Return to December 2000Table of Contents BEDSIDE PHYSIOLOGY Hidden rhythms in the heart rate record: a primer on neurocardiology Ernest L. Fallen, MD Dr. Fallen is Professor of Medicine with the Division of Cardiology, Department of Medicine, McMaster Univer- sity, Hamilton, Ont. Medical subject headings: cardiovascular physiology; Cheyne- Stokes respiration; electrocardiography; Fourier analysis; heart rate; neurotransmitters; nonlinear dynamics; syncope, vasovagal Clin Invest Med 2000;23(6):387-94. © 2000 Canadian Medical Association Introduction fying loss of physiologic regulation. Simply stated, the more variable are successive interbeat intervals Although we appear to think and act in the time do- the healthier is the person. How do we know this? main events, the neurobiologic processes that govern our existence are played out in the frequency do- Physiologic correlates of heart rate variability main.* In other words, our brain is constantly emit- ting signals that oscillate at varying frequencies. The Contained within any heart rate record (ECG) are frequency content of these neural rhythms reveals hidden rhythms that reflect autonomic regulation of marked variability even under steady state conditions sinus node function.1–3 Before describing these in healthy people. It is only when these oscillations rhythms, it should be recalled that the sinoatrial node begin to lose their variability (i.e., become more peri- is the target of competing bombardment by both odic), that central control of individual organ func- sympathetic and vagal efferent nerve impulses. The tion is endangered. Consider the following bedside resultant interbeat interval (the inverse of heart rate) observations: the patient with end-stage heart or lung at any point in time is simply the net effect, or bal- disease whose breathing pattern alternates between ance, between vagal neurotransmission (acetyl- hyperpnea and apnea (Cheyne-Stokes respiration); choline) and sympathetic neurotransmission (norepi- the patient with advanced biventricular failure whose nephrine) at the neuroeffector junction. The arterial pulse volume waxes and wanes with succes- competition occurs at both ends. At the sinus node sive beats (pulsus alternans); or the patient with criti- site, there is what is called accentuated antagonism cal coronary artery stenosis whose T-wave morphol- between release of the neurotransmitters. This means ogy on the electrocardiogram (ECG) begins to that the strength (amount) of neurotransmitter re- alternate from beat to beat prior to ventricular fibrilla- lease from either a vagal or a sympathetic stimulus is tion. In all 3 instances, the common denominator is a highest when the opposing agonist is already domi- shift from complex variability of the beat-to-beat os- nant.4 In any event, the release of either neurotrans- cillatory rhythm to a periodicity akin to a simple sine mitter is said to be proportional to so-called auto- wave, a dangerous pattern signi- nomic tone (i.e., the rate of nerve impulses as viewed along the course of an efferent nerve in the time domain). To understand what is happening at *The time domain refers to events in unit time. The frequency domain refers to the combination of frequencies that underlie the central site it is important to distinguish between events such as the beat-to-beat heart rate variability. tonic and phasic properties of nerve conduction be- Clin Invest Med • Vol 23, no 6, December 2000 387 Fallen cause we are more concerned with heart rate as a sympathetic and vagal input to all or any portion of measure over time than with “instantaneous” heart the signal.5 rate, an implausible term. Frequency domain measures such as the power If we examine a recording of nerve impulses we spectrum density of HRV provide information on would first see a train of repetitive bursts of electrical how the variance or power in the heart rate signal spikes. This is often referred to as its tonic activity. distributes as a function of time.3 Methods include On more careful inspection an oscillatory or phasic the nonparametric fast Fourier transform1 (e.g., pattern to the signal can be discerned, a property indi- Blackman–Tukey method) and the parametric autore- cating that the train of nerve impulses has a fre- gressive model technique.2 The fundamental principle quency modulation not dissimilar to that of an FM is based on a conversion or demodulation of a time radio signal. If we then visualize a succession of R series such as a continuous series of R–R interval in waves in a continuous ECG record as containing spe- the time domain (a tachygram) into its frequency cific oscillatory or phasic patterns we can dial in the components (Fig. 1). What first appears as a random various FM bands that make up the power spectrum.3 series of disconnected R-R spikes on the time axis is It turns out that these hidden phasic rhythms re- miraculously transformed into a smooth spectrum flect autonomic control of sinoatrial function.1,2 By with 2 distinct peaks. The low frequency band (0.05 applying mathematical algorithms using standard to 0.15 Hz) represents primarily sympathetic modula- signal-processing techniques we can easily tease out tion in the resting supine state,7 whereas the high fre- and separate the relative influences of sympathetic quency band (0.15 to 0.35 Hz) is essentially respira- and vagal modulation of heart rate and blood pres- tory driven sinus arrhythmia, a manifestation of vagal sure variability.5 modulation, as this band is completely abolished with high dose atropine.1 Hence, the ratio of the low fre- Methodologies quency to high frequency power provides an expres- sion of sympathovagal balance at any steady state The 2 principal approaches for measuring heart rate junction in time. The major advantage of the fre- variability (HRV) are by time domain and frequency quency domain approach, therefore, is to provide a domain analyses. In the time domain method, simple window through which there is a better view of phys- descriptive statistics are used to obtain a measure of iologic control of sinoatrial function compared with HRV from many interbeat intervals.6 For example, time domain measures. Its principal disadvantage is from 24 hours of a continuous ECG recording the requirement for strict steady state conditions to (Holter monitoring) the standard deviation of all suc- obtain, making it valid for short-term recordings (2.5 cessive conducted R–R intervals originating in the to 5 minutes) only.5 There are other more complex sinus node can be derived; the SDNN. These de- methods, including nonlinear analyses such as the scriptive statistics mainly capture the overall vari- various methods based on chaos theory8 and time ance or total power contained within the 24-hour in- variant techniques.9,10 The latter permit noise-free terbeat signal. There are other time domain measurements during physiologic perturbations. To measures, known as differencing parameters, that date, the clinical utility of these intriguing nonlinear mainly reflect vagal activity. These include the root methods awaits further testing and application. mean square standard deviation of successive R-R interval differences and the proportion of successive A word on clinical utility interbeat intervals that vary more than 50 millisec- onds. The advantages of time domain methods are It has been shown that a patient recovering from an their simplicity, reproducibility and proven prognos- acute myocardial infarction (MI) with an SDNN less tic power in certain clinical disorders (vida infra). than 50 milliseconds has a fourfold increase in the Their disadvantages include non-steady state of the risk of sudden death within 1 year compared with signal in the free living ambulatory setting and the the patient whose SDNN is greater than 100 mil- difficulty in ascertaining the relative contributions of liseconds (Fig. 2). Thus, a low HRV, whether mea- 388 Clin Invest Med • Vol 23, no 6, décembre 2000 Neurocardiac regulatory mechanisms sured in the time or frequency domain, stratifies the established risk factors in post-MI patients. Yet, post-MI patient into a group at high risk for malig- whereas HRV has reasonable prognostic power, its nant arrhythmias.11,12 This information is not only accuracy as a screening tool is weak if employed as prognostic but offers incremental value with other a single test. With regard to sudden cardiac death, Fig. 1: Steps in the derivation of the power spectrum of heart rate (HR) variability. The R-R interval of the electrocardiogram (ECG) is transformed into a tachygram (instantaneous HR series) from which the fre- quency components are computed, yielding the power spectrum by either the Blackman–Tukey (fast Fourier transform) or the smoother autoregressive modelling technique. Clin Invest Med • Vol 23, no 6, December 2000 389 Fallen there is a wealth of laboratory evidence supporting distinct spectral components of the heart rate signal the association of autonomic dysregulation with a indicating at least partial reinnervation.15,16 To what predisposition to lethal arrhythmias.13,14 A few at- extent early reinnervation is clinically beneficial re- tempts have been made (without success so far) to mains to be explored, but at least we have a simple forecast the onset of sustained ventricular tachycar- noninvasive method by which to monitor these pa- dia or fibrillation by spot-checking for a specific tients. In patients with diabetic autonomic neuropathy change in HRV before the event. One confounding there is, as with heart failure, a significant reduction variable is the progressive age-dependent decline in of HRV.17 Moreover, there is a failure to increase the HRV in healthy people, creating a greater and more low-frequency power during orthostasis, signifying widespread data overlap in the elderly. Another limi- severe impairment of baroreceptor sensitivity. HRV tation is the poor signal-to-background noise ratio can be modified by a variety of therapeutic interven- noted in some disease states such as advanced con- tions.