White matter changes in psychosis risk relate to development and are not impacted by the transition to psychosis

Abstract

Subtle alterations in white matter microstructure are observed in youth at clinical high risk (CHR) for psychosis. However, the timing of these changes and their relationships to the emergence of psychosis remain unclear. Here, we track the evolution of white matter abnormalities in a large, longitudinal cohort of CHR individuals comprising the North American Prodrome Longitudinal Study (NAPLS-3). Multi-shell diffusion magnetic resonance imaging data were collected across multiple timepoints (1-5 over 1 year) in 286 subjects (aged 12-32 years): 25 CHR individuals who transitioned to psychosis (CHR-P; 61 scans), 205 CHR subjects with unknown transition outcome after the 1-year follow-up period (CHR-U; 596 scans), and 56 healthy controls (195 scans). Linear mixed effects models were fitted to infer the impact of age and illness-onset on variation in the fractional anisotropy of cellular tissue (FA_T) and the volume fraction of extracellular free water (FW). Baseline measures of white matter microstructure did not differentiate between HC, CHR-U and CHR-P individuals. However, age trajectories differed between the three groups in line with a developmental effect: CHR-P and CHR-U groups displayed higher FA_T in adolescence, and 4% lower FA_T by 30 years of age compared to controls. Furthermore, older CHR-P subjects (20+ years) displayed 4% higher FW in the forceps major (p < 0.05). Prospective analysis in CHR-P did not reveal a significant impact of illness onset on regional FA_T or FW, suggesting that transition to psychosis is not marked by dramatic change in white matter microstructure. Instead, clinical high risk for psychosis-regardless of transition outcome-is characterized by subtle age-related white matter changes that occur in tandem with development.

Figure 1.

Flowchart depicting the selection of suitable scans.

Figure 2.. Age-related change in white matter microstructure.

Panel A presents curves modeling age-related change of white matter tissue anisotropy (FA_T) and free water (FW) in healthy controls (HC; blue curve; β₁ + β₄ × Age), clinical high-risk individuals with unknown transition outcomes (CHR-U; green curve; β₁ + β₂ + (β₄ + β₅) × Age) and in CHR individuals who transitioned to psychosis (CHR-P; red curve; β₁ + β₂ + (β₄ + β₆) × Age). Shaded areas reflect 95% confidence intervals estimated with bootstrapping (1,000 samples). Panel B presents the percent of age-related change per year (i.e., slope) in FA, FA_T and FW averaged across each of the 8 ROIs, in HC (β₄), CHR (β₄+ β₅) and CHR-P (β₄+ β₆) groups. The FA_T slope in whole-brain white matter significantly differed between CHR-P and HC (FDRp<0.05, as indicated by the asterisk). Negative percentages indicate age-related decline in dMRI values. Abbreviations: inferior fronto-occipital fasciculus (IFOF), inferior longitudinal fasciculus (ILF), superior longitudinal fasciculus (SLF), uncinate fasciculus (UF) and white matter (WM).

Figure 3.. Significant group differences across ages in white matter microstructure.

Graphs present between-group differences in fractional anisotropy (FA), tissue FA (FA_T) and free water (FW) as a function of age, quantified by β₃/β₁ × 100% in clinical high-risk with unknown transition outcomes (CHR-U) and by β₄/β₁ × 100% in CHR individuals who transitioned to psychosis (CHR-P). Negative percent (%) differences indicate lower values in CHR-U and CHR-P relative to healthy comparison (HC) subjects. Average dMRI measures were quantified yearly using age centering between 12 and 32 years. Age epochs at which the average dMRI measure significantly differed between the CHR groups and comparison subjects are denoted with red/green squares (RAWp<0.05), as determined by the significance of the β₃ and β₄ regression coefficients, respectively. Abbreviations: superior longitudinal fasciculus (SLF).

Figure 4.. Association between neurobehavioral and dMRI phenotypes.

Panel A presents a schematic of canonical correlation analysis (CCA) to examine multivariate neurobehavior-dMRI covariation at baseline. CCA maximizes covariation between the canonical score for neurobehavioral measures (U), and the canonical score for dMRI measures (V). Panel B presents correlations between each of the 49 neurobehavioural phenotypes with V and Panel C presents correlations between each of the 24 dMRI measures with U. Error bars indicate 95% confidence intervals estimated with bootstrapping (1,000 samples) and black squares demarcate significant correlations, after FDR correction across the neurobehavioural and dMRI phenotypes respectively. Abbreviations: psychopathology (PSYCH), inferior fronto-occipital fasciculus (IFOF), inferior longitudinal fasciculus (ILF), superior longitudinal fasciculus (SLF), uncinate fasciculus (UF), white matter (WM), global assessment of functioning (GAF); global functioning (GF), premorbid adjustment scale (PAS) and childhood brain injury (CBI).