We conducted a longitudinal, multicentre, case-control study in two different centres from the same geographical region, the Hospital Clínic of Barcelona and the Hospital Sant Joan de Déu. This study was approved by the Research Ethics Committee from both centres, and was carried out in accordance with the latest version of the Declaration of Helsinki (2013). Prior to study participation, parents or legal guardians of all participants provided written informed consent, and adolescents older than 12 years gave written informed assent.
Participants
A total of 108 adolescents aged 10 to 17 years, including 53 help-seeking adolescents at CHR-P and 55 HC, were evaluated at baseline and at 18-month follow-up. Participants at CHR-P were included if they presented with attenuated positive, negative symptoms and/or brief limited intermittent psychotic symptoms [24, 25], and/or had a 1st/2nd degree relative with psychotic disorder or a diagnosis of schizotypal disorder, plus a recent decline in functioning [26], evaluated with the Structured Interview for Prodromal Symptoms, and the Scale Of Prodromal Symptoms (SIPS/SOPS) [24]. Exclusion criteria for both groups encompassed previous personal history of psychosis, autism spectrum disorder, intellectual disability with impaired functioning, neurological comorbidity, history of cranioencephalic trauma with loss of consciousness, and/or counterindication for undergoing MRI scanning. Detailed information about the recruitment procedure can be found elsewhere [27].
Clinical and cognitive assessment
Clinical evaluations were carried out at baseline, 6, 12 and 18-month follow-up for adolescents at CHR-P, and at baseline and 18-month follow-up for HC. Baseline assessment included collection of socio-demographic data, administration of the SIPS/SOPS, the Kiddie Schedule for Affective Disorders and Schizophrenia for school-age children–present and lifetime version (K-SADS-PL) [28] to assess for possible psychiatric comorbidity; and the global assessment of functioning (GAF) scale, global functioning (GF): social and role scales [29]. A measurement of general cognitive capacity was estimated using the Vocabulary, Similarities, Block Design, and Matrix Reasoning subtests of the Wechsler Intelligence Scale for Children or the Wechsler Adult Intelligence Scale [30, 31], depending on the age at assessment [32]. Cannabis use was recorded as a dichotomous variable, according to consumption in the prior three months; and detailed medication history was recorded from each participant at the time of scanning; as well as doses of antipsychotic drugs, that were transformed into chlorpromazine equivalents [33].
At 6 and 12-month follow-up individuals at CHR-P were evaluated with the SIPS/SOPS to assess for possible transition to psychosis. Participants were classified as CHR-P-T if they scored 6 points in one or more subitems of the SOPS positive scale. At 18-month follow-up all participants repeated the baseline evaluation including the clinical and cognitive assessment.
Neuroimaging acquisition and preprocessing
High-resolution T1-weighted images were obtained on a 3 Tesla Siemens Magnetom Trio Tim or a Magneton Prisma scanner at the Hospital Clinic of Barcelona (2300 ms repetition time, 3.01 ms echo time, 1 mm slice thickness, 900 ms inversion time, 240 mm field of view, 256 x 256 matrix size and 9º flip angle, voxel size 0.94x0.94x1 mm³) and on a 1.5 Tesla General Electric Signa HDxt scanner at Hospital Sant Joan de Déu (9556 ms repetition time, 3.88 ms echo time, 1.2 mm slice thickness, 600 ms inversion time, 256 mm field of view, 192 x 192 matrix size and 8º flip angle, voxel size 0.94x0.94x1.2 mm³) at baseline and at 18-month follow-up, or at the time of transition to psychosis.
Surface-based morphometry was used to reconstruct the white and pial cortical surface for each participant, using automated procedures included in the FreeSurfer 6.0 image analysis suite (http://surfer.nmr.mgh.harvard.edu/). The technical details of these procedures are described elsewhere [34–36]. This processing allows generating CT measures, calculated as the closest distance between the white and pial surfaces [35].
For longitudinal analyses, images where automatically processed with the longitudinal stream in FreeSurfer [37]. Specifically, an unbiased within-subject template space and image was created using robust, inverse consistent registration. Processing steps included skull stripping, transformation into Talairach space, atlas registration as well as spherical surface maps and parcellations of the cerebral cortex, employing the Desikan-Killiany brain atlas [38], which were then initialized with common information from the within-subject template, significantly increasing reliability and statistical power.
Images were visually inspected at every stage and excluded if they presented gross quality defects or relevant artefacts. The ENIGMA cortical quality control protocol 2.0 (https://enigma.ini.usc.edu/protocols/imaging-protocols) was conducted, which consists on visual checks of the cortical segmentations of every scan, and region-by-region removal of values with errors in segmentation.
Additionally, an inter-site reliability study between centres was performed to assess compatibility for CT measures (Appendix 1).
Statistical analyses
Statistical analyses of socio-demographic and clinical data were performed with IBM SPSS Statistics v.25 (https://www.ibm.com/es-es/analytics/spss-statistics-software). T-test, analysis of variance (ANOVA) and chi-square were employed to study between group differences. Significance was set at P < .05.
The model assumptions for the normal distribution were checked using visual inspection of residual plots. Baseline differences in CT between groups were analysed using the general lineal model (GLM) within the mri_glmfit function, implemented in FreeSurfer 6.0. To study longitudinal between-group differences in CT, we performed a longitudinal two stage model. First, vertex-wise estimates of the longitudinal CT change were expressed as the symmetrized percentage of change (CTspc), which measures the rate of change of CT with respect to the average. Second, between-group differences of CTspc were analysed with the mri_glmfit function. For every analysis, sex, age and scanner were included as covariates, and a cluster-wise correction for multiple comparisons was performed, based on Monte Carlo Z simulation, in order to obtain a measure of the distribution of the maximum cluster size under the null hypothesis. Significance was set at a threshold of cluster-wise P (CWP) < .05. Peak coordinates were reported in Talairach space (X,Y,Z). To check the reliability of the results between centres, an intraclass correlation coefficient (ICC) was calculated in terms of absolute agreement and consistency for regions showing significant differences between groups (Table S1 in the Supporting Information).
Predictors of longitudinal changes in cortical thickness
The values of CTspc were extracted from the clusters showing significant differences between groups. A multiple linear regression was performed to test if relevant clinical variables (SOPS positive, SOPS negative, GF role and GF social) at baseline could predict longitudinal changes within regions showing significant differences in the CHR-P-T group. Sex, age and scanner were also fitted to the model. Simple linear regression was also used to study the association between age at baseline and CTspc, to investigate whether age at onset modulated longitudinal structural changes in CT, as seen in youth with schizophrenia [17, 18].
Effect of antipsychotic treatment, cannabis use and intelligence
To study the effect of antipsychotic treatment on regions showing significant differences in CT between groups, we regressed out the effect of baseline dose of antipsychotic medication, converted to chlorpromazine equivalents, on CTspc, using a linear model.
To examine the potential effect of cannabis use, and general cognitive capacity on the results, we ran a new GLM model including the fixed effects for these variables independently. For all analyses, statistical significance level was set at CWP < .05.