Vernon Bond Jr ¹, Bryan H Curry ², Krishna Kumar ³, Sudhakar Pemminati ⁴, Vasavi R Gorantla ⁵, Kishan Kadur ⁶, Richard M Millis ^{6 *}

Acupuncture treatments are safe and effective for a wide variety of diseases involving autonomic dysregulation. Heart rate variability (HRV) is a noninvasive method for assessing sympathovagal balance. The low frequency/high frequency (LF/HF) spectral power ratio is an index of sympathovagal influence on heart rate and of cardiovascular health. This study tests the hypothesis that from rest to 30% to 50% of peak oxygen consumption, the nonlinear Conte-Zbilut-Federici (CZF) method of computing the LF/HF ratio is a more reliable index of changes in the HRV than linear methods are.

The subjects of this study were 10 healthy young adults. Electrocardiogram RR intervals were measured during 6-minute periods of rest and aerobic exercise on a cycle ergometer at 30% and 50% of peak oxygen consumption (VO_2peak).

The frequency domain CZF computations of the LF/HF ratio and the time domain computations of the standard deviation of normal-to-normal intervals (SDNN) decreased sequentially from rest to 30% VO_2peak (P < 0.001) to 50% VO_2peak (P < 0.05). The SDNN and the CZF computations of the LF/HF ratio were positively correlated (Pearson’s r = 0.75, P < 0.001). fast Fourier transform (FFT), autoregressive (AR) and Lomb periodogram computations of the LF/HF ratio increased only from rest to 50% VO_2peak.

Computations of the LF/HF ratio by using the nonlinear CZF method appear to be more sensitive to changes in physical activity than computations of the LF/HF ratio by using linear methods. Future studies should determine whether the CZF computation of the LF/HF ratio improves evaluations of pharmacopuncture and other treatment modalities.

acupuncture, autoregressive, autonomic modulation, exercise, fast Fourier transform, heart rate variability

This is an Open-Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/) which permits unrestricted noncommercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Evidence is emerging that acupuncture and pharmacopuncture are efficacious treatments for cardiovascular disease, as well as a variety of other diseases involving autonomic dysregulation [1-11]. Frequency-domain (spectral) methods for computing heart rate variability (HRV) are commonly used to evaluate the power of sympathetic- parasympathetic (sympathovagal) autonomic neural modulations of heart rates. Low-frequency power (LF) is a measure of mainly sympathetic influence, with some vagal baroreceptor influences, over relatively long time intervals while high-frequency power (HF) is a measure, mainly, of the vagal influences of respiratory sinus arrhythmia (RSA) over short time intervals [12]. The significance of very low frequency (VLF) power is disputable, but appears to be a mixture of sympathetic modulations with metabolic, thermogenic and emotional influences [13]. LF is usually evaluated in the frequency range 0.04 − 0.15 Hz, HF in the range 0.15 − 0.4 Hz [14, 15] and VLF in the range 0.0033 − 0.04 Hz [13, 14]. The LF/HF ratio is used to indicate the amount of sympathovagal modulation of the instantaneous heart rate [14, 16].

Linear spectral HRV analyses include the nonparametric fast Fourier transform (FFT) and the parametric autoregressive (AR) methods [17]. The AR spectrogram is smoother than the FFT spectral power density, thereby providing the advantages of better resolution with longer time segments and easily distinguishable peaks [18]. Combining frequency-domain methods with time-domain methods, the latter largely based on the standard deviation or root mean square of the standard deviation of the electrocardiogram RR interbeat intervals, have led to many breakthroughs in understanding HRV [18]. Altered HRV, measured by using both time-domain and frequency-domain analyses, are now known to occur physiologically. Aging has been characterized by no change in the overall LF/HF ratio, with complementary decrements in both the LF and the HF, and hypertension is characterized by an increase in the magnitude of the LF component and a decrease in the magnitude of the HF component compared to those components in normotensive individuals [19, 20]. Individuals with ischemic cardiovascular disease, including myocardial infarction, show similar signs of sympathetic upregulation as those with hypertension [19, 20]. Inaccuracies in the computation of the LF/HF ratio are known to occur when data are not segmented appropriately according to the physiological state and can be effectively eliminated by using the Lomb periodogram computation of the LF/HF ratio [21, 22].

Inaccuracies can be amplified by physical activity [23] and might produce erroneous interpretations with respect to evaluating sympathovagal modulations of the heart rate associated with exercise [24]. The present study was, therefore, designed to test the hypothesis that, across different levels of physical activity from rest to 50% of peak oxygen consumption (VO_2peak), a nonlinear method of computing the LF/HF ratio may overcome the methodological inconsistencies associated with linear computations of that ratio for quantifying the autonomic effects of acupuncture and pharmacopuncture therapies.

Ten healthy, young-adult, African-American participants, 6 females and 4 males, volunteered for this study. All subjects were free of cardiac and pulmonary disease. The anthropometric and the physiological characteristics of the study subjects are presented in (Table 1). Before participation, each subject was familiarized with the experimental procedure and risks. Approval was obtained from the Institutional Review Board at Howard University, and written informed consent was obtained from all participants.

The subjects performed two laboratory sessions of exercise approximately one week apart in the upright position on an electronically-braked bicycle ergometer (Ergoline 800, Sensormedics Corp., Loma Linda, CA). In the first session, the subjects performed an incremental test to determine the peak oxygen uptake (VO_2peak). The initial power was set at 20 W and was increased by 20 W every 3 minutes to volitional fatigue. VO_2peak was defined as the VO₂ obtained during the last minute of the incremental exercise test. The VO_2peak test results were used to determine the subjects’ specific work rates corresponding to 30% and 50% VO_2peak for the second session. During the second session, the subjects rested on the ergometer for 10 minutes, followed by 10 minutes of work both at 30% VO_2peak and at 50% VO_2peak.

Height and weight were measured using standard laboratory procedures. Percent body fat was determined using Hologic QDR 4500/W dual-energy X-ray absorptiometry scans (Hologic, Inc., Marlborough, MA). Beat-to-beat electrocardiogram RR intervals were recorded using a BIOPAC MP150 high-speed data-acquisition system (BIOPAC Systems, Inc., Aero Camino, CA), with a sampling frequency of 1,000 Hz from the ECG signal, providing an accuracy of 1 ms for each RR interval. RR interval recordings were visually inspected for undesirable beats and noise by using BIOPAC Systems Acknowledge 9.6 Software. The RR intervals for the last 6 minutes at rest and during submaximal work at 30% VO_2peak and at 50% VO_2peak were analyzed. Expired gas fractions of VO₂ and VCO₂ and minute ventilation were measured using the Physio-Dyne Max II metabolic system (Physio Dyne Instrument Corp., Massapequa, NY). Expired gases were measured and averaged every 60 seconds. Before the VO_2peak test, the O₂ and the CO₂ analyzers were calibrated with medical grade 21.0% and 5% O₂ and CO₂, respectively. Calibration of the turbine flow meter of the metabolic system was performed with a 3-L syringe.

A spectral power density analysis of HRV was used to derive measures of autonomic modulation [25]. Frequency- domain HRV features followed the guidelines of the European Task Force [12]. HRV in the time domain was evaluated by using the standard deviation of the normal-normal electrocardiogram RR intervals (SDNN), and that in the frequency domain was evaluated by using the linear FFT, AR and Lomb periodogram methods and the nonlinear conte-zbilut-federici (CZF) method to compute the LF/ HF ratio by using Nevrokard HRV software (Nevrokard, HRV, Medistar, Ljubljana, Slovenia). The power spectrum of electrocardiogram RR intervals within the 0.15- to 0.4- Hz range was defined as the HF component and primarily represented the parasympathetic (vagal) modulation of respiratory sinus arrhythmia. The LF component (0.04 − 0.15 Hz) is considered to be a mixture of both vagal and sympathetic signals [26, 27].

Differences between the time-domain SDNN and the frequency- domain linear FFT, AR, and Lomb periodogram methods and the nonlinear CZF method for computing the LF/HF ratio, expressed in normalized units (nu), were evaluated at the following three levels of steady-state physical activity: resting conditions, aerobic exercise under 30% VO_2peak conditions, and aerobic exercise under 50% VO_2peak conditions.

Descriptive statistics were expressed as means ± standard errors. The statistical significances of the differences between the methods of computing the LF/HF ratio were evaluated by using an analysis of variance (ANOVA) at each level of physical activity. Statistical significance was set at P ≤ 0.05.

The anthropomorphic and the physiological characteristics of the study group show that they represent healthy, sedentary, young adults (Table 1). The effects of using the FFT, AR, Lomb periodogram, and CZF methods to compute the HRV parameters at each of the three levels of physical activity were determined (Table 2). CZF computations of the LF/HF ratio were found to be significantly smaller sequentially across the resting control (0.62 ± 0.04), the 30% VO_2peak (0.41 ± 0.03, P < 0.01 vs. control) and the 50% VO_2peak (0.34 ± 0.02, P < 0.001 vs. control and P < 0.05 vs. 30% VO_2peak) conditions. The activity-related changes in the CZF-computed LF/HF ratio were also found for the SDNN values from rest (62 ± 9 ms) to 30% VO_2peak (30 ± 4 ms, P < 0.001 vs. control) to 50% VO_2peak (25 ± 4 ms, P < 0.0001 vs. control and P < 0.05 vs. 30% VO_2peak). The SDNN values for the LF/HF ratio and the ratios computed by using the CZF method were significantly correlated (Pearson’s r = 0.75, P < 0.001). The LF/HF ratios at 50% VO_2peak computed by using the FFT, AR, and Lomb periodogram methods were significantly greater than the value obtained using the control computations (FFT and AR P < 0.01, Lomb P < 0.05), but were not significantly different from the LF/HF ratio computed for the 30% VO_2peak condition P > 0.1).

Whereas all the linear computations of the LF/HF ratio (FFT, AR and Lomb) were unable to differentiate between the two levels of exercise, 30% and 50% VO_2peak, the AR computations were the only linear computations that differentiated between the resting control and 30% VO_2peak conditions. The SDNN time-domain measurement and the nonlinear CZF computation of the LF/HF ratio, both indicators of the RR interbeat interval variability, differentiated between the three levels of activity.

Current noninvasive, linear computations of the LF/HF ratio, a useful index of sympathovagal balance, are often inconsistent from day to day in the same individuals under the same environmental conditions and are often exaggerated during aerobic exercise [21-24]. Exercise stress testing is the most widely-used method for assessing cardiovascular health, and the ability to quantify the LF/HF ratio during exercise stress testing may provide a robust tool for assessing the effects of acupuncture, pharmacopuncture, and other treatment modalities. The main purpose of this study was to determine whether a novel nonlinear (CZF) method for computing the LF/HF ratio, a reliable measure of sympathetic-parasympathetic influences on the heart rate that has been extended to a general assessment of autonomic neural signaling, was superior to the linear FFT, AR and Lomb methods.

This is the first study to compare FFT, AR, Lomb and CZF computations of the LF/HF ratio across the range of activity from rest to 30% to 50% of peak oxygen consumption. The main finding of this study is that only CZF computations of the LF/HF ratio, which decreased sequentially, thereby indicating less variability of the RR interbeat interval, were significantly different across the three levels of activity. FFT, AR and Lomb computations of the LF/HF ratio increased from rest to 50% of peak oxygen consumption, but were not significantly different from 30% to 50% of peak oxygen consumption, thereby reliably indicating increased sympathetic modulation of heart rates only from rest to 50% of peak oxygen consumption. This finding on linear computations of the LF/HF ratio agrees with that of a study comparing only FFT and AR computations of the LF/HF ratio in healthy, young adults across four conditions of physical activity: resting to 20% of peak oxygen consumption to recovery and resting to 40% of peak oxygen consumption to recovery [23].

In the present study, the mean LF/HF ratio at rest, computed by using the FFT, AR and the Lomb methods, were skewed toward greater sympathetic modulation than expected, as evidenced by three of the ten study subjects exhibiting a resting LF/HF ratio ≥ 3. This higher-than-expected level of sympathetic modulation at rest was accepted because it did not translate to higher-than-normal blood pressures. A resting LF/HF ratio < 1 is expected during both preprandial, overnight fasting and postprandial states resting in bed for a group of healthy young adults similar to those in the present study [28, 29]. In the present study, the subjects were all postprandial, and this condition was not controlled. The resting LF/HF ratio was also not controlled because inclusion of subjects exhibiting a wide range of resting LF/HF ratios was thought to make these results relevant to both research laboratories and medical clinics. Moreover, some normal study subjects were likely to have increased sympathetic modulations because of a “whitecoat syndrome” [30]. Sympathetic influences are reported to be exaggerated in an African-American subpopulation like the one comprising the present study group [31].

The present study evaluated the different linear computations of the LF/HF ratio in normalized units, which served to eliminate the computational variations associated with expressing HRV spectral powers in raw ms². Normalized units are used to express the LF and the HF on a proportional scale that compensates for the within- and the across-subject variability found for measures of raw ms² spectral powers, are reported to have smaller coefficients of variation than raw ms² units, and are more repeatable between studies [32]. The calculated LF/HF ratios and LF values were presented in data tables so that one could appreciate how activity-related changes in the LF, commonly used as the main indicator of sympathetic modulation, affected the computations of the LF/HF ratio. Because of the inverse proportionality between the LF and the HF expressed in normalized units, increases in the LF in normalized units are equivalent to decreases in the HF in normalized units, and vice versa [32]. Thus, the LF/HF ratio is thought to be a more useful index of autonomic (sympathetic- parasympathetic, sympathovagal) interactive modulation of the heart rate than the LF, a specific indicator of sympathetic modulation, and the HF, a specific indicator of vagal parasympathetic modulation, [32].

Linear FFT, AR and Lomb computations of the LF/HF ratio are reported to be reliable HRV spectral indicators of the power of the sympathovagal interactive modulations, with greater values being indicative of greater sympathetic influences [33]. On the other hand, the LF/HF ratio computed by using the nonlinear CZF method has been shown to be a spectral indicator of the variability in the RR interbeat intervals [33]. Decreases in the variability of RR intervals, also indicated in this study by using the time-domain SDNN measure, together with increases in the LF/HF ratio found by using FFT and AR computations, are widely- used indicators of autonomic responsiveness to aerobic exercise [23]. In the present study, the nonlinear CZF computations of the LF/HF ratio were found to decrease sequentially and significantly from rest to 30% to 50% of peak oxygen consumption. The time-domain SDNN computations of the LF/HF ratio exhibited the same activity- related pattern of change as the CZF computations, and those measurements were significantly correlated. We also observed that the differences between the CZF-computed values of the LF/HF ratio and the values of the LF/HF ratio in the time-domain SDNN computations between rest and 30% of peak oxygen consumption were significantly greater than those between 30% and 50% of peak oxygen consumption. This finding is consistent with previous research indicating that the largest cardiovascular perturbations occurred at the beginning of aerobic exercise [32]. On the other hand, the LF/HF ratios computed by using the linear FFT, AR and Lomb methods were found to be increased significantly only from rest to 50% of peak oxygen consumption. These findings suggest only the variability in the RR interbeat intervals may be measured reliably, not the power of the associated sympathovagal modulations from rest to 30% to 50% of peak oxygen consumption. These findings have important implications for HRV evaluations involving autonomic dysregulation and cardiovascular disease commonly performed using exercise stress testing protocols such as the one employed in the present study [24]. Exercise stress testing is a widely-used method for assessing patients and evaluating the effects of pharmacological and alternative medical treatments for cardiovascular disease on the cardiovascular system. Improvements in exercise stress testing, such as increased HRV, are important clinical outcomes of acupuncture treatments [34].

Greater sensitivity to activity from rest to 30% to 50% of peak oxygen consumption was demonstrated by using nonlinear CZF computations of the LF/HF ratio rather than linear FFT, AR, and Lomb periodogram computations. We conclude that sequential decrements in the variability of electrocardiogram RR interbeat intervals associated with changes in activity from rest to 30% to 50% of peak oxygen consumption can be reliably measured by using CZF computations of the LF/HF ratio. Future studies should determine whether the nonlinear CZF method improves assessment of the changes in the sympathovagal influences on heart rate and other autonomic functions that result from acupuncture, pharmacopuncture, and other medical treatments.

This work was supported by National Institutes of Health, National Center for Research Resources, Research Centers in Minority Institutions Program (NIH/NCRR/RCMI) Grant 2G12RR003048 to Howard University.

The authors declare that there are no conflict of interest.

1. Kim JM, Jeon HJ, Kim HJ, Cho CK, Yoo HS. Bee venom pharmacopuncture: an effective treatment for complex regional pain syndrome. J Pharmacopuncture. 2014;17(4):66-9.
2. Yook T, Yu J, Lee H, Song B, Kim L, Roh J, et al. Comparing the effects of distilled rehmannia glutinosa, wild ginseng and Astragali radix pharmacopuncture with heart rate variability (HRV): a randomized, sham-controlled and double-blind clinical trial. J Acupunct Meridian Stud. 2009;2(3):239-47.
3. Sun J, Li X, Yang C, Wang Y, Shi F, Gao Y, et al. Transcutaneous electrical acupuncture stimulation as a countermeasure against cardiovascular deconditioning during 4 days of head-down bed rest in humans. Acupunct Med. 2015;33(5):381-7.
4. Chien LW, Chen FC, Hu HY, Liu CF. Correlation of electrical conductance in meridian and autonomic nervous activity after auricular acupressure in middle-aged women. J Altern Complement Med. 2014;20(8):635-41.
5. Yang ZK, Wu ML, Xin JJ, He W, Su YS, Shi H, et al. Manual acupuncture and laser acupuncture for autonomic regulations in rats: observation on heart rate variability and gastric motility. Evid Based Complement Alternat Med. 2013;2013:ID276320.
6. Choi EJ, Yun YH, Yoo SY, Kim KS, Park JS, Choi IH. Autonomic conditions in tinnitus and implications for Korean medicine. Evid Based Complement Alternat Med. 2013;2013:ID402585.
7. Li QQ, Shi GX, Xu Q, Wang J, Liu CZ, Wang LP. Acupuncture effect and central autonomic regulation. Evid Based Complement Alternat Med. 2013;2013:ID267959.
8. Ngai SP, Jones AY. Changes in skin impedance and heart rate variability with application of acu-TENS to BL 13 (Feishu). J Altern Complement Med. 2013;19(6):558- 63.
9. Backer M, Schaefer F, Siegler N, Balzer S, Michalsen A, Langhorst J, et al. Impact of stimulation dose and personality on autonomic and psychological effects induced by acupuncture. Auton Neurosci. 2012;170(1- 2):48-55.
10. Sun J, Sang H, Yang C, Dong H, Lei C, Lu Y, et al. Electroacupuncture improves orthostatic tolerance in healthy individuals via improving cardiac function and activating the sympathetic system. Europace. 2013;15(1):127-34.
11. Chu H, Li MH, Juan SH, Chiou WY. Effects of transcutaneous electrical nerve stimulation on motion sickness induced by rotary chair: a crossover study. J Altern Complement Med. 2012;18(5):494-500.
12. No authors. Heart rate variability: standards of measurement, physiological interpretation, and clinical use. task force of the European society of cardiology and the North american society of pacing and electrophysiology. Eur Heart J. 1996;17(3):354-81
13. Tripathi KK. Very low frequency oscillations in the power spectra of heart rate variability during dry supine immersion and exposure to non-hypoxic hypobaria. Physiol Meas. 2011;32(6):717-29.
14. Kleiger RE, Stein PK, Bigger JT Jr. Heart rate variability: measurement and clinical utility. Ann Noninvasive Electrocardiol. 2005;10(1):88-101.
15. Alian AA, Galante NJ, Stachenfeld NS, Silverman DG, Shelley KH. Impact of central hypovolemia on photoplethysmographic waveform parameters in healthy volunteers part 2: frequency domain analysis. J Clin Monit Comput. 2011;25(6):387-96.
16. Perini R, Veicsteinas A. Heart rate variability and autonomic activity at rest and during exercise in various physiological conditions. Eur J Appl Physiol. 2003;90(3- 4):317-25.
17. Parati G, Saul JP, Di Rienzo M, Mancia G. Spectral analysis of blood pressure and heart rate variability in evaluating cardiovascular regulation. a critical appraisal. Hypertension. 1995;25(6):1276-86.
18. Mainardi LT, Bianchi AM, Cerutti S. Time-frequency and time-varying analysis for assessing the dynamic responses of cardiovascular control. Crit Rev Biomed Eng. 2002;30(1-3):175-217.
19. Otsuka K, Murakami S, Kubo Y, Yamanka T, Mitsutake G, Ohkawa S, et al. Chronomics for chronoastrobiology with immediate spin-offs for life quality and longevity. Biomed Pharmacother. 2003;57(S1):1-18.
20. Kerut EK, McKinnie JJ, Giles TD. Modern evaluation of the hypertensive patient: autonomic tone in cardiovascular disease and the assessment of heart rate variability. Blood Press Monit. 1999:4(S1):7-14.
21. Clifford GD, Tarassenko L. Quantifying errors in spectral estimates of HRV due to beat replacement and resampling. IEEE Trans Biomed Eng. 2005;52(4):630-8.
22. Clifford GD, Tarassenko L. Segmenting cardiac-related data using sleep stages increases separation between normal subjects and apnoeic patients. Physiol Meas. 2004;25(6):27-35.
23. Mendonca GV, Fernhall B, Heffernan KS, Pereira FD. Spectral methods of heart rate variability analysis during dynamic exercise. Clin Auton Res. 2009;19(4):237- 45.
24. Simula S, Vanninen E, Hedman A, Lehto S, Kuikka J, Hartikainen J. Myocardial (123) I-metaiodobenzylguanidine washout and heart rate variability in asymptomatic subjects. Ann Noninvasive Electrocardiol. 2012;17(1):8-13.
25. Sloan RP, DeMeersman RE, Shapiro PA, Bagiella E, Chernikhova D, Kuhl AS, et al. Blood pressure variability responses to tilt are buffered by cardiac control. Am J Physiol. 1997;273(3):1427-31.
26. Bernardi L, Ricordi L, Lazzari P, Solda P, Calciati A, Ferrari MR, et al. Impaired circadian modulation of sympathovagal activity in diabetes. a possible explanation for altered temporal onset of cardiovascular disease. Circulation. 1992;86(5):1443-52.
27. Millis RM, Austin RE, Bond V, Farugue M, Goring KL, Hickey BM, et al. Effects of high-carbohydrate and high- fat dietary treatments on measures of heart rate variability and sympathovagal balance. Life Sci. 2009;85(3-4):141-5.
28. Millis RM. Austin RE, Hatcher MD, Bond V, Farugue MU, Goring KL, et al. Association of body fat percentage and heart rate variability measures of sympathovagal balance. Life Sci. 2010;86(5-6):153-7.
29. Fagard RH, Stolarz K, Kuznetsova T, Seidlerova J, Tikhonoff V, Grodzicki T, et al. Sympathetic activity, assessed by power spectral analysis of heart rate variability, in white-coat, masked and sustained hypertension versus true normotension. J Hypertens. 2007;25(11):2280-5.
30. Labinson PT, Giacco S, Gift H, Mansoor GA, White WB. The importance of the clinical observer in the development of a white-coat effect in African-American patients with hypertension. Blood Press Monit. 2008;13(3):139-42.
31. Burr RL. Interpretation of normalized spectral heart rate variability indices in sleep research: a critical review. Sleep. 2007;30(7):913-9.
32. Eriksen M, Waaler BA, Walloe L, Wesche J. Dynamics and dimensions of cardiac output changes in humans at the onset and at the end of moderate rhythmic exercise. J Physiol. 1990;426:423-37.
33. Conte EW, Pieralice M, Laterza V, Losurdo A, Santacroce N, Conte S, et al. Traditional and a new methodology for analysis of heart rate variability: a review by physiological and clinical experimental results. International Journal of Research and Reviews in Applied Sciences. 2012;13(1):206-93.
34. Li P, Ayannusi O, Reid C, Longhurst JC. Inhibitory effect of electroacupuncture (EA) on the pressor response induced by exercise stress. Clin Auton Res. 2004;14(3):182-8.