Heart Rate Variability and Cardiac Diseases

Fig. 10.1

Kaplan-Meier survival curves from the Multicenter Post-Infarction Study demonstrating a decreased survival among patients with an SDNN of <50 ms (Reprinted from Kleiger et al. [2] with permission)

The heart rate turbulence (HRT) implicates the baroreflex-mediated autonomic influence on the sinoatrial node [5, 6]. In healthy subjects a ventricular premature ectopic beat provokes a biphasic reaction of an early acceleration and late deceleration of the heart rate, while in high-risk subjects such a reaction is diminished or even completely nonexistent. Schmidt et al. [7] showed that the HRT is even a stronger risk predictor than the traditional time- or frequency-domain variables using the Multicenter Post-Infarction Program patients.

10.1.1.2 After the Reperfusion Era

Timely reperfusion after an acute MI facilitates the cardiomyocyte salvage and decreases the cardiac morbidity and mortality. The crude annual mortality from the above studies decreased from approximately 10 % [2, 3] to 2–3 % in the current studies [8–10], probably due to the multidisciplinary approach in acute MIs [11]. Data on the risk stratification in recent MI patients using traditional and new methods have been reported. The GISSI study evaluated the prognostic value of the HRV in 567 patients with an acute MI who underwent thrombolytic therapy [8]. They had 52 deaths (9.2 %) during a 1000-day follow-up. They found that the time-domain analysis (SDNN, NN50+, and rMSSD) identified high-risk groups, but the relative risk of these measures was at most 3. The ATRAMI study, where 63 % of the patients received reperfusion therapy after an acute MI, showed that patients with an SDNN of <70 ms or baroreflex sensitivity of <3.0 ms per mm Hg carried a significant multivariate risk of cardiac mortality (3.2 [95 % CI 1.42–7.36] and 2.8 [1.24–6.16], respectively) after adjusting for the left ventricular ejection fraction (LVEF) and ventricular ectopy [9]. The DIAMOND substudy assessed the use of various fractal analysis methods of HRV from 24-h Holter recordings to predict death in 446 patients with an acute MI and a depressed left ventricular function [12]. They found that the short-term scaling exponent α1 was the most powerful measurement for predicting both arrhythmic and non-arrhythmic death.

Autonomic dysfunction occurring early after an MI has been shown to improve over time [13].

Data regarding the denervation during the early phase of an acute MI and reinnervation during the late phase of an MI have been demonstrated in humans [14, 15]. Furthermore, early revascularization, the use of β-blockers, and angiotensin-converting enzyme inhibitor ameliorate the remodeling and modulate the arrhythmia substrate following an MI [11]. The above three studies [8, 9, 12] confirmed the ability of the HRV analysis for risk stratification in the early phase after revascularization, whereas recent results demonstrated that the HRV measured during the late phase after an acute MI (10–14 weeks) is more predictive of mortality than those measured during the early phase after an MI (2–4 weeks) [16]. Recently, Huikuri et al. examined the relationship between the magnitude of recovery of the HRV or HRT recorded during the early and late phases of an MI. They found that an attenuated recovery of the VLF and LF components of the HRV and HRT was associated with a risk for serious arrhythmias but not with non-arrhythmic death [17]. Early measurement of the HRV to identify those at high risk should likely be repeated later in order to assess the risk of fatal arrhythmic events in the late phase.

Recently, several new HRV measurements have been proposed and their predictive values have been reported. Bauer et al. [18] explored a novel HRV measurement, termed the acceleration capacity (AC) and deceleration capacity (DC) of the heart rate. They developed and validated the predictive accuracy of the AC and DC using several large post-MI cohorts. They noted that an impaired DC of ≤2.5 ms was an efficient predictor of mortality in patients with a reduced LVEF (<30 %) as well as a preserved LVEF. They subsequently tested the incremental value of the DC adding it to the HRT in other cohorts of post-MI patients [19]. They found that patients with both an abnormal HRT and abnormal DC (called severe autonomic failure [SAF]) had a higher mortality irrespective of the LVEF.

More recently, Kisohara et al. [20] proposed a modified AC and DC, which were calculated with the time scales of T (window size defining heart rate) and s (wavelet scale) from 1 to 500 s and compared their prognostic values with the conventional measures proposed by Bauer (AC (conv) and the DC (conv)) that were calculated with (T, s) = [1,2 (beat)] [18] (detailed explanation will be in the Part II, Chap. 7). They used a cohort of 708 post-MI patients, with a crude annual mortality of 2.7 %. They found that a decrease in the DC for the minute-order long-term heart rate dynamics is a strong predictor of mortality, and their predictive power was independent of the AC (conv) and DC (conv). This finding bolsters the role of a new DC in the risk assessment among the post-MI patients and provides impetus for further study of the incremental value of the DC in a reduced and preserved LVEF population, respectively.

Kiyono et al. [21] previously reported a new HRV measure termed the non-Gaussianity index (lambda: λ) that characterizes an increased probability of large heart rate deviations from its trend (detailed explanation will be in the Part II, Chap. 9) [21]. A previous study reported that an increased λ is an independent mortality predictor among patients with chronic heart failure [22]. Hayano et al. [23] recently examined the predictive value of λ in 670 post-MI patients. During a median follow-up period of 25 months, 45 (6.7 %) patients died (32 cardiac and 13 noncardiac deaths). An increased λ predicted exclusively cardiac death after adjustments for the covariates. The combination of an increased λ and abnormal HRT provided the best predictive model for cardiac death (Fig. 10.2).

Fig. 10.2

Kaplan-Meier curves for the cardiac death after an AMI. The patients were stratified by a λ25 s > 0.6 and by the combination of a λ25 s > 0.6 and abnormal HRT. AMI acute myocardial infarction, HRT heart rate turbulence (Reprinted from Hayano et al. [23] with permission)

10.1.2 Heart Failure

The epidemiology of heart failure (HF) has been changing, with an increasing proportion of patients being diagnosed with HF and a preserved left ventricular ejection fraction (HFpEF), in addition to the classic HF with a reduced left ventricular ejection fraction (HFrEF). Compared to those with HFrEF, the patients with HFpEF are more often female, older, less likely to have coronary artery disease, and more likely to have hypertension [24]. Despite advancement in the heart failure treatment, the mortality rate of a HFpEF is similar to that of an HFrEF [24]. The mode of death depends on the New York Heart Association functional class. Patients with mild to moderate HF more frequently die suddenly, while those with severe HF are more likely to die of pump failure [25].

10.1.2.1 Risk Stratification in Heart Failure with Reduced Left Ventricular Ejection Fraction

Numerous studies have examined the prognostic role of various HRV measurements in those with an HFrEF for years [26]. Although a decreased time-domain HRV analysis, mostly the SDNN, has been considered as an independent risk marker for all-cause death or pump failure death [27–31], data on predicting sudden cardiac death is not consistent [32].

Much greater controversy exists in regard to the spectral analysis for the risk stratification in patients with HFrEF. Galinier et al. [29] showed that ischemic heart disease and a reduced LF power during the daytime were associated with sudden death. La Rovere [33] analyzed the short-term spectral components (8 min) during controlled breathing. They developed and validated the ability of a reduced short-term LF power to predict sudden death. Guzetti et al. [34] showed that a decreased VLF power at night was independently related to death from progressive pump failure, while the reduction in the LF power during the night was linked to sudden death. The role of the HRV in predicting the mortality of HF patients was recently confirmed by the GISSI-HF Holter substudy [35]. They found that the LF, VLF, and an abnormal HRT were associated with arrhythmic events defined as sudden death or appropriate discharges from an implantable cardioverter-defibrillator (ICD). These studies indicated that a spectral analysis for the risk stratification in patients with HFrEF seems to be a better marker of sudden death compared to a time-domain analysis. Finally, Maestri et al. [36] showed that nonlinear HRV indices provide important prognostic information on top of clinical data. They found that two variables (from empirical mode decomposition and symbolic dynamics families) added prognostic information to the clinical model.

10.1.2.2 Risk Stratification in Heart Failure Patients with a Preserved Left Ventricular Ejection Fraction

The HFpEF patients are not covered by current indications for ICDs. In this patient population, the HRT seems to be more useful in the risk stratification of sudden death rather than the HRV. The aforementioned GISSI-HF Holter substudy [35] demonstrated that an abnormal HRT in patients with an LVEF of >30 % was associated with a fourfold increased risk of arrhythmic events. In the HFpEF patients (LVEF >35 %) from the MUSIC study, a combination of an abnormal HRT, decreased HRV, and impaired QT interval dynamics provided the highest accuracy in identifying high-risk groups for the all-cause mortality and sudden death [37]. Further studies are required to test whether nonlinear HRV indices provide prognostic information using a large cohort of patients with HFpEF.

10.1.3 Nonischemic Heart Disease

Patients with nonischemic cardiomyopathy are at an increased risk of sudden death and heart failure death, with a 12–20 % estimated mortality rate at 3 years [38–40]. They have frequent and high-grade ventricular ectopy and a reduced HRV like those with ischemic heart disease. Currently, various risk stratifiers have been reported, but available data, especially in nonischemic patients, are too small to provide clear recommendations. A recent meta-analysis by Goldberger et al. [41] examined the prognostic role in arrhythmic outcomes, in 6088 patients with nonischemic dilated cardiomyopathy from 45 studies. They reviewed and validated the predictive value of various tests including autonomic tests, echocardiographic data, electrophysiologic studies, arrhythmias, surface ECG findings, signal-averaged electrocardiograms, vectrocardiograms, and T-wave alternans. They found that the measurements of depolarization (fragmented QRS) and repolarization (T-wave alternans) had a higher odds ratio, but none of the autonomic tests (HRV, HRT, and baroreflex sensitivity) were significant predictors of arrhythmic outcomes.

10.1.4 Therapeutic Implications of the HRV (ICD Implantations)

The contemporary therapeutic use for HRV analysis in post-MI or HF patients is the selection of patients for prophylactic ICD therapy [42–44]. Presently, no definitive recommendations can be given for the clinical use of noninvasive ECGs or Holter-derived risk markers in selecting candidates for prophylactic ICD therapy.

The contemporary practice is almost solely reliant on the LVEF for the risk stratification; some studies demonstrated the incremental value of the HRV in patients with a decreased LVEF in prospective ICD studies. In DINAMIT [45], 675 recent MI patients (day 6 to day 40) with an LVEF ≤35 % and SDNN ≤70 ms (or mean 24 h heart rate of ≥80 bpm) were randomized to receive or not receive an ICD. There was no significant difference in the survival rate between the groups. The ICD reduced the arrhythmic mortality but increased the non-arrhythmic mortality. In the ALIVE study [46], 3717 post-MI patients with a depressed LVEF were randomized to a placebo and azimilide and classified them into low- and high-risk groups on the basis of the triangular index of the HRV. By a multivariate analysis, a low HRV was associated with the all-cause mortality, but it did not predict the arrhythmic mortality. These data showed that linear HRV measurements in addition to a depressed LVEF may not be a useful marker to isolate patients who will benefit from ICD therapy.

The majority of the studies assessing the prognostic power of the HRV in post-MI or HF patients have used all-cause mortality as the endpoint. Some studies have suggested that a reduced HRV may be specifically related to arrhythmic events and sudden death. The CHARISMA trial assessed the use of a combination of the HRV, HRT, ambient arrhythmias, signal-averaged electrocardiograms, T-wave alternans, and programmed electrical stimulation (PES) to predict fatal arrhythmias when performed 6 weeks after an acute MI [47]. A total of 312 patients with a mean LVEF of 31 % received an implantable ECG loop recorder to detect ventricular fibrillation (VF) or symptomatic sustained ventricular tachycardia (VT). They found that an adjusted hazard ratio of the reduced VLF component and induction of sustained monomorphic VT during the PES was 7.0 and 4.8, respectively. Although multiple HRV measurements have been investigated to achieve a better selection, nonlinear HRV measurements have been neither analyzed in randomized controlled ICD trials nor tested as to whether they harbor a significant power to serve as a practical risk predictor. Recently, Au-yeung et al. [48] performed a retrospective analysis on the HRV measurements and occurrence of VT/VF and sudden death in the heart failure patients enrolled in the SCD-HeFT trial [40]. They found that a combination of the fractal exponent (α1 and α2) and HRT or LF/HF ratio, number of premature ventricular complexes, or triangular index of the HRV could discriminate survivors and those who had appropriate discharges from their ICDs. Of note, the high negative predictive value from these data identifies patients who are not likely to experience ICD therapies with a 97.3 % certainty.

Sudden death due to VT/VF is a major mechanism of sudden death, and pulseless electrical activity (PEA) and asystole occupy the rest [49]. There has been a progressive decrease in the prevalence of sudden death presenting as VF in recent publications, and a proportional increase in those presenting with non-shockable rhythms [50–52]. It is plausible that, in addition to the differences in the clinical characteristics and arrhythmic substrate, these two distinct mechanisms of sudden death make it difficult to predict the onset of sudden death with a high certainty (Fig. 10.3).

Fig. 10.3

Why is it difficult to predict sudden death? In addition to the differences in the clinical characteristics and arrhythmic substrate, two distinct mechanisms of sudden death (VT/VF and PEA or asystole) make it hard to predict the onset of sudden death with a high certainty. VT ventricular tachycardia, VF ventricular fibrillation, PEA pulseless electrical activity, LVEF left ventricular ejection fraction

To date, the only reliable metric to predict the benefit from an ICD is a severely depressed LVEF; however, the predictive value of the LV function is relatively low [53]. HRV has a low positive predictive value, but a high negative predictive value in predicting sudden death. Therefore, we suppose that a multivariable risk modeling should incorporate a Holter-based HRV analysis to refine the ICD patient selection, especially to exclude patients who are unlikely to benefit from ICD therapy. On the other hand, a risk model that incorporates a high positive predictive value can identify those that would benefit from ICD therapy. The clinical utility of noninvasive tests will be assessed in an ongoing randomized trial of ICDs (REFINE-ICD; NCT identifier 00673842).

10.2 Other Indications for the HRV

10.2.1 Heart Rate Variability Just Before Sudden Death

If we can recognize specific precursors before the initiation of ventricular tachycardia or ventricular fibrillation (VT/VF) by analyzing the HRV, these findings could be helpful in implementing algorithms for an ICD to improve the diagnosis and therapy of VT/VF. To date, several studies reported characteristic changes in the time- and frequency-domain measurements before the onset of VT/VF. Shusterman et al. [54] analyzed the HRV indices in 53 patients enrolled in the ESVEM trial that included ischemic heart disease patients who had a history of cardiac arrest, documented VF, sustained VT, or syncope. Only 10 % of the patients were on β-blockers. They found that the VLF, LF, and HF declined 30 min before sustained VT compared to that observed in the controls. In addition, the LF and LF/HF ratio declined just before the onset of the VT. Osaka et al. [55] performed frequency-domain analyses in 34 patients experiencing VF, cardiac arrest, or an acute MI and compared that to an age-, sex-, and disease-matched cohort of 191 patients. They found that several hours before the onset of a cardiac event, a transient decrease in the LF/HF was followed by an increase in the LF/HF. However, conflicting results were reported in which neither time- nor frequency-domain analyses failed to find characteristic changes in the RR dynamics before the VT/VF [56]. Recent studies using nonlinear methods found a significant difference in the RR dynamics before the occurrence of VT/VF. Makikallio et al. [57] examined the heart rate dynamics in post-MI patients who had VT/VF during Holter ECGs, where they found that a short-term fractal scaling exponent (α) obtained by a detrended fluctuation analysis and the slope (β) of the power-law regression line of the RR interval dynamics were the most powerful independent predictors differentiating patients with VT/VF from the controls. Because the deviations in the short-term slopes from the 1/f curve occur 15 min to 1 h before the onset of VF, a fractal-like signal behavior may reflect changes in the autonomic modulation of the heart. In another report by Wessel et al. [58], they analyzed the linear and nonlinear parameters using 1,000 beat-to-beat intervals stored in an ICD. They found that the time- and frequency-domain parameters were similar between the patients with and without VT/VF, but the symbolic dynamics and finite-time growth well discriminated those patients.

10.2.2 Atrial Fibrillation

Atrial fibrillation (AF) remains the most common, sustained arrhythmia and is independently associated with a higher risk of an ischemic stroke, as well as a poorer quality of life, higher hospitalization rates, and excess mortality [59]. Most of the HRV studies, however, excluded AF patients from the analysis. In early reports by Stein et al. [60] and Frey et al. [61], they demonstrated that an SD of a 5 min average ventricular response interval during a 24 h recording (SDAVI) was associated with mortality in chronic AF. A previous study by Yamada et al. [62] demonstrated the prognostic values of the linear and nonlinear dynamics of the heart rate in patients with chronic AF. Data on 107 participants showed that reduced nonlinear measurements by means of the Shannon entropy and approximate entropy were independent predictors for death after an adjustment for the clinical variables. The conventional linear measurements, however, were not significant predictors (Fig. 10.4).