Shedding light on the onset of psychiatric illness: looking through a developmental lens

Background

Understanding the mechanisms driving the onset of major psychiatric illnesses such as depression, bipolar disorder and schizophrenia has been far more challenging than expected. Contributing factors include that symptoms associated with each of these disorders overlap during different phases of emerging illness¹ and over the course of established illness.² In addition, while there is evidence supporting specificity of familial segregation,³ at least some genetic risk factors appear to be shared⁴ and treatment response crosses currently defined diagnostic boundaries.⁵ 

But does it follow, as some propose,⁶ that psychiatric illnesses are really different manifestations of a common origin—in other words that different psychiatric illnesses derive from pluripotent or shared beginnings and specific outcomes depend on mediating and moderating influences? Epidemiological and high-risk prospective studies provide compelling evidence to the contrary. In fact, a developmental perspective has illuminated clear distinctions in the early trajectories between illness such as schizophrenia and bipolar disorder.⁷ Further longitudinal studies of children over the peak risk period have highlighted the inadequacy of current cross-sectional approaches to psychiatric diagnosis that do not take into account the developmental and emergent course of psychopathology against the backdrop of family history and other predictive risk factors.⁸

Using bipolar disorder as an example, longitudinal studies of high-risk offspring of affected parents followed through childhood into adulthood have clearly demonstrated that psychopathology is in evolution—a moving target. What first manifests as anxiety or sleep problems in high-risk children might evolve into a depressive disorder in adolescence followed by an index hypomanic or manic episode in emerging adulthood.⁹ However, this evolving developmental trajectory differs between distinct end-stage illnesses. For example, clinical and neuroimaging evidence supports that the trajectory into schizophrenia is consistent with a neurodevelopmental process,¹⁰ which is distinctively different from the prototypical developmental trajectory of classical bipolar disorder.⁷

A longitudinal developmental approach is crucial to advance understanding of the mechanisms leading to different end-stage psychiatric illnesses. In this Commentary, we discuss two recently published papers both of which focus on mapping developmental trajectories of psychopathology with associated genetic or neural correlates and specific outcomes in adulthood.

Recent papers

A study recently published in JAMA Psychiatry authored by Rice and colleagues¹¹ prospectively investigated the developmental trajectories of depressive symptoms and the association with genetic risk liability. The study was based on data from the Avon Longitudinal Study of Parents and Children including over 7500 adolescents.

Self-reported depressive symptoms were assessed up to six times between 10.5 and 18.5 years of age using the short version of the Mood and Feelings Questionnaire. Polygenic risk scores for individual subjects were estimated for major depression, attention deficit hyperactivity disorder (ADHD) and schizophrenia. In addition, clinical phenotypes were assessed including childhood ADHD, social communication problems and pragmatic language difficulties identified by age 7, psychotic experiences at age 12 and 17 years, and family history of severe depression and schizophrenia.

Latent class growth analysis revealed three distinct trajectories of depression symptom scores including persistently low (73.7%), a later adolescent onset with first elevated probability at 16.5 years (17.3%) and an early adolescent onset with first elevated probability at 12.5 years (9.0%). The proportion of women in the later and early adolescent–onset trajectories was comparably high and over 70%.

In assessing polygenic risk scores, the typical later (post-pubertal) adolescent-onset depressive trajectory that persisted into adulthood was associated with elevated genetic risk for depression only, while the early-onset depressive trajectory was associated with all genetic risk scores assessed. A multivariate analysis showed that the early-onset depressive trajectory was most strongly associated with genetic risk for ADHD, while the association with genetic risk for psychotic disorders remained significant but not for major depression. Furthermore, higher rates of childhood neurodevelopmental disorders (ADHD, pragmatic language and social communication problems) and psychotic experiences at age 12 differentiated the early-onset from the later adolescent–onset depressive trajectory class.

This study provides evidence of distinct developmental trajectories of depressive symptoms distinguished by age of onset. Further, findings support differential genetic underpinnings associated with these distinctive depressive trajectories and underscore the importance of considering heterogeneity of psychopathology.

Main strengths of this study include the large representative birth cohort sample, prospective repeated measures design, rigorous methodology to determine polygenic risk scores and cross-validation of differential phenotypic associations between trajectory classes.

A common trade-off with large representative cohort studies is reliance on self-report measures to index psychopathology, as well as study attrition. A related concern is the possible under-reporting in male youth contributing to biased sex-specific results and the use of clinical threshold cut-offs of self-report measures, which poses the risk of misclassification. Further, non-random missing data were observed, where missingness was associated with depression scores, age, sex, and polygenic risk scores for ADHD and schizophrenia. The authors are commended for their rigorous approach to missing data through sensitivity analyses and inverse probability weighting, but the attrition biases may still add complexity to the interpretation of findings.

This study is consistent with accumulating evidence from other clinical, family and high-risk studies of heterogeneity in psychopathology as indicated by aetiologically distinct trajectories of depressive symptoms in youth.^{12 13} Specifically, depressive symptoms in early childhood have been associated with psychosocial adversity and shared genetic risk factors and clinical phenotypes indexing neurodevelopmental disorders and psychosis. On the other hand, post-pubertal adolescent-onset depression is more strongly associated with genetic risk for mood disorders which persist into adulthood.

A second paper by Luby and colleagues¹⁴ describes a longitudinal investigation designed to unpack cross-sectional associations between reduced volumes in orbitofrontal and striatal brain regions (associated with reward processing) and anhedonic symptoms in depressed adolescents, and an association been structural variations in these regions and adolescent substance abuse. The sample comprised 193 children ages 3–6 years at baseline oversampled for symptoms of depression. Participants completed up to three scans and had a final research assessment between 13 and 18 years of age 3 years after the final scan to assess substance use.

The main findings include a positive association between anhedonia and orbital frontal volumes in childhood followed by a steeper decline in volume and thickness of this region with age in those with higher anhedonic symptoms—although decreases occurred for all levels of anhedonia. Orbital frontal cortical volumes and thickness at age 12 and maturational trajectory over time was negatively associated with alcohol and/or marijuana use 3 years after the last scan.

Implications of this longitudinal study of preschoolers into adolescence includes evidence that anhedonia in childhood may influence the maturation of the orbitofrontal cortex. Authors discuss the finding that both steeper rates of growth and subsequent thinning in brain maturation might signal a high-risk state for subsequent psychopathology. In this study, such changes in the maturational trajectory of the orbital frontal cortex over childhood predicted a higher frequency of alcohol and marijuana use in adolescence, but not with depression.

The strength of this study lies in achieving three repeated scans over a developmentally sensitive period along with the parallel collection of repeated phenotypic information. Other methodological strengths include the use of the same scanner and interpretation of images by blind raters.

Attrition is a weakness, and compared with the Rice and colleagues’ manuscript above, this paper provided less information as to the nature of potential biases incurred through loss to follow-up. Further, assessments of children relied on parent and self-report after age 9 years. While the ability to link structural brain data to later substance use was commendable, the study was not designed to address this question. Reliance on a structured interview (CIDI) administered by trained raters at one time point to assess substance use is a major weakness. Ideally, substance use would have been assessed on the basis of a clinical or semistructured research assessment, as well as mapped prospectively using a more granular self-report measure. Finally, substance use was assessed in early adolescence when participants were 13–18 years of age. Given that substance abuse peaks over late adolescence and into early adulthood, many of the participants might have been counted as false negatives (assessed too early in development).

Concluding remarks

Longitudinal studies over childhood and adolescence make it possible to map the developmental trajectory of emergent psychopathology in parallel with multilevel markers (clinical, neurological, biological and behavioural) and risk exposures. This developmental approach allows for the disentanglement of cause from effect and advances the level of analysis from correlation to causation. In this way, complex pathways leading to illness onset, including dynamic interactions and modifying influences, can be characterised and time-varying covariates accounted for.

The papers reviewed here provide complimentary examples of a developmental approach to the identification of distinct trajectories of clinical symptoms and brain maturation mapped to end-stage disorders. Both illustrate the potential power of this approach to understand complex and evolving pathways to illness onset and persistence. However, given the substantial phenotypic and aetiological heterogeneity, along with the dynamic nature of emergent psychiatric illnesses, careful clinical assessment of the outcome of interest seems essential. Periodic comprehensive clinical assessments are preferred over structured interviews and could be supplemented by remote capture methods involving digital technologies to capture high-frequency or fine-grained phenotypic information over time.¹⁵ Remote capture approaches also offer a potential solution to alleviate biases from attrition, an inevitable and problematic design feature of trajectory-based studies through efficient and user-friendly assessments that can be completed outside the research setting. Moving forward, a developmental lens focusing on homogeneous populations and adopting deep phenotyping methods should shed light on illness onset, novel targets for preventive and early intervention, and advance individualised risk prediction.