Altered Autonomic Activity with Atrial Fibrillation as Demonstrated by Non-invasive Autonomic Monitoring

Register or Login to View PDF Permissions
Permissions× For commercial reprint enquiries please contact Springer Healthcare:

For permissions and non-commercial reprint enquiries, please visit to start a request.

For author reprints, please email
Average (ratings)
No ratings
Your rating


Independent, simultaneous measures of parasympathetic (P) and sympathetic (S) activity can provide insight into the physiological processes underlying arrhythmia. The P and S method provides a more specific and sensitive measure of P and S activity than methods based solely on heart rate variability. Serial P and S assessment (ANX 3.0, ANSAR Medical Technologies, Inc., Philadelphia, PA) was performed on 270 patients (79 males, 191 females) diagnosed with atrial fibrillation (AF). No coronary artery disease (CAD) was indicated in these patients. At the time of the assessment, all patients were treated for AF. Based on the P and S assessment, sympathovagal balance (SB=S/P) was computed. If patients were found to have low SB (SB <0.4), anticholinergic therapy was added. For high SB (SB >3.0), S blockade was added. For normal SB (0.4 < SB < 3.0), therapy was not changed. Within the cohort, 161 patients were assessed both in and out of AF. Of these, 131 (81.3%) reported fewer episodes and symptoms with the selected therapy. Within the same cohort, 43 (15.9%) reported a total absence of episodes and symptoms; 39 (90.7%) of these were found to be free of AF by a comprehensive cardiac work-up. Of the group with reduced symptoms, the average number of ectopic beats dropped from 162.2±56.1 to 7.25±8.77, but their mean heart rate and blood pressure did not change significantly. The average SB of the group changed from 3.15±2.98 (abnormally high SB) to 1.78±1.91 (normal SB) after the change in therapy was introduced. From the study, it appears that it is possible to treat the P and S components based on clinical SB classifications of low, normal, and high.

Disclosure:Samanwoy Ghosh-Dastidar, PhD, is an employee of ANSAR. Joe Colombo, PhD, is an employee and co-owner of ANSAR. The remaining authors have no conflicts of interest to declare.



Correspondence Details:Joe Colombo, PhD, 240 South Eighth Street, Philadelphia, PA 19107. E:

Copyright Statement:

The copyright in this work belongs to Radcliffe Medical Media. Only articles clearly marked with the CC BY-NC logo are published with the Creative Commons by Attribution Licence. The CC BY-NC option was not available for Radcliffe journals before 1 January 2019. Articles marked ‘Open Access’ but not marked ‘CC BY-NC’ are made freely accessible at the time of publication but are subject to standard copyright law regarding reproduction and distribution. Permission is required for reuse of this content.

Heart rate variability (HRV) is considered to be a means of assessing autonomic function. However, HRV alone provides mixed measures of autonomic function1–11 and, as a result, cannot provide much insight into the autonomic involvement in arrhythmia.1,6,10 When HRV analysis is employed in conjunction with respiratory activity (RA) analysis, it has been shown to independently, simultaneously, and non-invasively quantify activity in the parasympathetic (P) and sympathetic (S) branches of the autonomic nervous system (ANS).12–16 With independent, simultaneous measures of the two ANS branches, a true measure of sympathovagal balance (SB=S/P) becomes available. Therefore, it is suggested by the authors that the analysis of HRV along with RA may provide insight into autonomic involvement in arrhythmia.


The analysis of HRV along with RA forms the basis of P and S monitoring, which resolves several physiological and mathematical inconsistencies observed with methods based solely on HRV, as described in several papers.12–19 One issue especially relevant to the problem of P and S assessment in atrial fibrillation (AF) is that of noise reduction. Methods based on HRV alone use unequal and fixed frequency ranges, which causes the ratio of the low-frequency power to the high-frequency power to be unequally weighted. The unequal influence of noise combined with the wider high-frequency window makes it difficult to eliminate noise from the ratio.1 P and S assessment uses a narrower adaptive frequency band to assess the higher-frequency activity.16,18 The P and S terms comprising the ratio (SB) have similar weights, thus facilitating noise rejection during the computation.

In addition, the signals used for P and S monitoring, i.e. the instantaneous heart rate (derived from the heart beat interval) and RA, are inherently non-stationary. The degree of non-stationarity in the instantaneous heart rate is increased by arrhythmia. The fast Fourier transform (FFT) that has traditionally been used for spectral analysis to obtain frequency domain measures based solely on HRV1 is not suitable for analyzing non-stationary signals unless the signals are of sufficiently long duration. This implies that during FFT analysis the effect of events such as arrhythmia is distributed over large segments of the signal, which compromises the quality of the information in the signal.

In recent years, time–frequency analysis using continuous wavelet transforms (CWT) have become the method of choice for analyzing non-stationary signals.16–18 CWT analysis can accurately characterize sharp transients in the signal and can localize these sharp transient events to the time of occurrence. As a result, effects of events such as arrhythmia can be isolated, thus retaining the integrity of the remainder of the signal, which yields clear autonomic information.

The additional autonomic information may provide deeper insight into the processes of arrhythmia, and may facilitate therapy plans customized for the patient’s physiology, thereby improving outcomes. AF is the specific arrhythmia analyzed in this preliminary study. We investigated whether SB could indicate an autonomic component to AF. We followed patients with AF and no coronary artery disease (CAD). If SB was abnormal, low-dose ANS therapy was titrated in addition to the existing therapy. If SB was normal, no additional therapy was added.


Serial P and S assessment (ANX 3.0, ANSAR Medical Technologies, Inc., Philadelphia, PA) was performed on 270 patients (79 males, 191 females, average age 56.2±15.7 years) from four ambulatory cardiology clinics diagnosed with AF. No CAD was indicated in these patients. On average, the assessments were performed 6.23±0.57 months apart and patients had 3.1±0.22 tests over the three-year period of this study. The P and S assessment included six challenges: (A) a five-minute resting (initial) baseline; (B) a one-minute paced breathing exercise (paced at six breaths per minute) to challenge the P nervous system; (C) a one-minute baseline; (D) a series of five short Valsalva maneuvers (15 seconds or less) over one minute and 35 seconds to challenge the S nervous system; (E) a two-minute baseline; and (F) a quick postural change (stand) followed by five minutes of quiet head-up posture (standing).

Non-invasive, quantitative assessment of both ANS branches is independently and simultaneously accomplished by means of concurrent CWT-based time–frequency analyses of HRV and RA. HRV is derived from the electrocardiogram (ECG) and synchronized with RA, which is derived from the impedance plethysmography signal. Central vagal influence on HR is determined first as a frequency in the RA spectrum. This frequency is then translated to the HRV spectrum to locate P (vagal) activity and compute its magnitude. The S activity is then computed as the power of a lower-frequency range in the HRV spectrum. The time–frequency analysis technique, CWT, accounts for the time and frequency shifts inherent in the non-stationary P and S signals. SB is computed as the ratio of the average S to average P activity over the time period of analysis, i.e. four seconds.

At the time of assessment, all patients were treated for AF. If patients were found to have an abnormal SB, low-dose autonomic therapy was added to their anti-arrhythmic medication depending on the patient history. Sample therapy plans for patients within this population are as follows: for low SB (SB <0.4), indicating P excess (PE), tiotropium or nortriptyline was added as an anticholinergic; for high SB (SB >3.0), indicating S excess (SE), carvedilol, an angiotensin-converting enzyme (ACE) inhibitor, or an angiotensin receptor blocker (ARB) was added as a S-blocker. For normal SB (0.4


Out of the 270 patients included in this cohort (see Table 1), 22 (8.15%) were in AF during all the assessments. The mean heart rate and blood pressure were 86.2±21.9bpm and 121.5/71.3±13.2/5.0mmHg, respectively. There were 87 (32.22%) patients who were never in AF during an assessment. These patients had a mean heart rate and blood pressure of 68.5±10.7bpm and 128.6/76.1±17.0/9.9mmHg, respectively. Of the remaining 161 (270–22–87) who had some assessments in AF and some without AF, 131 (81.3%) reported fewer episodes and symptoms subsequent to the addition of low-dose ANS therapy. Of the same 161, 43 (15.9%) reported a total absence of episodes and symptoms subsequent to the addition of low-dose ANS therapy; 39 (90.7%) of these were found to be free of AF by comprehensive cardiac work-up. In the group that reported fewer symptoms, the average number of ectopic beats dropped from 162.2±56.1 to 7.25±8.77 after the change in therapy. Their mean heart rate and blood pressure did not change significantly: mean heart rate changed from 76.7±16.3 to 74.5±18.2bpm, and mean blood pressure changed from 125.0/76.0±18.5/10.0 to 128.1/74.9±24.1/9.8mmHg.

The average SB of the group that was in AF during every assessment was 0.76±0.13, indicating a mild (yet normal) P tendency. The average SB for the group that was never in AF during an assessment was 1.39±1.21, indicating a mild (yet normal) S tendency. In the group that reported fewer symptoms, the average SB changed from 3.15±2.98 (abnormally high SB) to 1.78±1.91 (normal SB) after the change in therapy.


It is accepted that frequency–domain analysis of HRV alone yields valid measures of autonomic activity.1 However, these measures do not provide accurate information on how much of that autonomic activity is P and how much is S.2–11 It should be noted that these measures do not directly measure ANS activity. Instead, they are computed to provide an estimate of the effect of ANS activity on the instantaneous heart rate (IHR). Herein lies the limitation of measures based solely on HRV.

Since the ANS has two branches (P and S), the system can be conceptualized as having two independent variables. IHR is one dependent variable in the system. Regardless of whether the measure is time– or frequency–domain, it is mathematically impossible to compute the values of the two independent variables when only the dependent variable is known. A second such relationship is needed.

In the early 1980s, researchers at the Massachusetts Institute of Technology (MIT)12–15 proposed that the second relationship should be the influence of respiration on P (vagal) activity. Cardio-vagal activity follows the respiratory rhythm, and the primary P influence in the human body is mediated by the vagus nerve. Based on these facts, it is deduced that in the frequency domain, the respiratory rate can be used to correctly identify the frequency band corresponding to P activity. Once the P frequency band is identified, the S frequency band can be easily identified as the range between 0.04Hz and the lower limit of the P band, or 0.15Hz, whichever is lower.1

The Fourier transform is very commonly used for the frequency–domain or spectral analysis of signals in general, but the technique suffers from a significant shortcoming: the Fourier transform is based on the assumption that the signal is periodic and infinite in duration. As a result, features that are localized in time cannot be represented accurately. Moreover, the transform is computed over the entire duration of the signal, which leads to an averaging effect on the features. Therefore, information about the time dependence of the signal is completely lost. These shortcomings are a significant part of the reason HRV cannot provide much insight into autonomic involvement in arrhythmia.1,6,10

In order to achieve some degree of time localization, most practical signal processing applications involve a windowed FFT that employs a fixed window size. The window is usually implemented as a sliding window and is used to analyze the signal piece-wise in overlapping sections. The selection of this appropriate window size is important. If the window size is too large, the spectral signatures of localized events can be missed. If the size is too small, the assumption of infinite signal duration is violated, leading to an inaccurate analysis. The appropriate window size becomes an especially significant issue for physiological signals such as IHR due to their non-stationary nature.18 When the signal is non-stationary (i.e. the statistical characteristics of the signal change over the time period of analysis), a fixed window cannot accurately characterize all aspects of the signal. In such cases, using sufficiently long durations may yield better estimates as long as excessive averaging does not obscure signal details. With non-stationary signals, the criteria for ‘sufficiently long’ change over time within the signal being analyzed. This would require the window to change during the analysis.

The CWT can be scaled and shifted in an infinite number of ways throughout the analysis, thereby enabling an accurate characterization of details within the signal at multiple levels of resolution. The wavelet basis acts as a window that adapts itself to the signal in real time, which essentially eliminates the fixed window limitations described earlier for Fourier transforms. The wavelet is stretched to characterize components of the signal that have a lower frequency and compressed to characterize those that have a higher frequency. This enables accurate identification of spectral signatures of local events and their localization in time. Moreover, all this becomes possible with much shorter signal durations, which considerably reduces the length of the observation times and therefore the clinical testing time, while improving the accuracy of any real-time online analysis of the signal.

The key to monitoring the two branches of the ANS in real time is the ability to not only capture sharp transients, but also identify their time localization. Without this information on time localization, a true real-time analysis is not possible. As explained earlier, the window size for the FFT must be sufficiently large in order to accurately characterize the frequency components. Now, if there is a frequency change of interest within this window, it is not possible to identify the precise time of the change. Moreover, due to the averaging effect, a sharp transient change is redistributed over the entire window, which implies that the estimated magnitude of the change may also be questionable. The CWT is able to adapt much faster and more easily to the non-stationarity in the signal. Therefore, the CWT can quickly and accurately identify and quantify localized responses, such as changes in P and S activity during clinical and physiological challenges.16,18,19

The example provided in Figure 1 is an HRV spectral curve obtained from the resting baseline period of a paroxysmal AF patient in fibrillation at the time of testing (see Figure 2). The patient is a 49-year-old female, 157.5cm tall, weighing 54.5kg. She experienced 193 ectopic (arrhythmic) beats throughout the 15.5-minute ANS assessment (see the cardiogram in Figure 2); the letters A to F correspond to the six challenges of the test as separated by the vertical broken lines (see ‘Methods’ section for more detail). The patient presents with amlodapine, avapro, lasix, metoprolol, plavix, and warfarin on board. The frequency ranges for methods based solely on HRV are presented in Figure 1 as a dark gray area for the low-frequency (LF) measure and a light gray area for the high-frequency (HF) measure. For analysis using the P and S method, areas are represented as blue for the P measure and red (overlapping the dark gray area) for the S measure. This spectrum is selected as a representative four-second segment from the five-minute resting baseline (section A in Figure 2).

This selection represents the patient’s average resting baseline response. The LF and HF (measures based on HRV alone) results are 849msec2 and 4,468msec2, respectively. These measures are based on the beat to beat interval (heart beat interval or HBI) signal in milliseconds derived from the ECG. These two measures yield an L/H ratio of 0.19. This ratio would suggest a significant P component to the arrhythmia and perhaps lead to a change in therapy plan according to the criteria proposed in the ‘Methods’ section. The P and S method results in an S activity level of 11.88bpm2 and a P activity level of 13.40bpm2. Both measures are considered to be high,16 indicating cardiac stress. The P and S measures result in an SB of 0.89, indicating normal autonomic balance and suggesting no autonomic contribution to the arrhythmia.

The differences between the two methods are apparent from Figure 1. The LF (based on HRV alone) and S (based on the P and S method) measures are equal. The dark gray and the red areas for the two methods exactly overlap. This is a coincidence due to the fact that the patient’s respiratory frequency is sufficiently high to separate the S and P activities appropriately into the LF and HF areas. However, the broad, fixed frequency band for the HF captures much more activity than is considered as P activity.12–15 This additional activity is related to the ‘noise’ associated with the arrhythmia and the reason why methods based solely on HRV are unable to provide much insight into autonomic involvement in arrhythmia. This noise inflates the power of the HRV spectrum and affects the LF and HF as well as the S and P measures. Since the HF band (0.25Hz bandwidth) is significantly wider than the LF band (0.11Hz bandwidth), the L/H ratio is significantly weighted toward the HF measure, resulting in a false-positive for P excess. The more equally matched P and S measures from the P and S method (S derived from a 0.11Hz-bandwidth window in this example and P derived from a 0.12Hz-bandwidth window in this example) help to reduce the effect of the noise on the SB measure. If there was indeed a P component to the arrhythmia, the deep breathing segment of the ANS assessment study (section B in Figure 2) would be markedly abnormal, showing significant spike activity, as observed throughout the rest of the study. That is not the case, which further highlights the lack of a P component to the AF in this patient.


The P and S method provides a more specific and sensitive measure of P and S activity than the methods based on HRV alone. SB obtained from P and S assessment based on a CWT-based concurrent analyses of HRV and RA may determine the presence of autonomic components in AF. From the study, it appears that it is possible to treat the P and S components based on clinical SB classifications of low, normal, and high. Low SB (SB <0.4) indicates PE and a possible P component to the arrhythmia, which can potentially be treated with low-dose anticholinergics. High SB (SB >3.0) indicates SE and a possible S component to the arrhythmia, which can potentially be treated with low-dose adrenergic blockade. Normal SB indicates proper autonomic balance and suggests that there may not be any autonomic component to the arrhythmia. Further longitudinal study of paroxysmal AF patients, while the patient is both in and out of AF, is recommended. In this way, each individual patient is his or her own control. This study has been proposed and will be the subject of further research.


  1. Malik M, The Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology, Heart rate variability, standards of measurement, physiological interpretation, and clinical use, Circulation, 1996;93:1043–65.
    Crossref | PubMed
  2. Alcalay M, Izraeli S, Wallach R, et al., Paradoxical pharmacodynamic effect of atropine on parasympathetic control: study by spectral analysis of heart rate fluctuations, Clin Pharmacol Ther, 1992;52:518–27.
    Crossref | PubMed
  3. Brown TE, Beightol LA, Koh J, Eckberg DL, Important influence of respiration on human R–R interval power spectra is largely ignored, J Appl Physiol, 1993;75(5): 2310–17.
  4. Hayano J, Mukai S, Sakakibara M, et al., Effects of respiratory interval on vagal modulation of heart rate, Am J Physiol Heart Circ Physiol, 1994;267(36):H33–40.
  5. Eckberg DL, Sympathovagal balance. A critical appraisal, Circulation, 1997;96(9):3224–32.
    Crossref | PubMed
  6. Low P (ed.), Clinical Autonomic Disorders, 2nd Edition, New York: Lippincott Press, 1997.
  7. Eckberg DL, Physiological basis for human autonomic rhythms, Ann Med, 2000;32:341–9.
    Crossref | PubMed
  8. Badra LJ, Cooke WH, Hoag JB, et al., Respiratory modulation of human autonomic rhythms, Am J Physiol Heart Circ Physiol, 2001;280(6):H2674–88.
  9. Cammann H, Michel L, How to avoid misinterpretation of heart rate variability power spectra?, Comp Methods Programs Biomed, 2002;1:15–23.
    Crossref | PubMed
  10. Freeman R, Assessment of cardiovascular autonomic function, Clin Neurophysiol, 2006;117(4):716–30.
    Crossref | PubMed
  11. Grossman P, Taylor EW, Toward understanding respiratory sinus arrhythmia: relations to cardiovagal tone, evolution and biobehavioral functions, Biol Psychol, 2007;74:263–85.
    Crossref | PubMed
  12. Akselrod S, Gordon D, Ubel FA, et al., Power spectrum analysis of heart fluctuations: a quantitative probe of beat to beat cardiovascular control, Science, 1981;213:220–2.
    Crossref | PubMed
  13. Akselrod S, Gordon D, Madwed JB, et al., Hemodynamic regulation: investigation by spectra analysis, Am J Physiol, 1985;249:H867–75.
  14. Akselrod S, Gordon D, Madwed JB, et al., Hemodynamic regulation in SHR: investigation by spectral analysis, Am J Physiol, 1987;253:H176–83.
  15. Akselrod S, Spectral analysis of fluctuations in cardiovascular parameters: a quantitative tool for the investigation of autonomic control, Trends Pharmacol Sci, 1988;9:6–9.
    Crossref | PubMed
  16. Aysin B, Aysin E, Effect of respiration in heart rate variability (HRV) analysis, IEEE Engineering in Medicine and Biology Society Conference, New York, NY, 2006.
  17. Olivera M, Santos-Bento M, Xavier R, et al., Wavelet analysis for the evaluation of cardiovascular autonomic nervous response to postural change in healthy subjects in relation to age, Second Joint Meeting of the European Federation of Autonomic Societies and the American Autonomic Society, Vienna, Austria, Clinical Autonom Res, 2007;17(5):301.
  18. Aysin B, Aysin E, Colombo J, Comparison of HRV analysis methods during orthostatic challenge: HRV with respiration or without? IEEE Engineering in Medicine and Biology Conference, Lyons, France, 2007.
  19. Arora RR, Bulgarelli RJ, Ghosh-Dastidar S, Colombo J, Autonomic mechanisms and therapeutic implications of postural diabetic cardiovascular abnormalities, J Diabetes Sci Technol, 2008;2(4):568–71.