Bipolar disorder (BD) is a highly severe and chronic mental condition with a lifetime prevalence of 2–4% for its most common subtypes (i.e., type I/II). Given the chronic and episodic course of the disease the prediction of recurrent manic and depressive episodes is a major, though not adequately resolved clinical and research question. On a phenomenological level the disease is characterised by phases of (hypo)mania, a state of elevated mood, increased energy, risk-taking and reduced sleep and phases of depression, best described by feelings of sadness, hopelessness, loss of energy and reduced sensitivity to positive outcomes. It has been proposed that increased emotional reactivity, deficient emotion regulation and impulse control as well as motivational dysregulation are important mechanisms underlying these symptoms (Wessa et al. Reference Wessa, Kanske and Linke2014).

BD is considered among the ten most important causes of disease burden worldwide (Whiteford et al. Reference Whiteford, Degenhardt, Rehm, Baxter, Ferrari, Erskine, Charlson, Norman, Flaxman, Johns, Burstein, Murray and Vos2013) and has been judged the most expensive behavioural health care diagnosis. Hence, it is an important finding that the number of previous episodes and hospitalisations due to full-blown manic and depressive episodes predicts poor functional outcome, more severe cognitive impairments and worse prognosis, as indicated by higher rates of suicidal behaviour (Table 1, Ref. 1, 2). Therefore targeted and tailored improved prevention with respect to relapse into illness episodes would reduce the financial impact of this disease, decrease illness burden for patients and their families and improve patients’ social, occupational and cognitive functioning.

Table 1. Example studies of predictors for and early signs of bipolar illness episodes

NAP = non-affective psychosis; HC = healthy controls.

The research on inter-episode prodromal symptoms in BD is half a century old and ever since, a few predictors and a number of consistent early signs of recurrent episodes have been identified. Most robustly, relapse is predicted by significant life events (Christensen et al., Reference Christensen, Gjerris, Larsen, Bendtsen, Larsen, Rolff, Ring and Schaumburg2003), but social rhythm disturbances have also been linked to the initiation of recurrent manic episodes (Levenson et al. Reference Levenson, Nusslock and Frank2013). Further, previous research suggests several early predictive signs for the onset of mania (e.g., reduced need for sleep, increased energy, racing thoughts, overspending and distractibility), depression (e.g., sadness; slowness of movements, talking less and loss of energy) or both illness phases (mood liability, irritability) (Table 1, Ref. 5).

Most of these studies are focused on retrospective self-report data, which imply two important methodological limitations: (1) recall biases and (2) restriction of introspective abilities. To prevent recall biases, real-time reports of symptoms by so-called ‘ambulatory assessment (AA) methods’ have been proposed to prospectively capture the dynamics of symptoms in chronic mental disorders (Trull & Ebner-Priemer, Reference Trull and Ebner-Priemer2013). In the long run, their application is therefore considered worthwhile in predicting recurrent symptoms and preventing full-blown episodes. In BD research, a handful of studies used such AA methods, suggesting increased intra-individual variability particularly in positive affect and self-esteem measures in BD patients, which distinguished these patients not only from healthy controls but also from patients with major depressive disorder (MDD) (Table 1, Ref. 6). Such increased variability could also explain seemingly contradictory findings of both generally reduced positive affect and greater positive emotionality in BD patients compared with controls (Table 1, Ref.7, 8). Further, BD patients showed a marked decrease in positive affect when stressed whereas MDD patients displayed a significant increase of negative affect to stress (Table 1, Ref. 9). These studies provide an important insight into the dynamics of symptoms in BD. However, so far no published AA study has used a prospective design relating to these dynamics recurrence of affective episodes.

Further, we might overcome the issue of limited capability to introspection mentioned above using cognitive, neuropsychological or even neurobiological measures as predictors of recurrent episodes. This has already been done in two studies revealing neuropsychological measures that predict functional outcome or recovery in BD patients (Table 1, Ref. 3). Moreover, the authors showed that patients with impairments in at least one cognitive domain had higher risk to experience a recurrent episode in a shorter time interval (Table 1, Ref. 4). Based on these results, the combination of neuropsychological measures and (repeated) ambulatory assessment appears extremely promising to detect mechanisms predicting the switch into an affective episode.

On a neurobiological level, aberrancies in prefrontal–limbic–striatal networks have been proposed to be the correlate of various aetiological mechanisms of the disease (Phillips & Swartz, Reference Phillips and Swartz2014), such as disturbed emotional reactivity, impaired emotion regulation (Wessa et al. Reference Wessa, Kanske and Linke2014) and impulse control (Swann et al. Reference Swann, Lijffijt, Lane, Steinberg and Moeller2009). However, longitudinal data concerning these neural alterations capturing the dynamics of BD is rare. The few existing studies indicate decreased prefrontal activation and/or decreased negative connectivity between prefrontal and limbic brain regions during elevated mood (e.g., Strakowski et al. Reference Strakowski, Eliassen, Lamy, Cerullo, Allendorfer, Madore, Lee, Welge, DelBello, Fleck and Adler2011; Cerullo et al. Reference Cerullo, Fleck, Eliassen, Smith, DelBello, Adler and Strakowski2012), whereas depression is associated with increased activity and connectivity between brain regions mediating emotional appraisal such as insula and amygdala (e.g., Cerullo et al. Reference Cerullo, Fleck, Eliassen, Smith, DelBello, Adler and Strakowski2012). This pattern is also in line with the assumption that deficits in impulse control known to be mediated by prefrontal brain structures that represent a main feature of mania although increased impulsivity has also been observed across illness phases (e.g., Swann et al. Reference Swann, Lijffijt, Lane, Steinberg and Moeller2009). To this end, a very recent longitudinal study investigated a group of bipolar patients in euthymic, manic and depressed states. Here, independent of the symptomatic state, all patients showed decreased activation of the cognitive control network during an emotional conflict task (Rey et al. Reference Rey, Desseilles, Favre, Dayer, Piguet, Aubry and Vuilleumier2014), with a stronger decrease during mania. During euthymia, this decreased response to conflict was only observed during difficult as compared to all task levels in manic and depressed patients.

The present findings underline that we currently face a gap in BD research that only begins to consider the dynamics of recurrent episodes. ‘Since the biology, as the symptoms, fluctuates in time’ a paradigm shift towards neurobiological assessment at multiple time points over the course of the disease is indispensible, as pointed out in a recent editorial of this journal (Vol. 23, p. 211). Despite considerable challenges when investigating large cohorts of patients over the course of bipolar illness, future research should adopt prospective study designs including behavioural and neuroimaging measures underlying cognitive, emotional and motivational deficits in BD.