Validation of the Polar V800 heart rate monitor and comparison of artifact correction methods among adults with hypertension.

Heart rate variability (HRV) measurements via ambulatory monitors have become common. We examined the validity of recording R-R intervals using the Polar V800™ compared to 12-lead electrocardiograms (ECG) among middle-aged (44.7±10.1years); overweight to obese (29.8±4.3 kg.m-2) adults (n = 25) with hypertension (132.3±12.2/ 84.3±10.2 mmHg). After resting for 5-min in the supine position, R-R intervals were simultaneously recorded using the Polar V800™ and the 12-lead ECG. Artifacts present in uncorrected (UN) R-R intervals were corrected with the Kubios HRV Premium (ver. 3.2.) automatic (AC) and threshold-based (TBC) correction, and manual correction (MC) methods. Intra-class correlation coefficients (ICC), Bland-Altman limits of agreement (LoA), and effect sizes (ES) were calculated. We detected 71 errors with the Polar V800™ for an error rate of 0.85%. The bias (LoAs), ES, and ICC between UN and ECG R-R intervals were 0.69ms (-215.80 to +214.42ms), 0.004, and 0.79, respectively. Correction of artifacts improved the agreeability between the Polar V800™ and ECG HRV measures. The biases (LoAs) between the AC, TBC, and MC and ECG R-R intervals were 3.79ms (-130.32 to +137.90ms), 1.16ms (-92.67 to +94.98ms), and 0.37ms (-41.20 to +41.94ms), respectively. The ESs of AC, TBC, and MC were 0.024, 0.008, and 0.002, and ICCs were 0.91, 0.95, and 1.00, respectively. R-R intervals measured using the Polar V800™ compared to 12-lead ECG were comparable in adults with hypertension, especially after the artifacts corrected by MC. However, TBC correction also yielded acceptable results.

Introduction

Hypertension is the most common, costly, and preventable cardiovascular disease risk factor [1] in the US with nearly one in two American adults having hypertension [2]. Modifiable (e.g., obesity, physical inactivity, smoking, dyslipidemia) and non-modifiable (e.g., age, gender, family history) risk factors can increase the likelihood of developing hypertension [2]. However, the underlying etiology for the development of hypertension is largely unknown. One of the hypotheses regarding the initiation, progression, and maintenance of hypertension is alterations in the neural control of blood pressure (BP) [3]. These neural alternations generally emerge as excessive sympathetic stimulation and parasympathetic inhibition of the autonomic nervous system (ANS) [4]. The sinoatrial node integrates the inputs from the ANS to adjust heart rate (HR) in response to the constantly changing internal and external environment to maintain homeostasis [4]. This adjustment causes an oscillatory pattern in the HR resulting from beat-to-beat fluctuations in the time period between sequential heartbeats, termed heart rate variability (HRV). HRV is used to non-invasively measure ANS modulation.

Due to the recent technological advances in signal analysis, users can now utilize accessible, simple, and affordable heart rate (HR) monitors to measure HRV rather than the gold standard criterion of the 12-lead electrocardiogram (ECG). Indeed, a PubMed search showed the number of publications associated with HRV rose from 393 in 1995, the year of the first Polar HR monitor (Polar Vantage NV™) for measurement of HRV, to 1726 in 2019. Nevertheless, artifacts (i.e., random distortion of usual R-R intervals) with technical (e.g., excessive noise which causes incorrect detection of R peak) and physiological (e.g., ectopic beat) origins can substantially impair the ability of HR monitors to accurately detect R-R intervals that are used to calculate the time, frequency, and non-linear domains of HRV [5-7].

Adults with hypertension may have a higher number of artifacts in HRV recordings than healthy individuals. Low R-wave amplitude may be observed in patients with the early stages of hypertension before left ventricular hypertrophy develops, a typical response to chronic high BP [8]. Further, increased chest wall size and epicardial adipose tissue can lead to diminished R-wave amplitude among patients with hypertension since more than 80% of this population are overweight to obese [9]. Moreover, arrhythmias (e.g., premature ventricular contractions) can disrupt R-wave morphology leading to HRV recording errors. Arrhythmias are common among hypertensives due to the disturbances in sinus and atrioventricular nodes and electrolyte imbalances caused by antihypertension medications (e.g., diuretics) [10]. Since HRV is measured using R-R intervals, low R-wave amplitude and distorted R-wave morphology can affect HR monitors’ ability to correctly identify R-R intervals. Thus, correcting the artifacts may be even more important in this clinical population to acquire valid HRV measures and accurately interpret the autonomic status of these patients.

The manual correction (MC) of artifacts with visual identification of the R-R intervals combined with the choice of a proper correction method is the most accurate method for obtaining valid HRV measures [6, 7]. However, the MC is laborious and requires multiple individuals and long hours of training to properly utilize. Various HRV software packages, therefore, offer easy-to-use automatic correction methods to edit artifacts, making them more attractive than the MC. Kubios is the most popular HRV software that is used to automatically correct artifacts [11]. The most recent version of the software (Kubios HRV Premium [ver. 3.2]) provides two options for correcting technical and physiological artifacts, the: 1) automatic correction (AC) and 2) threshold-based correction (TBC) methods. The AC uses a time series with differences between sequential R-R intervals to identify artifacts. The TBC makes comparisons of each R-R interval against a local R-R interval average and corrects those that exceed or fall behind the local R-R interval average more than a selected threshold value [11].

Compromised autonomic modulation is a key element in the pathology of hypertension [12]. Over time, excess sympathetic activation combined with reduced parasympathetic influence may result in left ventricular remodeling and atherosclerosis due to increased inflammation potentially as a consequence of excess sympathetic activation in patients with hypertension, increasing the risk of microvascular (e.g., vascular dementia) and macrovascular (e.g., myocardial infarction) injuries [12]. In fact, significantly lower HRV (indicative of increased sympathetic activity, or decreased parasympathetic influence, or a combination of both) has been shown in studies of patients with controlled hypertension and left ventricular hypertrophy as well as in patients with uncontrolled hypertension compared to those with controlled hypertension without left ventricular hypertrophy [13-15]. Therefore, examining HRV in hypertensives with the aim of understanding autonomic function as an additional guide to treatment may be of notable clinical significance, potentially leading to improved prognosis of this population.

Typically, ECGs are used to measure HRV for analyzing the autonomic activity of clinical populations. However, patients report discomfort with ECGs due to their cumbersome and irritating adhesive electrodes and large number of leads, which may result in low patient compliance with treatment [16]. Validation of a portable, leadless, lightweight, and easy-to-use HRV monitor without skin preparation and electrode placement among adults with hypertension can help clinicians to optimize the efficacy of therapies for the treatment of hypertension and to improve patient experience. This can also improve the effectiveness of researchers collecting HRV data and evaluating the mechanisms underlying a patient’s hypertension. The Polar HR monitors including S810™ [17, 18], S810i™ [19], RS800CX™ [20], RS800GX™ [21], and the current version V800™ HR monitor [22-24] have been validated among healthy individuals. HR monitors, including the latest version of Polar V800™, have not been validated in adults with hypertension. Thus, the aims of this study were to examine 1) the validity of the Polar V800™ compared to the gold standard ECG and 2) the accuracy of AC and TBC methods of Kubios HRV Premium (ver. 3.2) and MC in correcting artifacts among adults with hypertension.

Materials and methods

Study population

This study was approved by the University of Connecticut and Hartford Hospital Institution of Review Boards. Subjects completed written informed consent approved by the Institutional Review Boards of the University of Connecticut and Hartford Hospital. Consistent with the Institutional Review Board approved inclusion and exclusion criteria of the other studies being conducted in the laboratory from October 2016 to May 2018, subjects in this study were recruited from the surrounding community with direct mailings and posting of flyers, media advertisements, social media, previous studies, and from places of work and college campuses with the posting of flyers, listservs, class announcements, and newsletters. Adults (n = 25) with hypertension, defined as having a systolic BP SBP≥130-<160 mmHg and/or diastolic BP DBP≥80-<100 mmHg and not on antihypertensive medication, or on antihypertensive medication with an SBP<160 mmHg and DBP<100 mmHg [2] were enrolled. The subjects’ physical activity levels ranged from being sedentary to regularly physically active (i.e., 150 min a week moderate or 75 min a week vigorous intensity exercise). They were free of cardiovascular, pulmonary, renal, metabolic, or other chronic diseases and depression; had not smoked for at least 6 months prior to entry; and consumed less than two alcoholic drinks daily. Subjects with a medical history of cancer-related lymphedema, those trying to gain or lose weight and women who were pregnant, or planning to become pregnant, or were lactating were also excluded.

Study overview

Subjects were recruited on a rolling basis from other studies conducted in the laboratory from October 2016 to May 2018 from which their baseline data were analyzed for the purpose of this study. All subjects attended one laboratory session between 6:00 and 11:00 am. Prior to the resting HRV measurement, subjects’ chests were cleaned for the ECG attachment of the 12-lead ECG electrodes. Later, the chest strap (H7) of the Polar V800™ HR monitor was placed just below the pectoralis major muscles. Then, resting HRV was measured in the supine position. The R-R interval time series from ECG was stored in CASE™ (ver. 6.6) GE Healthcare system while those from Polar V800™ HR monitor was automatically stored in the Polar Flow web service.

Procedures

Subjects were informed to consume a light breakfast 2–3 hours and abstain from drinks containing alcohol and caffeine at least 12 hr before testing. Body mass index (BMI kg.m-2) was measured from body weight and height with a calibrated balance beam scale, and waist circumference (WC) was measured using a Gullick tape at the narrowest part of the torso [25]. Resting BP was measured following the American Heart Association’s standard procedures with an automated BPTRU monitor (BPTRU Medical Devices; Coquitlam, Canada). The measurements of BMI, WC, and BP preceded the HRV recording.

The ECG electrodes were placed in the Mason-Likar configuration using the GE Stress System (CASE, Milwaukee, WI) [26]. The ECG signal was checked to ensure that it was consistent and free of noise. After the signal was confirmed to be acceptable, the Polar V800™ H7 chest strap was fitted below the pectoralis major muscles and applied as described by the manufacturer. Subjects were in the supine position for 10 min with the last 5 min recorded for the analysis in a quiet, low light, and temperature-controlled room. During the recording, subjects were instructed to relax as much as possible but not sleep, move, or talk. Breathing frequency was paced at 12 breaths min–1 using a metronome to control for the respiratory influences on HRV measures (particularly HF or respiratory band).

Following the HRV measurement, a maximal cardiopulmonary graded exercise stress test (GEST) was performed using the Balke protocol on a Trackmaster treadmill (Full Vision, Inc., Newton, KS) to determine peak oxygen consumption (VO2peak) which was examined in the statistical analysis for its potential influence on HRV measures.

R-R interval processing

Initially, the 12-lead ECG and Polar V800™ HR monitor were started to simultaneously record R-R intervals at an acquisition rate of 1000 Hz. Cardiology XML files obtained from the ECG were imported into Kubios HRV Premium (ver. 3.2) (The Biomedical Signals Analysis and Medical Imaging Group, University of Kuopio, Finland) in order to export the R-R intervals detected automatically by its built-in QRS detection algorithm. Kubios HRV Premium (ver. 3.2) marked each R-wave with “+” sign that could be moved or removed to correct falsely detected R-waves. The detected R-R intervals were manually inspected on the ECG tachogram to ensure there were no false R-wave detections. If an R-wave was falsely determined, one of the authors (BTC) replaced it by moving “+” signs on the correct R-wave. The detected R-R intervals from ECG were then checked by another author (DAG) to ensure data accuracy and saved in a space delimited ASCII text file. In addition, the R-R intervals recorded by Polar V800™ HR monitor were exported from the Polar Flow web service (ver. 2.3; Polar Electro Oy, Kempele, Finland) in a space delimited ASCII text file.

Artifact identification

First, the R-R intervals from the ECG and Polar V800™ HR monitors were imported into the same spreadsheet side-by-side. Next, the R-R intervals from both devices were synchronized by inserting 0 ms interval to ease the comparison between the ECG and uncorrected (UN) V800™ HR monitor R-R intervals (these were subsequently removed prior to analysis). The technical artifacts from the Polar V800™ HR monitor recordings or physiological artifacts were then identified by comparing the R-R intervals from both devices. An artifact from the V800™ HR monitor was identified when the differences between ECG and Polar V800™ HR monitor R-R intervals were greater than 20 ms. Subsequently, the differences between the R-R intervals of ECG and Polar V800™ HR monitor (i.e., type of artifacts) were assigned one of seven error types of the error identification and correction guideline developed by Gamelin et al. [18] that was recently updated by Giles et al. [22] Table 1. Artifact identification was performed by the first author (BTC) and each step of this process was checked by a co-author (DAG).

Table 1

Types, description, and correction methods of artifacts.

Type	Description	Correction
T1	A positive or negative single interval difference greater than 20 msa	Artifact noted but not corrected
T2	A long interval followed by a short interval while the two R-R intervals either side were less than 20 ms	Long and short intervals averaged
T3	A short interval followed by a long interval while the two R-R intervals on either side were less than 20 ms	Short and long intervals averaged
T4	When the HR monitor missed R-R intervals equivalent to two or three ECG R-R intervals	Artificially long R-R interval divided by the number of missed R-waves (beats)b
T5	When the HR monitor detected extra short R-R intervals equivalent to one ECG R-R interval	Two extra short R-R intervals added together to approach corresponding ECG R-R interval
T6a	When the HR monitor entirely missed the R-R interval, while there was no difference in the HR monitor time stampa	Artifact noted but not corrected
T6b	When the HR monitor entirely missed the R-R interval and there was a larger than expected gap in the HR monitor time stamp	The known interval subtracted from the difference in time stamp

aArtifact is undetectable without simultaneous ECG recording.

bR-R interval of 0 ms was placed to maintain time synchrony between both devices.

Artifact correction

Artifacts were manually and automatically corrected after their identification. All artifact corrections were performed by the first author (BTC) and each step of this process was checked by a co-author (DAG). The MC was used only for T2, T3, T4, T5, and T6b artifacts present in R-R intervals derived from the Polar V800™ following the guidelines provided in Table 1. The T1 and T6a artifacts in the Polar V800™ R-R intervals, however, were not corrected since it is not possible to identify both artifacts without simultaneous ECG recordings The corrected Polar V800™ R-R intervals by the MC were then imported into Kubios HRV Premium (ver.3.2) to calculate the HRV measures. All non-sinus beats present in the ECG signal (N = 7) were replaced by interpolated R-R intervals from adjacent R-R intervals.

The automatic correction (AC) and threshold-based correction (TBC) methods of Kubios HRV Premium (ver. 3.2) were used to automatically correct the artifacts. For the AC, “Automatic Correction” under the “R-R Interval Series Option” was selected and then the “Apply” button was selected to correct the artifacts. For the TBC, an appropriate threshold among very low (0.45 sec), low (0.35 sec), medium (0.25 sec), strong (0.15 sec), very strong (0.05 sec), and custom under “R-R Interval Series Option” was selected and then the “Apply” button was selected to correct the artifacts. When the artifacts were corrected with the TBC, the lowest level of correction was chosen to prevent potential overcorrection. When artifacts were detected by both methods, they were automatically replaced through the cubic spline interpolation by the software package. After the artifact corrections were made, the R-R interval time series was then considered normal and defined as N-N intervals [27].

Calculation of HRV measures

Following the identification and correction of the artifacts, an identical segment of N-N intervals was selected from the ECG and Polar V800™ for the calculation of the various HRV domain measures. Commercial Kubios HRV Premium (ver. 3.2) analyzed the selected segments to obtain the time, frequency, and non-linear domains of HRV measures.

The time domain measures quantify the amount of variability within the sample and they included: standard deviation of normal R-R intervals (SDNN), root mean square of successive differences in normal R-R intervals (RMSSD), and percentage of successive normal R-R intervals greater than 50 ms (pNN50%). Frequency domain or power spectral density analysis reveal the content of signal’s power (variance) versus frequency and can be analyzed by the autoregressive and fast Fourier transformation methods [27]. A fast Fourier transformation was performed to quantify power spectrum density into absolute low frequency (LF ms2: 0.04–0.15 Hz) and absolute high frequency (HF ms2: 0.15–0.40 Hz) domains. In addition, normalized LF (LF nu), normalized HF (HF nu), and LF/HF ratio were calculated. Non-linear domain measures represent the chaotic heartbeat dynamics caused by the complex interactions between several regulatory systems. We included sample entropy (SampEn) as a measure of non-linear methods. SampEn measures the signal complexity or irregularity where large SampleEn suggests high irregularity and smaller SampleEn a more regular signal [11].

Measurement of R-wave amplitude

The amplitude of the R-wave from the subjects’ ECG trace was measured in mm (10 mm = 1 mV) from the PR segment to the top of the R wave and averaged for each subject’s lead [28]. Lead II was chosen for the R-wave amplitude measurement because it lies close to the cardiac axis (i.e., the overall direction of electrical movement), allowing for the best views of R-waves [29].

Statistical analysis

The independent sample t-test was used to report the mean differences between the subject characteristics of men and women. The magnitude of difference between the R-R intervals and HRV domain measures from the ECG and Polar V800™ HR monitor were calculated by measuring the effect size (ES) as the mean difference over the standard deviation of the difference [30]. The ES was defined as trivial when ES <0.2, small when ES was between ≥0.2 and <0.5, moderate when ES between ≥0.5 and <0.8, and large when ≥0.8 [31]. The intraclass correlation coefficient (ICC) with the 95% confidence interval (CI) assessed the concurrent validity (or interchangeable agreement) of the R-R intervals and HRV measures [32]. We selected a two-way mixed model with absolute agreement and reported absolute measures (mean of two measurements). The ICC was defined as poor when ICC was <0.50, moderate when ICC was between ≥0.50 and <0.75, good when ICC was between ≥0.75 and <0.90, and excellent when ICC was ≥0.90 [33]. Bland-Altman plots were created for the ECG R-R intervals versus the Polar V800™ HR monitor for the uncorrected (UN), AC, TBC, and MC R-R intervals. The 95% limits of agreement (LoA) for lower (-1.96) and upper limit (+1.96) were calculated as follows: 1) lower limit: mean difference—(SD of difference x 1.96); and 2) upper limit: (SD of difference x 1.96) + mean difference [34]. Homoscedasticity and heteroscedasticity were inspected through a histogram and Q-Q plot. The RMSSD, pNN50%, LFms2, and HFms2 were logarithmically transformed before the calculation of LoA ranges. Finally, the sub-group analyses of gender, BMI, medication use, and aerobic capacity were performed using the same statistical methods as in all subjects to study their potential impacts on the agreement between Polar V800™ and ECG. Gender and medication use were grouped into “male and female” and “medication and non-medication”. Further, BMI and aerobic capacity were divided into “overweight (25.0–29.9 kg/m2)” and “obesity (>30 kg/m2)” and “very poor to poor (15.6–37.5 ml/kg/min)”, “fair to good (32.0–41.5 ml/kg/min)”, and “excellent (48.6–51.7 ml/kg/min)”, respectively [25]. All statistical analyses were conducted using SPSS (ver. 24; Chicago, IL, USA).

Results

Subject characteristics

Participants were 17 men (68%) and 8 women (28%) between 18 and 55 years who were overweight to obese with hypertension and of very poor to good cardiorespiratory fitness for men and women of their age Table 2. Of these, six subjects were taking antihypertension medication that included diuretics (n = 1), angiotensin II receptor antagonists (n = 3), and β-Blockers (n = 2) S1 Dataset.

Table 2

Subjects characteristics of the total sample and by sex (mean ± SD).

Variable	Total Sample (n = 25)	Men (n = 17)	Women (n = 8)
Age (year)	44.7±10.1	42.5±10.3	49.4±8.4
Height (cm)	172.3±11.0	176.9±9.6a	162.6±5.4
Weight (kg)	93.7±30.6	90.9±16.2	86.4±17.2
Body Mass Index (kg.m-2)	29.8±4.3	28.2±3.9	32.0±5.3
Waist Circumference (cm)	99.9±12.1	95.9±10.7	105.1±14.0
Heart Rate (bpm)	73.6±11.0	71.8±11.9	78.4±7.2
Systolic Blood Pressure (mmHg)	132.3±12.2	129.6±11.8 a	140.9±7.9
Diastolic Blood Pressure (mmHg)	84.3±10.2	84.4±11.3	86.5±7.5
Maximum Oxygen Consumption (VO2peak)	33.5±9.8	38.0±8.2a	24.1±5.5

aSignificant differences at P < 0.05.

Agreement of R-R intervals

The R-wave amplitude of ECG lead II was 0.26±0.09 mV, which represents a depressed R-wave amplitude defined as R-wave amplitude of 0.5 mV or less compared to healthy individuals (29). There were 71 artifacts among the 8325 total R-R intervals yielding an error rate of 0.85%. The most often detected type of error among the total number of artifacts included T4 (49x; 69%) followed by T5 (7x; 9.9%), T1 (5x; 7.0%), T3 (5x; 7.0%), T6a (2x; 2.8%), T6b (2x; 2.8%), and T2 (1x; 1.4%). The length of HRV measurement for eight subjects was less than 5 min due to the loss of connection between the HR monitor and the strap for unknown reasons. Therefore, the average length of the HRV measurement was 4.6±0.9 min and number of R-R intervals was 333.0±70.5.

Fig 1 contains the Bland-Altman plots of the LoA and agreement between the R-R intervals from the ECG and the UN, AC, TBC, and MC R-R intervals from the Polar V800™ HR monitor. The UN R-R intervals from the Polar V800™ HR monitor were corrected using the AC and TBC methods of Kubios Premium (ver. 3.2) and MC. While the bias (0.69 ms) and ES (0.004) were small, the UN R-R intervals resulted in the widest range of LoA (from -215.80 to 214.42 ms) (Fig 1a). The AC method using Kubios HRV Premium (ver. 3.2) resulted in corrected R-R intervals with higher bias (3.79 ms), range of LoA (from -130.32 to 137.90 ms), and ES (0.024) (Fig 1b) than the bias (1.16 ms), range of LoA (from -92.67 to 94.98 ms), and ES (0.008) of the TBC method (Fig 1c). The MC method produced corrected R-R intervals with the smallest bias (0.37 ms), tightest range of LoA (-41.20 to 41.94 ms), and smallest effect size (0.002) (Fig 1d). Furthermore, an improvement in the ICC of the UN R-R intervals occurred depending on the method of correction that was used. The ICC of UN R-R intervals went from 0.79 (95% CI 0.78–0.80) to 0.91 (95% CI 0.90–0.91) in the AC R-R intervals, 0.95 (95% CI 0.95–0.95) in the TBC R-R intervals, and 1.00 (95% CI 1.00–1.00) in MC R-R intervals S2 Dataset.

Fig 1

Bland-Altman plots for the ECG and the UN Polar V800™ HR monitor R-R intervals (a) Bland-Altman plots for the ECG and the AC Polar V800™ HR monitor R-R intervals (b) Bland-Altman plots for the ECG and the TBC Polar V800™ HR monitor R-R intervals (c) Bland-Altman plots for the ECG and the MC Polar V800™ HR monitor R-R intervals (d) ECG: electrocardiogram; HR: heart rate; UN: uncorrected R-R intervals; AC: R-R intervals corrected by automatic correction method of the Kubios HRV Premium (ver 3.2); TBC: R-R intervals corrected by threshold-based correction method of the Kubios HRV Premium (ver 3.2); MC: R-R intervals corrected manually.

Agreement of HRV measures

Tables 3, 4, 5 and 6 contain the comparisons of the HRV measures calculated from UN, AC, TBC, and MC Polar V800™ HR monitor R-R intervals versus those calculated from ECG R-R intervals using the Kubios Premium (ver. 3.2) software.

Table 3

Comparison of HRV measures calculated from UN Polar V800™ and ECG R-R intervals (mean ± SD).

HRV Measure	ECG (mean±SD)	Polar UN (mean±SD)	Bias (LoA)	ICC (95% CI)	Effect Size
SDNN (ms)	55.5±26.7	90.5±62.2	-34.95 (-143.58 to 73.69)	0.27 (-0.07–0.57)	0.730
RMSSD (ms)	38.3±29.4	92.8±96.3	-57.51 (-223.81 to 114.78)	0.21 (-0.11–0.52)	0.765
pNN50 (%)	13.6±16.2	15.1±15.9	-1.50 (-6.14 to 3.21)	0.98 (0.95–0.99)	0.091
LF (ms2)	1215±1352.1	6924.2±16947.2	-5709.13 (-38536.30 to 27118.05)	0.53 (-0.97–0.57)	0.475
HF (ms2)	740.8±1262.8	4861.0±6863.9	-4120.18 (-17136.40 to 8896.033)	0.07 (-0.21–0.39)	0.835
LF (nu)	67.3±18.7	59.7±20.3	7.53 (-30.13 to 45.20)	0.49 (0.14–0.73)	0.387
HF (nu)	32.7±18.6	40.2±20.2	-7.53 (-45.20 to 30.13)	0.49 (0.14–0.73)	0.387
LF/HF Ratio	3.6±4.0	3.0±4.2	0.67 (-3.37 to 4.71)	0.87 (0.72–0.94)	0.163
SampEn	1.5±0.2	1.2±0.5	0.33± (-0.58 to 1.24)	0.39 (-0.18 to 0.71)	0.836

ECG: electrocardiogram; UN: uncorrected R-R intervals; ICC: intra-correlation coefficient; LoA: limits of agreement; SDNN: standard deviation of normal-to-normal N-N intervals; RMSSD: root mean square of the successive difference of intervals; pNN50%: the percentage of successive normal cardiac inter-beat intervals greater than 50 msec; LF: low frequency; HF: high frequency; SampEn: Sample Entropy; SD: standard deviation.

Table 4

Comparison of HRV measures calculated from Kubios Premium (ver. 3.2) AC Polar V800™ and ECG R-R intervals (mean ± SD).

HRV Measure	ECG (mean±SD)	Polar AC (mean±SD)	Bias (LoA)	ICC (95% CI)	Effect Size
SDNN (ms)	55.5±26.7	57.1±28.1	-1.60 (-23.27 to 20.06)	0.92 (0.83–0.96)	0.058
RMSSD (ms)	38.3±29.4	36.2±25.4	2.13 (-25.57 to 29.84)	0.87 (0.73–0.94)	0.078
pNN50 (%)	13.6±16.2	13.1±16.1	0.51 (-7.28 to 8.29)	0.97 (0.93–0.99)	0.031
LF (ms2)	1215±1352.1	1214.2±1360.9	0.80 (-369.45 to 371.06)	0.99 (0.98–1.00)	0.001
HF (ms2)	740.8±1262.8	749.1±1239.5	-8.27 (-1115.84 to 1099.293)	0.90 (0.79–0.96)	0.007
LF (nu)	67.3±18.7	66.7±19.6	0.57 (-24.29 to 25.44)	0.79 (0.57–0.90)	0.030
HF (nu)	32.7±18.6	33.3±19.6	-0.57 (-25.44 to 24.29)	0.79 (0.73–0.95)	0.030
LF/HF Ratio	3.6±4.0	3.5±3.9	0.12 (-1.76 to 2.00)	0.97 (0.94–0.99)	0.031
SampEn	1.5±0.2	1.5±0.3	0.02 (-0.21 to 0.21)	0.81 (0.57–0.91)	0.039

AC: R-R intervals corrected by automatic correction method of Kubios HRV Premium (ver. 3.2).

Table 5

Comparison of HRV measures calculated from Kubios Premium (ver. 3.2) TBC Polar V800™ and ECG R-R intervals (mean ± SD).

HRV Measure	ECG (mean±SD)	Polar TBC (mean±SD)	Bias (LoA)	ICC (95% CI)	Effect Size
SDNN (ms)	55.5±26.7	56.1±27.6	-0.55 (-10.59 to 9.49)	0.98 (0.96–0.99)	0.020
RMSSD (ms)	38.3±29.4	37.8±27.5	0.53 (-9.57 to 10.63)	0.98 (0.96–0.99)	0.019
pNN50 (%)	13.6±16.2	13.8±16.2	-0.16 (-1.33 to 1.02)	0.99 (0.98–1.00)	0.010
LF (ms2)	1215±1352.1	1209.2±1352.5	5.86 (-275.86 to 287.59)	1.00 (0.99–1.00)	0.004
HF (ms2)	740.8±1262.8	703.7±1201.6	37.07 (-460.59 to 534.74)	0.98 (0.95–0.99)	0.030
LF (nu)	67.3±18.7	67.3±18.0	-0.02 (-4.04 to 4.01)	0.99 (0.99–1.00)	0.001
HF (nu)	32.7±18.6	32.7±18.0	0.02 (-4.01 to 4.04)	0.99 (0.99–1.00)	0.001
LF/HF Ratio	3.6±4.0	3.5±3.9	0.12 (-0.79 to 1.02)	0.99 (0.98–1.00)	0.029
SampEn	1.5±0.2	1.5±0.3	0.02 (-0.35 to 0.40)	0.93 (0.85–0.97)	0.039

TBC: R-R intervals corrected by threshold-based correction method of Kubios HRV Premium (ver. 3.2).

Table 6

Comparison of HRV measures calculated from MC R-R intervals between the ECG and polar (mean ± SD).

HRV Measure	ECG (mean±SD)	Polar MC (mean±SD)	Bias (LoA)	ICC (95% CI)	Effect Size
SDNN (ms)	55.5±26.7	55.3±26.7	0.21 (-0.70 to 1.12)	1.00 (1.00–1.00)	0.009
RMSSD (ms)	38.3±29.4	38.1±29.1	0.18 (-1.82 to 2.18)	1.00 (0.99–1.00)	0.006
pNN50 (%)	13.6±16.2	13.7±16.1	-0.11 (-1.52 to 1.30)	1.00 (0.99–1.00)	0.007
LF (ms2)	1215±1352.1	1208.2±1351	6.85 (-65.77 to 79.48)	1.00 (1.00–1.00)	0.005
HF (ms2)	740.8±1262.8	730.9±1267.1	9.94 (-44.20 to 64.07)	1.00 (1.00–1.00)	0.008
LF (nu)	67.3±18.7	67.6±18.6	-0.35 (-1.68 to 0.98)	1.00 (1.00–1.00)	0.019
HF (nu)	32.7±18.6	32.3±18.6	0.35 (-0.98 to 1.68)	1.00 (1.00–1.00)	0.019
LF/HF Ratio	3.6±4.0	3.7±4.0	-0.04 (-0.23 to 0.14)	1.00 (1.00–1.00)	0.011
SampEn	1.5±0.2	1.5±0.2	0.01 (-0.11 to 0.12)	0.97 (0.93–0.99)	0.001

MC: R-R intervals corrected manually.

The UN resulted in the largest biases and the widest LoA ranges for all HRV measures compared to the AC, TBC, and MC methods, except for pNN50% (ES: 0.091; ICC: 0.98) and LF/HF ratio (ES: 0.163; ICC: 0.87) Table 3. The UN ES ranged from 0.091 to 0.835, indicating the magnitude of difference between the HRV measures calculated from UN Polar V800™ HR monitor and ECG R-R intervals was trivial to large, and had five out of 9 HRV domain measures between 0.2≤ and <0.5, indicating a moderate difference. Additionally, the UN ICCs ranged from 0.070 to 0.98 suggesting poor to excellent agreement, but had six out of nine HRV measures below <0.5, indicating poor agreement. Between the two correction methods of Kubios HRV Premium (ver. 3.2), the TBC resulted in smaller biases than the AC for SDNN, RMSSD, pNN50%, LFnu, HFnu, LF/HF ratio, but the AC produced smaller biases than TBC for LFms2, HFms2 and SampEn Tables 4 and 5. Additionally, when the TBC was compared to the AC method, the TBC produced tighter LoA ranges and smaller ES (<0.031 versus <0.085) for all HRV domain measures except for SampEn.

Moreover, the TBC method exhibited ICCs varying from 0.96 to 1.00, while the AC had ICCs ranging from 0.79 to 0.99, indicating that the agreeability slightly improved from good-to-excellent to excellent when the correction was made with the TBC rather than AC. The MC produced the smallest biases for SDNN, RMSSD, pNN50%, LF/HF, and SampEn compared to the AC and TBC methods but resulted in larger biases for LFms2 and HFms2 than AC, and for LFms2, LFnu, and HFnu than TBC Table 6. Additionally, when the MC was compared to the AC and TBC methods, the MC resulted in the tightest LoA ranges and smallest ES (<0.012) for all HRV measures. Furthermore, the MC had ICC of 0.99 or 1.00 for all HRV measures indicating excellent agreement for all HRV measures S3 Dataset.

Agreement of HRV measures separated by medication use, gender, BMI, and aerobic capacity

The magnitude of biases, LoA ranges, ICCs, and effect sizes of HRV measures of the subjects with medication use S2A–S2D Table, males S2A–S2D Table, those with overweight BMI S3A–S3D Table, and those with very poor-to-poor aerobic capacity S4A–S4D Table were similar to those of non-medication, female, obesity, and excellent aerobic capacity between the Polar V800™ and ECG. Also, as was in all subjects, the correction of artifacts substantially improved the disagreement seen between UN HRV measures of the Polar V800™ and ECG. Moreover, the accuracy of artifact correction methods in correcting R-R intervals were similar to those reported in all subjects (AC<TBC≤AC) S4–S7 Datasets.

Discussion

We compared uncorrected and corrected R-R intervals and HRV measures derived from the Polar V800™ HR monitor to the gold standard 12-lead ECG, with the aim of determining their level of agreement among adults with hypertension. In addition, we sought to determine the accuracy of the MC, AC and TBC methods of Kubios HRV Premium (ver.3.2) in correcting artifacts among this sample. Artifact correction was required due to the large biases, LoA ranges, ES, and poor ICC of HRV measures between the ECG and uncorrected Polar V800™ HR monitor data with the exceptions of pNN50% and LF/HF ratio. The results demonstrate that the MC consistently produced the lowest biases, LoA ranges, ES, and highest ICC for all correction methods followed by TBC and AC in turn. The results showed our subjects had a depressed R-wave amplitude that might have resulted from the clustering effect of being overweight to obese and having hypertension [8, 29], increasing the likelihood of the occurrence of artifacts in this population.

The rate of the occurrence of artifacts in the Polar V800 R-R interval time series in the current study population (0.85%) was higher than those reported by Giles et al. [22] (0.08%) and [23] (0.10%) who originally validated the Polar V800™ among healthy adults (n = 20 [3 women and 17 men]; n = 18 [all men]) with a normal BMI (18.5–24.9 kg.m-2). The data acquisition and analyses methods were the same between our study and the other two studies. This suggests that the performance of the Polar V800™ in the detection of R-R intervals somewhat declined in adults with hypertension. Typically, a low R-wave amplitude is associated with overweight or obesity [29] due to the electrical insulating role of excess epicardial and subcutaneous adipose tissue on chest wall of individuals with these conditions [9]. Additionally, Bacharova et al. [8] found an association between low R-wave amplitude among adults in the early stages of hypertension without structural changes in myocardium. Therefore, the low R-wave amplitude of 0.26±0.09 mV we found in the subjects in our study could be the result of them being overweight to obese, having hypertension, or both. This may also explain the higher occurrence of technical artifacts in this study compared to those reported by Giles et al. [22] and [23] among healthy adults with normal weight and BP as diminished R-wave amplitude increases the chance of artifact occurrence [7].

We identified the artifacts in the Polar V800™ R-R interval time series following the guidelines established by Giles et al. [22]. The most common type of error we encountered with the Polar V800™ uncorrected R-R intervals was T4 of 69%, which is often attributed to a decrease in R-wave amplitude due to a loss or decrease in connection between the chest strap and skin [22]. However, the depression of R-wave amplitude due to the subjects in our study being overweight to obese and having hypertension may have contributed to the greater occurrence of T4 errors compared to other studies involving healthy subjects [18, 22, 23]. Of note, the Polar V800TM uncorrected T1 and T6a errors can be seen on the corrected Bland–Altman plots (Fig 1D) as outliers, a similar finding reported by Giles et al. [22]. T1 and T6 errors are not visible without simultaneous ECG recording.

The uncorrected R-R intervals resulted in large biases and LoA ranges, poor ICC, and large ES, necessitating the correction of artifacts prior to calculation of HRV measures, as is recommended [27]. The MC was the most valid method of correction with the lowest bias, tightest LoA ranges, and highest ICC compared to the AC and TBC methods. Time, frequency, and non-linear HRV measures calculated from the Polar V800™ R-R intervals corrected by the MC and those calculated from ECG showed excellent agreement with consistent small biases, tight LoA ranges, trivial ES, and excellent ICC, similar to levels of agreement found in previous studies with the Polar V800™ [22-24].

The MC of artifacts in R-R interval time series is time consuming, thus automated tools have been developed for the identification and correction of such data. Kubios HRV Premium is possibly the most extensively used software package for this purpose. However, we observed the TBC of Kubios HRV Premium (ver. 3.2) was not able to properly correct T4 error containing ≥2 missed beats, the same issue detected previously by [23] with Kubios HRV (ver. 2.2). More specifically, the TBC replaced the T4 error containing ≥2 missed beats with one interval rather than the required ≥2, leading to HRV measures with wider LoA ranges than the MC. Therefore, correcting artifacts with the TBC may be the more convenient option for users who find the MC complicated and laborious.

The AC of Kubios HRV was the least accurate correction method with the largest biases, LoA ranges, ES and lowest ICCs compared to the MC and TBC, which may be explained by two critical issues we identified with the AC method. First, the AC was not able to properly correct T4 errors containing >2 missed beats, rather the erroneous beat was replaced with one interval rather than the required >2 intervals. Second, the AC always overcorrected three intervals at the beginning and end of an R-R interval time series whether there were artifacts or not. Considering even a single distortion in R-R intervals results in exaggerated measures of HRV [6, 35, 36] overcorrecting six intervals despite no artifacts present may be largely the reason for the poorest performance of the AC compared to the MC and TBC.

This is the first study to validate the Polar V800™ HR monitor in detecting R-R intervals and producing agreeable HRV measures among adults with hypertension who were also overweight or obese. The occurrence of technical artifacts was greater in our sample than in healthy individuals with normal weight and BP, necessitating the need for artifact correction in this clinical population. General knowledge and prior experience of HRV data and artifact correction are essential to ensure optimum data quality regardless of the selection of artifact correction method. Ideally, users should choose MC as a first option for correcting Polar V800™ artifacts following the current artifact correction guidelines, as MC results in the tightest LoA ranges, smallest ES, and excellent ICC compared to the automatic correction methods of Kubios HRV Premium (ver. 3.2.). However, Kubios HRV Premium (ver. 3.2.) methods may be of greater practical utility for high throughput situations where the correction of artifacts using MC is not practical. In that case, the TBC should be selected as the preferred option over AC since TBC produced smaller biases than the AC for all HRV measures except LF ms2 and HFms2 as well as tighter LoAs, and smaller ES for all HRV measures. Therefore, if MC is not feasible, TBC should be the method of choice when choosing a method of correction for Polar V800™ artifacts among adults with hypertension. Finally, Kubios should improve its automatic correction methods by addressing the previously mentioned issues with TBC and AC to optimize HRV data quality among this population.

Individual subject characteristics.

(XLSX)

Click here for additional data file.

Individual UN, AC, TBC, and MC R-R intervals.

(XLSX)

Click here for additional data file.

Individual HRV measures calculated from the UN, AC, TBC, and MC R-R intervals.

(PDF)

Click here for additional data file.

Individual HRV measures separated by medication use calculated from UN, AC, TBC, and MC R-R intervals.

(XLSX)

Click here for additional data file.

Individual HRV measures separated by gender calculated from UN, AC, TBC, and MC R-R intervals.

(XLSX)

Click here for additional data file.

Individual HRV measures separated by BMI calculated from UN, AC, TBC, and MC R-R intervals.

(XLSX)

Click here for additional data file.

Individual HRV measures separated by aerobic capacity calculated from UN, AC, TBC, and MC R-R intervals.

(XLSX)

Click here for additional data file.

(A-D) Comparison of HRV measures separated by medication use calculated from UN, AC, TBC, and MC R-R intervals (mean ± SD).

(PDF)

Click here for additional data file.

(A-D) Comparison of HRV measures separated by gender calculated from UN, AC, TBC, and MC R-R intervals (mean ± SD).

(PDF)

Click here for additional data file.

(A-D) Comparison of HRV measures separated by BMI calculated from UN, AC, TBC, and MC R-R intervals (mean ± SD).

(PDF)

Click here for additional data file.

(A-D) Comparison of HRV measures separated by aerobic capacity calculated from UN, AC, TBC, and MC R-R intervals (mean ± SD).

(PDF)

Click here for additional data file.

23 Jul 2020

PONE-D-20-14879

Validation of the Polar V800TM Heart Rate Monitor and Comparison of Artifact Correction Methods among Adults with Hypertension

PLOS ONE

Dear Dr. Cilhoroz,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Sep 06 2020 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Daniel Boullosa

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: No

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Review of manuscript PONE-D-20-14879 entitled "Validation of the Polar V800TM Heart Rate Monitor and Comparison of Artifact Correction Methods among Adults with Hypertension" for the PLOS ONE.

The Polar V800 sport tester have been verified in a group of healthy subjects and athletes. This manuscript validated the Polar V800 for a group of overweight and obese subjects. I still lack such a study in the literature. The results of this manuscript are important because it is known that overweight and obese subjects have poor heart rate variability (HRV) compared to healthy subjects. Therefore, HRV of overweight and obese subjects should be monitored, preferably at home. Small, lightweight and cheap devices for ECG recording, such as a sport tester, enable the development of telemedicine. In addition, this manuscript compared several methods of artifact detection and correction. In my opinion, this important issue is not sufficiently discussed in the literature. I therefore consider this manuscript to be an interesting and it adds knowledge in the literature. However, there are some issues in the manuscript that should be solved. Please see my comments below.

Page 5: You have stated that all relevant data are within the manuscript and its Supporting Information files. However, I cannot find any data files from which the results are based.

Page 8: I propose to replace “… beyond that of healthy populations.” with “… beyond that of patients”. The use of medical ECG device has usually been limited to the diagnostics of patients with suspected cardiovascular condition.

Page 9: You should mention the negative factors that are associated with hypertension – excessive body mass and/or low physical activity.

Page 9: You should describe what HRV is in a few sentences.

Page 9: I propose to replace “… with technical (e.g. missed beats) …” with “… with technical (e.g. excessive noise which causes incorrect detection of R peak) ...”

Page 9: I propose to replace “… and physiological (e.g. non-sinus beats) …” with “… and physiological (e.g. ectopic beat) …”

Page 10: I do not think atherosclerosis is the result of poor ANS activity. This is more due to a change in distribution of different types of cholesterol and hormonal imbalances.

Page 13: “breaths/min–1” is incorrect. Please replace with “breaths⋅min–1” or “breaths/min”.

Page 14: Is peak oxygen consumption really a potential covariate of sport tester validity? How was it tested in this manuscript? I have found very limited information in the result section.

Page 14: The subtitle “R-R Interval Recording” is rather “R-R Interval Processing”. Please consider rephrasing.

Page 14: “BTC” should be replaced by “one of the authors (BTC)”.

Page 14: So, the correct position of the R peak was performed by only one author? Given that R-R intervals obtained from a medical ECG device were used as a gold standard in this manuscript, I consider that a single author checking is not sufficient.

Page 14: Please describe the MC method in more detail. Who performed it? In which kind of software? Did the person have RR intervals available from the medical ECG device or not?

Page 16: “… in both signals (N = 3)” does it mean three occurrences of artifacts in all group of RR recordings? Or does it mean that recordings from three subjects contains artifacts?

Page 17: “unpredictable heartbeat dynamics” should be replaced by ”chaotic heartbeat dynamics”. A chaotic system is theorethically predictable, but it is a complex nonlinear system, so in practise the chaotic system looks stochastic (random). Mathematical theory dealing with nonlinear HRV indexes considers ANS to be chaotic, not stochastic. See, for example, Ernst G (2014). Heart rate variability. London: Springer.

Page 17: I agree that Poincaré plot is one of the graphical methods used to study chaotic systems. Poincaré plot can therefore be considered as a nonlinear method. However, I cannot agree that SD1, SD2, and SD2/SD1 can be considered as nonlinear indexes of HRV. It was shown that these indexes can be calculated from SDNN and SDSD (Sassi R, Cerutti S, Lombardi F, Malik M, Huikuri H V, Peng C-K, … Yamamoto Y (2015). Advances in heart rate variability signal analysis: Joint position statement by the e-cardiology ESC working group and the european heart rhythm association co-endorsed by the asia pacific heart rhythm society. Europace, 17, 1341–1353). Therefore these indexes should be considered as time domain indexes. Sample entropy is nonlinear HRV index suitable for short-term (≈5 min) RR recordings. Therefore, I suggest removing the indexes SD1, SD2, and SD1/SD2 and adding sample entropy.

Page 18: ”inter correlation coeficient” should be replaced by “intraclass correlation coefficient”.

Page 18: What type of ICC was used? There are several types of ICC. Please specify the ICC you used according to Koo & Li (2016) or a similar source.

Page 18: You did not specify which variables were treated by the logarithmic transformation. Please add this information.

Page 18/19: “The main findings of gender, BMI, medication use, and aerobic capacity sub-groups were similar to those in all subjects. Moreover, the removal of subjects using B-blockers from the analyses did not have an impact on the findings.” Please be more specific or remove these sentences.

Page 19, Table 2: There is a significant difference in systolic blood pressure. However, no statistical procedure is described in the statistical analysis subsection. Is it a two-sample t-test?

Page 19: What does “Agreeability and interchangeability” mean? I suggest to use the term agreement.

Page 19: Why is the average length of the HRV measurement 4.6±0.9 min? I expect a value of 5 min.

Page 19, Table 6: The results of 1.00 (1.00 – 1.00) for ICC (95% CI) look strange because ICC = 1 means a perfect agreement that cannot be reached in practise. In your manuscript, the ICC could not be equal to 1 because Figure 1D shows several discrepancies and the associated LoAs are not zero. Maybe it is due to rounding. Please try to write the ICC value for more decimal digits if the ICC values I close to 1.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Jakub Krejci

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

3 Sep 2020

Reviewers’ comments to Authors:

Reviewer #1:

Comments to the Author:

The Polar V800 sport tester have been verified in a group of healthy subjects and athletes. This manuscript validated the Polar V800 for a group of overweight and obese subjects. I still lack such a study in the literature. The results of this manuscript are important because it is known that overweight and obese subjects have poor heart rate variability (HRV) compared to healthy subjects. Therefore, HRV of overweight and obese subjects should be monitored, preferably at home. Small, lightweight and cheap devices for ECG recording, such as a sport tester, enable the development of telemedicine. In addition, this manuscript compared several methods of artifact detection and correction. In my opinion, this important issue is not sufficiently discussed in the literature. I therefore consider this manuscript to be an interesting and it adds knowledge in the literature. However, there are some issues in the manuscript that should be solved. Please see my comments below.

Thank you for your supportive comment!

Page 5: You have stated that all relevant data are within the manuscript and its Supporting Information files. However, I cannot find any data files from which the results are based.

We thank the reviewer for their input about the missing supporting data files. All supporting information including individual subject characteristics, HRV measures, R-R intervals used to create figures as well as individual HRV measures separated by medication use, gender, BMI, and aerobic capacity (to show the potential influence of medication use, gender, BMI, and VO2peak on the agreement between ECG and Polar V800TM) were added as supporting information with following names:

Supporting Information

S1 Dataset. Individual Subject Characteristics.

S2 Dataset. Individual UN, AC, TBC, and MC R-R intervals.

S3 Dataset. Individual HRV measures calculated from the UN, AC, TBC, and MC R-R intervals.

S1A-S1D Tables. Comparison of HRV measures separated by medication use calculated from UN, AC, TBC, and MC R-R intervals (mean ± SD).

S2A-S2D Tables. Comparison of HRV measures separated by gender calculated from UN, AC, TBC, and MC R-R intervals (mean ± SD).

S3A-S3D Tables. Comparison of HRV measures separated by BMI calculated from UN, AC, TBC, and MC R-R intervals (mean ± SD).

S4A-S4D Tables. Comparison of HRV measures separated by aerobic capacity calculated from UN, AC, TBC, and MC R-R intervals (mean ± SD).

S4 Dataset. Individual HRV measures separated by medication use calculated from UN, AC, TBC, and MC R-R intervals.

S5 Dataset. Individual HRV measures separated by gender calculated from UN, AC, TBC, and MC R-R intervals.

S6 Dataset. Individual HRV measures separated by BMI calculated from UN, AC, TBC, and MC R-R intervals.

S7 Dataset. Individual HRV measures separated by aerobic capacity calculated from UN, AC, TBC, and MC R-R intervals.

Please note that the findings of subgroup analyses of HRV measures separated by medication use, gender, BMI, and aerobic capacity (S1A-S1A Tables; S2A-S2D Tables; S3A-S3D Tables, and S4A-S4D Tables) were added as supporting files to justify our following statement in page 18/19 “The main findings of gender, BMI, medication use, and aerobic capacity sub-groups were similar to those in all subjects. Moreover, the removal of subjects using B-blockers from the analyses did not have an impact on the findings.” Therefore, we also added the individual HRV measures of sub-group analyses as supporting information files (S4 Dataset; S5 Dataset; S6 Dataset; and S7 Dataset).

We have also cited our supplemental information in our manuscript text as “S1 Dataset” under “Subject Characteristics”, “S2 Dataset” under “Agreement of R-R Intervals”, “S3 Dataset” under “Agreement of HRV Measures”, S1A-S1A Tables; S2A-S2D Tables; S3A-S3D Tables, and S4A-S4D Tables” as well as “S4-S7 Datasets” under “Agreement of HRV Measures Separated by Medication Use, Gender, and Aerobic Capacity”

Page 8: I propose to replace “… beyond that of healthy populations.” with “… beyond that of patients”. The use of medical ECG device has usually been limited to the diagnostics of patients with suspected cardiovascular condition.

We acknowledge the confusing statement about the use of heart rate monitors in the first sentence of the abstract. Therefore, we have revised the first sentence of the abstract which originally stated:

“The apparent ease of measuring heart rate variability (HRV) via ambulatory monitors has extended their use beyond that of healthy populations.”

Which now reads as:

“Heart rate variability (HRV) measurements via ambulatory monitors have become common.”

Page 9: You should mention the negative factors that are associated with hypertension – excessive body mass and/or low physical activity.

We have made the following changes to state the risk factors associated with hypertension after the first sentence of introduction:

1) Added (second sentence): “Modifiable (e.g., obesity, physical inactivity, smoking, dyslipidemia) and non-modifiable (e.g., age, gender, family history) risk factors can increase the likelihood of developing hypertension [2].”

2) Changed: “The underlying etiology for the development of hypertension is largely unknown. One of the hypotheses regarding the initiation, progression, and maintenance of hypertension is alterations in the neural control of blood pressure (BP) [3].”

to

“However, the underlying etiology for the development of hypertension is largely unknown. One of the hypotheses regarding the initiation, progression, and maintenance of hypertension is alterations in the neural control of blood pressure (BP) [3].”

Page 9: You should describe what HRV is in a few sentences.

We have made the following changes to describe HRV after the fourth sentence of the first paragraph of introduction:

1) Added (fifth sentence): “The sinoatrial node integrates the inputs from the ANS to adjust heart rate (HR) in response to the constantly changing internal and external environment to maintain homeostasis [4]).”

2) Added (sixth sentence): “This adjustment causes an oscillatory pattern in the HR resulting from beat-to-beat fluctuations in the time period between sequential heartbeats, termed heart rate variability (HRV).”

3) Changed (last sentence of the first paragraph of introduction): “Heart rate variability (HRV) is used to non-invasively measure ANS modulation.”

to

“HRV is used to non-invasively measure ANS modulation.”

Page 9: I propose to replace “… with technical (e.g. missed beats) …” with “… with technical (e.g. excessive noise which causes incorrect detection of R peak) ...”

As per the reviewer’s request, we have made the following changes in third sentence of the second paragraph of introduction in page 9:

1) Deleted: “(e.g., missed beats)”

2) Added: “(e.g., excessive noise which causes incorrect detection of R peak)”

Page 9: I propose to replace “… and physiological (e.g. non-sinus beats) …” with “… and physiological (e.g. ectopic beat) …”

As per the reviewer’s request, we have made the following changes in third sentence of the second paragraph of introduction in page 9:

1) Deleted “(e.g., non-sinus beats)”

2) Added: “(e.g., ectopic beat)”

Page 10: I do not think atherosclerosis is the result of poor ANS activity. This is more due to a change in distribution of different types of cholesterol and hormonal imbalances.

We thank the reviewer for this insightful comment. Though we agree the roles of the changes in the distribution of different types of cholesterol and hormonal balances in the formation of atherosclerosis we also would like to point out different physiological pathways shown in literature to contribute to the development of atherosclerosis. Atherosclerosis is classified as an inflammatory disease, showing that inflammation is the crucial component of this condition (Libby P. History of discovery: inflammation in atherosclerosis. Arterioscler Thromb Vasc Biol2012;32:2045–2051). The literature suggests that increased expression of inflammation does not occur as a standalone condition but also the result of sympathetic activation. In fact, El-Armoche et al. (2009) argue that during the stress response (physical or psychological) the sympathetic system stimulates myocytes through the beta-adrenergic signaling system resulting in the activation of proinflammatory cytokines (e.g., interleukine 18). These authors, therefore, infer that chronic stimulation of the beta-adrenergic system triggered by excess sympathetic activation afflicts detrimental cardiotoxic impact on endothelial function, contributing to the development and progression of atherosclerosis (El-Armouche A , Eschenhagen T. β-Adrenergic stimulation and myocardial function in the failing heart. Heart Fail Rev2009;14:225–241). Moreover, Pizzi et al. (2010) and Ulleryd et al. (2017) contend that ANS dysfunction can be the cause of alterations in endothelial functions in which the increased level of proinflammatory cytokines may be the mediators of the effect of ANS dysfunction on atherosclerosis formation and progression (Pizzi C, Manzoli L, Mancini S, Bedetti G, Fontana F, Costa GM Atherosclerosis. 2010 Sep; 212(1):292-8; Ulleryd MA, Prahl U, Börsbo J, Schmidt C, Nilsson S, Bergström G, Johansson ME).

Consequently, we retain stating the potential link between atherosclerosis and excess sympathetic activation. To make our statement clearer for the readers, however, we have made the following change in the second sentence of the paragraph:

1) Changed (page 10): “Over time, excess sympathetic activation combined with reduced parasympathetic influence may result in left ventricular remodeling and atherosclerosis in patients with hypertension, increasing the risk of microvascular (e.g., vascular dementia) and macrovascular (e.g., myocardial infarction) injuries [12].”

to

“Over time, excess sympathetic activation combined with reduced parasympathetic influence may result in left ventricular remodeling and atherosclerosis due to increased inflammation potentially as a consequence of excess sympathetic activation in patients with hypertension, increasing the risk of microvascular (e.g., vascular dementia) and macrovascular (e.g., myocardial infarction) injuries [12].”

Page 13: “breaths/min–1” is incorrect. Please replace with “breaths⋅min–1” or “breaths/min”.

As per the reviewer’s request, we have made the following change in the last sentence of procedures in page 13:

1) Changed: “breaths/min-1”

to

“breathsꞏmin-1”

Page 14: Is peak oxygen consumption really a potential covariate of sport tester validity? How was it tested in this manuscript? I have found very limited information in the result section.

We thank the reviewer for this important question regarding the potential role of VO2peak as a covariate in agreement of R-R interval detection between Polar V800TM and 12-lead ECG. Our reason to include VO2peak as a covariate derives from the findings of CARDIO-FIT study by Pathak et al. (2015) Impact of CARDIOrespiratory FITness on Arrhythmia Recurrence in Obese Individuals with Atrial Fibrillation: The CARDIO-FIT Study. The authors argue that overweightness and obesity beget arrythmias, which are considered artifacts in the context of HRV pre-processing. Pathak et al. (2015) hypothesized a potential connection between cardiorespiratory fitness and arrythmia occurrence as the positive correlation between cardiorespiratory fitness and cardiac health is well-established in the literature. To test their hypothesis, they divided 308 overweight or obese adults into low, adequate, and high cardiorespiratory fitness groups based on baseline VO2peak and followed them for four years to measure their fitness levels' impact on arrhythmia recurrence. After 4-years of follow up, 84% in high aerobic capacity group no longer had arrhythmias, compared with 76% in the average and 17% in the low aerobic capacity groups. Our reason to include VO2peak as a potential covariate was merely to explore whether cardiorespiratory fitness status can impact the occurrence of artifacts among individuals with hypertension and thus ensure the agreement between Polar V800TM and ECG was not confounded by any external variable.

As far as the statistical testing of this variable as a potential covariate, we formed aerobic capacity subgroups based on fitness quartiles, separating the aerobic capacity group into 1) excellent aerobic fitness, 2) fair to good aerobic fitness, and 3) very poor to poor aerobic fitness according to the 10th edition of the American College of Sports Medicine’s Guidelines for Exercise Testing and Prescription (2018). All statistical analyses that were performed in all subjects (n=25) were also carried out for different subgroups of aerobic capacity. The results of these analyses can be seen as supporting information.

Page 14: The subtitle “R-R Interval Recording” is rather “R-R Interval Processing”. Please consider rephrasing.

As per the reviewer’s request, we have made the following changes in the subtitle in page 14:

1) Deleted: “R-R Interval Recording”

2) Added: “R-R Interval Processing”

Page 14: “BTC” should be replaced by “one of the authors (BTC)”.

As per the reviewer’s request, we have made the following change in under R-R interval Processing subtitle of the methods in page 14:

1) Changed: “BTC”

to

“one of the authors (BTC)”

2) Added: “…checked by another author (DAG) to ensure data accuracy and…..”

Page 14: So, the correct position of the R peak was performed by only one author? Given that R-R intervals obtained from a medical ECG device were used as a gold standard in this manuscript, I consider that a single author checking is not sufficient.

We thank the reviewer for pointing out the missing information concerning the identification of correct R peak position. The first author of our paper (BTC) performed artifact identification as well as corrections first and then his findings were checked and confirmed by a co-author (DAG) who has published two validation papers at rest and during exercise using the Polar V800TM among healthy individuals. Giles D, Draper N, Neil W; Validity of the Polar V800 heart rate monitor to measure RR intervals at rest. Eur J Appl Physiol 2016;116(3):563-571; Giles DA, Draper N. Heart Rate Variability During Exercise: A Comparison of Artefact Correction Methods. J Strength Cond Res 2018 March 01;32(3):726-735. We, therefore, made the following changes in “Artifact Identification (pg 14)” and “Artifact Correction (pg 15)” sections to show that two authors were involved in the processing of HRV data:

1) Added (pg 14): “Artifact identification was performed by the first author (BTC) and each step of this process was checked by a co-author (DAG).”

Page 14: Please describe the MC method in more detail. Who performed it? In which kind of software? Did the person have RR intervals available from the medical ECG device or not?

We thank the reviewer for raising this concern regarding the details of the MC method. As stated previously, all artifact identification and correction processes were performed by the first author (BTC) and then confirmed by a co-author (DAG). Following the identification of artifacts as described under the “Artifact Identification” section of the methods, the artifacts were manually corrected following the guideline provided in Table 1 (Types, description, and correction methods of artifacts). The corrected R-R intervals were then imported into Kubios HRV Premium (ver.3.2) to obtain the HRV measures. All correction methods, including MC were used to correct artifacts present in the R-R intervals derived from the Polar V800TM and not from the ECG. The R-R intervals obtained from ECG were only corrected in the case of physiological artifacts (e.g., ectopic beats), which were detected on ECG strips. All physiological artifacts that were present in three subjects (N=7) were replaced by interpolated R-R intervals from adjacent R-R intervals. We, therefore, made the following changes in page 14 to describe the MC method in more details:

1) Added: “All artifact corrections were performed by the first author (BTC) and each step of this process was checked by a co-author (DAG).”

2) Deleted: “For the MC, non-sinus beats in both signals (N=3) were replaced by interpolated R-R intervals from adjacent R-R intervals.”

3) Changed: “Corrections were made only for T2, T3, T4, T5, and T6b since it is not possible to identify T1 and T6a artifacts without simultaneous ECG recording.”

to

“The MC was used only for T2, T3, T4, T5, and T6b artifacts present in R-R intervals derived from the Polar V800TM following the guidelines provided in Table 1. The T1 and T6a artifacts in the Polar V800TM R-R intervals, however, were not corrected since it is not possible to identify both artifacts without simultaneous ECG recordings.”

4) Added: “The corrected Polar V800TM R-R intervals by the MC were then imported into Kubios HRV Premium (ver.3.2) to calculate HRV measures. All non-sinus beats present in the ECG signal (N=7) were replaced by interpolated R-R intervals from adjacent R-R intervals.”

Page 16: “… in both signals (N = 3)” does it mean three occurrences of artifacts in all group of RR recordings? Or does it mean that recordings from three subjects contains artifacts?

We thank the reviewer for his comment and acknowledge the confusing statement about the number of non-sinus beats. We attempted to report the number of subjects with physiological artifacts, which were detected in the ECG strips of three subjects (thus N=3). However, we decided that reporting the total number of non-sinus beats (N=7) can be more revealing for the readers. Thus, we have made the following changes:

1) Deleted: ““For the MC, non-sinus beats in both signals (N=3) were replaced by interpolated R-R intervals from adjacent R-R intervals.”

2) Added: “The corrected Polar V800TM R-R intervals by the MC were then imported into Kubios HRV Premium (ver.3.2) to calculate HRV measures. All non-sinus beats present in the ECG signal (N=7) were replaced by interpolated R-R intervals from adjacent R-R intervals.”

Page 17: “unpredictable heartbeat dynamics” should be replaced by ”chaotic heartbeat dynamics”. A chaotic system is theorethically predictable, but it is a complex nonlinear system, so in practise the chaotic system looks stochastic (random). Mathematical theory dealing with nonlinear HRV indexes considers ANS to be chaotic, not stochastic. See, for example, Ernst G (2014). Heart rate variability. London: Springer.

We thank the reviewer for this insightful comment about the nature of the heartbeat dynamics. As per the reviewer’s suggestion, we have made the following change:

1) Changed: “Non-linear domain measures represent the unpredictable heartbeat dynamics caused by the complex interactions between several regulatory systems and they included: standard deviation 1 (SD1), SD2, and SD2/SD1 ratio. The non-linear measures were analyzed as a Poincaré plot, a graph in which each N-N interval is plotted against the next N-N interval, making a scatter plot [28]”

to

“Non-linear domain measures represent the chaotic heartbeat dynamics caused by the complex interactions between several regulatory systems.

Page 17: I agree that Poincaré plot is one of the graphical methods used to study chaotic systems. Poincaré plot can therefore be considered as a nonlinear method. However, I cannot agree that SD1, SD2, and SD2/SD1 can be considered as nonlinear indexes of HRV. It was shown that these indexes can be calculated from SDNN and SDSD (Sassi R, Cerutti S, Lombardi F, Malik M, Huikuri H V, Peng C-K, … Yamamoto Y (2015). Advances in heart rate variability signal analysis: Joint position statement by the e-cardiology ESC working group and the european heart rhythm association co-endorsed by the asia pacific heart rhythm society. Europace, 17, 1341–1353). Therefore these indexes should be considered as time domain indexes. Sample entropy is nonlinear HRV index suitable for short-term (≈5 min) RR recordings. Therefore, I suggest removing the indexes SD1, SD2, and SD1/SD2 and adding sample entropy.

We thank the reviewer for providing evidence to show the connection between time domain measures of SDNN and SDSD and SD1, SD2, and SD/SD1. We agree with the potential rationale to remove SD1, SD2, and SD/SD1. Therefore, we have made the following changes:

1) Changed (page 17): “Non-linear domain measures represent the chaotic heartbeat dynamics caused by the complex interactions between several regulatory systems, and they included: standard deviation 1 (SD1), SD2, and SD2/SD1 ratio.”

to

“Non-linear domain measures represent the chaotic heartbeat dynamics caused by the complex interactions between several regulatory systems.”

2) Added (page 17): “We included sample entropy (SampEn) as a measure of non-linear methods. SampEn measures the signal complexity or irregularity where large SampleEn suggests high irregularity and smaller SampleEn a more regular signal [11].”

3) Deleted (page 17): “The non-linear measures were analyzed as a Poincaré plot, a graph in which each N-N interval is plotted against the next N-N interval, making a scatter plot [28]. The analysis consisted of placing an ellipse to the plotted points [28]. The SD of the distance of each plotted N-N interval from the y= x axis (SD1 or ellipse’s width), the SD of each plotted N-N interval from the y = x + average N-N interval (SD2 or ellipse’s length), and SD2/SD1 ratio were then calculated [28].”

4) Changed (page 21): “The UN ES ranged from 0.091 to 0.835, indicating the magnitude of difference between the HRV measures calculated from UN Polar V800TM HR monitor and ECG R-R intervals was trivial to large, and had five out of 11 HRV domain measures between 0.2≤ and <0.5, indicating a moderate difference. Additionally, the UN ICCs ranged from 0.070 to 0.98 suggesting poor to excellent agreement, but had seven out of 11 HRV measures below <0.5, indicating poor agreement. Between the two correction methods of Kubios HRV Premium (ver. 3.2), the TBC resulted in smaller biases than the AC for SDNN, RMSSD, pNN50%, LFnu, HFnu, LF/HF ratio, SD1, SD2, and SD2/SD1 but the AC produced smaller biases than TBC for LFms2,HFms2 and SampEn (Tables 4 and 5). Additionally, when the TBC was compared to the AC method, the TBC produced tighter LoA ranges and smaller ES (<0.031 versus <0.085) for all HRV.”

to

“The UN ES ranged from 0.091 to 0.835, indicating the magnitude of difference between the HRV measures calculated from UN Polar V800TM HR monitor and ECG R-R intervals was trivial to large, and had five out of nine HRV domain measures between 0.2≤ and <0.5, indicating a moderate difference. Additionally, the UN ICCs ranged from 0.070 to 0.98 suggesting poor to excellent agreement, but had six out of nine HRV measures below <0.5, indicating poor agreement. Between the two correction methods of Kubios HRV Premium (ver. 3.2), the TBC resulted in smaller biases than the AC for SDNN, RMSSD, pNN50%, LFnu, HFnu, LF/HF ratio, but the AC produced smaller biases than TBC for LFms2,HFms2 and SampEn (Tables 4 and 5). Additionally, when the TBC was compared to the AC method, the TBC produced tighter LoA ranges and smaller ES (<0.031 versus <0.085) for all HRV domain measures except for SampEn.”

5) Deleted (page 22):

SD1 (ms) 27.1±20.8 65.7±68.2 -38.62 (-158.55 to 81.31) 0.21 (-0.11-0.52) 0.765

SD2 (ms) 72.7±33.6 106.5±62.1 -33.80 (-137.14 to 69.53) 0.37 (-0.01-0.66) 0.677

SD2/SD1 Ratio 3.4±1.4 2.7±1.6 0.66 (-1.45 to 2.76) 0.68 (0.32-0.85) 0.442

UN Table 3

SD1 (ms) 27.1±20.8 25.6±18 1.51 (-18.12 to 21.14) 0.87 (0.73-0.94) 0.078

SD2 (ms) 72.7±33.6 75.6±36.4 -2.97 (-26.11 to 20.17) 0.94 (0.87-0.97) 0.085

SD2/SD1 Ratio 3.4±1.4 3.4±1.3 -0.07 (-1.01 to 0.87) 0.94 (0.86-0.97) 0.055

AC Table 4

SD1 (ms) 27.1±20.8 26.7±19.5 0.38 (-6.78 to 7.53) 0.98 (0.96-0.99) 0.019

SD2 (ms) 72.7±33.6 73.8±35.7 -1.09 (-14.39 to 12.21) 0.98 (0.96-0.99) 0.031

SD2/SD1 Ratio 3.4±1.4 3.3±1.4 0.03 (-0.74 to 0.80) 0.96 (0.91-0.98) 0.021

TBC Table 5

SD1 (ms) 27.1±20.8 27.0±20.6 0.13 (-1.29 to 1.54) 1.00 (1.00-1.00) 0.006

SD2 (ms) 72.7±33.6 72.5±33.5 0.21 (-0.55 to 0.97) 1.00 (1.00-1.00) 0.006

SD2/SD1 Ratio 3.4±1.4 3.4±1.4 0.01 (-0.14 to 0.15) 1.00 (1.00-1.00)0.005

MC Table 6

6) Added:

SampEn 1.5±0.2 1.2±0.5 0.33± (-0.58 to 1.24) 0.39 (-0.18 to 0.71) 0.836

UN Table 3

SampleEn: Sample Entropy

SampEn 1.5±0.2 1.5±0.3 0.02 (-0.21 to 0.21) 0.81 (0.57-0.91) 0.039

AC Table 4

SampEn 1.5±0.2 1.5±0.3 0.02 (-0.35 to 0.40) 0.93 (0.85-0.97) 0.039

TBC Table 5

SampEn 1.5±0.2 1.5±0.2 0.01 (-0.11 to 0.12) 0.97 (0.93-0.99) 0.001

MC Table 6

Page 18: ”inter correlation coeficient” should be replaced by “intraclass correlation coefficient”.

A per the reviews request, we have made the following change:

1) Changed: “The inter class correlation coefficient (ICC) with the 95% confidence interval (CI) assessed the concurrent validity (or interchangeable agreement) of the R-R intervals and HRV measures [33].”

to

“The intraclass correlation coefficient (ICC) with the 95% confidence interval (CI) assessed the concurrent validity (or interchangeable agreement) of the R-R intervals and HRV measures [33].”

Page 18: What type of ICC was used? There are several types of ICC. Please specify the ICC you used according to Koo & Li (2016) or a similar source.

We thank the reviewer for this comment. We followed the ICC guideline published Koo & Li (2016) to choose appropriate type for our study. We specify the type as well as model and definition of ICCs reported in our study as recommended by the authors. Type: average measures (averaged measures of k raters for each subject) that shows agreement across the raters; model: two way-mixed that assumes a random effect of subjects but a fixed effect of rater; and definition: absolute agreement that shows whether two HRV scores from the Polar V800TM and ECG exactly match. Therefore, we have made the following change in page 18 regarding the type, model, and definition of the ICC analysis in our study:

1) Added: “We selected a two-way mixed model with absolute agreement and reported average measures (mean of two measurements).”

Page 18: You did not specify which variables were treated by the logarithmic transformation. Please add this information.

We thank the reviewer for this comment. As per the reviewer’s comment, we have made the following change in page 18:

1) Changed: “In the case of heteroscedasticity in the HRV domain measures, the data were logarithmically transformed before the calculation of LoA ranges.”

to

“The RMSSD, pNN50%, LFms2, and HFms2 were logarithmically transformed before the calculation of LoA ranges.”

Page 18/19: “The main findings of gender, BMI, medication use, and aerobic capacity sub-groups were similar to those in all subjects. Moreover, the removal of subjects using B-blockers from the analyses did not have an impact on the findings.” Please be more specific or remove these sentences.

We thank the reviewer for this comment. We aimed to study the potential impacts of gender, BMI, medication, and aerobic capacity by separating these variables into sub-groups of male and female (gender), overweight and obese (BMI), medication, non-medication (medication use) and very poor to poor, fair to good, and excellent (aerobic capacity). All statistical analyses that were performed in all subjects were also performed for different subgroups of gender, medication use, BMI, and aerobic capacity. The results of our sub-group analyses as well as the individual data, which the results were based on can be seen under “Supporting Information”. Moreover, we have made the following changes in “Statistical Analyses” and “Results” sections to clarify the role of gender, BMI, medication use, and aerobic capacity on the agreement of the Polar V800TM and ECG in producing R-R intervals:

1) Changed (statistical analyses section): “Finally, different sub-groups of gender, BMI, medication use (particularly β-Blockers), and aerobic capacity were also analyzed using the same statistical analyses in the main findings to rule out their potential impacts on the main findings.”

to

“Finally, the sub-group analyses of gender, BMI, medication use, and aerobic capacity were performed using the same statistical methods as in all subjects to study their potential impacts on the agreement between Polar V800TM and ECG. Gender and medication use were grouped into “male and female” and “medication and non-medication”. Further, BMI and aerobic capacity were divided into “overweight (25.0-29.9 kg/m2)” and “obesity (>30 kg/m2)” and “very poor to poor (15.6-37.5 ml/kg/min)”, “fair to good (32.0-41.5 ml/kg/min)”, and “excellent (48.6-51.7 ml/kg/min)”, respectively [25].”

2) Deleted (results): “The main findings of gender, BMI, medication use, and aerobic capacity sub-groups were similar to those in all subjects. Moreover, the removal of subjects using B-blockers from the analyses did not have an impact on the findings.”

3) Added (results):“Agreement of HRV Measures Separated by Medication Use, Gender, BMI, and Aerobic Capacity (title)”

4) Added (results): “The magnitude of biases, LoA ranges, ICCs, and effect sizes of HRV measures of the subjects with medication use S1A-S1D Tables, males S2A-S2D Tables, those with overweight BMI S3A-S3D Tables, and those with very poor-to-poor aerobic capacity S4A-S4D Tables were similar to those of non-medication, female, obesity, and excellent aerobic capacity between the Polar V800TM and ECG. Also, as was in all subjects, the correction of artifacts substantially improved the disagreement seen between UN HRV measures of the Polar V800TM and ECG. Moreover, the accuracy of artifact correction methods in correcting R-R intervals were similar to those reported in all subjects (AC<TBC≤AC) S4-S7 Datasets.”

Page 19, Table 2: There is a significant difference in systolic blood pressure. However, no statistical procedure is described in the statistical analysis subsection. Is it a two-sample t-test?

We thank the reviewer for pointing out the missing statistical information. As per the reviewer’s suggestion we have made the following changes to clarify the statistical method used to investigate the group differences between males and females:

1) Added: “The independent sample t-test was used to report the mean differences between the subject characteristics of men and women.”

Page 19: What does “Agreeability and interchangeability” mean? I suggest to use the term agreement.

As per the request of the reviewer we have made the following change throughout our study:

1) Changed “Agreeability and interchangeability”

to

“Agreement”

Page 19: Why is the average length of the HRV measurement 4.6±0.9 min? I expect a value of 5 min.

We thank the reviewer for his comment. We aimed to obtain 5 min of HRV measurement but unfortunately, the signal from the chest strap to heart rate monitor was interrupted in eight subjects for unknown reasons, which made the total measurement duration less than 5 min. We have made the following changes to clarify the reason for 4.6 min HRV measurement:

1) Added (results): “The length of HRV measurement for eight subjects was less than 5 min due to the loss of connection between the HR monitor and the strap for unknown reasons.”

2) Changed (results): “The average length of the HRV measurement was 4.6±0.9 min and number of R-R intervals was 333.0±70.5.”

to

“Therefore, the average length of the HRV measurement was 4.6±0.9 min and number of R-R intervals was 333.0±70.5.”

Page 19, Table 6: The results of 1.00 (1.00 – 1.00) for ICC (95% CI) look strange because ICC = 1 means a perfect agreement that cannot be reached in practise. In your manuscript, the ICC could not be equal to 1 because Figure 1D shows several discrepancies and the associated LoAs are not zero. Maybe it is due to rounding. Please try to write the ICC value for more decimal digits if the ICC values I close to 1.

We thank the reviewer for his comment. The reason for the discrepancies seen on Figure 1D was because we did not correct T1 and T6a errors as neither error is visible without simultaneous ECG recording. We have made the following change to explain the discrepancies seen on Figure1D:

1) Added (discussion, paragraph 3): “Of note, the Polar V800TM uncorrected T1 and T6a errors can be seen on the corrected Bland–Altman plots (Figure 1D) as outliers, a similar finding reported by Giles et al. [22]. T1 and T6 errors are not visible without simultaneous ECG recording.”

We obtained an exact ICC score of 1.00 (1.00-1.00) for many HRV variables produced especially by MC R-R intervals. We repeated the ICC analyses but still received the same results of 1.00 (1.00-1.00), particularly for MC HRV results. We can explain the reason for the ICCs of 1.00 (1.00-1.00) based on the McGraw et al. (1996) paper. These authors define the intra-class correlation as the ratio of the variance attributable to subjects divided by the total of subject and error variances. They argue that if the ratings by the raters are in perfect agreement, which appears to be the case especially when the artifacts were corrected by MC in our study then there will not be any within-subject variation and thus no error variance. Consequently, that ratio will involve a variance term for subjects divided by a variance term for subjects plus a variance term for error, making the latter exactly 0. Therefore, the ratio will turn out 1.00 (1.00-1.00) based on the aforementioned definition of ICC. McGraw, K. O., & Wong, S. P. (1996). Forming inferences about some intraclass correlation coefficients. Psychological Methods, 1(1), 30–46.

Submitted filename: Response to Reviewers.docx

Click here for additional data file.

23 Sep 2020

Validation of the Polar V800TM Heart Rate Monitor and Comparison of Artifact Correction Methods among Adults with Hypertension

PONE-D-20-14879R1

Dear Dr. Cilhoroz,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Please, address the minor issues raised by the reviewer during the proofreading.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Daniel Boullosa

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: I have read the revised manuscript and the response to reviewers. I appreciate that the authors provided raw data as supplementary material and added sample entropy to the results. I performed my own ICC calculations and got the same results as the authors. It is still debatable that 1.00 (1.00–1.00) is the correct notation. Now I consider this acceptable, because it is, for example, 0.999396 (0.998646 ± 0.999733) rounded to two decimal places.

I am satisfied with the manuscript and authors’ responses. I have no further request.

I have two typographic notes.

1) On page 24, there is “breaths min-1” where there is no dot between ”breaths” and “min”. The dot is probably visible in Word processor but lost in the pdf file.

2) Please unify the style of the units in the whole manuscript. As an example, please compare kg.m-2 in Table 2 and kg/m2 in the text above Table 2.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Jakub Krejčí

29 Sep 2020

PONE-D-20-14879R1

Validation of the Polar V800 Heart Rate Monitor and Comparison of Artifact Correction Methods among Adults with Hypertension

Dear Dr. Cilhoroz:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Daniel Boullosa

Academic Editor

PLOS ONE