Predicting bipolar disorder: Vitally important,clinically difficult

From any viewpoint – health, quality of life and economically – it is axiomatic that prevention is a more worthy medical/clinical goal compared to treatment. This is certainly true for major psychiatric disorders such as bipolar disorder. Thus, identifying those individuals likely to become bipolar before the illness emerges in full form would be an enormous boon to patients, their families, and physicians alike. That, of course, is the hope and promise of the enormous amount of current work attempting to identify the genes that are associated with mood disorders in general and bipolar disorder specifically. However, unless and until genetic studies allow us to predict with some accuracy which individuals are likely to become bipolar, we must use the tools currently available. Beyond family history, these include the many variables reviewed in the paper of Pfenning et al. (2017) in this month’s journal.

As expected, the authors describe a literature that is far too meager given the topic’s importance. They conclude with the equally expected conclusion that far more data are required before any clinical features and endophenotypes such as characteristic cognitive abnormalities and laboratory (e.g. neuroimaging findings) predictors can be considered to be clinically useful. The clinical features that seem to be associated with the later emergence of bipolar disorder are childhood anxiety, sleep problems, and subthreshold hypomanic symptoms.

Conceptually, one can divide clinical predictors into homotypic and heterotypic subtypes. Homotypic predictors are descriptively related to the later disorder. Thus, subsyndromal hypomanic symptoms predict the later emergence of bipolar disorder. Yet, this simply describes the frequent clinical observation across all medical specialties that mild forms of a disorder often emerge before more severe forms. In medicine, angina may precede and ‘predict’ a later myocardial infarction. As Pfennig et al. acknowledge, these findings confound predictors with outcomes—those with mild bipolar disorder are simply likely to develop a more severe version of the same disorder later.

Heterotypic predictors are those that are at least somewhat dissimilar to the later disorder. For bipolar disorder, a history of childhood anxiety is phenomenologically different than the core symptoms of bipolar disorder and might be an important clinical predictor in a high-risk population (e.g. children of a bipolar parent). The problem is that many of these heterotypic symptoms are diagnostically nonspecific and also predict other forms of psychopathology other than bipolar disorder. As an example, a recent study found that irritability, fear, and/or anxiety predicted the development of unipolar depression in a cohort of high-risk children and adolescents (Rice et al., 2017). To be sure, with a longer follow-up (only 4 years in the Rice et al. study), it may be that many of these individuals might later develop bipolar disorder. For now, though, childhood anxiety and sleep difficulties may simply be markers/predictors of psychopathology in general.

Although Pfennig et al. review cognitive test results as predictors, they did not include two relevant papers which examined premorbid intellectual functioning as predictors of bipolar disorder (Reichenberg et al., 2002; Tiihonen et al., 2005). In one study, those conscripted into the Finnish Defense forces who on follow-up (mean = 7.1 years) had developed bipolar disorder showed poorer performance on visuospatial reasoning tests (Tiihonen et al., 2005). Consistent with the nonspecific nature of predictors, these findings also predicted other psychotic disorders such as schizophrenia. Surprisingly, in this study, higher scores on arithmetic reasoning also predicted bipolar disorder. In contrast, using data from the Israeli Draft Board registry, no differences were found in intellectual tests in recruits between those who later developed nonpsychotic bipolar in a follow-up up to 11 years compared to a control population (Reichenberg et al., 2002). Whether the discrepant findings reflect the nonpsychotic nature of the Israeli bipolar cohort (which may therefore have selected for a milder patient population) or the specific cognitive tests administered is unclear.

A single case example that illustrates the nonspecific nature of psychopathology predictors is that of Robert Lowell, one of the towering figures of 20th century poetry who had severe bipolar I disorder with multiple hospitalizations for psychotic mania (and multiple courses of electroconvulsive therapy (ECT) to treat his manias in the pre-lithium era). In the recent biography of Lowell, highlighting the complex interactions between his poetry and his bipolar disorder, Jamison (2017) describes some of Lowell’s childhood traits-long before his first manic episode—as thuggish, restless, wild, brash, irritable, and contrary. In today’s world, he might be described as having had oppositional defiant disorder (ODD). It is hard not to believe that his childhood mood dysregulation was unrelated to his later bipolar disorder. Yet equally clearly, these same traits are seen in the majority of ODD children who do not grow up to be bipolar.

Thus, for now, the predictors described in the paper by Pfennig et al. should be considered generic markers of psychopathology rather than specific antecedents of bipolar disorder.