Heart rate variability (HRV) is gaining popularity in biometric wearables as a means to create more in-depth insights including measurements of stress, sleep, training status, and much more. In general, decreases in heart rate variability indicates some kind of negative general stress.

Valencell is continually asked by our customers how heart rate variability is measured and what can be done with the resulting measurements. Valencell believes any biometric indicators should be supported by scientific validation of those measurements to ensure they are providing real insights that users can count on. So to help answer these questions, we conducted a comprehensive scientific literature review around heart rate variability to understand what use cases and applications have been proven with scientific rigor. The results of that review are below.

Heart Rate Variability

Assessment of the beat-to-beat variability or heart rate variability (HRV) can be an indicator of “stress” (physical and/or mental) on the body.  The amount of variation in time (milliseconds) of the beat-to-beat interval is controlled by the nervous system with less variation under greater control of the sympathetic nervous system (the sympathetic nervous system is part of the central nervous system and is more active during times of stress) and more variation under greater control of the parasympathetic nervous system (the parasympathetic nervous system is part of the central nervous system and is more active during times of relaxation) (Plews et al., 2012).  The beat-to-beat interval can be seen in Figure 1 as either an R-R interval (RRi) from an ECG or P-P interval (PPi) from a pulse wave measured by PPG.  HRV is the statistical analysis of the RRi/PPi and can yield insight on one’s reaction to external or internal physical or emotional stimuli (Castaldo et al.).

RRi and PPi

Figure 1. ECG and PPG waveforms, with RRi comparison (Schäfer et al.)

Stress is a state of physical, mental, and emotional strain on the body. One should not think of stress as a single event or experience but rather an accumulation of all circumstances occurring over time. Stressors from all sources have an additive effect on the body and will eventually lead to decreased mental and/or physical performance if not properly balanced with relaxation and/or recovery. These stressors include, but are not limited to, changes in the following: physical workload, sleep, diet, disease states, daily career and life tasks and interactions.  The impact of the stressors on the body is monitored by HRV and related to fatigue (mental and work capacity) and readiness to perform mental and work tasks. However, HRV cannot identify the source of stress on the body and quantify, for example, the amount of stress caused by mental stress vs. the amount of stress caused by physical stress.

The following table outlines select use cases for heart rate variability supported by scientific literature:

HRV use cases

Data Collection and Handling for RRi to HRV

The individual baseline of RRi/PPi and related HRV metrics is not the same for all individuals and it cannot be assumed that data between individuals will show the same physiological response to similar stressors. Because of high day-to-day variability in environmental and homeostatic factors, results have been varied with regards to the influence of stress on HRV (Halson).  To overcome this variability, it is helpful to analyze individual baseline HRV metrics. To help resolve challenges with the minute-to-minute and day-to-day variation, Halson recommends utilizing weekly and/or 7-day rolling averages, techniques which would have higher validity and statistical value than daily recordings of HRV. Plews has recommended utilizing a change in baseline measures of greater or less than 0.5 of an individual’s Coefficient of Variance (CV) as a worthwhile change (Plews et al., 2012).

Regardless of the techniques used to monitor or determine meaningful changes in HRV, RRi/PPi data needs to be carefully analyzed in preparation for calculations.  Figure 2 outlines proposed steps in this process (Task force, 1996).

HRV analysis steps

Figure 2. Flow chart summarizing individual steps used in order to obtain data for HRV analysis (Task force, 1996)

Time Domain Analysis of HRV from RRi (PPi)

It has been shown that mathematical manipulation of RRi can provide information about different stress states, whether they are positive or negative changes from an individual’s baseline.  One commonly used method is the log-transformed square root of the mean sum of the squared differences between R-R intervals (ln rMSSD).  The benefits of using ln rMSSD include its resistance to breathing frequency influence, its ability to capture levels of parasympathetic activity over a short time frame, and its relative ease to calculate with R-R data (Halson).

Frequency Domain Analysis of HRV from RRi (PPi)

There are also frequency domain methods of analysis, which utilize transforms to get the interval data into the frequency domain.  Then, one can interpret the data as it relates to low vs. high frequencies, with low frequencies (LF), 0.04-0.15 Hz, reflecting both vagal and sympathetic activities and high frequencies (HF), 0.15-0.4 Hz, reflecting variations in parasympathetic activity (Earnest et al.).  Figure 3 showcases the difference frequency bands in the spectrum data from HRV.

FFT spectrum

Figure 3. FFT Spectrum data of very low frequency, low frequency, high frequency, and total frequency data.

Time Domain or Frequency Domain?

In terms of whether to choose frequency domain analysis over time domain analysis, it is ideal to execute analyses for both, since the RRi data will include the requisite information for both domains.  However, time domain analysis can be useful when trying to display changes in trend as they relate to time.  This can show how HRV changes when subjects themselves experience different things in the time domain, and display how relevant those changes were graphically.  It also has been shown that the frequency domain might change based on breathing frequency, while time domain analysis isn’t as susceptible to frequency changes in the subject (Penttilä et al.).

Nonlinear Analysis

Poincaré plots are a way to view HRV in a non-linear manner, as interactions between the parasympathetic and sympathetic nervous systems may be non-linear. Thus, Poincaré plots are potentially more indicative of underlying physiological occurrences than frequency analysis.

The points on the Poincaré plots represent durations between an R-R interval and its succeeding R-R interval.  The line with slope 1 (45-degree angle) represents equivalent intervals; therefore, points above the line are longer heart periods, while points between the line represent shorter periods (Quintana et al.).  Different shapes and scattering of the plots are meant to represent different states of health for the patient.  In general, healthy participants have comet shaped plots that disperse with longer beats.  Patients with heart failure have torpedo, fan, or complex shaped patterns (Quintana et al.).

In general, the more spread out these Poincaré plots are, the better their HRV state is.  This can be seen more definitively in Figure 4.

Poincare plot of HRV

Figure 4.  Poincaré Plot Graphic of Non-overtrained and Overtrained States.  Each of the dots are composed of two R-R intervals: RRn (horizontal axis) and RRn+1 (vertical axis). Place where SD1 and SD2 lines cross each other represent mean value of all R-R intervals. Line-of-identity is presented by SD2 line and all dots along the line-of-identity corresponding to long-term variability (i.e., sympathetic and parasympathetic modulation) while all perpendicular to line-of-identity (SD1) corresponding to short-term variability (i.e., parasympathetic modulation). (Makivic et al.)

Use Cases and Published References

Monitor Fitness Change

Recovery of RMSSD returns to baseline faster in highly trained subjects than in trained subjects. In highly trained subjects, RRi values after running return to baseline within 5-10 minutes in highly trained athletes, while trained athletes took > 90 minutes to return to baseline after the same interval session (Seiler et al.)

Monitoring Training Status

Changes in Ln rMSSD when examined in a 7-day rolling average indicate an individual’s readiness for training, with decreases marking a fatigued condition.  (Plews et al., 2013)

LF in overtrained athletes was elevated in rest supine position, after heavy endurance training.  (Uusitalo et al.)

Pre-OT state in athletes demonstrate a significant decrease or increase in time domain HRV indices RMSSD and SDNN.  This study found that severe changes from the baseline that lasted >2 weeks, both increases and decreases, indicated an athlete’s transition into NFOR.  (Tian et al.)

Mental Stress

RRi (Lackner 2011, Papousek 2010, Schubert 2009, Taelman 2011, Tharion 2009, Visnovcova 2014, Vuksanovic 2007), RMSSD (Li 2009, Li 2009, Taelman 2011, Tharion 2009), pNN50 (Taelman 2011, Taelman 2009), SDRR (Schubert 2009, Taelman 2011, Tharion 2009, Visnocova 2014), and HF (Hjortskov 2004, Li female 2009, Li male 2009, Papousek 2010, Taelman 2011, Tharion 2009, Traina 2011, Visnovcova 2014, Vuksanovic 2007) decrease with increasing mental stress.  LF (Hjortskov 2004, Lackner 2011, Papousek 2010, Taelman 2011, Tharion 2009, Traina 2011, Visnovcova 2014, Vuksanovic 2007) and LF/HF (Hjortskov 2004, Papousek 2010, Schubert 2009, Tharion 2009, Traina 2011, Vuksanovic 2007, Kofman 2006) increase with increasing mental stress.  (exact citations can be found in Castaldo et al.)

Sleep/Fatigue

RR-interval PSD in 0.02-0.08 Hz range correlated strongly with PVT (psychomotor vigilance task) lapses compared to other frequency bands.  (Chua et al.)

Decreased LF and LF/HF (increased parasympathetic activity) were seen prior to driving errors.  (Michail et al., Task Force)

Slow-wave sleep (SWS) is characterized by higher nonlinear HRV.  Nonlinear HRV was analyzed by using Sample Entropy (SampEn), which estimated the entropy of the RR interval time series as a measure of system complexity.  REM is associated with increased linear regimes of HRV in all frequency components (VLF, LF, HF).  (Vigo et al.)

Awakening when in the supine position resulted in a significant increase in the LF/HF ratio (P < .01) in the healthy subjects, but no significant changes in HRV were observed after awakening in patients with CAD.  (Huikuri et al., 1994)

Chronic Illness and/or Physiological Indication

More changes in RRi (number of changes in RR interval over a particular magnitude that are occurring per unit time) were experienced by healthy subjects than by diabetic patients with cardiac parasympathetic damage.  Essentially, pNN50 was higher in healthier subjects than in diabetic patients with cardiac parasympathetic damage.  (Ewing et al.)

HRV (24-hour SDNN) during the early phase of AMI (acute myocardial infarction) is decreased and is significantly related to clinical and hemodynamic indexes of severity. (Casolo et al.)

Awakening when in the supine position resulted in a significant increase in the LF/HF ratio (P < .01) in the healthy subjects, but no significant changes in HRV were observed after awakening in patients with CAD.  (Huikuri et al., 1994)

No significant circadian rhythms in any of the normalized spectral components of HRV were observed in patients with CAD, and the night-day difference in LF/HF ratio was smaller in the patients with CAD than in the healthy subjects (0.5 +/- 1.4 versus 1.8 +/- 0.7, P < .001).  (Huikuri et al., 1994)

All HRV measures (i.e. 2-hour SDNN, pNN50, rMSSD, VLF-100, LF, HF, TP-100) except LF/HF ratio were significantly associated with risk for a cardiac event (P=.0016 to .0496). A one–standard deviation decrement in 2 hour SDNN was associated with a hazard ratio of 1.47 for new cardiac events (95% confidence interval of 1.16 to 1.86).  (Tsuji et al.)

All frequency domain measures of HRV, i.e., total power (p < 0.001), high-frequency power (p < 0.05), low-frequency power (p < 0.01), very-low-frequency power (p < 0.01), and ultralow-frequency power (p < 0.05), were significantly lower before the onset of sustained ventricular tachycardia (VTach) than before nonsustained (VTach).  (Huikuri et al., 1993)

Extended Poincaré analysis identified paroxysmal atrial fibrillation (PAF) accurately in 90% of subjects, while conventional ECG analysis detected 14% of subjects to have PAF.  Both types of analyses correctly identified permanent atrial fibrillation.  (Duning et al.)

Reference Details

Casolo, G C, P Stroder, C Signorini, F Calzolari, M Zucchini, E Balli, A Sulla and S Lazzerini.  Heart rate variability during the acute phase of myocardial infarction.  Circulation 85 (1992): 2073-2079.

Castaldo, Rossana, Paolo Melillo, and Leandro Pecchia. “Acute Mental Stress Assessment via Short Term HRV Analysis in Healthy Adults: A Systematic Review.” IFMBE Proceedings 6th European Conference of the International Federation for Medical and Biological Engineering (2015): 1-4.

Chua, Eric Chern-Pin, Wen-Qi Tan, Sing-Chen Yeo, Pauline Lau, Ivan Lee, Ivan Ho Mien, Kathiravelu Puvanendran, and Joshua J. Gooley. “Heart Rate Variability Can Be Used to Estimate Sleepiness-related Decrements in Psychomotor Vigilance during Total Sleep Deprivation.” Sleep (2012): n. pag.

Duning, Thomas, P. Kirchhof, H. Wersching, T. Hepp, and R. Reinhardt. “Extended Electrocardiographic Poincare Analysis (EPA) for Better Identification of Patients with Paroxysmal Atrial Fibrillation.” Journal of Clinical & Experimental Cardiology 02.02 (2011): n. pag.

Earnest, C. P., R. Jurca, T. S. Church, J. L. Chicharro, J. Hoyos, and A. Lucia. “Relation between Physical Exertion and Heart Rate Variability Characteristics in Professional Cyclists during the Tour of Spain.” British Journal of Sports Medicine 38.5 (2004): 568-75

Ewing DJ, Neilson JMM, Traus P. “New method for assessing cardiac parasympathetic activity using 24-hour electrocardiograms”. British Heart Journal 52 (1984): 396-402.

Halson, Shona L. “Monitoring Training Load to Understand Fatigue in Athletes.” Sports Medicine 44.S2 (2014): 139-47.

Huikuri, H. V., M. J. Niemela, S. Ojala, A. Rantala, M. J. Ikaheimo, and K. E. Airaksinen. “Circadian Rhythms of Frequency Domain Measures of Heart Rate Variability in Healthy Subjects and Patients with Coronary Artery Disease. Effects of Arousal and Upright Posture.” Circulation 90.1 (1994): 121-26.

Huikuri, H. V., J. O. Valkama, K. E. Airaksinen, T. Seppanen, K. M. Kessler, J. T. Takkunen, and R. J. Myerburg. “Frequency Domain Measures of Heart Rate Variability before the Onset of Nonsustained and Sustained Ventricular Tachycardia in Patients with Coronary Artery Disease.” Circulation 87.4 (1993): 1220-228.

Makivić, Bojan, Marina Djordjević Nikić, and Monte S. Willis. “Heart Rate Variability (HRV) as a Tool for Diagnostic and Monitoring Performance in Sport and Physical Activities.” Journal of Exercise Physiology 16.3 (2013): 103-30. American Society of Exercise Physiologists, June 2013.

Michail, Emmanouil, Athina Kokonozi, Ioanna Chouvarda, and Nicos Maglaveras. “EEG and HRV Markers of Sleepiness and Loss of Control during Car Driving.” 2008 30th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (2008): n. pag.

Penttilä, Jani, Antti Helminen, Tuomas Jartti, Tom Kuusela, Heikki V. Huikuri, Mikko P. Tulppo, Rene Coffeng, and Harry Scheinin. “Time Domain, Geometrical and Frequency Domain Analysis of Cardiac Vagal Outflow: Effects of Various Respiratory Patterns.” Clinical Physiology 21.3 (2001): 365-76.

Plews, Daniel J., Paul B. Laursen, Andrew E. Kilding, and Martin Buchheit. “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 (2012): 3729-741.

Plews, Daniel J., Paul B. Laursen, Jamie Stanley, Andrew E. Kilding, and Martin Buchheit. “Training Adaptation and Heart Rate Variability in Elite Endurance Athletes: Opening the Door to Effective Monitoring.” Sports Medicine 43.9 (2013): 773-81

Quintana, Daniel S., and James A. J. Heathers. “Considerations in the Assessment of Heart Rate Variability in Biobehavioral Research.” Frontiers in Psychology 5 (2014): n. pag.

Schäfer, Axel, and Jan Vagedes. “How Accurate Is Pulse Rate Variability as an Estimate of Heart Rate Variability?” International Journal of Cardiology 166.1 (2013): 15-29.

Seiler, Stephen, Olav Haugen, and Erin Kuffel. “Autonomic Recovery after Exercise in Trained Athletes.” Medicine & Science in Sports & Exercise 39.8 (2007): 1366-373.

Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. Electrophysiology. “Heart Rate Variability : Standards of Measurement, Physiological Interpretation, and Clinical Use.” Circulation 93.5 (1996): 1043-065. Web.

Tian, Ye, Zi-Hong He, Jie-Xiu Zhao, Da-Lang Tao, Kui-Yun Xu, Conrad P. Earnest, and Lars R. Mc Naughton. “Heart Rate Variability Threshold Values for Early-Warning Nonfunctional Overreaching in Elite Female Wrestlers.” Journal of Strength and Conditioning Research 27.6 (2013): 1511-519.

Tsuji, H., M. G. Larson, F. J. Venditti, E. S. Manders, J. C. Evans, C. L. Feldman, and D. Levy. “Impact of Reduced Heart Rate Variability on Risk for Cardiac Events: The Framingham Heart Study.” Circulation 94.11 (1996): 2850-855.

Uusitalo, A. L., A. J. Uusitalo, and H. K. Rusko. “Heart Rate and Blood Pressure Variability During Heavy Training and Overtraining in the Female Athlete.” International Journal of Sports Medicine 21.1 (2000): 45-53.

Vigo, Daniel E., Javier Dominguez, Salvador M. Guinjoan, Mariano Scaramal, Eduardo Ruffa, Juan Solernó, Leonardo Nicola Siri, and Daniel P. Cardinali. “Nonlinear Analysis of Heart Rate Variability within Independent Frequency Components during the Sleep-wake Cycle.” Autonomic Neuroscience 154.1-2 (2010): 84-88.