Associations between multimodal brain age and the emergence of internalising symptoms in adolescence

563 

Abstract Submission 

Niamh MacSweeney^1,2,3, Dani Beck^1,2,3, Lucy Whitmore⁴, Kathryn Mills⁴, Lars Westlye⁵, Tilmann von Soest¹, Christian Tamnes^1,2,3

¹PROMENTA Research Center, Department of Psychology, University of Oslo, Oslo, Norway, ²NORMENT, Division of Mental Health and Addiction, Oslo University Hospital & Institute of Clinical Medicine, University of Oslo, Oslo, Norway, ³Division of Mental Health and Substance Abuse, Diakonhjemmet Hospital, Oslo, Norway, ⁴University of Oregon, Eugene, OR, ⁵NORMENT, Division of Mental Health and Addiction, Oslo University Hospital & Institute of Clinical M, Oslo, Norway

Niamh MacSweeney  
PROMENTA Research Center, Department of Psychology, University of Oslo|NORMENT, Division of Mental Health and Addiction, Oslo University Hospital & Institute of Clinical Medicine, University of Oslo|Division of Mental Health and Substance Abuse, Diakonhjemmet Hospital
Oslo, Norway|Oslo, Norway|Oslo, Norway

Dani Beck, Dr  
PROMENTA Research Center, Department of Psychology, University of Oslo|NORMENT, Division of Mental Health and Addiction, Oslo University Hospital & Institute of Clinical Medicine, University of Oslo|Division of Mental Health and Substance Abuse, Diakonhjemmet Hospital
Oslo, Norway|Oslo, Norway|Oslo, Norway

Lucy Whitmore  
University of Oregon
Eugene, OR

Kathryn Mills  
University of Oregon
Eugene, OR

Lars Westlye  
NORMENT, Division of Mental Health and Addiction, Oslo University Hospital & Institute of Clinical M
Oslo, Norway

Tilmann von Soest  
PROMENTA Research Center, Department of Psychology, University of Oslo
Oslo, Norway

Christian Tamnes  
PROMENTA Research Center, Department of Psychology, University of Oslo|NORMENT, Division of Mental Health and Addiction, Oslo University Hospital & Institute of Clinical Medicine, University of Oslo|Division of Mental Health and Substance Abuse, Diakonhjemmet Hospital
Oslo, Norway|Oslo, Norway|Oslo, Norway

Adolescence is the period of greatest risk for the onset of internalising symptoms, particularly in females(1,2). How variations in patterns of brain maturation relate to the emergence of these difficulties remains unclear due to inconsistent findings in recent longitudinal studies(3–6). Prior work has focused on group-level differences in brain maturation using unimodal approaches, limiting the quantification of deviations from typical development at an individual level. Brain age prediction allows us to assess individual deviations from age-expected patterns (i.e., brain age gap (BAG)), indirectly informative of brain maturational patterns, and how this relates to the development of internalising difficulties.

Using the Adolescent Brain Cognitive Development (ABCD) Study data (N = ~11,880, 9-10-years at baseline), we examined cross-sectional and longitudinal associations between multi-modal brain age models trained independently on T1-weighted (T1), diffusion tensor (DTI), and resting-state functional (rs-fMRI) MRI data, and self-reported youth internalising symptoms. Although we expected deviations from typical brain development to be associated with greater internalising symptoms, we did not hypothesise about the directionality of this relationship given that both accelerated and delayed patterns of brain maturation have been reported. However, we expected these associations to be stronger in females due to the higher incidence of mood difficulties in females.

Age prediction was calculated for each MRI modality separately using XGBoost regression(7). We used ~50% of the ABCD data (baseline and 2-year follow-up: N= ~9,000 observations) as the test sample, and ~50% for model training (N = ~7,200) and ten-fold cross-validation (N=~1,200). In the training set, R², RSME, and MAE values were used to assess model accuracy and an age-bias correction was applied. Within each modality, BAG was calculated by subtracting chronological age from predicted age, producing T1-BAG, DTI-BAG, and fMRI-BAG for each participant. Internalising symptoms were measured using the Brief Problem Monitor Internalising Scale at 3-year follow-up.

For each MRI modality, univariate latent-change score models(8) were estimated to test 1) whether baseline BAG was associated with later internalising symptoms, and 2) to what extent change in BAG between timepoints (ΔBAG) related to internalising symptoms. We ran multi-group analyses to explore sex differences. We also accounted for earlier internalising symptoms at 6-month follow-up in our models.

As shown in Figure 1, the brain age model performance for DTI was the most accurate (r=0.66, p<0.001, 95% CI=[0.65, 0.67], MAE=0.71), followed by T1 (r=0.59, p<0.001, 95% CI=[0.57, 0.60], MAE=0.79) and rs-fMRI (r=0.41, p <0.001, 95% CI=[0.39, 0.43], MAE=0.89).

Although baseline T1-BAG was not significantly associated with later youth internalising difficulties, ΔT1-BAG was significantly associated with internalising symptoms at follow-up (ß =0.13, p<0.001). Similar results were found for the fMRI-BAG model, whereby ΔfMRI-BAG (ß=0.04, p=0.007), but not baseline fMRI-BAG, was related to later internalising symptoms. These associations remained significant when accounting for earlier internalising difficulties. Our multi-group analyses showed that these effects were significant in females only. The DTI-BAG measures were not found to be associated with youth internalising difficulties.

Our findings suggest that the rate of structural and functional brain maturation is associated with the emergence of mood difficulties in female youth. Female youth that exhibit faster brain ageing over time (ΔBAG), but not older-looking brains to begin with, may be at an increased risk for the onset of internalising symptoms. This pattern of brain maturation could reflect the beginning of a negative developmental trajectory and thus, may be a key window of opportunity for intervention.

Psychiatric (eg. Depression, Anxiety, Schizophrenia) ¹

Normal Brain Development: Fetus to Adolescence

Multivariate Approaches ²

Affective Disorders

Development

FUNCTIONAL MRI

MRI

Multivariate

STRUCTURAL MRI

WHITE MATTER IMAGING - DTI, HARDI, DSI, ETC