Mapping the developmental trajectory of bipolar disorder: Importance of prerequisite groundwork

Investigating the pathways to bipolar disorder at both clinical and biobehavioral levels is extremely important to advance research and clinical practice. This work has its origins in longitudinal studies of children of depressed parents (Weissman et al., 1987) and in studies of youth at clinical risk of developing psychosis (Yung et al., 2004). Longitudinal prospective studies of well-defined high-risk groups over the peak risk period are essential for understanding the complex interplay of risk processes in determining illness onset and in identifying opportunities for early intervention. Given heritability estimates around 80%, children of parents with confirmed bipolar disorder are an informative and identifiable high-risk group. Methodologically, there are challenges, including phenotypic and etiological heterogeneity, to contend with. That is, the nature of the bipolar trait (as indicated by a spectrum of bipolar-related disorders segregating in relatives of bipolar patients) and differences in the range of early clinical antecedents of emerging bipolar disorder points to considerable phenotypic heterogeneity. Furthermore, there is evidence of bipolar subtypes with potentially different etiology, pathophysiology, clinical course and treatment response underscoring the importance of etiologic heterogeneity. Several international offspring studies have described the emerging clinical trajectory and early course of bipolar disorder and are investigating the associated changes in multi-level biopsychosocial risk factors—in some cases based on prospective observations approaching up to 20 years (Duffy et al., 2014). However, most have not considered the influence of heterogeneity on outcomes.

Despite differences in recruitment and assessment methods among independent bipolar offspring studies, their findings have been remarkably convergent. Specifically, that (1) mania (using adult criteria) is rare prior to puberty, even in individuals from multigenerational families with highly penetrant illness; (2) bipolar disorder usually debuts as depressive episodes that are indistinguishable from unipolar depression in mid-adolescence; and (3) bipolar-related mood episodes are not uncommonly presaged by childhood clinical antecedents, including sleep, anxiety, and sub-affective disorders which likely have a different meaning in these high-risk children than in children without a family history of bipolar disorder. But individual offspring studies have varied in reported rates of comorbid disorders, quality of remission, and range of antecedent conditions. Presumably, these differences reflect variation in parent bipolar subtype, health of the other parent, comorbidity of the bipolar parent, early risk and trauma exposures of the offspring, and the social context of rearing—all of which are subject to cultural factors.

Outcome differences can be expected to be increasingly more challenging to reconcile across different risk populations and study approaches. In fact, the early course of bipolar disorder charted from prospective studies of high-risk offspring is strikingly different from that derived from studies of clinical samples of children diagnosed with pediatric bipolar disorder. Pediatric or very-early-onset bipolar disorder is characterized by pre-pubertal mania, high rates of comorbid neurodevelopmental disorders (such as attention-deficit/hyperactivity disorder [ADHD] and pervasive developmental disorder), an ultra rapid cycling course with poor quality of remission, and substantially reduced global functioning. These features are far less prominent in familial high-risk populations. In part, this discrepancy reflects that few psychiatric symptoms have phenotypic specificity and often span diagnostic boundaries. Therefore, sometimes mania-like symptoms may not be related to a bipolar diathesis or reflect any psychiatric disorder at all. In fact, individual symptoms akin to symptoms of mania are readily endorsed by adolescents in the general population and are often regarded as normal variants (e.g. risk-taking behavior, falling in love). At the other end of the severity spectrum, even definitive manic symptoms in the context of an acute psychotic episode are unable to predict the long-term stability of the bipolar diagnosis (Ruggero et al., 2010). It seems that without context, or if not considered alongside other risk factors (e.g. confirmed family history of bipolar disorder), mood symptoms per se have poor prognostic value.

At the heart of these problems, it is increasingly clear that our current approach to mood disorder diagnosis is woefully inadequate, especially for the purpose of mapping illness trajectories. Our current diagnoses are largely cross-sectional and as such rely heavily on the immediate interpretation of what are in fact complex clinical presentations known to vary as the individual undergoes development and the illness evolves. Failure to adopt a longitudinal perspective, and without taking into consideration other risk factors, our diagnoses are necessarily subject to change. Furthermore, as mentioned earlier, there is substantial heterogeneity within individual diagnostic constructs. For example, evidence points to the existence of several bipolar ‘subtypes’ ranging from psychotic bipolar disorder through classical episodic manic-depressive illness to soft-spectrum bipolar disorder thought to be characterized by chronic mood instability (Grof et al., 2009). These subtypes are associated with different clusters or spectra of disorders segregating in relatives, clinical course, treatment response, comorbidities and neural correlates. This is a problem because it suggests that even well-designed and properly executed prospective longitudinal studies will not be able to successfully map prototypical bipolar subtype trajectories. High-risk studies will need to go back further, tracing the origins of illness starting from well-characterized bipolar parents. This is important and necessary because as in other areas of medicine mapping complex disease (i.e. dementia, cancer), it is likely that each kind of bipolar illness trajectory is associated with differing and possibly quite separate etiological pathways, biomarkers and treatment targets. Nevertheless, this remains a worthwhile endeavor because once prototypical trajectories are mapped, these can be meaningfully compared to the trajectories of other illnesses and intermediate phenotypes identified.

Thus, in psychiatry, we are at critical crossroads. There is reasonable agreement that our current diagnostic approach does not inform risk prediction, support biomarker research or make possible individualized treatment. But instead of taking stock and devising methods to advance, we seem to be moving backward and echoing a previous well-worn (although never fully, nor indeed satisfactorily, resolved) argument between ‘lumpers’ and ‘splitters’. That is, there are those who say that given the substantial overlap and clinical heterogeneity within many non-classical presentations, we should consider a unitary clinical course or staging model that accommodates both mood and psychotic disorders versus those who argue for an alternative approach that has been used productively in other medical disciplines to differentiate between illnesses and illness subtypes using all validated predictors (i.e. longitudinal course, family history, treatment response, quality of remission and any available biomarkers). The forum of the current debate is within the concept of staging—in particular, its application to psychiatric disorders. In recent years, the staging of bipolar disorder has emerged as a particularly hot topic. But even attempting to apply staging to bipolar disorder seems somewhat premature because much of the prerequisite groundwork—for example, defining and differentiating more homogeneous illness trajectories within the heterogeneous bipolar disorder construct—is still very much a work in progress.

In a recent issue of the Australian and New Zealand Journal of Psychiatry, Pfennig and colleagues summarized key findings in a narrative review of prospective studies to help advance our understanding of the early clinical course of evolving bipolar disorder. Given the intensified research in this area over the past few years, such a review is certainly timely. In their review, the authors consider studies of different populations and offspring of bipolar parents and community cohorts and include samples clinically at risk for psychosis. The review aggregates findings across these groups at variable risk, age, sex proportion and stage of illness. This approach can be highly informative if these differences are considered when interpreting the findings, but it is important to keep in mind that the identified antecedents are only useful predictors in the specific population studied (i.e. those at familial risk). Based on their review of the literature, the authors discuss a proposed staging model of bipolar disorder—translated directly from studies of youth at clinical risk for developing psychosis. However, as highlighted earlier, this is not an approach that has been applied consistently, and it should be borne in mind that its validity and utility are still very much under debate. Subsuming two broad heterogeneous diagnostic constructs, such as mood and psychotic disorders, under one illness model dilutes findings specific to individual illness trajectories and risks losing any potential signals among noise. Thus, although we broadly support the concept of mapping illness trajectories with a view to ultimately staging bipolar disorder, we feel that the field as a whole, and in particular researchers targeting bipolar disorder, should take careful aim before ‘firing’ and that much more reconnaissance is needed in order to fully appreciate the bipolar landscape.