Empirical investigation into the structure of psychopathology supports a dimensional conceptualisation of what constitutes mental disorders over the classic categorical approach that is the current psychiatric nosology (Caspi et al., 2014; Kotov et al., 2017). In parallel, investigation into neurobiological mechanisms of psychopathology find robust evidence that brain patterns are shared across mental disorder diagnoses (Goodkind et al., 2015; Opel et al., 2020; Sha et al., 2019). Moreover, most mental disorders onset in the first three decades of life (Kessler et al., 2007), a time associated with large-scale reorganisation and maturation of the brain (Paus et al., 2008). Capitalising on these observations, it seems reasonable to assume that in order to understand psychopathology and what causes it, we must first understand what is shared and what is distinct across different forms of psychopathology.
In this thesis, we aimed to disentangle neurobiological correlates of different forms of psychopathology. We used multivariate statistics to investigate brain-behaviour associations related to dimensional and categorical measures of psychopathology. Specifically, given the important context of development, we performed this work in a sample of children and adolescents between the age of 5 and 21. Most of the participants of this sample had at least one mental disorder diagnosis. We then performed out-of-sample validation of our findings in three different samples of children and adolescent from the general population.
The main findings of this thesis will be integrated and discussed in the context of what is shared and what is distinct across different forms of psychopathology. In paper I, we investigated shared associations across measures of brain structure based on magnetic resonance imaging and measures of mental health, cognitive, and socio-environmental factors. We found evidence for two latent dimensions or “modes”: one reflecting physical and cognitive maturation, and another reflecting a cross-diagnostic pattern linking social and cognitive troubles with reduced white matter surface area. Of note, these patterns were consistent across diagnostic groups. In paper II, we narrowed the focus down to the investigation of shared associations across measures of brain function and mental health measures only. Specifically, we utilised both categorical and dimensional approaches to psychopathology to identify their shared associations with functional magnetic resonance imaging resting-state functional connectivity. We found evidence for a shared pattern relating functional connectivity to five dimensions of psychopathology, recapitulating the psychopathology hierarchy. Autism-spectrum disorder was the only diagnostic category to exhibit a specific brain functional connectivity pattern. In addition, we identified a connectivity pattern related to a categorical cross-diagnostic case-control pattern (i.e., no diagnosis versus all diagnoses) and a dimensional cross-diagnostic case-control pattern (i.e., allowing the diagnoses to cluster by their covariance with functional connectivity resulted in this pattern). To further expand on this relationship between brain measures and measures of psychopathology, we then investigated in paper III whether the categorical patterns identified in paper II were sensitive to questionnaires measuring psychopathology in yet another independent Norwegian cohort. Here, we found that the categorical connectivity patterns replicated, but that they were not sensitive to symptom load in the validation sample.
The findings of this thesis should be interpreted within the constraints of the chosen methodology. Notably, the data that forms the basis of the work is cross-sectional, thereby limiting any inferences to be made regarding change or developmental trajectory. Moreover, although out-of-sample validation was performed, generalisability of the findings was only partly demonstrated. Other considerations pertain to known limitations of functional brain imaging methodology, multivariate statistics, and studying developmental and clinical samples.
In sum, this thesis highlights the utility of multivariate statistics in disentangling brain-psychopathology relationships, as well as bridging the relevance of such associations from population-based studies to a clinical developmental sample. Importantly, the results suggest that the overarching associations are shared across diagnostic boundaries. Future studies may attempt to validate these shared brain-behaviour patterns across more comparable samples. The findings of this thesis support the notion that shared, transdiagnostic dimensions are more plausible operationalisations of psychopathology than categorical diagnoses. Although this has wide applications for the psychiatric nosology and the conceptualisation of psychopathology, an important caveat is that the design and methodology of this thesis do not permit inferences regarding aetiology or mechanistic insight. Towards achieving this goal, however, this thesis provides evidence that similarities show proclivity over differences in brain-behaviour associations relevant for a wide range of psychopathology in youth.