Lack of Evidence for Regional Brain Volume or Cortical Thickness Abnormalities in Youths at Clinical High Risk for Psychosis: Findings From the Longitudinal Youth at Risk Study

Abstract

There is cumulative evidence that young people in an “at-risk mental state” (ARMS) for psychosis show structural brain abnormalities in frontolimbic areas, comparable to, but less extensive than those reported in established schizophrenia. However, most available data come from ARMS samples from Australia, Europe, and North America while large studies from other populations are missing. We conducted a structural brain magnetic resonance imaging study from a relatively large sample of 69 ARMS individuals and 32 matched healthy controls (HC) recruited from Singapore as part of the Longitudinal Youth At-Risk Study (LYRIKS). We used 2 complementary approaches: a voxel-based morphometry and a surface-based morphometry analysis to extract regional gray and white matter volumes (GMV and WMV) and cortical thickness (CT). At the whole-brain level, we did not find any statistically significant difference between ARMS and HC groups concerning total GMV and WMV or regional GMV, WMV, and CT. The additional comparison of 2 regions of interest, hippocampal, and ventricular volumes, did not return any significant difference either. Several characteristics of the LYRIKS sample like Asian origins or the absence of current illicit drug use could explain, alone or in conjunction, the negative findings and suggest that there may be no dramatic volumetric or CT abnormalities in ARMS.

Introduction

Adolescents and young adults in the putative prodrome of psychotic illness—variously labeled as being at “ultra high risk” (UHR), “clinical high risk” (CHR), or in an “at-risk mental state” (ARMS)—experience distressing subthreshold psychotic symptoms and have a 30–43% risk of transition to psychosis over a 36-month period.¹ These individuals are typically identified through clinical assessment of help-seeking individuals who present (1) attenuated or (2) brief and intermittent psychotic symptoms, or (3) a decrease in global functioning combined with a genetic risk for psychosis.^2,3

Structural magnetic resonance imaging (MRI) brain studies have featured prominently in attempts to identify biomarkers of ARMS. In general, this work has shown baseline grey matter volume (GMV) reductions in frontal, temporal, and limbic areas of ARMS individuals.^4–10 Although the results of ARMS MRI research, typically obtained in small samples, are heterogeneous and contradictory,^11,12 many of the identified brain changes are similar to those seen in patients with established schizophrenia.^13,14 Some GMV reductions, particularly in frontolimbic areas, have been confirmed to be statistically robust through meta-analysis¹⁵ and multicentre investigations.¹⁶

In parallel to GMV findings, only 4 whole-brain studies compared cortical thickness (CT) between ARMS individuals and controls and their results were divergent. One study reported cortical thinning in several brain regions, including frontal, temporal, and limbic areas¹⁷ while 3 studies did not report any cortical thinning significant at the whole-brain level in a larger sample of ARMS individuals when compared at baseline with healthy controls (HC).^18–20

Fewer studies have investigated alterations of white matter volume (WMV) in ARMS but their findings are consistent with what has been reported for GMV. They reported smaller WMV in fronto-temporo-limbic areas^5,6,21 as well as a global reduction of WM growth over time²² in ARMS compared to HC.

While baseline comparisons between ARMS and HC are useful for identifying putative biomarkers of young people in need of care, the majority of ARMS individuals do not transition to frank psychosis (ARMS-NT), spurring attempts to identify ARMS individuals at incipient risk of psychosis onset (ARMS-T). At the whole-brain level, gray matter differences associated with transition to psychosis have been localized in the same fronto-temporo-limbic regions that also distinguish the overall ARMS group (regardless of transition) from HC.^4,6,23,24 More precisely, baseline GMV reductions in ARMS-T when compared with ARMS-NT were especially consistent in the frontoinsular and superior temporal regions.¹⁵

All these studies recruited ARMS samples from North America, Europe, and Australia. There are few structural brain MRI studies performed in ARMS samples from Asia and all were conducted in small cohorts.^17,25,26 Nevertheless, establishing consistency across different ethnic groups represents a critical step in the development of any putative biomarkers.

An additional advantage of such research in Asian countries is the very low prevalence of cannabis and other drug use.²⁷ Substance use is more frequent in patients with psychotic disorders in Western countries²⁸ and could be a problematic confound for ARMS research in Western populations.^29,30 Substance use, and cannabis in particular, have been associated with structural changes in at-risk populations.^31–34

We used both voxel-based morphometry (VBM) and surface-based morphometry (SBM) analyses to run a comprehensive and not regionally biased whole-brain investigation of baseline GMV, WMV, and CT alterations in a relatively large sample of 69 ARMS individuals with minimum antipsychotics or substance use recruited from Singapore as part of the LYRIKS.³⁵ Given the good statistical power offered by our large sample size, we hypothesized that we should reproduce some of the GMV, WMV and CT alterations in the frontal and temporal lobes as reported by previous whole-brain studies.

Methods and Materials

Participants

Our sample comprised 75 ARMS subjects and 40 HC between 14 and 29 years old, matched for age, gender, handedness, and educational level. The participants were part of the LYRIKS, in Singapore. ARMS subjects were recruited from programs targeted at identifying individuals at-risk for developing psychosis run by the Institute of Mental Health, and from various community mental health agencies. Details of the recruitment strategy were previously reported.³⁶ In brief, we adopted an active approach of recruiting individuals from various psychiatric clinics and community mental health agencies, and a passive approach of self-referrals from print and social media advertisements. ARMS subjects met inclusion criteria for the prodromal state of schizophrenia in accordance with the comprehensive assessment of at-risk mental states (CAARMS).³ CAARMS assessments were performed by experienced psychometricians that were trained at ORYGEN in Melbourne. Interrater reliability was established and monthly supervisions were conducted throughout the study period to guarantee diagnostic validity. At-risk participants had no history of psychiatric, neurological or serious medical disorders, or mental retardation; and were not on antipsychotic medications. We excluded anyone with a current substance abuse as defined by the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV). About 6 AMRS subjects and 1 HC had a past history of substance use disorder (table 1). Six ARMS subjects and 8 HC were excluded from the original sample due to the use of a different T1-weighted structural MRI sequence (n = 10) or the presence of gross structural abnormalities or movement artifacts (n = 4). The demographics and clinical information of the remaining 69 ARMS and 32 HC are detailed in table 1. Out of 69 ARMS subjects, 33 had a concomitant diagnostic of depression and/or anxiety and 37 were medicated with antidepressants, mostly selective serotonin reuptake inhibitor (SSRI, n = 28), but also non-SSRI (n = 7) or both SSRI and non-SSRI in association (n = 2). During 28-month follow-up, 7 ARMS subjects converted to psychosis and 13 withdrew from the study, leaving a final sample of 56 ARMS-NT and 7 ARMS-T at baseline.

Table 1.

Demographic, Clinical, and Anatomical Characteristics of Participants

	ARMS Subjects (SD)	Healthy Controls (SD)	Difference (P value)
Count	69	32
Age	21.52 (3.49)	22.97 (3.94)	.07
Gender			.15
Male (%)	68	53
Female (%)	32	47
Handedness			.64
Right handed (%)	84	91
Left handed (%)	7	3
Ambidextrous (%)	9	6
Ethnicity			.13
Chinese (%)	67	56
Malay (%)	23	16
Indian (%)	6	19
Other %	4	9
Education
PSLE	196.3 (47.75)	206.1 (31.34)	.48
Baseline clinical scores
CAARMS positive	16.33 (7.35)	—
GRD (%)	30	—
APS (%)	81	—
BLIPS (%)	7	—
CDSS	5.42 (4.61)	—
BAI	20.74 (11.16)	—
Comorbidities
Depression and/or anxiety (%)	48	0
Past history SUD
Alcohol (%)	6	3	.56
Illicit drug (%)	3	0	.33
Brain volumes
VBM-ICV (ml)	1502.18 (141.05)	1448.24 (118.67)	.59
SBM-ICV (ml)	1465.61 (146.64)	1410.48 (152.81)	.31
SBM-total GM (ml)	685.71 (55.49)	663.55 (47.09)	.79
SBM-total WM (ml)	470.84 (52.12)	460.79 (47.80)	.38
Hippocampi (ml)	8.73 (0.76)	8.72 (0.61)	.09
Ventricles (ml)	14.91 (6.88)	12.60 (5.54)	.22

Notes: Percentages were rounded to the nearest integer. All ARMS and control subjects belong to the 3 major ethnicities in Singapore (Chinese, Malay, and Indian), except 2 ARMS (Javanese and Eurasian), and 2 controls (Javanese and Israeli). APS, attenuated psychotic symptoms; BAI, Beck anxiety inventory; BLIPS, brief limited intermittent psychotic symptoms; CAARMS, comprehensive assessment of at-risk mental states; CDSS, Calgary depression scale for schizophrenia; GRD, genetic risk and deterioration syndrome; GM, gray matter; ICV, intracranial volume; PSLE, primary school leaving examination; SBM, surface-based morphometry; SUD, substance-use disorder; VBM, voxel-based morphometry; WM, white matter.

Additional exclusion criteria for controls were: (1) history of severe head injury, (2) personal history of psychotic disorder, and (3) personal history of other neuropsychiatric disorder. Controls did not have any family history of neuropsychiatric disorders, except, 3 controls had a first-degree relative with a history of depression, 2 had a second-degree relative with history of schizophrenia (n = 1) or depression (n = 1). In both the ARMS and HC groups, Primary School Leaving Examination (PSLE) scores, which are the result of a standardized multidisciplinary test of scholastic achievement, were used as a measure of educational level. Written informed consent was provided by all participants aged 21 and above or from a legally acceptable representative for participants under 21 with participant’s assent. Ethics approval for this study was provided by the National Healthcare Group’s Domain Specific Review Board.

Image Acquisition

T1-weighted structural MRI data were obtained from a 3T Siemens Trio Tim scanner (Siemens, Erlangen, Germany) at the Center for Cognitive Neuroscience, Duke-NUS Graduate Medical School, Singapore. The principal sequence relevant to this study was a T1-weighted 3D magnetization-prepared rapid-acquisition gradient echo (MPRAGE) sequence optimized for gray-white matter contrast. It was identical to that used by the Alzheimer’s Disease Neuroimaging Initiative ADNI consortium.³⁷ Parameters were as follows: TR = 2300ms, TE = 2.98ms, TI = 900ms, flip angle = 9°, BW = 240 Hz/pixel, FOV = 256×240mm, matrix = 256×240; resulting voxel dimensions: 1.0×1.0×1.0mm, acquisition time 5min 03 s. Parallel imaging was used to improve the signal-to-noise ratio instead of shortening the scan time. We obtained a single high-quality image instead of averaging 2 or more rapidly acquired images. Images were inspected for motion artifact at the time of acquisition and scanning was repeated as necessary. Images were reviewed for any gross pathological findings.

Voxel-Based Morphometry

Every scan was visually checked to exclude the presence of artifacts or gross anatomical abnormalities that could impact image pre-processing. Voxel-wise analyses of brain GMV and WMV differences were conducted using the Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra (DARTEL) procedure³⁸ implemented in SPM8 (http://www.fil.ion.ucl.ac.uk/spm/software/spm8/) running under MATLAB 2009b (http://www.mathworks.com.au/products/matlab/). Briefly, each participant’s T1-weighted anatomical scan was segmented into distinct tissue compartments and spatially normalized via a nonlinear algorithm using a unified procedure.³⁸ A study-specific template was generated by normalizing each participant’s segmented gray or white matter image to a common space. Native-space gray or white matter images were then spatially normalized to this template. Jacobian modulation of voxel intensities was employed to preserve GMV or WMVs. The images were smoothed with an 8-mm full-width half maximum Gaussian kernel prior to statistical analysis.

The General linear model (GLM) was used to test for group differences in volume at each voxel, as implemented in Randomize (http://fsl.fmrib.ox.ac.uk/fsl/ fslwiki/Randomise). All results were corrected for multiple comparisons Type I error with a nonparametric cluster-size based procedure.^39,40 A voxel-wise threshold was initially set to 0.001 to compromise between sensitivity to spatially extended vs focal and intense differences. Then, a cluster-size threshold was calculated via permutation testing (10000 permutations). We compared baseline GMV and WMV between ARMS group and HC group, while covarying for age, gender, intracranial volume (ICV), handedness, and ethnicity.

Surface-Based Morphometry

The semiautomated CT measurements were performed using FreeSurfer v5.1.0 (http://surfer.nmr.mgh.harvard.edu/; Martinos Imaging Centre, Charlestown, MA), as described by Dale, Fishl and colleagues.^41,42

The white (ie, gray-white matter boundary) and pial (ie, gray-cerebrospinal fluid boundary) surfaces were visually inspected and edited, where necessary, using standard procedures (http://surfer.nmr.mgh.harvard.edu/fswiki/Edits), blind to diagnostic status. Surfaces for each participant were registered to a study-specific template and smoothed using a Gaussian kernel of 25mm prior to group analysis.

We used a GLM implemented in Freesurfer to estimate group differences in CT at each vertex of the cerebral surface while controlling for the effect of age, gender, handedness, and ethnicity. Right and left hemispheres were tested separately. False Discovery Rate (FDR) P < .05 was used for multiple comparisons correction.

Volume-of-Interest Measurements

We derived 5 volume-of-interest (VOI) measurements from the Freesurfer analysis: ICV, total GMV, total WMV, hippocampal volume, and ventricular volume. ICV was calculated using a validated method described elsewhere.⁴³ Total ventricular volume was defined as the total volume of lateral ventricles, third ventricle, fourth ventricle, and fifth ventricle.

Statistical analyses were performed with the Statistical Package for the Social Sciences, version 21 (SPSS 21.0, IBM Corp. Armonk, NY, USA). Differences in cerebral volumes were tested using one-way analysis of covariance (ANCOVA) with age, gender, handedness, ethnicity, and ICV as covariates.

Results

Demographics and Volume-of-Interest Differences

There was no group difference in sociodemographics (age, gender, handedness, ethnicity, and educational level) or past history of substance use disorder (table 1). No group difference in ICV, total GMV, total WMV, hippocampal volume, or ventricular volume between ARMS and HC was observed (table 1).

GMV and WMV Differences Between ARMS Subjects and Healthy Controls

We found no regional GMV or WMV differences between ARMS and HC (ie, voxel-wise cluster-forming threshold of P < .001 and P < .05 corrected at the cluster level). Lowering the initial voxel-wise cluster-forming threshold to P < .01 did not return significant group differences either (P < .05 corrected at the cluster level).

At a voxel-wise threshold of P < .001 and k > 10 voxels (uncorrected at the cluster level), we found one cluster of increased GMV on the right precentral gyrus (k = 88 voxels, t peak = 3.64, Montreal Neurological Institute [MNI] coordinates = 4, 9, 44) and a second cluster of decreased GMV on the right frontal inferior gyrus (k = 17 voxels, t peak = 3.58, MNI coordinates = 46, 15, 21) in ARMS when compared with HC.

Cortical Thickness Differences Between ARMS Subjects and Healthy Controls

We found no regional CT differences between ARMS and HC at P < .05 (FDR corrected). At a voxel-wise cluster-forming threshold of P < .001 (uncorrected at the cluster level), we found one cluster of increased CT on the right frontal pole in ARMS when compared with HC (k = 230 vertices, t peak = 3.78, MNI coordinates = 21, 69, −2).

Conversion to Psychosis

We found no significant difference between HC and ARMS-T, or between ARMS-T and ARMS-NT concerning GMV, WMV, CT, or VOI analyses based on the same set of thresholds. For the VBM analysis, lowering the initial voxel-wise cluster-forming threshold to P < .01 (P < .05 corrected at the cluster level) did not return significant group differences either.

Comorbid Depression and Anxiety Disorders

To investigate structural differences that could be related to anxiodepressive disorders and that affect a large proportion of AMRS individuals, we compared GMV, WMV, CT, and VOI between ARMS with a concomitant diagnostic of depression and/or anxiety (n = 33) and ARMS without (n = 36). We found no significant differences. An additional comparison of GMV, WMV, CT, and VOI between ARMS individuals with antidepressant (n = 37) and those without (n = 32) found no significant difference either.

Discussion

Although there is evidence for the involvement of frontal, temporal, and limbic areas in ARMS for psychosis, the sample size of previous studies is often modest and findings mainly concern ARMS samples from Western countries. In this study, we examined brain structural changes in a large sample of 69 ARMS subjects recruited in Singapore, and for which potential biases introduced by drug use, including antipsychotics and cannabis, were well controlled. Comparison of regional GMV, WMV, and CT as well as ventricular and hippocampal volumes between ARMS individuals and HC revealed no significant differences. The further analysis of the same structures between ARMS-T and ARMS-NT as well as between ARMS-T and HC did not return any positive result either.

Regional reductions of GMV in ARMS subjects are the most common findings in whole-brain VBM studies.^15,44 Only 3 whole-brain VBM studies reported negative findings but their ARMS sample was either unusually young (12–18 years old)^20,22 or small (n = 14).²⁶ Concerning CT, only one previous study¹⁸ used the same preprocessing technique (Freesurfer), while 3 others^17,19,20 used a different algorithm: CLASP⁴⁵ or voxel-based CT.⁴⁶ Their findings were divergent, reporting either extended¹⁷ or no CT differences at the whole-brain level^18–20 in ARMS subjects when compared with HC at baseline. Our results are consistent with the absence of cross-sectional difference between ARMS subject and HC at the whole-brain level reported by the 3 largest studies.^18,19,22 Additional comparison of hippocampal volumes between ARMS and HC showed no significant difference as well. Reduced hippocampal volume is a frequent finding from region-of-interest studies in ARMS samples^47–51 and has been shown to be statistically significant at the whole-brain level in 1 VBM study,⁴ although some inconsistences have also been reported.^52,53 The higher sensitivity of manual tracing methods to detect volumetric changes in medial temporal structures could explain our inability to replicate hippocampal volume reduction often reported by manually traced region-of-interest studies in ARMS samples. However, Freesurfer automated segmentation performance has been shown to produce volumetric data that were very close to those obtained with the “gold standard” manual tracing method.⁵⁴

The sensitivity of our analyses did not improve when specifically comparing ARMS-T with HC or ARMS-NT. These additional group comparisons were clearly underpowered due to the small number of subject in the ARMS-T group (n = 7). However, a recent well powered study has also reported the absence of structural abnormalities in ARMS-T when compared with ARMS-NT at the whole brain level.¹⁹

The absence of relationship between clinical high-risk status (regardless of later transition or nontransition to psychosis) and brain structure might be attributed to unique characteristics of LYRIKS. Understanding the local pathways to care for the ARMS subjects is an important area of work, and efforts are currently underway. In a previous publication, we found that LYRIKS sample, was comparable to other samples from the UK or Australia concerning social and clinical profiles.³⁵ Accordingly, clinical characteristics reported in table 1 (ie, CAARMS ratings, grouping and comorbidities) are also comparable to those from OASIS and PACE samples,⁵⁵ although the rate of conversion to psychosis (ie, 10% at 28 months) is probably among the lowest reported.¹ However, ethnicity differences might be contributing to the negative findings as most participants in the LYRIKS sample have Asian origins. Another interesting difference could be the relative lack of drug use, including cannabis and/or antipsychotics in our sample. Half the ARMS individuals were pharmacologically treated for depression and/or anxiety and both the medication and the affective disorder could potentially impact brain structure. Last, the relatively conservative whole-brain approach could explain divergences with other region-of-interest studies. These 4 points are developed below.

Ethnicity

It is widely recognized that the expression of psychotic symptoms varies among ethnic groups.^56,57 Although these disparities seem more related to psychosocial inequalities than to ancestry differences,⁵⁸ it raised the idea that ethnical differences could be instructive regarding the pathogenesis of schizophrenia.⁵⁹ Accordingly, a structural MRI study reported an effect of ethnicity on gray-matter findings following a first episode of psychosis.⁶⁰ These neuroimaging findings should be interpreted with caution regarding the modest sample size and the abundance of possible confounds, nevertheless, they suggest that some neuroanatomical features of psychosis could be specific to the ethnic group under investigation. In general, it is not very likely that our negative findings are attributable to the ethnical characteristics of our sample alone. Nevertheless, a different genetic background may modify the susceptibility of the brain to different etiological factors⁶¹ and could impact the neuroanatomical correlates of the pathophysiological process.

Drugs

Singapore has the second lowest annual prevalence of cannabis-use worldwide (0.005 in 2006)⁶² and no participant in our sample reported current illicit drug use. While most neuroimaging studies in ARMS excluded subjects with current and/or past substance abuse and/or dependence regarding the DSM or the International Classification of Diseases (ICD), they possibly included cannabis users as long as they did not fulfill the criteria for abuse or dependence. Only few studies specified the proportion of cannabis users in their sample but the reported rate can be as high as 35% for current use^9,63 and up to 70% for a history of cannabis use.¹⁰ In these previous studies, the prevalence of cannabis use did not statistically differ between ARMS subjects and controls, suggesting that neuroimaging findings were not driven by cannabis use only. Nevertheless, this does not exclude the possibility that cannabis use could act as a risk-modifying factor by interacting with other risk factors like genetics and have more dramatic consequences in the group of ARMS than in HCs.^64,65 Accordingly, 3 recent studies in early psychosis have shown that the amount of gray matter loss in the cingulate cortex was either positively correlated with cannabis use^34,66 or restricted to cannabis users only.⁶⁷ Moreover, the hippocampus is rich in endocannabinoid receptors and hippocampal volume reduction has been strongly associated with cannabis use in a recent meta-analysis,⁶⁸ suggesting that the absence of hippocampal atrophy in our sample may be partly related to the relative lack of cannabis use.

Antipsychotics are another potential confounding factor because they have been shown to alter GMV in schizophrenia after both continued⁶⁹ and short-term treatment⁷⁰ administration. In this study, we can exclude the potential influence of antipsychotic treatment on our results as only 3 subjects received a very small dose (< 15mg week of haloperidol equivalent). However, the absence of antipsychotic use is unlikely to explain our negative findings, given the results of a recent meta-analysis indicating an effect of antipsychotics on GMV in the opposite direction (ie, antipsychotics reverse the GMV reductions associated with a greater risk of transition to psychosis).¹⁵

Affective Comorbidity

Approximately half of ARMS individuals in our sample had a comorbid depressive and/or anxious disorder, a proportion that is comparable with other ARMS samples.⁵⁵ Disentangling emerging psychosis with concomitant mood disturbances from depression or anxiety with attenuated psychotic symptoms is challenging from both a clinical and neuroanatomical point of view. Similarly to psychosis, affective disorders may also show neuroanatomical features within medial prefrontal and medial temporal structures⁷¹ and this could represent an important source of confound for neurostructural findings in ARMS. Accordingly, a recent study showed that comorbid depression and anxiety may contribute to GMV reduction in the anterior cingulate cortex in ARMS.⁷² In our sample, we did not find any effect of comorbid depression and/or anxiety or antidepressant treatment on regional GMV, WMV, CT, or VOI. However, we cannot exclude that antidepressant treatment may have interfered with the natural course of ARMS individuals.^73,74

Whole-Brain Analysis

We made the initial choice of a whole-brain analysis because it is a common and well accepted statistical approach for both VBM and SBM analyses. Moreover, in the context of an excess of significance in the neuroimaging literature,^75,76 the whole-brain approach limits the risk of publication bias toward positive findings that is thought to be partially responsible for the lack of reliable biomarkers in psychiatry despite intense research in neuroimaging.⁷⁷ Indeed, region-of-interest studies are directed towards regions that can be easily anatomically delimited or regions of theoretical importance, which intrinsically depend on results from previous studies, thereby inflating the risk of confirmation bias. We completed the initial whole-brain approach with the individual analysis of 2 VOIs (ie, ventricles and hippocampus) that are commonly implicated among structural findings in psychosis but are the best assessed individually, using volumetric information from the subcortical segmentation in Freesurfer. Instead of running additional region-of-interest analyses in the hypothesized frontotemporal and limbic regions, we examined the group difference using P < .001 uncorrected, at the voxel or the vertex level for both the VBM and SBM analyses, respectively. In the context of the literature, neither the direction of the trend (ie, increased GMV or CT), nor the location of the clusters (ie, precentral gyrus, frontal pole) advocate in favor of true differences between ARMS and HC. For inclusion of these data in a meta-analysis, GMV, WMV, or CT for a specific region are available on request to the corresponding author (J.Z).

Our results might be limited by the cross-sectional design of the study. Cannon and colleagues have recently reported greater GM loss over time in several frontal areas of ARMS-T when compared with ARMS-NT or HC, although they observed no CT differences between all 3 groups when compared cross-sectionally at baseline.¹⁸

Last, our analysis was limited to anatomical changes in gray and white matter segments. Two functional MRI studies in the same ARMS sample have previously reported alterations in task-based activations⁷⁸ as well as abnormalities in functional-connectivity at rest⁷⁹ when compared with HC. This suggests that, in our sample, (1) there might be very little structural change in ARMS or (2) VBM and SBM analyses may not be sensitive to detect subtle structural differences. Functional or diffusion MRI studies might reveal more insights on the pathophysiology changes in youths at high clinical risk for psychosis.

Conclusion

Taken together, this comprehensive cross-sectional analysis of regional volumes and CT was conducted in a relatively large sample of ARMS subjects, mainly free of possibly important confounds including antipsychotic medication and substance abuse. Only few whole-brain studies have examined brain structural changes in an ARMS sample of comparable size, particularly in Asian populations.⁸⁰ We found no evidence of regional GMV, WMV, or CT differences between ARMS and HC, ARMS-T and HC, or ARMS-T and ARMS-NT at baseline. The small number of ARMS transitioning to psychosis and the absence of longitudinal analysis of brain changes over-time are clear limitations, especially in light of recent findings suggesting progressive structural changes in ARMS despite the absence of baseline differences with HC.¹⁸ Nevertheless, our negative findings suggest that there may be no dramatic alterations of regional brain volumes or CT in ARMS when the incidence of possible confounds is limited.

Funding

National Research Foundation Singapore under the National Medical Research Council Translational and Clinical Research Flagship Program (NMRC/TCR/003/2008); Ministry of Health, Singapore; Swiss National Science Foundation ((ID 139872 and 145598 to P.K.); Swiss Society for Medicine and Biology Scholarships (ID 148384 to P.K.); National Health and Medical Research Council Project Grant (ID 1050504 and 1066779 to A.F.); Australian Research Council Future Fellowship (ID FT130100589 to A.F.).