OUP user menu

Relationship between cardiac autonomic function and sustained ventricular tachyarrhythmias in patients with an implantable cardioverter defibrillators

Irma Battipaglia, Lucy Barone, Luca Mariani, Fabio Infusino, Romolo Remoli, Giulia Careri, Gaetano Pinnacchio, Pierpaolo Tarzia, Gaetano Antonio Lanza, Filippo Crea
DOI: http://dx.doi.org/10.1093/europace/euq408 1725-1731 First published online: 19 November 2010

Abstract

Aims Low left ventricular ejection fraction (LVEF) is the main indication of implantable cardioverter defibrillators (ICD) in patients with dilated cardiomyopathy (DCM) for the primary prevention of sudden cardiac death, but ICD therapy at follow-up occurs in a minority of patients. We investigated whether heart rate variability (HRV) may improve risk stratification in DCM patients.

Methods and results We studied 42 patients (age 67.3 ± 3.5; 37 males) who had undergone ICD implant for either idiopathic or ischaemic DCM (LVEF <40%) 34.6 ± 19.7 months prior to the study (range 6–84). Patients underwent 24 h electrocardiographic Holter monitoring, and HRV was assessed over 2 hours in the afternoon showing stable sinus rhythm. Left ventricular ejection fraction was measured by two-dimensional echocardiography. The serum levels of C-reactive protein and N-terminal pro-B-type natriuretic peptide (NT-proBNP) were also obtained. The primary endpoint was the occurrence of appropriate ICD shocks in the 6 months preceding the study. The occurrence of appropriate ICD discharge from ICD implant was considered as a secondary endpoint. In the last 6 months, appropriate ICD shocks had occurred in seven patients (17%). There were no differences between patients with and without ICD shocks in clinical variables, as well as in LVEF and in C-reactive protein and NT-proBNP serum levels. In contrast, most HRV parameters were significantly depressed in patients with, compared with those without, ICD shocks; the most significant difference was shown for the average of the standard deviations of RR intervals in all consecutive 5 min segments (n ¼ 12) within the 2 h (26.7 ± 9 vs. 39.7 ± 14 ms; P= 0.02) in the time domain and for LF amplitude (8.4 ± 3 vs. 14.8 ± 7 ms; P= 0.02) in the frequency domain. Implantable cardioverter defibrillator discharge had occurred in 11 patients (26%) since ICD implant (average 35 months). No clinical or laboratory variable showed significant differences between patients with or without ICD discharge, except very low-frequency (VLF) amplitude (23.8 ± 7 vs. 30.8 ± 10.6 ms, respectively; P= 0.049).

Conclusion In ICD patients with reduced LVEF, several depressed HRV indices were significantly associated with appropriate ICD shocks in the previous 6 months, and VLF amplitude was the only variable significantly associated with ICD shocks recorded since ICD implant. These data suggest that full HRV analysis might be helpful for improving risk stratification for life-threatening ventricular arrhythmias and ICD indication in patients with DCM.

  • Implantable cardioverter defibrillator
  • Shock
  • Dilated cardiomyopathy
  • Ventricular tachyarrhythmias
  • Heart rate variability

Introduction

The implantable cardioverter defibrillator (ICD) has significantly improved survival in patients with an increased risk of sudden cardiac death (SCD).14 Although resuscitation from tachyarrhythmic SCD, in the absence of acute, reversible disease, is a clear indication for ICD, at present, the only accepted indication for ICD for the primary prevention of SCD is the documentation of depressed left ventricular (LV) function, usually LV ejection fraction (LVEF) <35%.5

Epidemiological data of treated patients, however, have clearly shown that most patients who receive an ICD according to this indication do not have any appropriate ICD discharge, even during long-term follow-up.6,7 Thus, the identification of patients who actually have a significantly high probability of SCD and would therefore benefit from ICD implant would be warranted.

Implantable cardioverter defibrillators are indeed expensive devices; furthermore, they are not without side effects. Besides possible implant-related complications, indeed, ICDs may impair quality of life through some limitations in patients' activities and inappropriate ICD discharges.810

Several possible tests and parameters have been proposed for improving risk stratification of patients with depressed LVEF in the attempt to select patients at low and high risk of SCD, including standard electrocardiographic (ECG) parameters (QRS duration and QT interval),11,12 ventricular late potentials,13 T-wave alternans,14,15 and measures of cardiac autonomic function.1620 Unfortunately, controversial data have often been reported for any of these methods, which, therefore, have not gained sufficient universal trust to be included in the clinical decision process for ICD indication.

Among variables assessing cardiac autonomic tone, heart rate variability (HRV) has only sporadically been investigated as the predictor of SCD, but studies have usually been focused on just one or few parameters.1620 Furthermore, there have been no studies specifically designed to answer the question of whether HRV can identify patients with appropriate ICD discharge therapy.

In this study, we aimed at assessing whether, in patients with reduced LVEF due to either ischaemic or idiopathic dilated cardiomyopathy (DCM), who were implanted with an ICD, HRV shows any association with the occurrence of life-threatening ventricular tachyarrhythmias requiring appropriate device intervention.

Methods

Patients

From October 2007 to June 2009, we prospectively studied 42 consecutive ICD patients (37 males, 5 females, age 67.3 ± 3.5 years) who performed routine controls of their device at our hospital and fulfilled the following inclusion criteria: (i) depressed LV function (LVEF <40%), related to either ischaemic or idiopathic DCM; (ii) sinus rhythm at a basal ECG recording; (iii) New York Heart Association (NYHA) class I–III; (iv) stable haemodynamic conditions in the last 6 months or since ICD implantation (in the case of <6-month implant).

Patients were excluded in the case of heart disease different from ischaemic or idiopathic DCM, in the case of a totally pacemaker (PM)-dependent rhythm, or recent episodes of clinical instability (besides ICD discharge), requiring hospital admission or changes in medical therapy for worsening of heart failure (e.g. diuretics, ACE-inhibitors, etc.).

Clinical assessment

A detailed clinical history, including drug therapy, was collected and a two-dimensional (2D) echocardiographic Doppler examination was performed in each patient. Left ventricular ejection fraction was measured on apical four- and two-chamber views using the Simpson method.

Holter monitoring

All patients underwent 24 h ECG Holter monitoring using a three-channel digital recorder (Oxford Medilog FD5). The bipolar chest lead CM5-CM1 and modified aVF were always monitored.

Heart rate variability was analysed, after careful revision of recording by an expert operator, using the Oxford Medilog Excel 3 system software (Oxford Instrument, Abingdon).

In the attempt to obtain constant sinus rhythm for HRV analysis, the atrial PM of patients was programmed at 40 b.p.m. Despite that, several patients continued to show intermittent atrial PM rhythm (mainly during the night). Thus, we decided to perform HRV analysis only on 2 consecutive hours of the 24 h Holter recording which did not show any atrial PM beat. The 2 h was always chosen in the afternoon, in a period between 3 and 8 p.m. (possibly from 4:30 to 6:30 p.m.), as we verified that a continuous 2 h ECG recording without any PM intervention could be obtained for all patients only during this period of the day.

Heart rate variability was assessed both in the time domain and in the frequency domain. Time-domain variables included: (i) SDNN (standard deviation of all RR intervals in the 2 h); and (ii) SDNNi [average of the standard deviations of RR intervals in all consecutive segments of 5 min (n= 12) in the 2 h].

Spectral analysis of RR intervals was done using the fast Fourier transformation (spectral resolution of 0.0005 Hz). Frequency-domain variables included RR interval amplitude in the: (i) very low-frequency (VLF, 0.0033–0.04 Hz); (ii) low frequency (LF, 0.04–0.15 Hz); and (iii) high frequency (HF, 0.15–0.40 Hz). The LF/HF ratio was also calculated.

C-reactive protein and B-type natriuretic peptide measures

A venous blood sample was collected in all patients. Blood was centrifuged at a rate of 1273g for 20 min. Then, serum aliquots of 500 μL were obtained and stored at −80°C until assayed.

C-reactive protein serum levels were measured using a high-sensitivity nephelometric method (BN100, Behring Diagnostic, Milano, Italy). N-terminal pro-B-type natriuretic peptide (NT-proBNP) was measured using an immunological test (pro-BNP II Elecsys kit, Roche Diagnostics Spa, Milano, Italy).

Clinical endpoints

The primary endpoint of the study was the occurrence of appropriate ICD discharges on episodes of sustained ventricular tachycardia or ventricular fibrillation, as derived from ICD interrogation, in the 6 months preceding the study.

The choice of this time period was based on the hypothesis that cardiac autonomic state (as assessed by HRV) could be better associated with arrhythmic events over a relatively short time period, as significant changes in sympatho-vagal balance during longer follow-up periods might modify its predictive value over time.21

Accordingly, to better define the association of HRV with ICD therapy, the occurrence of appropriate ICD discharges during the whole period from implantation to the time of the study was also assessed. The comparison of the association of HRV with ICD shocks occurring in the previous 6 months and those occurring over the whole period of follow-up from ICD implant allowed to verify the potential difference in HRV predictive value over relatively short and long follow-up periods.

Statistical analysis

Continuous variables were compared by unpaired t-test. As HRV frequency-domain parameters showed a skewed distribution, they were transformed into their natural logarithm for analysis. Comparisons of categorical variables were done by Fisher's exact test. Data are reported as average ± standard deviation or proportion. Statistical significance was attributed for two-tailed P< 0.05.

Results

Primary endpoint

Implantable cardioverter defibrillator interrogation showed that appropriate ICD shocks in the last 6 months had occurred in seven patients (17%). Four patients had two ICD interventions and three patients one ICD discharge. No inappropriate discharges were recorded in the device memory during this period in any patient.

There were no significant differences in the main clinical characteristics between patients with and without ICD shocks in the 6 months before the study (Table 1). Similarly, LVEF at 2D echocardiography and serum levels of both C-reactive protein and NT-proBNP did not differ between the two groups (Table 2).

View this table:
Table 1

Main clinical findings of patients with or without appropriate ICD shocks in the 6 months preceding the study

ICD shocks (n= 7)No ICD shocks (n= 35)P-value
Age (years)65.3 ± 5.867.7 ± 7.80.44
Gender (M/F)1 F, 6 M4 F, 31 M1
Cardiovascular risk factors
 Diabetes3 (43%)12 (34%)0.69
 Hypertension6 (86%)29 (83%)1
 Smoking1 (14%)5 (14%)1
 Familiar history of SCD4 (57%)24 (69%)0.67
 Obesity (BMI >30 kg/m2)1 (14%)6 (17%)1
 High cholesterol levels4 (57%)29 (83%)0.15
Diagnosis
 Ischaemic DCM5 (71%)26 (74%)1
 Idiopathic DCM2 (29%)9 (26%)
Indications to ICD
 Primary prevention5 (71%)27 (77%)1
 Secondary prevention2 (29%)8 (23%)
Device
 ICD only4 (57%)15 (43%)0.68
 ICD-biventricular PM3 (43%)20 (57%)
NYHA class
 I–II4 (57%)21 (60%)1
 III3 (43%)14 (40%)
Drug therapy
 β-Blockers (%)6 (91%)34 (97%)0.99
 ACE-inhibitors (%)6 (91%)28 (80%)0.67
 Diuretics (%)5 (71%)23 (66%)0.35
 Digoxin (%)2 (28%)3 (8.6%)0.84
 Amiodarone (%)1 (14%)4 (11%)0.66
FU from ICD implant (months)39.1 ± 3033.7 ± 170.58
  • BMI, body mass index; DCM, dilated cardiomyopathy; FU, follow-up; NYHA, New York Heart Association; PM, pacemaker; SCD, sudden cardiac death.

View this table:
Table 2

Main instrumental and laboratory findings of patients with or without appropriate ICD shocks in the 6 months preceding the study

ICD shocks (n= 7)No ICD shocks (n= 35)P-value
QRS complex width (ms)145.9 ± 36146.5 ± 220.95
LVEF (%)32.4 ± 634.2 ± 100.67
Ventricular arrhythmias at Holter ECG
 PVCs/hour94.5 ± 100120.0 ± 1820.35
 Number of NSVT1.4 ± 1.35.4 ± 160.84
Laboratory assays
 C-reactive protein (mg/L)3.4 ± 2.74.1 ± 6.40.45
 NT-proBNP (pg/mL)841 ± 51833 ± 2810.58
  • ECG, electrocardiogram; LVEF, left ventricular ejection fraction; NSVT, non-sustained ventricular tachycardia; NT-proBNP, N-terminal pro-B-type natriuretic peptide; PVCs, premature ventricular complexes.

In contrast, several HRV parameters were found to be significantly lower in patients with, compared with those without, appropriate ICD discharge (Table 3). The most significant differences were observed for SDNNi in the time domain (26.7 ± 9 vs. 39.7 ± 14 ms, respectively; P= 0.02) and for LF amplitude in the frequency domain (8.4 ± 3 vs. 14.8 ± 7 ms, respectively; P= 0.02).

View this table:
Table 3

Heart rate variability parameters in patients with or without ICD shocks in the previous 6 months

ICD shocks (n= 7)No ICD shocks (n= 35)P-value
RR (ms)772.9 ± 161804.05 ± 1070.52
SDNN (ms)52.6 ± 14.571.7 ± 230.04
SDNNi (ms)26.7 ± 939.7 ± 140.02
VLF amplitude (ms)21.5 ± 730.46 ± 100.04
LF amplitude (ms)8.4 ± 314.8 ± 70.02
HF amplitude (ms)8.2 ± 413.6 ± 100.06
LF/HF1.1 ± 0.21.2 ± 0.50.67

Values of SDNNi <44.0 ms identified all patients with ICD shocks. Specifically, ICD shocks occurred in 7 of 29 patients with SDNNi <44.0 ms and in none of 13 patients with SDNNi values ≥44.0 ms. Thus, sensitivity, specificity, and positive and negative predictive values for ICD shocks of SDNNi <44.0 ms were 100, 37, 24, and 100%, respectively.

Similarly, values of LF amplitude <13 ms identified all patients with ICD shocks. Implantable cardioverter defibrillator shocks occurred in 7 of 27 patients with LF amplitude <13.0 ms and in none of 15 patients with values ≥13.0 ms. Thus, sensitivity, specificity, and positive and negative predictive values for ICD shocks of LF amplitude <13.0 ms were 100, 43, 26, and 100%, respectively.

Implantable cardioverter defibrillator shocks from implant to study

One or more appropriate ICD shocks occurred from implant to follow-up (34.6 ± 19.7 months; range, 6–84) in 11 out of 42 patients (26%). The results showed that no clinical or laboratory variable acquired at the time of the study was associated with appropriate ICD therapy during the whole follow-up (Table 4). Heart rate variability parameters showed lower values in patients with, compared with those without, ICD discharges, but statistical significant difference was achieved only for VLF (Table 5). The values of VLF amplitude <33.5 ms identified all patients with ICD discharge over the whole follow-up. Implantable cardioverter defibrillator shocks occurred in 11 of 31 patients with VLF amplitude <33.5 ms and in none of 11 patients (26.2%) with values ≥33.5 ms.

View this table:
Table 4

Main clinical findings of ICD patients with or without appropriate ICD shocks during the whole follow-up period (average 34.6 months)

ICD shocks (n= 11)No ICD shocks (n= 31)P-value
Age (years)68.8 ± 7.266.7 ± 7.70.44
Gender (M/F)1 F, 10 M4 F, 27 M1
Cardiovascular risk factors
 Diabetes4 (36.4%)11 (35.5%)1
 Hypertension10 (91%)25 (80.6%)0.65
 Smoking1 (9%)5 (16%)1
 Familiar history of SCD7 (63.6%)21 (67.7%)1
 Obesity (BMI >30 kg/m2)1 (9%)6 (19.4%)0.65
 High cholesterol levels7 (63.6%)26 (84%)0.21
Diagnosis
 Ischaemic DCM8 (72.7%)23 (74%)1
 Idiopathic DCM3 (27.3%)8 (26%)
Indications to ICD
 Primary prevention6 (54.5%)26 (84%)0.09
 Secondary prevention5 (45.5%)5 (16%)
Device
 ICD only7 (63.6%)12 (38.7%)0.18
 ICD-biventricular PM4 (36.4%)19 (61.3%)
NYHA class
 I–II6 (54.5%)19 (61.3%)0.73
 III5 (45.5%)12 (38.7%)
Drug therapy
 β-Blockers (%)10 (91%)30 (96.8%)0.46
 ACE-inhibitors (%)10 (91%)24 (77.4%)0.66
 Diuretics (%)7 (63.6%)21 (67.7%)1
 Digoxin (%)2 (18.2%)3 (9.7%)0.59
 Amiodarone (%)2 (18.2%)3 (9.7%)0.59
FU from ICD implant (months)38.1 ± 25.433.5 ± 17.40.59
QRS complex width (ms)143.5 ± 32.4147.5 ± 21.80.72
LVEF (%)35.0 ± 9.933.5 ± 100.68
C-reactive protein (mg/L)2.7 ± 2.34.4 ± 6.70.23
NT-proBNP (pg/mL)1588.6 ± 2731691 ± 1900.91
Arrhythmias on Holter ECG
 No. of PVCs/hour84.3 ± 116104.1 ± 1530.66
 No. of NSVT1.64 ± 2.16 ± 170.19
  • BMI, body mass index; DCM, dilated cardiomyopathy; FU, follow-up; NYHA, New York Heart Association; PM, pacemaker; SCD, sudden cardiac death; ECG, electrocardiogram; LVEF, left ventricular ejection fraction; NSVT, non-sustained ventricular tachycardia; NT-proBNP, N-terminal pro-B-type natriuretic peptide; PVCs, premature ventricular complexes.

View this table:
Table 5

Heart rate variability parameters in patients with or without ICD shocks during the whole follow-up period (average 34.6 months)

ICD shocks (n= 11)No ICD shocks (n= 31)P-value
RR (ms)782 ± 133805 ± 1110.58
SDNN (ms)62.4 ± 20.770.7 ± 23.50.3
SDNNi (ms)31.8 ± 1139.7 ± 14.50.11
VLF amplitude (ms)23.8 ± 730.8 ± 10.60.049
LF amplitude (ms)11.5 ± 5.714.5 ± 7.70.24
HF amplitude (ms)11 ± 6.313.4 ± 10.40.46
LF/HF1.1 ± 0.21.2 ± 0.50.41

Discussion

In this study, in patients with an ICD, the only variable associated with appropriate device discharge for ventricular tachyarrhythmias in the previous 6 months was the presence of depressed values of HRV, compatible with a derangement of sympatho-vagal balance towards sympathetic predominance. No other clinical or laboratory finding, including LVEF and serum NT-proBNP and C-reactive protein levels, showed significant association with recent ICD discharge in our patients.

Heart rate variability was also significantly associated with the occurrence of appropriate ICD shocks since ICD implant in our small group of patients. The association, however, only concerned VLF and was just below the statistical significance, suggesting that HRV may have a limited predictive value for arrhythmic events over long-term follow-up.

Implantable cardioverter defibrillator implant has significantly improved survival in patients at risk of SCD, due to its effectiveness in the treatment of life-threatening ventricular tachyarrhythmias.14 After MADIT-II results were available,4 the only accepted indication to ICD has become the evidence of depressed LVEF (usually 30–35%).5 Nevertheless, the data from clinical trials and registries have made evident that this exclusive indication has clear limits, as most patients do not receive any appropriate shocks at follow-up,6,7 making ICD implant in the whole population of patients with LV dysfunction limitedly cost-effective.22 Moreover, ICD implant is not without risk and may also impair quality of life.810

Therefore, it would be desirable to identify variables able to select patients at high or low risk of SCD among these patients, in order to optimize cost-effectiveness and the benefit–risk ratio of ICD.

Several methods and variables have been proposed,1120 but, after initial enthusiasm, the evidence of controversial results has impeded their use on a large scale for this purpose and none, at present, are actually recommended in clinical guidelines.5

A derangement of the cardiac sympatho-vagal function towards a predominance of sympathetic activity has been shown to predict cardiovascular mortality in several clinical settings, including depressed LV function.2326 Furthermore, some studies have found that HRV also predicts the occurrence of SCD independent of LVEF.1620 Although cardiac autonomic function can be assessed by several methods, including heart rate turbulence27 and baroreceptor sensitivity,16 most studies have used HRV analysis. However, there have been no specific studies aimed at assessing whether HRV can be helpful to stratify for SCD and life-threatening arrhythmic events in patients who are candidates to ICD implant.

In the DINAMIT trial, in fact, the use of low SDNN (<70 ms) in addition to low LVEF (<35%) to select patients for trial enrolment was unable to identify patients who could benefit from ICD, and ICD discharge occurred in 17.8% of the patients.28 However, this study presented some limitations. First, it included patients with a recent acute myocardial infarction (AMI), which may have limited the possibility of a correct risk stratification by HRV; indeed, HRV is known to progressively improve up to 6 months after AMI;21,29 thus, the assessment of HRV in a stabilized phase of the disease could be more appropriate, as also recently suggested by the REFINE study.30 Secondly, patients were selected also according to a high heart rate (80 b.p.m.), which is only a mild predictor of risk, not independent of HRV.18 Finally, only SDNN was considered as an HRV variable in the DINAMIT study, whereas other variables, in particular frequency-domain parameters, might have a more significant predictive value for SCD occurrence.1620

In this study, we evaluated several HRV parameters and showed that depressed HRV was able to predict patients with ICD discharge in the previous 6 months. Interestingly, VLF and LF amplitudes seemed the most helpful HRV parameters to reliably predict arrhythmic events, whereas SDNN had limited power.

On the other hand, HRV was less predictive of ICD discharge over the whole long follow-up period from device implant (about 35 months), although VLF amplitude still showed significant association with the endpoint. This suggests that HRV assessment can be more helpful to predict short/mid-term risk and that a periodic reassessment of HRV can be mandatory for its optimal use as a risk predictor.

It should be noticed that HRV assessment was associated with a high negative predictive value for ICD shocks, whereas its ability to predict arrhythmic events was limited, thus suggesting that HRV might be more helpful to identify patients at very low risk of events that might not have need for ICD implant.

It is worth noting that in contrast to HRV parameters, no predictive power for ICD discharge could be detected for C-reactive protein and NT-proBNP serum levels in our patients.

C-reactive protein and other inflammatory variables have been suggested to predict cardiovascular events and mortality in ischaemic as well as in heart failure patients31,32 and some observational retrospective data suggested that it could predict SCD.33,34 However, C-reactive protein is a non-specific inflammatory marker and its relation with SCD can be biased by confounding factors. In particular, several recent studies have shown that the autonomic nervous system can significantly modulate inflammatory reactions, with vagal impairment being associated with increased inflammatory activity.35,36 Accordingly, depressed HRV has been shown to be associated with increased C-reactive protein serum levels, as well as with other markers of inflammation, in several clinical studies.3638 Thus, the association of C-reactive protein with SCD might be in some cases mediated by the relation of C-reactive protein with impaired autonomic activity.39

Similarly, NT-proBNP and other natriuretic peptides are strong predictors of cardiovascular mortality in several clinical settings, including heart failure and LV dysfunction.40,41 Furthermore, a few studies have reported that NT-proBNP can also predict SCD.42,43 However, natriuretic factors seem to reflect better the severity of LV dysfunction rather than an increased tendency to life-threatening arrhythmias.

Limitations of the study

Some limitations of our study should be acknowledged. The number of patients was small and, therefore, our data should be confirmed in larger populations. Future studies should also better clarify whether the predictive role for ICD discharge of HRV apply to patients with both ischaemic and non-ischaemic DCM; indeed, although HRV in our study predicted the endpoint independently from the underlying disease, only 11 patients (26%) had non-ischaemic DCM.

Heart rate variability parameters in this study were derived from only 2 afternoon hours of ECG recording; this might have introduced some bias due to differences in activity among patients. However, the life style of our patients was largely similar. Furthermore, although how the predictive value of short-period HRV compares with 24 h HRV needs further investigation, the ability of 2 h HRV to predict events in our study is in keeping with previous reports that showed a significant prognostic value of HRV indices derived from as short periods as from 2 h to 2 only minutes.44,45

As sudden death displays a circadian rhythm with the highest rate in the early morning,46 the predictive role of HRV might have been even stronger if HRV had been measured over 2 hours in the morning; however, the choice of the period was obliged in our study, as the afternoon hours were those that allowed HRV assessment in all patients thanks to stable sinus rhythm.

Conclusion

Our data show that in ICD patients with reduced LVEF, several depressed HRV indices were significantly associated with appropriate ICD shocks in the previous 6 months, and VLF amplitude might also be associated with ICD shocks over longer follow-up. These data suggest that full HRV analysis would merit assessment in large studies to better define its utility for improving risk stratification for life-threatening ventricular arrhythmias and ICD indication in this kind of patient.

Conflict of interest: none declared.

References

View Abstract