Cognitive heterogeneity is a key predictor of differential functional outcome in patients with bipolar disorder

Bipolar disorder BD is an exceptionally complex illness. Although it is defined by its episodic course, with periods of symptomatic recovery, these periods of relative remission are often characterized by subthreshold affective symptoms and incomplete functional recovery(Sanchez-Moreno et al., 2009). While it is true that some BD patients function very well throughout their lives, others struggle with significant disability. The vast range in functioning is paralleled by heterogeneous clinical course (e.g. differential recurrence rates, non-uniform treatment response). There are recognized clinical subtypes (BD I vs BD II, psychotic vs non-psychotic), which reflect historical attempts to address diagnostic and clinical heterogeneity; however, these subtypes are neither based on neurobiological evidence nor do they provide a roadmap for predicting outcomes(American Psychiatric Association, 2013).

Beyond the subthreshold affective symptoms that are common in remitted BD patients, among the most persistent symptoms are cognitive deficits. At the group level, cognitive deficits are qualitatively similar to those seen in schizophrenia (SZ), albeit less severe(Burdick et al., 2011); however, group-level comparisons inherently fail to account for the substantial heterogeneity in cognitive profiles within the disorder. Approximately 30–50% of euthymic BD patients present as “neuropsychologically normal”(Martino et al., 2008), while an equal proportion are as profoundly impaired as patients with a diagnosis of SZ. This heterogeneity has hindered our ability to identify causal mechanisms by neglecting important differences at the level of the individual patient. Importantly, cognitive deficits have been consistently shown to have a deleterious impact on clinical and functional outcomes (Depp et al., 2012), making them a high priority intervention target(Burdick et al., 2007). It has become increasingly clear that it will be necessary to empirically parse the disease heterogeneity if we are to effectively identity the modifiable clinical and biological risk factors that contribute to poor outcome. If successful, this approach will provide the field with potential therapeutic mechanistic targets for future intervention and prevention efforts designed to promote full recovery.

An example: We have successfully applied alternative classification approaches (e.g. cluster analyses) using cognitive performance, an objective measure of brain function, as a classifier. We identified three cognitively-homogeneous, discrete BD subtypes, equally distributed as 1) cognitively intact; 2) selectively impaired; and 3) globally impaired. These subtypes do not simply recapitulate existing clinical subtypes (e.g. BD I vs II) but represent a novel classification(Burdick et al., 2014). Our work has been replicated and across several studies, these cognitively-derived subtypes are strongly associated with level of community functioning(Green et al., 2020). Cross-sectionally, we have shown that these cognitively-derived subtypes differ on key clinical features (e.g. premorbid IQ, episode recurrence rate, sleep disruption, childhood trauma)(Burdick et al., 2014) and the degree to which genetic risk for disease contributes to outcome(Russo et al., 2017), pointing to subtype-specific risk factors for poor outcome, many of which are modifiable. Based on these findings and convergent data, we hypothesize that there is a subgroup of patients with BD who have a course of illness characterized by early-life risk factors and neurodevelopmental abnormalities which contribute to poor cognitive outcomes, while another subgroup is marked by a decline in cognitive functioning that is a more direct result of the expression of the disease and occurs after its onset(Millett and Burdick, 2021). In support of this hypothesis, using proxy measures of cognitive decline in this cross-sectional cohort, we find that the 3 subtypes have differential cognitive trajectories and varying levels of peripheral inflammation (C reactive protein and tumor necrosis factor levels), supporting the hypothesis that BD is characterized by a neuroprogressive course but only in some patients. In our hands, approximately 2/3 of BD patients evidence some decline, whilst the other 1/3 show resilience to this progression. These data further suggest that this cognitively-driven classification approach may help to identify those at greatest risk for a declining course of illness, most directly relevant to cognitive decline. Although cognitive decline is correlated with functional decline, a parallel approach has identified subtypes within BD that were derived using measures of functional outcome (rather than cognitive performance). This study showed three distinct functional profiles and reported that the strongest predictors of functional disability subgroup status were subthreshold affective symptoms and cognitive impairment(Solé et al., 2018).

Of note, almost all of the studies conducted to date that have utilized this type of classification approach are based upon cross-sectional data, including ours. There is a surprising dearth of longitudinal data in BD, in general, with even fewer studies including the key outcomes that contribute directly to patient’s quality of life (e.g., cognition, community function). A rare exception is a recently published study that showed three distinct functional trajectories in a large cohort of patients with BD, with two of the three subgroups evidencing persistent functional impairment over time. Although measures of cognition were not included in this analysis, several factors predicted a more severe functional disability versus milder impairment including: a higher rate of prior hospitalizations, childhood maltreatment, subthreshold depressive symptoms, sleep disturbances, an elevated body mass index, and more psychotropic medications at baseline(Godin et al., 2020). Additional studies of this nature are sorely needed to confirm and expand upon the promising results from this and the more numerous cross-sectional analyses.

By identifying the subtype of patients who are at greatest risk for poor functional outcomes and elucidating the specific, modifiable risk factors that contribute to disability in those patients we allow for both a) the development of novel biological interventions to prevent disability; and b) the proper allocation of the necessary, but unfortunately limited, extra resources (e.g. wrap-around social support) to those with the greatest need. Future studies are necessary to validate and expand upon this work in larger cohorts using a longitudinal design.