Keywords

1 Introduction

Determining the physiological correlates of psychiatric disorders and assessing the potential clinical utility of physiological measures are major pursuits in biological psychiatry research (Insel et al. 2010, 2013). The impetus for such research is the promise of objective measures for assessment and predicting response to treatments, which would improve diagnostic validity and inform treatment selection (Insel et al. 2010, 2013). To facilitate this goal, the National Institute of Mental Health has proposed Research Domain Criteria (or RDoC), which provide researchers with a novel framework in which research can be presented (Morris and Cuthbert 2012). In this context, physiological measures provide an important tool through which diagnosis could be improved and treatment options tailored. Currently, the assessment and management of bipolar spectrum disorders (BSDs) often entails misdiagnosis following a significant delay in detection (Hirschfeld et al. 2003; Suppes et al. 2001), the inability to predict course (Crowe et al. 2012; Malhi et al. 2012, 2013), trial and error in treatment selection (Malhi et al. 2009; Sachs 2013), and a failure to implement long-term management strategies (Keck 2006; Malhi et al. 2009). Therefore, improvement in the assessment and treatment of those with BSDs is critical given the great personal (Bonnín et al. 2012; Merikangas et al. 2007) and socioeconomic burden associated with BSDs (Merikangas et al. 2007). Critically, these disorders are associated with the highest risk of suicide of any mental disorder (Nock et al. 2009), highlighting the need for early and accurate detection with improved diagnosis and a more personalized approach to effective treatment. Initiation of successful treatment early in the course of the disorder will undoubtedly reduce morbidity (Baldessarini et al. 2003; Post et al. 2010) and improve treatment outcomes (Berk et al. 2011; Ketter et al. 2006; Malhi 2012). Given the potential for these needs to be met with the translation of physiological measures into clinical practice (Morris and Cuthbert 2012), identification of physiological markers that could be employed in assessment and treatment selection remains an ambitious but worthy goal.

BSDs represent a cluster of disorders characterized by extreme changes in mood (Malhi et al. 2012). Depression, mania, hypomania, and euthymia (periods of remission) are phases of illness that are subjectively experienced by patients, and objectively determined by clinicians (Tohen et al. 2009). These include cyclothymia, Bipolar I disorder, Bipolar II disorder, and Bipolar Disorder Not Elsewhere Classified (NEC) and are partitioned from major depression on the basis of cycling into mood elevation such as (hypo)mania. The spectrum is conceptualized as increasing in severity and burden from Bipolar Disorder NEC, Cyclothymia, Bipolar II Disorder, through to Bipolar I Disorder (Merikangas et al. 2007), but in reality, this does not always hold true. Currently, there are no physiological tests that can be employed to assist with the detection, assessment, and diagnosis of BSDs. The physiological correlates implicated in cognitive and emotional disturbances underlying BSDs and the different phases of illness have thus far been investigated using neuroimaging and peripheral physiology techniques. Studies have revealed disturbances in prefrontal and limbic network neutral activity (see Strakowski et al. 2012), neural activity states and early neural processing (see Degabriele and Lagopoulos 2009), and within the autonomic nervous system (ANS; Gruber et al. 2011; Lee et al. 2012), respectively. Understandably, most studies have considered these neural and autonomic activation characteristics separately. Hence we now consider the neural and autonomic characteristics of BSDs in the context of an embodied (see Craig 2009; Niedenthal 2007; Price et al. 2011) disturbance affecting both the brain and body so as to develop a novel framework (see Fig. 1) in which the physiological correlates of the disorder can be investigated.

In addition to characterizing the physiological correlates of the BSDs, we further characterize the correlates—or markers—of treatment effect, and predictors of response to treatment. Patients with BSDs are typically prescribed pharmacological treatment (including lithium, antipsychotics [e.g., olanzapine, risperidone, quetiapine], and anticonvulsants [e.g., valproate, carbazepine lamotrigine)] either alone or in combination) but first-line treatment is often ineffective (Malhi et al. 2009) reflecting the fact that information gained from clinical assessment alone is insufficient for planning and implementing treatment. If physiological measures could anticipate treatment efficacy, then the trial and error involved in first-line treatment strategies may be diminished or subsided altogether. Therefore, the need to identify potential physiological markers of BSDs and to ascertain their clinical utility is imperative.

2 Measuring the Physiological Correlates of Bipolar Spectrum Disorders

Studies on BSDs have most often utilized neuroimaging techniques such as functional magnetic resonance imaging (fMRI), positron emission tomography (PET), and single-photon emission computed tomography (SPECT) and findings from studies using these neuroimaging methods have been largely consistent (Strakowski et al. 2005). The next most employed measures have been electroencephalography (EEG) and event-related potential (ERP). While neuroimaging methods enable localization of regional responses to cognitive and emotional tasks across the whole brain (Friston et al. 1998), EEG and ERP have higher temporal resolution such that fluctuations in neural activity states (Davidson 1998, 2004) and early information processing can be examined (Donchin and Coles 1988), respectively. In terms of peripheral physiological measures, cardiovascular measures including heart rate (HR) and heart rate variability (HRV) gathered using electrocardiography (ECG), and galvanic skin response (GSR) have been widely employed. Within these, HR is a measure of overall ANS arousal (see Duschek et al. 2013; Lopes and White 2006) under tonic inhibitory control by the parasympathetic nervous system (PNS; Saul 1990; Thayer et al. 2009), high frequency HRV measures PNS activity (see Duschek et al. 2013; Lopes and White 2006), and galvanic skin response (GSR) measures reflect sympathetic nervous system (SNS) activity (see Dawson et al. 2007). Notably, only GSR has been employed in combination with other measures for multimodal investigations. Building toward a more embodied perspective of BSDs, we delineate, examine, and then integrate the cognitive neuropsychological (brain; fMRI and EEG, and ERP) “neurocorrelates” and the peripheral physiological (body; HR and HRV, and GSR) correlates.

A methodological issue that has been discussed in most physiological studies of BSDs is heterogeneity with respect to previous and current treatment effects, clinical course, and comorbidities. Though it is often not feasible to exclude or stratify patients on all these bases, future research should report and, where possible, take these factors into account. Another self-evident issue that has challenged researchers is the difficulty in recruiting and testing manic bipolar patients in laboratory settings, explaining why there are relatively fewer studies of this phase of illness (Small et al. 1999). With technological advances, future research will likely take advantage of less invasive, ambulatory sensors including smartphone (Heathers 2013) and sensorized clothing (e.g., Mariani et al. 2012; Quintana et al. 2012; Siegel 2013), which will enable the collection of longitudinal physiological data across illness and treatment phases within BSD patients.

3 Neurocorrelates of Bipolar Spectrum Disorders and Their Treatment

3.1 Functional Neuroimaging Correlates

3.1.1 Characterization and Differentiation

3.1.1.1 Characterization

Characterization of the functional neuroanatomy of BSDs has been extensive, systematic, and consistent for the last two decades (see Table 1). The results and conclusions obtained across neuroimaging modalities—including fMRI, PET, and SPECT—have been largely consistent and non-modality specific (see Strakowski et al. 2005). Overall, BSDs display prefrontal cortex (PFC) hypoactivity and limbic hyperactivity during emotional and cognitive tasks, and these findings correlate with trait and state emotional lability and mood disturbances in BSDs (see Strakowski et al. 2012). In addition, bipolar disorder is characterized by dysfunctional connectivity among ventral prefrontal networks and limbic brain regions, particularly the amygdala (Blond et al. 2012; Chen et al. 2011; Houenou et al. 2011; Strakowski et al. 2012; Townsend et al. 2012) indicating both difficulty in regulating mood alongside a dysfunction of emotion processing. Impaired PFC regulation subsequently leads to a loss of neurological emotional homeostasis, emotional lability, and mood disturbances (Strakowski et al. 2012). It is posited that a disruption of frontal regulatory networks allows for extreme mood states, switching among mood states, and mixed states (Strakowski et al. 2012). These abnormalities have been conceptualized as dysfunction within oscillatory mechanisms, which perhaps worsen over time, and result in the many manifestations of the illness (Schneider et al. 2012).

Table 1 Meta-review of fMRI assessment in BSD characterization and differentiation
Full size table

Interestingly, bipolar patients have consistently decreased frontal activation across the ventrolateral PFC (VLPFC), a region critical for emotional processing and mood regulation (Blond et al. 2012; Chen et al. 2011; Houenou et al. 2011; Strakowski et al. 2012; Townsend and Altshuler 2012), and consistently decreased inferior frontal gryus (IFG) activity, specifically the right IFG (R IFG), a region associated with regulatory inhibition (Hajek et al. 2013). Furthermore, bipolar patients have also increased activity within limbic regions including parahippocampus, hippocampus, and amygdala and basal ganglia (Blond et al. 2012; Chen et al. 2011; Houenou et al. 2011; Strakowski et al. 2012; Townsend and Altshuler 2012), which may underpin abnormalities of primary emotion processing. Hence, decreased IFG activity in bipolar disorder which is seen during both cognitive and emotional processing, and increased limbic activation that is seen during emotional processing (Chen et al. 2011; Hajek et al. 2013; Houenou et al. 2011; Strakowski et al. 2012) may reflect trait-based correlates of BSDs.

With respect to the manic phase of bipolar disorder, IFG activity is decreased in mania but not in euthymic or depressed states, and limbic activation increases are not associated with mood states (Chen et al. 2011; Houenou et al. 2011; Strakowski et al. 2012). However, amygdala activation varies as a function of mood state and the valence of the emotional stimuli: hyperactivity to emotional stimuli in mania; hyperactivity to negative stimuli and hypoactivity to positive stimuli in depression, and normalized activations in euthymia (Townsend and Altshuler 2012). However, it is important to note that many of these findings are preliminary and may be contingent on additional factors such as the tasks used and disorder phenotype, but they do suggest that manic and depressed phases of bipolar disorder can be differentiated on the basis of altered IFG activity and valence-mood congruent activation of the amygdala (Strakowski et al. 2012; Townsend and Altshuler 2012). With respect to response inhibition, the manic phase is associated with reduced performance, associated with decreased R IFG and medial frontal gyrus (MFG) activation and increased bilateral basal ganglia activation (Hajek et al. 2013; Houenou et al. 2011; Strakowski et al. 2012). In the euthymic phase, response inhibition is not dysfunctional, although activity in left superior temporal and right MFG is increased and basal ganglia activation decreased (Hajek et al. 2013; Houenou et al. 2011; Strakowski et al. 2012). Therefore, euthymic patients compensate for reduced inhibitory IFG activity with increased activation of adjacent cortical areas, thereby yielding normalized inhibitory functions (Hajek et al. 2013). During euthymia, recovery of frontal control, along with compensation from other brain regions, temporarily restores neurological emotional homeostasis (Strakowski et al. 2012). However, the underlying functional abnormalities in the VLPFC networks leave the risk for emotional and cognitive disruption, leading to manic, depressed, or mixed phases, even under minor stress (Strakowski et al. 2012).

3.1.1.2 Differentiation

Over-activation in the medial temporal lobe during tasks involving emotion or memory may differentiate patients with bipolar disorder from patients with schizophrenia (Whalley et al. 2012). However, differential diagnosis with fMRI has been less accurate with bipolar disorder than schizophrenia (Whalley et al. 2012). While promising preliminary findings have been reported suggesting that BSDs may be distinguished from unipolar major depression in small samples (Diler et al. 2013; Grotegerd et al. 2013; Marchand et al. 2013), large-scale studies are needed to determine the sensitivity and specificity of these findings.

3.1.2 Treatment

Effective treatment would be expected to normalize the state-based and trait-based VLPFC-limbic network disturbances correlated with BSDs using fMRI. In comparison to the fMRI correlates characterizing and differentiating BSDs, fMRI correlates of treatment are understudied. In the last decade, eight controlled studies examining the impacts of treatment administration in BSDs on cognitive and emotional stimuli have been published, with each demonstrating some sort of normalization effect (Table 2). Lithium appears to have prophylactic effects on cognition after 14 days’ treatment and acts on frontal regions in the euthymic phase of BSDs with little impact during the depressed phase (Silverstone et al. 2005). After 12 weeks of lamotrigine administration during the euthymic phase, there are increases in the prefrontal and cingulate regions, thereby normalizing the activity of circuitry involved in emotion regulation (Haldane et al. 2008; Jogia et al. 2008). In the depressed phase, 8 weeks of lamotrigine administration reduces amygdala reactivity to negative stimuli, with greater reductions in reactivity being correlated with reductions in depression symptoms after 8 weeks (Chang et al. 2008). When patients are given a 4-week course of antipsychotics and then a 14-week course of lamotrigine, decreases in mania symptoms following treatment are associated with increased VLPFC and dorsolateral (DLPFC) activity during cognitive-emotional (Pavuluri et al. 2010b) and response inhibition tasks (Pavuluri et al. 2010a). In subsyndromal patients, there are no consistent differences after 12 weeks of valproate treatment (Chang et al. 2009). Finally, a study investigating the effect of psychotherapy showed normalization of IFG hypoactivity after 12 weekly sessions (Favre et al. 2013); however, it was difficult to differentiate the effect of psychotherapy from improvement with the medication patients were already receiving (Favre et al. 2013).

Table 2 Literature review findings of fMRI assessment in BSD treatment
Full size table

In sum, there appears to be normalization of the cognitive and emotional neural networks implicated in BSDs with treatment when patients present in the depressed or manic phases of illness. Additionally, prophylactic treatment appears to affect these networks. However, the majority of investigations thus far have small sample sizes (less than 20 patients) and many did not have a control group for comparison (e.g., Chang et al. 2008; Haldane et al. 2008). Additionally, some studies investigated adolescents (e.g., Chang et al. 2008, 2009; Pavuluri et al. 2010a, b), who are likely to have fundamentally different responses compared to adults. Nevertheless, there are promising findings from this relatively new line of research that should encourage future research with larger samples, across different treatments. In doing so, clinically useful fMRI treatment markers for predicting treatment response and treatment monitoring may be determined.

3.2 EEG and ERP Correlates

3.2.1 Characterization and Differentiation

Within the EEG and ERP literature, there has been one systematic review on the correlates that characterize and differentiate BSDs (Degabriele and Lagopoulos 2009). Here, we updated this review (Table 3). There have been 22 studies examining the BSD characteristics with EEG and ERP measures. Overall, there appears to be measurable correlates in frequency band, ERP component, and sleep EEG characteristics (see Degabriele and Lagopoulos 2009), and network properties (Kam et al. 2013; Kim et al. 2013) that can characterize and differentiate BSD phases and BSDs from unipolar depression and schizophrenia. Single electrode EEG (see Iacono et al. 1983), clinical EEGs (see Cook et al. 1986; Small et al. 1999), and associating EEG data with neuroanatomical abnormalities from computerized tomography images (see Dewan et al. 1988) are yet to produce in measurable correlates characterizing and differentiating BSDs.

Table 3 Literature review findings of EEG assessment in BSD characterization and differentiation
Full size table

BSD studies employing EEG data show that differential power at specific frequency bands, which are associated with different activity states, are correlated with traits and states of BSDs. Specifically, differential activity in the alpha band between the frontal lobe hemispheres, frontal alpha asymmetry—an index associated with behavioral motivation (Davidson 1998, 2004)—correlates with BSD phases. In the depressed phase, increased right-dominant, withdrawal-related, frontal alpha asymmetry, relative to controls, is characteristic at rest (Nusslock et al. 2012). Additionally, bipolar disorder patients with decreased functional network integration and decreased optimal balance of network segregation in functional fronto-central and centro-parietal networks had higher depression scores (see Kim et al. 2013). In the manic phase, various frequency characteristics can be observed during rest (see Clementz et al. 1994; Kano et al. 1992), with increased ‘busy thinking’ related, beta activity correlating with increased mania symptoms (Kam et al. 2013). Increased left-dominant, goal striving frontal alpha asymmetry, relative to controls, appears to be characteristic of mania (Harmon-Jones et al. 2008; Nusslock et al. 2012), opposing the activity characterizing the depressed phase. In the hypomanic phase, increased left-dominant frontal alpha asymmetry at rest is also observed (Harmon-Jones et al. 2008; Nusslock et al. 2012), an effect that also correlates with hypomanic personality (Peterson and Harmon-Jones 2008; Wyczesany et al. 2010). In the euthymic phase, bipolar patients appear to have more normalized frequency characteristics and frontal alpha asymmetry relative to controls (Nusslock et al. 2012), though some residual frequency characteristics remain suggesting some trait-based cognitive dysfunction (El-Badri et al. 2001).

ERP components, such as the commonly reported P300, provide an opportunity to determine whether early information-processing is impaired in patient samples (see Degabriele and Lagopoulos 2009; Donchin and Coles 1988; Kemp et al. 2009). In studies with bipolar patients in no specific phase, results show ERP component differences characteristic of disturbance in early executive functions (Hall et al. 2007; Muir et al. 1991; O’Donnell et al. 2004; Souza et al. 1995) and accentuation of the early processing of positive stimuli (Degabriele et al. 2011). Furthermore, some ERP components appear to be heritable, endophenotypes for bipolar disorder (Hall et al. 2007). Therefore, early processing deficits appear to be measureable BSD traits.

Clinical course from unipolar to bipolar disorder has been predicted using EEG sleep components (Rao et al. 2002), a finding that may be associated with increasing or differential chronobiological disturbances in BSDs (see Malhi and Kuiper 2013). Bipolar I conversion from cyclothymia and bipolar II is reliably classified by increased manic-related, left-dominant frontal asymmetry at rest (Nusslock et al. 2012). Finally, bipolar disorder can be differentiated from unipolar disorder with specific ERP components (Muir et al. 1991) and network properties (Koles et al. 1994), and from schizophrenia with specific ERP components (O’Donnell et al. 2004; Souza et al. 1995) and network properties (Kam et al. 2013).

In summary, there are identifiable EEG and ERP characteristics that correlate with the states and symptoms of depression and mania that may differentiate the two poles of the illness: withdrawal or negative valence-related right-hemispheric dominance for the depressed phase and approach or positive valence-related left-hemispheric dominance for the manic phase, each related to changes in network properties. In addition, frequency band and early processing disturbances consistently appear to be trait-based characteristics of bipolar disorder, which can differentiate it from unipolar disorder. Furthermore, early processing and network disturbances differentiate bipolar disorder from schizophrenia. Given that classification and predicting clinical course using EEG has been successfully examined, future directions in EEG and ERP research should concern classification of different phases and differential diagnosis using the aforementioned characteristics.

3.3 Treatment

Within the EEG and ERP literature, there has been one review examining correlates of lithium treatment effect (Ikeda and Kato 2003). Here, we updated this review and review other treatments (Table 4). The EEG and ERP characteristics that correlate with the phases and symptoms and trait-based early processing characteristics of BSDs would be expected to normalize with effective treatment. Indeed, all six studies reviewed show that EEG and ERP can detect treatment-related changes with commonly prescribed treatment. Although clinical EEG did not appear to be a useful measure for characterizing BSDs, it appears that existing EEG abnormalities (including, spikes and irregular beta, theta, and slow alpha activity) is a predictor of 3-month lithium treatment non-response (Ikeda et al. 2002). Additionally, clinical EEG and ERP changes with anticonvulsant medication predict treatment response (Gerez and Tello 1992).

Table 4 Literature review findings of EEG assessment in BSD treatment
Full size table

On the one hand, studies (Schulz et al. 2000; Small et al. 1989, 1998) that show frequency band component changes with treatment do not interpret these changes in the context of changes in neural activity or cognitive-emotional processing, correlating with normalization of symptoms. Instead, these studies discuss the potential utility of frequency band component changes as a tool to monitor and measure treatment responses. On the other hand, studies (Howells et al. 2012; Schulz et al. 2000; Small et al. 1989, 1998) showing laterality effects suggest that the greater left-dominant frontal approach-related activity characterizing mania may be normalized in treatment responders, relative to treatment non-responders. After 20 weeks of lithium treatment, relative alpha power in the right centro-parietal region is decreased (Schulz et al. 2000). Lithium, carbamazepine, and risperidone treatment non-response are correlated with higher left fronto-temporal amplitudes than responders in the fast delta, theta, and beta bands at baseline (Small et al. 1998). After lithium treatment, beta1 and left delta, theta, and beta2 increase, and treatment response correlates with increases in delta (Small et al. 1989). Lithium plasma level is correlated with increased theta power.

After carbamazepine administration, delta activity in the anterior regions is increased, with more right-sided increases, and theta is decreased (Small et al. 1989). Mindfulness-based cognitive therapy appears to decrease right frontal beta at rest and normalize P300-like ERP components in already medicated euthymic patients (Howells et al. 2012). Although the specificity of changes due to the therapy is uncertain, this was interpreted as improvement in attentional readiness and attenuation of non-relevant information processing (Howells et al. 2012). Although network disturbances appear to characterize and differentiate BSDs (as described in the previous section), the impacts of treatment on these networks are yet to be investigated.

Overall, there appears to be measurable EEG and ERP correlates of general treatment response that normalize phase and symptom-based characteristics of BSDs; however, the specificity of these effects to a particular medication or phase remains uncertain. Future directions would be to examine the specificity of medication effects and consequent treatment responses on EEG, ERP components, and network properties at each illness phase. Additionally, the relationships between frequency band component changes and cognitive-emotional changes with treatment should be determined. These developments would lay the foundations for investigation into the clinical utility of EEG and ERP components as markers of treatment response in BSDs.

4 Peripheral Physiological Correlates of Bipolar Spectrum Disorders and their Treatment

4.1 Cardiovascular Correlates

4.1.1 Characterization and Differentiation

A systematic literature review of cardiovascular correlates in BSD is presented in Table 5. Eight studies have examined cardiovascular measures and suggest that BSDs are associated with a higher heart rate (HR), reflecting increased arousal and reduced PNS function (see Duschek et al. 2013; Lopes and White 2006) can characterize BSDs. This is important as high resting HR is associated with an increased risk of suicide (Lemogne et al. 2011), which may be related to high suicidality in BSDs (Nock et al. 2009). Studies further suggest that lower HRV can characterize BSDs under tonic and phasic conditions. High frequency HRV measures reflects activity within the PNS branch of the ANS (see Duschek et al. 2013; Lopes and White 2006). PNS activity at the heart during emotional responding is associated with engagement of executive PFC control on the limbic system, and thus afferent and efferent brainstem nuclei linked to the heart (Duschek et al. 2013). Studies of BSDs suggest that low HRV are correlated with BSD traits. This is important given that low HRV is associated with poor mental and physical health and psychological flexibility in the face of stress, increasing the risk of cardiovascular disease and mental disorder, and overall morbidity and mortality (see Duschek et al. 2013; Kemp and Quintana 2013).

Table 5 Literature review findings of cardiovascular assessment in BSD characterization and differentiation
Full size table

Studies suggest that euthymic patients have higher HR (Iacono et al. 1983) and lower HRV (Cohen et al. 2003; Lee et al. 2012) at rest than controls, reflecting disturbed capacity to adapt and regulate autonomic arousal. Similarly, in manic (Henry et al. 2010) and subsyndromal (Lee et al. 2012) depressed bipolar patients, there appears to be increased HR and decreased HRV at rest, relative to controls. Furthermore, decreases in HRV appear to be related to both mania (Henry et al. 2010) and depression severity (Lee et al. 2012; Migliorini et al. 2011). In controls at risk for mania, however, there have been findings of increased HRV during emotional films, relative to controls at low risk (Gruber et al. 2008). There has been no comparison between, or within patients, differentiating BSD phases, except for one pilot study showing that HRV seemed to differ from controls when a patient was in a depressed state, rather than in a euthymic state (see Migliorini et al. 2011). Studies that compare bipolar to unipolar (Iacono et al. 1983) and schizophrenia (Henry et al. 2010) patients have not revealed any differential findings. However, many studies have encountered measurement problems—including poor consistency of findings across different HRV measures and short recording times—when compared against established guidelines (Task Force of the European Society of Cardiology the North American Society of Pacing Electrophysiology [Task Force] 1996). These issues have been highlighted in the “Comment” column in Table 5. Therefore, the results of these studies should be interpreted with caution. Given the resurgence of interest in this type of research, future directions for research into cardiovascular measures in the characterization and differentiation of BSDs should replicate findings across phases of illness between, and within, patients and other diagnostic groups, in accordance with established guidelines (Task Force 1996). Furthermore, given that cardiovascular measures are some of the least costly, time-consuming, invasive, and most mobile (see Heathers 2013; Mariani et al. 2012; Quintana et al. 2012) of commonly employed measures, the potential clinical utility of HR and HRV in BSD assessment and monitoring should be explored.

4.1.2 Treatment

A review of the cardiovascular literature focusing on the treatment of BSDs is presented in Table 6. The cardiovascular findings of the increased HR and decreased HRV characterizing BSD phases would expect to be normalized with effective treatment. For example, lithium is known to decrease depolarization at the sinoatrial node of the heart (Chong et al. 2001); thus, it could be predicted that lithium administration increases HRV, in the context of the wider literature (Duschek et al. 2013; see Lopes and White 2006). However, no study has directly investigated this possibility. Decreased depolarization at the sinoatrial node could be a direct effect of lithium or a more indirect, neural-ANS integration effect.

Table 6 Literature review findings of cardiovascular assessment in BSD treatment
Full size table

The impacts of commonly prescribed pharmacological treatment have been documented as manipulation checks in studies that characterize BSDs using cardiovascular measures. Additionally, there are some other potentially relevant findings from pharmacological HRV studies. Both these types of findings are provided in Table 6. Tricyclic antidepressants are known to decrease HRV in unipolar patients (Kemp et al. 2010) and in a mixed sample of bipolar and unipolar patients (Paclt et al. 2003). Promisingly, there appears to be some effect of lithium on HRV (Henry et al. 2010), consistent with decreased depolarization at the sinoatrial node. This finding should be followed up in trials that investigate the impact of lithium, along with other commonly prescribed medications, on tonic and phasic changes in HR and HRV across BSD phases and in healthy controls. The potential clinical utility of these HR and HRV measures for treatment monitoring in BSDs should be explored. This is important given cardiovascular measures provide insights into suicidality (Lemogne et al. 2011), health and wellbeing (Duschek et al. 2013; Kemp and Quintana 2013), and morbidity and mortality (Kemp and Quintana 2013).

4.2 GSR and Multimodal Correlates

4.2.1 Characterization and Differentiation

Here, we present the first review of multimodal methods to study the physiological correlates of BSDs (Table 7). If BSDs can be characterized and differentiated by embodied dysfunction in brain-body integrated systems, multimodal studies would be expected to show this dysfunction. GSR has been employed only in the context of multimodal physiology measurement. In contrast to the parasympathic HRV measures, GSR is used as a measure of tonic SNS activation at rest or during experimental tasks, whereby task-related phasic changes in skin conductance can be compared from a baseline (see Dawson et al. 2007 for full explanation). Increased magnitude of GSRs—associated with the strength of SNS activation—but not the number of responses has been shown in BSDs relative to controls (Iacono et al. 1983; Malhi et al. 2005). Given that studies employing GSR phasic responses in response to brief emotional stimuli could not make this differentiation (Gruber et al. 2008, 2011), GSR magnitude appears to be a trait marker of bipolar disorder. Though simultaneous GSR, EEG, and HR and HRV measures (Iacono et al. 1983)—along with measurement of positive facial emotion expression (Gruber et al. 2008, 2011)—have been utilized, results from these simultaneous measures have been largely inconsistent within studies and thus have not provided complementary information. However, this may be due to methodology (see “Comment” column in Table 7), including lack of spatial resolution using single electrode EEG (e.g., Iacono et al. 1983) and short recording times for HRV measures (e.g., Gruber et al. 2008, 2011).

Table 7 GSR and multimodal assessment in BSD characterization and differentiation
Full size table

With the improvement of MR-compatible GSR systems and techniques, preliminary research on BSD with simultaneous fMRI and GSR measurement in BSDs has been conducted (e.g., Malhi et al. 2005). This work suggests that bipolar patients may have cognitive deficits related to arousal and appraisal of emotional stimuli given simultaneous VLPFC hypoactivity and increased SNS activity during an emotional stroop task (Malhi et al. 2005). The future research should examine whether simultaneous and integrative neurological and peripheral physiological measurement provides further insight into BSDs. Should integrative measurements be employed, the neural responses when the central and peripheral ANSs (both PNS and SNS) are concurrently active could be measured (see Gray et al. 2009), and thus embodiment of cognitive and emotional processes can be observed.

With GSR-fMRI integration, for example, neural activity during periods of SNS activity can be partitioned (Gray et al. 2009). This is promising given that PFC activity—the disturbances of which are implicated in BSDs—is positively associated with GSR amplitude (Critchley et al. 2000). Future work will employ these methods given that BSDs may be characterized by a lack of inhibition of the VLPFC on increased amygdala activity which relates to increased SNS activity, which then may be related to the disengagement of the PFC regions involved in appraisal during emotional stimuli, in accordance with aforementioned neural-ANS integration accounts (see Damasio 1996; Porges 2007; Thayer et al. 2009) and preliminary findings (e.g., Malhi et al. 2005). In addition to neural-SNS integration, dysfunctional prefrontal engagement of the inhibitive PNS might be involved in BSDs. Thus, the future work in BSDs should explore the integration of the PNS with integrative measurement (e.g., fMRI-HRV measurement; see Thayer et al. 2012). Finally, in order to provide a complete account of the neutral-autonomic integration, longitudinal integrative measurement (e.g., fMRI-GSR-HRV; EEG-GSR-HRV) should be employed. A more complete embodied account and multi-modal measurement techniques may provide objective measures of the mechanisms through which emotional lability and mood disturbance occurs in BSDs. Consideration of an embodied psychophysiological mechanism, not just examining individual physiological correlates, may provide the clinically useful, objective physiological measures needed to characterize and differentiate BSDs.

4.2.2 Treatment

Although there have been studies that employ multimodal physiology to characterize BSDs, there are yet to be multimodal investigations of treatment effects and responses. Additionally, treatment studies employing GSR are yet to be conducted. Research has already employed simultaneous fMRI-GSR measurement (e.g., Malhi et al. 2005); thus, determining whether treatment normalizes these correlates is a likely next step. In doing so, researchers may be able to illustrate the manner in which treatment impacts the embodied neural-SNS integration of cognitive and emotional stimuli, which in turn may be related to improvement in depressed and manic symptoms, and prophylaxis in the euthymic phase.

5 An Embodied Framework for the Psychophysiology of Bipolar Spectrum Disorders

Here, we interpret the physiological correlates of BSDs in line with embodied, neural-autonomic integration perspectives (see Craig 2009; Niedenthal 2007), particularly with respect to the crucial role for the body in emotion, motivation, and cognition (see Price et al. 2011). Characterization of BSD as an embodied disturbance is gaining some attention with consideration of molecular biological correlates across circadian, homeostatic, and stress systems (see Malhi et al. 2012; Malhi and Kuiper 2013). A previous review of BSDs (Green et al. 2007) has discussed the potential impact of the neurological dysfunction on autonomic arousal systems; however, the exact disturbance remains to be determined. Initial support for an embodied perspective on BSD psychophysiological correlates originates from preliminary multimodal investigations (e.g., Malhi et al. 2005). These studies may provide the objective physiological measures needed to characterize and differentiate, and make treatment decisions in patients with BSDs.

After having provided overviews of the cardiovascular, EEG and ERP, fMRI, and multimodal literatures that characterize and differentiate BSDs, we incorporated these areas into a simplified, working embodied framework for the physiology of BSDs (see Fig. 1). Accordingly, we suggest that BSDs may be considered as a cluster of disorders characterized by dysfunctional embodied cognitive and emotional neural integration with the ANS, resulting in emotional lability and mood disturbances. In a simplified illustration, dysfunctional VLPFC connectivity between the amygdala impacts the neural-autonomic integration of SNS and PNS activity, through the brainstem, resulting in the core cognitive and emotional dysfunctions of BSDs, which is the target of treatment. Dysfunctional neural-autonomic integration will lead to decreased control over primary visceromotor activity and decreased flexibility for adaptation to stress: SNS hyperactivation and PNS hypoactivation. Dysregulated primary viscerosensory feedback from the periphery will then be inadequately regulated by hypoactivation of the medial frontal regions involved in the appraisal of the stimuli, resulting in emotional lability and mood disturbance characteristic of BSDs. It is perhaps through bidirectional projections between brain and body that underpin core cognitive and emotional dysfunction in BSDs: VLPFC may fail to regulate the amygdala effectively, increasing SNS activity through visceral efferent pathways, leading to a state of ANS rigidity, which subsequently impacts on the brain through afferent feedback, resulting in the core dysfunctions. An embodied approach to better understanding BSDs may, in time, provide clinically useful physiological measures needed to improve assessment and make treatment decisions in patients with BSDs.

Fig. 1
figure 1

Embodied framework for the psychophysiology of bipolar spectrum disorders (BSDs). This schematic provides a working framework, which enables the characterization of BSDs in an embodied manner. It details the dysfunctional neural-autonomic integration of cognitive and emotional processes and responses, which result in emotional lability and mood disturbances. Within this framework, this dysfunction would be normalized by effective treatment. With respect to noninvasive multimodal measurement of the physiological correlates, cortical neural activity can be measured (but not limited to the measures reviewed; see in red) using electroencephalogram EEG) ( and functional magnetic resonance imaging (fMRI), the temporal information processing stream from the cortex with event-related potential (ERP), subcortical neural activity with fMRI, sympathetic nervous system activity with galvanic skin response (GSR), and parasympathetic nervous system activity with high frequency heart rate variability (HRV)

Full size image

5.1 Novel Predictions Within the Embodied Framework

Within the simple, embodied framework for the physiology of BSDs, novel predictions regarding BSD psychophysiology arise. Within BSD patients during the euthymic phase in comparison to other phases, increased VLPFC activity would be associated with higher PNS and low SNS activity (e.g., greater HRV; lower GSR). However, in comparison to healthy controls, the characteristic decreased VLPFC activity would be associated with lower PNS and higher SNS activity (e.g., lower HRV; greater GSR). We further predict that neural and peripheral responses in response to positive, approach-related stimuli will differ from those in response to negative, withdrawal-related stimuli, and that these responses during mania will differ from those in the depressed phase. We predict that manic phase will be associated with decreased VLPFC and R IFG activity, in addition to increased left-dominant alpha asymmetry, lower PNS and higher SNS activity (e.g., lower HRV; higher GSR), and in turn medial PFC disengagement involved in appraisal of goal-oriented positive stimuli, relative to negative stimuli. By contrast, during the depressed phase, we predict the opposite: decreased VLPFC and R IFG, increased right-dominant alpha asymmetry, low PNS and higher SNS activity (e.g., lower HRV; higher GSR), and in turn medial PFC disengagement from appraisal of withdrawal-oriented negative stimuli, relative to positive stimuli. We predict that effective treatments will act on neural-autonomic integration, thereby normalizing these differences.

6 Conclusions and Future Directions

In conclusion, there are measurable physiological correlates of BSDs and their responses to treatment. Of particular promise, discernible fMRI and EEG and ERP correlates that characterize and differentiate BSDs, and responses to treatment in BSD patients are beginning to emerge. However, the ability to use these correlates to aid classification of BSDs and to improve treatment selection and prediction of response requires further validation. Additionally, cardiovascular correlates of BSDs and treatment response are still in the initial stages of investigation, but again early findings hold promise especially when considered within the context of the wider literature regarding autonomic disturbances and their relationship with suicidality, and longer term morbidity and mortality. Other than the aforementioned future directions specific to each physiological modality, a number of general future initiatives for research are recommended.

From having developed a simplified integrative embodied framework of BSDs, a major future direction for research into the physiological correlates of BSDs will be to employ simultaneous recording techniques in order to determine dysfunction in the neural-autonomic integration of cognitive and emotional responses. Such an approach will aid our understanding of the adverse impact of BSDs on brain and body function and facilitate the characterization of treatment targets. Building on previous work on neural-autonomic integration in BSDs (e.g., Malhi et al. 2005), future work could partial-out the inhibitory and excitatory neural activation associated with the activity of both the inhibitory PNS and excitatory SNS, respectively. Such work would illustrate the manner in which BSD patients attend to, process, respond to, regulate responses, and recover from cognitive and emotional stimuli, accounting for the bidirectional nature of the central and peripheral ANSs. In doing so, the differential cognitive and emotional inhibition and activation changes associated with the manic and depressed phases may be better integrated and understood. Further, longitudinal and simultaneous multimodal physiological assessment is likely to better differentiate phases of illness, characterize clinical trajectory, and provide insights into the chronobiological changes associated with phase and course in BSDs (see Malhi et al. 2012; Malhi and Kuiper 2013).

With the development of a simplified integrative framework reflecting the current state of the literature, future work may be guided toward examining the embodied, cognitive, and emotional dysfunction that is associated with BSD emotional lability and mood disturbance in accordance with the aforementioned novel predictions. Taking an embodied account of psychophysiological mechanisms, and not just examining the physiological correlates of the dysfunctional parts, is likely to yield more clinical meaningful and objective physiological measures that will in turn improve assessment and therapeutic decision making in patients with BSDs. Given complexity of assessment and the multitude of treatment considerations associated with BSDs, and the fact that they continue to exert an overwhelming burden because of their prevalence and poor response to treatment, there remains a critical need to continue with such endeavors.

7 Conflict of Interest

The authors Tim Outhred and Andrew H. Kemp are supported by an Australian Postgraduate Award and an International Research Professorship from the University of São Paulo, respectively. Gin S. Malhi has received research support from AstraZeneca, Eli Lilly, Organon, Pfizer, Servier, and Wyeth. He has been a speaker for AstraZeneca, Eli Lilly, Janssen Cilag, Lundbeck, Pfizer, Ranbaxy, Servier, and Wyeth. He has been a consultant for AstraZeneca, Eli Lilly, Janssen Cilag, Lundbeck, and Servier.