Forty-seven health seeking participants aged 14 to 29 years were recruited from a specialised tertiary referral service for assessment and early intervention of mental health problems in young people [24
] as part of a longitudinal study of youth mental health at the Brain and Mind Research Institute (BMRI), Sydney, Australia. Inclusion criterion for this sub-study were: i) persons aged 12-30 years seeking professional help primarily for significant anxiety, depressive, hypomanic or psychotic symptoms and, ii) willingness to participate in other neurobiological and longitudinal research with the BMRI related to clinical outcomes [26
]. As such, this cohort represents a selected sub-set of a much broader cohort (n = 1483, 1260 of whom were aged 12-25 years) who presented to our services for clinical care during the same time period [25
]. In addition 33 healthy control subjects (age range 15 to 28) were also recruited from the general population.
Subjects were excluded if they did not have sufficient English-language skills or had insufficient intellectual capacity to participate in the neuropsychological aspects of the concurrent studies [27
]. The Human Research Ethics Committee of the University of Sydney approved this study, and all patients gave prospective written informed consent for their clinical data to be used for research purposes. Parental consent was obtained for patients under 18 years of age.
All patients who entered the services were assessed and managed by medically and/or psychologically-trained health professionals [28
]. In this sub-study, an independent psychiatrist or trained research psychologist also conducted a standardized clinical interview, focusing on assessment of the detailed criteria developed for formal application of our clinical staging framework [13
]. For those who are assessed in clinical environments, discrete categories (stages 1 to 4; whereby, stage 1a = 'help-seeking'; stage 1b = 'attenuated syndrome'; stage 2 = 'discrete disorder'; stage 3 = 'persistent or recurrent illness'; and stage 4 = 'chronic debilitating illness') are described. A key point of differentiation, however, occurs between those early 'sub-threshold' syndromes (classed as stage 1a or 1b) and the onset (stage 2) of a more discrete disorder. Importantly, entry to stage 2 was not simply analogous to, or defined by, meeting existing DSM or ICD criteria for a specific mood or psychotic disorder. At the time of scan 23 patients were identified to be at the 'attenuated syndrome' [stage 1b] (15 females) and 24 were rated to be within the 'discrete disorder' [stage 2] (N = 13; 5 females) or 'persistent or recurrent illness' [stage 3] (N = 11; 5 females) stage. The stage 2/3 cohorts included patients with a diagnosis of bipolar disorder, depression and psychosis. To ensure consistent inter-rater reliability estimates, two approaches were applied: i) first, we compared the determination of clinical stage by the original treating clinician with that assigned by the consensus pairs working from the records. The key comparison was for the degree of concordance between ratings of 'sub-syndromal or attenuated states' (stage 1) versus later established disorders (stages 2-4); ii) second, the ratings derived from a consensus pair were compared with that derived by an independent rater from our research team.
All imaging was conducted at the BMRI imaging facility using a 3T GE Discovery MR750 scanner (GE Medical Systems, Milwaukee, WI). Structural images were acquired using an 8-channel phased array head coil using a T1-weighted magnetization prepared rapid gradient-echo (MPRAGE) sequence producing 196 sagittal slices (TR = 7.2 ms; TE = 2.8 ms; flip angle = 10°; matrix 256 × 256; 0.9 mm isotropic voxels). For each subject, two T1-weighted MRI scans were obtained in single scan sessions, and individual structural whole brain MRI's were averaged to increase signal-to-noise ratio.
An unbiased optimised VBM analysis was carried out using FSL (FMRIB Software Library, http://www.fmrib.ox.ac.uk/fsl
, Smith et al., 2004) with no a priori information regarding possible loci of structural changes in the grey matter. The FSL-VBM analysis pipeline was then conducted as follows. Firstly, brain tissue was extracted prior to tissue-type segmentation using BET [29
]. All T1-weighted images were then transformed into standard space using a limited degrees-of-freedom non-linear model to ensure maximum spatial alignment and images were corrected for non-uniformity. The FAST4 [30
] tool was then applied to segment tissues according to their type. The segmented grey matter partial volume images were then aligned into the MNI standard space by applying the affine registration tool FLIRT [31
] and nonlinear registration FNIRT [33
] methods, which use a B-spline representation of the registration warp field [35
]. An averaged study-specific template was then created using all the unsmoothed images to which the grey matter partial volume images were re-registered and these registered images were then modulated to correct for local expansion and contraction by the Jacobian of the warp field. Rigorous visual inspection was used to ensure the quality of brain image extraction, segmentation, and registration for each averaged structural image. The modulated, segmented images were then smoothed with an isotropic Gaussian kernal with 3 mm standard deviation (FWHM 7.05 mm). The most common quality control issue encountered was suboptimal brain extraction using the BET tool. This was evident in brain-extracted images, which had unwanted skull and neck tissue included as part of the final processed image. In such instances it was necessary to re-run the brain extraction step using the "bias field correction and neck cleanup" option in BET.
Finally, permutation-based non-parametric testing, using a 5000 permutation set was used in a voxel-wise GLM [36
] for contrasting differences between (i) healthy controls versus stage 1 patients (help-seeking/attenuated syndrome), (ii) healthy controls versus stage 2/3 (discrete disorder/persistent or recurrent illness) which depicts a more established disease process and finally (iii) stage 1 versus stages 2/3. To identify regionally specific associations that were not confounded by differences in age and total intracranial volume across the three groups, both these variables were entered into the design matrix as confounding covariates. Family-wise error (FWE) correction [37
] was used to correct the threshold for multiple comparisons across space and threshold-free cluster enhancement was employed to assess cluster significance [38
]. The significant p-value with the FWE corrected threshold was set at p < 0.05.