PR O G RE S S I N C ARDI O V A S CU L A R D I S EA S E S 5 6 (2 0 1 3) 15 3–1 5 9
Available online at www.sciencedirect.com
ScienceDirect www.onlinepcd.com
Heart Rate Variability in Risk Stratification of Cardiac Patients Heikki V. Huikuria,⁎, Phyllis K. Steinb a
Institute of Clinical Medicine, Department of Internal Medicine, University of Oulu, and University Hospital of Oulu, Oulu, Finland Division of Cardiology, Washington University School of Medicine, St. Louis, MO, USA
b
A R T I C LE I N F O
AB ST R A C T
Keywords:
Heart rate (HR) variability has been extensively studied in cardiac patients, especially in
Mortality
patients surviving an acute myocardial infarction (AMI) and also in patients with
Myocardial infarction
congestive heart failure (CHF) or left ventricular (LV) dysfunction. The majority of studies
Heart failure
have shown that patients with reduced or abnormal HR variability have an increased risk of mortality within a few years after an AMI or after a diagnosis of CHF/LV dysfunction. Various measures of HR dynamics, such as time-domain, spectral, and non-linear measures of HR variability have been used in risk stratification. The prognostic power of various measures, except of those reflecting rapid R–R interval oscillations, has been almost identical, albeit some non-linear HR variability measures, such as short-term fractal scaling exponent have provided somewhat better prognostic information than the others. Abnormal HR variability predicts both sudden and non-sudden cardiac death. Because of remodeling of the arrhythmia substrate after AMI, early measurement of HR variability to identify those at high risk should likely be repeated later in order to assess the risk of fatal arrhythmia events. Future randomized trials using HR variability/ turbulence as one of the pre-defined inclusion criteria will show whether routine measurement of HR variability/turbulence will become a routine clinical tool for risk stratification of cardiac patients. © 2013 Elsevier Inc. All rights reserved.
The purpose of this review will be to identify cardiac patients in whom heart rate (HR) variability, by itself or in combination with other measures, has potential or proven clinical utility. Studies will be reviewed that employ a large variety of HR variability measures, with the majority derived from easily-obtainable outpatient 24-h recordings. Other sources of HR variability data include short recordings that could potentially be performed in an office setting. Clinical utility is, of course, a broad concept. In this review, we will consider a measure to have clinical utility if it identifies patients at higher risk of an adverse outcome, differentiates those with
more advanced from those with less advanced disease, identifies those who would benefit from interventions or identifies those who are benefitting from or being harmed by an intervention. HR variability has been applied in a huge number of studies. Currently, a PubMed search on the term results in over 17,000 citations, well beyond the possible scope of any review. Therefore, we have limited our review to applications of HR variability in populations with a prior myocardial infarction (AMI), congestive heart failure (CHF) or left ventricular (LV) dysfunction.
Statement of Conflict of Interest: see page 158. ⁎ Address reprint requests to Heikki Huikuri, MD, Institute of Clinical Medicine, Department of Internal Medicine, University of Oulu, P.O. Box 5000 (Kajaanintie 50), FIN-90014 University of Oulu, Finland. E-mail address: [email protected] (H.V. Huikuri). 0033-0620/$ – see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.pcad.2013.07.003
154
PR O GRE S S I N C ARDI O VAS CU L AR D I S EAS E S 5 6 ( 2 0 13 ) 15 3–1 59
Abbreviations and Acronyms AMI = acute myocardial infarction ATRAMI = Autonomic Tone and Reflexes After Myocardial Infarction
What makes HRV normal or abnormal?
The term HR variability encompasses a large number of measures of different types. Time BHAT = Beta-Blocker Heart domain measures reAttack Trial flect “how much” HR CABG = coronary artery bypass variability there is. In grafting general, if such measures are extremely CARISMA = Cardiac Arrhythmias low, it can be assumed and RIsk Stratification after that there is true autoMyocardial infArction nomic dysfunction, but CHF = congestive heart failure higher values for these various measures could DEFINITE = Defibrillators in reflect more healthy Non-Ischemic Cardiomyopathy Treatment Evaluation autonomic function or an unhealthy, highly DIAMOND = Danish Investigairregular heart rate pattions of Arrhythmia and tern (erratic rhythm). Mortality ON Dofetilide Frequency domain meaDYNAMIT = Defibrillator in sures which quantify Acute Myocardial Infarction Trial the amount of variance in HR at different unHF = high frequency derlying frequencies HR = heart rate have the same properties. Extremely low ICD = implantable cardioverter values are associated defibrillator with a lack of autoLF = low frequency nomic modulation of HR, but higher values, LV = left ventricular without examination LVEF = left ventricular ejection of their underlying orfraction ganization using power spectral plots or other MI = myocardial infarction graphical methods, canMPIP = Multicenter Postnot be assumed to Infarction Project reflect better HR variREFINE = Risk Estimation After ability. Moreover, higher Infarction, Noninvasive values for certain freEvaluation quency domain HR variability measures can SDNN = standard deviation of be caused by some unN–N intervals derlying disorders, e.g. sleep apnea. The use of ratio measures of HR variability like the low frequency/high frequency (LF/HF) ratio confounds the problem, because if either LF or HF is capturing an abnormally organized HRV pattern, the ratio will be meaningless as a surrogate for autonomic function. Non-linear measures like the short-term fractal scaling exponent, which reflects the fractal characteristics of rapid R–R interval oscillations measured by the detrended fluctuation analysis method, and the power law slope, appear to better differentiate between healthy and unhealthy organization of the heart’s rhythm and have proved to be more sensitive for risk stratification than some of the standard measures. HR turbulence, which is covered in detail
in another review, also appears to be sensitive to autonomic dysfunction. It is likely that the optimal characterization of normal or abnormal HR variability would be based on a combination of these different HR variability parameters rather than on a single one.
Heart rate variability after myocardial infarction The prognostic significance of HR variability has been extensively studied in patients who have survived an AMI. Table 1 summarizes the largest studies in this field. In 1978 Wolf et al. showed that patients with a low magnitude of short-term HR variability (no sinus arrhythmia on 30 consecutive RR intervals on admitting ECG) had a poor prognosis after AMI.1 In the late 1980s, the Multicenter Post-Infarction Project (MPIP) confirmed the predictive value of reduced HRV by measuring 24-h standard deviation of N–N intervals (SDNN) over 24 h following AMI.2 Patient with an SDNN 100 ms. Adjustment for covariates did not explain this association, although relative risk declined to 2.8. Other studies (including a reanalysis of MPIP) have shown that spectral measures of HR variability, mainly the ultra-low, very-low, and low frequency spectral components are reduced in survivors of AMI and that decreased values are associated with an increased risk of all-cause mortality.3 Because accurate Holter scanning is relatively labor intensive, the group at St. George's hospital in London experimented with global, geometric indices of HR variability (HRV index) that were derived from the histogram of NN intervals. Decreased values for these measures were also found to be associated with higher risk of mortality.4 In addition, this group suggested that markedly decreased SDNN for a single, stable 5-min period could pre-select patients in whom 24-h Holter recordings using geometric methods would provide significant risk stratification.5 The ATRAMI (Autonomic Tone and Reflexes After Myocardial Infarction) trial was a multicenter observational study performed about 10 years after MPIP. The ATRAMI investigators confirmed that reduced 24-h SDNN (< 70 ms in this case) is associated with a relative risk of 3.2 for mortality during 21 months of follow up.6 Furthermore, the combination of low SDNN and LVEF < 35% carried a relative risk of 6.7 for mortality when compared with patients with both higher SDNN and LVEF. A more recent study, however, in the era of modern therapy showed a very low incidence of severely depressed HR variability (10 % for SDNN < 50 ms). All patients in the study were treated with direct percutaneous coronary intervention within 12 h of their event. Although 4-year survival was 80% in this low SDNN group compared to 92% among those with higher SDNN, positive predictive value was low.7 The majority of early studies focused on measurement of HR variability relatively early after AMI.1–6 Measures of HR variability appeared to be more depressed at the early phase of AMI with substantial improvement during recovery.8 Despite this autonomic recovery, measurement of HR
Table 1 – Summary of main studies assessing the prognostic significance of heart rate variability after acute myocardial infarction. Study
HRV Measurement
Number of Patients
Mean Follow-Up Time
Primary Endpoint
SDNN
808
34 months
all-cause mortality
Bigger JT Jr et al., 1992
Frequency domain measures
715
2.5 years
all-cause death, cardiac death, arrhythmic death
Bigger JT Jr et al., 1993
Frequency domain measures
331
3 years
all-cause mortality
Fei L et al., 1996
SDNN and HRV index
700
1 year
all-cause mortality
Bigger JT Jr et al., 1996
Power–law behavior
715
3 years
all-cause mortality
La Rovere MT et al., 1998
SDNN
21 months
Huikuri HV et al., 2000
Spectral, fractal and time domain measures
Cardiac death and non-fatal cardiac arrest all-cause mortality
Mäkikallio TH et al., 2005
1284 446
685 days
Spectral, fractal and, time-domain measures
2130
1012 days
Sudden cardiac death, non-sudden cardiac death
Stein PK et al., 2005
Spectral, fractal, and time-domain measures
740
362 days
all-cause mortality
Perkiömäki JS et al., 2006
Spectral, fractal and time-domain measures Spectral, fractal and time-domain measures SDNN
675
30 months
Non-fatal coronary events
312
2 years
ECG documental VT or VF
412
4.3 years
all-cause mortality
Huikuri HV et al., 2009 Erdogan et al., 2008
Increased mortality, if SDNN < 50 ms All frequency domain measures of HRV predicted an increased risk of primary endpoint All frequency domain measures predicted the risk of primary endpoint Geometric HRV index predicted mortality better than SDNN Power–law regression between log(power) and log(frequency) predicted mortality SDNN < 70 ms predicted increased cardiac mortality Short-term fractal scaling exponent was the most powerful predictor of mortality All HRV measures predicted sudden and non-sudden cardiac death Fractal measures predicted all-cause mortality Several measures of HRV predicted primary endpoint Many HRV measures predicted the primary endpoint SDNN < 50 ms predicted mortality
Comment Landmark study VLF spectral power was the strongest predictor Holter recording were performed one year after AMI
Power–law regression analyzed from ULF and VLF spectral areas Multicenter study Post-AMI patients with depressed ejection fraction HRV predicted equally both sudden and non-sudden cardiac death. Largest HRV study Short-term fractal scaling exponent was more powerful predictor than other HRV measures
VLF power was the strongest predictor positive predictive power of SDNN was low
PR O G RE S S I N C ARDI O V A S CU L A R D I S EA S E S 5 6 (2 0 1 3) 15 3–1 5 9
Kleiger RE et al., 1987
Main Results
Abbreviations: AMI = acute myocardial infarction, ECG = electrocardiogram, HRV = heart rate variability, SDNN = standard deviation of N–N intervals, ULF = ultra-low frequency, VF = ventricular fibrillation, VLF = very-low frequency, VT = ventricular tachycardia.
155
156
PR O GRE S S I N C ARDI O VAS CU L AR D I S EAS E S 5 6 ( 2 0 13 ) 15 3–1 59
variability has been shown to yield similar prognostic information when the 24-h HR variability measurements were performed one year after AMI or in middle-aged women 3–6 months after hospitalization for acute coronary syndrome.9,10 However, more recently, perhaps due to the increased use of beta-blockers, a study of 600 patients 5–7 days post-MI and a follow up study of 400 of them one year after MI showed that while 24-h time or frequency domain HRV measures improved after MI, none of them predicted cardiac mortality at either time point.8 At the same the, short-term fractal scaling exponent, the power law slope and HR turbulence values remained stable during recovery and predicted cardiac death at both time points.8 Another factor to be considered is that at the time of the earliest studies of the predictive value of HRV for mortality post-MI, e.g. MPIP, post-MI mortality was about 30% on follow up, whereas currently it is far lower (< 4%), necessitating larger trials to test any putative risk marker. The majority of the post-MI studies performed in the 1980's and 1990's used 24-h time- and frequency domain as well as geometric measures of HR variability for risk stratification of patients after AMI. Although all statistical, geometrical, and spectral measures of HR variability differ in their manner of computation and analysis, these methods are fundamentally based on moment statistics and describe the magnitude of HR variability or of its underlying components. It is not therefore surprising that most of the measures that have been shown to have prognostic value have relatively close mutual correlation, and that there are only minor differences in prognostic power between them. It must be kept in mind though that measures of beat-to-beat HR variability which are supposed to reflect parasympathetic control of heart rate (HF power, pNN50, rMSSD) have rarely shown a strong association with outcome. This is likely because of the high prevalence of erratic rhythm in these populations resulting in higher measures for these parameters that do not reflect better parasympathetic function. More recent studies have applied methods based in nonlinear analysis methods, which provide very different and perhaps complementary information on HR dynamics compared to traditional statistical methods.11 Methods based on nonlinear dynamics have provided better prognostic information than that obtained by using more traditional methods. Analysis of l/f characteristics, i.e. the inverse power– law slope, which describes the slope of the spectral powers in the ultra-low and very-low frequency areas has provided prognostic information beyond the traditional HR variability measures.12 Decreased short-term fractal scaling exponent measured by the detrended fluctuation analysis method, a measure of greater randomness in short-term R–R interval fluctuations, has turned out to be a powerful non-linear index in risk stratification, a more powerful prognostic tool than traditional HR variability indexes in various post-AMI populations. In the DIAMOND-MI (Danish Investigations of Arrhythmia and Mortality ON Dofetilide) trial, short-term fractal exponent < 0.75 identified post-AMI patients at a relative risk of 3.0 for mortality and was more strongly related to outcome than traditional time and frequency domain measures.12 Moreover, reduced short-term fractal exponent
was related to both arrhythmic and non-arrhythmic mortality. In another study of consecutive patients with acute MI, both short-term fractal exponent < 1.025 (hazard ratio = 2.0) and power law slope ≤ − 1.507 (hazard ratio = 1.9) were independently associated with recurrent non-fatal coronary events.13 Another measure of increased randomness of heart rate patterns, measured as the ratio of the axes of a Poincare plot fitted to the plot of each N–N interval vs. the next, also predicted mortality in the Cardiac Arrhythmia Suppression Trial.14 Patients in this trial were at variable times post-MI but were selected for having significant ventricular arrhythmias. The majority of the studies assessing the prognostic power of HR variability after AMI have used all-cause mortality as the end-point. Some studies, using various definitions of arrhythmic events in their designs, have suggested that reduced or altered HR variability may be specifically related to arrhythmic events and sudden cardiac death. However, the clinical applicability of measurement of HR variability as a predictor of fatal arrhythmic events was challenged in a randomized prophylactic implantable cardioverter defibrillator (ICD) trial, the DYNAMIT (Defibrillator in Acute Myocardial Infarction Trial) which used reduced HR variability combined with reduced left ventricular ejection fraction measured early (within two weeks) after AMI as an inclusion criterion for the trial, but could not show any mortality benefit from ICD therapy in these presumably high-risk patients.15 The more recent multi-center CARISMA (Cardiac Arrhythmias and RIsk Stratification after Myocardial infArction) study showed that reduced HR variability measured relatively late (six weeks after AMI) rather than early after AMI, especially the very-low frequency spectral component, was the most powerful index among many other non-invasive risk markers in predicting the ventricular fibrillation or sustained ventricular tachycardia diagnosed by implantable arrhythmia devices.16 A further sub analysis of the CARISMA and another recent similar study (REFINE; Risk Estimation After Infarction, Noninvasive Evaluation) study confirmed that HR variability and HR turbulence yield more powerful prognostic information for arrhythmic events when measured late (6–10 weeks after AMI) rather than early (within two weeks) after AMI in the current treatment era.17 Thus, lack of recovery from reduced autonomic function after AMI was a stronger predictor of adverse events than decreased autonomic function at the early phase after AMI. The results from these two studies suggest that the measurement of HR variability for prognostic purposes should be performed late after AMI (> 6 weeks after AMI) rather than at the early phase after the acute event. Observational studies using various end-points have shown that reduced HR variability also predicts non-sudden cardiac death. The largest study in this field showed that reduced HR variability/turbulence is in fact a stronger predictor of non-sudden than sudden cardiac death.18 In this analysis, HR variability/turbulence, measured early after the AMI, was a better predictor of mortality among the patients with left ventricular ejection fraction (LVEF > 40%) than in those with reduced LVEF. Often forgotten in this focus on risk stratification, is the negative predictive value of HR variability measures. Patients with normal HR variability measures, even if they are post-
PR O G RE S S I N C ARDI O V A S CU L A R D I S EA S E S 5 6 (2 0 1 3) 15 3–1 5 9
AMI, are at very low risk of mortality. However, as has been found with other non-invasive risk markers, the positive predictive accuracy and sensitivity of abnormal HR variability for adverse outcomes have remained low in most observational studies. The sensitivity, specificity and predictive values of HR variability in predicting future events depend partly on defining the cutoff points of abnormal HR variability measures. The cutoff points vary between the studies and depend on the study populations, such as post-AMI patients with and without CHF, usage of beta-blocking medication etc., which influence the HR variability measures, and there is no universal consensus on an algorithm for determining the ideal cutoffs for different HR variability measures in different populations. Age and presence of diabetes and even clinical depression are factors influencing HR variability and should be considered in defining the cutoff values. HR variability has been combined with other risk markers in order to improve the risk stratification in some studies.6 The problem with combining various risk markers, however, is the loss of sensitivity despite an improved positive predictive accuracy. Little is known about the effect of interventions on HR variability in post-MI patients and about whether interventions that are associated with improvements in HR variability are associated with better outcomes or vice versa. One intervention, coronary artery bypass surgery (CABG) is associated with markedly decreased HR variability that slowly improves over weeks and months. It is likely that any predictive value for long-term outcomes contained in traditional time and frequency domain HR variability is lost once the patient has undergone CABG surgery.19 However, two studies have shown that measurements of non-linear HR variability parameters after CABG surgery did identify those at higher risk of a complicated post-surgical course.20,21 Only one trial, the BHAT (Beta-Blocker Heart Attack Trial) examined change in HR variability with therapy and the subsequent relationship of HR variability to mortality.22 In BHAT, patients randomized to propranolol had a significantly greater increase in HF power at 6-weeks and after 21 months of follow up. In the same study, higher HR variability at 6 weeks and assignment to the beta-blocker arm were independent predictors of a combined endpoint of mortality and non-fatal cardiac events. In a meta-analysis of the effect of pharmacologic, behavioral and exercise strategies on HR variability, Nolan et al.23 concluded that these secondary prevention strategies have a moderate but significant positive effect on HR variability. Considered together, the available data show that abnormal HR variability, measured by various time and frequencydomain measures, and non-linear measures from 24-h R–R interval recordings are general risk markers for common modes of cardiac death: arrhythmic, vascular, and hemodynamic after AMI. HR variability measured early after AMI seems to provide more powerful information on the risk of early nonsudden cardiac death, especially due to progressive heart failure. Because of remodeling of the arrhythmia substrate and autonomic function after AMI, early measurement of HR variability to identify those at high risk of sudden cardiac death should likely be repeated later (6–8 weeks) after AMI.
157
Future research in this era should focus on randomized trials that take into consideration the timing of HR variability/ turbulence measurement in their study designs. Information from such trials may identify the populations in which the clinical applicability of routine measurement of HR variability is warranted, the measurements most strong predictive of outcomes in these populations, and the optimal timing of such measurements. Furthermore, the results from REFINE and CARISMA studies suggest that measurement of the evolution of HR variability over time may have greater prognostic value. Measurement of the dynamic changes HR variability over time in association with prescribed therapies in various populations, perhaps on an individual basis, may also add to the clinic applications of HR variability measurements.
Heart rate variability in heart failure/left ventricular dysfunction Traditional analysis of HR variability has shown that its magnitude is generally reduced in CHF patients, and the detrended fluctuation analysis method has shown altered fractal characteristics of HR dynamics in CHF. The UK-Heart study focused on the prognostic significance of HR variability in outpatients with NYHA class I–III CHF and showed that reduced SDNN predicts mortality.24 Annual mortality rates for the study population were 5.5% for SDNN > 100 ms, 12.7% for SDNN = 50 to 100 ms and 51.4% for SDNN < 50 ms. Increased mortality was mainly due to progressive heart failure rather than sudden death among the CHF patients with reduced SDNN. However, in the DIAMOND-CHF study, which enrolled patients admitted to the coronary care unit with new or worsening CHF, only the reduced short-term fractal scaling exponent was independently associated with mortality in a population that was followed for a mean of 1.8 years and had a 42% mortality.25 Furthermore, the prognostic value of HR variability in this population was stronger in class II than in class III and IV patients. Several studies including primarily post-AMI patients with depressed LVEF have yielded quite similar results, i.e., reduced HR variability and altered short-term fractal HR behavior predict cardiac and all-cause mortality.11 Studies including patients with non-ischemic cardiomyopathy have also generally yielded similar results. In the DEFINITE (Defibrillators in Non-Ischemic Cardiomyopathy Treatment Evaluation) trial, SDNN was evaluated by tertiles.26 There were no deaths during 3-year follow up among those with SDNN > 113 ms (highest tertile). Among those with SDNN 81–113 ms 7% died and among those with SDNN < 81 ms 10% died. Among excluded patients (atrial fibrillation or > 25% ectopic beats) 17% died. The CARISMA and REFINE studies mentioned above included post-AMI patients with depressed LVEF (< 40%) after AMI and showed that HR variability/ turbulence measured late, but not early after AMI, predict fatal or near fatal arrhythmic events in these patients.16 The DINAMIT study also showed that HR variability measured early after AMI among the patients with LV dysfunction does not have a significant role in selecting the patients for ICD
158
PR O GRE S S I N C ARDI O VAS CU L AR D I S EAS E S 5 6 ( 2 0 13 ) 15 3–1 59
therapy.15 Similar to post-AMI patients, the clinical applicability of measurement of HR variability in patients with CHF remains uncertain because despite some evidence that an intervention like exercise training27 can improve at least parasympathetic measures of HR variability, there is a lack of randomized trials showing that the prognosis of patients with impaired HR variability can be improved by medical or device intervention. The exception might be the case of cardiac resynchronization therapy (CRT). Among 509 patients who received a CRT-D device for primary or secondary prevention of sudden death or who received a CRT pacemaker, use of CRT was associated with a significant increase in SDANN monitored by the device.28 SDANN is a measure of circadian activity, and an increase in SDANN could reflect an increase in functional status due to the device. However, SDANN ≤ 65 ms, at baseline or SDANN ≤ 76 ms 4 weeks after implantation were the strongest predictors of need for transplant or mortality, and there was a significant negative correlation between the reduction in LV end systolic volume with CRT and SDANN at baseline and at week 4. Thus, HR variability measured after implantation might identify CRT patients who are less likely to benefit and who remain at higher risk for adverse events.
Summary and recommendations Future research in this era should focus on randomized trials that take into consideration the timing of HR variability/ turbulence measurement in their study designs. Information from such trials may identify the populations in which the clinical applicability of routine measurement of HR variability is warranted, the measurements most strongly predictive of outcomes in these populations, and the optimal timing of such measurements. Furthermore, the results from REFINE and CARISMA studies suggest that measurement of the evolution of HR variability over time may have greater prognostic value. Measurement of the dynamic changes HR variability over time in association with prescribed therapies in various populations, perhaps on an individual basis, may also add to the clinic applications of HR variability measurements. Before the completion of such trials, patients with low HR variability/turbulence after AMI should receive the best available medical and invasive therapy, rehabilitation, and follow-up, even if routine measurements of HR variability/turbulence cannot yet be recommended in clinical practice guidelines.
Statement of Conflict of Interest All authors declare that there are no conflicts of Interest.
Acknowledgments Funded in part by the Sigrid Juselius Foundation, and the Finnish Foundation for Cardiovascular Research, Helsinki, Finland.
REFERENCES
1. Wolf MM, Vargios GA, Hunt D, Sloman JG. Sinus arrhythmia in acute myocardial infarction. Med J Aust. 1978;2:52–3. 2. Kleiger RE, Miller JP, Bigger JT, Moss AJ, the Multicenter Postinfarction Research Group. Decreased heart rate variability and its association with increased mortality after myocardial infarction. Am J Cardiol. 1987;59:256–62. 3. Bigger Jr 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;85:164–71. 4. Malik M, Farrell T, Cripps T, Camm AJ. Heart rate variability in relation to prognosis after myocardial infarction: selection of optimal processing techniques. Eur Heart J. 1989;10:1060–74. 5. Fei L, Copie X, Malik M, Camm AJ. Short- and long-term assessment of heart rate variability for risk stratification after acute myocardial infarction. Am J Cardiol. 1996;77:681–4. 6. La Rovere MT, Bigger Jr JT, Marcus FI, et al. Baroreflex sensitivity and heart-rate variability in prediction of total cardiac mortality after myocardial infarction. Lancet. 1998;351: 478–84. 7. Erdogan A, Coch M, Bilgin M, et al. Prognostic value of heart rate variability after acute myocardial infarction in the era of immediate reperfusion. Herzschrittmacherther Elektrophysiol. 2008;19:161–8. [Epub 2009 Feb 11]. 8. Jokinen V, Tapanainen JM, Seppänen T, Huikuri HV. Temporal changes and prognostic significance of measures of heart rate dynamics after acute myocardial infarction in the betablocking era. Am J Cardiol. 2003;92:907–12. 9. Bigger Jr JT, Fleiss JL, Rolnitzky LM, Steinman RC. Frequency domain measures of heart period variability to assess risk of late myocardial infarction. J Am Coll Cardiol. 1993;21: 729–36. 10. Janszky I, Ericson M, Mittleman MA, et al. Heart rate variability in long-term risk assessment in middle-aged women with coronary heart disease: the Stockholm Female Coronary Risk Study. J Intern Med. 2004;255:13–21. 11. Mäkikallio TH, Tapanainen JM, Tulppo MP, Huikuri HV. Clinical applicability of heart rate variability analysis by methods based on nonlinear dynamics. Cardiac Electrophysiol Rev. 2002;6:250–5. 12. Huikuri HV, Mäkikallio TH, Peng C-K, et al. Fractal correlation properties of R–R interval dynamics and mortality in patients with depressed left ventricular function after an acute myocardial infarction. Circulation. 2000;101:47–53. 13. Perkiömäki JS, Jokinen V, Tapanainen J, Airaksinen KE, Huikuri HV. Autonomic markers as predictors of nonfatal acute coronary events after myocardial infarction. Ann Noninvasive Electrocardiol. 2008;13:120–9. 14. Stein PK, Domitrovich PP, Huikuri HV, Kleiger RE, Cast Investigators. Traditional and nonlinear heart rate variability are each independently associated with mortality after myocardial infarction. J Cardiovasc Electrophysiol. 2005;16(1): 13–20. 15. Hohnloser SH, Kuck KH, Dorian P, et al. Prophylactic use of an implantable cardioverter–defibrillator after acute myocardial infarction. N Engl J Med. 2004;351:2481–8. 16. Huikuri HV, Raatikainen MJP, Moerch-Joergensen R, et al. Prediction of fatal or near fatal cardiac arrhythmia events in patients with depressed left ventricular function after an acute myocardial infarction. Results of the Cardiac Arrhythmias and Risk Stratification after Acute Myocardial Infarction (CARISMA) Study. Eur Heart J. 2009;30:689–98. 17. Huikuri HV, Exner DV, Kavanagh KM, et al. Attenuated recovery of heart rate turbulence early after myocardial
PR O G RE S S I N C ARDI O V A S CU L A R D I S EA S E S 5 6 (2 0 1 3) 15 3–1 5 9
18.
19.
20.
21.
22.
23.
infarction identifies patients at high risk of fatal or near-fatal arrhythmic events. Heart Rhythm. 2010;7:229–35. Mäkikallio TH, Barthel P, Schneider R, et al. Prediction of sudden cardiac death after acute myocardial infarction; role of Holter monitoring in the modern treatment era. Eur Heart J. 2005;26:762–9. Stein PK, Domitrovich PP, Kleiger RE, CAST Investigators. Including patients with diabetes mellitus or coronary artery bypass grafting decreases the association between heart rate variability and mortality after myocardial infarction. Am Heart J. 2004;147:309–16. Laitio T, Huikuri H, Kentala E, et al. Correlation properties and complexity of perioperative RR-interval dynamics in coronary artery bypass surgery patients. Anesthesiology. 2000;93:69–80. de Godoy MF, Takakura IT, Correa PR, Machado MN, Miranda RC, Brandi AC. Preoperative nonlinear behavior in heart rate variability predicts morbidity and mortality after coronary artery bypass graft surgery. Med Sci Monit. 2009;15:CR117–22. Lampert R, Ickovics JR, Viscoli CJ, Horwitz RI, Lee FA. Effects of propranolol on recovery of heart rate variability following acute myocardial infarction and relation to outcome in the Beta-Blocker Heart Attack Trial. Am J Cardiol. 2003;91:137–42. Nolan RP, Jong P, Barry-Bianchi SM, Tanaka TH, Floras JS. Effects of drug, biobehavioral and exercise therapies on heart
24.
25.
26.
27.
28.
159
rate variability in coronary artery disease: a systematic review. Eur J Cardiovasc Prev Rehabil. 2008;15:386–9. Nolan J, Batin PD, Andrews R, et al. Prospective study of heart rate variability and mortality in chronic heart failure: results of the United Kingdom Heart Failure Evaluation and Assessment of Risk Trial (UK-Heart). Circulation. 1999;98:1510–6. Mäkikallio TH, Huikuri HV, Hintze U, et al. Fractal analysis and time- and frequency-domain measures of heart rate variability as predictors of mortality in patients with heart failure. Am J Cardiol. 2001;87:178–82. Rashba EJ, Estes NA, Wang P, et al. Preserved heart rate variability identifies low-risk patients with nonischemic dilated cardiomyopathy: results from the DEFINITE trial. Heart Rhythm. 2006;3:281–6. Piotrowicz E, Baranowski R, Piotrowska M, Zieliński T, Piotrowicz R. Variable effects of physical training of heart rate variability, heart rate recovery, and heart rate turbulence in chronic heart failure. Pacing Clin Electrophysiol. 2009;32(Suppl 1):S113–5. Landolina M, Gasparini M, Lunati M, et al. Heart rate variability monitored by the implanted device predicts response to CRT and long-term clinical outcome in patients with advanced heart failure. Eur J Heart Fail. 2008;10:1073–9. [Epub 2008 Oct 1].
Available online at www.sciencedirect.com
ScienceDirect www.onlinepcd.com
Heart Rate Variability in Risk Stratification of Cardiac Patients Heikki V. Huikuria,⁎, Phyllis K. Steinb a
Institute of Clinical Medicine, Department of Internal Medicine, University of Oulu, and University Hospital of Oulu, Oulu, Finland Division of Cardiology, Washington University School of Medicine, St. Louis, MO, USA
b
A R T I C LE I N F O
AB ST R A C T
Keywords:
Heart rate (HR) variability has been extensively studied in cardiac patients, especially in
Mortality
patients surviving an acute myocardial infarction (AMI) and also in patients with
Myocardial infarction
congestive heart failure (CHF) or left ventricular (LV) dysfunction. The majority of studies
Heart failure
have shown that patients with reduced or abnormal HR variability have an increased risk of mortality within a few years after an AMI or after a diagnosis of CHF/LV dysfunction. Various measures of HR dynamics, such as time-domain, spectral, and non-linear measures of HR variability have been used in risk stratification. The prognostic power of various measures, except of those reflecting rapid R–R interval oscillations, has been almost identical, albeit some non-linear HR variability measures, such as short-term fractal scaling exponent have provided somewhat better prognostic information than the others. Abnormal HR variability predicts both sudden and non-sudden cardiac death. Because of remodeling of the arrhythmia substrate after AMI, early measurement of HR variability to identify those at high risk should likely be repeated later in order to assess the risk of fatal arrhythmia events. Future randomized trials using HR variability/ turbulence as one of the pre-defined inclusion criteria will show whether routine measurement of HR variability/turbulence will become a routine clinical tool for risk stratification of cardiac patients. © 2013 Elsevier Inc. All rights reserved.
The purpose of this review will be to identify cardiac patients in whom heart rate (HR) variability, by itself or in combination with other measures, has potential or proven clinical utility. Studies will be reviewed that employ a large variety of HR variability measures, with the majority derived from easily-obtainable outpatient 24-h recordings. Other sources of HR variability data include short recordings that could potentially be performed in an office setting. Clinical utility is, of course, a broad concept. In this review, we will consider a measure to have clinical utility if it identifies patients at higher risk of an adverse outcome, differentiates those with
more advanced from those with less advanced disease, identifies those who would benefit from interventions or identifies those who are benefitting from or being harmed by an intervention. HR variability has been applied in a huge number of studies. Currently, a PubMed search on the term results in over 17,000 citations, well beyond the possible scope of any review. Therefore, we have limited our review to applications of HR variability in populations with a prior myocardial infarction (AMI), congestive heart failure (CHF) or left ventricular (LV) dysfunction.
Statement of Conflict of Interest: see page 158. ⁎ Address reprint requests to Heikki Huikuri, MD, Institute of Clinical Medicine, Department of Internal Medicine, University of Oulu, P.O. Box 5000 (Kajaanintie 50), FIN-90014 University of Oulu, Finland. E-mail address: [email protected] (H.V. Huikuri). 0033-0620/$ – see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.pcad.2013.07.003
154
PR O GRE S S I N C ARDI O VAS CU L AR D I S EAS E S 5 6 ( 2 0 13 ) 15 3–1 59
Abbreviations and Acronyms AMI = acute myocardial infarction ATRAMI = Autonomic Tone and Reflexes After Myocardial Infarction
What makes HRV normal or abnormal?
The term HR variability encompasses a large number of measures of different types. Time BHAT = Beta-Blocker Heart domain measures reAttack Trial flect “how much” HR CABG = coronary artery bypass variability there is. In grafting general, if such measures are extremely CARISMA = Cardiac Arrhythmias low, it can be assumed and RIsk Stratification after that there is true autoMyocardial infArction nomic dysfunction, but CHF = congestive heart failure higher values for these various measures could DEFINITE = Defibrillators in reflect more healthy Non-Ischemic Cardiomyopathy Treatment Evaluation autonomic function or an unhealthy, highly DIAMOND = Danish Investigairregular heart rate pattions of Arrhythmia and tern (erratic rhythm). Mortality ON Dofetilide Frequency domain meaDYNAMIT = Defibrillator in sures which quantify Acute Myocardial Infarction Trial the amount of variance in HR at different unHF = high frequency derlying frequencies HR = heart rate have the same properties. Extremely low ICD = implantable cardioverter values are associated defibrillator with a lack of autoLF = low frequency nomic modulation of HR, but higher values, LV = left ventricular without examination LVEF = left ventricular ejection of their underlying orfraction ganization using power spectral plots or other MI = myocardial infarction graphical methods, canMPIP = Multicenter Postnot be assumed to Infarction Project reflect better HR variREFINE = Risk Estimation After ability. Moreover, higher Infarction, Noninvasive values for certain freEvaluation quency domain HR variability measures can SDNN = standard deviation of be caused by some unN–N intervals derlying disorders, e.g. sleep apnea. The use of ratio measures of HR variability like the low frequency/high frequency (LF/HF) ratio confounds the problem, because if either LF or HF is capturing an abnormally organized HRV pattern, the ratio will be meaningless as a surrogate for autonomic function. Non-linear measures like the short-term fractal scaling exponent, which reflects the fractal characteristics of rapid R–R interval oscillations measured by the detrended fluctuation analysis method, and the power law slope, appear to better differentiate between healthy and unhealthy organization of the heart’s rhythm and have proved to be more sensitive for risk stratification than some of the standard measures. HR turbulence, which is covered in detail
in another review, also appears to be sensitive to autonomic dysfunction. It is likely that the optimal characterization of normal or abnormal HR variability would be based on a combination of these different HR variability parameters rather than on a single one.
Heart rate variability after myocardial infarction The prognostic significance of HR variability has been extensively studied in patients who have survived an AMI. Table 1 summarizes the largest studies in this field. In 1978 Wolf et al. showed that patients with a low magnitude of short-term HR variability (no sinus arrhythmia on 30 consecutive RR intervals on admitting ECG) had a poor prognosis after AMI.1 In the late 1980s, the Multicenter Post-Infarction Project (MPIP) confirmed the predictive value of reduced HRV by measuring 24-h standard deviation of N–N intervals (SDNN) over 24 h following AMI.2 Patient with an SDNN 100 ms. Adjustment for covariates did not explain this association, although relative risk declined to 2.8. Other studies (including a reanalysis of MPIP) have shown that spectral measures of HR variability, mainly the ultra-low, very-low, and low frequency spectral components are reduced in survivors of AMI and that decreased values are associated with an increased risk of all-cause mortality.3 Because accurate Holter scanning is relatively labor intensive, the group at St. George's hospital in London experimented with global, geometric indices of HR variability (HRV index) that were derived from the histogram of NN intervals. Decreased values for these measures were also found to be associated with higher risk of mortality.4 In addition, this group suggested that markedly decreased SDNN for a single, stable 5-min period could pre-select patients in whom 24-h Holter recordings using geometric methods would provide significant risk stratification.5 The ATRAMI (Autonomic Tone and Reflexes After Myocardial Infarction) trial was a multicenter observational study performed about 10 years after MPIP. The ATRAMI investigators confirmed that reduced 24-h SDNN (< 70 ms in this case) is associated with a relative risk of 3.2 for mortality during 21 months of follow up.6 Furthermore, the combination of low SDNN and LVEF < 35% carried a relative risk of 6.7 for mortality when compared with patients with both higher SDNN and LVEF. A more recent study, however, in the era of modern therapy showed a very low incidence of severely depressed HR variability (10 % for SDNN < 50 ms). All patients in the study were treated with direct percutaneous coronary intervention within 12 h of their event. Although 4-year survival was 80% in this low SDNN group compared to 92% among those with higher SDNN, positive predictive value was low.7 The majority of early studies focused on measurement of HR variability relatively early after AMI.1–6 Measures of HR variability appeared to be more depressed at the early phase of AMI with substantial improvement during recovery.8 Despite this autonomic recovery, measurement of HR
Table 1 – Summary of main studies assessing the prognostic significance of heart rate variability after acute myocardial infarction. Study
HRV Measurement
Number of Patients
Mean Follow-Up Time
Primary Endpoint
SDNN
808
34 months
all-cause mortality
Bigger JT Jr et al., 1992
Frequency domain measures
715
2.5 years
all-cause death, cardiac death, arrhythmic death
Bigger JT Jr et al., 1993
Frequency domain measures
331
3 years
all-cause mortality
Fei L et al., 1996
SDNN and HRV index
700
1 year
all-cause mortality
Bigger JT Jr et al., 1996
Power–law behavior
715
3 years
all-cause mortality
La Rovere MT et al., 1998
SDNN
21 months
Huikuri HV et al., 2000
Spectral, fractal and time domain measures
Cardiac death and non-fatal cardiac arrest all-cause mortality
Mäkikallio TH et al., 2005
1284 446
685 days
Spectral, fractal and, time-domain measures
2130
1012 days
Sudden cardiac death, non-sudden cardiac death
Stein PK et al., 2005
Spectral, fractal, and time-domain measures
740
362 days
all-cause mortality
Perkiömäki JS et al., 2006
Spectral, fractal and time-domain measures Spectral, fractal and time-domain measures SDNN
675
30 months
Non-fatal coronary events
312
2 years
ECG documental VT or VF
412
4.3 years
all-cause mortality
Huikuri HV et al., 2009 Erdogan et al., 2008
Increased mortality, if SDNN < 50 ms All frequency domain measures of HRV predicted an increased risk of primary endpoint All frequency domain measures predicted the risk of primary endpoint Geometric HRV index predicted mortality better than SDNN Power–law regression between log(power) and log(frequency) predicted mortality SDNN < 70 ms predicted increased cardiac mortality Short-term fractal scaling exponent was the most powerful predictor of mortality All HRV measures predicted sudden and non-sudden cardiac death Fractal measures predicted all-cause mortality Several measures of HRV predicted primary endpoint Many HRV measures predicted the primary endpoint SDNN < 50 ms predicted mortality
Comment Landmark study VLF spectral power was the strongest predictor Holter recording were performed one year after AMI
Power–law regression analyzed from ULF and VLF spectral areas Multicenter study Post-AMI patients with depressed ejection fraction HRV predicted equally both sudden and non-sudden cardiac death. Largest HRV study Short-term fractal scaling exponent was more powerful predictor than other HRV measures
VLF power was the strongest predictor positive predictive power of SDNN was low
PR O G RE S S I N C ARDI O V A S CU L A R D I S EA S E S 5 6 (2 0 1 3) 15 3–1 5 9
Kleiger RE et al., 1987
Main Results
Abbreviations: AMI = acute myocardial infarction, ECG = electrocardiogram, HRV = heart rate variability, SDNN = standard deviation of N–N intervals, ULF = ultra-low frequency, VF = ventricular fibrillation, VLF = very-low frequency, VT = ventricular tachycardia.
155
156
PR O GRE S S I N C ARDI O VAS CU L AR D I S EAS E S 5 6 ( 2 0 13 ) 15 3–1 59
variability has been shown to yield similar prognostic information when the 24-h HR variability measurements were performed one year after AMI or in middle-aged women 3–6 months after hospitalization for acute coronary syndrome.9,10 However, more recently, perhaps due to the increased use of beta-blockers, a study of 600 patients 5–7 days post-MI and a follow up study of 400 of them one year after MI showed that while 24-h time or frequency domain HRV measures improved after MI, none of them predicted cardiac mortality at either time point.8 At the same the, short-term fractal scaling exponent, the power law slope and HR turbulence values remained stable during recovery and predicted cardiac death at both time points.8 Another factor to be considered is that at the time of the earliest studies of the predictive value of HRV for mortality post-MI, e.g. MPIP, post-MI mortality was about 30% on follow up, whereas currently it is far lower (< 4%), necessitating larger trials to test any putative risk marker. The majority of the post-MI studies performed in the 1980's and 1990's used 24-h time- and frequency domain as well as geometric measures of HR variability for risk stratification of patients after AMI. Although all statistical, geometrical, and spectral measures of HR variability differ in their manner of computation and analysis, these methods are fundamentally based on moment statistics and describe the magnitude of HR variability or of its underlying components. It is not therefore surprising that most of the measures that have been shown to have prognostic value have relatively close mutual correlation, and that there are only minor differences in prognostic power between them. It must be kept in mind though that measures of beat-to-beat HR variability which are supposed to reflect parasympathetic control of heart rate (HF power, pNN50, rMSSD) have rarely shown a strong association with outcome. This is likely because of the high prevalence of erratic rhythm in these populations resulting in higher measures for these parameters that do not reflect better parasympathetic function. More recent studies have applied methods based in nonlinear analysis methods, which provide very different and perhaps complementary information on HR dynamics compared to traditional statistical methods.11 Methods based on nonlinear dynamics have provided better prognostic information than that obtained by using more traditional methods. Analysis of l/f characteristics, i.e. the inverse power– law slope, which describes the slope of the spectral powers in the ultra-low and very-low frequency areas has provided prognostic information beyond the traditional HR variability measures.12 Decreased short-term fractal scaling exponent measured by the detrended fluctuation analysis method, a measure of greater randomness in short-term R–R interval fluctuations, has turned out to be a powerful non-linear index in risk stratification, a more powerful prognostic tool than traditional HR variability indexes in various post-AMI populations. In the DIAMOND-MI (Danish Investigations of Arrhythmia and Mortality ON Dofetilide) trial, short-term fractal exponent < 0.75 identified post-AMI patients at a relative risk of 3.0 for mortality and was more strongly related to outcome than traditional time and frequency domain measures.12 Moreover, reduced short-term fractal exponent
was related to both arrhythmic and non-arrhythmic mortality. In another study of consecutive patients with acute MI, both short-term fractal exponent < 1.025 (hazard ratio = 2.0) and power law slope ≤ − 1.507 (hazard ratio = 1.9) were independently associated with recurrent non-fatal coronary events.13 Another measure of increased randomness of heart rate patterns, measured as the ratio of the axes of a Poincare plot fitted to the plot of each N–N interval vs. the next, also predicted mortality in the Cardiac Arrhythmia Suppression Trial.14 Patients in this trial were at variable times post-MI but were selected for having significant ventricular arrhythmias. The majority of the studies assessing the prognostic power of HR variability after AMI have used all-cause mortality as the end-point. Some studies, using various definitions of arrhythmic events in their designs, have suggested that reduced or altered HR variability may be specifically related to arrhythmic events and sudden cardiac death. However, the clinical applicability of measurement of HR variability as a predictor of fatal arrhythmic events was challenged in a randomized prophylactic implantable cardioverter defibrillator (ICD) trial, the DYNAMIT (Defibrillator in Acute Myocardial Infarction Trial) which used reduced HR variability combined with reduced left ventricular ejection fraction measured early (within two weeks) after AMI as an inclusion criterion for the trial, but could not show any mortality benefit from ICD therapy in these presumably high-risk patients.15 The more recent multi-center CARISMA (Cardiac Arrhythmias and RIsk Stratification after Myocardial infArction) study showed that reduced HR variability measured relatively late (six weeks after AMI) rather than early after AMI, especially the very-low frequency spectral component, was the most powerful index among many other non-invasive risk markers in predicting the ventricular fibrillation or sustained ventricular tachycardia diagnosed by implantable arrhythmia devices.16 A further sub analysis of the CARISMA and another recent similar study (REFINE; Risk Estimation After Infarction, Noninvasive Evaluation) study confirmed that HR variability and HR turbulence yield more powerful prognostic information for arrhythmic events when measured late (6–10 weeks after AMI) rather than early (within two weeks) after AMI in the current treatment era.17 Thus, lack of recovery from reduced autonomic function after AMI was a stronger predictor of adverse events than decreased autonomic function at the early phase after AMI. The results from these two studies suggest that the measurement of HR variability for prognostic purposes should be performed late after AMI (> 6 weeks after AMI) rather than at the early phase after the acute event. Observational studies using various end-points have shown that reduced HR variability also predicts non-sudden cardiac death. The largest study in this field showed that reduced HR variability/turbulence is in fact a stronger predictor of non-sudden than sudden cardiac death.18 In this analysis, HR variability/turbulence, measured early after the AMI, was a better predictor of mortality among the patients with left ventricular ejection fraction (LVEF > 40%) than in those with reduced LVEF. Often forgotten in this focus on risk stratification, is the negative predictive value of HR variability measures. Patients with normal HR variability measures, even if they are post-
PR O G RE S S I N C ARDI O V A S CU L A R D I S EA S E S 5 6 (2 0 1 3) 15 3–1 5 9
AMI, are at very low risk of mortality. However, as has been found with other non-invasive risk markers, the positive predictive accuracy and sensitivity of abnormal HR variability for adverse outcomes have remained low in most observational studies. The sensitivity, specificity and predictive values of HR variability in predicting future events depend partly on defining the cutoff points of abnormal HR variability measures. The cutoff points vary between the studies and depend on the study populations, such as post-AMI patients with and without CHF, usage of beta-blocking medication etc., which influence the HR variability measures, and there is no universal consensus on an algorithm for determining the ideal cutoffs for different HR variability measures in different populations. Age and presence of diabetes and even clinical depression are factors influencing HR variability and should be considered in defining the cutoff values. HR variability has been combined with other risk markers in order to improve the risk stratification in some studies.6 The problem with combining various risk markers, however, is the loss of sensitivity despite an improved positive predictive accuracy. Little is known about the effect of interventions on HR variability in post-MI patients and about whether interventions that are associated with improvements in HR variability are associated with better outcomes or vice versa. One intervention, coronary artery bypass surgery (CABG) is associated with markedly decreased HR variability that slowly improves over weeks and months. It is likely that any predictive value for long-term outcomes contained in traditional time and frequency domain HR variability is lost once the patient has undergone CABG surgery.19 However, two studies have shown that measurements of non-linear HR variability parameters after CABG surgery did identify those at higher risk of a complicated post-surgical course.20,21 Only one trial, the BHAT (Beta-Blocker Heart Attack Trial) examined change in HR variability with therapy and the subsequent relationship of HR variability to mortality.22 In BHAT, patients randomized to propranolol had a significantly greater increase in HF power at 6-weeks and after 21 months of follow up. In the same study, higher HR variability at 6 weeks and assignment to the beta-blocker arm were independent predictors of a combined endpoint of mortality and non-fatal cardiac events. In a meta-analysis of the effect of pharmacologic, behavioral and exercise strategies on HR variability, Nolan et al.23 concluded that these secondary prevention strategies have a moderate but significant positive effect on HR variability. Considered together, the available data show that abnormal HR variability, measured by various time and frequencydomain measures, and non-linear measures from 24-h R–R interval recordings are general risk markers for common modes of cardiac death: arrhythmic, vascular, and hemodynamic after AMI. HR variability measured early after AMI seems to provide more powerful information on the risk of early nonsudden cardiac death, especially due to progressive heart failure. Because of remodeling of the arrhythmia substrate and autonomic function after AMI, early measurement of HR variability to identify those at high risk of sudden cardiac death should likely be repeated later (6–8 weeks) after AMI.
157
Future research in this era should focus on randomized trials that take into consideration the timing of HR variability/ turbulence measurement in their study designs. Information from such trials may identify the populations in which the clinical applicability of routine measurement of HR variability is warranted, the measurements most strong predictive of outcomes in these populations, and the optimal timing of such measurements. Furthermore, the results from REFINE and CARISMA studies suggest that measurement of the evolution of HR variability over time may have greater prognostic value. Measurement of the dynamic changes HR variability over time in association with prescribed therapies in various populations, perhaps on an individual basis, may also add to the clinic applications of HR variability measurements.
Heart rate variability in heart failure/left ventricular dysfunction Traditional analysis of HR variability has shown that its magnitude is generally reduced in CHF patients, and the detrended fluctuation analysis method has shown altered fractal characteristics of HR dynamics in CHF. The UK-Heart study focused on the prognostic significance of HR variability in outpatients with NYHA class I–III CHF and showed that reduced SDNN predicts mortality.24 Annual mortality rates for the study population were 5.5% for SDNN > 100 ms, 12.7% for SDNN = 50 to 100 ms and 51.4% for SDNN < 50 ms. Increased mortality was mainly due to progressive heart failure rather than sudden death among the CHF patients with reduced SDNN. However, in the DIAMOND-CHF study, which enrolled patients admitted to the coronary care unit with new or worsening CHF, only the reduced short-term fractal scaling exponent was independently associated with mortality in a population that was followed for a mean of 1.8 years and had a 42% mortality.25 Furthermore, the prognostic value of HR variability in this population was stronger in class II than in class III and IV patients. Several studies including primarily post-AMI patients with depressed LVEF have yielded quite similar results, i.e., reduced HR variability and altered short-term fractal HR behavior predict cardiac and all-cause mortality.11 Studies including patients with non-ischemic cardiomyopathy have also generally yielded similar results. In the DEFINITE (Defibrillators in Non-Ischemic Cardiomyopathy Treatment Evaluation) trial, SDNN was evaluated by tertiles.26 There were no deaths during 3-year follow up among those with SDNN > 113 ms (highest tertile). Among those with SDNN 81–113 ms 7% died and among those with SDNN < 81 ms 10% died. Among excluded patients (atrial fibrillation or > 25% ectopic beats) 17% died. The CARISMA and REFINE studies mentioned above included post-AMI patients with depressed LVEF (< 40%) after AMI and showed that HR variability/ turbulence measured late, but not early after AMI, predict fatal or near fatal arrhythmic events in these patients.16 The DINAMIT study also showed that HR variability measured early after AMI among the patients with LV dysfunction does not have a significant role in selecting the patients for ICD
158
PR O GRE S S I N C ARDI O VAS CU L AR D I S EAS E S 5 6 ( 2 0 13 ) 15 3–1 59
therapy.15 Similar to post-AMI patients, the clinical applicability of measurement of HR variability in patients with CHF remains uncertain because despite some evidence that an intervention like exercise training27 can improve at least parasympathetic measures of HR variability, there is a lack of randomized trials showing that the prognosis of patients with impaired HR variability can be improved by medical or device intervention. The exception might be the case of cardiac resynchronization therapy (CRT). Among 509 patients who received a CRT-D device for primary or secondary prevention of sudden death or who received a CRT pacemaker, use of CRT was associated with a significant increase in SDANN monitored by the device.28 SDANN is a measure of circadian activity, and an increase in SDANN could reflect an increase in functional status due to the device. However, SDANN ≤ 65 ms, at baseline or SDANN ≤ 76 ms 4 weeks after implantation were the strongest predictors of need for transplant or mortality, and there was a significant negative correlation between the reduction in LV end systolic volume with CRT and SDANN at baseline and at week 4. Thus, HR variability measured after implantation might identify CRT patients who are less likely to benefit and who remain at higher risk for adverse events.
Summary and recommendations Future research in this era should focus on randomized trials that take into consideration the timing of HR variability/ turbulence measurement in their study designs. Information from such trials may identify the populations in which the clinical applicability of routine measurement of HR variability is warranted, the measurements most strongly predictive of outcomes in these populations, and the optimal timing of such measurements. Furthermore, the results from REFINE and CARISMA studies suggest that measurement of the evolution of HR variability over time may have greater prognostic value. Measurement of the dynamic changes HR variability over time in association with prescribed therapies in various populations, perhaps on an individual basis, may also add to the clinic applications of HR variability measurements. Before the completion of such trials, patients with low HR variability/turbulence after AMI should receive the best available medical and invasive therapy, rehabilitation, and follow-up, even if routine measurements of HR variability/turbulence cannot yet be recommended in clinical practice guidelines.
Statement of Conflict of Interest All authors declare that there are no conflicts of Interest.
Acknowledgments Funded in part by the Sigrid Juselius Foundation, and the Finnish Foundation for Cardiovascular Research, Helsinki, Finland.
REFERENCES
1. Wolf MM, Vargios GA, Hunt D, Sloman JG. Sinus arrhythmia in acute myocardial infarction. Med J Aust. 1978;2:52–3. 2. Kleiger RE, Miller JP, Bigger JT, Moss AJ, the Multicenter Postinfarction Research Group. Decreased heart rate variability and its association with increased mortality after myocardial infarction. Am J Cardiol. 1987;59:256–62. 3. Bigger Jr 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;85:164–71. 4. Malik M, Farrell T, Cripps T, Camm AJ. Heart rate variability in relation to prognosis after myocardial infarction: selection of optimal processing techniques. Eur Heart J. 1989;10:1060–74. 5. Fei L, Copie X, Malik M, Camm AJ. Short- and long-term assessment of heart rate variability for risk stratification after acute myocardial infarction. Am J Cardiol. 1996;77:681–4. 6. La Rovere MT, Bigger Jr JT, Marcus FI, et al. Baroreflex sensitivity and heart-rate variability in prediction of total cardiac mortality after myocardial infarction. Lancet. 1998;351: 478–84. 7. Erdogan A, Coch M, Bilgin M, et al. Prognostic value of heart rate variability after acute myocardial infarction in the era of immediate reperfusion. Herzschrittmacherther Elektrophysiol. 2008;19:161–8. [Epub 2009 Feb 11]. 8. Jokinen V, Tapanainen JM, Seppänen T, Huikuri HV. Temporal changes and prognostic significance of measures of heart rate dynamics after acute myocardial infarction in the betablocking era. Am J Cardiol. 2003;92:907–12. 9. Bigger Jr JT, Fleiss JL, Rolnitzky LM, Steinman RC. Frequency domain measures of heart period variability to assess risk of late myocardial infarction. J Am Coll Cardiol. 1993;21: 729–36. 10. Janszky I, Ericson M, Mittleman MA, et al. Heart rate variability in long-term risk assessment in middle-aged women with coronary heart disease: the Stockholm Female Coronary Risk Study. J Intern Med. 2004;255:13–21. 11. Mäkikallio TH, Tapanainen JM, Tulppo MP, Huikuri HV. Clinical applicability of heart rate variability analysis by methods based on nonlinear dynamics. Cardiac Electrophysiol Rev. 2002;6:250–5. 12. Huikuri HV, Mäkikallio TH, Peng C-K, et al. Fractal correlation properties of R–R interval dynamics and mortality in patients with depressed left ventricular function after an acute myocardial infarction. Circulation. 2000;101:47–53. 13. Perkiömäki JS, Jokinen V, Tapanainen J, Airaksinen KE, Huikuri HV. Autonomic markers as predictors of nonfatal acute coronary events after myocardial infarction. Ann Noninvasive Electrocardiol. 2008;13:120–9. 14. Stein PK, Domitrovich PP, Huikuri HV, Kleiger RE, Cast Investigators. Traditional and nonlinear heart rate variability are each independently associated with mortality after myocardial infarction. J Cardiovasc Electrophysiol. 2005;16(1): 13–20. 15. Hohnloser SH, Kuck KH, Dorian P, et al. Prophylactic use of an implantable cardioverter–defibrillator after acute myocardial infarction. N Engl J Med. 2004;351:2481–8. 16. Huikuri HV, Raatikainen MJP, Moerch-Joergensen R, et al. Prediction of fatal or near fatal cardiac arrhythmia events in patients with depressed left ventricular function after an acute myocardial infarction. Results of the Cardiac Arrhythmias and Risk Stratification after Acute Myocardial Infarction (CARISMA) Study. Eur Heart J. 2009;30:689–98. 17. Huikuri HV, Exner DV, Kavanagh KM, et al. Attenuated recovery of heart rate turbulence early after myocardial
PR O G RE S S I N C ARDI O V A S CU L A R D I S EA S E S 5 6 (2 0 1 3) 15 3–1 5 9
18.
19.
20.
21.
22.
23.
infarction identifies patients at high risk of fatal or near-fatal arrhythmic events. Heart Rhythm. 2010;7:229–35. Mäkikallio TH, Barthel P, Schneider R, et al. Prediction of sudden cardiac death after acute myocardial infarction; role of Holter monitoring in the modern treatment era. Eur Heart J. 2005;26:762–9. Stein PK, Domitrovich PP, Kleiger RE, CAST Investigators. Including patients with diabetes mellitus or coronary artery bypass grafting decreases the association between heart rate variability and mortality after myocardial infarction. Am Heart J. 2004;147:309–16. Laitio T, Huikuri H, Kentala E, et al. Correlation properties and complexity of perioperative RR-interval dynamics in coronary artery bypass surgery patients. Anesthesiology. 2000;93:69–80. de Godoy MF, Takakura IT, Correa PR, Machado MN, Miranda RC, Brandi AC. Preoperative nonlinear behavior in heart rate variability predicts morbidity and mortality after coronary artery bypass graft surgery. Med Sci Monit. 2009;15:CR117–22. Lampert R, Ickovics JR, Viscoli CJ, Horwitz RI, Lee FA. Effects of propranolol on recovery of heart rate variability following acute myocardial infarction and relation to outcome in the Beta-Blocker Heart Attack Trial. Am J Cardiol. 2003;91:137–42. Nolan RP, Jong P, Barry-Bianchi SM, Tanaka TH, Floras JS. Effects of drug, biobehavioral and exercise therapies on heart
24.
25.
26.
27.
28.
159
rate variability in coronary artery disease: a systematic review. Eur J Cardiovasc Prev Rehabil. 2008;15:386–9. Nolan J, Batin PD, Andrews R, et al. Prospective study of heart rate variability and mortality in chronic heart failure: results of the United Kingdom Heart Failure Evaluation and Assessment of Risk Trial (UK-Heart). Circulation. 1999;98:1510–6. Mäkikallio TH, Huikuri HV, Hintze U, et al. Fractal analysis and time- and frequency-domain measures of heart rate variability as predictors of mortality in patients with heart failure. Am J Cardiol. 2001;87:178–82. Rashba EJ, Estes NA, Wang P, et al. Preserved heart rate variability identifies low-risk patients with nonischemic dilated cardiomyopathy: results from the DEFINITE trial. Heart Rhythm. 2006;3:281–6. Piotrowicz E, Baranowski R, Piotrowska M, Zieliński T, Piotrowicz R. Variable effects of physical training of heart rate variability, heart rate recovery, and heart rate turbulence in chronic heart failure. Pacing Clin Electrophysiol. 2009;32(Suppl 1):S113–5. Landolina M, Gasparini M, Lunati M, et al. Heart rate variability monitored by the implanted device predicts response to CRT and long-term clinical outcome in patients with advanced heart failure. Eur J Heart Fail. 2008;10:1073–9. [Epub 2008 Oct 1].
