Prognostic Significance of Heart Rate Variability in Post–Myocardial Infarction Patients in the Fibrinolytic Era
The GISSI-2 Results
Background Studies performed before the introduction of fibrinolysis showed that a low heart rate variability (HRV) is associated with higher mortality in post–myocardial infarction (MI) patients. We evaluated whether HRV adds information relevant to risk stratification in patients treated with fibrinolysis as well.
Methods and Results From 24-hour ECG recordings obtained at discharge in patients treated with recombinant tissue-type plasminogen activator or streptokinase, we measured several time-domain indexes of HRV: standard deviation (SDNN), root-mean-square of successive differences (RMSSD), and number of RR interval increases >50 ms (“NN50+”). The prognostic value of HRV for total and cardiovascular mortality was assessed. Of 567 patients with valid recordings, 52 (9.1%) died during the 1000 days of follow-up, 44 (7.8%) of cardiovascular causes. All indexes of low HRV were able to identify patients (16% to 18% of total population) with a higher total mortality (20.8% to 24.2% versus 6.0% to 6.8%, depending on index used). The independent predictive value of low HRV was confirmed by the adjusted analysis with the following relative risks: NN50+, 3.5 (95% CI, 1.9 to 6.7); SDNN, 3.0 (95% CI, 1.55 to 5.9); and RMSSD, 2.8 (95% CI, 1.5 to 5.3). Advanced age, previous MI, Killip class at entry, and use of digitalis were also independent predictors. Similar data were obtained for cardiovascular mortality.
Conclusions Time-domain indexes of HRV retain their independent prognostic significance even in post-MI patients of all ages treated with fibrinolysis.
Prognosis after MI has improved steadily with modern treatment, particularly fibrinolytic therapy. There is still a sizable mortality, however, both during the acute phase and in the longer term, and not all factors relating to prognosis have been elucidated. Since the landmark work of Kleiger et al,1 several studies in the past decade have evaluated the role of a low HRV in predicting subsequent mortality after MI.2 3 4 5 Most studies, however, were completed before the introduction of routine fibrinolysis in the treatment of acute MI. In an attempt to reassess the prognostic value of several risk variables in post-MI patients in the fibrinolytic era,6 7 in this study we evaluated whether HRV, measured with several time-domain indexes, would help to obtain a better risk stratification at discharge for patients treated with fibrinolysis during the acute phase of MI.
The study population was taken from patients in the GISSI-2 trial who had survived until discharge from hospital. The GISSI-2 protocol has been described in detail elsewhere.8 It was a large multicenter trial of different fibrinolytic agents given during the acute phase of MI and involved 12 490 patients. From each center, a smaller number of patients was randomly selected for quality control tests, including a 24-hour ECG recording and two-dimensional echocardiogram. Information on each patient was available with regard to the type of fibrinolytic therapy received (streptokinase 1 500 000 U over 30 to 60 minutes or TPA 100 mg over 3 hours, with or without subcutaneous heparin 12 500 U twice daily within 6 hours of the onset of symptoms); the clinical history of chronic ischemic heart disease (previous angina or previous MI), diabetes, hypertension, and hypercholesterolemia; events and complications during and after the acute phase of MI; evaluation of left ventricular function by echocardiography; and other drug therapy given. The ECG recordings obtained at discharge and sent by the participating centers for quality control of arrhythmia counts (about 10% of total population) were used for the analysis of HRV as described below. The patients were then followed for up to 1000 days after their acute episodes. Follow-up was by the attending physician for the first 180 days as required by the main GISSI-2 protocol, and thereafter through the local census office of the town of residence of the patient whenever information on the cause of death was available. The investigators were blind with regard to the mortality results until after the HRV analysis had been completed.
Measurement of HRV
All tapes were played through a Pathfinder arrhythmia analyzer (Reynolds Medical Ltd); by use of the time reference track on each tape, the replay speed was controlled within 0.5%. The indexes used in this study were the SDNN; the RMSSD; and NN50+, the index developed by two of us (J.M.M.N., D.J.E.) several years ago9 10 that represents an analogue to the pNN50 later derived by Bigger et al,2 3 with the differences that the results are expressed in absolute numbers (“counts” or “NN50”) instead of relative (“pNN50”) to the total number of RR intervals and that only increases by 50 ms, “NN50+,” not the decreases, are taken into account. During analysis, the ECG waveform was constantly monitored. The 24-hour RR interval tachograms were also scanned visually by the operator, with manual editing. The total time excluded from the analysis either automatically or manually was aggregated, and the percentage of the tape actually analyzed was determined. We excluded recordings not in sinus rhythm and those of <22 hours' duration to avoid any major confounding effect due to day versus night difference in HRV and those with <90% of recordings suitable for analysis.
As in other retrospective studies on the same topic,2 3 patients were divided post hoc into “low-HRV” and “high-HRV” groups according to a cutoff point that allowed the best discrimination for subsequent mortality in univariate analysis between the two groups, with search for limits for dichotomization restricted to 10% low, 90% high and 35% low, 65% high. The χ2 statistic was used to test the statistical significance of differences for low- and high-HRV groups in the prevalence of concomitant risk factors for mortality.
Data for mortality up to 1000 days of follow-up after discharge from the hospital were evaluated with odds ratios and their 95% CIs for mortality in patients with low versus high HRV (Mantel-Haenszel-Peto procedure).
The Cox proportional-hazards model was used to assess the independent prognostic value of each HRV index on subsequent mortality. The following variables were considered in the model: sex; age (<70 and ≥70 years); previous MI; Killip class at entry; signs or symptoms of late left ventricular dysfunction; history of diabetes and of hypertension; non–Q-wave MI; presence of frequent ventricular arrhythmias (>10 PVCs per hour); ineligibility for exercise test; use at discharge of β-blockers, digitalis, and antiarrhythmic agents; and mean heart rate (≤75 versus >75 bpm) at Holter recording. These variables were selected because they were shown, on the basis of previous data, to be independently associated with total mortality by the main analysis of the GISSI-2 database7 or were presumed to affect the data of HRV. The results of the multivariate analysis are presented in terms of relative risks with their 95% CIs.
A total of 567 patients had 24-hour ECGs recorded at discharge (median, 13 days; range, 4 to 24 days after the acute MI) that provided valuable data. For these recordings, the median duration of the tapes was 23 hours and the median percentage of recordings analyzed was 97%. The other tapes were excluded because of recording quality too poor to analyze, nonsinus rhythm, too many ectopic beats, or short tape. The clinical characteristics of the 567 patients are shown in Table 1⇓. These did not differ significantly from those of patients whose tapes were rejected, and they were representative of the GISSI-2 population that underwent 24-hour ECG recording at discharge.
HRV and Prognosis
During the 1000-day period of follow-up, 52 patients (9.2%) died, 44 of them of cardiovascular causes. Information on mode of death (sudden versus nonsudden) was available only in a minority of the patients, so that data according to mode of death were not analyzed. As shown in Table 1⇑, people who died during follow-up were usually older, more frequently had had a previous MI or suffered a transmural MI, and had signs of congestive heart failure at discharge. There was no relation between indexes of HRV and interval between onset of MI and thrombolysis or date of recordings. Table 2⇓ shows the prevalence of low HRV in different subgroups of patients for the three indexes according to the cutoff values that best discriminated, on univariate analysis, patients belonging to the two groups with “low HRV” and “high HRV.” Total mortality for patients with low versus high HRV were as follows: NN50+, 24.2% versus 6.3% (odds ratio, 8.55; 95% CI, 3.9 to 18.6); RMSSD, 20.8% versus 7.9% (odds ratio, 5.4; 95% CI, 2.5 to 11.5); and SDNN, 23.7% versus 6.3% (odds ratio, 8.0; 95% CI, 3.7 to 17.3). Receiver operator characteristic curves for all three indexes were similar (Fig 1⇓). Survival curves for patients with low and high HRV according to the three indexes are shown in Fig 2⇓.
After correction for the confounding roles of sex; age; previous MI; Killip class at entry; late left ventricular dysfunction; diabetes; hypertension; non–Q-wave MI; presence of ventricular arrhythmias; use at discharge of β-blockers, digitalis, and antiarrhythmic agents; and average heart rate, all the indexes used maintained their prognostic significance. Relative risks (95% CI) for the three indexes were as follows: NN50+, 3.5 (1.9 to 6.7); SDNN, 3.0 (1.5 to 5.9); and RMSSD, 2.8 (1.5 to 5.3). In the same models, other variables such as age >70 years, previous MI, and use of digitalis at discharge were also significant. When the analysis was limited to patients who died of cardiovascular causes (n=44), similar data were obtained: relative risks (95% CI) for patients with low HRV were NN50+, 4.0 (2.0 to 7.7); RMSSD, 3.8 (1.9 to 7.1); and SDNN, 2.6 (1.3 to 5.3). Thus, both broad-based and parasympathetic-specific indexes of HRV were independent predictors of total and cardiovascular mortality.
In this study, we demonstrated that a low HRV measured at discharge by several time-domain indexes retains its independent prognostic significance in patients treated with fibrinolysis during the acute phase and may be useful for risk stratification of post-MI patients.
HRV as a Marker of Autonomic Balance
The critical role of the parasympathetic and sympathetic nervous systems in influencing vulnerability to lethal arrhythmias in ischemic heart disease has been well established both in experimental animals and in the clinical setting,11 suggesting that markers of autonomic nervous system imbalance may be useful for risk stratification of post-MI patients.1 2 3 4 5 Since frequency of heartbeat is under tight control by both limbs of the autonomic nervous system, several indexes were developed to quantify parasympathetic and sympathetic tone by measuring the different components of the beat-by-beat variability of heart rate.12 The time-domain indexes used in this study were a broad-band index of autonomic modulation (SDNN) and more specific indexes of parasympathetic tone (NN50+ and RMSSD).13 All three indexes of HRV were lower in patients with major risk factors, including advanced age, larger infarcts, and signs of left ventricular dysfunction at entry or at discharge (Table 2⇑), thus confirming that the pathophysiological derangements accompanying or consequent to infarction strongly influence the activity of the autonomic nervous system.
HRV and Prognosis After MI: Previous Investigations
Kleiger et al1 reported a significant association between low HRV (measured as SD of RR intervals in 24 hours) and a higher mortality in a retrospective study on a database of 815 recordings from patients enrolled in the Multicenter Post-Infarction Program. They identified a subgroup with low HRV (SD, <50 ms) with a 3.4-fold increase in the risk of death in univariate analysis compared with the patients with higher HRV. This increase remained significant (relative risk, 1.7) after adjustment for several risk variables, including age, New York Heart Association class, rales in the coronary care unit, EF, and presence of frequent (≥10 PVCs per hour) ventricular arrhythmias. Since the index used in that study, the SD of 24-hour RR intervals, is a relatively nonspecific index of autonomic function, the same database was later reanalyzed (with minor differences in number of recordings for technical reasons) with more sophisticated indexes of HRV, both in the frequency domain2 and the time domain,3 to test the hypothesis that a better identification of high-risk individuals would be obtained with the use of more specific techniques. All these studies have confirmed and extended the value of HRV as an index of risk for mortality after MI, but the adjusted relative risks of total mortality were quite similar to those observed in the previous study, with values between 1.6 and 2.3 for frequency-domain variables and 1.3 and 2.1 for time-domain variables. The significance of HRV as a risk factor has also been supported in other studies of post-MI patients. Kleiger et al4 showed in a study performed on the database of patients enrolled in the Multicenter Diltiazem Post-Infarction Trial that SD predicts mortality, and Farrell et al5 showed that their “triangular index,” a broad-based measure of 24-hour RR-interval dispersion, also predicted cardiac mortality or arrhythmic events in a series of consecutive patients admitted at one hospital. In all these studies, HRV was strongly linked to ventricular function. The relative impact on prognosis of measures of HRV and of ventricular function, such as EF, has been very variable, depending on population characteristics, cutoff points for EF and HRV, and additional variables included in the multivariate model. For example, in the study by Kleiger et al,1 EF <30% appeared to be somewhat more potent in predicting mortality than SD in the multivariate analysis; conversely, in the study by Farrell et al,5 EF <40% was excluded from the multivariate model for both total events and arrhythmic events.
HRV and Prognosis After MI: the Present Study
Our study differs from previous investigations in two main respects. First, all patients of our study were treated with fibrinolysis (either TPA or streptokinase) during the acute phase. Second, no age limits were used in the enrollment of patients, so that the data obtained in this study are applicable to post-MI patients of all ages. On the other hand, a limitation of our study is represented by the fact that mode of death (sudden versus nonsudden) was not retrieved. This may theoretically reduce the prognostic power of a test that would identify patients at risk for arrhythmic death. However, in the two articles by Bigger et al,2 3 the relative risks of several frequency- and time-domain indexes were very consistent independently of the type of death (all deaths, cardiac, or arrhythmic) taken as end point, suggesting that, in the clinical setting, the prognostic power of HRV indexes may be unaffected by the mode of death under scrutiny. A second limitation of our study is that we used only time-domain indexes of HRV and did not evaluate the significance of frequency-domain indexes12 of HRV. Thus, we cannot draw any conclusion as to the comparative efficiency of frequency-domain indexes in identifying patients at risk.
In the present study, we found that subjects with low HRV had a severalfold increased risk of dying during the follow-up of 1000 days, which was similar for all three time-domain indexes used. The population identified as “high risk” in our study, according to the value of HRV, represented 16% to 17% of total population. These values lie between the 10.1% identified as being at risk of death on the basis of ultralow-frequency power in the study by Bigger et al2 and the 27.1% identified as being at risk on the basis of their HRV indexes in the study by Farrell et al.5 The multivariate analysis indicated that, for all indexes of HRV, this increase was only partially due to other risk variables associated with low HRV, such as advanced age and history of previous MI, whereas measures of left ventricular dysfunction were no longer significant. These data, which are consistent with those obtained by Farrell et al,5 give some further insights into the complex relationship and relative importance of HRV and of left ventricular dysfunction for prognosis. Thus, all time-domain indexes used in the study had an independent prognostic power in identifying patients likely to die. The cutoff value for SDNN that was used to separate groups at low and high risk of mortality in our study (70 ms) was higher than the 50 ms identified in the study by Kleiger et al.1 Using a 50-ms cutoff in our study reduces the population identified at risk to 7.8% of total patients without modifying the prognostic power (odds ratio, 7.8 for 50 ms versus 8.0 for 70 ms). Differences in the population under analysis (inclusion of patients with advanced age, widespread use of thrombolysis, etc) may underline these slight discrepancies. Also, since these data were obtained in a retrospective analysis with post hoc choice of best cutoff point, they warrant prospective investigation in patients now suffering acute MI and treated with even newer therapies such as novel fibrinolytic regimens14 and ACE inhibitors.15
An accurate risk stratification of patients before discharge from the hospital has been recognized for a long time as a critical step in optimizing the management of patients after MI both in terms of individual outcome and in terms of social costs of diagnostic and therapeutic interventions. In this respect, information obtained from noninvasive, simple tests will likely remain the focus of the diagnostic approach to the universe of post-MI patients. Since the development of adequate hardware and software for analyzing Holter recordings, the measurement of HRV from 24 ECG recordings is viewed as an important addition for a better risk stratification of post-MI patients.
Results from this study indicate that this technique may be useful even in the fibrinolytic era,16 at a time in which the value of several diagnostic tools established in the 1970s and 1980s has been reassessed.6 7 Future prospective studies will be needed to test the hypothesis that mortality of the patients identified as being at high risk may be reduced by appropriate pharmacological interventions.
This study was supported in part by a grant from the British Heart Foundation. Dr Zuanetti is the recipient of a Maurelio Caniato fellowship award. We thank Gail Borthwick for careful analysis of the ECG recordings and Luisa Galbiati for secretarial assistance.
Selected Abbreviations and Acronyms
|HRV||=||heart rate variability|
|NN50+||=||number of RR interval increases >50 ms|
|PVC||=||premature ventricular contractions|
|RMSSD||=||root-mean-square of successive differences|
|SDNN||=||SD of sinus RR intervals|
|TPA||=||tissue-type plasminogen activator|
↵1 *A complete list of the investigators can be found in Reference 8.
2 Presented in part at the 65th Scientific Sessions of the American Heart Association, New Orleans, La, November 16-19, 1992.
- Received October 12, 1995.
- Revision received January 17, 1996.
- Accepted February 1, 1996.
- Copyright © 1996 by American Heart Association
Bigger JT, Fleiss JL, Steinman RC, Rolnitzky LM, Kleiger RE, Rottman JN. Frequency domain measures of heart period variability and mortality after myocardial infarction. Circulation. 1992;85:164-171.
Kleiger RE, Miller JP, Bigger JT, Rolnitzky LM, and the Multicenter Diltiazem Postinfarction Trial (MDPIT) Research Group. Cycle length variability predicts cardiac mortality independent of diltiazem therapy. J Am Coll Cardiol. 1990;15:190A. Abstract.
Farrell TG, Bashir Y, Cripps T, Malik M, Poloniecki J, Bennett ED, Ward DE, Camm AJ. Risk stratification for arrhythmic events in postinfarction patients based on heart rate variability, ambulatory electrocardiographic variables and the signal-averaged electrocardiogram. J Am Coll Cardiol. 1991;18:687-697.
Maggioni AP, Zuanetti G, Franzosi MG, Rovelli F, Santoro E, Staszewsky L, Tavazzi L, Tognoni G. Prevalence and prognostic significance of ventricular arrhythmias after acute myocardial infarction in the fibrinolytic era: GISSI-2 results. Circulation. 1993;87:312-322.
Volpi A, De Vita C, Franzosi MG, Geraci E, Maggioni AP, Mauri F, Negri E, Santoro E, Tavazzi L, Tognoni G. Determinants of 6-month mortality in survivors of myocardial infarction after thrombolysis: results of the GISSI-2 data base. Circulation. 1993;88:416-429.
Ewing DJ, Neilson JMM, Travis P. New method for assessing cardiac parasympathetic activity using 24 hour electrocardiograms. Br Heart J. 1984;52:396-402.
Neilson JM. High speed analysis of ventricular arrhythmias from 24 hour recordings. Comput Cardiol. 1974;55-61.
Schwartz PJ, Priori SG. Sympathetic nervous system and cardiac arrhythmias. In: Zipes DP, Jalife J, eds. Cardiac Electrophysiology: From Cell to Bedside. Philadelphia, Pa: WB Saunders Co; 1990:330-342.
Malliani A, Pagani M, Lombardi F, Cerutti S. Cardiovascular neural regulation explored in the frequency domain. Circulation. 1991;84:482-492.
Myers GA, Martin GJ, Magid NM, Barnett PS, Shaad JW, Weiss JS, Lesch M, Singer DH. Power spectral analysis of heart rate variability in sudden cardiac death: comparison to other methods. IEEE. 1986;BME-33:1149-1156.