Aerobic Threshold Identification in a Cardiac Disease Population Based on Correlation Properties of Heart Rate Variability

index of heart rate (HR) variability correlation properties, the short-term scaling exponent alpha1 of detrended uctuation analysis (DFA a1) has shown potential to delineate the rst ventilatory threshold (VT1). This study aims to extend this concept to a group of participants with cardiac disease. Sixteen volunteers with stable coronary disease or heart failure performed an incremental cycling ramp to exhaustion PRE and POST a 3-week training intervention. Oxygen uptake (VO 2) and HR at VT1 were obtained from a metabolic cart. An ECG was processed for DFA a1 and HR. The HR variability threshold (HRVT) was de ned as the VO 2, HR or power where DFA a1 reached a value of 0.75. Mean VT1 was reached at 16.82 5.72 mL/kg/min, HR of 91.3 11.9 bpm and power of 67.8 17.9 watts compared to HRVT at 18.02

from a metabolic cart. An ECG was processed for DFA a1 and HR. The HR variability threshold (HRVT) was de ned as the VO 2, HR or power where DFA a1 reached a value of 0.75. Mean VT1 was reached at 16.82 5.72 mL/kg/min, HR of 91.3 11.9 bpm and power of 67.8 17.9 watts compared to HRVT at 18.02 7.74 mL/kg/min, HR of94.7 14.2 bpm and power of 73.2 25.0 watts. Linear relationships were seen between modalities, with Pearson's r of 0.95 (VO 2), 0.86 (HR) and 0.87 (power). Bland–Altman assessment showed mean differences of 1.20 mL/kg/min, 3.4 bpm and 5.4 watts. Mean peak VO 2and VT1 did not change after training intervention. However, the correlation between PRE to POST change in VO 2at VT1 with the change in VO 2at HRVT was signi cant (r = 0.84,p< 0.001). Reaching a DFA a1 of 0.75 was associated with the VT1 in a population with cardiac disease. VT1 change after training intervention followed that of the HRVT, con rming the relationship between these parameters. Keywords: ventilatory threshold; HRV; detrended uctuation analysis; DFA a1; endurance exercise 1. Introduction Training zone classi cation is a cornerstone for exercise intensity distribution study and implementation [1]. Although there may be different schools of thought on what type of distribution is “optimal” (polarized vs. pyramidal vs. threshold) all models are de ned by having the major portion of training done at a low exercise intensity in the various elds of application [1]. Examination of training intensity distribution in other subject populations such as those with ischemic heart disease or congestive heart failure also indicates the importance of proper workload modulation and low-intensity (aerobic) exercise [2–4]. To identify these training zones, a set of boundaries have been developed. In the classic three-zone model, separation is determined by lactate or gas exchange markers, with the lowest-intensity zone (zone 1) below the aerobic threshold as denoted by the rst lactate (LT1) or ventilatory threshold (VT1) [5,6]. Unfortunately, both gas exchange [7] and lactate threshold [8,9] methods of identi cation are subject to various types of inaccuracies as well

been developed. In the classic three-zone model, separation is determined by lactate or gas exchange markers, with the lowest-intensity zone (zone 1) below the aerobic threshold as denoted by the rst lactate (LT1) or ventilatory threshold (VT1) [5,6]. Unfortunately, both gas exchange [7] and lactate threshold [8,9] methods of identi cation are subject to various types of inaccuracies as well as the need for special equipment and operators [10]. In lieu J. Clin. Med.2021,10, 4075.

J. Clin. Med.2021,10, 4075 2 of 13 of these issues, several heart rate variability (HRV) indexes have been proposed as a means of de ning the low-intensity threshold transition [11,12]. However, the use of time and frequency domain HRV measures has been shown to have issues with precision and reliability [13]. Recently, encouraging data have indicated that a non-linear index of fractal correlation properties of heart rate (HR) time series based on the short-term scaling exponent alpha1 of detrended uctuation analysis (DFA a1) can indicate zone 1 transition in a population of healthy recreational athletes [14]. Whether that observation extends to other, more diverse, populations is unknown. Given the importance of proper intensity-zone assessment in cardiac rehabilitation outcome and guidance [2,3,15,16] an assessment of agreement between a DFA-a1-based HRV threshold (HRVT) with the VT1 obtained by gas exchange would be of great interest. If DFA a1 behavior is similar both in athletes and in those with cardiac disease, this could lead to more widespread interest in using this HRV index as a modality for training guidance in the low to moderate intensity areas for the purpose of intensity distribution in therapy and rehabilitation. DFA a1 behavior during exercise has been described previously from both a theoretical and a practical standpoint [17,18]. In brief, this index represents the degree of self-similarity and fractal-like composition of a series of cardiac interbeat intervals, providing information about organismic demands and network physiology during exercise [19]. The underlying mechanism behind the change in this index with increasing exercise intensity has been attributed to alterations in the parasympathetic–sympathetic balance on the sinoatrial node as well as interactions of different organismic subsystems and (non-neural) in uencing factors [12,20–22]. DFA a1 is a dimensionless index with a wide dynamic range centered around the VT1 [18]. At low exercise intensity, DFA a1 values remain in a well-correlated, fractal pattern that progresses to a more uncorrelated behavior past the aerobic threshold. In a previous investigation involving a diverse population of male runners, it was shown that a DFA a1 value of 0.75 reached on an incremental exercise ramp closely corresponded

dynamic range centered around the VT1 [18]. At low exercise intensity, DFA a1 values remain in a well-correlated, fractal pattern that progresses to a more uncorrelated behavior past the aerobic threshold. In a previous investigation involving a diverse population of male runners, it was shown that a DFA a1 value of 0.75 reached on an incremental exercise ramp closely corresponded to the VT1 intensity measured by gas exchange [14]. A signi cant advantage of DFA a1 behavior as opposed to other HRV indexes used to de ne the aerobic threshold revolves around the underlying method of determination. Several time and frequency domain- related HRV parameters (standard deviation of NN intervals, SDNN; standard deviation 1 from Poincar²plot analysis, SD1; high frequency power of frequency domain HRV analysis, HF power) do decline with increasing exercise effort to reach a nadir near the VT1 [12]. By utilizing incremental exercise ramps, observation of when that nadir occurs can be helpful in threshold de nition. However, many subjects do not display easily noted nadirs [13] and this technique requires a near-maximal-effort exercise ramp to be performed. A major bene t of DFA a1 methodology for intensity assessment is that the VT1-related threshold does not rely on nadir identi cation. As exercise intensity climbs, there is a near linear decline from values of 1.0 to 0.5 or below, centered around the VT1, enabling calculation of heart rate, oxygen uptake (VO2) or power at a prede ned DFA a1 of 0.75. Further intensity increases cause continued DFA a1 decline to values well below 0.5, indicating work rates above the VT1. Therefore, the additional dynamic range provides numeric con rmation of exceeding a zone 1 target. Since the index is dimensionless, no prior normalization or ramp testing should be needed in a mixed-intensity exercise session to assess time spent in a low-intensity zone [18]. Well-established metrics of intensity such as cycling power need calibration with either gas exchange or blood lactate concentration to indicate relative load and intensity zone [10]. Although many authorities recommend intensity zones based on percentage of maximum heart rate [23], others have appropriately pointed

be needed in a mixed-intensity exercise session to assess time spent in a low-intensity zone [18]. Well-established metrics of intensity such as cycling power need calibration with either gas exchange or blood lactate concentration to indicate relative load and intensity zone [10]. Although many authorities recommend intensity zones based on percentage of maximum heart rate [23], others have appropriately pointed out that these values do not necessarily match those derived from gas exchange or lactate methods [10]. Conversely, it has been shown in healthy subjects that the DFA a1 index appears to be already standardized to exercise load status, with values lower than 0.75 representing intensity above the VT1.

J. Clin. Med.2021,10, 4075 3 of 13 Hence, the purpose of this study is to explore two major questions. First, does the DFA-a1-derived HRVT agree with the VT1 determined by gas exchange in subjects with congestive heart failure or ischemic cardiac disease? Second, does the change in HRVT cor- respond to the change in the VT1 after a standard cardiac rehabilitation exercise program? 2. Materials and Methods 2.1. Participants Sixteen volunteers with either stable congestive heart failure (CHF) or previous exer- tional angina (CAD) were evaluated before and after an exercise rehabilitation program. All subjects underwent baseline cardiac catheterization, had an unchanged serial resting ECG and were deemed clinically stable for 3 weeks prior to each incremental exercise ramp testing session. Participants with CAD had ejection fractions above 50% and had normal cardiac contractile function. The etiology of the CHF was of either ischemic and/or dilated cardiomyopathy with ejection fractions below 40%. All participants were on beta blockade therapy except for one (Table). No participant had symptoms of unstable angina or evidence of signi cant ventricular arrythmia, atrial brillation, supraventricular arrhyth- mias or exercise-test-related dif culty. The study protocol complied with the Declaration of Helsinki and was reviewed and accepted by the ethical committee of Tours (France; no. 2005–2025). All participants were informed about the study procedure and risks and gave their written informed consent. Table 1.Demographic data of all participants (n= 16). Age (Years) Ht (cm) BW (kg) HR PEAK(bpm) VO 2PEAK(mL/kg/min) Etiology 52 173 71 117 17.36 CHF 59 168 73 109 23.05 CAD 55 176 89 141 31.42 CAD 62 172 85 129 28.96 CAD 56 167 72 131 42.50 CAD 55 176 94 112 24.84 CAD 54 178 94 95 25.89 CAD 59 163 58 110 24.93 CAD 41 176 58 128 51.23 CAD 53 171 96 131 27.21 CHF 44 173 67 133 33.05 CAD 70 174 81 108 21.52 CHF 64 170 74 114 35.67 CHF 40 182 89 162 27.88 CHF * 62 175 97 132 28.53 CAD 58 175 67 131 22.28 CHF 55 ( 8) 173 ( 5) 79.0 ( 13)

24.93 CAD 41 176 58 128 51.23 CAD 53 171 96 131 27.21 CHF 44 173 67 133 33.05 CAD 70 174 81 108 21.52 CHF 64 170 74 114 35.67 CHF 40 182 89 162 27.88 CHF * 62 175 97 132 28.53 CAD 58 175 67 131 22.28 CHF 55 ( 8) 173 ( 5) 79.0 ( 13) 124 ( 16) 29.15 ( 8.42) - Age: current age, Ht: height, BW: body weight, HRPEAK: peak heart rate on baseline ramp test, VO2PEAK: peak oxygen uptake reached on baseline ramp test, Etiology: cause of cardiac disease, either congestive heart failure (CHF) or ischemic coronary artery disease (CAD) (* denotes no beta blocker therapy usage); mean ( standard deviation, SD) in last row. 2.2. Exercise Testing Protocol Participants performed an incremental cycling ramp in an upright position on a cycle ergometer with careful monitoring for symptoms (ERG 900, GE Medical System, CASE Exercise Testing System Case, Milwaukee, WI, USA). Ramp protocol consisted of an initial cycling power of 20 watts followed by a 10 watt increase every minute until exhaustion. Peak effort was con rmed by failure of oxygen uptake (VO2) and/or HR to increase with further increases in work rate. Testing procedures were the same pre- and post-exercise intervention.

J. Clin. Med.2021,10, 4075 4 of 13 2.3. Exercise Training Intervention The exercise training intervention consisted of cycle ergometer sessions (30 min per day, 5 times per week), at an intensity approximating the participants' VT1 heart rate, which was measured during the rst incremental ramp test. In addition, gymnastic (callisthenic) sessions of 50 min per day were performed either in a conventional fashion or immersed in a swimming pool at an intensity at or below the VT1 [24]. The participants also attended counseling sessions centered around secondary cardiovascular prevention, stress management, nutritional guidance and smoking cessation. No exercise sessions were devoted to training above VT1-related intensity. 2.4. Gas Exchange Testing Gas exchange parameters were measured breath-by-breath using a Vmax Spectra system (SensorMedics Corporation, Yorba Linda, CA, USA). VO2, carbon dioxide output (VCO2) and HR were imported into Microsoft Excel 365 for detailed analysis. Cycling power was measured continuously and reported for each gas exchange measurement. Graphing of the above parameters were done to derive VT1, VO2PEAKand VO2vs. time. VT1 was determined by the excess CO2method using the breath-by-breath data without averaging [25] with two experienced observers con rming VT1 results. If no consensus was found, an additional opinion was sought. VO2at the time of VT1 was based on linear regression of the VO2over time relationship excluding a VO2plateau, if present. Both VO2PEAKand peak cycling power (PPEAK) were obtained from an average of each participant's VO2or cycling power over the last 60 s of the ramp test. Power at VT1 was calculated from the 60 s average power centered at the time of VT1 [26]. Peak heart rate (HRPEAK) was de ned as the average of the three highest heart rates measured during the nal 60 s of the test. 2.5. RR Measurements and Calculation of DFA-a1-Derived Threshold An ECG system (VISTA Holter NOVACOR, Rueil, Malmaison, France) with a sampling rate of 200 Hz was used to record the participant's ECG/RR times series. The RR data were extracted as a text le and imported into Kubios HRV Software Version 3.4.2 (Biosignal Analysis and Medical Imaging Group, Department of Physics, University of

RR Measurements and Calculation of DFA-a1-Derived Threshold An ECG system (VISTA Holter NOVACOR, Rueil, Malmaison, France) with a sampling rate of 200 Hz was used to record the participant's ECG/RR times series. The RR data were extracted as a text le and imported into Kubios HRV Software Version 3.4.2 (Biosignal Analysis and Medical Imaging Group, Department of Physics, University of Kuopio, Kuopio, Finland). Kubios' preprocessing settings were at the default values including the RR detrending method, which was kept at “Smoothn priors” (Lambda = 500). The DFA a1 window width was set to 4 n 16 beats. The RR series was then corrected by the Kubios “automatic method”, and relevant HRV parameters were exported as text les for further analysis. Percent of artifacts during pertinent measurement windows was usually less than 2% and never above 5%. DFA a1 was calculated from the RR data series using 2 min time windows with repeat computation every 5 s throughout the test. The moving time window measurement was used to better delineate rapid changes in the DFA a1 index over the course of the test. For the detection of HRVT, a DFA a1 value of 0.75 was chosen based on a previous study in recreational athletes [14]. This value is also the midpoint between a fractal, well-correlated behavior of the HR time series of 1.0 (seen with very light exercise) and an uncorrelated value of 0.5, which represents white noise, random behavior (seen with high-intensity exercise). Plotting of DFA a1 vs. time was then performed, generally showing a reverse sigmoidal curve with a stable area above 1.0 at low work rates, a rapid, near linear drop reaching below 0.5 at higher intensity, then attening without major change. The procedure used to indicate at what level of cycling intensity (as per VO2or HR) the DFA a1 would cross a value of 0.75 has been detailed previously [14]. Note should be made that the HR at DFA a1 0.75 (based on ECG data) was compared to the HR at VT1 obtained from the metabolic cart data. Power at DFA a1 = 0.75 (power at HRVT)

what level of cycling intensity (as per VO2or HR) the DFA a1 would cross a value of 0.75 has been detailed previously [14]. Note should be made that the HR at DFA a1 0.75 (based on ECG data) was compared to the HR at VT1 obtained from the metabolic cart data. Power at DFA a1 = 0.75 (power at HRVT) was calculated from the 60 s average power centered at the time DFA a1 reached 0.75.

J. Clin. Med.2021,10, 4075 5 of 13 2.6. Statistics Statistical analysis was performed for the key variables, VO2PEAK, PPEAK, HRPEAK, VO2, HR and power at VT1, VO2, HR and power at HRVT. Conventional methods were used for the calculation of means and standard deviations (SDs). Normal distribution of data was checked by Shapiro–Wilk's testing. Agreement against the established standard VT1-related parameters was assessed using Pearson's r correlation coef cient, standard error of estimate (SEE), coef cient of determination (R 2 ) and Bland–Altman plots with limits of agreement [27]. The magnitude of Pearson's r correlations was evaluated as follows: 0.3 r < 0.5, low; 0.6 r < 0.8, moderate and r 0.8, high [28]. Paired t-testing was used for comparison of PRE vs. POST intervention. For all tests, the statistical signi cance was accepted asp 0.05. Cohen's d was used to denote effect sizes (small effect = 0.2, medium effect = 0.5, large effect = 0.8) [29]. Analysis was performed using Microsoft Excel 365 with Real Statistics Resource Pack software (Release 6.8) and Analyse-it Software (Version 5.66). 3. Results 3.1. Comparison of VT1 and HRVT Regression analysis for comparisons of VT1 vs. HRVT (such as their respective VO2, HR and power values) calculated from all ramp tests performed (both PRE and POST) are shown in Figure. Strong correlations were seen between VT1-based determinations and those derived from HRVT. Pearson's r was calculated at 0.95, 0.86 and 0.87 (allp< 0.001) for gas-based VO2(mL/kg/min), HR and power comparisons, respectively. Bland–Altman evaluation for each metric comparison is shown in Figure. Mean differences between gas-exchange-based VT1 and HRVT were 1.20 mL/kg/min for VO2, 3.4 bpm for HR and 5.4 watts for power. Limits of agreement (1.96 SD) are also listed in Figure. 3.2. PRE vs. POST Training Comparisons After exercise intervention (PRE vs. POST) HRPEAKand PPEAKincreased markedly, which was statistically signi cant with high effect sizes (Table). VO 2PEAKshowed no signi cant change. Analysis of individual changes in VT1- and HRVT-related parameters did show variation (Figure). Plotting the change in VT1 VO 2vs. the change in HRVT VO2showed both the presence of heterogeneity of

The study explores the relationship between heart rate variability and ventilatory thresholds in cardiac patients.