A literature review of heart rate variability in depressive and bipolar disorders

Introduction

Changes in autonomic nervous system (ANS) activity have been linked to a variety of mood disorders, including bipolar disorders, depressive disorders, anxiety disorders and schizophrenia (Alvares et al., 2015; Gorman and Sloan, 2000; Henry et al., 2010; Kemp et al., 2014a; Moon et al., 2013; Outhred et al., 2014; Valenza et al., 2014), and consequently a variety of cardiac disorders (Kemp and Quintana, 2013; Stapelberg et al., 2012; Stein and Kleiger, 1999) and other general medical disorders (Cygankiewicz and Zareba, 2013; Xhyheri et al., 2012). In addition, vagal nerve stimulation (VNS) has shown promise as a treatment for severe depressive disorders (Daban et al., 2008; George et al., 2007; Rizvi et al., 2011). Regulation of parasympathetic activity (identified as cardiac vagal tone) has also been connected to regulation of emotional reactivity, psychological flexibility and social engagement, as well as reduced prefrontal cortical activity (Geisler et al., 2013; McCraty and Childre, 2010; Porges, 2009; Sgoifo et al., 2015; Thayer et al., 2009, 2012; Thayer and Lane, 2000). This review examines the current literature which explores the associations of the ANS with both major depressive disorders (MDDs) and bipolar disorders, using heart rate variability (HRV) as a tool.

Autonomic dysregulation and HRV

A useful, safe, convenient and non-invasive measure of autonomic activity is HRV. HRV can be measured as the variations in time between consecutive R waves on an electrocardiograph (ECG) (‘Time domain measures’) and decomposition of the wave form of ECG-RR intervals (interbeat intervals or RRIs) into frequency power bands using spectral analysis (‘Frequency domain measures’) (Rajendra Acharya et al., 2006; Voss et al., 2006). A number of parameters are used to measure HRV (Kleiger et al., 2005; Reyes del Paso et al., 2013; Shaffer et al., 2014; Xhyheri et al., 2012). While there is wide agreement about the interpretation of these measures, there remains some debate about the precise implications of each parameter (Heathers, 2014; Quintana and Heathers, 2014). The most commonly used time domain measures are the standard deviation of normal to normal RRIs (SDNN), the root mean square of normal to normal interval differences (RMSSD) and the proportion of RRIs of more than 50 millisecond (pNN50). The SDNN is predominantly (but not exclusively) indicative of sympathetic activity (increasingly with the duration of ECG recordings), while the RMSSD and pNN50 are indicative of parasympathetic activity. The most common frequency domain measures are the high-frequency (HF) power band (HF = 0.15–0.4 Hz) and low-frequency (LF) power band (LF = 0.04–0.15 Hz). The HF measure is indicative of parasympathetic activity and the LF measure of a mix of sympathetic and parasympathetic activity. Total power (TP) is a broad measure of autonomic activity. HRV varies inversely with the heart rate and directly with the interbeat interval.

HRV provides a convenient tool for investigating the activity of the ANS in the pathophysiology of major mood disorders, as well as the genesis of cardiovascular and other complications associated with mood disorders and their treatment (Hanson et al., 2013; Kaplan et al., 1991; Kemp et al., 2012; Kemp and Quintana, 2013; Soares-Miranda et al., 2014; Stapelberg et al., 2012). Measurement of HRV has the potential to offer a diagnostic tool through the identification of individuals suffering some forms of mental illness (Brunoni et al., 2013; Carney et al., 2000; Kemp et al., 2010; Moon et al., 2013; Stapelberg et al., 2012; Valenza et al., 2014). It may offer even more specificity for melancholic/somatic subtypes of depressive illness (Kemp et al., 2014b; Messerotti Benvenuti et al., 2015), but this requires further confirmation.

The activity of the ANS is sensitive to age and gender (Thayer et al., 1998; Voss et al., 2013), physical activity (Perini and Veicsteinas, 2003; Routledge et al., 2010), alcohol consumption (Kemp and Quintana, 2013; Romanowicz et al., 2011; Sagawa et al., 2011), nicotine (Munjal et al., 2009), food and water intake (Quintana and Heathers, 2014), circadian rhythm (Boudreau et al., 2013; Tobaldini et al., 2013), heart rate (prominently around 60 bpm or less) (Billman, 2013; Sacha and Pluta, 2008), blood pressure (Kleiger et al., 2005) and several medications (Hanson et al., 2013; Kemp et al., 2014a; Kemp and Quintana, 2013; O’Regan et al., 2014; Stapelberg et al., 2012). HRV is also highly sensitive to respiration, including respiratory rate, respiratory depth and respiratory sinus arrhythmia (RSA) (Heathers, 2014; Perakakis et al., 2009; Quintana and Heathers, 2014; Schafer and Vagedes, 2013; Shaffer et al., 2014). Regulating the frequency and depth of respiration reduces this contribution to HRV (Quintana and Heathers, 2014). However, the number and complexity of the confounding variables which contribute to HRV make the collection of data from this parameter difficult and interpretation open to question. Designing research methodology to minimize these confounding variables and analysing the data to adjust for unavoidable confounding variables are critically important challenges in this area of study.

The ANS functions in a non-linear fashion, with the parasympathetic system approaching chaotic activity (Voss et al., 2006). Non-linear measures offer, potentially, more valid interpretations of ANS activity than the linear measures discussed above and are less sensitive to confounding variables (Kaplan et al., 1991; Kemp et al., 2010; Voss et al., 2006).

HRV and bipolar disorder

Research into mania has revealed considerable evidence of reduction in many HRV parameters and particularly parasympathetic activity, although the findings are not fully consistent. Henry et al. (2010) studied HRV in 23 hospitalized subjects with current mania and 23 healthy controls, using brief ECG recordings. The severity of illness was considered to be ‘moderate’, using the criteria of Leucht et al. (2005) and Lukasiewicz et al. (2013), and the age range was between 18 and 55 years, inclusive. Subjects with mania exhibited non-significantly higher mean heart rates but reductions in SDNN. In contrast, the HF score was significantly higher in the manic patients, although so was the LF/HF ratio. Measures of the RMSSD and pNN50 were lower in the manic patients compared with the controls. Apart from a small decrease in LF/HF ratio with lithium use, there were no significant differences in measures related to mood stabilizer or antipsychotic use. Similarly, Moon et al. (2013) studied HRV in 41 subjects with bipolar affective disorder (manic phase), 34 with MDD, 35 with schizophrenia and 27 healthy control subjects. The manic subjects were reported to have Young Mania Rating Scale (YMRS) mean scores of 19.15 ± 8.58. They were all taking mood stabilizers and/or atypical antipsychotic medications. A brief ECG was recorded and the SDNN, RMSSD, TP, LF and HF measures were all lower in the manic subjects. Only the HF scores were reduced in MDD. There were no significant differences in Approximate Entropy in either group. The authors suggested that the relatively ‘extreme affective lability’ in the bipolar patients may have contributed to the changes in HRV reported. In 2014, Chang et al. (2014) found that 61 unmedicated patients with mania, compared with 183 healthy control subjects, exhibited lower mean and variance of RRIs (similar in significance to the SDNN) and lower LF and HF spectral power. However, the LF/HF ratios were higher, and both YMRS total scores were positively correlated with LF/HF ratio and negatively correlated with HF measures. These observations suggest that during unmedicated mania, HRV was significantly disrupted and the trend was towards higher sympathetic and lower parasympathetic activity.

Chang et al. (2015) examined HRV using 5-minute, supine ECG recordings (with standardized respiratory rate) in 116 unmedicated subjects with bipolar II disorder (currently depressed), 591 with MDD (also unmedicated) and 421 control subjects. The bipolar and unipolar depressed subjects were significantly older than the controls. The measures of HRV revealed that the bipolar subjects exhibited lower total variance, HF, LF and very-low-frequency (VLF) scores than the depressed subjects, and both were lower than the control subjects. They suggested that HRV measures might assist separation of unipolar and bipolar depression, but these findings await confirmation.

Outhred et al. (2014) summarized the association of bipolar spectrum disorder with disturbances in prefrontal and limbic network neural activity, early neural processing and ANS dysregulation. They commented upon the significant technical variations and limitations in the research they reviewed, as well as the diagnostic heterogeneity and medication use, but noted significant evidence of reduced autonomic activity in these disorders. Quintana et al. (2015) found significant reduction in 33 subjects suffering from bipolar disorders and 47 suffering from schizophrenia, with varying severity of symptoms, compared with 212 healthy controls.

HRV and euthymia in bipolar disorders

Cohen et al. (2003) recorded HRV using brief ECG recordings in 39 euthymic bipolar I and 39 healthy control subjects. The SDNN, TP and LF/HF ratios were significantly lower in the bipolar subjects than controls, and HF power was higher, suggesting a reduction in the sympathovagal ratio. Re-analysis of the data using non-linear mathematical procedures found no significant differences between the groups (Todder et al., 2005).

Gruber et al. (2011) recorded parasympathetic activity using HF scores in 23 euthymic bipolar subjects, compared with 24 controls, and reported a significant difference. The data showed identical mean values but with large variations. Latalova et al. (2013) reported in a ‘Letter to the Editor’ a study of HRV in 23 subjects with remitted bipolar affective disorder (subtype unspecified). Brief ECG recordings were collected and HRV analysis included calculation of time domain variables, frequency domain variables and ‘non-linear analyses’. A reduction in ‘most’ (unspecified) parameters of ANS function was reported, but multiple methodological deficiencies left this report difficult to interpret.

Lee et al. (2012) addressed the association of ‘sub-syndromal’ bipolar depression and HRV. HRV measures from brief ECG recordings in bipolar subjects (10 males and 23 females) and healthy controls (23 males and 36 females) were employed. The Montgomery–Asberg Depression Rating Scale (MADRS) score in bipolar subjects was 3.8 ± 3.7 (mean ± standard deviation), and the Clinical Global Impression–Severity Scale (CGI-S) score was 1.6 ± 0.7 (very low). The bipolar patients were found to have lower HRV measures using the SDNN, pNN50 and log TP. Negative correlations were found between the CGI-S score and SDNN, RMSSD, pNN50, LF and HF powers. They concluded that sub-syndromal bipolar depression (not full remission) was accompanied by reduced HRV.

More recently, Voggt et al. (2015) used 30-minute ECG recordings in 90 predominantly medicated bipolar I and II subjects combined, deemed ‘clinically stable’, and 62 healthy control subjects, to measure differences in HRV. They found that the SDNN, LF and HF parameters were lower in the bipolar subjects. The contribution by medication use (a very significant confounding variable) was not included in their multivariate analysis (multi-linear regression).

HRV and depressive disorder

The research on HRV in depressive disorders has already been reviewed by a number of authors, but it is helpful to summarize those findings and update the literature since their publication. The evidence for parasympathetic dysregulation in depressive disorders was critically assessed by Rottenberg (2007) in a meta-analysis of 13 cross-sectional studies. He concluded that depression only contributed a small to medium effect size impact upon cardiac vagal control, and only explained about 2% of the variance, with a ‘small to medium’ overall effect size (Cohen’s d = 0.332). However, the overall difference between depressed subjects and healthy controls was significant (t = 4.26 and p < 0.001) and the range of Cohen’s d for individual studies was from 0.83 to −0.078. Subsequently, Kemp et al. (2010) performed a meta-analysis of 18 studies, which examined HRV in 673 patients with MDD and 407 healthy control subjects. Those with depression had lower HRV time-based measures, higher ‘high-frequency’ power measures and higher Valsalva ratios than controls. Non-linear measures (probably more reliable than linear measures) indicated a negative correlation of depression severity and HRV. Tricyclic antidepressants were associated with decreased HRV, while selective serotonin reuptake inhibitors (SSRIs), mirtazapine and nefazodone, had no significant effects. In 2012, Stapelberg et al. (2012) reviewed the literature relating to HRV in MDD and found further evidence consistent with reduced parasympathetic activity, although they noted that the results were confounded by the use of medications.

Kemp et al. (2012) used brief ECG sampling in non-medicated depressed subjects, including 24 without anxiety, 24 with generalized anxiety disorder and 14 with panic disorder or post-traumatic stress disorder, compared with 94 age- and sex-matched control subjects. They found significantly reduced RMSSD and HF measures in the depressed subjects with comorbid generalized anxiety. The LF/HF ratios were higher in those subjects, but this is of uncertain significance (Billman, 2013; Reyes del Paso et al., 2013; Shaffer et al., 2014).

Wang et al. (2013), in a carefully designed study, examined a number of HRV variables in 53 MDD subjects (of at least moderate severity) and 53 healthy subjects, matched for age and sex. Using 24-hour ECG recordings, they found SDNN, SDANN (standard deviation of the average normal to normal intervals), RMSSD and pNN50 measures, as well as HF power, were all significantly lower in the depressed subjects. The LF/HF ratios positively correlated with severity of depression, but, as noted above, this ratio is of uncertain significance. Brunoni et al. (2013) observed the effects of treatment for depression using sertraline or direct electric current therapy (tDCS) upon HRV, with 118 moderately depressed subjects and matched controls. The lower HRV measures in their depressed subjects did not change with positive clinical response to either treatment modality. They suggested that reduction in HRV in depressive disorder may be a biomarker of depression. Sgoifo et al. (2015) came to a similar conclusion in their review of the literature regarding autonomic dysregulation and depressive disorders. These conclusions await further confirmation.

Ahrens et al. (2008) found no differences in HRV measures (mainly frequency) in 22 recurrently depressed female subjects, in remission, compared with 20 control subjects (aged 51 ± 1.7 years). The ECG recordings were very brief and HRV measures limited to frequency data. The results are therefore limited in significance. However, Ha et al. (2015a) in a study of 30 predominantly female, ‘newly’ depressed and medication-free subjects (mean age: 65 ± 5 years) exhibited lower high, low and total frequency HRV measures, as well as lower SDNN, RMSSD and pNN50, than 30 matched controls.

Shinba (2013) also studied the reactivity of HRV during depressive episodes in a group of 22 unmedicated depressed subjects, compared with 47 age- and gender-matched control subjects, during a cognitive challenge task. They found an increase in HF power and LF/HF ratio during and after the challenge, suggesting increased parasympathetic and possibly increased sympathetic activity in depressed subjects during a cognitive stress. However, the number of depressed subjects was small and the use of a cognitive challenge makes this study significantly different from those performed at rest.

In a study using 15,105 subjects and carefully considered methodology, Kemp et al. (2014a) found that depressive illness was not associated with reduction in measures of HRV, but recognized that the study population was heterogeneous and confounding variables may have distorted their findings in this group. Subsequently, Kemp et al. (2014b) confirmed the significance of the heterogeneity by finding lower HRV (measured as RMSSD) in subjects suffering from MDD with melancholia (n = 40) than in a non-melancholic cohort (n = 32). Messerotti Benvenuti et al. (2015) found a significant association between somatic symptoms of depression (using the Beck Depression Inventory II) and reduced HRV, measured by SDNN, NN50 and TP (but not RMSSD). However, their study population consisted of 62 healthy college students, and only the 25 students deemed to suffer dysphoric symptoms exhibited this association. Cognitive and affective symptoms of depression were not associated with changes in HRV. Their findings are of interest as they confirm an association of reduced HRV with somatic symptoms of depression (some similarities to melancholia) but are limited by the absence of actual depressive disorders in their subject cohort.

Ha et al. (2015b) reported that in 30 antidepressant-free subjects aged more than 60 years, compared with 30 healthy age- and sex-matched subjects, recently diagnosed depression was associated with reduced LF and HF power, TP, VLF power, SDNN, RMSSD and pNN50 measures. The severity of depression was moderate, with MADRS scores of 24.6 ± 5.

HRV, sleep and mood disorders

As disruption of sleep is highly significant in major mood disorders, promotes relapse, contributes significantly to impaired function (Boland and Alloy, 2013; Boudebesse et al., 2014; Goodwin et al., 2007; Harvey et al., 2005; Murray and Harvey, 2010; Pandi-Perumal and Kramer, 2010; Paykel et al., 1995; St-Amand et al., 2013; Yatham and Maj, 2010) and HRV exhibits a circadian rhythm (Boudreau et al., 2011, 2012, 2013), the associations of HRV with sleep and mood disorders are considered.

In healthy sleep, the sympathovagal balance is tipped towards parasympathetic activity around sleep onset and during sleep stages I, II, III and IV (Bianchi and Mendez, 2013; Busek et al., 2005; Cabiddu et al., 2012; Chouchou and Desseilles, 2014). Conversely, the sympathovagal balance is reversed during rapid eye movement (REM) sleep (Busek et al., 2005; Chouchou and Desseilles, 2014). All autonomic activity falls during stages III and IV, but perhaps with a greater fall in sympathetic than parasympathetic activity (Boudreau et al., 2013; Goldberger et al., 2001).

The sympathovagal imbalance reported in patients with mood disorders during 24-hour, bedtime and sleep studies may in part be related to the relative increase in REM-stage sleep duration reported in these disorders (Palagini et al., 2013; Pandi-Perumal and Kramer, 2010; Paykel et al., 1995). Yang et al. (2011) reported a study of ambulant 24-hour HRV measures (including multiscale entropy) in 52 non-medicated patients with MDD, 47 non-medicated patients with primary insomnia and 88 matched control subjects. The ‘sleep’ process was included in the ECG record of the ‘bedtime’ period. Subjects with current MDD and primary insomnia exhibited significant reductions in HRV indices (consistent with reduced parasympathetic activity) and reduced physiologic complexity, compared with the control subjects. The differences in HRV measures also correlated with poor sleep quality (Pittsburgh Sleep Quality Index and the Epworth Sleepiness Scale), but not with the measures of depressive symptoms (Beck Depression Inventory and Hamilton Depression Rating Scale). This confirmed that insomnia contributed to adverse changes in autonomic function, independent of depressive symptoms.

Migliorini et al. (2011) reported a study of HRV during sleep in one subject with bipolar affective disorder (aged 37 years, type I disorder) and eight healthy control subjects (aged 18–45 years). The study followed a complex design, with the bipolar subject observed over four nights and the control subjects for one night each. The bipolar subject’s mean heart rate, SDNN and RMSSD were lower than the mean of all control values. In addition, the percentage of REM sleep was consistently increased, as would be expected from previous research (Palagini et al., 2013; Paykel et al., 1995; Steiger and Kimura, 2010). These findings were consistent with previous findings of reduced autonomic activity and increased REM sleep in bipolar subjects, but the research design and small numbers limited their significance.

Mariani et al. (2012) reported a ‘bedtime’ HRV study of 12 patients suffering from bipolar disorders, aged between 18 and 65 years, in clinical states defined as euthymic, depressed, hypomanic or mixed. They were compared with 102 healthy control subjects. They found that ‘most’ (unspecified) RMSSD scores and measures of Sample Entropy for bipolar subjects were below the first quartile and/or median value of the control group. This study is very difficult to interpret because of the small number of bipolar subjects, the mixture of mental states being assessed, the heterogeneous nature of the bipolar diagnoses, the lack of data regarding medication use, the lack of detail regarding results and the uncertain comparability of multiple observations from a number of subjects. However, the trend in these findings was again towards a reduction in HRV in subjects suffering from bipolar mood disorders, during the ‘bedtime’ period.

Medication effects in HRV

Research has provided considerable evidence that antidepressant medications can exert significant effects upon HRV, but the differences between antidepressants are noteworthy. The effects of antipsychotic, anticonvulsant medications, lithium and benzodiazepines appear less intense than antidepressants, but the research findings exhibit significant variation.

Therapeutic interventions to enhance HRV

A number of treatment interventions have been considered to enhance HRV and thus inhibit the adverse effects of illness such as depressive and bipolar disorders. Exercise has shown promise (Soares-Miranda et al., 2014; Tonello et al., 2014), but varying efficacy and differences in effect between resistance exercise and aerobic exercise (Kingsley and Figueroa, 2014) and time of day leave the benefits of exercise uncertain. HRV biofeedback has shown some positive effects upon HRV (Prinsloo et al., 2014; Wheat and Larkin 2010), but the evidence is still limited. Acupuncture (Chung et al., 2014) has been employed with some encouraging but limited results, while the benefits of Yoga remain uncertain (Posadzki et al., 2015). Mindfulness may also enhance HRV, but this remains uncertain (Nijjar et al., 2014).

Conclusion

The ANS is central to physiological health and is regulated by central nervous system components which are fundamental to psychological functions. It is increasingly clear that major unipolar and bipolar mood disorders are associated with significant disruption of ANS activity. Whether these changes precede these disorders or persist during remission are important questions to be addressed. They also have implications for a variety of physiological functions such as cardiac function, energy regulation and gastrointestinal motility.

The changes in the operation of the ANS also offer a window into the neurophysiological changes which occur in the central ANS (insular cortex, cingulate cortex, medial and supraorbital prefrontal cortex, with links to hypothalamus, thalamus and periaqueductal grey). The ANS is heavily involved in the stress (or demand) response mechanisms and maintenance of optimal non-stress or vegetative functions. While sympathetic activity is essential for adequate responses to stress/demand, parasympathetic activity balances the potentially adverse effects of sympathetic drive (e.g. excessive arousal and reduced activity of non-essential functions) optimizing the physiological functioning of the ANS. As parasympathetic activity is more rapid and flexible than sympathetic drive, this component of the ANS provides significant modulation of function. A variety of medications, including many psychotropic drugs, also appear to adversely impact autonomic function, but the degree of effect varies widely among medications.

HRV offers a relatively simple, well-tolerated, non-invasive and inexpensive method for studying some of these pathophysiological processes. The results of research into HRV and mood disorders suggest a clear trend towards evidence of a negative association and particularly the dysregulation of parasympathetic activity. However, there has been considerable variation in these results and concerns about the validity of the interpretations made of the variables derived, making definitive conclusions difficult. Careful attention to the management of confounding variables through more substantial sample sizes, careful subject selection (particularly subjects not taking relevant medications), use of non-linear measures, appropriate statistical analyses (such as Propensity Score Matching and Inverse Probability Weighting) (Hogan and Lancaster, 2004; Kemp et al., 2014a, 2014b; Pearl, 2000; Voss et al., 2006; Williamson and Forbes, 2014) and greater consistency in research methodology would facilitate our knowledge of major mood disorders and their adverse effects upon this critically important component of central nervous system function.

Despite these limitations, the study of HRV has already contributed meaningfully to our understanding of the pathophysiology of mood disorders, their adverse effects upon a variety of physiological functions and the adverse effects of psychotropic medications. When combined with clinical and other psychophysiological research techniques, HRV provides a valuable tool for expanding our knowledge and management of these common and disabling mental disorders.