Heart Rate Variability to Assess Autonomic Function Phyllis K. Stein, Ph.D. Research Assistant Professor of Medicine and Director, HRV Lab Washington University School of Medicine, St. Louis, MO
Heart Rate Variability to Assess Autonomic Function
Phyllis K. Stein, Ph.D.Research Assistant Professor of Medicine and Director, HRV LabWashington University School of
Medicine,St. Louis, MO
PART I
Understanding ECGs and How the Heart Works
Overview of Blood
Circulation
The Heartbeat
Valves
Valves
Electrical Pathways
Action Potential Basics
1 2 3 4 5Resting voltage
Resting voltage
Cardiac Action Potential
Components of the ECG
ECG Measurements
Autonomic Nervous System Effects on the Heart
Parasympathetic Nervous System (PNS),
inhibits cardiac action potentials
Sympathetic Nervous System (SNS),
stimulates cardiac action potentials
Single Channel Normal ECG
p wave
QRS complex
t wave
A Normal 12 Lead ECG
Atrial Premature Contraction (APC)
Abnormal p wave
Early QRS
Atrial Bigeminy
Atrial Fibrillation (AF)
Normal ECG with Ventricular Premature Contractions (VPCs)
VPCs
Right Bundle Block (RBB)
Wide QRS peak
Dangerously Abnormal ECGS
Ventricular Tachycardia (VT)
Ventricular Fibrillation (VF)
Keywords
• Atrium• Ventricle• SA node• AV node• ECG Components• P wave• QRS complex• T wave • Sympathetic Nervous
System
• Parasympathetic Nervous System
• Vagal• APC or SVE• Bigeminy• VPCs• VT• VF
PART IIHolter and Other Continuous ECG
Data
Patient wearing a Holter device.
Heart Rate Variability (HRV) Lab Analyzes Data from Continuous
Electronically-Stored ECGs
Holter Monitor2 or 3 channels of SimultaneousECG signals
Cassette Tape
Flash Card
Continuous ECG Data Also Obtained from Overnight Sleep Studies
• Sleep studies have many channels of data including ECG
• Data stored on a hard disk and file exported to a CD
• One channel is ECG
Analysis of Stored ECG Signals
• Continuous ECG signal is digitized and loaded on the Holter scanner
• Holter scanner is a computer with special commercial software that can process ECGs
• Many other computer algorithms exist that can display and measure things from ECGs
The Job of the Holter Scanner
• Read and display the stored ECG
• Identify the peak of each beat
• Accurately label each beat as normal, APC or VPC
• Measure the time between the peaks of each beat
• Create a report describing the recording
• Export the results as a “beat file”
The QRS File
• MARS scanner exports “QRS” files.
• QRS file is a list of every detected event on the tape, with the time after the next event.
• Events can be normal beats, APCs, VPCs or just noise.
• QRS file is in binary format, so we need to convert it to something we can read.
Digitized ECG Format
• .MIT Format– Binary format– Consists of a .HDR file and .SIG file
• .RAW file– Binary format– Does not contain any header info– Can be reloaded onto MARS like tape
• .NAT file– Actual file on MARS– Can be reloaded into MARS “slot” and restore all original
data and analyses
The .MIB file• QRS file from the MARS scanners are
saved to “HRV.”• “HRV” is the name of the Sun computer
that does all HRV calculations.• QRS file is converted to MIB file and stored
on “HRV.”• .MIB= machine-independent beatfile• Heart rate variability is calculated from
the .MIB file
Example of the Beginning of a .MIB File
• # 13:46:03.726• Study code=8050MJP OK,1• Record number code=8050MJP1• Start time=13:41:00• First beat=13:46:03.726• Start date=02-May-03• Samples per second=128• Marquette conversion date=Thu Jun 10 13:19:17 2004• Marquette hardware revision=508 833 523 4.00 0.25• End header• Q0.000000000• Q687.500000000• Q617.187500000• Q656.250000000• Q656.250000000• Q656.250000000• Q648.437500000• Q656.250000000• Q656.250000000• Q687.500000000• Q625.000000000• Q656.250000000• Q656.250000000• Q656.250000000• Q656.250000000
header
Files Generated from the .MIB File
• All heart rate variability calculations are made and exported to an EXCEL spreadsheet with one row per subject
• Heart rate tachograms -beat-by-beat plots of heart rate vs. time
• HRV power spectral plots - graphical representation of HRV
• HRV Poincaré plots - graphical representations of HR patterns
Part of an HRV Spreadsheet
ID avnnT avnnD avnnN pnn50T pnn50D pnn50N
1A36181 1010.034 988.613 1043.868 5.559 6.188 4.36
1A49681 999.295 988.617 1016.784 1.295 2.018 0.586
1A75451 846.611 849.501 836.082 0.482 0.4 0.572
1B74381 810.154 813.078 780.171 9.725 10.264 4.494
1B74391 725.69 710.065 777.362 6.451 5.553 12.008
1B74401 866.626 821.987 930.132 15.402 8.237 35.138
1B76181 674.383 703.628 646.714 0.933 1.38 0.398
1B76191 817.108 826.079 789.545 2.274 3.173 1.034
• x-axis = time in minutes (0-10 minutes)
• y-axis for each 10-min plot is H (0-100 bpm in 5 cm)
• “x-axis” is mean HR for that 10-min segment
Heart Rate Tachogram
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10
Tim e (M in.)
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10
00:09:00
00:19:00
00:29:00
00:39:00
00:49:00
00:59:00
0-100 bpm
“x-axis”
Hourly HRV Power Spectral Plots (much reduced in size)
Hourly Poincaré plots(much reduced in size)
Keywords
• Holter• Scanner• Beat file• QRS File• Binary• .MIB• Header
• Recognize:– Tachograms– Power spectral plots– Poincaré plots
Part III
HRV in Detail
Background (HRV)
• Decreased heart rate variability
• Abnormal heart rate variability
• Identify patients with autonomic abnormalities who are at increased risk of arrhythmic events.
Simplified Model of Cardiovascular Autonomic
Control
Renin angiotensinsystem
Heart Rate Cardiac outputBlood pressure
Parasympathetic Nervous system
SympatheticNervous system
How HRV Reflects the Effect of the Autonomic Nervous System
of the Heart
HR Fluctuations
• Fluctuations in HR (HRV) are mediated by sympathetic (SNS) and parasympathetic (PNS) inputs to the SA node.
• Rapid fluctuations in HR usually reflect PNS control only (respiratory sinus arrhythmia).
• Slower fluctuations in HR reflect combined SNS and PNS + other influences.
Rapid Fluctuations in HR Are Vagally Mediated
• “Rapid” fluctuations in HR are at >10 cycles/min (respiratory frequencies)
• Vagal effect on HR mediated by acetylcholine binding which has an immediate effect on SA node.
• If HR patterns are normal, rapid fluctuations in HR are vagally modulated
Acetylcholine Binding
The Acetylcholine Neurotransmitter binds to a receptor on a muscle once released from a
neuron.
Slower Fluctuations in HR Reflect Both SNS and Vagal Influences
• “Slower” fluctuations in HR are <10 cycles per min.
• SNS effect on HR is mediated by norepinephrine release which has a delayed effect on SA node
• Both SNS and vagal nerve traffic fluctuate at >10 cycles/min, but the time constant for changes in SNS tone to affect HR is too long to affect HR at normal breathing frequencies.
NE blinds to the beta-receptor (Alpha subunit of G-protein).
After binding, G protein links to second messenger (adenyl cyclase) which converts ATP to cAMP. cAMP activates protein kinase A which breaks ATP to ADP+phosphate which phosphorylates the pacemaker channels and increases HR
Sympathetic activation takes too long to affect RSA
Assessment of HRV
Approach 1
•Physiologist’s Paradigm
HR data collected over short period of time (~5-20 min), with or without interventions, under carefully controlled laboratory conditions.
Approach 2
Clinician’s/Epidemiologists’s Paradigm
Ambulatory Holter Recordings usually collected over 24-hours or less, usually on outpatients.
Assessment of HRV
Approaches 1 and 2 can be combined
Longer-term HRV-quantifies changes in HR over periods of >5min.
Intermediate-term HRV-quantifies changes in HR over periods of <5 min.
Short-term HRV-quantifies changes in HR from one beat to the next
Ratio HRV-quantifies relationship between two HRV indices.
HRV Perspectives
Sources of Heart Rate Variability
• Extrinsic– Activity - Sleep Apnea– Mental Stress - Smoking– Physical Stress
• Intrinsic Periodic Rhythms– Respiratory sinus arrhythmia– Baroreceptor reflex regulation– Thermoregulation– Neuroendocrine secretion– Circadian rhythms– Other, unknown rhythms
Ways to Quantify HRV
Approach 1: How much variability is there?Time Domain and Geometric Analyses
Approach 2: What are the underlying rhythms? What physiologic process do they represent? How much power does each underlying rhythm have?
Frequency Domain Analysis
Approach 3: How much complexity or self-similarity is there?
Non-Linear Analyses
Time Domain HRV
• SDNN-Standard deviation of N-N intervals in msec (Total HRV)
• SDANN-Standard deviation of mean values of N-Ns for each 5 minute interval in msec (Reflects circadian, neuroendocrine and other rhythms + sustained activity)
Longer-term HRV
• SDNNIDX-Average of standard deviations of N-Ns for each 5 min interval in ms (Combined SNS and PNS HRV)
• Coefficient of variance (CV)-
SDNNIDX/AVNN. Heart rate
normalized SDNNIDX.
Time Domain HRV
Intermediate-term HRV
Time Domain HRV
• rMSSD-Root mean square of successive differences of N-N intervals in ms
• pNN50-Percent of successive N-N differences >50 ms
Calculated from differences between successive N-N intervals
Reflect PNS influence on HR
Short-term HRV
Geometric HRV
HRV Index-Measure of longer-term HRV
From Farrell et al, J am Coll Cardiol 1991;18:687-97
Examples of Normal and AbnormalGeometric HRV
Frequency Domain HRV
• Based on autoregressive techniques or fast Fourier transform (FFT).
• Partitions the total variance in heart rate into underlying rhythms that occur at different frequencies.
• These frequencies can be associated with different intrinsic, autonomically-modulated periodic rhythms.
What are the Underlying Rhythms?
One rhythm5 seconds/cycle or12 times/min
5 seconds/cycle= 1/5 cycle/second
1/5 cycle/second= 0.2 Hz
What are the Underlying Rhythms?
Three Different Rhythms
High Frequency = 0.25 Hz (15 cycles/minLow Frequency = 0.1 Hz (6 cycles/min)Very Low Frequency = 0.016 Hz (1 cycle/min)
Ground Rules for Measuring Frequency Domain HRV
• Only normal-to-normal (NN) intervals included• At least one normal beat before and one normal beat
after each ectopic beat is excluded• Cannot reliably compute HRV with >20% ectopic
beats
• With the exception of ULF, HRV in a 24-hour recording is calculated on shorter segments (5 min) and averaged.
Longer-Term HRV
• Total Power (TP)
Sum of all frequency domain components.
• Ultra low frequency power (ULF)
At >every 5 min to once in 24 hours. Reflects circadian, neuroendocrine, sustained activity of subject, and other unknown rhythms.
Frequency Domain HRV
Intermediate-term HRV
• Very low frequency power (VLF)
At ~20 sec-5 min frequencyReflects activity of renin-angiotensin system, vagal activity, activity of subject.Exaggerated by sleep apnea. Abolishedby atropine
• Low frequency power (LF)
At 3-9 cycles/minBaroreceptor influenceson HR, mediated by SNS and vagal
influences. Abolished by atropine.
Frequency Domain HRV
Short-term HRV
• High frequency power (HF)
At respiratory frequencies
(9-24 cycles/minute, respiratory sinus arrhythmia but may also include non-respiratory sinus arrhythmia). Normally abolished by atropine.
Vagal influences on HR with normal patterns.
Frequency Domain HRV
Frequency Domain HRV
• LF/HF ratio-may reflect SNS:PNS balance under some conditions.
• Normalized LF power= LF/(TP-VLF)-correlates with SNS activity under some conditions.
• Normalized HF power=HF/(TP-VLF)-proposed as a measure of relative vagal control of HR. Increased for abnormal HRV.
Ratio HRV
0.20 Hz 0.40 Hz0
LF peak
HF peak
24-hour average of 2-min power spectral plots in a healthy adult
Relationship of Time and Frequency Domain HRV
SDNN Total Power
SDANN Ultra Low Frequency Power
SDNNIDX Very Low Frequency Power Low Frequency Power
pNN50 High Frequency PowerrMSSD
Non-Linear HRV• Non-linear HRV characterize the structure
of the HR time series, i.e., is it random or self-similar.
• Increased randomness of the HR time series is associated with worse outcomes in cardiac patients.
• Non-linear HRV measures are not available from commercial Holter systems.
• Most commonly used measure of randomness is the short-term fractal scaling exponent (DFA1 or α1). Decreased DFA1 increased randomness of the HR.
• Another index is power law slope, a measure of longer term self-similarity of HR. Decreased slope worse outcome.
• Normal DFA1 is about 1.1. DFA1<0.85 is associated with higher risk.
Non-Linear HRV
Detrended Fluctuation Analysis (DFA)
Power Law Slope
Comparison of Normal and Highly Random HRV Plots