Processing metadata

rt rate variability (HRV) is frequently applied in sport-speci c settings. The rising use of freely accessible applications for its recording requires validation processes to ensure accurate data. It is the aim of this study to compare the HRV data obtained by the Polar H10 sensor chest strap device and an electrocardiogram (ECG) with the focus on RR intervals and short-term scaling exponent alpha 1 of Detrended Fluctuation Analysis (DFA a1) as non-linear metric of HRV analysis. A group of 25 participants performed an exhaustive cycling ramp with measurements of HRV with both recording systems. Average time between heartbeats (RR), heart rate (HR) and DFA a1 were recorded before (PRE), during, and after (POST) the exercise test. High correlations were found for the resting conditions (PRE: r = 0.95, rc= 0.95, ICC 3,1= 0.95, POST: r = 0.86, rc= 0.84, ICC 3,1= 0.85) and for the incremental exercise (r > 0.93, rc> 0.93, ICC 3,1> 0.93). While PRE and

(RR), heart rate (HR) and DFA a1 were recorded before (PRE), during, and after (POST) the exercise test. High correlations were found for the resting conditions (PRE: r = 0.95, rc= 0.95, ICC 3,1= 0.95, POST: r = 0.86, rc= 0.84, ICC 3,1= 0.85) and for the incremental exercise (r > 0.93, rc> 0.93, ICC 3,1> 0.93). While PRE and POST comparisons revealed no differences, signi cant bias could be found during the exercise test for all variables (p< 0.001). For RR and HR, bias and limits of agreement (LoA) in the Bland–Altman analysis were minimal (RR: bias of 0.7 to 0.4 ms with LoA of 4.3 to 2.8 ms during low intensity and 1.3 to 0.5 ms during high intensity, HR: bias of 0.1 to 0.2 ms with LoA of 0.3 to 0.5 ms during low intensity and 0.4 to 0.7 msduring high intensity). DFA a1 showed wider bias and LoAs (bias of 0.9 to 8.6% with LoA of 11.6 to 9.9% during low intensity and 58.1 to 40.9% during high intensity). Linear HRV measurements derived from the Polar H10 chest strap device show strong agreement and small bias compared with ECG recordings and can be recommended for practitioners. However, with respect to DFA a1, values in the uncorrelated range and during higher exercise intensities tend to elicit higher bias and wider LoA. Keywords:HRV; RR intervals; DFA a1; chest strap; wearable; endurance exercise 1. Introduction Heart rate variability (HRV) is believed to re ect autonomic nervous system activity by non-invasively measuring the time and pattern between consecutive R-waves in the electrocardiogram (ECG) [1,2]. It encompasses a wide range of application elds including medical and sport-speci c settings [3,4]. Here, studies mainly apply HRV in the context of physical exercise and training monitoring purposes and for the optimization of the training process, mostly by taking resting measurements with focus on standard linear parameters like the root mean square of successive differences (RMSSD) [4–6]. The gold standard measure for the quanti cation of RR intervals and assessment of HRV is the electrocardiogram (ECG) [7]. However, it has become common practice to

and training monitoring purposes and for the optimization of the training process, mostly by taking resting measurements with focus on standard linear parameters like the root mean square of successive differences (RMSSD) [4–6]. The gold standard measure for the quanti cation of RR intervals and assessment of HRV is the electrocardiogram (ECG) [7]. However, it has become common practice to use mobile systems (e.g., apps, wearables) and chest straps for data recording, storage, analysis and/or export which offers superior practicability with respect to cost, ease to use, portability and Sensors2022,22, 6536.

Sensors2022,22, 6536 2 of 13 interpretation [8–10]. However, an evaluation of these applications and sensor devices in different populations is required in order to ensure the validity of the RR interval measurements for the subsequent HRV interpretation [9,11,12]. A frequently used chest strap device is the Polar H10 (Polar Electro Oy, Kempele, Finland) which has already proven validity to assess RR intervals correctly during rest and physical exercise conditions [13]. In this report, there was a substantially better signal quality during high-intensity activities for the Polar H10 in comparison to a 3-lead ECG Holter monitor which led the authors to the conclusion that simple chest strap devices might be recommended as the gold standard for RR interval assessments when strong body movements are present [13]. However, the investigated sample was very homogeneous in its characteristics (healthy, lean, and physically t volunteers with an average age of 25 years) and the sample size was small (N = 10), so that a generalization of this conclusion should be regarded with caution. A more recent study evaluated the Polar H10 against a photoplethysmography technology (Welltory, New York, NY, USA) and a 12-channel ECG during resting condition in supine and in seated positions [10]. There were no differences observed in the results of the Polar H10 in comparison to the ECG when Kubios HRV software was used to obtain HRV data. Nevertheless, the comparison focused solely on RMSSD and the sample consisted of professional road cyclists, con rming the validity only for these speci cities. In addition to the requirement to perform population-based validations, to date no validation has been conducted for non-linear measures of HRV. It is possible that agreement with ECG measurement is also dependent on the particular index investigated, which is why there is also the need for metric-speci c validations [9]. The non-linear short- term scaling exponent alpha 1 of Detrended Fluctuation Analysis (DFA a1) has shown its suitability to describe the complex cardiac autonomic regulation during various exercise intensities, modalities, and environmental conditions while possessing a wide dynamic range throughout all intensity zones in contrast to standard linear parameters

why there is also the need for metric-speci c validations [9]. The non-linear short- term scaling exponent alpha 1 of Detrended Fluctuation Analysis (DFA a1) has shown its suitability to describe the complex cardiac autonomic regulation during various exercise intensities, modalities, and environmental conditions while possessing a wide dynamic range throughout all intensity zones in contrast to standard linear parameters [14,15]. In this context, recent studies show the potential of this parameter to demarcate physiological threshold boundaries by means of xed DFA a1 values at the aerobic (~0.75) and anaerobic threshold (~0.5) for usage in performance monitoring and exercise intensity distribution in endurance sports [16,17]. Additionally, exercise associated DFA a1 dynamics may also give insights about physiological status in terms of monitoring of fatigue during low-intensity exercise [18]. Therefore, the importance of a validation study with focus on the recording method and DFA a1 becomes evident, since researchers and recreational athletes frequently use this device for data collection and therefore rely on appropriate measurements. The present study compared selected HRV data (with inclusion of DFA a1) obtained by the Polar H10 chest strap device using the Elite HRV application for data recording, storage and export vs. a 12-channel ECG and analyzed by Kubios HRV Software version 3.5.0 (Biosignal Analysis and Medical Imaging Group, Department of Physics, University of Kuopio, Kuopio, Finland [19]) during resting and exercise conditions in a group of female and male recreational athletes. 2. Materials and Methods 2.1. Participants Twenty- ve participants (men: n = 14, age: 40 14 years, height: 178.1 9.0 cm, body weight: 82.2 14.8 kg; women: n = 11, age: 34 10 years, height: 169.1 4.3 cm, body weight: 67.8 9.5 kg) volunteered to take part in this study after being recruited via word of mouth as via the internet. Entry criteria included adults (>18 years) of either sex and of any tness level without previous medical history, current medications or recent illness. Participants were asked to abstain from caffeine, alcohol, tobacco, and vigorous exercise24 hbefore testing and provided written informed consent. Ethical approval for the study was obtained by the University of

word of mouth as via the internet. Entry criteria included adults (>18 years) of either sex and of any tness level without previous medical history, current medications or recent illness. Participants were asked to abstain from caffeine, alcohol, tobacco, and vigorous exercise24 hbefore testing and provided written informed consent. Ethical approval for the study was obtained by the University of Hamburg, Department of Psychology and Movement Science, Germany (reference no.: 2021_400) and is in line with the principles of the Declaration of Helsinki. The complete methodical procedure can be seen in Figure.

Sensors2022,22, 6536 3 of 13Sensors 2022, 22, x FOR PEER REVIEW 3 of 15 Participants were asked to abstain from caffeine, alcohol, tobacco, and vigorous exercise 24 h before testing and provided written informed consent. Ethical approval for the study was obtained by the University of Hamburg, Department of Psychology and Movement Science, Germany (reference no.: 2021_400) and is in line with the principles of the Decla- ration of Helsinki. The complete methodical procedure can be seen in Figure 1. Figure 1. Flow chart of the methodical procedure. 2.2. Exercise Protocol and Data Acquisition An incremental ramp was performed on a mechanically braked cycle (Ergoselect 4 SN, Ergoline GmbG, Bitz, Germany) with cadence kept between 60–80 rpm. The protocol consisted of a three-minute initial workload of 50 Watts followed by a 1 watt increase every 3.6 s (equivalent to 50 Watts/3 min) until the volunteer’s voluntary exhaustion. Re- cordings of heart rate (HR) and RR intervals were taken continuously during the exercise test, as well as prior (PRE; after a period of habituation and session preparation) and post (POST) exercise by means of 3-min supine rest condition measurement intervals with two devices at the same time: (1) 12-channel ECG CardioPart 12 Blue (AMEDTEC Mediz- intechnik Aue GmbH, Germany; sampling rate: 500 Hz; desktop software: AMEDTEC ECGpro version 5.10.002), and (2) Polar H10 sensor chest strap device (Polar Electro Oy, Kempele, Finland; sampling rate: 1000 Hz; app software: Elite HRV App, Version 5.5.1). Placement of the ECG electrodes and chest strap device is pictured in Figure 2. Figure 1.Flow chart of the methodical procedure. 2.2. Exercise Protocol and Data Acquisition An incremental ramp was performed on a mechanically braked cycle (Ergoselect 4 SN, Ergoline GmbG, Bitz, Germany) with cadence kept between 60–80 rpm. The protocol consisted of a three-minute initial workload of 50 Watts followed by a 1 watt increase every 3.6 s (equivalent to 50 Watts/3 min) until the volunteer's voluntary exhaustion. Recordings of heart rate (HR) and RR intervals were taken continuously during the exercise test, as well as prior (PRE; after a period of habituation and session preparation)

60–80 rpm. The protocol consisted of a three-minute initial workload of 50 Watts followed by a 1 watt increase every 3.6 s (equivalent to 50 Watts/3 min) until the volunteer's voluntary exhaustion. Recordings of heart rate (HR) and RR intervals were taken continuously during the exercise test, as well as prior (PRE; after a period of habituation and session preparation) and post (POST) exercise by means of 3-min supine rest condition measurement intervals with two devices at the same time: (1) 12-channel ECG CardioPart 12 Blue (AMEDTEC Medizintechnik Aue GmbH, Germany; sampling rate: 500 Hz; desktop software: AMEDTEC ECGpro version 5.10.002), and (2) Polar H10 sensor chest strap device (Polar Electro Oy, Kempele, Finland; sampling rate: 1000 Hz; app software: Elite HRV App, Version 5.5.1). Placement of the ECG electrodes and chest strap device is pictured in Figure.Sensors 2022, 22, x FOR PEER REVIEW 4 of 15 Figure 2. Placement of 12-channel ECG electrodes and Polar H10 chest strap device. V1 = 4th inter- costal space at the right border of the sternum, V2 = 4th intercostal space at the left border of the sternum, V3 = midway between locations V2 and V4, V4 = at the mid-clavicular line in the 5th inter- costal space, V5 = at the anterior axillary line in the same horizontal level as V4, V6 = at the mid- axillary line on the same horizontal level as V4 and V5. Limb leads of the right arm (RA) and left arm (LA) outwardly on the shoulders and right leg (RL) and left leg (LL) leads at the lower edge of the ribcage. Breath-by-breath pulmonary gas exchanges were recorded throughout the ramp us- ing a metabolic analyzer (Quark CPET, module A-67-100-02, COSMED Deutschland GmbH, Fridolfing, Germany; desktop software: Omnia version 1.6.5). Prior to testing, the gas analyzers were calibrated according to the manufacturer’s instructions. The protocol was terminated when the participant fell below a cadence of 60 rpm or due to self-deter- mination. The following criteria served for the assumption of exhaustion: (A) heart rate >90% of the maximum predicted heart rate (prediction model according to [20]:

software: Omnia version 1.6.5). Prior to testing, the gas analyzers were calibrated according to the manufacturer’s instructions. The protocol was terminated when the participant fell below a cadence of 60 rpm or due to self-deter- mination. The following criteria served for the assumption of exhaustion: (A) heart rate >90% of the maximum predicted heart rate (prediction model according to [20]: 208 − (0.7 × age) and (B) respiratory quotient > 1.1. Maximum oxygen uptake (VO 2max) and maximum HR (HR max) were defined as the average VO2 and HR over the last 30 s of the test. 2.3. Data Processing After extraction of the 12-channel ECG data (converted from exported.xml files) and RR data (exported from the Elite HRV app) as text files, import into Kubios HRV Premium Software version 3.5.0 (Biosignal Analysis and Medical Imaging Group, Department of Physics, University of Kuopio, Kuopio, Finland, [19]) was conducted. For the 12-channel ECG, lead 2 was used as the comparable lead to the chest strap device [21]. Preprocessing settings were set to the default values including the RR detrending method which was kept at “smoothness priors” (Lambda = 500). The RR series was then corrected by the Ku- bios HRV Premium “automatic method” [22]. For DFA a1 calculation window width was set to 4 ≤ n ≤ 16 beats [23]. During rest conditions a 2-min time window (00:30–02:30 min:s) was chosen for the analysis. During the incremental exercise time-varying analysis was adjusted to a 2-min window width and 20-s grid interval for the moving window, so that the exported.csv files from Kubios HRV Premium Software contained the HRV metrics of interest (RR, HR, DFA a1) recalculated every 20 s. Data sets with artefacts >5% were ex- cluded from analysis based on [24]. 2.4. Statistics Figure 2. Placement of 12-channel ECG electrodes and Polar H10 chest strap device. V1 = 4th intercostal space at the right border of the sternum, V2 = 4th intercostal space at the left border of the sternum, V3 = midway between locations V2 and V4, V4 = at the mid-clavicular line in the 5th intercostal space, V5

[24]. 2.4. Statistics Figure 2. Placement of 12-channel ECG electrodes and Polar H10 chest strap device. V1 = 4th intercostal space at the right border of the sternum, V2 = 4th intercostal space at the left border of the sternum, V3 = midway between locations V2 and V4, V4 = at the mid-clavicular line in the 5th intercostal space, V5 = at the anterior axillary line in the same horizontal level as V4, V6 = at the mid-axillary line on the same horizontal level as V4 and V5. Limb leads of the right arm (RA) and left arm (LA) outwardly on the shoulders and right leg (RL) and left leg (LL) leads at the lower edge of the ribcage.

Sensors2022,22, 6536 4 of 13 Breath-by-breath pulmonary gas exchanges were recorded throughout the ramp using a metabolic analyzer (Quark CPET, module A-67-100-02, COSMED Deutschland GmbH, Fridol ng, Germany; desktop software: Omnia version 1.6.5). Prior to testing, the gas analyzers were calibrated according to the manufacturer's instructions. The protocol was terminated when the participant fell below a cadence of 60 rpm or due to self-determination. The following criteria served for the assumption of exhaustion: (A) heart rate >90% of the maximum predicted heart rate (prediction model according to [20]: 208 (0.7 age) and (B) respiratory quotient > 1.1. Maximum oxygen uptake (VO2max) and maximum HR (HRmax) were de ned as the average VO2and HR over the last 30 s of the test. 2.3. Data Processing After extraction of the 12-channel ECG data (converted from exported.xml les) and RR data (exported from the Elite HRV app) as text les, import into Kubios HRV Premium Software version 3.5.0 (Biosignal Analysis and Medical Imaging Group, Department of Physics, University of Kuopio, Kuopio, Finland, [19]) was conducted. For the 12-channel ECG, lead 2 was used as the comparable lead to the chest strap device [21]. Preprocessing settings were set to the default values including the RR detrending method which was kept at “smoothness priors” (Lambda = 500). The RR series was then corrected by the Kubios HRV Premium “automatic method” [22]. For DFA a1 calculation window width was set to 4 n 16 beats [23]. During rest conditions a 2-min time window (00:30–02:30 min:s) was chosen for the analysis. During the incremental exercise time-varying analysis was adjusted to a 2-min window width and 20-s grid interval for the moving window, so that the exported.csv les from Kubios HRV Premium Software contained the HRV metrics of interest (RR, HR, DFA a1) recalculated every 20 s. Data sets with artefacts >5% were excluded from analysis based on [24]. 2.4. Statistics Results are presented as mean standard deviation (SD). Normal distribution of data was checked by Shapiro–Wilk testing and visual inspection of data histograms. Agreement of the variables obtained by means of the two devices during the three

interest (RR, HR, DFA a1) recalculated every 20 s. Data sets with artefacts >5% were excluded from analysis based on [24]. 2.4. Statistics Results are presented as mean standard deviation (SD). Normal distribution of data was checked by Shapiro–Wilk testing and visual inspection of data histograms. Agreement of the variables obtained by means of the two devices during the three conditions (PRE, incremental exercise, POST) was evaluated using linear regression, Pearson's r correla- tion coef cient (r), Lin's Concordance Correlation Coef cient (rc), Intraclass Correlation Coef cient (ICC3,1), coef cient of determination (R 2 ), standard error of estimate (SEE) and Bland–Altman plots with limits of agreement (LoA) [25]. The size of Pearson's r correlation coef cient was evaluated as follows; 0.3 r < 0.5 low; 0.6 r < 0.8 moder- ate and r 0.8 high [26]. Computation of rcwas conducted using a S1Macro for SPSS (https://doi.org/10.1371/journal.pone.0239931.s002 sents a modi cation of Pearson's r correlation coef cient, in that it assesses not merely the distance of data points to the line of best t, but also how far this line deviates from the line of perfect agreement, as represented by the 45-degree line through the origin. Sizes of 0.8 are rated at almost perfect agreement [27]. Bland–Altman mean differences for data comparisons were expressed as either absolute or percentage bias (difference/mean 100 ). Analyze-it software (Version 5.66) was used for its automatic computation. In the case of normal distribution, paired t-test was used for the comparison of data, whereas Wilcoxon Signed Ranks Test was applied in case of violation of the precondition. In the case of non- normally distributed data, medians and estimates of the median differences along with the 95% con dence intervals (Hodges-Lehmann estimator) were additionally calculated using a SPSS Syntax code (Hodges-Lehmann Con dence Interval for Median difference|Raynald's SPSS Tools; [28]). Effect sizes were calculated with Cohen's d (d) and its respective thresh- olds (small effect = 0.20, medium effect = 0.50, large effect = 0.80; [29]. For all tests, the statistical signi cance was accepted asp 0.05. Analysis was performed using IBM ® SPSS ® Statistics