During a 4-year period, all new papers concerning the clinical applicability of heart rate variability in fetal, neonatal, and adult medicine were collected (with the help of Current Contents, ISI, Philadelphia). For this review we selected papers from this collection and, if necessary, gathered less recent but relevant papers.

Physiology of Heart Rate Variability

Because of continuous changes in the sympathetic-parasympathetic balance, the sinus rhythm exhibits fluctuations around the mean heart rate. Frequent small adjustments in heart rate are made by cardiovascular control mechanisms (Figure 1). This results in periodic fluctuations in heart rate. The main periodic fluctuations found are respiratory sinus arrhythmia and baroreflex-related and thermoregulation-related heart rate variability [4, 5].

View larger version (25K): [in this window] [in a new window]   Figure 1. Scheme of the cardiovascular control mechanisms responsible for the main periodic fluctuations in heart rate. The exact pathways through the brainstem are not visualized.

Due to inspiratory inhibition of the vagal tone, the heart rate shows fluctuations with a frequency equal to the respiratory rate [6]. The inspiratory inhibition is evoked primarily by central irradiation of impulses from the medullary respiratory to the cardiovascular center. In addition, peripheral reflexes due to hemodynamic changes and thoracic stretch receptors contribute to respiratory sinus arrhythmia [4]. Fluctuations with the same frequency occur in blood pressure and are known as Traube-Hering waves [7]. Respiratory sinus arrhythmia can be abolished by atropine or vagotomy [4, 8] and is parasympathetically mediated.

The so-called 10-second rhythm in heart rate originates from self-oscillation in the vasomotor part of the baroreflex loop. These intrinsic oscillations result from the negative feedback in the baroreflex [9] and are accompanied by synchronous fluctuations in blood pressure (Mayer waves) [7]. The frequency of the fluctuations is determined by the time delay of the system. They are augmented when sympathetic tone is increased [10-12] and they decrease with sympathetic or parasympathetic blockade [4, 12].

Peripheral vascular resistance exhibits intrinsic oscillations with a low frequency [13, 14]. These oscillations can be influenced by thermal skin stimulation [15] and are thought to arise from thermoregulatory peripheral blood flow adjustments. The fluctuations in peripheral vascular resistance are accompanied by fluctuations with the same frequency in blood pressure and heart rate [15] and are mediated by the sympathetic nervous system.

Heart Rate Variability Measurement

Heart rate variability can be assessed in two ways: by calculation of indices [16] based on statistical operations on R-R intervals (time domain analysis) or by spectral (frequency domain) analysis of an array of R-R intervals [4]. Both methods require accurate timing of R waves. The analysis can be performed on short electrocardiogram (ECG) segments (lasting from 0.5 to 5 minutes) or on 24-hour ECG recordings.

Time Domain Analysis

Two types of heart rate variability indices are distinguished in time domain analysis Figure 2, top). Beat-to-beat or short-term variability (STV) indices represent fast changes in heart rate. Long-term variability (LTV) indices are slower fluctuations (fewer than 6 per minute). Both types of indices are calculated from the R-R intervals occurring in a chosen time window (usually between 0.5 and 5 minutes). An example of a simple STV index is the standard deviation (SD) of beat-to-beat R-R interval differences within the time window. Examples of LTV indices are the SD of all the R-R intervals, or the difference between the maximum and minimum R-R interval length, within the window. With calculated heart rate variability indices, respiratory sinus arrhythmia contributes to STV, and baroreflex- and thermoregulation-related heart rate variability contribute to LTV.

View larger version (19K): [in this window] [in a new window]   Figure 2. Example of an adult heart rate trace.Top. Each point represents the instantaneous heart rate calculated from the accessory R-R interval length. Beat-to-beat or short-term variability (STV) and long-term variability (LTV) are indicated. Bottom. Heart rate power spectrum of the heart rate trace visualized in the top panel. The three main components (peaks) of the power spectrum correspond to the main periodic fluctuations in heart rate: thermoregulation-related heart rate variability: very-low-frequency (VLF) fluctuations, fewer than 3 per minute or 0.05 Hz; baroreflex-related heart rate variability: low-frequency (LF) fluctuations, approximately 6 per minute or 0.1 Hz; and respiratory sinus arrhythmia: high-frequency (HF) fluctuations, equal to the respiratory rate, for example, 19/min or 0.32 Hz. The area under each peak corresponds to the amount or amplitude of each specific fluctuation present in the original heart rate trace.

Twenty-four-hour ECG recordings are frequently used by cardiologists to calculate heart rate variability. For instance, the SD of all R-R intervals within the 24-hour recording, or the mean of the SD of R-R intervals within successive 5-minute periods, is calculated [17-19] (Table 1). These 24-hour indices of heart rate variability also encounter ultradian rhythms (that is, with a period length greater than 1 hour) in heart rate.

Frequency Domain Analysis

Since spectral analysis was introduced as a method to study heart rate variability [5, 20], an increasing number of investigators prefer this method to that of calculation of heart rate variability indices Figure 2, bottom). The main advantage of spectral analysis of signals is the possibility to study their frequency-specific oscillations [7, 21, 22]. Thus not only the amount of variability but also the oscillation frequency (number of heart rate fluctuations per second) can be obtained. Spectral analysis involves decomposing the series of sequential R-R intervals into a sum of sinusoidal functions of different amplitudes and frequencies by the Fourier transform algorithm. The result can be displayed (power spectrum) with the magnitude of variability as a function of frequency [23]. Thus the power spectrum reflects the amplitude of the heart rate fluctuations present at different oscillation frequencies (see Figure 2, bottom).

Spectral analysis can be performed on a short-lasting heart rate trace of 0.5 minute to several minutes. The individual R-R intervals are obtained by R-wave detection. The subsequent array of R-R intervals must be free of artifacts (for example, missed or spurious R waves). To perform a Fourier function on a time-limited signal, the signal must be periodic and stationary [7]. The series of time intervals between consecutive R waves can be treated as if these intervals are equally spaced (a function of R-R interval length against R-R interval number). The Fourier transformation will then result in a spectrum with power as a function of frequency expressed in cycles per beat. The expression can be transformed in Hertz by dividing by mean R-R interval length. To obtain a real data sequence of events equally spaced in time, the sequential R-R intervals are considered as a function of time, interpolated, and subsequently sampled. To obtain a stationary signal, a detrending procedure must be performed. This can be done by subtracting a least-square polynomial approximation from the original signal or by high-pass filtering [7].

Respiratory sinus arrhythmia gives a spectral peak around the respiratory frequency, the baroreflex-related heart rate fluctuations are found as a spectral peak around 0.1 Hz in adults [4], and the thermoregulation-related fluctuations are found as a peak below 0.05 Hz (see Figure 2, bottom).

Measurement Conditions

Heart rate variability can be studied under spontaneous conditions or with provocation; for example, standing or head-up tilt (increase in sympathetic tone) or deep breathing at a rate of 6 breaths per minute (increase in respiration-related heart rate variability).

A 24-hour Holter ECG recording is part of the routine investigations following an acute myocardial infarction. In most of the studies concerning postinfarction patients, therefore, heart rate variability has been established using these 24-hour ECG recordings. In other fields of medicine, for example, regarding diabetic autonomic neuropathy, short-lasting ECG records (ranging from 0.5 to 10 minutes) have been used to calculate spectral and nonspectral heart rate variability indices. These short-lasting measurements were nearly always performed under standardized conditions with and without autonomic nervous system stimulation (that is, tilt and deep breathing).

Commercially Available Equipment

Commercial devices to assess 24-hour heart rate variability are now available. The conventional tape recorders for Holter monitoring may show variations in tape speed that may cause erroneous STV results [24]. Therefore speed control is necessary with the help of a timing track, that is, simultaneously recorded, crystal-generated timing pulses. The only study that we know of that evaluates commercially available heart rate variability equipment is the study of Molgaard and colleagues [24] concerning the Pathfinder II system. This system corrects for tape speed errors and has a high accuracy of QRS detection but contains no correction for artificially long R-R intervals [24]. The effect of artificially long R-R intervals depends on the heart rate variability index used.

Maturational and Physiologic Influences on Heart Rate Variability

Maturity of the Autonomic Nervous System

The amount of heart rate variability is influenced by physiologic and maturational factors. Maturation of the sympathetic and vagal divisions of the autonomic nervous system results in an increase in heart rate variability with gestational age [25, 26] and during early postnatal life [26, 27]. Heart rate variability decreases with age. This decline starts in childhood [28, 29]. Infants have a high sympathetic activity that decreases quickly between ages 5 and 10 years [28]. The influence of provocation on heart rate variability (that is, standing and fixed breathing) is more pronounced at younger ages [29]. In adults, an attenuation of respiratory sinus arrhythmia with advancing age usually predominates [30-32]. In contrast, Jennings and Mack [33] found a heart rate variability decline that was more pronounced in low-frequency heart rate variability than in respiratory sinus arrhythmia. However, it has also been found that the ratio between high- and low-frequency heart rate variability is stable with advancing age [29, 34]. Thus the parasympathetic-sympathetic balance does not change.

Respiratory Frequency

Respiratory sinus arrhythmia increases when respiratory frequency approaches the frequency of the intrinsic baroreflex-related heart rate fluctuations (entrainment). Therefore, respiratory sinus arrhythmia in adults is maximal at a breathing rate of 6 per minute (0.1 Hz). Respiratory rate greater than this frequency is negatively correlated with the amount of respiratory sinus arrhythmia [35, 36].

Renin-Angiotensin System

It has been proved that angiotensin-converting enzyme blocking augments the low-frequency heart rate oscillations in dogs. Therefore it has been suggested that the renin-angiotensin system lowers LTV due to attenuation of fluctuations in peripheral vascular resistance [4].

Behavioral State

Investigations in the fetus and newborn have revealed that during rapid eye movement (REM) sleep LTV is increased and STV is decreased compared to during non-REM sleep [26, 37, 38]. These differences between REM and non-REM sleep are due mainly to a shift in the vagal-sympathetic balance from a higher sympathetic tone during REM sleep to a higher vagal tone during non-REM sleep [39]. In addition, the slower and more regular breathing in non-REM sleep (more respiratory sinus arrhythmia, thus more STV) contributes to the differences found.

Adult heart rate variability has been investigated primarily in awake adults because this enables investigators to instruct the participants to breath at fixed frequencies. Heart rate variability studies in adults have revealed that body posture influences heart rate variability. In the upright position, baroreflex-related heart rate variability is enhanced due to an increased sympathetic tone. Respiratory sinus arrhythmia is augmented in the supine position [12].

Heart Rate and Circadian Rhythm

A positive correlation between STV and R-R interval length has been reported [35, 40]. This correlation is probably caused by an increase in parasympathetic tone related to R-R interval elongation. In adult heart rate variability, a circadian rhythm has been found concomitant with night-day fluctuations in sympathetic and vagal activity. Thus respiratory sinus arrhythmia increases during the night and decreases in the morning. In contrast, the baroreflex-related heart rate variability decreases during the night and exhibits a rapid early morning increase [41].

Fetal and Neonatal Heart Rate Variability

In obstetrics it has been noticed that acute hypoxia resulted in an increase in heart rate variability [42], whereas chronic hypoxemia was accompanied by low heart rate variability [43, 44]. Low heart rate variability is associated with low Apgar scores [45] and pH [46] at birth and has been attributed to depression of the central nervous system [47] because a persistent fixed fetal heart rate pattern was also described in anencephaly [48] and fetal decerebration [49]. Unfortunately, the reduction in heart rate variability appears to be a rather late sign of fetal compromise [50].

In asphyxiated newborns, diminished heart rate variability is also found [51, 52]. Transient loss of heart rate variability indicates a good prognosis and might be caused by cerebral edema, whereas sustained loss of heart rate variability predicts neurologic sequelae or neonatal death and is probably due to irreversible damage to the brain or brain stem. Severe neonatal respiratory distress syndrome is accompanied by a reduction in low-frequency heart rate variability, caused by transient depression of the medulla oblongata due to elevated PCO₂ levels and acidosis [53-56]. If the respiratory distress improves, heart rate variability increases [55]. A recent report has cited reduction in LTV in newborns with a clinically significant patent ductus arteriosus [55, 57]. This might be ascribed to a marginal oxygen supply of the myocardium that limits fluctuations in heart rate [58]. Loss of heart rate variability also has been found in infants with periventricular hemorrhage [59]. Damage of vasomotor areas in the medulla oblongata due to increased intracranial pressure has been suggested to be the cause of the heart rate variability loss. In infants who subsequently died of the sudden infant death syndrome, a higher heart rate and lower heart rate variability have been found [60, 61]. The explanation might be a decrease in vagal tone [62] or reduced motility in infants at risk [63].

Cardiovascular Disease

Acute Myocardial Infarction

A predominance of sympathetic activity and a reduction in parasympathetic cardiac control has been found in patients with acute myocardial infarction [64]. Sympathetic activity decreases the fibrillation threshold and predisposes to ventricular fibrillation. Vagal activity increases the threshold and appears to protect against malignant ventricular tachyarrhythmias [65-68]. The degree of respiratory sinus arrhythmia shows a linear relation with parasympathetic cardiac control [22, 69] and thus can be used as a prognostic tool in patients who have had a myocardial infarction.

One of the first studies relating heart rate variability to death in cardiac patients was done by Wolf and colleagues [70]. In this study the variance of R-R interval length in short segments of ECG recordings (30 R-R intervals) was calculated. A relative low variance was accompanied by a higher risk (3.8) for in-hospital death. In later studies 24-hour ECG recordings were used to obtain a measure of overall heart rate variability. These studies have also proved that myocardial infarction lowers beat-to-beat heart rate variability and that diminished heart rate variability is associated with an increased risk for ventricular fibrillation and sudden cardiac death [17, 18, 71, 72] (see Table 1). Anterior wall myocardial infarction results in a more profound reduction in heart rate variability than inferior wall infarction [73].

Spectral analysis has been used to investigate postinfarction heart rate variability in only a few studies. Lombardi and colleagues [74] described a relative increase in low-frequency and decrease in high-frequency heart rate fluctuations 2 weeks after myocardial infarction. During the first hours after the event, a sympathetic predominance can already be found by spectral analysis of heart rate [22]. An augmented low-frequency component of the power spectrum still exists 1 month after the infarction [75], but during the following months a gradual normalization in power distribution between the low-and high-frequency components and thus in sympathovagal interaction occurs [74, 76]. When assessed by the nonspectral heart rate variability indices, the differences in heart rate variability between postinfarction patients and age-matched controls also decrease with advancing postinfarction time [72].

Heart rate variability is an independent predictor of death when other known postinfarction risk variables (for example, prolongated left ventricular ejection fraction, ventricular arrhythmias, and clinical variables) are considered [18, 71]. Heart rate variability has a higher association with risk for death than other variables obtained by Holter monitoring (for example, mean heart rate and ventricular arrhythmias) [71]. Heart rate variability also appears to be a better predictor of arrhythmic complications than prolongation of the ejection fraction [77]. However, although the risk for arrhythmic events and sudden cardiac death is significantly increased in patients with an abnormally low heart rate variability, this risk remains relatively low. Therefore low heart rate variability has a rather low positive predictive value in mass screening (< 20%) [18] and is of limited value for risk stratification [78]. In combination with other risk factors (for example, late potentials), the positive predictive value increases to more than 30% [18].

Another autonomic nervous system measure that seems to be associated with a good prognosis after myocardial infarction is the baroreflex function. Determination of the baroreflex slope with pharmacologic manipulation of the blood pressure (epinephrine, nitroprusside) gives an indication about the strength of vagal reflexes, and strong vagal reflexes reduce the vulnerability to myocardial infarction [66]. A low baroreflex sensitivity in postinfarction patients is accompanied by an increased mortality rate (sensitivity, 71.4%; specificity, 93.0%) [65].

Which method of heart rate variability measurement should be used? Myers and colleagues [79] compared the effectiveness of several nonspectral heart rate variability indices with spectral analysis of heart rate to segregate patients at low and high risk for sudden cardiac death. They concluded that power spectral analysis was more effective. However, Bigger and associates [80] and Kleiger and coworkers [1] described a good correlation between nonspectral and spectral measures of heart rate variability, showing that these measures are interchangeable. The advantage of the nonspectral heart rate variability indices is that they are simple and inexpensive to compute.

The widespread availability of equipment for 24-hour ECG recording in cardiology units has resulted in heart rate variability measures calculated for a period of 24 hours. In other fields of medicine, shorter recording times are usually used to assess heart rate variability. As explained earlier, the fastest heart rate fluctuations (respiratory sinus arrhythmia) are vagally mediated, whereas slower fluctuations are mediated by both vagal and sympathetic divisions of the autonomic nervous system. Indices of heart rate variability calculated during a 24-hour period include not only the heart rate rhythms caused by respiration, blood pressure control, and thermoregulation but also slower diurnal rhythms. One could question, therefore, whether heart rate variability determined in short recordings under standardized conditions is a better predictor of sympathovagal balance, and thus of death, than 24-hour heart rate variability. The low 24-hour heart rate variability in high-risk postinfarction patients appears to be attributable to a faster heart rate with smaller differences between adjacent R-R intervals, a smaller difference in daytime and night-time heart rate, and fewer short-lasting heart rate accelerations [81]. The difference between daytime and night-time respiratory sinus arrhythmia is decreased as well [82]. Thus the reduction in 24-hour heart rate variability is due to diminished vagal nervous activity, amplified by an attenuated nocturnal increase in vagal tone. Studies by Malik and colleagues [83-85] comparing heart rate variability measurements based on different ECG recording lengths have revealed that high- and low-risk patients are most significantly distinguished when both short and very long components of heart rate variability are included. This is due to a greater circadian fluctuation in sympathetic and parasympathetic influence on heart rate in low-risk compared with high-risk patients. Therefore the components of heart rate variability due to diurnal rhythms must be included in the heart rate variability measurement when it is used to determine long-term prognoses after myocardial infarction.

Other Cardiovascular Diseases

In congestive heart failure and coronary artery disease, heart rate variability is also altered due to attenuated vagal and augmented sympathetic tone [86]. A diminished vagal but relatively preserved sympathetic modulation of heart rate was found by Saul and colleagues [87] and by Binkley and associates [88], using spectral analysis of short-lasting ECG recordings (15- and 4-minute, respectively), in patients with congestive heart failure. Airaksinen and colleagues [89] also found a diminished vagal activity in patients with coronary artery disease. Respiratory sinus arrhythmia was attenuated during a deep breathing test (1 minute breathing at 6 breaths per minute). Heart rate variability based on 24-hour monitoring is also decreased in congestive heart failure and coronary artery disease [90, 91]. Hayano and associates [92] reported a negative correlation between the extent of coronary atherosclerosis and respiratory sinus arrhythmia measured by spectral analysis of 5-minute recordings in patients who had coronary artery angiography. However, other investigators could not detect a significant correlation between heart rate variability impairment and extent of coronary artery disease [89, 93] or severity of congestive heart failure as assessed by left ventricular ejection fraction, cardiac output, pulmonary wedge pressure, or clinical classification [94]. Therefore, heart rate variability is a marker of sympathovagal balance or imbalance rather than of severity of disease. Decreased heart rate variability in these patients also seems to be a predictor of death [93] (see Table 1).

Heart transplantation results in extrinsic denervation of the transplanted heart. Only intracardiac and humoral mechanisms modulate heart rate variability of the recipient. Heart rate power spectra are flat, indicating dissociation of the donor heart from the recipient's central nervous system [95, 96]. Sands and associates reported an inexplicable increase in heart rate variability in patients developing allograft rejection [95]. However, this was not found by others [96, 97]. Fallen and associates [96] reported reappearance of normal heart rate variability in the power spectrum of a recipient, without histologic signs of rejection, 33 months after transplantation. Both respiratory sinus arrhythmia and baroreflex-related heart rate variability were present. They suggested that re-innervation occurred in this patient.

In essential hypertension an enhanced sympathetic activity and a reduced vagal activity [98] have been confirmed by spectral analysis of short-lasting ECG registrations. The altered sympathovagal balance of cardiac control results in less (vagally mediated) respiratory sinus arrhythmia and more baroreflex-related (0.1 Hz) heart rate variability in patients with essential hypertension compared with normotensive controls [10, 99]. In contrast, sympathetic stimulation by tilt [10, 99] or lower-body negative pressure [100], which normally results in an increase in low-frequency heart rate oscillations, gives an attenuated response in hypertensive persons. A reduced day-night oscillation in sympathetic activity is also found in these patients [101]. Therefore essential hypertension results in enhanced sympathetic activity with reduced responsiveness of neural regulatory mechanisms.

Neurologic Disease

Disorders of the central and peripheral nervous system have effects on heart rate variability. The vagally and sympathetically mediated fluctuations in heart rate may be independently affected by some disorders. Vallbona first described the effect of brain damage on heart rate variability in 1965 [102]. All normal cyclic changes in heart rate are reduced in the presence of severe brain damage [103]. Heart rate variability is less accurate than the Glasgow Coma Scale in predicting outcome, but it is easily accessible and may provide information about the patient's neurologic status [104]. In serial determinations, the rate of return of normal heart rate variability may reflect the subsequent state of neuronal function.

Kuroiwa and colleagues [105, 106] described the influence of autonomic dysfunction of central origin in patients with parkinsonism, spinocerebellar degeneration, and the Shy-Drager syndrome. Patients with symptoms due to impaired parasympathetic activity (bladder-bowel disturbance) exhibited diminished short-term variability (that is, the SD of 100 successive R-R intervals). In multiple sclerosis the STV, as measured by the SD of beat-to-beat R-R interval differences [107] or by the heart rate range during deep breathing [108], is also attenuated, suggesting cardiovascular autonomic dysfunction.

Patients with polyneuropathy due to diabetes mellitus, chronic alcoholism, or the Guillain-Barre syndrome also have decreased heart rate variability. In chronic alcohol abuse, autonomic neuropathy develops due to insufficient nutrition (vitamin B₁₂ deficiency) [109]. Short-term variability (SD of beat-to-beat R-R interval differences within 30 minutes) is decreased in alcohol-dependent men with or without established vagal neuropathy according to standard cardiac autonomic function tests (for example, Valsalva maneuver and deep breathing test) [109, 110]. Due to functional recovery of the vagus nerve in patients with a relative short duration of drinking history, STV might improve with prolonged abstinence [111]. Persistent low heart rate variability with abstinence is more often found in persons with alcoholism with complications (for example, the Wernicke-Korsakoff syndrome or leg paresthesia) caused by damage in both the peripheral and central autonomic nervous system [110].

Patients with the Guillain-Barre syndrome show severe but temporary impairment of autonomic function resulting in a reduction in STV (SD of R-R interval differences) that closely follows the clinical course of the disease. Remarkably, the recovery of neurologic symptoms is preceded by the recovery of the STV [112, 113]. The decreased STV is probably due to a conduction block of the vagus nerve (segmental demyelinization) [114].

Other examples of peripheral neurologic dysfunctions that can be measured by heart rate variability analysis are parasympathetic autonomic dysfunction due to environmental neurotoxic agents (for example, organic solvents or lead) [115, 116] or vibratory tool operation [117, 118].

The effect of neurologically complete traumatic quadriplegia on spectral analysis of heart rate is informative. Quadriplegic patients have normal respiratory sinus arrhythmia, but baroreflex-related heart rate [119] and blood pressure [120] fluctuations are lacking. The disappearance of the baroreflex-related heart rate variability is caused by interruption of the spinal pathways linking supraspinal cardiovascular centers with the peripheral sympathetic outflow. The preservation of respiratory sinus arrhythmia supports the predominantly central origin of these heart rate fluctuations.

Diabetic Autonomic Neuropathy

Autonomic neuropathy is a common complication of diabetes [121]. The diagnosis of neuropathy is usually established by cardiovascular autonomic function tests like the Valsalva maneuver, deep breathing test, and isometric handgrip test [122, 123]. It has been stated that parasympathetic damage occurs more commonly than sympathetic damage [121, 122, 124]. The clinical significance of early and reliable detection of diabetic autonomic neuropathy, as possible with heart rate spectral analysis, is obvious because the 5-year mortality rate in diabetic patients with neuropathy is increased four to five times compared with diabetic patients without neuropathy [125, 126].

Heart rate variability indices used in studies concerning diabetic neuropathy have been derived from the autonomic function tests [127, 128]. For instance, the mean difference between maximal and minimal heart rate during six deep breaths with a frequency of six per minute (respiratory sinus arrhythmia or STV) is commonly used.

The decreased beat-to-beat variability during deep breathing in diabetic neuropathy was first reported by Wheeler and Watkins [129] and confirmed by many others [130, 131]. In studies comparing cardiac autonomic function tests and heart rate variability indices (based on both short [5-minute] and 24-hour ECG recordings), it was established that in diabetic patients without abnormal function tests, heart rate variability is also lowered [130, 132]. It was concluded that cardiac (parasympathetic) autonomic activity might be diminished in diabetic patients before clinical symptoms of neuropathy become evident [130].

Spectral analysis of heart rate is also a sensitive method for early detection of autonomic neuropathy. It allows a better discrimination between sympathetic and vagal influences than nonspectral heart rate variability analysis. With spectral analysis it became apparent that not only respiratory sinus arrhythmia (and thus the parasympathetic system) but also slower heart rate fluctuations are affected by autonomic neuropathy. Van Den Akker and associates [133] found a shift in the central frequency of baroreflex-related heart rate variability from 0.1 to 0.065 Hz in the heart rate power spectrum (derived from 10-minute ECG registration). This phenomenon was described earlier by Kitney and colleagues [134]. The frequency shift in baroreflex-related heart rate variability can be explained by an increase in the time delay in the blood pressure control loop. However, the conduction velocity of the postganglionic sympathetic unmyelinated C fibers in diabetic patients with autonomic neuropathy is not altered [135]. An increase in reaction time by an increase in the sympathetic delay in the neuromuscular junction or a smaller number of sympathetic fibers involved might explain the delay in the peripheral limb of the baroreflex. The findings of Van Den Akker and associates are in agreement with the results of Mackay [136], who found a maximal respiratory sinus arrhythmia (defined as heart rate range during deep breathing) at a breathing frequency of 6.3/min in controls, 5.4/min in diabetic patients without clinical evidence of neuropathy, and 4.5/min in diabetic patients with neuropathy. In accordance with the physiology of heart rate variability, a maximum respiratory sinus arrhythmia in normal adults is found at approximately 6 breaths per minute because of interaction with the intrinsic or baroreflex-related heart rate fluctuations, which have a 10-second rhythm (0.1 Hz). The breathing frequencies that elicited the highest amount of respiratory sinus arrhythmia found by Mackay correspond to spectral peaks of 0.11, 0.09, and 0.075 Hz, respectively. This finding indicates, although not recognized as such by Mackay, a shift in the frequency of baroreflex-related heart rate fluctuations as found by Van Den Akker and colleagues and thus impairment of the sympathetic nervous system. Bernardi and colleagues [137] also found a maximal respiratory sinus arrhythmia at a breathing frequency of 6/min (0.1 Hz) in controls and 5/min (0.08 Hz) in diabetics. More recent studies based on spectral analysis have shown that in diabetic autonomic neuropathy the impairment of the sympathetic pathways is at least as severe as the parasympathetic dysfunction [138-141].

Renal Failure

As with diabetic patients, patients with chronic renal failure may have dysfunctions of the autonomic nervous system. In patients with renal failure, autonomic function tests have been done [142, 143], followed by heart rate variability indices [144] and spectral analysis of heart rate [145]. Although autonomic function tests revealed predominant impairment of the parasympathetic nervous system [143], spectral analysis exhibited a strong reduction in the heart rate power spectrum at all frequency ranges, both sympathetically and parasympathetically mediated [145, 146]. The alterations in autonomic nervous system function that accompany renal failure may be due partially to circulating metabolic agents given that heart rate variability in uremic patients improves after hemodialysis [144].

Iatrogenic Influences on Heart Rate Variability

Medication

Heart rate variability can be significantly influenced by various groups of drugs. The influence of medication must be considered while interpreting heart rate variability. On the other hand, heart rate variability can be used to quantify the effects of certain drugs on the autonomic nervous system.

One of the first drugs that was investigated in relation to heart rate variability is atropine. A high dose of atropine abolishes STV (or respiratory sinus arrhythmia) and reduces LTV (or baroreflex- and thermoregulation-related heart rate variability) considerably. Propranolol, on the other hand, does not influence STV but attenuates LTV [4]. In this way the parasympathetic mediation of STV and the combined parasympathetic and sympathetic mediation of LTV has been proved.

Heart rate variability has been used recently to investigate the mode of action of drugs in adults. The vagomimetic effect on the heart of low doses of atropine has been demonstrated by calculation of the SD of R-R interval length (LTV) and of beat-to-beat R-R interval differences (STV) during sympathetic (tilt) and parasympathetic (deep breathing) stimulation. Thus low doses of atropine increase both STV and LTV, whereas fast infusion attenuates both heart rate variability indices, especially during deep breathing [147]. Scopolamine (a vagomimetic drug) increases STV and LTV. This increase is most pronounced in the vagally mediated, respiration-related heart rate variability [148].

The effects of ß-blockers and calcium channel blockers on the heart rate variability have been studied in postinfarction and hypertensive patients [149-151]. With spectral analysis it is possible to unravel and quantitate sympathetic and parasympathetic activities of these drugs and thus explain their protective effects in cardiac diseases. In normotensive adults, the ß-adrenergic blocker atenolol appears to augment vagally mediated fast fluctuations in heart rate [152]. Guzzetti and colleagues [150] studied the effect of atenolol in patients with essential hypertension. They found not only an increase in high-frequency fluctuations but also a decrease in the sympathetically mediated low-frequency oscillations. This decrease in sympathetic activity was also noticed in postinfarction patients using metoprolol [149] and in patients with heart failure using acebutolol [151]. Thus ß-blockers are able to restore the sympathetic-parasympathetic balance in cardiovascular disease.

Calcium channel blockers show various effects on heart rate variability. Diltiazem reduces low-frequency heart rate variability to the same extent as ß-blockers in postinfarction patients [149]. In contrast, nifedipine does not reduce cardiac sympathetic activity. This explains why in secondary prevention trials after myocardial infarction a reduction in mortality rate by ß-blockers, but not consistently by calcium channel blockers, is found. No effect of nifedipine [153], but a beneficial effect of diltiazem [154] in preventing, has been found in postinfarction patients.

Other drugs that affect heart rate variability are sedatives, analgesics, and anesthetics. The negative influence on heart rate variability of these drugs is predominantly established through central nervous system depression. (For a review the reader is referred to Scheffer [155] and Petrie [156].) Diazepam was one of the first sedatives studied in relationship with heart rate variability [157]. A dose-dependent attenuation in vagal tone assessed by the amount of respiratory sinus arrhythmia has been described [158]. Other benzodiazepines are also found to attenuate respiratory sinus arrhythmia [159]. They are assumed to influence the parasympathetic nervous system through a central mechanism (that is, interaction with GABA receptors that alter vagal tone) [158].

The early detection of side effects of some drugs might be investigated by heart rate variability analysis. For instance, in drug-induced cardiomyopathy a diminished respiratory sinus arrhythmia has been described before clinical symptoms develop [160].

Artificial Ventilation

Influence of artificial ventilation on heart rate has been described in patients with diminished spontaneous respiratory sinus arrhythmia, that is, in preterm infants with the respiratory distress syndrome [161], in decerebrated children [162], and in adults under anesthesia [163]. The effect of intermittent positive-pressure ventilation on heart rate is due to changes in intrathoracic pressure that activate the baroreflex. The result is a decrease in heart rate during mechanical lung inflation, in contrast to the centrally initiated increase in heart rate related to spontaneous inspiration [163].

Conclusions and Future Lines of Research

Heart rate variability has proved to be a valuable tool to investigate the sympathetic and parasympathetic function of the autonomic nervous system, especially in diabetic and postinfarction patients. Spectral analysis of heart rate has clarified the nature of diabetic autonomic neuropathy and of other neurologic disorders that encounter the autonomic nervous system. In the future, individual therapy adjustments to aim at the most favorable sympathetic-parasympathetic balance in postinfarction patients might be possible with the help of heart rate variability analysis. This analysis seems to be easily applicable in adult medicine, where physiologic influences, such as respiratory rate and behavior, can be standardized. Unfortunately, many different methods to calculate nonspectral heart rate variability indices and many different definitions for low- and high-frequency heart rate power have been used in heart rate variability studies. No consensus exists about the best heart rate variability assessment. Although nearly all indices correlate well with each other, the use of different indices makes clinical studies difficult to compare. Only one study that evaluates commercially available equipment to measure heart rate variability has been published.

Future studies are not only necessary to evaluate the commercially available devices; large prospective studies are also necessary to determine sensitivity, specificity, and predictive value of heart rate variability regarding death or morbidity in cardiac and diabetic patients. First, consensus should be obtained concerning the most favorable heart rate variability measurement procedure, that is, recording length, test circumstances (for example, time of the day, posture of the patient, breathing rate), and heart rate variability computation methods. Second, reference values for different ages are needed. Finally, the predictive value should be obtained in large longitudinal studies. An attempt to perform such a multicenter prospective study in postmyocardial infarction patients was begun in 1991 (Autonomic Tone and Reflexes After Myocardial Infarction, ATRAMI) [68].

Spectral analysis of heart rate variability can be a powerful tool to assess autonomic nervous system function. It is not only useful when studying the pathophysiologic processes in certain diseases but also may be used in daily clinical practice. Therefore it is worthwhile to continue investigating the clinical applicability of heart rate variability.