Schizophrenia Bulletin Open https://doi.org/10.1093/schizbullopen/sgac040 The Relationship Between Grey Matter Volume and Clinical and Functional Outcomes in People at Clinical High Risk for Psychosis Stefania Tognin*,1,2,22, , Anja Richter1,22, , Matthew J. Kempton1,3, Gemma Modinos1, , Mathilde Antoniades1, Matilda Azis1, , Paul Allen1,4, Matthijs G. Bossong1, Jesus Perez5, Christos Pantelis6, , Barnaby Nelson7,8, Paul Amminger7,8, Anita Riecher-Rössler9, , Neus Barrantes-Vidal10, , Marie-Odile Krebs11, Birte Glenthøj12,13, Stephan Ruhrmann14, Gabriele Sachs15, , Bart P. F. Rutten16, Lieuwe de Haan17, Mark van der Gaag18,19, EU-GEI High Risk Study Group21, Lucia R. Valmaggia20, and Philip McGuire1,3 1 Department of Psychosis Studies, Institute of Psychiatry, Psychology & Neuroscience, King’s College London, UK; 2South London and Maudsley NHS Foundation Trust, London, UK; 3National Institute for Health Research (NIHR) Biomedical Research Centre (BRC), UK; 4Department of Psychology, University of Roehampton, London, UK; 5CAMEO Early Intervention in Psychosis Services, Cambridgeshire and Peterborough NHS Foundation Trust, Cambridge, UK; 6Department of Psychiatry, Melbourne Neuropsychiatry Centre, University of Melbourne & Melbourne Health, Carlton South, Victoria, Australia; 7Orygen, Parkville, Victoria, Australia; 8 Centre for Youth Mental Health, The University of Melbourne, Parkville, Victoria, Australia; 9University of Basel, Faculty of Medicine, Basel, Switzerland; 10Departament de Psicologia Clínica i de la Salut (Universitat Autònoma de Barcelona), Fundació Sanitària Sant Pere Claver (Spain), Center for Biomedical Research in Mental Health (CIBERSAM), Madrid, Spain; 11University of Paris, GHU-Paris, Sainte-Anne, C’JAAD, Hospitalo-Universitaire department SHU, Inserm U1266, Institut de Psychiatrie (CNRS 3557), Paris, France; 12 Center for Neuropsychiatric Schizophrenia Research, CNSR, and Center for Clinical Intervention and Neuropsychiatric Schizophrenia Research, CINS, Mental Health Services Capital Region of Denmark, Mental Health Center Glostrup, Glostrup, Denmark; 13 Department of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark; 14 Department of Psychiatry and Psychotherapy, Faculty of Medicine and University Hospital University of Cologne, Cologne, Germany; 15Department of Psychiatry and Psychotherapy, Medical University of Vienna, Vienna, Austria; 16Department of Psychiatry and Neuropsychology, Faculty of Health, Medicine and Life Sciences, School for Mental Health and Neuroscience (MHeNS), Maastricht University, Maastricht, The Netherlands; 17Early Psychosis Department, Amsterdam UMC, Amsterdam, The Netherlands; 18 Department of Clinical Psychology and Amsterdam Public Mental Health Research Institute, Faculty of Behavioural and Movement Sciences, VU University, Amsterdam, The Netherlands; 19Department of Psychosis Research, Parnassia Psychiatric Institute, The Hague, The Netherlands; 20Department of Psychology, Institute of Psychiatry, Psychology & Neuroscience, King’s College London, UK ; 21See Appendix for EU-GEI High Risk Study Group Author list.; 22These authors equally contributed to this work. *To whom correspondence should be addressed; Room M5.09, Department of Psychosis Studies, Institute of Psychiatry, Psychology & Neuroscience, King’s College London, Room M5.09, PO Box 67, De Crespigny Park, London SE5 8AF, UK; tel: +44 (0)20 7848 0415, fax: +44 (0)20 7848 0976; e-mail: Stefania.tognin@kcl.ac.uk Objective: To examine the association between baseline gyrus, cingulate gyrus, gyrus rectus, inferior temporal gyrus, alterations in grey matter volume (GMV) and clinical and parahippocampal gyrus, striatum, and hippocampus). Using functional outcomes in people at clinical high risk (CHR) for voxel-based morphometry, we analysed the relationship be- psychosis. tween GMV and clinical and functional outcomes. Methods: 265 CHR individuals and 92 healthy controls were Results: Within the CHR sample, a poor functional out- recruited as part of a prospective multi-center study. After come (GAF < 65) was associated with relatively lower a baseline assessment using magnetic resonance imaging GMV in the right striatum at baseline (P < .047 after (MRI), participants were followed for at least two years to de- Family Wise Error correction). There were no significant termine clinical and functional outcomes, including transition associations between baseline GMV and either subsequent to psychosis (according to the Comprehensive Assessment of remission or transition to psychosis. an At Risk Mental State, CAARMS), level of functioning (according to the Global Assessment of Functioning), and symptomatic remission (according to the CAARMS). GMV Conclusions: In CHR individuals, lower striatal GMV was measured in selected cortical and subcortical regions was associated with a poor level of overall functioning of interest (ROI) based on previous studies (ie orbitofrontal at follow-up. This finding was not related to effects of © The Author(s) 2022. Published by Oxford University Press on behalf of the University of Maryland's school of medicine, Maryland Psychiatric Research Center. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial License (https://creativecommons.org/ licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com Page 1 of 11 S. Tognin et al antipsychotic or antidepressant medication. The failure over time.18 However, McIntosh and colleagues19 found to replicate previous associations between GMV and later no differences in brain structure (either at baseline or psychosis onset, despite studying a relatively large sample, over time) between those CHR individuals who were still is consistent with the findings of recent large-scale multi- experiencing symptoms at follow-up and those who were center studies. not. This latter study involved relatives who were at ge- netic risk for developing schizophrenia, and it is possible Key words: clinical high risk for psychosis/clinical out that GMV differences observed in other studies might comes/functioning/remission/transition to psychosis/grey have been driven by different underlying psychopatho- matter volume/structural magnetic resonance imaging logical mechanisms. In order to clarify the relationship between GMV in CHR individuals and adverse outcomes, we used MRI Introduction to assess a large sample of people at CHR, who were A key research goal in studies of people at Clinical High followed up clinically for at least two years to determine Risk (CHR) for psychosis is to identify measures that outcomes defined in terms of transition to psychosis, could help to predict clinical and functional outcomes in level of functioning, and remission from the CHR state. this population. Grey matter volume (GMV) at presen- Participants were assessed at nine international sites tation, as assessed using Magnetic Resonance Imaging using the same acquisition sequence (Alzheimer’s Disease (MRI), was the first biological metric to be assessed in Neuroimaging Initiative [ADNI] sequence).20,21 We this population, and over the last 20 years, several studies adopted a Region of Interest approach (ROIs) to inves- have investigated the relationship between baseline GMV tigate whether GMV of CHR individuals was associated and clinical outcomes in CHR samples.1–7 Early studies in- with transition to psychosis (transition vs nontransition, volved relatively small samples and focused on later tran- CHR-T vs CHR-NT), level of functioning (good vs sition to psychosis as the sole outcome measure.7–9 More poor functioning, CHR-PF vs CHR-GF), and sympto- recently, multi-center studies included larger samples and matic remission (remission vs nonremission, CHR-R vs expanded outcomes to include remission from the CHR CHR-NR). Based on findings from the studies discussed state and level of functioning at follow up.10–13 Despite above,1,13,15-18 we hypothesized that transition to psychosis, this body of work, findings have been inconsistent, and poor functioning, and nonremission would be associated the extent to which MRI measurements of GMV differ to decreased GMV within our selected ROIs. Based on between CHR individuals with distinct clinical outcomes mostly inconsistent findings for these outcomes, in each is still unclear. case, we tested the null hypothesis that GMV at baseline Although a number of early studies reported that would not be related to each outcome at follow up. Most reductions in regional GMV at baseline were associated of the selected ROIs were based on prior studies on our with the subsequent onset of psychosis, the topographical outcomes of interest, however, as (1) imaging correlates location of the findings varied. For example, GMV studies of functional outcome and symptom remission are rela- comparing CHR who transitioned to those who did not, tively new areas of research and (2) the current sample size reported lower volume of the left parahippocampal did not allow for a full agnostic approach, we decided to gyrus,1 the left anterior cingulate cortex (ACC),6 and also explore brain regions in which volumetric differences the bilateral insula,4 while other studies did not iden- have been observed between CHR and healthy controls tify significant associations,2,3 and some found increases (HC). The final list of ROIs comprised the following: or- in hippocampal volume.5 Similarly, among more recent bitofrontal gyrus, cingulate gyrus, gyrus rectus, inferior larger scale studies, while some have described a relation- temporal gyrus, parahippocampal gyrus, striatum, and ship between lower regional GMV at presentation and hippocampus.1,15,16,18,22,23 For completeness, we are also later transition,1 others have not.14 reporting GMV differences between the CHR-T and Relatively few MRI studies have examined associations CHR-NT, CHR-PF and CHR-GF, and CHR-NR and between baseline GMV and other clinically meaningful CHR-R subgroups at the whole brain level and GMV outcomes in CHR cohorts, such as level of functioning differences between the CHR and HC subgroups within at follow up or remission from the CHR state. A poor the selected ROIs. functional outcome has been linked to reduced frontal, limbic, and cerebellar,15 frontal, cingulate, and tem- Methods poral,13 and striatal GMV at baseline.16 Nonremission from the CHR state has been associated with reduced Sample GMV in frontal, cingulate and temporal cortex,13,15 and From July 2010 to August 2015, 234 CHR individuals in orbitofrontal cortex and the pallidum.17 A longitudinal and 92 HC aged 18–35 years were scanned at nine centers study found that nonremission was related to a greater using 3-Tesla MRI scanners and the same ADNI-2 decline in CA1 volume, a hippocampal subfield, which T1 protocol20,21 (Amsterdam, the Netherlands; Basel, was further associated with increasing symptom severity Switzerland; Cologne, Germany; Copenhagen, Denmark; Page 2 of 11 Gray Matter Volume and Adverse Outcomes in Psychosis Risk London, United Kingdom; Melbourne, Australia; Paris, the timeframe considered was the time since the last France; The Hague, the Netherlands; Vienna, Austria) as follow-up. Data on baseline demographic characteris- part of the EU-GEI High Risk Study20 (see Table S1). The tics such as age, sex, and ethnicity were obtained from London site included an additional 31 CHR individuals the Medical Research Council Sociodemographic recruited as part of another study and scanned with Schedule.28 The presence of psychosis and other Axis the same scanner and sequence. Participants recruited I comorbidities was assessed using the SCID-I.29 An es- in Amsterdam and The Hague were scanned at the timate of the IQ was measured using the shortened ver- same center in Amsterdam, thus they were merged and sion of the Wechsler Adult Intelligence Scale.30,31 considered as one site. The final sample comprised 265 Participants were invited for face-to-face follow-up individuals at CHR. To reduce variability introduced by assessments at 6 months, 1 year, and 2 years after small samples, we only included sites that contributed baseline. Where face-to-face meetings were not pos- neuroimaging data for at least five CHR participants. sible, participants were followed up for 2 years using Scanner information, details of the image acquisition se- available clinical records, and this follow-up was ex- quence at each site, and information on included CHR tended when additional clinical data was available. individuals at each site for each analysis are reported in Transition to psychosis was determined using the Supplemental Tables S1 and S2. CAARMS25 and the SCID-I,29 with the clinical di- Exclusion criteria for all participants were past or agnosis of a psychotic disorder confirmed by the present diagnosis of neurological disorders according participant’s clinical team. Level of functioning was to DSM-IV criteria,24 estimated IQ lower than 60, or assessed using the GAF scale.27 Remission from the any MRI contraindication such as presence of metal CHR state was defined as when an individual no implants, pacemakers, or pregnancy. In addition, CHR longer met CAARMS inclusion criteria at the last fol- were excluded if they had a past or present diagnosis low-up assessment available. of psychotic disorder. HC were excluded if they were meeting CHR criteria, or if they had personal or familiar history of a psychiatric disorder according to DSM-IV Data Analysis criteria.24 Demographic and Clinical Data. Demographic and clinical Ethical approval for the study was obtained sepa- data were analyzed using SPSS, version 25.0 (IBM Corp.). rately at each individual site and informed written con- Group differences were investigated using t-test for contin- sent was obtained from all the participants at the single uous variables and χ 2 test for categorical variables. centers. The study was conducted in accordance with the Declaration of Helsinki. MRI Preprocessing. Voxel Based Morphometry (VBM) as implemented in the Statistical Parametric Mapping software (SPM12; http://www.fil.ion.ucl.ac.uk/spm) Clinical Assessment was employed to investigate GMV differences between CHR individuals were defined and assessed using the groups. First, all Sstructural images were visually checked Comprehensive Assessment of At Risk Mental State for artifacts, and poor-quality images were excluded. (CAARMS25). CHR individuals were included if they Assessment of image quality resulted in the exclusion of were meeting at least one of the following criteria: (1) three participants (one with large congenital cyst, one with Attenuated Psychotic Symptoms (APS), subthreshold distortion from brace, one with movement artefact). Second, frequency, and/or subthreshold intensity; (2) Brief T1 images were preprocessed with a unified segmentation Limited Intermittent Psychotic Symptoms (ie positive procedure32 implemented in SPM12. All images were seg- psychotic symptoms at psychotic intensity and frequency mented into grey matter (GM), white matter (WM), and which last for less than 7 days and resolved spontane- cerebrospinal fluid (CSF) partitions. Subsequently, a fast ously); (3) positive family history for psychosis or schiz- diffeomorphic image registration algorithm (DARTEL) otypal personality disorder plus chronic low functioning was used to warp the GM partitions into a new study- (<50) or a marked decline (ie 30% drop) in psychoso- specific reference space with an isotropic spatial resolu- cial functioning over the past 12 months (Genetic Risk tion of 1.5 mm3.33–35 The warped grey partitions were then and Deterioration Syndrome: GRD) as assessed with affine transformed into the Montreal Neuroanatomical the SOFAS.26 At baseline, all participants were assessed Imaging (MNI) space. An additional “modulation” step36 by trained raters using the CAARMS,25 the Global was used to scale the grey matter probability values by Assessment of Functioning (GAF27), and the SOFAS.26 the Jacobian determinants of the deformations to ensure Raters were trained in the use of the CAARMS and that the total amount of grey matter in each voxel was GAF prior to the study and completed online training conserved after the registration. As a final step, the grey videos every 12 months from study onset to assess matter probability values were smoothed using an 8 mm interrater reliability (see Supplemental Methods, Table full-width at half-maximum (FWHM) Gaussian kernel. S3). When assessing CAARMS and GAF at follow-up, Proportional scaling of the total intracranial volume was Page 3 of 11 S. Tognin et al used to identify regionally specific changes that were not For completeness, we reported GMV differences be- confounded by global differences. tween the CHR-T and CHR-NT, CHR-PF and CHR-GF, and CHR-NR and CHR-R subgroups at the whole brain ROI Analysis level, in the Supplementary Materials (3. Supplemental results). GMV differences between CHR and HC within An ROI approach was used to investigate group the selected ROIs are also reported in the Supplementary differences in a priori brain areas in which volumetric Materials (3. Supplemental results). In all analyses, scan- abnormalities have been identified in previous studies ning site, age, and sex were modeled as covariates of no on remission, functional outcomes, and comparisons interest. of CHR individuals and HC. These comprised sets of In order to exclude that significant results were con- bilateral cortical and subcortical regions which were founded by the use of antipsychotic or antidepressant examined in two anatomical masks. The first mask medication or substance use (ie cannabis) we run two ad- included the following cortical regions: orbitofrontal ditional analyses. Firstly, as the number of participants gyrus, cingulate gyrus, gyrus rectus, inferior tem- taking antipsychotic medication was relatively low poral, and parahippocampal cortex. 1,22,23 The second (N = 28), we rerun all the analyses removing these included the following subcortical regions: the hip- participants. Secondly, as the number of participants pocampus and the striatum (ie caudate, pallidum, taking antidepressants or using cannabis was relatively putamen). 15,16,18 The masks were created using the high (N = 81 and N = 90), we modeled these as covariates WFU Pickatlas toolbox implemented in SPM12. As of no interest. Results are reported in Supplementary these regions of interest covered a large portion of Materials. the brain, which would require a very conservative correction for multiple comparisons, we decided to use each mask individually, and then as a merged Results single mask. Demographic and Clinical Data A small volume correction for ROI analyses was applied. Within the masks, statistical inferences were made using Of the 262 CHR individuals studied at baseline (after a threshold of P < .05 after Family-Wise Error (FWE) excluding 3 participants due to poor quality of images), correction for multiple comparisons. In all analyses, scan- 234 (89.3%) were antipsychotic medication-naïve, 28 ning site, age, and sex were modeled as covariates of no (10.7%) were taking antipsychotic medication and 81 interest. Ethnicity was added as a covariate of no interest (30.9%) were taking antidepressants. During the follow in the comparison between HC and CHR as there was up period, 51 participants (19%) transitioned to psy- a significant difference between groups. Results at P < chosis (CHR-T), and 211 (81%) did not (CHR-NT). .001 uncorrected are not reported in the main text but are The mean time from baseline to transition in those available in the supplemental material. individuals who transitioned to psychosis was 360.1 (SD = 409.9 days). Of the 164 CHR individuals for whom GAF data was available at the latest possible follow-up Clinical and Functional Outcomes at Follow-up (mean time: 11.92 months [SD = 6.66 months], 63 had a In order to examine MRI correlates of transition to psy- good functional outcome (38%, CHR-GF, mean score at chosis, functional outcome, and clinical remission we follow-up: 73.23 [SD = 7.44], mean time: 13.50 months performed three full factorial analyses of co-variance [SD = 7.33 months]) and 101 had a poor outcome (62%, (ANCOVA). In the first ANCOVA, CHR individuals CHR-PF, mean score at follow-up: 52.02 [SD = 9.69], mean who had transitioned to psychosis during follow-up time: 10.96 months [SD = 6.06 months]). Information on (CHR-T) were compared to CHR individuals who did remission was available for 133 individuals, of which 57 not transition to psychosis (CHR-NT). In the second were in remission (43%, CHR-R) at follow up and 76 still ANCOVA we compared CHR individuals with a rela- met CHR criteria (57%, CHR-NR). The characteristics of tively good or poor level of functioning at the last avail- these subgroups are shown in table 1. There were no sig- able follow-up assessment for that participant. Following nificant differences between the CHR-PF and CHR-GF, previous studies in CHR samples,37–40 a GAF score of 65 CHR-NR and CHR-R, and the CHR-T and CHR-NT or more at follow-up was used as the threshold for a good subgroups or between CHR individuals and HC in age, functional outcome (CHR-GF), with a score below 65 sex, or handedness. However, the CHR-PF subgroup defined as poor (CHR-PF). A third ANCOVA compared had a significantly lower baseline GAF score than the CHR individuals who still met CAARMS criteria at the CHR-GF subgroup (P = .01). In addition, compared to last available follow-up (nonremission, CHR-NR) with HC, CHR individuals completed significantly fewer years CHR individuals who did not (remission, CHR-R). of education (P < .001), and differed in terms of eth- ROI analyses were then applied to the results of these nicity (P < .001; see table 2). Finally, significantly more ANCOVAs. CHR-NT individuals (n = 15) were taking antipsychotic Page 4 of 11 Table 1. Baseline Demographic and Clinical Characteristics of Participants CHR-NT CHR-T CHR-GF CHR-PF CHR-R CHR-NR P (n = 211) (n = 51) P Value (n = 63) (n = 101) P Value (n = 57) (n = 76) Value Sex, N Male 109 29 .50 29 55 .29 30 37 .65 Female 102 22 34 46 27 39 Age in years, mean (SD) 22.7 (4.8) 22.3 (4.0) .60 22.4 (4.1) 23.2 (5.1) .31 23.0 (4.1) 23.4 (5.5) .64 Education in years, mean (SD) 14.6 (2.9) 14.0 (3.0) .30 15.1 (2.7) 14.7 (3.2) .44 15.5 (2.7) 14.8 (3.0) .18 Ethnicity Caucasian 154 37 .73 50 70 .55 38 54 .34 Black 20 7 5 15 10 7 Asian 3 1 1 2 0 2 Others 34 6 8 14 9 13 Handedness Right 127 32 .42 42 68 .06 42 54 .52 Left 20 3 9 5 6 11 Baseline GAF score, mean (SD) 54.7 (9.9) 52.7 (9.1) .31 56.6 (10.3) 52.8 (9.1) .01 55.2 (9.1) 54.8 (9.5) .81 Antipsychotic medication, N Yes 15 13 <.001 3 11 .17 5 5 .64 No 196 38 60 90 52 71 Antidepressant medication, N Yes 68 13 .35 22 30 .49 23 22 .17 No 143 38 41 71 34 54 Cannabis, N 74 16 .89 22 32 .88 23 14 .06 Yes No 128 29 42 74 35 39 Days to transition, mean (SD) — 360.1 (409.9) NA — — — — — — Note: CHR, clinical high risk; CHR-GF, clinical high risk with good overall functioning (GAF ≥ 65); CHR-NR, clinical high risk non-remission; CHR-NT, clinical high risk non-transition; CHR-PF, clinical high risk with poor overall functioning (GAF < 65); CHR-R, clinical high risk remission; CHR-T, clinical high risk transition; GAF, Global Assessment of Functioning Scale (range: 0–100), the lower the score the more impaired overall functioning; NA, not available; SD, standard deviation. Significant differences are highlighted in bold. Page 5 of 11 Gray Matter Volume and Adverse Outcomes in Psychosis Risk S. Tognin et al medication at baseline compared to CHR-T individuals Discussion (n = 13, P < .001). The main finding from the present study was that a smaller right striatal GMV at baseline was associated with MRI Results a relatively poor functional outcome in CHR individuals. GMV and Transition to Psychosis. The ROI analyses However, we did not find evidence that alterations in found no significant differences between the CHR-T GMV at baseline were linked to either later onset of psy- (n = 51) and CHR-NT (n = 211) subgroups (P < .05, chosis or to remission from the CHR state. FWE-corrected threshold). Results did not change after An association between altered striatal volume and removing CHR participants on antipsychotic medi- functional outcome in people at high risk for psychosis cation or adjusting for antidepressant medication use or cannabis use. Differences evident at a threshold of Table 2. Baseline Demographic and Clinical Characteristics of the Whole CHR Sample and Healthy Controls P < .001 uncorrected are reported in the supplemen- tary results. No differences were found at the whole Healthy Control CHR Group P brain level. Group (n = 92) (n = 262) Value GMV and Functional Outcome. Subcortical mask: compared to the CHR-GF subgroup (n = 63), the Sex, N CHR-PF subgroup (n = 101) had less GMV in the right Male 47 138 .0.83 Female 45 124 striatum (Montreal Neurological Institute [MNI] coor- Age in years, 23.7 (4.2) 22.6 (4.8) .09 dinates x = 6, y = 20, z = −3, z score = 3.79, k = 186, mean (SD) P = .047 FWE-corrected threshold; see figure 1). After Education in 16.0 (2.9) 14.6 (2.9) <.001 excluding individuals that were taking antipsychotic years, mean (SD) medication (CHR-GF = 3; CHR-PF = 11), the signifi- Ethnicity Caucasian 65 191 <.001 cant effect in the right striatum remained unchanged Black 10 27 (MNI coordinates x = 6, y = 20, z = −3, z score = 3.84, Asian 11 4 k = 246, P = .040 FWE-corrected). Antidepressant med- Others 6 40 ication and cannabis use did not influence the finding in Antipsychotic medication, N the right striatum. Yes 0 28 NA No 92 234 Cortical and merged masks: no significant GMV Antidepressant medication, N differences were found (P < .05, FWE-corrected Yes 0 81 NA threshold). No differences were found at the whole No 92 181 brain level. Cannabis use, N GMV and Remission From the CHR State. The ROI Yes 25 90 .96 No 43 157 analyses showed no significant differences between the Handedness CHR-NR (n = 76) and CHR-R subgroups (n = 57) (P Right 66 159 .28 < .05, FWE-corrected threshold). Findings remained Left 6 23 unchanged after removing CHR individuals on antipsychotics or adjusting for antidepressant or can- Note: CHR, clinical high risk; NA, not available; SD, standard nabis use. No differences were found at the whole brain deviation. level. Significant differences are highlighted in bold. Fig. 1. Differences between clinical high risk individuals with poor overall functioning (CHR-PF) and those with good overall functioning (CHR-GF). The CHR-PF group had less grey matter volume than did CHR-GF in the right striatum (P < .05 FWE-corrected). Effects are displayed at P < .001 uncorrected for illustration purposes. T-values are indicated by color bars (for color figure refer to online version). Page 6 of 11 Gray Matter Volume and Adverse Outcomes in Psychosis Risk is consistent with data from previous studies. De Wit differences in participant recruitment and hence clinical and colleagues16 found that lower baseline subcortical characteristics of an already very heterogenous pop- volumes predicted long-term level of functioning in ulation might have led to different results.48 Only a few CHR individuals. Eggins and colleagues41 reported that MRI studies have investigated a relation between GMV young people presenting with depressive, manic, and psy- and nonremission from the CHR state, showing reduced chotic symptoms had reduced caudate volume relative volumes in frontal, cingulate, temporal cortex,13 in the or- to healthy controls, and that this was correlated with im- bitofrontal cortex, pallidum,17 and CA1,18 while another paired functioning. A recent multi-center MRI study in study, consistent with our present findings, found no sig- 116 CHR individuals also found an association between nificant association.19 GMV at baseline and functional outcome.10 However, the In the present study, because the total number of relationship found in this study, was with cortical rather participants was relatively large, and the numbers in the than the striatal volume, and was specific to the level of respective outcome subgroups were medium size (ranging social, as opposed to overall functioning. Finally, studies 37–192), the absence of significant associations with tran- using other neuroimaging modalities in CHR individuals sition and remission seems unlikely to simply be due to have previously linked a poor functional outcome to a lack of power. Nevertheless, although our participants increased prefrontal and striatal activation, to lower all met the same inclusion and exclusion criteria, the thalamic glutamate levels,40 and to altered interactions CHR state is an intrinsically heterogeneous category.49 between hippocampal resting activity and striatal dopa- While all participants were assessed at presentation, mine synthesis capacity.38 individuals may contact clinical services at different Level of functioning is increasingly used as an out- stages of the CHR state. In an effort to reduce and con- come measure in studies of CHR individuals, as it is trol the heterogeneity within the CHR state, some recent clinically meaningful and complements other outcomes studies modified the standard inclusion criteria, such that such as transition to psychosis and remission, which pri- samples are enriched for individuals with a particularly marily depend on measurements of symptom severity.42,43 high risk of transition.48,50 Because some of the sites in In particular, some individuals who, from a symptomatic the present study contributed a relatively small number perspective, may be categorized as having a relatively of individuals, we could not model site as factor, but good outcome can still have a very low level of overall controlled for site variability by increasing the number functioning, which translates into a high level of clinical of covariates of no interest. The latter approach might need.12 However, because it has only been employed as an have reduced the chance of detecting small yet significant outcome measure relatively recently, its biological basis in effects. CHR individuals is still poorly understood. A number of previous MRI studies have reported associations between later transition to psychosis and Strengths and Limitations altered GMV in cortical brain regions, thalamus, hip- A methodological strength of the present study is that pocampus, and basal ganglia,1,4–7,44–46 while other studies all sites acquired MRI data using the same ADNI-2 pro- have not found a significant relationship.2,3,37,47 Early MRI tocol,21 which was specifically designed for multi-center studies in CHR groups involved relatively small samples, studies. Nevertheless, variability can still arise through and the number of individuals in the subgroup that de- the use of scanners that differ in manufacturer, model, veloped psychosis was even smaller. More recent studies and age.51,52 This can be assessed by scanning a group of have been able to examine larger samples by recruiting the same “travelling” volunteers at each site,52,53 with post- from multiple sites. However, the findings have again been hoc calibration methods used to attenuate these effects.48 inconsistent. One previous multi-center study utilizing a However, this was not possible in the present study. relatively large sample was conducted by Mechelli and A further strength is that we minimized the potentially colleagues.1 The study assessed n = 182 CHR individuals confounding impact of site differences by restricting the and reported an association between later transition and analysis to those sites that had contributed imaging data reduced parahippocampal volume. Participants were from at least five CHR individuals. Although this led to scanned at five sites, with each using a different protocol, the exclusion of one site in the functional (5 individuals and all but one acquiring data at 1.5T which might have excluded) and one site in the remission analyses (2 led to different findings compared to those reported individuals excluded) (see Supplementary Materials), here. However, more recently, Cannon and colleagues,14 this approach is likely to have increased the power to de- in a study of n = 274 individuals, found no relation- tect true statistical effects, and by including a moderate ship between MRI measures at baseline and later tran- number of covariates have reduced the risk of Type II sition. These data were collected at eight sites, all using errors.54–56 the same ADNI protocol at 3T.21 It is possible that dif- During the follow up period, as is typical for CHR ferent acquisitions sequences and scanner manufacturers individuals, participants received needs-based clin- might have led to different results.48 It is also possible that ical care involving case management, but in most cases Page 7 of 11 S. Tognin et al (89.3%) not any active treatment (such as antipsychotic Funding medication). While its content may have varied between The European Network of National Schizophrenia sites and individuals, a recent network meta-analysis Networks Studying Gene-Environment Interactions suggests that there is no evidence that this form of clinical (EU-GEI) Project is funded by grant agreement contact significantly alters long term clinical outcomes.57 HEALTH-F2-2010-241909 (Project EU-GEI) from the This study has also a number of limitations. Firstly, European Community Seventh Framework Programme. GAF was used to measure functioning. While GAF is Additional support was provided by a Medical Research still widely used in research settings, it has been criticized Council Fellowship to M.K. (grant MR/J008915/1). for its lack of reliability58 and limited ability to meaning- G.M. is supported by a Sir Henry Dale Fellowship, jointly fully assess all areas of functioning (as opposed to eg the funded by the Wellcome Trust and the Royal Society WHODAS59). To improve reliability within and across (202397/Z/16/Z). C.P. was supported by a National EUGEI sites, regular interrater reliability exercises, in- Health and Medical Research Council (NHMRC) Senior cluding key measures such as the CAARMS and the Principal Research Fellowship (1105825), NHMRC GAF, were carried out before60 and throughout the du- L3 Investigator Grant (1196508). B.N. was supported ration of the study. The GAF reliability score in the by an NHMRC Senior Research Fellowship (1137687) final year of the study was 0.83, indicating good relia- and a University of Melbourne Dame Kate Campbell bility (see Table S3). Secondly, in the present study we Fellowship. The Melbourne site was supported by subdivided groups using GAF scores of >65 and <65 NHMRC-EU Grant (ID: 567215). N.B.V. was supported as thresholds following the approach used in previous by Ministerio de Ciencia, Innovación e Universidades studies in CHR samples.37,38,40,61 While we acknowledge (PSI2017-87512-C2-1-R). that GAF could be analyzed as a continuous variable, an advantage of using the cut-off is that it facilitates com- parison of our data with those from with other studies Acknowledgments that used the same approach. Thirdly, while longitudinal We would like to thank all participants who took part in analyses would be helpful to elucidate the trajectories of the study. our outcomes of interest, the high number of missing outcome data at different follow-up time points did not Conflict of Interest allow for these analyses to be carried out. Missing data, and consequently reduced sample sizes, are particularly None. problematic when conducting imaging analyses which require modest to large samples. Finally, while multi- Appendix center recruitment is necessary to ascertain large samples, this also increases the possible confounding effects of *EU-GEI High Risk Study Group Author site differences in MRI acquisition and clinical ratings. Philip McGuire1,3 However, we minimized the latter by using the same Lucia R. Valmaggia20 scanning protocol,21 and conducting regular interrater Maria Calem1 reliability exercises both before60 and throughout the du- Mathilde Antoniades1 ration of the study. Sara Pisani1 Gemma Modinos1 Lieuwe de Haan17 Conclusion Mark van der Gaag18,19 Our findings suggest that in people at CHR for psychosis, Eva Velthorst17 relatively reduced right striatal volume is associated with Tamar C. Kraan17 a poor functional outcome. The absence of associations Daniella S. van Dam17 with either symptomatic remission or transition to psy- Nadine Burger17 chosis, despite a relatively large sample size, raises the Barnaby Nelson7,8 possibility that other neuroimaging measures62–65 or pe- Patrick McGorry7,8 ripheral blood markers,66,67 might be better predictors of G Paul Amminger7,8 outcomes in CHR individuals, particularly when these Christos Pantelis6 are integrated in multi-modal models.68 We plan to inves- Athena Politis7 tigate this in further studies in the EU-GEI sample. Joanne Goodall7 Anita Riecher-Rössler9 Stefan Borgwardt9 Supplementary Data Erich Studerus9 Supplementary data are available at Schizophrenia Rodrigo Bressan23 Bulletin Open online. Ary Gadelha23 Page 8 of 11 Gray Matter Volume and Adverse Outcomes in Psychosis Risk Elisa Brietzke23 A cross-sectional high resolution region of interest magnetic Graccielle Asevedo23 resonance imaging study. J Psychiatr Res 2010;44(7):447–453. Elson Asevedo23 6. Rothlisberger M, Riecher-Rossler A, Aston J, Fusar-Poli P, Radu EW, Borgwardt S. Cingulate volume abnormalities in Andre Zugman23 emerging psychosis. Curr Pharm Des 2012;18(4):495–504. Neus Barrantes-Vidal10 7. Pantelis C, Velakoulis D, McGorry PD, et al. Neuroanatomical Tecelli Domínguez-Martínez10 abnormalities before and after onset of psychosis: a Anna Racciopi10 cross-sectional and longitudinal MRI comparison. Lancet Thomas R. Kwapil10 2003;361(9354):281–288. Manel Monsonet10 8. Borgwardt SJ, McGuire PK, Aston J, et al. Structural Lídia Hinojosa10 brain abnormalities in individuals with an at-risk mental state who later develop psychosis. Br J Psychiatry Suppl Mathilde Kazes11 2007;51:s69–s75. Claire Daban11 9. Borgwardt SJ, McGuire PK, Aston J, et al.. Reductions in Julie Bourgin11 frontal, temporal and parietal volume associated with the Olivier Gay11 onset of psychosis. Schizophr Res 2008;106(2-3):108–114. Célia Mam-Lam-Fook11 10. Koutsouleris N, Kambeitz-Ilankovic L, Ruhrmann S, et al. Marie-Odile Krebs11 Prediction models of functional outcomes for individuals in Dorte Nordholm12 the clinical high-risk state for psychosis or with recent-onset depression: a multimodal, multisite machine learning ana- Lasse Randers12 lysis. JAMA Psychiatry 2018;75(11):1156–1172. Kristine Krakauer12 11. Addington J, Cornblatt BA, Cadenhead KS, et al. At clin- Louise Glenthøj12 ical high risk for psychosis: outcome for nonconverters. Am J Birte Glenthøj12,13 Psychiatry 2011;168(8):800–805. Merete Nordentoft12 12. Rutigliano G, Valmaggia L, Landi P, et al. Persistence or Stephan Ruhrmann14 recurrence of non-psychotic comorbid mental disorders Dominika Gebhard14 associated with 6-year poor functional outcomes in pa- tients at ultra high risk for psychosis. J Affect Disord Julia Arnhold14 2016;203:101–110. Joachim Klosterkötter14 13. Cropley VL, Lin A, Nelson B, et al. Baseline grey matter Gabriele Sachs15 volume of non-transitioned “ultra high risk” for psychosis Iris Lasser15 individuals with and without attenuated psychotic symptoms Bernadette Winklbaur15 at long-term follow-up. Schizophr Res 2016;173(3):152–158. Harald Aschauer15 14. Cannon TD, Chung Y, He G, et al. Progressive reduction in Philippe A Delespaul16 cortical thickness as psychosis develops: a multisite longitu- dinal neuroimaging study of youth at elevated clinical risk. Bart P. Rutten16 Biol Psychiatry 2015;77(2):147–157. Jim van Os1,16 15. Reniers RL, Lin A, Yung AR, et al. Neuroanatomical pre- 23 LiNC - Lab Interdisciplinar Neurociências Clínicas, dictors of functional outcome in individuals at ultra-high risk Depto Psiquiatria, Escola Paulista de Medicina, for psychosis. Schizophr Bull 2017;43(2):449–458. Universidade Federal de São Paulo – UNIFESP. 16. de Wit S, Ziermans TB, Nieuwenhuis M, et al. Individual prediction of long-term outcome in adolescents at ultra-high risk for psychosis: applying machine learning techniques to References brain imaging data. Hum Brain Mapp 2017;38(2):704–714. 17. de Wit S, Wierenga LM, Oranje B, et al. Brain develop- 1. Mechelli A, Riecher-Rossler A, Meisenzahl EM, et al. ment in adolescents at ultra-high risk for psychosis: lon- Neuroanatomical abnormalities that predate the onset gitudinal changes related to resilience. Neuroimage Clin of psychosis: a multicenter study. Arch Gen Psychiatry 2016;12:542–549. 2011;68(5):489–495. 18. Ho NF, Holt DJ, Cheung M, et al. Progressive decline in 2. Yucel M, Wood SJ, Phillips LJ, et al. Morphology of hippocampal CA1 volume in individuals at ultra-high- the anterior cingulate cortex in young men at ultra-high risk for psychosis who do not remit: findings from the lon- risk of developing a psychotic illness. Br J Psychiatry gitudinal youth at risk study. Neuropsychopharmacology 2003;182:518–524. 2017;42(6):1361–1370. 3. Velakoulis D, Wood SJ, Wong MT, et al. Hippocampal 19. McIntosh AM, Owens DC, Moorhead WJ, et al. and amygdala volumes according to psychosis stage and Longitudinal volume reductions in people at high genetic risk diagnosis: a magnetic resonance imaging study of chronic of schizophrenia as they develop psychosis. Biol Psychiatry schizophrenia, first-episode psychosis, and ultra-high-risk in- 2011;69(10):953–958. dividuals. Arch Gen Psychiatry 2006;63(2):139–149. 20. van Os J, Rutten BP, et al.; European Network of National 4. Takahashi T, Wood SJ, Yung AR, et al. Insular cortex gray Networks studying Gene-Environment Interactions in matter changes in individuals at ultra-high-risk of developing Schizophrenia (EU-GEI). Identifying gene-environment psychosis. Schizophr Res 2009;111(1–3):94–102. interactions in schizophrenia: contemporary challenges 5. Buehlmann E, Berger GE, Aston J, et al.. Hippocampus for integrated, large-scale investigations. Schizophr Bull abnormalities in at risk mental states for psychosis? 2014;40(4):729–736. Page 9 of 11 S. Tognin et al 21. initiative. IAAMaAsdn, available at: http://adni.loni.usc.edu/ of young people with affective and psychotic disorders. methods/mri-tool/mri-acquisition/. Accessed May 31, 2022. Psychiatry Res Neuroimaging 2018;271:8–16. 22. Gong Q, Dazzan P, Scarpazza C, et al. A neuroanatomical 42. Carrion RE, McLaughlin D, Goldberg TE, et al. Prediction signature for schizophrenia across different ethnic groups. of functional outcome in individuals at clinical high risk for Schizophr Bull 2015;41(6):1266–1275. psychosis. JAMA Psychiatry 2013;70(11):1133–1142. 23. Nenadic I, Dietzek M, Schonfeld N, et al. Brain structure 43. Schmidt A, Cappucciati M, Radua J, et al. Improving prog- in people at ultra-high risk of psychosis, patients with first- nostic accuracy in subjects at clinical high risk for psychosis: episode schizophrenia, and healthy controls: a VBM study. systematic review of predictive models and meta-analytical se- Schizophr Res 2015;161(2–3):169–176. quential testing simulation. Schizophr Bull 2017;43(2):375–388. 24. Association. AAP. Diagnostic and Statistical Manual of 44. Koutsouleris N, Meisenzahl EM, Davatzikos C, et al. Use of Mental Disorders IV (DSM-IV). Washington DC: American neuroanatomical pattern classification to identify subjects in Psychiatric Press Inc; 1994. at-risk mental states of psychosis and predict disease transi- 25. Yung AR, Yuen HP, McGorry PD, et al. Mapping the onset of tion. Arch Gen Psychiatry 2009;66(7):700–712. psychosis: the Comprehensive Assessment of At-Risk Mental 45. Koutsouleris N, Borgwardt S, Meisenzahl EM, Bottlender R, States. Aust N Z J Psychiatry 2005;39(11–12):964–971. Moller HJ, Riecher-Rossler A. Disease prediction in the 26. Goldman HH, Skodol AE, Lave TR. Revising axis V for at-risk mental state for psychosis using neuroanatomical DSM-IV: a review of measures of social functioning. Am J biomarkers: results from the FePsy study. Schizophr Bull Psychiatry 1992;149(9):1148–1156. 2012;38(6):1234–1246. 27. Hall RC. Global assessment of functioning. A modified scale. 46. Mittal VA, Walker EF, Bearden CE, et al. Markers of Psychosomatics 1995;36(3):267–275. basal ganglia dysfunction and conversion to psychosis: neurocognitive deficits and dyskinesias in the prodromal 28. Mallett R. Sociodemographic Schedule. London, UK: Section period. Biol Psychiatry 2010;68(1):93–99. of Social Psychiatry, Institute of Psychiatry; 1997. 47. Hannan KL, Wood SJ, Yung AR, et al. Caudate nucleus 29. First M, Spitzer R, Gibbon M, Williams J. Structured volume in individuals at ultra-high risk of psychosis: a Clinical Interview for DSM-IV axis I disorders (SCID). New cross-sectional magnetic resonance imaging study. Psychiatry York, NY: New York State Psychiatric Institute Biometrics Res 2010;182(3):223–230. Research; 1995. 48. Tognin S, van Hell HH, Merritt K, et al. Towards precision 30. Blyler CR, Gold JM, Iannone VN, Buchanan RW. Short form medicine in psychosis: benefits and challenges of multimodal of the WAIS-III for use with patients with schizophrenia. multicenter studies-PSYSCAN: translating neuroimaging Schizophr Res 2000;46(2–3):209–215. findings from research into clinical practice. Schizophr Bull 31. Velthorst E, Levine SZ, Henquet C, et al. To cut a short test 2020;46(2):432–441. even shorter: reliability and validity of a brief assessment of 49. Fusar-Poli P, Bonoldi I, Yung AR, et al. Predicting psychosis: intellectual ability in schizophrenia--a control-case family meta-analysis of transition outcomes in individuals at high study. Cogn Neuropsychiatry 2013;18(6):574–593. clinical risk. Arch Gen Psychiatry 2012;69(3):220–229. 32. Ashburner J, Friston KJ. Unified segmentation. Neuroimage 50. Addington J, Liu L, Brummitt K, et al. North American 2005;26(3):839–851. Prodrome Longitudinal Study (NAPLS 3): methods and 33. Ashburner JA. fast diffeomorphic image registration algo- baseline description. Schizophr Res 2022;243:262–267. rithm. Neuroimage 2007;38(1):95–113. 51. Jovicich J, Czanner S, Han X, et al. MRI-derived measure- 34. Ashburner J, Friston K. Multimodal image coregistration ments of human subcortical, ventricular and intracranial and partitioning--a unified framework. Neuroimage brain volumes: reliability effects of scan sessions, acquisition 1997;6(3):209–217. sequences, data analyses, scanner upgrade, scanner vendors 35. Ashburner J, Friston KJ. Computing average shaped tissue and field strengths. Neuroimage 2009;46(1):177–192. probability templates. Neuroimage 2009;45(2):333–341. 52. Schnack HG, van Haren NE, Brouwer RM, et al. Mapping 36. Mechelli A, Price JC, Friston JK, John A. Voxel-based reliability in multicenter MRI: voxel-based morphometry and morphometry of the human brain: methods and applications. cortical thickness. Hum Brain Mapp 2010;31(12):1967–1982. Curr Med Imaging Rev 2005;1:105–113. 53. Cannon TD, Sun F, McEwen SJ, et al. Reliability of neuro- 37. Modinos G, Kempton MJ, Tognin S, et al. Association of ad- anatomical measurements in a multisite longitudinal verse outcomes with emotion processing and its neural sub- study of youth at risk for psychosis. Hum Brain Mapp strate in individuals at clinical high risk for psychosis. JAMA 2014;35(5):2424–2434. Psychiatry 2020;77(2):190–200. 54. Vierron E, Giraudeau B. Sample size calculation for 38. Modinos G, Richter A, Egerton A, et al. Interactions be- multicenter randomized trial: taking the center effect into tween hippocampal activity and striatal dopamine in people account. Contemp Clin Trials 2007;28(4):451–458. at clinical high risk for psychosis: relationship to adverse out- 55. Suckling J, Barnes A, Job D, et al. The neuro/PsyGRID comes. Neuropsychopharmacology 2021;46(8):1468–1474. calibration experiment: identifying sources of variance 39. Bossong MG, Antoniades M, Azis M, et al. Association of and bias in multicenter MRI studies. Hum Brain Mapp hippocampal glutamate levels with adverse outcomes in indi- 2012;33(2):373–386. viduals at clinical high risk for psychosis. JAMA Psychiatry 56. Hyatt CS, Owens MM, Crowe ML, Carter NT, Lynam DR, 2019;76(2):199–207. Miller JD. The quandary of covarying: a brief review 40. Allen P, Chaddock CA, Egerton A, et al. Functional outcome and empirical examination of covariate use in structural in people at high risk for psychosis predicted by thalamic glu- neuroimaging studies on psychological variables. Neuroimage tamate levels and prefronto-striatal activation. Schizophr Bull 2020;205:116225. 2015;41(2):429–439. 57. Davies C, Cipriani A, Ioannidis JPA, et al. Lack of evidence 41. Eggins PS, Hatton SN, Hermens DF, Hickie IB, Lagopoulos J. to favor specific preventive interventions in psychosis: a net- Subcortical volumetric differences between clinical stages work meta-analysis. World Psychiatry 2018;17(2):196–209. Page 10 of 11 Gray Matter Volume and Adverse Outcomes in Psychosis Risk 58. Vatnaland T, Vatnaland J, Friis S, Opjordsmoen S. Are GAF 64. Morgan SE, Young J, Patel AX, et al. Functional magnetic scores reliable in routine clinical use? Acta Psychiatr Scand resonance imaging connectivity accurately distinguishes 2007;115(4):326–330. cases with psychotic disorders from healthy controls, based 59. Federici S, Bracalenti M, Meloni F, Luciano JV. World on cortical features associated with brain network devel- Health Organization disability assessment schedule opment. Biol Psychiatry Cogn Neurosci Neuroimaging 2.0: an international systematic review. Disabil Rehabil 2021;6(12):1125–1134. 2017;39(23):2347–2380. 65. Guo S, Huang CC, Zhao W, et al. Combining multi-modality 60. Berendsen S, Kapitein P, Schirmbeck F, et al. Pre-training data for searching biomarkers in schizophrenia. PLoS One inter-rater reliability of clinical instruments in an international 2018;13(2):e0191202. psychosis research project. Schizophr Res 2021;230:104–107. 66. Dickens AM, Sen P, Kempton MJ, et al. Dysregulated lipid 61. Bossong MG, Antoniades M, Azis M, et al. Association of metabolism precedes onset of psychosis. Biol Psychiatry hippocampal glutamate levels with adverse outcomes in indi- 2021;89(3):288–297. viduals at clinical high risk for psychosis. JAMA Psychiatry 67. Pollak TA, Kempton MJ, Iyegbe C, et al. Clinical, cognitive 2019;76(2):199–207. and neuroanatomical associations of serum NMDAR auto- 62. Kempton MJ, McGuire P. How can neuroimaging facilitate antibodies in people at clinical high risk for psychosis. Mol the diagnosis and stratification of patients with psychosis? Psychiatry 2021;26(6):2590–2604. Eur Neuropsychopharmacol 2015;25(5):725–732. 68. Koutsouleris N, Dwyer DB, Degenhardt F, et al. Multimodal 63. Vieira S, Gong QY, Pinaya WHL, et al. Using machine machine learning workflows for prediction of psych- learning and structural neuroimaging to detect first epi- osis in patients with clinical high-risk syndromes and sode psychosis: reconsidering the evidence. Schizophr Bull recent-onset depression. JAMA Psychiatry 2021;78(2): 2020;46(1):17–26. 195–209. Page 11 of 11