging in major depressive disorder: results from the 68. Arnone, D., McIntosh, A. M., Ebmeier, K. P., Munafò, M. R. & Anderson, I. M. ENIGMA Major Depressive Disorder working group. Mol. Psychiatry 2020. Magnetic resonance imaging studies in unipolar depression: systematic https://doi.org/10.1038/s41380-020-0754-0. [Epub ahead of print]. review and meta-regression analyses. Eur. Neuropsychopharmacol. 22, 1–16 92. Cole, J. H., Marioni, R. E., Harris, S. E. & Deary, I. J. Brain age and other bodily (2012). ‘ages’: implications for neuropsychiatry. Mol. Psychiatry 24, 266–281 (2019). 69. Koolschijn, P. C. M. P., van Haren, N. E. M., Lensvelt-Mulders, G. J. L. M., Hulshoff 93. Le, T. T. et al. Effect of ibuprofen on BrainAGE: a randomized, placebo- Pol, H. E. & Kahn, R. S. Brain volume abnormalities in major depressive controlled, dose-response exploratory study. Biol. Psychiatry Cogn. Neurosci. disorder: a meta-analysis of magnetic resonance imaging studies. Hum. Brain Neuroimaging 3, 836–843 (2018). Mapp. 30, 3719–3735 (2009). 94. Kakeda, S. et al. Relationship between interleukin (IL)-6 and brain morphol- 70. Lemaitre, H. et al. Nonuniformity across cortical thickness. Neurobiol. Aging ogy in drug-naïve, first-episode major depressive disorder using surface- 33, 617.e1–617.e9 (2012). based morphometry. Sci. Rep. 8, 2–10. (2018). 71. Storsve, A. B. et al. Differential longitudinal changes in cortical thickness, 95. Sajedi, H. & Pardakhti, N. Age prediction based on brain MRI image: a survey. surface area and volume across the adult life span: regions of accelerating J. Med Syst. 43, 279 (2019). and decelerating change. J. Neurosci. 34, 8488–8498 (2014). 96. Alexander, A. L. et al. Characterization of cerebral white matter properties 72. Hutton, C., Draganski, B., Ashburner, J. & Weiskopf, N. A comparison between using quantitative magnetic resonance imaging stains. Brain Connect 1, voxel-based cortical thickness and voxel-based morphometry in normal 423–446 (2011). aging. Neuroimage 48, 371–380 (2009). 97. Song, S. K. et al. Dysmyelination revealed through MRI as increased radial 73. Jha, S. C. et al. Environmental influences on infant cortical thickness and (but unchanged axial) diffusion of water. Neuroimage 17, 1429–1436 (2002). surface area. Cereb. Cortex 29, 1139–1149 (2018). 98. Song, S. K. et al. Demyelination increases radial diffusivity in corpus callosum 74. Migliorini, R. et al. Anterior cingulate cortex surface area relates to behavioral of mouse brain. Neuroimage 26, 132–140 (2005). inhibition in adolescents with and without heavy prenatal alcohol exposure. 99. Chen, G. et al. Disorganization of white matter architecture in major Behav. Brain Res 292, 26–35 (2015). depressive disorder: a meta-analysis of diffusion tensor imaging with tract- 75. Tozzi L. et al. Interactive impact of childhood maltreatment, depression, and based spatial statistics. Sci. Rep. 6, 1–12. (2016). age on cortical brain structure: mega-analytic findings from a large multi-site 100. Choi, S. et al. Association of brain-derived neurotrophic factor DNA methy- cohort. Psychol. Med. 50, 1020–1031 (2020). lation and reduced white matter integrity in the anterior corona radiata in 76. Kelly, P. et al. Cortical thickness, surface area, and gyrification abnormalities in major depression. J. Affect. Disord. 172, 74–80 (2015). children exposed to maltreatment: neural markers of vulnerability? Biol. 101. Sanjuan, P. M., Thoma, R., Claus, E. D., Mays, N. & Caprihan, A. Reduced white Psychiatry 74, 845–852 (2013). matter integrity in the cingulum and anterior corona radiata in posttraumatic Schmaal et al. Translational Psychiatry (2020)10:172 Page 19 of 19 stress disorder in male combat veterans: a diffusion tensor imaging study. 124. Renteria, M. E. et al. Subcortical brain structure and suicidal behaviour in Psychiatry Res. Neuroimaging 214, 260–268 (2013). major depressive disorder: a meta-analysis from the ENIGMA-MDD working 102. Gur, R. E. & Gur, R. C. Sex differences in brain and behavior in adolescence: group. Transl. Psychiatry 7, e1116 (2017). findings from the Philadelphia Neurodevelopmental Cohort. Neurosci. Bio- 125. Pilar-Cuéllar, F. et al. Neural plasticity and proliferation in the generation of behav Rev. 70, 159–170 (2016). antidepressant effects: hippocampal implication. Neural Plast. 2013, 537265 103. Giedd, J. N., Raznahan, A., Mills, K. L. & Lenroot, R. K. Review: magnetic (2013). resonance imaging of male/female differences in human adolescent brain 126. Miller, K. L. et al. Multimodal population brain imaging in the UK Biobank anatomy. Biol. Sex. Differ. 3, 1–9 (2012). prospective epidemiological study. Nat. Neurosci. 19, 1523–1536 (2016). 104. Green, J. G. et al. Childhood adversities and adult psychiatric disorders in the 127. Pagliaccio D., Alqueza K. L., Marsh R., Auerbach R. P. Brain volume abnorm- national comorbidity survey replication I: associations with first onset of DSM- alities in youth at high risk for depression: adolescent brain and cognitive IV disorders. Arch. Gen. Psychiatry 67, 113–123 (2010). development study. J Am Acad Child Adolesc Psychiatry 2019. https://doi.org/ 105. McLaughlin, K. A. et al. Childhood adversities and first onset of psychiatric 10.1016/j.jaac.2019.09.032. [Epub ahead of print] disorders in a national sample of US adolescents. Arch. Gen. Psychiatry 69, 128. Castellanos, F. X., Di Martino, A., Craddock, R. C., Mehta, A. D. & Milham, M. P. 1151–1160 (2012). Clinical applications of the functional connectome. Neuroimage 80, 527–540 106. Kessler, R. C. et al. Childhood adversities and adult psychopathology in the (2013). WHO world mental health surveys. Br. J. Psychiatry 197, 378–385 (2010). 129. Yan, C. G. et al. Reduced default mode network functional connectivity in 107. Frodl, T. et al. Childhood stress, serotonin transporter gene and brain struc- patients with recurrent major depressive disorder. Proc. Natl Acad. Sci. USA tures in major depression. Neuropsychopharmacology 35, 1383–1390 (2010). 116, 9078–9083 (2019). 108. Dannlowski, U. et al. Limbic scars: long-term consequences of childhood 130. Fried, E. I. & Nesse, R. M. Depression sum-scores don’t add up: why analyzing maltreatment revealed by functional and structural magnetic resonance specific depression symptoms is essential. BMC Med. 13, 1–11. (2015). imaging. Biol. Psychiatry 71, 286–293 (2012). 131. Chekroud, A. M. et al. Reevaluating the efficacy and predictability of anti- 109. Teicher, M. H., Samson, J. A., Anderson, C. M. & Ohashi, K. The effects of depressant treatments: a symptom clustering approach. JAMA Psychiatry 74, childhood maltreatment on brain structure, function and connectivity. Nat. 370–378 (2017). Rev. Neurosci. 17, 652–666 (2016). 132. Drysdale, A. T. et al. Resting-state connectivity biomarkers define neuro- 110. Frodl, T. et al. Childhood adversity impacts on brain subcortical structures physiological subtypes of depression. Nat. Med. 23, 28–38 (2016). relevant to depression. J. Psychiatr. Res. 86, 58–65 (2017). 133. Dinga, R. et al. Evaluating the evidence for biotypes of depression: attempted 111. O’Doherty, J. P. Reward representations and reward-related learning in the replication of Drysdale et.al. 2017. NeuroImage Clin. 22, 101796 (2019). human brain: insights from neuroimaging. Curr. Opin. Neurobiol. 14, 769–776 134. Kaiser, R. H., Andrews-Hanna, J. R., Wager, T. D. & Pizzagalli, D. A. Large-scale (2004). network dysfunction in major depressive disorder: a meta-analysis of resting- 112. Haber, S. N. & Knutson, B. The reward circuit: linking primate anatomy and state functional connectivity. JAMA Psychiatry 72, 603–611 (2015). human imaging. Neuropsychopharmacology 35, 4–26 (2010). 135. Cuthbert, B. N. Research domain criteria: toward future psychiatric nosologies. 113. Chen, Y. & Baram, T. Z. Toward understanding how early-life stress repro- Dialogues Clin. Neurosci. 17, 89–97 (2015). grams cognitive and emotional brain networks. Neuropsychopharmacology 136. Eickhoff, S. B., Bzdok, D., Laird, A. R., Kurth, F. & Fox, P. T. Activation likelihood 41, 197–206 (2016). estimation meta-analysis revisited. Neuroimage 59, 2349–2361 (2012). 114. Sheridan, M. A. et al. Early deprivation disruption of associative learning is a 137. Salimi-Khorshidi, G., Smith, S. M., Keltner, J. R., Wager, T. D. & Nichols, T. E. developmental pathway to depression and social problems. Nat. Commun. 9, Meta-analysis of neuroimaging data: a comparison of image-based and 1–8 (2018). coordinate-based pooling of studies. Neuroimage 45, 810–823 (2009). 115. Baldwin, J. R., Reuben, A., Newbury, J. B. & Danese, A. Agreement between 138. Albajes-Eizagirre, A. & Radua, J. What do results from coordinate-based meta- prospective and retrospective measures of childhood maltreatment: a sys- analyses tell us? Neuroimage 176, 550–553 (2018). tematic review and meta-analysis. JAMA Psychiatry 76, 584–593 (2019). 139. Veer I., Waller L., Lett T., Erk S., Walter H. ENIGMA task-based fMRI: a work- 116. Holma, K. M. et al. Differences in incidence of suicide attempts between group studying the genetic basis of task-evoked brain activity. In: 25th bipolar I and II disorders and major depressive disorder. Bipolar Disord. 16, Organization of Human Brain Mapping Annual Meeting. 2019. 652–661 (2014). 140. Hibar, D. P. et al. Cortical abnormalities in bipolar disorder: an MRI analysis of 117. Schaffer, A. et al. International society for bipolar disorders task force on 6503 individuals from the ENIGMA Bipolar Disorder Working Group. Mol. suicide: meta-analyses and meta-regression of correlates of suicide attempts Psychiatry 23, 932–942 (2018). and suicide deaths in bipolar disorder. Bipolar Disord. 17, 1–16 (2015). 141. van Erp, T. G. M. et al. Subcortical brain volume abnormalities in 2028 indi- 118. Balcioglu, Y. H. & Kose, S. Neural substrates of suicide and suicidal behaviour: viduals with schizophrenia and 2540 healthy controls via the ENIGMA con- from a neuroimaging perspective. Psychiatry Clin. Psychopharmacol. 28, sortium. Mol. Psychiatry 21, 547–553 (2016). 314–328 (2018). 142. Hibar, D. P. et al. Subcortical volumetric abnormalities in bipolar disorder. Mol. 119. Bani-Fatemi, A. et al. Structural and functional alterations of the suicidal brain: Psychiatry 21, 1710–1716 (2016). an updated review of neuroimaging studies. Psychiatry Res. Neuroimaging 143. Boedhoe, P. S. W. et al. Distinct subcortical volume alterations in pediatric and 278, 77–91 (2018). adult OCD: a worldwide meta- and mega-analysis. Am. J. Psychiatry 174, 120. Domínguez-Baleón, C., Gutiérrez-Mondragón, L. F., Campos-González, A. I. & 60–69 (2017). Rentería, M. E. Neuroimaging studies of suicidal behavior and non-suicidal 144. Logue, M. W. et al. Smaller hippocampal volume in posttraumatic stress self-injury in psychiatric patients: a systematic review. Front. Psychiatry 9, 500 disorder: a multisite ENIGMA-PGC study: subcortical volumetry results from (2018). posttraumatic stress disorder consortia. Biol. Psychiatry 83, 244–253 (2018). 121. Cox Lippard, E. T., Johnston, J. A. & Blumberg, H. P. Neurobiological risk factors 145. Mackey, S. et al. Mega-analysis of gray matter volume in substance depen- for suicide insights from brain imaging. Am. J. Prev. Med. 47, S152–S162 dence: general and substance-specific regional effects. Am. J. Psychiatry 176, (2014). 119–128 (2019). 122. van Heeringen, K., Bijttebier, S., Desmyter, S., Vervaet, M. & Baeken, C. Is there a 146. Boedhoe P. et al. Subcortical brain volume, regional cortical thickness and neuroanatomical basis of the vulnerability to suicidal behavior? A coordinate- cortical surface area across attention-deficit/hyperactivity disorder (ADHD), based meta-analysis of structural and functional MRI studies. Front Hum. autism spectrum disorder (ASD), and obsessive-compulsive disorder (OCD). Neurosci. 8, 1–8 (2014). Am J Psychiatry 2020 [in press]. 123. Schmaal, L. et al. Imaging suicidal thoughts and behaviors: a comprehensive 147. Brouwer, R. M. et al. Genetic influences on individual differences in long- review of 2 decades of neuroimaging studies. Mol. Psychiatry 25, 408–427 itudinal changes in global and subcortical brain volumes: results of the (2020). ENIGMA plasticity working group. Hum. Brain Mapp. 38, 4444–4458 (2017).