Author + information
- Received June 16, 1998
- Revision received October 15, 1999
- Accepted December 29, 1999
- Published online April 1, 2000.
- Marek Malik, PhD, MD, FACC∗,* (, )
- A.John Camm, MD, FACC∗,
- Michiel J Janse, MD†,
- Desmond G Julian, MD, FACC‡,
- Gerald A Frangin, MD§,
- Peter J Schwartz, MD, FACC∥,
- on behalf of the EMIAT Investigators
- ↵*Reprint requests and correspondence: Dr. Marek Malik, Department of Cardiological Sciences, St. George’s Hospital Medical School, London SW17 0RE, United Kingdom
This substudy tested a prospective hypothesis that European Myocardial Infarct Amiodarone Trial (EMIAT) patients with depressed heart rate variability (HRV) benefit from amiodarone treatment.
The EMIAT randomized 1,486 survivors of acute myocardial infarction (MI) aged ≤75 years with left ventricular ejection fraction (LVEF) ≤40% to amiodarone or placebo. Despite a reduction of arrhythmic mortality on amiodarone, all-cause mortality was not changed.
Heart rate variability was assessed from prerandomization 24-h Holter tapes in 1,216 patients (606 on amiodarone). Two definitions of depressed HRV were used: standard deviation of normal to normal intervals (SDNN) ≤50 ms and HRV index ≤20 units. The survival of patients with depressed HRV was compared in the placebo and amiodarone arms. A retrospective analysis investigated the prospective dichotomy limits. All tests were repeated in five subpopulations: patients with first MI, patients on beta-adrenergic blocking agents, patients with LVEF ≤30%, patients with Holter arrhythmia and patients with baseline heart rate ≥75 beats/min.
Centralized Holter processing produced artificially high SDNN but accurate HRV index values. Heart rate variability index was ≤20 U in 363 (29.9%) patients (183 on amiodarone) with all-cause mortality 22.8% on placebo and 17.5% on amiodarone (23.2% reduction, p = 0.24) and cardiac arrhythmic mortality 12.8% on placebo and 4.4% on amiodarone (66% reduction, p = 0.0054). Among patients with prospectively defined depressed HRV, the largest reduction of all-cause mortality was in patients with first MI (placebo 17.9%, amiodarone 10.3%, 42.5% reduction, p = 0.079) and in patients with heart rate ≥75 beats/min (placebo 29.0%, amiodarone 19.3%, 33.7% reduction, p = 0.075). Among patients with first MI and depressed HRV, amiodarone treatment was an independent predictor of survival in a multivariate Cox analysis. The retrospective analysis found a larger reduction of mortality on amiodarone in 313 (25.7%) patients with HRV index ≤19 U: 23.9% on placebo and 17.1% on amiodarone (28.4% reduction, p = 0.15). This was more expressed in patients with first MI: 49.4% mortality reduction on amiodarone (p = 0.046), on beta-blockers: 69.0% reduction (p = 0.047) and with heart rate ≥75 beats/min: 37.9% reduction (p = 0.054).
Measurement of HRV in a large set of centrally processed Holter recordings is feasible with robust methods of assessment. Patients with LVEF ≤40% and depressed HRV benefit from prophylactic antiarrhythmic treatment with amiodarone. However, this finding needs confirmation in an independent data set before clinical practice is changed.
The European Myocardial Infarct Amiodarone Trial (EMIAT) was a randomized, double-blind placebo controlled clinical trial designed to determine whether amiodarone would reduce all-cause mortality, cardiac mortality and presumed arrhythmic death in patients surviving a recent myocardial infarction (MI) with a left ventricular ejection fraction (LVEF) ≤40% and age ≤75 years. The European Myocardial Infarct Amiodarone Trial enrolled 1,486 patients who were followed for a mean of 21 months. There was no difference between the two treatment groups with regard to all-cause mortality and cardiac mortality, but there was a significant 35% risk reduction in presumed arrhythmic deaths in patients treated with amiodarone (1). This outcome prompted several considerations.
The risk in patients with LVEF ≤40% comprises both arrhythmic and nonarrhythmic death, and some patients probably died from heart failure after having an arrhythmic death prevented by amiodarone. Others, who were similarly protected from an arrhythmic death, probably were not at high risk of dying from other causes, such as pump failure. This concept highlights the importance of developing a clinically feasible approach to select postinfarction patients at particularly high risk of predominantly arrhythmic death who would benefit from amiodarone therapy and in whom death from fatal arrhythmia would not be merely converted into death from nonarrhythmic cause.
Depressed heart rate variability (HRV) (2) is a recognized risk factor in survivors of acute MI (3–5) and, compared with depressed LVEF, predicts more specifically arrhythmic than nonarrhythmic mortality (6, 7). The design of EMIAT, therefore, included a hypothesis that a positive effect of amiodarone would be noted, particularly in the subgroup of patients with depressed HRV defined according to prospectively set criteria. Because of its association with arrhythmic events rather than with all-cause mortality, depressed HRV might identify postinfarction patients who would benefit from prophylactic antiarrhythmic therapy.
Having this in mind, this substudy of EMIAT evaluated the prospective hypothesis on the effects of amiodarone in patients with depressed HRV and investigated whether the prospectively set criteria were correct and whether the potential benefit of amiodarone might be improved by redesigning the definition of the patients at high risk of arrhythmia. The principle goal of the substudy was to confirm or reject the hypothesis that all-cause mortality in patients with depressed HRV is reduced on amiodarone.
The study utilized data collected during EMIAT and the analysis of prerandomization Holter recordings completed before breaking the code of EMIAT. The complete set of patients enrolled in the trial was analyzed, and no patient was excluded for any reasons other than nonavailability of data. The distinction between amiodarone and placebo arms of EMIAT was based on an intention-to-treat analysis, and cases of therapy discontinuation were not censored for this analysis.
Details of the trial can be found elsewhere (1). Briefly, patients were eligible for the trial if they were aged between 18 and 75 years and if LVEF, determined by multiple-gated nuclear angiography and assessed between days 5 and 21 after the index infarction, was ≤40%. The exclusion criteria of the trial (1) included documented bradycardia (<50 beats/min), second and third degree atrioventricular block and contraindications to amiodarone, such as thyroid dysfunction. Before randomization, a 12-lead electrocardiogram was made and a three-channel 24-h Holter recording was obtained. The majority of Holter recordings were obtained before hospital discharge; all recordings were obtained within 21 days of the index infarction.
For each patient enrolled in the trial, the complete follow-up period within the trial (mean 664 ± 107 days) was considered. Of the 205 trial patients who died, 31 died from a documented noncardiac cause. The other 174 deaths were assumed to be cardiac deaths. Of these, 126 were witnessed and 48 unwitnessed. Sudden death was defined as that occurring within 1 h of symptoms or in patients with no symptoms or stable symptoms and with no left ventricular failure. When unwitnessed, sudden death was defined as unexpected death in a patient known to be well (no progressive left ventricular failure) and with no symptoms or with stable cardiovascular symptoms. Of the 126 witnessed deaths, 44 were sudden, of which 41 were documented or considered to be arrhythmic. Of the 82 witnessed nonsudden deaths, six were arrhythmic. Of the 48 unwitnessed deaths, 36 were considered to be sudden and assumed to be caused by arrhythmia.
Deaths, as recorded in the database of EMIAT, were taken as the end points for this study. The classification of the deaths to noncardiac, cardiac nonarrhythmic and cardiac arrhythmic as decided by the event committee of EMIAT was not questioned in this study, and the recorded mortality categories were used.
Holter recordings and HRV assessment
The prerandomization Holter recordings were centrally analyzed using Marquette Laser Holter 8000 Systems (Marquette Medical Systems, Milwaukee, Wisconsin), and an RR interval datafile was produced for each recording listing the duration of individual RR intervals measured on a discrete scale with 1/128 s (≈7.8 ms) steps and the morphological classification of individual QRS complexes. The RR interval datafiles were transferred to another center that was responsible for assessing HRV in all recordings.
The Holter recording was considered suitable for HRV analysis if it contained ≥18 h of analyzable data defined as the sum of all noise-free RR intervals between sinus rhythm QRS complexes (that is, patients not having at least 18 h of analyzable sinus rhythm recording were excluded—e.g., patients with atrial fibrillation, pacemaker dependent cardiac rhythm, etc.). From each recording fulfilling this criterion, two HRV measures were obtained (2): standard deviation of normal to normal intervals (SDNN) was computed as the standard deviation of all analyzable sinus rhythm RR intervals, and HRV index was computed as the total number of analyzable sinus rhythm RR intervals divided by their modal frequency.
Because Holter readings performed in a central laboratory of the whole trial could not achieve the same precision as in smaller studies that do not have constraints imposed, the SDNN values were obtained after automatic “filtering” of the original RR intervals (3,8); otherwise analyzable RR intervals were excluded if they differed ≥20% from the immediately preceding analyzable RR intervals.
The analysis of the Holter recordings and the subsequent measurement of HRV was performed, and the HRV data recorded in the database of EMIAT before the code of the trial was disclosed.
In order to investigate the combined effect of depressed HRV with some other recognized risk stratifiers, and to study populations of patients in whom a positive effect of amiodarone might be expected based on previous observations in EMIAT substudies and other amiodarone trials (9–11), the following six patient categories were considered in this study:
1. All patients = all patients suitable for the study (i.e., for whom the HRV data were available),
2. patients with first MI = patients who had no history of a definite prior MI,
3. patients on beta-adrenergic blocking agents = patients who were treated with beta-blockers at the time of EMIAT enrollment,
4. patients with reduced LVEF = patients with LVEF ≤30%. A separate analysis of these patients was envisaged in the design of EMIAT,
5. patients with arrhythmia on Holter = arrhythmia signs were taken as either a mean of ventricular premature beats ≥10 per hour of the noise free section of the Holter recording or at least one triplet of ventricular premature beats within the entire noise free recording. A separate analysis of these patients was envisaged in the design of EMIAT,
6. patients with fast, baseline heart rate = patients with heart rate ≥75 beats/min recorded in the prerandomization short-term ECG recordings.
Patients with reduced LVEF, with signs of arrhythmia on Holter and with fast baseline heart rate were studied because these selectors are known to improve predictive value of HRV; patients with first MI and patients on beta-blockers were studied because previous substudies of EMIAT (9) suggested a possible positive effect of amiodarone in these populations.
To investigate the influence of the other risk factors on HRV measurements, the values of individual HRV indexes were compared between patients with and without a history of prior MI, on and off beta-blocker treatment at enrollment, with LVEF ≤30% and LVEF >30%, with and without arrhythmia signs on Holter and with resting heart rate ≥75 beats/min and <75 beats/min. Further, correlation coefficients were evaluated between individual HRV indexes and the values of LVEF and resting heart rate.
To ensure that the randomization of EMIAT did not lead to an artificial bias in the HRV values, individual HRV indexes were compared between patients of the placebo and amiodarone arms of each patient category considered in this study (as described in the previous section).
To investigate whether the preselection procedure of EMIAT (that is, selection of patients with LVEF ≤40% who are at an increased risk of follow-up mortality), as well as the selection of patients within the considered categories, influences the power of depressed HRV to predict adverse outcome, the values of individual HRV indexes were compared between patients who did and did not survive the follow-up of EMIAT. These comparisons were performed separately within the placebo and amiodarone arms of all patient categories considered.
The prospective hypothesis of EMIAT considered two definitions of depressed HRV: patients were classified as having depressed HRV if having (a) SDNN ≤50 ms or (b) HRV index ≤20 U (both definitions were envisaged to be used independently of each other).
For each of the definitions, the all-cause mortality and the arrhythmic death during follow-up were compared in the placebo and amiodarone arms of patients with depressed HRV and in the placebo and amiodarone arms of patients with preserved HRV. These comparisons were performed for all the patient categories considered.
Multivariate Cox analysis
To test whether amiodarone is an independent predictor of survival in patients with low HRV after adjusting confounding effects of other risk factors, multivariate Cox analysis was performed using amiodarone treatment, LVEF >30%, age ≤60 years, history of previous MI, baseline heart rate >75 beats/min, beta-blocker treatment and arrhythmia sign on Holter as predictors of total mortality during follow-up. The analysis was performed for the total population of the trial, patients with HRV index ≤20 U and patients with HRV ≤20 U belonging to individual subgroups, as described in the section “Patient Categories.” For the analysis in the categories, the risk predictor defining the category was omitted from the analysis.
In the complete population of patients for whom the HRV data were available, the values of SDNN and HRV index measures were sorted, and for each of these HRV indexes, two dichotomy limits alpha and beta were identified such that approximately 10% of patients had the HRV values <alpha and approximately 40% of patients had the HRV values <beta. The interval between alpha and beta was then divided into approximately 100 steps, and, for each value between alpha and beta, patients were identified with HRV not greater than the given value. For each subgroup of patients obtained in this way, the all-cause mortality was compared in the placebo and amiodarone arms. These comparisons were performed for all the patient categories considered. Based on the results of these comparisons, a retrospective readjustment of the prospectively defined dichotomy limits of SDNN and HRV index values was attempted to achieve the highest mortality reduction on amiodarone among patients with reduced HRV.
Data of continuous variables are presented as mean ± standard deviation. Comparisons of categorical variables were performed using Fisher exact test, comparisons of continuous variables using two-tail two-sample t test assuming unequal variances of compared samples, and survival characteristics were compared by log-rank test of Kaplan-Meier statistics. Excel 5.0 for Windows package was used for basic computations including the t test; other statistical tests were performed by the PC version of the SAS and Statistica packages. Although precise p values are presented in all cases, p ≤ 0.05 was considered as a statistical significance.
Of all the patients enrolled into EMIAT, a suitable Holter recording was available in 1,216. The remaining 270 patients included those for whom no RR interval data file was produced for technical reasons (e.g., Holter tape without an ECG signal, n = 225) and those for whom the RR interval data file did not fulfill the criteria for meaningful HRV assessment (n = 45). Of the RR interval files that were suitable for the analysis, 875 (72.0%), 1,061 (87.3%) and 1,179 (97.0%) had at least 23 h, 22 h and 20 h of analyzable sinus rhythm data, respectively.
Table 1shows the clinical and follow-up characteristics of patients with and without eligible Holter recordings. As no differences between these two groups were noted, it can be assumed that the patients with eligible Holter recordings are representative of the complete population of EMIAT.
Table 2shows all-cause and cardiac arrhythmic mortality in different categories of patients with eligible Holter recordings.
Heart rate variability assessment
Figure 1 shows scatter diagrams of correlations between individual HRV measures in the complete population of the study. In the complete population, the correlation coefficient between SDNN and HRV index values was r = 0.86. Although this correlation coefficient is highly statistically significant, it is not particularly strong, and Figure 1 shows that the individual measures do not replace each other. In ideal data, the scatter diagram between HRV index and SDNN should be a straight line (2). Here, however, the average quality of the RR interval data affects the statistical measures of HRV and, despite the “filtering” attempts, leads to artificially high SDNN values in a large number of cases. It should be noted that no SDNN value appears below the line of SDNN [ms] ≈2.5 × HRV index [U] that would represent a perfect relationship between the two measures.
Table 3shows the comparison of HRV measures between patients belonging and not belonging to the individual categories considered in this study. The correlation coefficients between LVEF and SDNN and HRV index were r = 0.19 and 0.22, respectively. The correlation coefficients between short-term baseline heart rate and these HRV measures were r = −0.39 and −0.42, respectively. Thus, higher HRV values were obtained in patients with slower heart rate (and consequently in patients on beta-blockers) and in patients with higher LVEF. History of previous MI did not influence the HRV values. Note that SDNN tends to be higher in patients with arrhythmia signs on Holter while the HRV index values do not differ between patients with and without arrhythmia sign. This is almost certainly because on average, the standard quality of Holter analysis is poorer in patients with arrhythmia, and this artificially increases the statistical measures. As seen in Table 3, HRV values were strongly associated with baseline heart rate dichotomized at 75 beats/min, probably because, in the patients with reduced LVEF, sympathetic overdrive is frequently the cause of both reduced HRV and increased heart rate.
While in all other categories, the randomization did not introduce any bias in terms of HRV values, a trend towards a possible bias was seen in patients with increased heart rate among whom patients randomized to amiodarone tend to have lower HRV values.
Table 4shows the comparison of HRV values between survivors and nonsurvivors in individual patient categories. Generally, depressed global 24-h HRV was statistically associated with mortality during follow-up in both placebo and amiodarone arms of individual patient categories. This association was weakest among patients who were on beta-blockers at the time of EMIAT randomization, probably because these patients represent a low risk group among other patients of the trial (the treatment with beta-blockers was not randomized in the trial and was based on clinical judgement). There were important practical differences between the individual methods for HRV assessment.
Table 5shows the number of patients in individual patient categories who fulfilled the prospectively set criteria of depressed HRV. Because of the influence of RR interval data quality, the criterion SDNN ≤50 ms was too tight and selected very few patients, indeed. On the contrary, the criterion HRV index ≤20 U performed as envisaged in the design of the prospective substudy and, in most patient categories, selected approximately 30% of patients. Only in patients with LVEF ≤30% and in patients with baseline heart rate ≥75 beats/min were approximately 40% patients selected.
Table 6shows the comparison of all-cause mortality in placebo and amiodarone arms of prospectively selected patients with depressed HRV in individual patient categories. Although the differences did not reach statistical significance, the placebo arm mortality of patients with HRV index ≤20 U was substantially higher than the amiodarone arm mortality in all patient categories. Corresponding curves of arrhythmia-free survival are shown in Figure 2. Table 6 also shows the comparisons of all-cause mortality in placebo and amiodarone arms of prospectively selected patients with preserved HRV. It should be noted that the potential benefit of amiodarone, which was previously noted in patients with first MI and in patients on beta-blockers (9), was almost exclusively manifested in patients with reduced HRV because the mortality levels on placebo and on amiodarone are practically identical in the subgroup of these patients with preserved HRV.
The reduction of the cardiac arrhythmic mortality on amiodarone was very dramatic in patients with HRV index ≤20 U. Although the cardiac nonarrhythmic mortality increased on amiodarone, the increase was lower than the reduction of arrhythmic mortality (only in patients on beta-blockers were both the cardiac arrhythmic and cardiac nonarrhythmic mortalities decreased).
Multivariate Cox analysis
Statistical significances of predictor association with all-cause mortality in individual Cox analyses performed are shown in Table 7. Compared with the total population, a trend is very obvious for amiodarone being an independent predictor of survival in patients with reduced HRV (HRV index ≤20 U). Statistical significance for amiodarone treatment being an independent predictor of survival has been reached among patients without a history of previous MI and with reduced HRV.
Of the complete population of the study, 138 (11.3%) and 463 (38.1%) patients had SDNN measurements ≤55 ms and ≤80 ms, respectively, and 136 (11.2%) and 471 (38.7%) patients had HRV indexes ≤15 U and ≤22 U, respectively. Consequently, the ranges of SDNN dichotomy limits from 55 to 80 ms and of HRV index dichotomy limits from 15 to 22 U were used in the retrospective study.
Figure 3 shows the dependence of mortality in placebo and amiodarone arms on the dichotomy limit used to define depressed HRV. In the total population stratified according to the values of the HRV index, both the cardiac arrhythmic and nonarrhythmic mortality increased in the placebo arm when selecting patients with more and more depressed HRV. On the contrary, only the cardiac nonarrhythmic mortality increased in the amiodarone arm while the arrhythmic mortality was not only substantially lower than in the placebo arm, but also independent of the depressed HRV. Similar trends were seen in other patient categories, and the difference between placebo and amiodarone arms was most pronounced in patients on beta-blockers at the time of EMIAT randomization.
The reduction of all-cause mortality on amiodarone as a function of HRV dichotomy is shown in Figure 4, which also shows the results of statistical comparisons of the survival in groups with differently depressed HRV. When selecting more and more patients with depressed HRV (that is, when loosening the criterion defining depressed HRV), the statistical comparisons become more and more powerful. In this, an “optimum” dichotomy limit of HRV index ≤19 U was identified (see Fig. 4). Table 8and Figure 5 show that when using this dichotomy limit, the mortality reduction on amiodarone becomes significant in patients with first MI and in patients on beta-blockers (the dichotomy selects approximately 25% of these patients) and almost reaches statistical significance in patients with increased baseline heart rate (selecting approximately 40% of these patients). Table 8 also shows that the mortality levels on placebo and amiodarone are much less different in patients with HRV index >19 units and that, similar to the observation made with the prospective substudy, the potentially positive effect of amiodarone in patients with first MI and in patients on beta-blockers is solely due to the mortality reduction among patients with HRV index ≤19 U.
A retrospective attempt to optimize the dichotomy of the SDNN measure does not yield any generally useful selection criterion. Up to the limit of SDNN ≤70 ms, the difference between placebo and amiodarone survival is statistically significant, while the difference in other patient categories remains well below statistical significance (Fig. 4). The criterion SDNN ≤70 ms selects approximately 24% of patients on beta-blockers. Again, this lack of any retrospective improvement of the SDNN dichotomy limit almost certainly reflects the quality of data that were used to compute the SDNN values.
Being the first prospectively set-up investigation of the relationship between depressed HRV and prophylactic treatment in a large population of survivors of acute MI, this study has implications for the assessment of HRV in large clinical populations, for its predictive value as well as for the identification of patients who might benefit from prophylactic antiarrhythmic treatment.
Implications of the study
The analysis of the placebo arm of the population of EMIAT showed that the predictive power of depressed HRV is preserved when patients with depressed LVEF are considered. The HRV values were significantly lower in nonsurvivors compared with survivors. The total follow-up mortality of 13.4% in the placebo arm of this substudy increased to 20.0%, 23.8% and 24.1% when selecting the lowest tertile, quartile and quintile of patients with depressed HRV, respectively (HRV index ≤21 U, ≤19 U and ≤17.5 U, respectively). The analysis also showed that, in the placebo arm, the mortality increase with depressed HRV values was mainly caused by an increase in the cardiac arrhythmic mortality. The nonarrhythmic mortality was also increased, but to a lesser extent.
The test of prospective risk stratification based on HRV values was successful. The criterion HRV index ≤20 U selected approximately 30% of the total population of the substudy with a 70% increase of the placebo arm all-cause mortality and a 90% increase of placebo arm arrhythmic mortality (compared with all-cause and arrhythmic mortality in the total placebo arm). Although the mortality reduction on amiodarone in the prospectively stratified population did not reach the level of statistical significance, the hypothesis that these patients with depressed HRV benefit from prophylactic antiarrhythmic treatment (e.g., with amiodarone) should not be denied because EMIAT was not powered for testing this hypothesis, and the reduction of all-cause mortality from 22.8% on placebo to 17.5% on amiodarone is impressive. On the contrary, we observed a mild increase of mortality on amiodarone (12.8%) compared with placebo (9.5%) in patients with prospectively defined preserved HRV.
The retrospective analysis suggested that an even higher reduction of all-cause mortality with antiarrhythmic treatment might be expected in patients at greater arrhythmia risk, e.g., when selecting 20%–25% patients with lowest HRV values. This is especially the case when depressed HRV is combined with other arrhythmia risk factors, such as arrhythmia signs on Holter (12–14) or increased heart rate (15,16).
Finally, from the technical point of view of HRV assessment, the study indicates that, in long-term Holter recordings of a large multicenter trial, the quality of data required to obtain the correct values of statistical measures of HRV is difficult to achieve. In such a setting, more robust methods for HRV assessment (such as the geometrical HRV index method) are probably more appropriate because they are not so highly dependent on precise localization and morphological classification of every single QRS complex. The usual misrecognitions of computerized Holter analysis are not only linked to improper handling of intervals containing ectopic beats but, possibly more frequently, to incorrect interpretations of myopotentials and tall T waves as QRS complexes as well as to omission of low voltage QRS complexes. This created both artificially short and artificially long RR intervals, which distorted the statistical measures of HRV despite filtering techniques that excluded obvious jumps in the RR interval series; it is know that these procedures of off-line post-process “data improvement” are frequently of little help (8).
Relation to previous reports
In many previous reports, depressed HRV was reported to be associated with adverse outcome in survivors of acute MI (3–5,17–19). This is confirmed by our observation, which shows that depressed HRV remains a risk predictor in a population prestratified according to other recognized risk factors (e.g., placebo all-cause mortality 22.8% vs. 9.5% in patients with prospectively defined reduced and preserved HRV—HRV index dichotomized at 20 U).
The prospective hypothesis of this study was based on previous observations showing that, compared with low LVEF, depressed HRV predicts arrhythmic postinfarction complications more strongly than nonarrhythmic events (6) and that patients can be selected who are at predominantly arrhythmic risk (20). The results of our retrospective substudy are in concordance with reports showing that dichotomy limits optimized in a univariate setting might need to be changed when the combination of more risk factors is considered (21). As the population of EMIAT was prestratified in respect to low LVEF, it is not surprising that the prospectively set dichotomy limit of HRV index ≤20 U that was taken from studies of a total postinfarction population (5) is not quite optimum and that other cut-off limits led to a more striking difference between placebo and amiodarone mortality.
The fact that depressed HRV becomes a more potent risk predictor when combined with other risk factors has been already observed (2,22,23). In particular, the combination of depressed HRV with increased heart rate has been reported to increase the predictive power of both factors (16), and similar findings have been noted for combinations of HRV with arrhythmia signs (20).
Beta-blockers and amiodarone have been reported to have synergistic or additive pharmacological effects with respect to their ability to suppress arrhythmias (24,25). The combination of the two drugs might be dangerous in patients with low heart rates or atrioventricular conduction disturbances (26). The combined effect of amiodarone and beta-blockers in EMIAT has been discussed in more detail elsewhere (9). In brief, the finding that amiodarone is particularly effective in patients on beta-blockers must be interpreted with caution because a positive interaction between amiodarone and beta-blockers was not envisaged prospectively. As the patients of EMIAT were not randomized with respect to treatment with beta-blockers, it is possible that patients prescribed beta-blockers on clinical grounds are those with lower risk of nonarrhythmic mortality in whom the antiarrhythmic action of amiodarone might be particularly effective. Also, in addition to an adrenergic inhibitory action, amiodarone has direct effects on the action potential of sinus nodal cells and, independently of beta-blocker treatment, it may have a positive effect due to the reduction of heart rate (27,28). At the same time, it is interesting to note that the potential benefit of the combination of amiodarone with beta-blockers is practically restricted to patients with reduced HRV.
Similar caution is needed in interpreting the observation that amiodarone is particularly effective in patients with first MI and that HRV is particularly potent in selecting those of these patients who benefit from amiodarone treatment. The differences in response to amiodarone between patients without and with a history of previous MI have not been seen in other studies of the drug, and the predictive power of HRV has not been previously observed to differ in patients with first and subsequent MI.
European Myocardial Infarct Amiodarone Trial was not designed and powered to test the prospective hypothesis of this substudy, and the prospective analysis did not reach statistical significance. In the light of this, our finding is clearly insufficient to change the clinical practice of prophylactic management of survivors of acute MI. The HRV data of other amiodarone trials in infarction survivors, where available, need to be urgently analyzed in a prospective fashion, preferably using the dichotomies of HRV index ≤20 U and ≤19 U. If the observations made here are reproduced, clinical practice may indeed need to be changed.
Retrospective subgroup analysis, even when based on baseline parameters and when supported by previously made observations, is, at best, hypothesis generating. Our retrospective analysis might, therefore, serve merely as a basis for future prospective trials where amiodarone or any other antiarrhythmic agent could be tested in patients with a recent MI. Those results of the retrospective analysis that are not supported by previous independent observations require further detailed analysis. This is particularly the case with the predictive power of depressed HRV in patients with first MI.
- European Myocardial Infarct Amiodarone Trial
- heart rate variability
- left ventricular ejection fraction
- myocardial infarction
- standard deviation of normal to normal intervals
- Received June 16, 1998.
- Revision received October 15, 1999.
- Accepted December 29, 1999.
- American College of Cardiology
- Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology
- Cripps T.R,
- Malik M,
- Farrell T.G,
- Camm A.J
- ATRAMI Investigators,
- La Rovere M.T,
- Mortara A,
- Bigger J.T Jr.,
- et al.
- Odemuyiwa O,
- Malik M,
- Farrell T,
- Bashir Y,
- Poloniecki J,
- Camm A.J
- Schwartz P.J,
- La Rovere M.T,
- Vanoli E
- ↵Janse MJ, Malik M, Camm AJ, Julian DG, Frangin GA, Schwartz PJ. Identification of post-acute myocardial infarction patients with potential benefit from treatment with amiodarone. Eur Heart J 1998;19:85–95.
- Bigger J.T Jr.,
- Fleiss J.L,
- Kleiger R,
- Miller J.P,
- Rolnitzky L.M
- Nul D.R,
- Doval H.C,
- Grancelli H.O,
- et al.
- Moss A.J,
- Davis H.T,
- De Camilla J,
- Bayer L.W
- Copie X,
- Hnatkova K,
- Staunton A,
- Fei L,
- Camm A.J,
- Malik M
- Farrell T.G,
- Bashir Y,
- Cripps T,
- et al.
- Bigger J.T,
- Fleiss J.L,
- Steinman R.C,
- Rolnitzky L.M,
- Kleiger R.E,
- Rottman J.N
- Bigger J.T,
- Fleiss J.L,
- Rolnitzky L.M,
- Steinman R.C
- Hartikainen J.A,
- Malik M,
- Staunton A,
- Poloniecki J,
- Camm A.J
- Redvood S,
- Odemuyiwa O,
- Hnatkova K,
- et al.
- Zuanetti G,
- Neilson J.M,
- Latini R,
- Santoro E,
- Maggioni A.P,
- Ewing D.J
- Marcus F.I
- Singh B.N
- Kodama I,
- Kamiya K,
- Toyama J