ORIGINAL RESEARCH

Heart Rate Variability for Risk Assessment of Myocardial Ischemia in

Patients Without Known Coronary Artery Disease: The HRV-DETECT
(Heart Rate Variability for the Detection of Myocardial Ischemia)
Study*
Ilan Goldenberg, MD;† Ronen Goldkorn, MD;† Nir Shlomo, MA; Michal Einhorn, MA; Jacob Levitan, PhD; Raphael Kuperstein, MD;
Robert Klempfner, MD; Bruce Johnson, PhD

Background-—Detecting significant coronary artery disease (CAD) in the general population is complex and relies on combined
assessment of traditional CAD risk factors and noninvasive testing. We hypothesized that a CAD-specific heart rate variability
(HRV) algorithm can be used to improve detection of subclinical or early ischemia in patients without known CAD.
Methods and Results-—Between 2014 and 2018 we prospectively enrolled 1043 patients with low to intermediate pretest
probability for CAD who were screened for myocardial ischemia in tertiary medical centers in the United States and Israel. Patients
underwent 1-hour Holter testing, with immediate HRV analysis using the HeartTrends DyDx algorithm, followed by exercise stress
echocardiography (n=612) or exercise myocardial perfusion imaging (n=431). The threshold for low HRV was identified using
receiver operating characteristic analysis based on sensitivity and specificity. The primary end point was the presence of
myocardial ischemia detected by exercise stress echocardiography or exercise myocardial perfusion imaging. The mean age of
patients was 61 years and 38% were women. Myocardial ischemia was detected in 66 (6.3%) patients. After adjustment for CAD
risk factors and exercise stress testing results, low HRV was independently associated with a significant 2-fold increased likelihood
for myocardial ischemia (odds ratio, 2.00; 95% CI, 1.41–2.89 [P=0.01]). Adding HRV to traditional CAD risk factors significantly
improved the pretest probability for myocardial ischemia.
Conclusions-—Our data from a large prospective international clinical study show that short-term HRV testing can be used as a
novel digital-health modality for enhanced risk assessment in low- to intermediate-risk individuals without known CAD.
Downloaded from http://ahajournals.org by on January 21, 2020

Clinical Trial Registration-—URL: http://www.ClinicalTrials.gov. Unique identifiers: NCT01657006, NCT02201017). ( J Am Heart

Assoc. 2019;8:e014540. DOI: 10.1161/JAHA.119.014540.)
Key Words: coronary artery disease • heart rate variability • risk prediction

C oronary artery disease (CAD) is a major public health

problem accounting for the majority of deaths in the
United States.1 Patients are frequently asymptomatic or
known disease is complex and relies on combined assess-
ment of traditional cardiovascular risk factors, clinical symp-
toms, and noninvasive testing. Exercise stress testing (EST) is
exhibit nontypical symptoms until ischemic heart disease the most commonly used modality for CAD assessment, but,
manifests itself as sudden cardiac death or myocardial because of its limited ability to correctly diagnose myocardial
infarction. Detecting significant CAD in patients without ischemia, it is not recommended for screening in the general

From the Division of Cardiology, University of Rochester Medical Center, Rochester, NY (I.G.); Leviev Heart Center and the Israeli Association for Cardiovascular Trials,
Sheba Medical Center, Tel Hashomer, Israel (I.G., R.G., N.S., M.E., R. Kuperstein, R. Klempfner); Tel Aviv University, Tel Aviv, Israel (I.G.); Ariel University, Ariel, Israel
(J.L.); Cardiac Rehabilitation Center, Mayo Clinic, Rochester, MN (B.J.).
Accompanying Data S1, Table S1, and Figure S1 are available at https://www.ahajournals.org/doi/suppl/10.1161/JAHA.119.014540
*This work was presented as an abstract at the American Heart Association Scientific Sessions 2020, Philadelphia, PA.
†
Dr Goldenberg and Dr Goldkorn contributed equally to this work.
Correspondence to: Ilan Goldenberg, MD, Cardiology Division, Department of Medicine, University of Rochester Medical Center, 265 Crittenden Boulevard CU
420653, Rochester, NY 14642. E-mail: [email protected]
Received September 7, 2019; accepted November 13, 2019.
ª 2019 The Authors. Published on behalf of the American Heart Association, Inc., by Wiley. This is an open access article under the terms of the Creative Commons
Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-
commercial and no modifications or adaptations are made.

DOI: 10.1161/JAHA.119.014540 LJournal of the American Heart Association 1

 HRV-DETECT Goldenberg et al

ORIGINAL RESEARCH
Clinical Perspective Methods
The data that support the findings of this study are available
What Is New? from the corresponding author upon reasonable request.
• This is the first study to evaluate short-term heart rate
variability (HRV) testing for the detection of myocardial Study Population
ischemia in individuals with low to intermediate pretest The HRV-DETECT study is composed of 2 parallel multicenter
probability for coronary artery disease (CAD) in the setting
prospective studies, designed to evaluate the independent
of a large prospective multicenter clinical study.
association between HRV testing, using the HeartTrends
• Low HRV, assessed using the HeartTrends DyDx algorithm,
algorithm, and the presence of myocardial ischemia, as
was shown to be independently associated with a 2-fold
increased risk for the presence of myocardial ischemia is detected by either positive exercise stress echocardiography
individuals without known CAD. (eSE) (study 1 [NCT02201017]) or positive exercise myocar-
• HRV testing was shown to provide incremental diagnostic dial perfusion imaging (eMPI) test (study 2 [NCT01657006]).
yield to traditional CAD risk factors and to exercise stress Between May 2014 and April 2018 we prospectively
testing for the detection of myocardial ischemia. enrolled 1151 patients without known CAD who underwent
eSE or eMPI in 4 tertiary medical centers from the United States
What Are the Clinical Implications? (Mayo Clinic in Arizona and Minnesota) and Israel (Sheba
• Short-term HRV testing with the HeartTrends DyDx algo- Medical Center and Shaarei Zedek Medical Center). Inclusion
rithm can be used as a novel digital-health modality for criteria were the presence of: (1) at least 1 CAD risk factor
enhanced detection of myocardial ischemia in patients (diabetes mellitus, hypertension, smoking, positive family
without known CAD. history, and/or dyslipidemia) in asymptomatic patients
• The test can be used in conjunction with traditional referred for cardiovascular risk assessment; or (2) chest pain
cardiovascular risk factors to identify individuals without
syndromes or equivocal/equivalent angina with low to inter-
known CAD who have an increased likelihood for the
mediate pretest probability for CAD. Detailed inclusion and
presence of myocardial ischemia.
exclusion criteria are presented in Table S1. Of the 1151
• In the era of wearable digital monitoring devices and
increased interest in personalized approaches to risk
enrolled patients, 67 had uninterpretable Holter results and 31
assessment, HRV may provide useful information to direct did not complete the EST. Thus, the final study sample for the
Downloaded from http://ahajournals.org by on January 21, 2020

lifestyle change and to monitor general health status. present analysis comprised 1043 patients, of whom 612 (59%)
underwent evaluation using eSE and 431 (41%) were evaluated
using eMPI.
population.2–5 Accordingly, additional noninvasive modalities The 2 studies were approved by the institutional review
with superior sensitivities are necessary for CAD screening boards of all participating centers and all patients provided
and evaluation in individuals without known disease. informed consent before enrollment.
Heart rate variability (HRV) values have been shown to be
low in patients with CAD, and low HRV has been shown to be
an independent predictor of cardiovascular mortality and Study Design
sudden cardiac death.6–13 Prior data from 2 studies suggest The design of the HRV-DETECT study is presented in Figure 1.
that low HRV, as assessed by the HeartTrends algorithm, may Eligible and consenting patients underwent a 1-hour period of
provide a higher sensitivity for CAD detection compared with digital Holter ECG recording for the purpose of accurate heart
conventional EST.14,15 rate recording. Application of ECG electrodes was performed by
The multicenter prospective HRV-DETECT (Heart Rate medical technicians following standard recommendations,
Variability for the Detection of Myocardial Ischemia) study using approved Holter stickers. The 1-hour Holter ECG data
was designed to evaluate the independent association of HRV were used for the HRV analysis by the HeartTrends algorithm.
with the presence of myocardial ischemia among 1043 Subsequently, all patients underwent standard EST together
patients without known CAD who underwent evaluation for with either echocardiography or myocardial perfusion imaging.
the presence of myocardial ischemia. We hypothesized that The decision to perform eSE or eMPI was left to the discretion of
HRV testing, using the new HeartTrends algorithm, may the referring physician. Both tests were performed according to
provide incremental risk stratification data to traditional established American College of Cardiology/American Heart
cardiovascular risk factors and EST for the detection of Association (AHA) clinical practice guidelines.3
subclinical or early ischemia in patients without known CAD, An independent core laboratory comprising 3 cardiologists
possibly permitting more timely detection and earlier inter- experienced in echocardiogram readings blinded to the
vention in this population. results of the HRV and the clinical characteristics of enrolled

DOI: 10.1161/JAHA.119.014540 LJournal of the American Heart Association 2

 HRV-DETECT Goldenberg et al

ORIGINAL RESEARCH
Subject idenficaon:
No known CAD; Low-to-intermediate
pre-test probability

Subjects meeng inclusion/exclusion

N = 1151 subjects

One-hour digital Holter recorded. Heart rate data encrypted,

Heart rate data extracted for HRV HRV analyzed offline &
analysis interpreted

HRV, EST, and CAD risk factors

EST data interpreted, blinded Final analysis
Perform Exercise Stress Test (EST) assessed as independent
to Echocardiogram/MPI sample
followed by Stress Echo or MPI predictors of a posive
and HRV results N= 1043
ECHO/MPI and long-term
cardiovascular events

Paent discharged with Stress

Echo/MPI result. No intervenon in
Excluded:
paent’s management
• Uninterpretable Holter (N=67)
• Incomplete EST (N=31)

Figure 1. Flow diagram of the HRV-DETECT (Heart Rate Variability for the Detection of Myocardial Ischemia) study design. CAD indicates
coronary artery disease; EST, exercise stress test; HRV, heart rate variability; MPI, myocardial perfusion imaging.
Downloaded from http://ahajournals.org by on January 21, 2020

patients adjudicated the EST and stress echocardiograms. indices, derive information from both the time and frequency
Similarly, a separate independent core laboratory comprising domains, as well as reflect increased randomness in the RR
3 cardiologists experienced in nuclear medicine adjudicated interval time series.
the myocardial perfusion imaging results. Analysis of the From the detrended RR time series the algorithm calculates
recorded Holter data by the HeartTrends algorithm was different multipoles––quadrupoles, octoupoles, and hexade-
performed in an offline fashion, blinded to the EST and capoles––and derives the new HRV parameter DyDx. Quadru-
imaging results. poles describe the overall distribution of data points in the
The results of the HeartTrends algorithm were not available Poincar
e Plot, ie, the shape of the plot. DyDx calculates the ratio
to the treating physicians and were not used to guide between the peak density on the y axis (Dy) and the x axis (Dx),
management. respectively. A more detailed description of the technical
aspects of the HeartTrends device is provided in Data S1.
HRV Algorithm
The Multipole method has been described in detail else- Definitions and Outcomes Measures
where.16,17 Briefly, the Multipole HRV analysis is a new way of EST was defined as positive per AHA guidelines.18,19 The main
investigating the Poincar
e Plot from complex time series. The ST-segment criteria included ≥1 mm of horizontal or downslop-
algorithm interprets the Poincar
e Plot as a 2-dimensional ing ST-segment depression ≥80 ms after the J point (as
body, where each data point in the plot is assigned a unit compared with the level of the PQ interval) for 3 consecutive
mass to describe the total mass distribution within the plot. beats or ST-segment elevation ≥1 mm in a non–Q-wave lead
The measures obtained from these kinds of analyses bear other than V1 or AVR. Additional non–ST-segment ECG and
intrinsic time dependence because of the construction of the clinical criteria for a positive EST test were implemented per
plot as opposed to SD of the normalized NN interval analysis, AHA guidelines for exercise stress testing.18,19
which does not include any time ordering (shuffling the RR Myocardial perfusion imaging was defined as positive when
intervals lead to the same value for SD of the normalized NN the amount of myocardial ischemia was >5% of the
interval). The Multipole method, as do other Poincar
e plot myocardium.

DOI: 10.1161/JAHA.119.014540 LJournal of the American Heart Association 3

 HRV-DETECT Goldenberg et al

ORIGINAL RESEARCH
Stress echocardiograms were assessed as follows: seg- in model 2, HRV was added to the CAD covariates of model 1;
mental wall motion was evaluated and scored using the and in model 3, the EST results and resting heart rate were
method by the American Society of Echocardiography4 as added to the covariates of model 2 to assess the independent
normal (score of 1), hypokinetic (score of 2), akinetic (score of association of HRV with the presence of myocardial ischemia
3), or dyskinetic (score of 4). A wall motion score index was after further adjustment for EST and heart rate. The most
derived by summation of individual segment scores divided by appropriate predictive model for myocardial ischemia was
the number of interpreted segments. Inadequately visualized selected using a stepwise algorithm based on Akaike informa-
segments were not scored. A final wall motion score index tion criterion. Pretest and posttest probabilities for the
was derived for rest and peak stress echocardiograms by presence of myocardial ischemia were calculated by assessing
consensus between the 2 observers. If consensus in reading the rate of myocardial ischemia among patients with 1, 2, and
stress echocardiography study could not be reached, the 3 CAD risk factors, before and after adding HRV as an
judgment of a third observer was obtained. A stress additional risk factor. Area under the curve analysis was used
echocardiography test was considered positive when new or to compare the effect of HRV on the improvement in the
worsening of preexisting wall motion abnormality was detection of myocardial ischemia with conventional CAD risk
observed. Only echocardiographic criteria (new or worsening factors and EST, with Delong testing used for detecting
wall motion abnormality) were considered as positive tests.4 differences in the area under the curve.
The HRV score, as assessed by the HeartTrends algorithm, Sensitivity analyses were performed by bootstrapping: (1)
was dichotomized using receiver operating characteristic the ROC analysis for the identification of the optimal HRV
(ROC) analysis based on sensitivity and specificity as low threshold; and (2) for the assessment of the association
versus high in the primary analysis, and was also assessed as between HRV and the presence of myocardial ischemia in
a continuous measure and categorized by quartiles in 2 models 2 and 3. For this purpose, 100 bootstrap samples
additional secondary analyses. were generated. Bland-Altman analysis was used to evaluate
the consistency of the HeartTrends algorithm derived from the
1-hour Holter data to the results derived from 20-minute
Statistical Analysis recordings. Analyses were performed using SAS software
We evaluated the association of the HeartTrends HRV (version 9.30; SAS Institute).
algorithm with the presence of significant myocardial
Downloaded from http://ahajournals.org by on January 21, 2020

ischemia as detected by a positive eSE or eMPI (defined

above). The sample size was calculated to show an indepen-
Results
dent association between HRV with the presence of signifi- The mean age of the 1043 study participants was 61 years
cant myocardial ischemia. With an expected event rate of 5%, (10 years) and 56% were men. All patients had low to
we estimated that a sample of 1000 patients would be intermediate pretest probability for CAD, with a relatively high
required to detect an odds ratio of 1.6 for HRV as a binary frequency of cardiovascular risk factors, including hyperten-
predictor of myocardial ischemia with >80% power, using 2- sion (45%), diabetes mellitus (17%), dyslipidemia (51%), and a
sided 0.05 level tests and assuming a 10% dropout rate and family history of CAD (49%).
uninterpretable tests. ROC analysis identified a DyDx HRV value of 2.6 as being
To identify the optimal decision threshold of HRV result for optimal for detection of myocardial ischemia among study
myocardial ischemia, ROC analysis was performed, and the patients. Thus, 433 study patients (42%) had low HRV (≤2.6)
best threshold was selected based on sensitivity and speci- and 610 study patients (58%) had high HRV (>2.6). Baseline
ficity. Baseline clinical characteristics of study patients were characteristics of study patients by low versus high HRV are
assessed by HRV. Continuous variables were compared using t shown in Table 1. Patients with low HRV had a significantly
test or Kruskal–Wallis test as appropriate for normal/nonnor- higher frequency of known CAD risk factors, including older
mal, distributed, continuous variables and expressed as age, hypertension, diabetes mellitus, and dyslipidemia, and
meanSD/median (interquartile range). Categorical variables were likely to be treated with cardiovascular medications.
were assessed using chi-square test or Fisher exact test, when
at least 1 of the cells in the table had an expected number <5.
Multivariate regression analysis was used to assess the
Association Between HRV and the Presence of
association of HRV, assessed as a continuous measure, with Myocardial Ischemia
the presence of myocardial ischemia, as detected by eSE and Myocardial ischemia, as detected by positive eSE or eMPI,
eMPI. Multivariate logistic regression modeling was used to was present in 66 (6.3%) patients. The frequency of myocar-
evaluate HRV as a categorical measure. Modeling was dial ischemia was significantly higher among patients with low
performed in 3 steps. Model 1 included only CAD risk factors; HRV (11%) as compared with those with high HRV (3%,

DOI: 10.1161/JAHA.119.014540 LJournal of the American Heart Association 4

 HRV-DETECT Goldenberg et al

ORIGINAL RESEARCH
Table 1. Baseline Clinical Characteristics by HRV analysis showed that low HRV was independently associated
with a 2-fold (P=0.01) increased likelihood for the presence of
Low HRV (≤2.6) High HRV (>2.6) myocardial ischemia after adjustment for CAD risk factors as
Variable (n=433) (n=610) P Value compared with high HRV (model 2 in Table 2). Furthermore,
HRV result, 1.78 (1.74–1.95) 3.21 (2.88–3.56) <0.001 when HRV was assessed as a continuous measure, multivari-
median (IQR) ate regression analysis showed that each single unit reduction
Age, median (IQR) 64 (57–70) 60 (51–67) <0.001 in HRV was independently associated with a corresponding
Age ≥65, y 236 (51) 186 (32) <0.001 52% (odds ratio, 1.52; 95% CI, 1.11–2.13 [P=0.02]) increased
likelihood for the presence of myocardial ischemia. Additional
Men 248 (54) 336 (58) 0.18
factors shown to be independently associated with the
Hypertension 242 (52) 231 (40) <0.001
presence of myocardial ischemia in the multivariate models
Diabetes mellitus 109 (24) 66 (11) <0.001 were age and a family history of CAD (Table 2).
Dyslipidemia 260 (56) 273 (47) 0.004
Family history 243 (53) 270 (47) 0.066
of CAD Use of HRV to Improve Pretest Probability for the
PVD 7 (2) 6 (1) 0.68 Detection of Myocardial Ischemia
Past TIA or CVA 6 (1) 9 (2) 0.68 We further assessed whether including HRV in the risk
Past/current smoker 240 (52) 303 (52) 0.93 assessment would improve the pretest probability for the
presence of myocardial ischemia based on traditional CAD
Medications
risk factors, including age, hypertension, and a family history
ACEIs 37 (8) 39 (7) 0.51
of CAD (Figure 2). This analysis showed that, among patients
ARBs 38 (8) 27 (5) 0.03 with none of the above CAD risk factors, the pretest
b-Blockers 75 (16) 66 (11) 0.03 probability for myocardial ischemia was 3%. Adding low
CCBs 44 (10) 39 (7) 0.13 HRV to the risk assessment increased the posttest probabil-
Statins 162 (35) 128 (22) <0.001 ity in patients with no CAD risk factors to 4.2%, whereas high
Diuretics 20 (4) 15 (3) 0.17 HRV reduced the posttest probability to 1.8%. Among
patients with 3 CAD risk factors, the pretest probability for
EST results
Downloaded from http://ahajournals.org by on January 21, 2020

myocardial ischemia was 12%. Adding low HRV to the risk

Ischemic 36 (8) 45 (8) 1.00
assessment in this higher-risk cohort increased the posttest
Borderline/ 46 (10) 67 (12) 0.48 probability to 16%, whereas high HRV reduced the posttest
ischemic
probability to 7%, suggesting that HRV can be used to
Noninvasive imaging test for myocardial ischemia*
improve conventional CAD risk assessment (Figure 2). Area
Definitely 47 (11) 19 (3) 0.002 under the curve analysis yielded consistent findings, demon-
ischemic
strating that adding HRV to individual CAD risk factors
Nonischemic 369 (85) 572 (93) (Figure 3A through 3C), or to all CAD factors combined
Possibly 18 (4) 19 (4) (Figure 3D), was associated with a significant improvement in
ischemic the sensitivity and specificity for the detection of myocardial
Data are shown as number (percentage) unless otherwise indicated. ACEIs indicates ischemia.
angiotensin-converting enzyme inhibitors; ARBs, angiotensin receptor blockers; CAD,
coronary artery disease; CCBs, calcium channel blockers; CVA, cerebrovascular accident;
EST, exercise stress test; HRV, heart rate variability; IQR, interquartile range; PVD, Comparison of Diagnostic Yield of EST and HRV
peripheral vascular disease; TIA, transient ischemic attack.
*Patients underwent either stress myocardial perfusion imaging or stress for the Detection of Myocardial Ischemia
echocardiography for the evaluation of ischemia.
Compared with HRV, EST was associated with a lower
sensitivity of 30% but with a higher specificity of 94% for the
P=0.002) (Table 1). Accordingly, low HRV was associated with detection of myocardial ischemia. Combined assessment of
a sensitivity of 71% for the detection of myocardial ischemia, a HRV and EST in the same multivariate model showed that
specificity of 60%, a positive predictive value of 11%, and a both tests were independently associated with the presence
negative predictive value of 97%. of myocardial ischemia (model 3 in Table 2). Furthermore,
Consistent with those findings, multivariate analysis area under the curve analysis showed that the addition of HRV
showed that low HRV was independently associated with to EST was associated with a significant improvement in the
the presence of myocardial ischemia. When dichotomized at sensitivity and specificity of EST for the detection of
the identified cutoff of 2.6, multivariate logistic regression myocardial ischemia (Figure 4).

DOI: 10.1161/JAHA.119.014540 LJournal of the American Heart Association 5

 HRV-DETECT Goldenberg et al

ORIGINAL RESEARCH
Table 2. Multivariate Logistic Regression Analysis: Independent Predictors of Myocardial Ischemia Detected by Noninvasive
Testing

Model 1 Model 2 Model 3

Variable OR (95% CI) P Value OR (95% CI) P Value OR (95% CI) P Value

Age (per y) 1.05 (1.02–1.07) 0.002 1.04 (1.01–1.07) 0.004 1.04 (1.01–1.07) 0.01
Men (vs women) 1.31 (0.78–2.21) 0.31 1.32 (0.79–2.24) 0.30 1.07 (0.61–1.89) 0.81
Hypertension 1.01 (0.58–1.73) 0.99 1.00 (0.58–1.71) 0.99 0.93 (0.52–1.64) 0.79
Diabetes mellitus 1.64 (0.88–2.95) 0.11 1.48 (0.79–2.67) 0.21 1.40 (0.72–2.61) 0.31
Family history of CAD 2.11 (1.26–3.65) 0.01 2.05 (1.22–3.55) 0.01 2.09 (1.21–3.73) 0.01
Positive HRV (≤2.57) 2.00 (1.41–2.89) 0.01 2.04 (1.46–2.92) 0.01
Resting heart rate 1.00 (0.98–1.03) 0.65
(per 1-unit increment)
Positive EST 7.03 (3.67–13.24) 0.01

CAD indicates coronary artery disease; EST, exercise stress test; HRV, heart rate variability; OR, odds ratio.

Model Stability and Sensitivity Analyses may also be adequate for data analysis. The full 60 minutes of
data were used for the current study.
Bootstrapping the ROC analysis for the identification of the
optimal threshold HRV showed a high correlation with the
identified threshold of 2.6. Bootstrapping also demonstrated Discussion
considerable stability in the association of HRV with the To our knowledge, this is the largest prospective multicenter
presence of myocardial ischemia. clinical study to evaluate the association of HRV with the
We also evaluated the consistency of the HRV results presence of myocardial ischemia in individuals without known
derived from the HeartTrends algorithm using 1-hour Holter CAD. Our findings provide several important clinical implica-
recording versus data derived from 20-minute recordings. tions regarding risk assessment for myocardial ischemia in
Downloaded from http://ahajournals.org by on January 21, 2020

Bland-Altman analysis of the Holter ECG data for mean this population. We have shown that: (1) low HRV is
difference and 95% CIs yielded a P value of 0.0003 after independently associated with the presence of myocardial
20 minutes (Figure S1), suggesting that this shorter interval ischemia after adjusting for known CAD risk factors; (2) the

25
Post-test probability with HRV (%)

20

15

Low HRV (DyDx)

10

High HRV (DyDx >2.6)

0
0 2 4 6 8 10 12 14 16 18

Pre-test probability by CAD risk factors (%)

Figure 2. Posttest probability for myocardial ischemia by heart rate variability (HRV). CAD indicates
coronary artery disease.

DOI: 10.1161/JAHA.119.014540 LJournal of the American Heart Association 6

 HRV-DETECT Goldenberg et al

ORIGINAL RESEARCH
A B

P=0.005 P=0.002

C D

P=0.005 P=0.047

Figure 3. Comparison of area under the curve (AUC) for coronary artery disease (CAD) risk factors before and after the addition of heart rate
Downloaded from http://ahajournals.org by on January 21, 2020

variability (HRV) by: (A) age; (B) diabetes mellitus; (C) family history of CAD; and (D) comparison of model 2 (CAD risk factors without HRV) and
model 3 (CAD risk factor+HRV)*. *P values were assessed using Delong tests.

addition of HRV to conventional CAD risk assessment is independent association between HRV and the risk of
associated with a significant improvement in the probability myocardial ischemia remains consistent even after adjust-
for the detection of myocardial ischemia; and (3) the ment for EST results. Furthermore, HRV was shown to
improve the diagnostic yield of EST for the detection of
myocardial ischemia. These findings suggest that HRV can be
used to improve risk stratification among patients with low to
intermediate pretest probability for CAD, providing incremen-
P=0.001
tal data to traditional cardiovascular risk factors and exercise
stress testing.

Traditional Risk Assessment for CAD

Cardiovascular disease results in 1 of every 3 deaths in the
United States, or 800 000 per year.20 CAD accounts for
more than half of all cardiovascular events in adults younger
than 75 years and is the leading cause of death.1 Among those
who die suddenly of CAD, more than half have no antecedent
symptoms.1 In addition, myocardial infarction is frequently
Figure 4. Comparison of area under the curve (AUC) for exercise silent, causing no recognized symptoms but negatively affect-
stress test (EST) before and after the addition of heart rate ing prognosis.21,22 This has resulted in screening programs
variability (HRV).* *P values were assessed using Delong tests. designed to identify CAD before it manifests clinically, with EST

DOI: 10.1161/JAHA.119.014540 LJournal of the American Heart Association 7

 HRV-DETECT Goldenberg et al

ORIGINAL RESEARCH
being the most widely used modality in screening programs.23 population but had a lower sensitivity and a higher specificity
However, EST was not shown to provide incremental benefits compared with HRV. These data are consistent with prior
to risk assessment that is based on traditional cardiovascular reports on the sensitivity and specificity of EST. In a meta-
risk factors, such as age, sex, lipid levels, blood pressure, analysis of 147 consecutively published reports involving
smoking status, and presence of diabetes mellitus, and its 24 074 patients who underwent both coronary angiography
benefits were not shown to outweigh its possible harms and and exercise testing revealed wide variability in sensitivity and
costs.24 Therefore, the American College of Physicians and specificity (mean sensitivity was 68%, with a range of 23% to
other groups recommend against screening low- and interme- 100% and an SD of 16%; mean specificity was 77%, with a
diate-risk adults without known CAD with EST or more range of 17% to 100% and an SD of 17%).27 Furthermore, in
advanced modalities such as myocardial perfusion imaging the few studies where workup bias was avoided by having
or stress echocardiography.24 Despite this, inappropriate patients agree to undergo both procedures, the approximate
cardiac testing in low-risk adults has been identified as one sensitivity and specificity of 1 mm of horizontal or downward
of the most overused clinical practices,25 possibly attributable ST depression were 50% and 90%, respectively.28,29 In the
to the fact that traditional CAD risk factors are considered by present study, EST was associated with a somewhat lower
clinical practitioners as insufficient for CAD risk assessment. sensitivity and a similar specificity compared with prior
Accordingly, additional, simpler modalities are needed for reports, possibly attributable to the lower pretest probability
improved CAD detection in patients without known disease. for CAD in our study population. In contrast, HRV was
associated with higher sensitivity of 71% and a negative
predictive value of 97%, but had a lower specificity of 60%.
HRV for CAD Risk Assessment Additionally, HRV testing was shown to significantly improve
HRV is an established cardiovascular risk factor. The associ- the diagnostic yield of EST (Figure 4). These data further
ation of HRV and prognosis, both for all-cause and cardio- support the incorporation of short-term HRV testing with
vascular mortality, has been studied using ECG at rest, with traditional cardiovascular risk factors for risk assessment in
exercise and in the ambulatory setting. A meta-analysis by individuals without known CAD, wherein a negative test can
Hillebrand and colleagues26 found that, using both resting and effectively be used as an additional noninvasive modality to
ambulatory ECG monitoring, lower HRV is associated with a rule out significant myocardial ischemia in an intermediate-
32% to 45% increased risk of first cardiovascular event in risk population and a positive test suggests the need for
Downloaded from http://ahajournals.org by on January 21, 2020

patients without known CAD. Additionally, elevated HRV additional CAD evaluation.
demonstrates a protective effect, with an increase in SD of
the normalized NN interval of 1% resulting in an 1%
reduction of fatal or nonfatal cardiovascular disease event. Limitations
The present study extends prior data on the association Despite the fact that the present study comprises a large
between HRV and CAD risk and shows that low HRV, as population of patients without known CAD who were
assessed by the novel HeartTrends algorithm, is significantly prospectively enrolled in a multicenter clinical study, the rate
associated with the presence of myocardial ischemia in a of the primary end point (a positive eSE or eMPI) was only 6%,
large population of individuals without known CAD. This possibly attenuating the statistical power to detect a statis-
association was also evident when HRV was assessed as a tically significant association with HRV and the presence of
continuous measure, wherein each 1-unit reduction in HRV myocardial ischemia. However, this event rate was expected
was independently associated with a corresponding 52% in individuals with low to intermediate pretest probability for
increased likelihood for the presence of myocardial ischemia CAD and formed the basis for the study’s sample size
after adjustment for CAD risk factors. Notably, in the present calculation. Furthermore, despite the relatively low event rate,
study low HRV, identified using ROC analysis for sensitivity our results remained statistically significant after multivariate
and specificity, was independently associated with a signif- adjustment for traditional CAD risk factors and the EST
icant 2-fold increased likelihood for the presence of myocar- results, further supporting the consistency of results on the
dial ischemia and improved the pretest probability for the independent association of HRV to the presence of myocar-
presence of ischemia among patients with 0 to 3 traditional dial ischemia.
CAD risk factors. HRV was shown to be attenuated in noncardiovascular
pathological conditions, including respiratory, neurologic, and
renal disease.6 Therefore, it is possible that low HRV may
HRV and EST for CAD Risk Assessment reflect the presence of a noncardiac pathology rather than
A positive EST was also independently associated with the myocardial ischemia. Accordingly, it is important to incorpo-
presence of myocardial ischemia in the present study rate the results of the test with the overall clinical status of

DOI: 10.1161/JAHA.119.014540 LJournal of the American Heart Association 8

 HRV-DETECT Goldenberg et al

ORIGINAL RESEARCH
the individual and in the context of the presence of additional Tomography, and Society for Cardiovascular Magnetic Resonance. Circulation.
2010;122:2748–2764.
traditional CAD risk factors. 4. Douglas PS, Khandheria B, Stainback RF, Weissman NJ, Peterson ED, Hendel RC,
It should also be noted that the present findings are Stainback RF, Blaivas M, Des Prez RD, Gillam LD, Golash T, Hiratzka LF, Kussmaul
WG, Labovitz AJ, Lindenfeld J, Masoudi FA, Mayo PH, Porembka D, Spertus JA,
applicable only to the present study population, comprising Wann LS, Wiegers SE, Brindis RG, Douglas PS, Hendel RC, Patel MR, Peterson ED,
patients with a low to intermediate pretest probability for the Wolk MJ, Allen JM; American College of Cardiology Foundation; American Society
of Echocardiography; American College of Emergency Physicians; American
presence of CAD without the presence of important comor- Heart Association; American Society of Nuclear Cardiology; Society for
bidities (such as cardiomyopathy, atrial fibrillation, and Cardiovascular Angiography and Interventions; Society of Cardiovascular
Computed Tomography; Society for Cardiovascular Magnetic Resonance.
moderate to severe pulmonary disease), which may be ACCF/ASE/ACEP/AHA/ASNC/SCAI/SCCT/SCMR 2008 appropriateness cri-
teria for stress echocardiography: a report of the American College of Cardiology
present in patients who are being evaluated for CAD. Foundation Appropriateness Criteria Task Force, American Society of Echocar-
diography, American College of Emergency Physicians, American Heart Asso-
ciation, American Society of Nuclear Cardiology, Society for Cardiovascular
Angiography and Interventions, Society of Cardiovascular Computed Tomogra-
phy, and Society for Cardiovascular Magnetic Resonance endorsed by the Heart
Conclusions Rhythm Society and the Society of Critical Care Medicine. Circulation.
2008;117:1478–1497.
HRV analysis has become an important tool in cardiology
5. Hendel RC, Berman DS, Di Carli MF, Heidenreich PA, Henkin RE, Pellikka PA,
because its measurements are noninvasive and easy to Pohost GM, Williams KA; American College of Cardiology Foundation
Appropriate Use Criteria Task Force; American Society of Nuclear Cardiology;
perform, have relatively good reproducibility, and provide American College of Radiology; American Heart Association; American Society
prognostic information on patients with heart disease. In the of Echocardiography; Society of Cardiovascular Computed Tomography;
Society for Cardiovascular Magnetic Resonance; Society of Nuclear Medicine.
era of wearable digital monitoring devices and increased ACCF/ASNC/ACR/AHA/ASE/SCCT/SCMR/SNM 2009 appropriate use cri-
teria for cardiac radionuclide imaging: a report of the American College of
interest in personalized approaches to risk assessment, HRV Cardiology Foundation Appropriate Use Criteria Task Force, the American
may provide useful information to direct lifestyle change and Society of Nuclear Cardiology, the American College of Radiology, the
American Heart Association, the American Society of Echocardiography, the
monitor general health status. Our study demonstrates that Society of Cardiovascular Computed Tomography, the Society for Cardiovas-
short-term HRV analysis using remote digital technology can cular Magnetic Resonance, and the Society of Nuclear Medicine. Circulation.
2009;119:e561–e587.
be used for improved risk assessment for myocardial 6. Xhyheri B, Manfrini O, Mazzolini M, Pizzi C, Bugiardini R. Heart rate variability
ischemia in individuals without known CAD, providing incre- today. Prog Cardiovasc Dis. 2012;55:321–331.
mental data to traditional cardiovascular risk factors. 7. Task Force of the European Society of Cardiology the North American
Society of Pacing Electrophysiology. Heart rate variability: standards of
measurement, physiological interpretation, and clinical use. Circulation.
1996;93:1043–1065.
8. Valkama JO, Huikuri HV, Koistinen J, Yli-M€ayry S, Juhani Airaksinen KE,
Sources of Funding Myerburg RJ. Relation between heart rate variability and spontaneous and
Downloaded from http://ahajournals.org by on January 21, 2020

induced ventricular arrhythmias in patients with coronary artery disease. J Am

The study was supported by an unrestricted research grant Coll Cardiol. 1995;25:437–443.
from Lev-El Diagnostics of Heart Disease, Jerusalem, Israel, to 9. Molgaard H, Sorensen KE, Bjerregaard P. Attenuated 24-h heart rate variability
in apparently healthy subjects, subsequently suffering sudden cardiac death.
the Israeli Association for Cardiovascular Trials. Clin Auton Res. 1991;1:233–237.
10. Bigger JT, Fleiss JL, Steinman RC, Rolnitzky LM, Kleiger RE, Rottman JN.
Frequency domain measures of heart period variability and mortality after
myocardial infarction. Circulation. 1992;85:164–171.
Disclosures 11. Dekker JM, Crow RS, Folsom AR, Hannan PJ, Liao D, Swenne CA, Schouten EG.
None. Low heart rate variability in a 2-minute rhythm strip predicts risk of coronary
heart disease and mortality from several causes: the ARIC study. Circulation.
2000;102:1239–1244.
12. Liao D, Cai J, Rosamond WD, Barnes RW, Hutchinson RG, Whitsel EA,
References Rautaharju P, Heiss G. Cardiac autonomic function and incident coronary heart
disease: a population-based case-cohort study: the ARIC study. Am J
1. Go AS, Mozaffarian D, Roger VL, Benjamin EJ, Berry JD, Borden WB, Bravata Epidemiol. 1997;145:696–706.
DM, Dai S, Ford ES, Fox CS, Franco S, Fullerton HJ, Gillespie C, Hailpern SM,
Heit JA, Howard VJ, Huffman MD, Kissela BM, Kittner SJ, Lackland DT, 13. Tsuji H, Larson MG, Venditti FJ, Manders ES, Evans JC, Feldman CL, Levy D.
Lichtman JH, Lisabeth LD, Magid D, Marcus GM, Marelli A, Matchar DB, Impact of reduced heart rate variability on risk for cardiac events: the
McGuire DK, Mohler ER, Moy CS, Mussolino ME, Nichol G, Paynter NP, Framingham Heart Study. Circulation. 1996;94:2850–2855.
Schreiner PJ, Sorlie PD, Stein J, Turan TN, Virani SS, Wong ND, Woo D, 14. Oieru D, Shlomo N, Moalem I, Rozen E, Naimushin A, Klempfner R, Goldenberg
Turner MB; American Heart Association Statistics Committee and Stroke I, Goldkorn R. A novel heart rate variability algorithm for the detection of
Statistics Subcommittee. Heart disease and stroke statistics—2013 update: myocardial ischemia—pilot data from a prospective clinical trial. Isr Med
a report from the American Heart Association. Circulation. 2013;127:e6– Assoc J. 2015;17:161–165.
e245.
15. Goldkorn R, Naimushin A, Shlomo N, Dan A, Oieru D, Moalem I, Rozen E,
2. Moyer VA; U.S. Preventive Services Task Force. Screening for coronary heart Gur I, Levitan J, Rosenmann D, Mogilewsky Y, Klempfner R, Goldenberg I.
disease with electrocardiography: U.S. Preventive Services Task Force Comparison of the usefulness of heart rate variability versus exercise
recommendation statement. Ann Intern Med. 2012;157:512–518. stress testing for the detection of myocardial ischemia in patients
3. Greenland P, Alpert JS, Beller GA, Benjamin EJ, Budoff MJ, Fayad ZA, Foster E, without known coronary artery disease. Am J Cardiol. 2015;115:1518–
Hlatky MA, Hodgson JM, Kushner FG, Lauer MS, Shaw LJ, Smith SC, Taylor AJ, 1522.
Weintraub WS, Wenger NK. 2010 ACCF/AHA guideline for assessment of 16. Olesen RM, Thomsen PE, Saermark K. Statistical analysis of the DIAMOND MI
cardiovascular risk in asymptomatic adults: executive summary. A report of study by the multipole method. Physiol Meas. 2005;26:591–598.
the American College of Cardiology Foundation/American Heart Association
Task Force on Practice Guidelines Developed in Collaboration With the 17. Lewkowicz ML, Puzanov N, Shnerb N, Særmark K. Description of complex time
American Society of Echocardiography, American Society of Nuclear Cardiol- series by multipoles. Physica A. 2002;311:260–274.
ogy, Society of Atherosclerosis Imaging and Prevention, Society for Cardio- 18. Fletcher GF, Ades PA, Kligfield P, Arena R, Balady GJ, Bittner VA, Coke LA, Fleg
vascular Angiography and Interventions, Society of Cardiovascular Computed JL, Forman DE, Gerber TC, Gulati M, Madan K, Rhodes J, Thompson PD,

DOI: 10.1161/JAHA.119.014540 LJournal of the American Heart Association 9

 HRV-DETECT Goldenberg et al

ORIGINAL RESEARCH
Williams MA; American Heart Association Exercise, Cardiac Rehabilitation, and 24. Chou R; High Value Care Task Force of the American College of Physicians.
Prevention Committee of the Council on Clinical Cardiology, Council on Cardiac screening with electrocardiography, stress echocardiography, or
Nutrition, Physical Activity and Metabolism, Council on Cardiovascular and myocardial perfusion imaging: advice for high-value care from the American
Stroke Nursing, and Council on Epidemiology and Prevention. Exercise College of Physicians. Ann Intern Med. 2015;162:438–447.
standards for testing and training: a scientific statement from the American 25. Kale MS, Bishop TF, Federman AD, Keyhani S. Trends in the overuse of
Heart Association. Circulation. 2013;128:873–934. ambulatory health care services in the United States. JAMA Intern Med.
19. Mensah GA, Brown DW. An overview of cardiovascular disease burden in the 2013;173:142–148.
United States. Health Aff (Millwood). 2007;26:38–48. 26. Hillebrand S, Gast KB, de Mutsert R, Swenne CA, Jukema JW, Middeldorp S,
20. Sheifer SE, Gersh BJ, Yanez ND III, Ades PA, Burke GL, Manolio TA. Prevalence, Rosendaal FR, Dekkers OM. Heart rate variability and first cardiovascular
predisposing factors, and prognosis of clinically unrecognized myocardial event in populations without known cardiovascular disease: meta-analysis and
infarction in the elderly. J Am Coll Cardiol. 2000;35:119–126. dose-response meta-regression. Europace. 2013;15:742–749.
21. Sigurdsson E, Thorgeirsson G, Sigvaldason H, Sigfusson N. Unrecognized 27. Gibbons RJ, Balady GJ, Beasley JW, Bricker JT, Duvernoy WF, Froelicher VF,
myocardial infarction: epidemiology, clinical characteristics, and the prognos- Mark DB, Marwick TH, McCallister BD, Thompson PD Jr, Winters WL, Yanowitz
tic role of angina pectoris. The Reykjavik Study. Ann Intern Med. FG, Ritchie JL, Gibbons RJ, Cheitlin MD, Eagle KA, Gardner TJ, Garson A Jr,
1995;122:96–102. Lewis RP, O’Rourke RA, Ryan TJ. ACC/AHA Guidelines for Exercise Testing. A
22. Chou R, Arora B, Dana T, Fu R, Walker M, Humphrey L. Screening report of the American College of Cardiology/American Heart Association
asymptomatic adults with resting or exercise electrocardiography: a review Task Force on Practice Guidelines (Committee on Exercise Testing). J Am Coll
of the evidence for the U.S. Preventive Services Task Force. Ann Intern Med. Cardiol. 1997;30:260–311.
2011;155:375–385. 28. Morise AP, Diamond GA. Comparison of the sensitivity and specificity of
23. Lauer M, Froelicher ES, Williams M, Kligfield P; American Heart Association exercise electrocardiography in biased and unbiased populations of men and
Council on Clinical Cardiology, Subcommittee on Exercise, Cardiac Rehabil- women. Am Heart J. 1995;130:741–747.
itation, and Prevention. Exercise testing in asymptomatic adults: a statement 29. DelCampo J, Do D, Umann T, McGowan V, Froning J, Froelicher VF.
for professionals from the American Heart Association Council on Clinical Comparison of computerized and standard visual criteria of exercise ECG
Cardiology, Subcommittee on Exercise, Cardiac Rehabilitation, and Prevention. for diagnosis of coronary artery disease. Ann Noninvasive Electrocardiol.
Circulation. 2005;112:771–776. 1996;1:430–442.
Downloaded from http://ahajournals.org by on January 21, 2020

DOI: 10.1161/JAHA.119.014540 LJournal of the American Heart Association 10

Supplemental Material
Downloaded from http://ahajournals.org by on January 21, 2020
 Data S1.

Heart Trends Device

The regulatory status: The previous software version was cleared for marketing in the USA

under 510k number K012825. The cleared intended use was for the analysis, summary and

reporting of up to 3 channels of prerecorded ambulatory ECG data. It was also intended to

provide measurements of MPW (Multipole Parameter Weighted) HRV.

Cleared Device description: The HeartTrends® software was employed as a measuring tool to

present Heart Rate Variability (HRV) to qualified clinician review, edit and assessment. It

provides measurements of the MPW HRV. Heart rate variability is a known method for

analyzing the changes in heart rate over the recorded duration. Subjects with a low variability

were suggested to be at increased risk for cardiac events. The HeartTrends device is based on a

novel algorithm for analyzing heart rate variability, which is constructed from the Multipole
Downloaded from http://ahajournals.org by on January 21, 2020

method, based on a physical-mathematical description of complex time series. The Multipole

method generates several parameters, multipoles, where every single one describes the HRV.

The HeartTrends device received CE mark (0344) on 17 July 2013 and valid through 2024, for

the product category: Software based medical devices for diagnostics-aid of ischemic heart

diseases using Heart Rate Variability analysis. The DyDx indicator value can be used as a

prognostic score to assist in diagnosis of coronary artery disease for which the physician renders

their own opinion. HeartTrends does not offer a diagnostic opinion to the user. HeartTrends is

intended to be used by qualified personnel in evaluating the subject in conjunction with the

subject's clinical history, symptoms, other diagnostic tests, as well as the professional's clinical

judgment.
 The changes between the FDA-cleared software version and the version to be tested in this study

were minimal; the base algorithm did not change, changes were made in the GUI dividing the

software into a Client-Server base application, and adaptation to more recent operating systems.

Additionally, there was a change in the intended use of the HeartTrends device, while used

before for ECG data recording, storing and analysis, it is now investigated for the diagnosis of

ischemic heart disease:

The software underwent full software validation and verification that complies with international

regulatory requirements.
Downloaded from http://ahajournals.org by on January 21, 2020
 Table S1. Inclusion and exclusion criteria for study participation.
Inclusion Criteria:
• Age ≥ 21
• Referral for EST in a subjects without known CAD
due to either one of the following two indications:
1. Chest pain syndrome or equivocal angina in
subjects with low to intermediate pretest
probability of CAD
2. At least one CAD risk factor (diabetes mellitus,
hypertension, smoking, positive family history,
and/or dyslipidemia) in asymptomatic subjects
referred for cardiovascular risk assessment.
• Willing and able to provide written informed consent

Exclusion Criteria:
• Acute Coronary Syndrome
• Established CAD
Downloaded from http://ahajournals.org by on January 21, 2020

• Atrial fibrillation or flutter

• Cardiac Pacemaker
• Clinical diagnosis of heart failure
• Severe COPD (FEV1< 50% predicted value)
• Active myocarditis, constrictive pericarditis, any
cardiomyopathy, cardiac or systemic amyloidosis
• Known drug or alcohol dependence or any other
factors which will interfere with the study conduct or
interpretation of the results or in the opinion of the
investigator are not suitable to participate;
• Any illness that might reduce life expectancy to less
than 1 year from screening
• Left bundle branch block (LBBB), significant intra-
ventricular conduction delay (IVCD) or significant
(>1mm) ST deviations on baseline ECG
 • Inability to perform an exercise stress test (i.e.
orthopedic or neurological limitations)
• Any significant valvular disease defined as:
Established valvular regurgitation or stenosis abnormality
above moderate severity
• BMI >35 kg/m2
• Recent (< 6 months) history of pulmonary embolism
Downloaded from http://ahajournals.org by on January 21, 2020
 Figure S1. Bland-Altman analysis of the Holter ECG comparing HeartTrends results from
1-hour recordings with data derived from 20 minute recordings.*

*The graph of tracks the P-Value over 10-minute time intervals. Results of the analysis also
yielded statistically significant P-Values of: 0.000297, 0.000303, 0.000284, 0.000248, 0.000127
for all the 10 to 50-minute intervals analyzed.
Downloaded from http://ahajournals.org by on January 21, 2020