Method

Participants comprised adolescents with a history of suicide attempt and MDD (n = 28); a history of MDD but no suicide attempt (n = 31); and no personal or family history of psychiatric disorder or suicide attempt (healthy controls, n = 41). Suicide attempts were defined by the Columbia Classification Algorithm of Suicide Assessment. ^{Reference Posner, Oquendo, Gould, Stanley and Davies4} Exclusion criteria included neurological disorders, anoxia history, head injuries, Wechsler verbal score <80, ^{Reference Wechsler5} current use of sedative medication, pregnancy, ineligibility for magnetic resonance imaging (MRI), bipolar disorder, psychosis, substance misuse or positive urine drug/saliva alcohol screen, and left-handedness (owing to participant recruitment for concurrent functional neuroimaging studies). Axis I disorders were assessed using the Kiddie-SADS – Present and Lifetime version (K-SADS-PL). ^{Reference Kaufman, Birmaher, Brent, Rao and Ryan6} Suicide attempt was assessed with the Suicide Intent Scale and Suicide History Form. ^{Reference Beck, Schuyler, Herman, Beck, Resnick and Lettieri7,Reference Oquendo, Halberstam, Mann and First8} Depression, anxiety, suicidal ideation and pubertal status were assessed respectively with the Beck Depression Inventory (BDI), the Screen for Childhood Anxiety Related Emotional Disorders (SCARED), the Suicidal Ideation Questionnaire (SIQ) and the Petersen Pubertal Development Scale. ^{Reference Beck, Ward, Mendelson, Mock and Erbaugh9–Reference Petersen, Crockett and Richards12} Past trauma was determined by psychiatrist assessment and/or the Child Trauma Questionnaire. ^{Reference Bernstein, Fink, Handelsman and Foote13} Positive trauma history included sexual or physical abuse, or trauma with risk of death or bodily harm.

Two MRI scanners were used: a 3 T Siemens Allegra for 49 scans and a 3 T Siemens Trio for 51 scans (Siemens, Erlangen, Germany). We acquired T ₁-weighted magnetisation-prepared rapid gradient echo (MPRAGE) structural images of 240 slices (thickness 0.8 mm): repetition time 1630 ms, echo time 2.48 ms, inversion time 800 ms, field of view 200 mm, flip angle 8°, matrix 256 × 256). Brain cortical thickness and grey- and white-matter volumes were measured using FreeSurfer version 5.1 for Linux (https://surfer.nrm.mgh.harvard.edu/). Topographical defects were automatically corrected and images normalised. Cortical thickness measures were computed as the distances between the grey/white-matter boundary and the pial surface. ^{Reference Fischl and Dale14} Cortical volumes were calculated from the surface mask. ^{Reference Fjell, Westlye, Amlien, Espeseth, Reinvang and Raz15} Two whole-brain surface-based analyses of covariance were completed in Qdec1.4 (FreeSurfer application) to examine the main effect of group on cortical thickness and volume respectively, with age, gender, total brain volume and scanner used as covariates. Groups were equally distributed across both scanners, and three participants were scanned in both scanners and compared. Monte Carlo simulation analyses were performed to correct for multiple voxel-wise comparisons in Qdec, and provided a cluster-wise significance threshold of P<0.05 for these analyses.

To examine the nature of between-group differences in cortical thickness and volume, values were extracted from all cortical regions identified in the above analysis of covariance (ANCOVA). Post hoc pairwise, between-group independent t-tests were conducted on these extracted values. Significance thresholds for comparisons were P<0.05 (two-tailed) and Bonferroni-corrected for the three post hoc comparisons. Exploratory correlational analyses to examine relationships between any cortical thickness and volume abnormalities and clinical variables were completed in SPSS version 20.0.