7  Applications in Clinical Practice and Healthcare

This chapter organizes the settings in which HRV is used, paying attention to the strength of the evidence.

flowchart LR
  subgraph C["Clinical (relatively strong evidence)"]
    MI["Prognosis after myocardial<br/>infarction and in heart failure"]
    DAN["Diabetic<br/>autonomic neuropathy"]
    NEO["Early detection of<br/>neonatal sepsis"]
  end
  subgraph R["Research & healthcare (developing)"]
    ST["Stress,<br/>mental health"]
    SL["Sleep"]
    SP["Exercise,<br/>conditioning"]
    BF["HRV<br/>biofeedback"]
  end
  subgraph W["Wearables"]
    WD["Smartwatches,<br/>rings, chest straps"]
  end
  C --> W
  R --> W

7.1 Clinical applications

7.1.1 Prognosis after myocardial infarction and in heart failure

The first major demonstration of the clinical value of HRV came from studies of patients after myocardial infarction.

  • Kleiger et al. (1987): among patients after myocardial infarction, the group with a 24-hour Holter SDNN below 50 ms had a relative risk of death about 5.3 times that of the group above 100 ms
  • Subsequent studies have repeatedly shown that reduced 24-hour HRV (especially SDNN and VLF) is an independent predictor of death and sudden death after myocardial infarction and in heart failure
  • Metrics derived from HRV, such as heart rate turbulence and deceleration capacity, are also used for prognosis
NotePredictive ≠ a treatment target

Low HRV predicts poor prognosis, but whether interventions that raise HRV improve prognosis is a separate question. HRV is currently positioned as one piece of reference information for risk stratification.

7.1.2 Diabetic autonomic neuropathy (CAN)

In diabetes, long nerve fibers are affected first, so the vagus (long) tends to be affected before the sympathetic nerves. Cardiovascular autonomic neuropathy (CAN) is associated with silent myocardial ischemia, orthostatic hypotension, and the risk of sudden death.

Assessment uses tests that examine heart-rate and blood-pressure responses to stimuli, typified by Ewing’s autonomic function tests.

Test Content Mainly assesses
Heart rate variation during deep breathing Difference in heart rate between inspiration and expiration during deep breathing at 6 breaths/min (E/I ratio) Vagus
Valsalva ratio RR ratio during straining (Valsalva maneuver) and after release Vagus + sympathetic
30:15 ratio on standing RR ratio of the 30th to the 15th beat after standing Vagus
Blood pressure change on standing Fall in systolic blood pressure on standing Sympathetic (vascular)

A reduction in resting short-term HRV (SDNN, RMSSD, etc.) is also used as an early indicator of CAN.

7.1.3 Early detection of neonatal sepsis

In neonatal intensive care units for very-low-birth-weight infants, a monitor (HeRO) was developed that analyzes in real time reduced heart-rate variability and the appearance of transient bradycardia (decelerations), indicating the risk of sepsis.

  • In a large randomized controlled trial (Moorman et al. 2011), displaying this monitor reduced mortality in very-low-birth-weight infants (about 10.2% → 8.1%)
  • One of the few examples in which an RCT has shown that HRV-related metrics improved clinical outcomes

7.1.4 Other clinical research

  • Sepsis, intensive care: association of reduced HRV with severity and prognosis
  • Epilepsy: detection of autonomic changes before and after seizures
  • Neurodegenerative diseases such as Parkinson’s disease: assessment of autonomic dysfunction
  • Pregnancy, fetus: fetal heart rate monitoring (loss of baseline variability of the fetal heart rate is a sign of fetal compromise)

7.2 Stress and mental health

  • Acute mental stress tasks (mental arithmetic, speech tasks, etc.) cause heart rate↑ and RMSSD/HF↓
  • Meta-analyses have shown that patients with depression or anxiety disorders tend to have low resting HRV (Kemp et al. 2010, etc.). However, the effects of antidepressants (especially tricyclics) need to be disentangled
  • Neurovisceral integration model (Thayer & Lane): posits a circuit in which the prefrontal cortex inhibits the amygdala and regulates the heart via the vagus, and regards resting HRV as an index of emotion-regulation capacity
CautionCaveats to “HRV tells you your stress”

HRV falls in the same way with physical load (exercise, illness, sleep deprivation, drinking, dehydration). You cannot conclude “there is mental stress” from a fall in HRV alone. It is reasonable to understand a wearable’s “stress score” as, in most cases, an estimate of physical load combining HRV and heart rate.

7.3 Sleep

  • During non-REM sleep (especially deep sleep) vagal activity is high and HRV is stable; during REM sleep it becomes unstable
  • Sleep-stage estimation in wearables uses changes in heart rate and HRV in addition to body movement (acceleration)
  • Nighttime HRV tends to be a “measurement under consistent conditions” suitable for day-to-day comparison, because during sleep there is little influence from posture, activity, talking, etc.

7.4 Exercise and conditioning

7.4.1 Adjusting training with morning HRV

Mainly in endurance sports, a method is used in which resting HRV each morning (lnRMSSD) is recorded and training intensity is adjusted.

flowchart LR
  M["Morning lnRMSSD<br/>(same time and posture)"] --> A["7-day moving average"]
  A --> B{"Within the individual's<br/>reference range<br/>(about mean ± 0.5 SD)?"}
  B -->|"Within range"| H["High-intensity<br/>training as planned"]
  B -->|"Below range"| L["Reduce intensity / rest"]

  • Kiviniemi et al. (2007): the group whose training intensity was adjusted on the basis of HRV improved maximal oxygen uptake as much as or more than the group that followed a predetermined plan
  • Because daily values fluctuate a lot, the key is to use the moving average and the individual’s range of variation as the reference (Plews et al.)
  • With excessive training (overreaching), HRV often falls, but a pattern in which it rises instead (excessive vagal dominance) has also been reported, so it is not simple

7.4.2 HRV during exercise

During exercise, vagal activity almost disappears and breathing speeds up, making the usual HRV metrics hard to interpret. On the other hand, there are attempts to estimate the aerobic threshold using DFA α1 (Chapter 5).

7.5 HRV biofeedback

A method of practicing to maximize HRV through slow breathing at about 6 breaths/min (the resonance frequency) (Chapter 3).

  • The respiratory fluctuation and the baroreflex resonate, enlarging HRV
  • Effects on anxiety and depressive symptoms, stress, blood pressure, etc. have been reported with small-to-moderate effect sizes (there are meta-analyses). On the other hand, study quality and control conditions vary
  • Note that the “rise in HRV” during practice is mainly a mechanical effect of the change in breathing, and does not in itself mean an increase in vagal activity

7.6 Wearable devices

Type Measurement Accuracy Features
Chest strap (ECG-based) ECG High Close to the HRV reference standard in sports and research
Smartwatch PPG (wrist) Relatively good at rest, vulnerable to movement Daytime spot measurements, nighttime measurements
Ring PPG (finger) Relatively good at rest at night Specialized for measurement during sleep
  • The “HRV” displayed by each company differs in metric (SDNN or RMSSD), timing of measurement (daytime spot, nighttime average, specific sleep stage), and algorithm. For example, one product displays SDNN while another displays nighttime RMSSD
  • Therefore, values cannot be compared between devices. The basic approach is to look at your own trend on the same device
  • Arrhythmia detection (atrial fibrillation notifications) also uses the irregularity of PPG inter-beat intervals, but its purpose differs from HRV analysis. HRV analysis presupposes sinus rhythm, and periods of atrial fibrillation are excluded
ImportantPrinciples for using HRV in healthcare
  1. Look at within-individual change (do not compare with others or between devices)
  2. Standardize measurement conditions (time of day, posture, during sleep, etc.)
  3. Don’t overreact to a single day’s value (use moving averages and reference ranges)
  4. Interpret it in combination with heart rate, sleep, and subjective well-being
  5. Do not use it for diagnosis. If you have concerning symptoms, see a medical professional

7.7 Summary

Application Evidence Key point
Prognosis after MI and in heart failure Strong (many cohorts) 24-hour SDNN, etc. A predictor, not a treatment target
Diabetic autonomic neuropathy Strong (standard tests) Responses to deep breathing, Valsalva, standing
Neonatal sepsis Mortality reduced in an RCT An example where HRV-related metrics improved clinical outcomes
Stress, mental health Moderate (many associations, causality/specificity unclear) Cannot be distinguished from physical load
Exercise, conditioning Moderate Use the within-individual moving average as the reference
HRV biofeedback Moderate (variable quality) Resonance-frequency breathing
Health management with wearables Developing No comparison between devices; look at your own trend