Free Access
Issue
Mov Sport Sci/Sci Mot
Number 106, 2019
Page(s) 27 - 35
DOI https://doi.org/10.1051/sm/2019006
Published online 28 May 2019

© ACAPS, 2020

1 Introduction

Analysis of heart rate variability (HRV) is commonly used to non-invasively evaluate heart rate control by the autonomic nervous system (“Heart rate variability. Standards of measurement, physiological interpretation, and clinical use. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology,” 1996). More than 30 years of research have helped us to better understand the physiological and pathophysiological repercussions of HRV indices, and provide guidelines and caveats regarding their use and interpretation (Al Haddad, Laursen, Chollet, Ahmaidi, & Buchheit, 2011; Berntson, et al., 1997).

R-R interval time series (RRts, sometimes referred as heart periods) is the raw material used for HRV analysis. The accuracy of RRts is critical to obtaining reliable HRV results and proper clinical interpretation (Berntson, et al., 1997; Merri, Farden, Mottley, & Titlebaum, 1990). The gold standard to obtain RRts is the electrocardiogram (ECG) (Bailey, et al., 1990; Uijen, de Weerd, & Vendrik, 1979) but technological solutions have been developed to facilitate R-R extractions (Capdevila, Moreno, Movellan, Parrado, & Ramos-Castro, 2012; Lu & Yang, 2009; Moreno, et al., 2015).

One key point in the accuracy of the RRts is the sampling frequency. Accordingly, sampling frequencies varying from 250–500 Hz up to 1000 Hz have been suggested and frequencies lower than 100 Hz should be avoided (Berntson, et al., 1997). However, sampling frequency is only one point that can influence RRts accuracy. Indeed, signal measurement, peak detection algorithms, digitization, data conditioning, or wireless transmission (if any) can also generate errors.

In recent years, several studies have focused on the evaluation of heart rate monitors dedicated to HRV analysis from different manufacturers such as Polar (Cassirame, Tordi, Mourot, Rakobowchuk, & Regnard, 2007; Gamelin, Baquet, Berthoin, & Bosquet, 2008), Suunto (Bouillod, Cassirame, Bousson, Sagawa Jr, & Tordi, 2015) or Hosand (Cassirame, Stuckey, Sheppard, & Tordi, 2013) in the context of sports science or health for investigation or monitoring. New tools based on mobile phones (Flatt & Esco, 2013, 2015; Heathers, 2013), infrared systems (Charlot, Cornolo, Brugniaux, Richalet, & Pichon, 2009; Lu & Yang, 2009) or webcams (Moreno, et al., 2015) have also been proposed. These studies reported differences or similarities as compared with the gold standard ECG method, together with the limits of agreement derived from Bland and Altman analysis (Bland & Altman, 1995). However, they do not inform potential users regarding the minimum accuracy required to obtain a proper periodogram for subsequent HRV analysis, especially when HRV is low (disease, exercise, post-exercise) as this condition is often associated with a reduced accuracy (e.g., Charlot, et al., 2009; Moreno, et al., 2015).

Most of the time, these studies were performed with healthy subjects under conditions of high HRV (supine position, rest).

The aim of the present study was to investigate how the accuracy of the RRts modifies the indices subsequently calculated from HRV analyses. We artificially created inaccuracies in RRts covering a wide range of R-R interval durations and HRV ranges and evaluated how the HRV-derived indices were subsequently affected. This allowed us to test whether the inaccuracies observed in the calculated indexes are dependent of the level of HRV. By doing this, we attempt to provide recommendations (independent of the device used) to guide scientists/clinicians who wish to perform appropriate HRV analyses.

2 Methodology

2.1 Subjects

Twenty-five healthy subjects (mean age ± SD: 35.4 ± 15.6 years; mean height: 174.2 ± 7.6 cm; and mean weight: 72.1 ± 4.7 kg) provided written informed consent to participate in this study, which was approved by the Regional Ethical Review Board and performed in accordance with the Declaration of Helsinki. They were all non-smokers, and none was taking any medication. Participants were requested to refrain from consuming caffeine or alcohol for at least 24 h and to eat a light meal 3 hours (h) prior to testing, as well as to avoid exercise for at least 48 h before the test sessions.

2.2 Experimental design

Seventy-five R-R interval time series of 512 inter-beat length were obtained using a standard ECG (BioAmp and Powerlab system, AD Instruments, Castell Hill, Australia) connected to a computer equipped with LabChart software (version 7) (Cassirame, Vanhaesebrouck, Chevrolat, & Mourot, 2017). The powerlab system recorded ECG signals with a sampling rate of 4000 Hz and applied an online digital filter at 50 Hz to eliminate noise from electronic disturbance. RRts were obtained in the supine and standing positions (25 in each position) during an active tilt-test and 25 were obtained during submaximal exercise. These three conditions were used to obtain a large range of RRts and subsequent HRV indices.

The active tilt-test is used for physiological/pathophysiological purposes (Bahjaoui-Bouhaddi, Henriet, Cappelle, Dumoulin, & Regnard, 1998; Mourot, Bouhaddi, Tordi, Rouillon, & Regnard, 2004). A 15 min supine resting period was respected, then measurements were taken for 8 min in the supine position, followed by 7 min standing (Schmitt, et al., 2013). Exercise was performed on a cycle ergometer (Monark 828E, Monark Exercise AB, Vansbro, Sweden) and consisted of 10 min at a fixed submaximal power output of 175 W. All tests took place in a quiet, dimly lit room (room temperature 22 to 24 °C).

Extraction of the RRts was performed with the “peak analysis” module of the Labchart software. The RRts of 512 inter-beat length were manually selected during a stable period for each condition. Care was taken to ensure that each segment was free of movement artefacts and sharp transients in the signal due to premature beats.

Each original RRts was artificially modified with a home-made software specifically created for this study (see below) to obtain new RRts with inaccuracy levels with limits of agreements of ±2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28 and 30 ms. Thus, HRV analyses were performed on 1200 different RRts (75 original + 75 × 15 degraded) to compare the results from the degraded series with the original results in each condition (supine, standing and exercise). We created new RRts with errors up to 30 ms because such large errors have previously been reported for tools such as oximeter or video based systems.

2.3 Re-editing of R-R intervals time series

To create RRts with poor accuracy, we developed a specific software that randomly generated errors on the original RRts, respecting several conditions to fit with the requirements of the Bland & Altman method (Bland & Altman, 1995), namely:

  • the distribution of errors had to be normal;

  • compared with original RRts, the degraded had to be close to 0 (from −0.05 to 0.05 ms);

  • for each limit of agreement (from ±2 to ±30 ms), the standard deviation of the errors had to be between 1 and 15 ms (with a tolerance of ±0.05 ms). Limits of agreement were calculated as follows: bias ±2 × SD. (Example: for a limit of agreement of ±4 ms; the SD of the errors had to be between 1.95 and 2.05 ms).

Moreover, we added the following specification: an error on a specific R-R interval had to have a direct impact on the next R-R interval. Indeed, if a specific R-peak is detected too early, the R-R will be shorter and the missing time carried over to the next R-R interval. Conversely, if a specific R-Peak is detected too late, extra time is subtracted from the next R-R value

2.4 HRV analysis

HRV analysis was performed on each sample using Kubios HRV analysis software (The Biomedical Signal and Medical Imaging Analysis Group, Department of Applied Physics, University of Kuopio, Finland; see Tarvainen, Niskanen, Lipponen, Ranta-Aho and Karjalainen, 2014 for a detailed description of the methods). The following indices were calculated and reported:

  • RRmean: average R-R (ms);

  • SDNN: standard deviation of sample (ms);

  • rMSSD: root mean square of successive standard deviations (ms);

  • pNN50: percent of successive intervals in which the change exceeds 50 ms;

  • LF: power from low frequency band (ms2) (0.04 to 0.15 hz);

  • HF: power from high frequency band (ms2) (> 0.15 to 0.40 hz);

  • LFnorm: percent of LF divided by (LF + HF);

  • HFnorm: percent of HF divided by (LF + HF);

  • LF/HF: ratio of LF to HF;

  • SD1 and SD2 from Poincaré plot (ms);

  • α1 and α2 from the detrended fractal analysis;

  • ApEn: approximate Entropy.

2.5 Statistical analysis

First, we tested within each condition (supine, standing, exercise) whether the HRV indices calculated from the altered RRts were significantly different from those calculated from the original series. The normality of the distribution of the data was verified with the Shapiro-Wilk test and since the distribution was found to be normal, a one-way analysis was used to compare HRV indices obtained from the 16 different R-R interval time series (one original + 15 reedited). Statistical analyses were performed using Sigmaplot software v12 (SAX Software, Karlsruhe, Germany).

Secondly, HRV is mainly vagally-driven and rMSSD could be used to represent the overall variability (Billman, 2013). It is also known that for a given series, the overall variability could influence the calculated HRV indices, even if the mean R-R duration is unchanged (Cassirame, Chevrolat, Tordi, & Mourot, 2015). Thus, we verified whether HRV indices calculated from the degraded RRts were dependent on rMSSD. For a more practical clinical application of the results, we restricted this verification to 5 levels of bias ranging from ±2 to ±10 ms, i.e. levels commonly found in the literature (Bouillod, et al., 2015; Cassirame, et al., 2013).

For each single HRV index, we plotted the percent of difference between indices calculated from original RRts and reedited RRts against the rMSSD of original RRts (see Fig. 1). The best mathematical model linking rMSSD and the percent of difference was determined by using the best trend procedure proposed by Microsoft Excel 2007 (Microsoft, Redmond, USA). This process allowed us to obtain mathematical relationship between the percentage of errors induced and the rMSSD of original series, for each index.

For the purposes of clarity, we present here only the relationships that were considered large (r > 0.5) based on the scale proposed by Hopkins, Marshall, Batterham and Hanin (2009). Finally, the mathematical models expressing the relationship between rMSSD and the percent of difference of each HRV index were used to calculate the rMSSD values below which the HRV index would be modified by 5, 10 and 15% (Hopkins, et al., 2009). These calculations were processed for each HRV parameter and their modification with limits of agreement of ±2, ±4, ±6, ±8 and ±10 ms.

thumbnail Fig. 1

Relationships between rMSSD of the sample used to calculate HRV indexes and error in the indexes: example for SDNN and SD1.

3 Results

HRV indices during the supine, standing and exercise conditions are presented in Tables 13, respectively, for each degraded RRts. The results showed that HRV indices calculated from measurement in the supine position were only slightly affected by poor accuracy of RRts, compared to the standing and exercise conditions. Indeed, in the supine position, significant differences in HRV indices compared to those derived from original RRts were observed only for bias larger than ±20 ms (ApEn, 9% or 0.1 n.u). During the standing condition, significant alterations of the indices were observed for limits of agreement of ±12 ms (see Tab. 2; for example ApEn: 15% or 0.14 n.u), while alterations appeared from ±4 ms during exercise (see Tab. 3; rMSSD: 16% or 1 ms; SD1: 15% or 0.6 ms).

Significant relationships were observed between rMSSD and the percent difference between indices calculated from original RRts and reedited RRts. These relationships indicated that a low rMSSD (i.e. RRts displaying low variability) was associated with large modifications of the HRV index calculated from the RRts. Representative examples for SDNN and SD1 are displayed in Figure 1. In these examples (the results for the other indices are not shown for the purpose of clarity), the curves are displayed in the same position, i.e. from low to high rMSSD: ±4 ms below, then ±6, ±8 and finally ±10 ms over. The results indicated that: 1) the larger the error’s magnitude, the larger the relative error in calculated HRV indexes, and most importantly; 2) the lower the rMSSD of the original RRts, the larger the relative error in the HRV indices. Based on these mathematical relationships, rMSSD values inducing 5, 10 and 15% modifications in a given HRV indexes were calculated and are presented in Table 4.

Table 1

Presentation of HRV analysis results in the supine position for each imprecision range.

Table 2

Presentation of HRV analysis results in the standing position for each imprecision range.

Table 3

Presentation of HRV analysis results in the exercise situation for each imprecision range.

Table 4

Presentation of rMSSD threshold by imprecision level for each HRV analysis index.

4 Discussion

The results of the present study indicate that: 1) inaccuracies in the RRts (as low as 4 ms) could lead to significant differences in the subsequently calculated HRV indices; and 2) the inaccuracy is dependent on the rMSSD level of the RRts used to obtain calculate the indices. Thus, using an accurate device to obtain RRts is of particular importance, especially when rMSSD is low as during exercise or acute stroke

Numerous studies using HRV analyses have been published during the last decades in many physiological/pathological conditions including exercise. The RRts obtained in these conditions could be inaccurate, depending on the devices used to obtain the RRts and/or technical disturbances of technical origins (e.g., movement of the subject that create noise and limits the ability to find the R peak with accuracy). This could introduce a bias into the subsequent HRV results, and finally represents a significant problem in their interpretation (Berntson, et al., 1997; “Heart rate variability. Standards of measurement, physiological interpretation, and clinical use. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology”, 1996). While previous studies focused on different specific settings to measure RRts such as sampling frequency (Berntson, et al., 1997), or how to correct artefacts once the recordings have been performed (Rincon Soler, Silva, Fazan, & Murta, 2018), to the best of our knowledge, this study is the first to investigate how the accuracy of the RRts modifies the indices subsequently calculated from HRV analyses. It is also the first to highlight a link between the variability of the original RRts (expressed as rMSSD) and the accuracy of the calculated indices.

Our results indicate that the level of inaccuracy of the RRts at which the HRV indices are significantly altered is dependent on the mean value of the RRts, which in turn is associated with the condition of the RRts recording (see Tabs. 13). Indeed, it can be observed that during supine resting, the HRV indices were significantly affected only when the limits of agreement were larger than ±20 ms, i.e. limits of agreement that have previously been reported, but are quite rare in the literature (Moreno, et al., 2015; Parak, et al., 2015). For example, the limits of agreements were ±5 ms in Kingsley, Lewis and Marson (2005), ±12 ms in Weippert et al. (2010) and ±2 ms in Bouillod et al. (2015). With limits of agreement of ±20 ms, we noted that technological measurements were based on video or optical sensor measurements with a low sampling rate (> 200 Hz), which does not guarantee accurate R peak detection and R-R interval measurement (Merri, et al., 1990). During standing, a significant alteration of the HRV indices was observed for limits of agreement of ±12 ms, and this fell to ±4 ms during exercise. Several studies have previously reported limits of agreement close to ±4 ms during exercise with devices used in the field of sport sciences (Cassirame, et al., 2007; Kingsley, et al., 2005). This means that in this condition, specific attention should be paid to the level of accuracy of the device, since significantly altered indices may subsequently be calculated, rendering proper interpretation of the autonomic nervous system activity virtually impossible.

The differences between the three conditions related mainly to a reduction in the mean RR and, most importantly, in the variability of the RRts depending on vagal activity. Overall, this clearly indicates that when a RRts has low rMSSD, which is usually the case when an evaluation of the ANS activity through HRV is performed, the accuracy of the device should be high in order to avoid a significant alteration of the subsequently calculated HRV indices, and consequently, misinterpretation in the subject’s evaluation.

To help user to evaluate the potential imprecision level of HRV indices, we calculated threshold values based on the limits of agreement (Tab. 4). For example, if a device has limits of agreement of ±6 ms, an error of 5% can be expected in the SDNN index if rMSSD is lower than 14.3 ms. Under the same circumstances, an error of 15% can be expected for the SDNN index for rMSSD lower than 9.6 ms. Another way to use Table 4 is to choose the appropriate device depending on the anticipated rMSSD value. For example, if a user wants to monitor training adaptations in healthy subjects under supine resting conditions (i.e., expected rMSSD values higher than 60 ms: Plews, Laursen, Kilding, & Buchheit, 2012), then the device could have limits of agreements as high as ±10 ms since no rMSSD value in Table 4 is higher than 60 ms. However, if the same user wants to monitor training adaptation using e.g. post-exercise situation (rMSSD values lower than 30 ms (Buchheit, et al., 2008), it is required to use a device with low limits of agreement. For example an error of 10% in the HF indexes will be expected if the device has limits of agreement of ±6 ms (rMSSD value threshold of 30.7 ms).

Taking the expected rMSSD into account in choosing the appropriate device is of particular importance, since several factors and situations can reduce rMSSD. Indeed, values of rMSSD as low as 18 ms have been reported in patients suffering from diabetes (Kudat, et al., 2006), myocardial infarction (Brateanu, 2015), or obesity (Zahorska-Markiewicz, Kuagowska, Kucio, & Klin, 1993). Some tests used for clinical examinations and diagnosis also reduce HRV, such as the orthostatic test used to assess autonomic function for example (Mourot, Bouhaddi, Perrey, Cappelle, et al., 2004), during which rMSSD values can be reduced by up to 65% when tilted (Mourot, Bouhaddi, Perrey, Cappelle, et al., 2004). During submaximal exercise in apparently healthy subjects values as low as 6 ms have been reported (Mourot, Bouhaddi, Perrey, Rouillon, & Regnard, 2004). HRV analysis during exercise is also performed to determine ventilatory thresholds (Cassirame, et al., 2014; Mourot, et al., 2011).

Initially, RRts were obtained from ECG recordings performed in laboratories or clinical environments and only ECG Holter monitoring was performed in an ambulatory setting. Thanks to technological improvements, RRts have more recently been obtained from heart rate monitors based on the measurement of the electrical activity of the heart (Achten & Jeukendrup, 2003). Today, other approaches based on photoplethysmography (Charlot, et al., 2009; Moreno, et al., 2015; Schäfer & Vagedes, 2013), oximetry (Johnston & Mendelson, 2005; Lu & Yang, 2009) or video (Capdevila, et al., 2012; Tarassenko, et al., 2014) are proposed to obtain RRts, allowing large access to HRV analysis. Furthermore, mobile phone solutions through chest belts or with optical measurements (Flatt & Esco, 2013, 2015; Heathers, 2013; Peng, Zhou, Lin, & Zhang, 2015; Scully, et al., 2012) are very promising in helping medical staff to monitor patients using telemedicine.

Nevertheless, the information provided by most systems on offer showed insufficient RRts accuracy (Capdevila, et al., 2012; Peng, et al., 2015; Tarassenko, et al., 2014) to perform valid HRV analysis, with limits of agreement over ±20 ms. In addition, it should be underlined that many validation studies were performed in the resting and supine conditions with healthy subjects (Parak, et al., 2015; Scully, et al., 2012; Vasconcellos, et al., 2015), i.e. in conditions of high rMSSD. The results of the present study showed that this context is clearly favourable to the evaluation of the device. However, for example, although the accuracy of optical systems has been strongly upgraded in the last decade, making it possible to obtain mean heart rate values similar to those of an ECG (Arberet, et al., 2013), studies have nonetheless shown that RRts are still not suitable for HRV analysis (Chuang, Ye, Lin, Lee, & Tai, 2015; Selvaraj, Jaryal, Santhosh, Deepak, & Anand, 2008).

Several parameters could help to improve and obtain accurate RRts. Given that RRts is a temporal measurement, the sampling rate of the device is important. Basic recommendations from the European Society of Cardiology and the North American Society of Pacing and Electrophysiology (“Heart rate variability. Standards of measurement, physiological interpretation, and clinical use. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology”, 1996) propose a sampling rate of at least 500 Hz. Furthermore, several authors have advocated a sampling rate of at least 1000 Hz in order to not bias the HRV analysis (Ellis, Zhu, Koenig, Thayer, & Wang, 2015; Hejjel & Roth, 2004; Singh, Manjit, & Vijay Kumar, 2014). Technically, video and plethysmographic system use sampling rates from 50 to 256 Hz (Chuang, et al., 2015; Moreno, et al., 2015; Parak, et al., 2015), i.e. far below the suggested minimum of 500 Hz. Several studies also stress the importance of using appropriate algorithms to detect R-peak (Ellis, et al., 2015; Peng, et al., 2015) and appropriate correction methods thereafter (Rincon Soler, et al., 2018).

5 Conclusion

HRV analysis is a non-invasive method to assess the activity of the autonomic nervous system. HRV provides interesting information for both clinical applications and research purposes, in patients as well as in healthy subjects. Recent technological improvements have enable new tools that are easier to use, more affordable, permitting HRV analysis without any background or technical skills. Nevertheless, the accuracy of the device is a crucial point, especially when analysis is performed on RRts with low variability. Overall, we demonstrate that HRV analysis performed on samples with low variability (rMSSD < 30 ms) is largely affected by the level of accuracy. Thus, users have to consider both the population and the situation when choosing the appropriate measurement device. If no information can be found about the accuracy of the system to be used, we highly recommend first checking the accuracy and determining the limits of agreement.

Acknowledgements

The authors express their gratitude to the participants, for their collaboration in the study. We also thank Fiona Ecarnot (EA3920, University Hospital Besancon, France) for editorial assistance.

References

  • Achten, J., & Jeukendrup, A.E. (2003). Heart rate monitoring: Applications and limitations. Sports Medicine (Auckland, N.Z.) , 33(7), 517–538. [CrossRef] [PubMed] [Google Scholar]
  • Al Haddad, H., Laursen, P.B., Chollet, D., Ahmaidi, S., & Buchheit, M. (2011). Reliability of resting and postexercise heart rate measures. International Journal of Sports Medicine , 32(8), 598–605. [CrossRef] [PubMed] [Google Scholar]
  • Arberet, S., Lemay, M., Renevey, P., Solà, J., Grossenbacher, O., Andries, D., Sartori, D., & Bertschi, M. (2013). Photoplethysmography-based ambulatory heartbeat monitoring embedded into a dedicated bracelet. In Computing in Cardiology Conference (CinC), 2013 (pp. 935–938). [Google Scholar]
  • Bahjaoui-Bouhaddi, M., Henriet, M.T., Cappelle, S., Dumoulin, G., & Regnard, J. (1998). Active standing and passive tilting similarly reduce the slope of spontaneous baroreflex in healthy subjects. Physiological Research , 47(4), 227–235. [PubMed] [Google Scholar]
  • Bailey, J.J., Berson, A.S., Garson, A., Jr, Horan, L.G., Macfarlane, P.W., Mortara, D.W., & Zywietz, C. (1990). Recommendations for standardization and specifications in automated electrocardiography: Bandwidth and digital signal processing. A report for health professionals by an ad hoc writing group of the Committee on Electrocardiography and Cardiac Electrophysiology of the Council on Clinical Cardiology, American Heart Association. Circulation , 81(2), 730–739. [CrossRef] [PubMed] [Google Scholar]
  • Berntson, G.G., Bigger, J.T., Jr, Eckberg, D.L., Grossman, P., Kaufmann, P.G., Malik, M., Nagaraja, H.N., Porges, S.W., Saul, J.P., Stone, P.H., & van der Molen, M.W. (1997). Heart rate variability: Origins, methods, and interpretive caveats. Psychophysiology , 34(6), 623–648. [CrossRef] [PubMed] [Google Scholar]
  • Billman, G.E. (2013). The LF/HF ratio does not accurately measure cardiac sympatho-vagal balance. Frontiers in Physiology , 4. doi: 10.3389/fphys.2013.00026. [Google Scholar]
  • Bland, J.M., & Altman, D.G. (1995). Comparing methods of measurement: Why plotting difference against standard method is misleading. Lancet , 346(8982), 1085–1087. [CrossRef] [PubMed] [Google Scholar]
  • Bouillod, A., Cassirame, J., Bousson, J.M., Sagawa Jr, Y., & Tordi, N. (2015). Acurácia do sistema Suunto para a análise da variabilidade da frequência cardíaca durante um teste de inclinação. Revista Brasileira de Cineantropometria e Desempenho Humano , 17(4), 409. doi: 10.5007/1980-0037.2015v17n4p409. [CrossRef] [Google Scholar]
  • Brateanu, A. (2015). Heart rate variability after myocardial infarction: What we know and what we still need to find out. Current Medical Research and Opinion , 31(10), 1855–1860. [CrossRef] [PubMed] [Google Scholar]
  • Buchheit, M., Millet, G.P., Parisy, A., Pourchez, S., Laursen, P.B., & Ahmaidi, S. (2008). Supramaximal training and postexercise parasympathetic reactivation in adolescents. Medicine and Science in Sports and Exercise , 40(2), 362–371. [CrossRef] [PubMed] [Google Scholar]
  • Capdevila, L., Moreno, J., Movellan, J., Parrado, E., & Ramos-Castro, J. (2012). HRV based health&sport markers using video from the face. Conference Proceedings: …Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual Conference, 2012 (pp. 5646–5649). [Google Scholar]
  • Cassirame, J., Tordi, N., Mourot, L., Rakobowchuk, M., & Regnard, J. (2007). L’utilisation d’un nouveau système d’enregistrement de fréquence cardiaque battement à battement pour l’analyse traditionnelle de variabilité de fréquence cardiaque. Science & Sports , 22(5), 238–242. [Google Scholar]
  • Cassirame, J., Stuckey, M.I., Sheppard, F., & Tordi, N. (2013). Accuracy of the Minicardio system for heart rate variability analysis compared to ECG. The Journal of Sports Medicine and Physical Fitness , 53(3), 248–254. [PubMed] [Google Scholar]
  • Cassirame, J., Tordi, N., Fabre, N., Duc, S., Durand, F., & Mourot, L. (2014). Heart rate variability to assess ventilatory threshold in ski-mountaineering. European Journal of Sport Science , 15(7), 615–622. [CrossRef] [PubMed] [Google Scholar]
  • Cassirame, J., Chevrolat, S., Tordi, N., & Mourot, L. (2015). Précision du périodogramme, quelles conséquences pour l’analyse de la variabilité de la fréquence cardiaque ? In 16e congrès de l’ACAPS. [Google Scholar]
  • Cassirame, J., Vanhaesebrouck, R., Chevrolat, S., & Mourot, L. (2017). Accuracy of the Garmin 920 XT HRM to perform HRV analysis. Australasian Physical & Engineering Sciences in Medicine , 40(4), 831–839. [CrossRef] [PubMed] [Google Scholar]
  • Charlot, K., Cornolo, J., Brugniaux, J.V., Richalet, J.P., & Pichon, A. (2009). Interchangeability between heart rate and photoplethysmography variabilities during sympathetic stimulations. Physiological Measurement , 30(12), 1357–1369. [CrossRef] [PubMed] [Google Scholar]
  • Chuang, C.-C., Ye, J.-J., Lin, W.-C., Lee, K.-T., & Tai, Y.-T. (2015). Photoplethysmography variability as an alternative approach to obtain heart rate variability information in chronic pain patient. Journal of Clinical Monitoring and Computing , 29(6), 801–806. [CrossRef] [PubMed] [Google Scholar]
  • Ellis, R.J., Zhu, B., Koenig, J., Thayer, J.F., & Wang, Y. (2015). A careful look at ECG sampling frequency and R-peak interpolation on short-term measures of heart rate variability. Physiological Measurement , 36(9), 1827–1852. [CrossRef] [PubMed] [Google Scholar]
  • Flatt, A.A., & Esco, M.R. (2013). Validity of the ithleteTM smart phone application for determining ultra-short-term heart rate variability. Journal of Human Kinetics , 39, 85–92. [CrossRef] [PubMed] [Google Scholar]
  • Flatt, A.A., & Esco, M.R. (2015). Smartphone-derived heart rate variability and training load in a female soccer team. International Journal of Sports Physiology and Performance , 10(8), 994–1000. [CrossRef] [PubMed] [Google Scholar]
  • Gamelin, F.-X., Baquet, G., Berthoin, S., & Bosquet, L. (2008). Validity of the polar S810 to measure R-R intervals in children. International Journal of Sports Medicine , 29(2), 134–138. [CrossRef] [PubMed] [Google Scholar]
  • Heart rate variability. Standards of measurement, physiological interpretation, and clinical use. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. (1996). European Heart Journal, 17(3), 354–381. [Google Scholar]
  • Heathers, J.A.J. (2013). Smartphone-enabled pulse rate variability: An alternative methodology for the collection of heart rate variability in psychophysiological research. International Journal of Psychophysiology: Official Journal of the International Organization of Psychophysiology , 89(3), 297–304. [CrossRef] [Google Scholar]
  • Hejjel, L., & Roth, E. (2004). What is the adequate sampling interval of the ECG signal for heart rate variability analysis in the time domain? Physiological Measurement , 25(6), 1405–1411. [CrossRef] [PubMed] [Google Scholar]
  • Hopkins, W.G., Marshall, S.W., Batterham, A.M., & Hanin, J. (2009). Progressive statistics for studies in sports medicine and exercise science. Medicine and Science in Sports and Exercise , 41(1), 3–13. [CrossRef] [PubMed] [Google Scholar]
  • Johnston, W., & Mendelson, Y. (2005). Extracting heart rate variability from a wearable reflectance pulse oximeter (pp. 157–158). IEEE. doi: 10.1109/NEBC.2005.1431971. [Google Scholar]
  • Kingsley, M., Lewis, M.J., & Marson, R.E. (2005). Comparison of Polar 810s and an ambulatory ECG system for RR interval measurement during progressive exercise. International Journal of Sports Medicine , 26(1), 39–44. [CrossRef] [PubMed] [Google Scholar]
  • Kudat, H., Akkaya, V., Sozen, A.B., Salman, S., Demirel, S., Ozcan, M., Atilgan, D., Yilmaz, M.T., & Guven, O. (2006). Heart rate variability in diabetes patients. The Journal of International Medical Research , 34(3), 291–296. [CrossRef] [PubMed] [Google Scholar]
  • Lu, G., & Yang, F. (2009). Limitations of oximetry to measure heart rate variability measures. Cardiovascular Engineering (Dordrecht, Netherlands) , 9(3), 119–125. [PubMed] [Google Scholar]
  • Merri, M., Farden, D.C., Mottley, J.G., & Titlebaum, E.L. (1990). Sampling frequency of the electrocardiogram for spectral analysis of the heart rate variability. IEEE Transactions on Bio-Medical Engineering , 37(1), 99–106. [CrossRef] [PubMed] [Google Scholar]
  • Moreno, J., Ramos-Castro, J., Movellan, J., Parrado, E., Rodas, G., & Capdevila, L. (2015). Facial Video-Based Photoplethysmography to Detect HRV at Rest. International Journal of Sports Medicine , 36(6), 474–480. [CrossRef] [PubMed] [Google Scholar]
  • Mourot, L., Bouhaddi, M., Perrey, S., Cappelle, S., Henriet, M.-T., Wolf, J.-P., Rouillon, J.D., & Regnard, J. (2004). Decrease in heart rate variability with overtraining: Assessment by the Poincaré plot analysis. Clinical Physiology and Functional Imaging , 24(1), 10–18. [CrossRef] [PubMed] [Google Scholar]
  • Mourot, L., Bouhaddi, M., Perrey, S., Rouillon, J.-D., & Regnard, J. (2004). Quantitative Poincare plot analysis of heart rate variability: Effect of endurance training. European Journal of Applied Physiology , 91(1), 79–87. [CrossRef] [PubMed] [Google Scholar]
  • Mourot, L., Bouhaddi, M., Tordi, N., Rouillon, J.-D., & Regnard, J. (2004). Short- and long-term effects of a single bout of exercise on heart rate variability: Comparison between constant and interval training exercises. European Journal of Applied Physiology , 92(4–5), 508–517. [CrossRef] [PubMed] [Google Scholar]
  • Mourot, L, Tordi, N., Bouhaddi, M., Teffaha, D., Monpere, C., & Regnard, J. (2011). Heart rate variability to assess ventilatory thresholds: Reliable in cardiac disease? European Journal of Cardiovascular Prevention and Rehabilitation: Official Journal of the European Society of Cardiology, Working Groups on Epidemiology & Prevention and Cardiac Rehabilitation and Exercise Physiology , 19(6), 1272–1280. [Google Scholar]
  • Parak, J., Tarniceriu, A., Renevey, P., Bertschi, M., Delgado-Gonzalo, R., & Korhonen, I. (2015). Evaluation of the beat-to-beat detection accuracy of PulseOn wearable optical heart rate monitor. Conference Proceedings: …Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual Conference, 2015 (pp. 8099–8102). [Google Scholar]
  • Peng, R.-C., Zhou, X.-L., Lin, W.-H., & Zhang, Y.-T. (2015). Extraction of Heart Rate Variability from Smartphone Photoplethysmograms. Computational and Mathematical Methods in Medicine , 2015. doi: 10.1155/2015/516826. [Google Scholar]
  • Plews, D.J., Laursen, P.B., Kilding, A.E., & Buchheit, M. (2012). Heart rate variability in elite triathletes, is variation in variability the key to effective training? A case comparison. European Journal of Applied Physiology , 112(11), 3729–3741. [CrossRef] [PubMed] [Google Scholar]
  • Rincon Soler, A.I., Silva, L.E.V., Fazan, R., & Murta, L.O. (2018). The impact of artifact correction methods of RR series on heart rate variability parameters. Journal of Applied Physiology (Bethesda, Md.: 1985) , 124(3), 646–652. [CrossRef] [Google Scholar]
  • Schäfer, A., & Vagedes, J. (2013). How accurate is pulse rate variability as an estimate of heart rate variability? A review on studies comparing photoplethysmographic technology with an electrocardiogram. International Journal of Cardiology , 166(1), 15–29. [CrossRef] [PubMed] [Google Scholar]
  • Schmitt, L., Regnard, J., Desmarets, M., Mauny, F., Mourot, L., Fouillot, J.-P., Coulmy, N., & Millet, G. (2013). Fatigue shifts and scatters heart rate variability in elite endurance athletes. PloS One , 8(8), e71588. [CrossRef] [PubMed] [Google Scholar]
  • Scully, C.G., Lee, J., Meyer, J., Gorbach, A.M., Granquist-Fraser, D., Mendelson, Y., & Chon, K.H. (2012). Physiological parameter monitoring from optical recordings with a mobile phone. IEEE Transactions on Bio-Medical Engineering , 59(2), 303–306. [Google Scholar]
  • Selvaraj, N., Jaryal, A., Santhosh, J., Deepak, K.K., & Anand, S. (2008). Assessment of heart rate variability derived from finger-tip photoplethysmography as compared to electrocardiography. Journal of Medical Engineering & Technology , 32(6), 479–484. [CrossRef] [PubMed] [Google Scholar]
  • Singh, B., Manjit, S., & Vijay Kumar, B. (2014). Sample entropy based HRV: Effect of ECG sampling frequency. Biomedical Science and Engineering , 2–4, 68–72. [Google Scholar]
  • Tarassenko, L., Villarroel, M., Guazzi, A., Jorge, J., Clifton, D.A., & Pugh, C. (2014). Non-contact video-based vital sign monitoring using ambient light and auto-regressive models. Physiological Measurement , 35(5), 807. doi: 10.1088/0967-3334/35/5/807. [CrossRef] [PubMed] [Google Scholar]
  • Tarvainen, M.P., Niskanen, J.-P., Lipponen, J.A., Ranta-Aho, P.O., & Karjalainen, P.A. (2014). Kubios HRV – heart rate variability analysis software. Computer Methods and Programs in Biomedicine , 113(1), 210–220. [CrossRef] [PubMed] [Google Scholar]
  • Uijen, G.J., de Weerd, J.P., & Vendrik, A.J. (1979). Accuracy of QRS detection in relation to the analysis of high-frequency components in the electrocardiogram. Medical & Biological Engineering & Computing , 17(4), 492–502. [CrossRef] [PubMed] [Google Scholar]
  • Vasconcellos, F.V.A., Seabra, A., Cunha, F.A., Montenegro, R.A., Bouskela, E., & Farinatti, P. (2015). Heart rate variability assessment with fingertip photoplethysmography and polar RS800cx as compared with electrocardiography in obese adolescents. Blood Pressure Monitoring , 20(6), 351–360. [CrossRef] [PubMed] [Google Scholar]
  • Weippert, M., Kumar, M., Kreuzfeld, S., Arndt, D., Rieger, A., & Stoll, R. (2010). Comparison of three mobile devices for measuring R-R intervals and heart rate variability: Polar S810i, Suunto t6 and an ambulatory ECG system. European Journal of Applied Physiology , 109(4), 779–786. [CrossRef] [PubMed] [Google Scholar]
  • Zahorska-Markiewicz, B., Kuagowska, E., Kucio, C., & Klin, M. (1993). Heart rate variability in obesity. International Journal of Obesity and Related Metabolic Disorders: Journal of the International Association for the Study of Obesity , 17(1), 21–23. [Google Scholar]

Cite this article as: Cassirame J, Chevrolat S, & Mourot L (2019) Effects of R-R time series accuracy on heart rate variability indexes. Mov Sport Sci/Sci Mot, 106, 27–35

All Tables

Table 1

Presentation of HRV analysis results in the supine position for each imprecision range.

Table 2

Presentation of HRV analysis results in the standing position for each imprecision range.

Table 3

Presentation of HRV analysis results in the exercise situation for each imprecision range.

Table 4

Presentation of rMSSD threshold by imprecision level for each HRV analysis index.

All Figures

thumbnail Fig. 1

Relationships between rMSSD of the sample used to calculate HRV indexes and error in the indexes: example for SDNN and SD1.

In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.