INTRODUCTION — Evaluation of beat-to-beat heart rate dynamics as a noninvasive, albeit indirect, probe of autonomic nervous system function is of interest from a number of basic perspectives, along with having potential translational applications in cardiology and other clinical areas. For example, a large body of clinical and experimental evidence indicates a central role for the autonomic nervous system in the triggering or sustaining of malignant ventricular arrhythmias [1]. Higher sympathetic activity unopposed by vagal activity promotes arrhythmia in a variety of ways:

●Reducing ventricular refractory period and the ventricular fibrillation threshold

●Promoting triggered activity afterpotentials

●Enhancing automaticity (see "Enhanced cardiac automaticity")

Vagal stimulation opposes these changes and reduces the effects of sympathetic stimulation by prolonging refractoriness, elevating the ventricular fibrillation threshold, and reducing automaticity.

Furthermore, the fundamental role of the autonomic nervous system in regulating inflammation, believed to underlie many disease processes, is increasingly being appreciated. Increased sympathetic activity promotes inflammation, and increased vagal activity moderates it [2,3].

There are currently three major noninvasive or minimally invasive assessment approaches to evaluating the functioning of the autonomic nervous system which provide complementary information about autonomic as well as non-autonomic regulatory mechanisms in health and disease:

●RR interval (RRI) or heart rate variability (HRV) from short-term or longer term monitoring

●Baroreflex sensitivity (BRS) testing

●Bedside autonomic function tests (eg, Valsalva maneuver, tilt testing, and other orthostatic challenges)

The utility of HRV measures for prediction of outcome or detection of changes in clinical status depends on their stability over time. HRV is influenced significantly by age, race, sex, physical fitness, clinical conditions, sleep/wake cycles and drug treatment, but most 24-hour HRV measures appear to be stable when measured on a day-to-day basis and over periods of days to weeks when there are no major intervening clinical events.

This topic will briefly review selected technical aspects and some clinical applications of HRV testing in adults. Although measurements are actually based on the RR intervals (cardiac interbeat interval obtained from a continuous ECG recording and usually from the normal sinus to normal sinus [NN] intervals), the term "HRV," instead of "RRV," has entered the lexicon and will be used here.

AUTONOMIC NERVOUS SYSTEM INTERACTIONS WITH CARDIAC RATE AND RHYTHM

Modulation of heart rate and the autonomic nervous system — The autonomic nervous system is the primary regulator of heart rate in the presence of sinus rhythm (figure 1). The intrinsic sinus node rate at rest (ie, the rate after pharmacologic or surgical denervation of the sinus node) is about 95 to 110 beats per minute [4]. Under normal supine resting conditions, there is little efferent sympathetic neural input to the sinoatrial node, and the concentration of circulating catecholamines is low; however, there is substantial efferent parasympathetic traffic on the vagus nerves, which slows the resting sinus node rate to about 55 to 75 beats per minute in healthy adults. Clinicians should keep in mind that resting heart rate is determined by both sympathetic and parasympathetic tone (ie, the basic firing rate of the nerves). The classical range given for "normal sinus rhythm" at rest of 60 to 100 beats per minute excludes some of the fittest individuals who have resting rates <60 beats per minute but includes some individuals with relatively higher rates (especially >85 to 90 per minute) associated with a variety of adverse outcomes in large population studies [5].

Heart rate variability (HRV) measures the fluctuations in the RR intervals (also referred to as NN intervals for presumed sinus beats) related primarily to autonomic control mechanisms, but it does not measure autonomic "tone" per se, despite the many erroneous claims to the contrary [6]. Instead, HRV reflects the extraordinarily complex, nonlinear interplay of all of the feedback loops, autonomic and nonautonomic, which regulate sinus node pacemaker activity and facilitate the matching of cardiac output to the needs of the body. Therefore, interpretations that attempt to equate specific HRV measures with explicit neuroautonomic mechanisms are, of necessity, vast simplifications. One notable example is the mis-use of low-to-high frequency power ratio (LF/HF ratio) as a reliable index of "sympatho-vagal" balance [7].

Analysis of HRV patterns from continuous electrocardiograms (ECGs) permits the identification and quantification of underlying physiologic rhythms. The strength of these rhythms is expressed by the magnitude of various frequency-domain HRV measures. When recordings of at least 24 hours are available, the predominant physiologic rhythm that accounts for the most HRV is the circadian rhythm, with relatively increased sympathetic activity associated with higher heart rates during the daytime and increased vagal activity associated with lower heart rates during the night [8]. During normal sleep, there are also prominent physiologic rhythms associated with each approximately 90-minute sleep cycle, and there is evidence that these rhythms persist during wake time, possibly in association with neuroendocrine rhythms, but such ultradian HRV measures have not found clinical applications at this point. (See 'HRV methodology, definitions, and normal values' below.)

Baroreflex activity (oscillations at relatively low frequencies that include so-called Mayer waves) also causes concomitant fluctuations in heart rate [9]. These baroreflex changes are most apparent with sudden standing or during tilt in healthy young adults. Finally, rhythmic fluctuations in the frequency of impulse conduction along the vagus nerves, modulated by the rate and depth of breathing, result, among supine subjects with adequate autonomic functioning, in substantial variations in RR intervals at higher frequencies known as respiratory sinus arrhythmia (RSA) [10,11]. RSA can be exaggerated during metronomic breathing or with meditation-related slow breathing rates.

RR interval fluctuations are also affected by mental or physical activities and responses to environmental stressors, which reduce the average frequency of impulses conducted along the vagus nerves and, as effort increases, increase the activity of the sympathetic nervous system. In some individuals, changes related to sleep-disordered breathing events, periodic limb movements, restless legs syndrome, or poor sleep can exaggerate HRV during sleep [12-15]. Non-autonomic factors (eg, neuroendocrine function) also play an important role in modulating HRV, and some HRV can be exaggerated by an abnormally irregular sinus rhythm due to sinus node dysfunction or subtle atrial ectopy (erratic supraventricular rhythm) that is often missed in routine assessments, especially in older individuals [16,17]. (See "Normal sinus rhythm and sinus arrhythmia".)

Triggering of cardiac arrhythmias and the autonomic nervous system — Both branches of the autonomic nervous system have an important role in the triggering or sustaining of malignant ventricular arrhythmias, particularly post-myocardial infarction (MI). This relationship partly explains the predictive value of abnormal HR variability for such events when the arrhythmogenesis is a consequence of alterations in autonomic functioning.

The sympathetic nervous system — Chronically increased sympathetic activity and elevated plasma catecholamines can be found in the setting of myocardial dysfunction [18,19]. These alter the electrophysiologic properties of the myocardium and promote arrhythmogenesis, regardless of the mechanism involved (enhanced automaticity, triggered activity, or reentry) [20,21]. Some of the arrhythmogenic effects of high sympathetic activity are related to the adverse effects of tachycardia, such as ischemia, while others result from heterogeneity of ventricular repolarization.

Increased sympathetic activity can cause tachycardia, which can result in myocardial ischemia, but myocardial ischemia, especially in the anterior wall, independent of the underlying mechanism, usually increases sympathetic activity while decreasing efferent vagal activity [22,23]. Additionally, myocardial ischemia triggers a release of norepinephrine from epicardial sympathetic nerves and an increase in its local myocardial concentrations due to the high extracellular potassium concentrations in the ischemic regions, resulting in regional heterogeneity of depolarization and repolarization, which is an important precondition for the development of reentrant activity and the precipitation of ventricular fibrillation [24].

Myocardial infarction can also involve the epicardial sympathetic nerves, producing regional sympathetic denervation of the myocardium distal to the infarct and further enhancing heterogeneity of repolarization and increasing the potential for reentry during times of enhanced sympathetic activity [25,26]. The noninfarcted tissue responds normally to sympathetic stimulation with a shortening of the ventricular refractory period, while denervated tissue fails to respond.

The parasympathetic nervous system — The parasympathetic nervous system, through vagal innervation, may exert important antiarrhythmic effects by reducing the heart rate and counteracting the proarrhythmic effects of sympathetic nervous system activity [27-30]. Furthermore, the parasympathetic nervous system appears to play a large role in regulating the inflammatory response, and RR variability is inversely related to the production of many inflammatory markers [31,32]. Our understanding of the action of the vagus nerve, however, has been extremely simplistic [33].

Physiologically, as the vagus emerges from the brainstem, it can be visualized anatomically as having two branches: a supradiaphragmatic branch providing neural pathways to organs above the diaphragm and a subdiaphragmatic branch providing pathways to organs below the diaphragm. The vagal motor fibers traveling through the supradiaphragmatic branch predominantly are myelinated and originate in the nucleus ambiguus. By contrast, the vagal motor fibers traveling through the subdiaphragmatic branch are predominantly unmyelinated and originate in the dorsal nucleus of the vagus. Both source nuclei are located in the brainstem with the nucleus ambiguus being ventral to the dorsal nucleus of the vagus. In close proximity to the dorsal nucleus of the vagus is the nucleus tractus solitarius, which serves as the nucleus of the sensory vagal pathways that travel through both branches of the vagus. Through the process of evolution and the transition from reptiles to mammals, there was a transition in how vagal motor pathways regulated the heart. Originally, the heart was regulated by the unmyelinated vagal fibers originating in the dorsal nucleus of the vagus, but with the appearance of mammals, the predominant vagal motor regulation of the heart shifted to myelinated vagal fibers originating in the nucleus ambiguus. This evolutionary process provided a layering of neural circuits involved in the regulation of the heart consistent with the Jacksonian principle of dissolution, which emphasized a disinhibition of phylogenetically older neural pathways as an adaptive response to illness, injury, and threat.

From this perspective, in humans (and other mammals), activation of the phylogenetically "older," unmyelinated, "dorsal" vagus mediates, among other things, the immobility and bradycardia responses associated with severe threat, and possibly some of the slower rhythms of HRV. Thus, the dorsal vagus can be lifesaving, but at the same time, in extreme situations, immobility and bradycardia can result in cardiac arrest and death. The activity of the evolutionarily "newer," myelinated mammalian "ventral" branch of the vagus nerve is hypothesized to be associated with respiratory sinus arrhythmia and a vagal "brake" on heart rate. Its activity is enhanced by safe social connections. This "newer" branch of the vagus is associated with the capability to rapidly release the vagal brake and increase heart rate in the presence of increased demand and also with the ability to recover rapidly to baseline. The conceptual framework for this schema, known as "polyvagal theory" [33], is beginning to be incorporated into a more nuanced understandings of the parasympathetic nervous system in the context of increased interest in clinical devices and interventions that appear to increase vagal activity.

As can be seen from the polyvagal perspective, it is simplistic to assume that "higher parasympathetic activity is always better" since excess vagal activation can result in syncope, heart block, and potentially, in some cases, increase dispersion of atrial refractoriness resulting in paroxysmal atrial fibrillation [34]. Abnormally enhanced vagal activity can also be associated with inferior-posterior type MIs.

HRV METHODOLOGY, DEFINITIONS, AND NORMAL VALUES — Heart rate variability (HRV) is derived from intervals between normal sinus heart beats (NNs) and can be quantified by many methods (table 1) [6,20,35-40], including

●Time domain measures

●Frequency domain measures

●Heart rate turbulence

In addition, newer methods based on nonlinear dynamics (complexity science) have been developed that are discussed elsewhere (www.physionet.org) [41].

It should be noted, however, that because the term "RR interval" is often used, closer scrutiny is needed to determine if results of a particular study are in fact based on NN intervals only. This understanding is increasingly relevant as more wearable devices become commercially available that purport to measure HRV, but may, in fact, derive this measurement indirectly from the pulse waveform. Furthermore, most clinical HRV measures are only interpretable for autonomic assessment purposes if the patient is predominantly in normal sinus rhythm and if the detection of the R-wave peaks and beat morphology is accurate, which can be problematic in a Holter recording ordered for clinical purposes and scanned by a technician who can only allocate limited effort before generating a clinical Holter report [17,42].

HRV measurements — Most Holter scanner manufacturers provide an HRV analysis feature; however, such analysis may not be included in the clinical report unless specifically requested by the ordering clinician. In addition, some manufacturers use different names for the same HRV variables. However, Holter scanners also can export a version of a "beat file," which is a list of the individual heart beats in the recording by morphology (eg, normal, premature ventricular or supraventricular, etc) and the time between them. The beat file can be used to accurately calculate HRV. It should be noted that the RR interval file, often available from pulse-based devices, is not equivalent to a beat file because it is not linked to an actual ECG signal, is heavily influenced by "noise" generated during physical activity, and lacks beat labels or independent ways to check on the accuracy and consistency of each beat's detection.

For more information on HRV measurement, open source software for HRV analysis and open access databases are freely available at the website of the NIH/NCRR Research Resource for Complex Physiologic Signals (www.physionet.org), along with tutorial material. Also, free HRV software is available for download from The Biosignal and Medical Analysis Group in Finland [43]. In 2015, the European Society of Cardiology, along with the European Heart Rhythm Association and the Asia Pacific Heart Rhythm Association, published a position statement that provides a compact summary of HRV techniques and applications [40].

Those who use HRV measures should also be aware that short-term variability of NN intervals does not necessarily reflect intact vagal modulation of the sinoatrial node. Indeed, anomalous (non-respiratory) sinus rhythm may confound measures of high frequency fluctuations in HRV, which are generally interpreted as due to healthy cardiac vagal modulation. Erratic sinus rhythm, described above, especially in older individuals and those with organic heart disease, may be a form of short-term HRV that is a marker of adverse cardiovascular risk [16,17]. An alternative conceptual construct, heart rate fragmentation, along with a set of metrics to quantify potentially nonvagal sources of short-term HRV, have been proposed [44-46]. The utility of this quantitative approach remains to be further validated.

Time domain HRV

●Average NN in milliseconds (and/or heart rate in beats per minute) – Not a measure of variability per se, but average heart rate is an important marker for cardiac autonomic function that is available from any commercial Holter software. Daytime and nighttime average heart rates are also generally available, and often hourly heart rates can be obtained as well, providing potential insights into circadian rhythms.

●SDNN in milliseconds – Standard deviation of NN (normal-to-normal RR) intervals over a 24-hour period reflects total HRV, but in the literature, SDNN is often reported over a brief (often five minutes) measurement period and results are erroneously interpreted by referring to studies from 24-hour measures.

●SDANN in milliseconds – Standard deviation of the average NN intervals for all of the five-minute intervals in a 24-hour continuous ECG recording. However, this measure is not meaningful in a brief recording, but because it is based on five-minute averaged heart rates, which are not affected by short-term variations, SDANN can provide an estimate of the strength of circadian HRV even when the patient is in atrial fibrillation.

●pNN50 and related – Percent NN intervals >50 ms different from the prior interval is often available on commercial Holter HRV reports. pNN50 is extremely sensitive to uneven beat detection and/or incorrect beat morphology labeling. Similar pNN statistics with lower thresholds (eg, pNN20) have also been described [47]. pNN625 is a heart rate corrected version of pNN50 where 625 represents 6.25 percent of the local average NN. pNN50, therefore, would be identical to pNN625 at an average NN of 800 ms, which corresponds to a heart rate of 75 bpm.

●rMSSD in milliseconds – Root mean square of differences between successive NN intervals; essentially the average absolute value of the change in NN interval between beats. rMSSD is also sensitive to uneven beat detection and/or incorrect beat morphology labeling.

It should be noted that when time domain measures are compared between individuals, neither the pNN statistics nor rMSSD tend to be normally distributed and natural log (in) transformation may be needed to permit parametric statistical comparisons. Moreover, especially in elderly individuals and patients with significant autonomic neuropathy (eg, severe diabetes), the pNN statistic can approach zero, creating major problems when log transformation is performed. Generally, in this situation, adding one to all values of pNN statistics solves that problem and permits parametric statistical analysis.

Frequency domain HRV

●Total power (TP) in ms² – TP captures the total variance in HRV. TP is sometimes reported as the total variance over 24 hours and sometimes as the 24-hour average of five-minute variances, an important distinction.

●Ultra-low frequency power (ULF) in ms² – ULF captures the magnitude of underlying rhythms in heart rate at frequencies of every five minutes to once in 24 hours and is not meaningful in a short recording.

●Very low frequency power (VLF) in ms² – VLF captures the magnitude of underlying oscillations in the heart rate pattern at frequencies between every 25 seconds and every five minutes (0.003 to 0.04 Hz). VLF power at night is exaggerated in the case of sleep-disordered breathing or periodic limb movements, both of which cause oscillations of heart rate patterns at frequencies within the VLF band.

●Low-frequency power (LF) in ms² – LF captures the magnitude of heart rate oscillations in the range of three to nine cycles per minute (0.04 to 0.15 Hz).

●High-frequency power (HF) in ms² – HF captures heart rate oscillations in the range of 9 to 24 cycles per minute, which is the range of typical adult respiratory frequencies (0.15 to 0.40 Hz). The HF band limits would need to be adjusted in recordings of infants and small children who normally have higher respiratory rates than adults.

●LF/HF ratio (unitless) –Often referred to as the "sympathovagal" balance. However, this term is something of a misnomer since lower frequency fluctuations may be related to both sympathetic and parasympathetic activity [7]. Also, during exercise, overall heart rate variability (including LF) decreases, yet sympathetic activity is high. The LF/HF ratio may be most useful during maneuvers like standing and tilt testing, or at rest during metronomic breathing.

●Normalized LF power (NLF) in percent – NLF captures the proportion of HRV accounted for by low frequency power. A similar and complemental measure, normalized HF power (NHF), can be calculated in the same manner.

For measures like VLF, LF, HF, LF/HF, and NLF, 24-hour HRV can be viewed as the averaged HRV from multiple shorter recording periods. Note that for statistical reasons, most frequency domain HRV, which is not normally distributed in the population, is natural log (ln) transformed. NLF and NHF are exceptions. Ln transformation normalizes the distribution and tends to bring outlier values back to the middle of the distribution, so at times, frequency domain HRV values can appear to provide better parametric statistical discrimination between normal and abnormal HRV than time domain HRV.

Heart rate turbulence — Heart rate "turbulence" (HRT) is a relatively recent HRV parameter that evaluates the oscillation (shortening then lengthening) in NN intervals associated with a temporary loss of cardiac output in the presence of a premature ventricular complex/contraction (PVC; also referred to a premature ventricular beat or premature ventricular depolarization) [48]. Thus, HRT can be thought of as a measure of the resilience of the autonomic nervous system in the face of a challenge. As such, HRT has been proposed as a useful marker for cardiac autonomic neuropathy in some populations. Two measures have been calculated, turbulence onset (TO) and turbulence slope (TS), using a composite of the responses to all PVCs on the recording.

●TO measures whether or not there is a brief tachycardia in the average of two normal heart beat intervals after, compared with before, the PVC. The absence of this response (zero or positive TO) would indicate a lack of vagally-modulated autonomic adaptation to the PVC. From a polyvagal perspective, as described above, TO would reflect the immediate ability of the vagal brake to be released in order to compensate for the acute loss of cardiac output associated with the PVC.

●TS measures the degree to which there is a predominantly baroreflex-mediated slower oscillation in heart rate (bradycardia, tachycardia, and return to baseline) after the PVC. A low value for TS (<2.5 ms/beat) is a strong marker for autonomic dysfunction among cardiac patients and suggests a potential inability of the cardiovascular system to recover appropriately after challenges [49]. Abnormal HRT has also been shown to be a risk marker in older adults without cardiac disease, although cut points for TS and elevated risk appear to be higher [50].

Although HRT measures, which generally require 24-hour Holter monitoring or at least five PVCs for calculation, are not available on most commercial Holter systems, abnormal heart rate turbulence may help identify autonomic dysfunction and increased risk of cardiovascular events among both cardiac patients and in low-risk population-dwelling older adults [49,51-53]. As an example, in a prospective cohort study of 1455 acute myocardial infarction survivors with HRT measurements who were followed for an average of 22 months, patients with abnormal TO and TS were at significantly greater risk of death (hazard ratio 5.9, 95% CI 2.9-12.2) [53]. Although abnormal HRT has been strongly associated with risk of sudden death, a more complete review of research findings for HRT in 2013 concluded that "to enhance its predictive value HRT should be evaluated in combination with other markers" [54]. The results of SCD-HeFT, in which abnormal TS was found to be one of several Holter-based risk factors for sudden cardiac death among patients with CHF, support this conclusion [55].

Physiologic interpretation of HRV measures — When HRV is measured over a five-minute period, SDNN, VLF, LF, and HF power, the LF/HF ratio, normalized LF power, rMSSD, and pNN50 values can be generated. SDNN in this case would be the total HRV in five minutes only and, as previously mentioned, is not related to circadian rhythm or to any paper likely cited in the manuscript that links decreased SDNN to adverse outcomes. HRV measures from five-minute or other short-term recordings, in some cases as short as one minute, have similar physiologic meanings to those from 24-hour recordings, which can be seen as the average of multiple short recordings, but they reflect only a very limited snapshot of autonomic function, and there are considerable circadian changes in autonomic function within a single 24-hour cycle [56].

Heart rate should be the first number to be evaluated to give the rest of the values some context. As noted, the same names are often used for variables derived both short-term (typically five-minute) and longer term (usually 24-hour) recordings, so it is important to remember the origin of the data. Thus, it is obvious that a mean heart rate of 80 bpm over 24 hours would not be the same as a mean heart rate of 80 bpm while sitting in the office.

●SDNN – SDNN captures total HRV, and low values on a 24-hour recording reflect a lack of circadian rhythm, a concerning sign. Higher values usually, though not necessarily, mean all is well, although a patient in atrial fibrillation might have very high values of SDNN, as would a patient whose heart rate increased throughout the recording rather than following a circadian rhythm. However, among cardiac patients, SDNN >100 ms has been associated with a markedly lower risk of mortality. For example in the UK-heart study, among patients with CHF, annual mortality was 5.5 percent for SDNN >100 ms versus 51.4 percent in patients with SDNN <50 ms [57].

●SDANN – SDANN is only meaningful in a longer recording and also captures circadian rhythm in a 24-hour recording [58]. The same is true of ULF power, which measures nearly the same thing. SDNN and SDANN measured over 24 hours are usually of a similar magnitude because of the predominant contribution of circadian rhythm to total HRV. Thus, SDNN and the smoothed-out SD of five minutes averaged NN version, SDANN, should be within approximately 20 or 30 ms of each other, with SDANN usually lower, and a wider discrepancy could raise questions about the underlying rhythm or the quality of scanning. Since 24-hour SDNN, SDANN, TP, and ULF are all primarily influenced by the magnitude of the circadian rhythm, they tend to provide very similar information to the clinician, but SDNN is the best known and the most intuitively obvious.

●VLF power – The physiologic interpretation of VLF power has not been well-studied. In healthy adults, VLF appears to reflect parasympathetic activity since it is abolished by atropine administration and unaffected by beta-blockade. It also appears to reflect the activity of the renin-angiotensin system since it is reduced by ACE inhibition [59,60]. At the same time, decreased VLF power has been strongly related to adverse outcomes. As is the case for most HRV measures, context is important. Numbers alone do not carry enough information. Thus, because it captures oscillations in the range of every 25 seconds to every five minutes, VLF power is also increased by abnormalities such as sleep disordered breathing events (although it could be argued that the ability to mount a strong autonomic response to them is a better clinical sign than being unable to mount such a response).

●LF power – LF power reflects the combined modulation of efferent parasympathetic (vagal) and efferent sympathetic nervous system activity and is modulated by baroreflex activity [61].

●HF power – HF power, under normal conditions, reflects modulation of efferent parasympathetic (vagal) activity by ventilation (respiratory sinus arrhythmia), but only in the presence of true sinus rhythm. When the power spectrum is plotted with the subject supine, HF power has a clearly visible peak, the center frequency of which reflects the predominant respiratory frequency. In the presence of very low respiratory rates (ie, below 10 per minute), respiration can actually modulate LF power, and the amount of HRV can increase dramatically because respiration is affecting both sympathetic and parasympathetic control of heart rate [62-64]. Also, very rapid respiration, like that seen among CHF patients, sharply decreases HF power amplitude, paradoxically reflecting loss of vagal function because there is not enough time for the full effect of increased parasympathetic activity during the exhalation phase to slow the heart rate even among healthy people [65]. However, across the usual range of breathing frequencies, changes in respiratory rate appear to have little effect on the amplitude of the HF peak [65].

●LF/HF ratio – The LF/HF ratio is often claimed to characterize "sympathovagal balance" or "relative sympathetic activity" because the LF band reflects modulation by both the sympathetic and parasympathetic arms of the autonomic nervous system and the HF band reflects parasympathetic activity. Although there are selected situations where this concept might usefully be applied, far too many situations exist where it cannot (eg, exercise, heart failure, the cold pressor test, etc), and claims about the meaning of this ratio should be interpreted with extreme caution [7,66].

Several caveats apply to the interpretation of HRV measures. The position of the patient during a five-minute recording will dramatically affect HRV since parasympathetic activity is markedly increased for most people when they are supine. The limitations of this five-minute resting assessment must also be appreciated. Just as a resting ECG can yield information about clear ECG abnormalities, it provides no information about abnormalities or adaptations when the patient is active. A similar limitation applies to supine five-minute HRV [6]. If it is markedly depressed, then certainly there is some autonomic abnormality present, but the effect of activity on HRV or whether circadian rhythm is abnormal, etc, will be much more informative than a resting snapshot.

Two other metrics defined in the table, rMSSD and pNN measures (table 1), are also usually reported from both five-minute and longer recordings [35,47]. Both characterize heart rate changes on a beat-to-beat basis and, since beat-to-beat changes in heart rate as opposed to longer term trends are mediated by changes in parasympathetic activity, the magnitude of either of these measures is taken as a surrogate of parasympathetic modulation of heart rate or, in the context of polyvagal theory, the activity of the myelinated vagus. For this reason, these measures should correlate strongly with each other and with HF power, which also measures parasympathetic activity as expressed by changes in the respiratory frequency band.

Normal values for HRV — One of the complicating factors in using HRV clinically, aside from the technical issues discussed above, is that unlike, for example, a blood pressure measurement, there is a considerable range of normal values even in people of the same age and gender, and considerable ranges of normal change with age. A review of existing studies of normal values for short-term HRV found huge discrepancies in values [67]. More is known about 24-hour HRV measures, which have been collected in a large number of clinical populations and in some nonclinical ones as well. However, few studies have considered the effect of race on 24-hour HRV, despite the fact that even among well-educated African-Americans with middle-class or higher income levels, health outcomes are significantly worse than in comparable Caucasian subjects.  

●One study compared HRV in healthy middle-aged adults, recent post-MI patients (from the 1980s), and patients who were one-year post-MI [68]. SDNN was highest in the healthy adults, intermediate in the one-year post-MI group, and lowest in the recent MI group, but there were no differences between groups in time domain measures of beat-to-beat HRV (rMSSD and pNN50), potentially because the organization of the heart rate patterns (true RSA versus erratic sinus rhythm) cannot be determined from traditional time domain measures. Importantly, the study reported that at the upper end of the distribution for the three groups, HRV did not separate healthy subjects from patients with either recent MI or one year post-MI. Thus, clinically, it is possible to find normal circadian HRV in a cardiac patient, probably reflecting the ability to sleep well. Conversely, consistent with the idea that very low HRV reflects a significant abnormality, only 2 of 274 normal adults had values for ULF power previously reported to identify CHD patients at high risk of death, and only three had values for LF power below the cutpoint for increased risk of mortality in CHD patients, thus supporting the possibility that finding a presumably healthy adult with very low 24-hour HRV would be clinically meaningful but also underscoring the fact the "normal" HRV values may be found in patients with known CVD.

●A 1998 study of HRV in 260 healthy adults found that HRV was both age and gender-dependent, with 95 percent confidence intervals reported for each decade [69]. Unfortunately, these limits are so broad as to be unusable clinically. For example, among adults in their 60's, normal limits for SDNN were defined as between 68 and 186 ms, and normal limits for 24-hour averaged heart rate were 52 to 99 bpm.

●The Cardiovascular Health Study examined changes in 24-hour-based frequency domain HRV over five years in 585 adults >65 years old [70]. HRV declined more in the group that was aged 65 to 69 at baseline, less in the group that was 70 to 74, and almost not at all in the group that was ≥75 years old. Although most HRV was slightly but significantly lower in the group with prevalent cardiovascular disease, there was no difference in the magnitude of the age-related change in HRV among those with or without CVD. However, heart rate, ratio-based and nonlinear HRV values continued to decline with advancing age [70].

Reproducibility of HRV measures — The utility of HRV measures for prediction of outcome or detection of changes in clinical status depends on their stability over time. HRV is influenced significantly by age, race, sex, physical fitness, clinical comorbidities and drug treatment, but most 24-hour HRV appears to be stable when measured on a day-to-day basis and over periods of days to weeks when there are no major intervening events [71-73]. Less is known about serial changes in HRV in the same individuals over longer periods of time, but as previously mentioned, these changes might be age- and measure-dependent [74]. Finally, the limitation of 24-hour recordings is being overcome by advances in technology. Whereas in the early days of clinical HRV data collection, subjects were asked to wear a somewhat cumbersome Holter device that recorded their ECG on reel-to-reel or later cassette tapes (two channels over 24 hours at a sampling rate of 128 Hz), current technology permits collection of multichannel, high-resolution ECGs via a small patch attached to the chest, over several days in a row, either using a built-in SSD card or uploaded to a web server. HRV can also be measured from stored signals from inpatient ICU monitoring. This has greatly expanded potential applications of inpatient and outpatient HRV to interventions, eg, dialysis or chemotherapy.

Reproducibility in patients with cardiac disease — The reproducibility of HR variability measurements in patients with cardiac disease has been found to be comparable to the high stability observed in normal subjects. The stability of measures of HRV makes it possible to distinguish real changes due to progression or regression of cardiac disease or drug effects from apparent changes due to random variation. Unfortunately, criteria for real and significant changes in HRV have not been established. However, if a patient goes from apparently normal HRV (eg, SDNN >120 ms) to markedly decreased HRV (eg SDNN <70 ms) on serial recordings, this would be a matter of concern. As examples:

●Patients with ventricular arrhythmias – In a study that evaluated the day-to-day stability of HRV in two groups of arrhythmia patients (a random sample of 40 patients in the ESVEM study who had sustained ventricular arrhythmias, and a random sample of 40 patients in the placebo group of CAPS who had nonsustained ventricular arrhythmias), there were no significant differences in measures of HRV between the two 24-hour recording sessions in either group [75].

●Patients with stable angina – In a study of 261 patients with chronic stable angina followed for three years, most HRV measures remained largely unchanged over time, although pNN50 and LF/HF were less stable [76]. Declines in most HRV measures, although not HF power, were seen in those who suffered a myocardial infarction during that period. However, as previously stated, the magnitude of traditional time and frequency domain HRV measures fails to capture changes in the organization of heart rate patterns, which may be captured by non-linear measures that were not determined in this study.

CLINICAL USES OF HRV IN CARDIOLOGY — The traditional clinical applications of HRV are based on a Task Force report published in 1996, in which the Task Force recommended two applications of HRV testing [35]:

●Prediction of risk of cardiac death or arrhythmic events post-myocardial infarction (MI)

●Detection and quantification of autonomic neuropathy in patients with diabetes mellitus

A 2015 position statement reiterated the potential role for HRV in post-MI patients and also suggested a role in patients with heart failure [40]. While HRV has been investigated and shows potential for risk stratification in a variety of clinical conditions, and HRV can be derived from a Holter recording ordered for other clinical indications, there remains no clear indication for the test in the routine management of most patients. However, the ability of HRV from Holter monitoring to identify patients with sleep-disordered breathing, and the high and increasing prevalence of this disorder in the population (eg, an estimated 17 percent and increasing among males 50 to 70 years of age, the same age group that is likely to be tested for cardiovascular disease) support the possibility that Holter monitoring could be indicated as a screening tool in cardiac patients [77]. In some research labs, where HRV is measured in various clinical studies, significant and unsuspected sleep-disordered breathing is frequently seen and findings reported to the primary investigator, who then notifies the patient and suggests further evaluation by a sleep lab [78].  

Prediction of mortality in the early post-MI period — The impact of HRV on prognosis post-MI, initially reported in the era prior to treatment with thrombolysis, has also been validated in patients with an MI treated with thrombolytic therapy [79-81]. Patients with reduced indices of HRV measured early following an MI (within 14 days) have a three- to fourfold greater risk of death within three years following an MI. However, the sensitivity of measures like SDNN <50 ms as predictors of mortality, originally estimated to be approximately 30 percent, has declined simply because improvements in post-MI treatment have resulted in markedly fewer patients having SDNN at those levels.

In a later study of 412 post-MI patients treated in the era of percutaneous revascularization and followed for 4.3 years, SDNN <50 ms was seen in only 7 percent of patients, and only 31 patients died in the entire cohort [82]. Even though an SDNN measurement of <50 ms was associated with a doubling of risk of mortality compared with SDNN >50 ms in the earlier landmark study, in the current study, SDNN <50 ms was only found in 3 of the 31 who died [79,82].

Another open question in this era of lower post-MI mortality is that of the optimal recording time after the event. It is possible that an early recording shortly after MI would identify one set of higher risk patients, and a later one, perhaps six weeks after MI, when patients have or have not recovered well, would provide more useful information.

Prediction of mortality in the late post-MI period — While reduced HRV in the early post-MI period is clearly associated with a worse prognosis among patients from the pre-PCI era, the substantial recovery of HR variability within the three months after myocardial infarction, particularly following an inferior infarct, raises a question as to whether recovery values for HR variability predict death [68,83]. Among the 68 placebo-treated patients entered into the Cardiac Arrhythmia Pilot Study (CAPS) who had 24-hour ECG recordings at baseline and at 3, 6, and 12 months after MI, there was a substantial increase in all measures of HRV between three weeks and three months [84]. On average, recovery of HRV was completed by three months post-MI; between 3 and 12 months, the values were stable for the group as a whole and for individual patients.

Several older studies suggested that reduced HRV months or years post-MI remains a predictor of adverse outcomes [85-88]. As an example, in a study of 292 patients admitted with an acute coronary syndrome between 1991 and 1994, reduced HRV seen on 24-hour Holter recordings three to six months after their event continued to have prognostic value [88].

Patients with stable CHD — An association between decreased HRV and the presence of significant coronary heart disease (CHD) has been suggested. Among 470 consecutive patients undergoing elective coronary angiography, patients with obstructive CHD (>50 percent stenosis) had significantly reduced HRV based on five-minute supine measurement, especially in the low frequency (LF) band (180 with versus 267 ms² without obstructive CHD) [89]. In a multivariate analysis using a cutoff of LF power = 250 ms², persons with LF power below 250 ms² were at significantly greater risk of obstructive CHD (adjusted OR 2.4, 95% CI 1.3-4.4) independent of baseline Framingham Risk Scores.

Patients with heart failure — Several studies have shown that patients with heart failure and/or cardiomyopathy have reduced HRV compared with controls, and that reduced HRV was associated with disease severity measures such as NYHA functional class, left ventricular diastolic dimension, reduced left ventricular ejection fraction, and peak O₂ consumption [57,90-96]. Prognostically, reductions in HRV have been shown to be independent predictors of overall mortality, mortality from heart failure, sudden cardiac death, ventricular arrhythmias, and the need for transplant [57,95].

Improvements in HRV are seen with effective heart failure therapy (eg cardiac resynchronized therapy [CRT]). Patients with improved HRV after CRT have been shown to have better outcomes than those in whom HRV is not improved [97,98]. (See "Cardiac resynchronization therapy in heart failure: Indications", section on 'Rationale for CRT'.)

A detailed review of current studies of HRV applications to heart failure patients is available elsewhere [99].

Patients with atrial fibrillation — Though the presence of atrial fibrillation (AF) precludes the use of standard NN variability measures, analysis of the variability and irregularity of the ventricular response interval (VRI) may provide useful information regarding prognosis of patients with chronic AF and likelihood of AF recurrence following cardioversion [100,101].

●In a series of 107 patients with chronic atrial fibrillation who underwent 24-hour ambulatory monitoring for VRI variability and irregularity and were followed for 33 months, reductions in all VRI variability and irregularity measures were associated with an increased risk for cardiac death [100].

●In a series of 93 patients who underwent cardioversion for AF and had RR variability measured in sinus rhythm, the AF recurrence rate at two weeks was significantly higher (73 versus 9 percent) in patients with reduced RR variability [101]. (See "Paroxysmal atrial fibrillation", section on 'Natural history'.)

Furthermore, despite the apparent lack of beat-to-beat autonomic control in AF, a circadian rhythm, measurable by SDANN which averages out the extreme short-term variability in RR intervals, persists and can be assessed, and there is evidence that decreased SDANN is an adverse sign in AF [58]. The clinical applicability of this, like other potential applications of HRV, has not yet been evaluated.

Reduced HRV in the general population — Reduced HRV has been associated with increased mortality and a greater risk of cardiac events in several population studies [50,102-105]. As examples:

●In a random sample of 900 patients enrolled in the ARIC (Atherosclerosis Risk in Communities) study who had time-domain measures of RR variability determined from a two-minute ECG rhythm strip, those with low RR variability had a significantly higher risk of cardiovascular death (adjusted relative risk [RR] 1.98, 95% CI 1.06-3.70) and all-cause death (adjusted RR 1.50, 95% CI ) compared with patients with an intermediate RR variability (SDNN 23.9 to 35.4 ms) [103].

●Among 1172 patients aged 65 years or greater from the CHS (Cardiovascular Health Study), even in the lowest stratum of Framingham Risk Score, abnormal values for HRV and HR turbulence identified a group with a significantly increased relative risk of mortality (RR 7.7 compared with those with normal HRV, 95% CI 3.7-16.0) [52].

●Among 4652 patients in the MESA (Multi-Ethnic Study of Atherosclerosis) population who had time-domain measures of RR variability determined from a 30-second ECG rhythm strip and who were followed for a median of 7.6 years, patients in the lowest tertile of HRV had a significantly greater risk of developing heart failure (adjusted hazard ratio 2.4, 95% CI 1.4-4.2) [104].

In a 2013 meta-analysis of eight studies involving 21,988 patients, patients with the lowest HRV as measured by SDNN had a significantly higher risk of first cardiovascular events compared with patients with the greatest HRV measurements (RR 1.35, 95% CI 1.1-1.7) [105]. Furthermore, combinations of HRV measures from different domains (time domain, frequency domain, non-linear, or HRT) may have better predictive value than HRV from a single domain because they capture different aspects of cardiac autonomic activity.  

HRV in combination with other risk factors — By itself, HRV has limited sensitivity and specificity to sufficiently identify high-risk patients, although as previously suggested, normal HRV has been shown to be found in low risk patients. At the same time, correlations between time or frequency domain measures of HRV and previously-identified postinfarction risk factors such as left ventricular ejection fraction (LVEF) and ventricular arrhythmias are remarkably weak [79]. This lack of correlation and the fact that measures of HRV are independent predictors of patients at high risk after myocardial infarction suggest the potential utility of combining abnormal HRV with other risk stratifiers to improve risk prediction, although doing so clearly limits the proportion of patients to whom this applies [68,80,106].

As an example, among 1284 patients with a recent MI (within 28 days) followed for an average of 21 months in the ATRAMI study in whom baroreflex sensitivity and SDNN were used for risk stratification, patients with abnormal values for both had a 17 percent risk of mortality, whereas patients with normal values for both had a 2 percent risk (of note, mortality for the entire cohort was only 3.5 percent) [80]. The small number of patients with reduced HRV and LVEF less than 35 percent had nearly a sevenfold increase in mortality compared with patients with preserved HRV and LVEF of 35 percent or greater. (RR 6.7, 95% CI 3.1-14.6). HRV has also been shown to add to existing risk scores for prediction of HF and of stroke in asymptomatic older adults [107].

In addition, many other conditions are associated with autonomic abnormalities, including alcoholism, chronic kidney disease, depression, severe hypertension, chemotherapy and other chemotoxins, sleep apnea, etc. So clinically, if the question is whether the patient has significant autonomic neuropathy, absent or very diminished HRV may provide useful adjunctive information. Turning this question around, the ability of serial HRV measures to identify patients who are having an adverse response to an intervention (eg, cardiotoxic chemotherapy) before they manifest clear deficits in cardiac function as reflected, for example, in reduced LVEF, is currently being investigated. As a final example, in infants, quantification of relative bradycardia and decreased heart rate variability, likely due to activation of the dorsal branch of the vagus, may provide early warning of infection-related conditions in very low birth weight infants. In preterm babies, the myelinated vagus has not yet become active, and RSA is minimal [108].

OUR APPROACH — Guidelines for the routine use of ambulatory monitoring to determine heart rate variability (HRV), based on the original Task Force recommendations from 1996, were also published in 1999 by a task force of the American College of Cardiology/American Heart Association [109]. While HRV has been used for risk stratification for all-cause, cardiac, and arrhythmic death post-myocardial infarction (MI), the precise role for HRV testing in clinical practice remains undefined and perhaps is a bit of a moving target due to changes in therapeutic approaches and also because understanding the message of HRV requires a real understanding of the context and often appraisal of the actual heart rate patterns and graphical plots derived from them. Some authors promote HRV for risk stratification, but others have been less enthusiastic [35,110].

With this in mind, we do not currently recommend the routine use of HRV testing outside of the clinical trial setting. We also believe that to further the field, existing and prospective datasets need to be analyzed with the goal of developing understandable and replicable algorithms that capture the rich and full clinical picture obtainable from these recordings.

Finally, the diagnostic and prognostic utility of ultra-short ECGs (eg, 10 to 30 seconds) or recordings for reliable estimates of short-term HRV parameters requires further study [111].

SUMMARY AND RECOMMENDATIONS

●There are three major noninvasive or minimally invasive assessment approaches to evaluating the autonomic nervous system which provide complementary information about autonomic as well as non-autonomic regulatory mechanisms in health and disease: heart rate variability (HRV), baroreflex sensitivity, and bedside autonomic function tests. (See 'Introduction' above.)

●The autonomic nervous system, including the sympathetic and parasympathetic nervous systems, has an important role in the triggering or sustaining of malignant ventricular arrhythmias. (See 'Triggering of cardiac arrhythmias and the autonomic nervous system' above.)

●HRV (derived from intervals between normal sinus beats NN intervals) can be measured by many methods, which can be categorized as time domain measures, frequency domain measures, nonlinear/complexity-based measures, and heart rate turbulence. (See 'HRV methodology, definitions, and normal values' above.)

●HRV appears to be highly consistent over 24 hours, despite marked differences among the five-minute intervals during a day. The same stability in HRV appears to be true for patients with ventricular arrhythmias, angina, and heart failure. (See 'Reproducibility of HRV measures' above.)

●HRV has been shown to be significantly decreased, compared with normal values, among post-myocardial infarction patients, although there is considerable inter-individual difference in this. This decreased HRV is a likely marker for autonomic dysregulation and likely reflects both decreased parasympathetic and increased sympathetic activity. Reduced HRV has also been associated with worse outcomes in patients with stable coronary heart disease, heart failure, and atrial fibrillation. Also, however, without accurate ECG analysis and an understanding of what HRV represents, HRV, especially short-term measures, can appear to be higher in the presence of either uneven beat detection or an erratic (fragmented) sinus rhythm. (See 'Clinical uses of HRV in cardiology' above.)

●We do not recommend the routine use of HRV testing based on clinical Holter scanning at the moment. Refinement of the conceptual framework and algorithms by which HRV is interpreted and measured lend support to the anticipation that HRV will have translational value in a variety of clinical settings. (See 'Our approach' above.)