RR Variability in Healthy, Middle-Aged Persons Compared With Patients With Chronic Coronary Heart Disease or Recent Acute Myocardial Infarction
Background The purpose of this investigation was to establish normal values of RR variability for middle-aged persons and compare them with values found in patients early and late after myocardial infarction. We hypothesized that presence or absence of coronary heart disease, age, and sex (in this order of importance) are all correlated with RR variability.
Methods and Results To determine normal values for RR variability in middle-aged persons, we recruited a sample of 274 healthy persons 40 to 69 years old. To determine the effect of acute myocardial infarction RR variability, we compared measurements of RR variability made 2 weeks after myocardial infarction (n=684) with measurements made on age- and sex-matched middle-aged subjects with no history of cardiovascular disease (n=274). To determine the extent of recovery of RR variability after myocardial infarction, we compared measurements of RR variability made in the group of healthy middle-aged persons with measurements made in 278 patients studied 1 year after myocardial infarction. We performed power spectral analyses on continuous 24-hour ECG recordings to quantify total power, ultralow-frequency (ULF) power, very-low-frequency (VLF) power, low-frequency (LF) power, high-frequency (HF) power, and the ratio of LF to HF (LF/HF) power. Time-domain measures also were calculated. All measures of RR variability were significantly and substantially lower in patients with chronic or subacute coronary heart disease than in healthy subjects. The difference from normal values was much greater 2 weeks after myocardial infarction than 1 year after infarction, but the fractional distribution of total power into its four component bands was similar for the three groups. In healthy subjects, ULF power did not change significantly with age; VLF, LF, and HF power decreased significantly as age increased. Patients with chronic coronary heart disease showed little relation between power spectral measures of RR variability and age. Patients with a recent myocardial infarction showed a strong inverse relation between VLF, LF, and HF power and age and a weak inverse relation between ULF power and age. ULF power best separates the healthy group from either of the two coronary heart disease groups. Differences in RR variability between men and women were small and inconsistent among the three groups.
Conclusions All measures of RR variability were significantly and substantially higher in healthy subjects than in patients with chronic or subacute coronary heart disease. The difference between healthy middle-aged persons and those with coronary heart disease was much greater 2 weeks after myocardial infarction than 1 year after infarction, but the fractional distribution of total power into its four component bands was similar for the healthy group and the two coronary heart disease groups. Values of RR variability previously reported to predict death in patients with known chronic coronary heart disease are rarely (≈1%) found in healthy middle-aged individuals. Thus, when measures of RR variability are used to screen groups of middle-aged persons to identify individuals who have substantial risk of coronary deaths or arrhythmic events, misclassification of healthy middle-aged persons should be rare.
In September 1993, the Cardiovascular Technology Assessment Committee of the American College of Cardiology published a position statement on heart rate (actually RR interval) variability for risk stratification.1 The committee concluded that, although RR variability has the potential to become an important prognostic tool, there are substantial unanswered questions that preclude RR variability from being a standard clinical test at present. One important area for future research proposed by the committee was the determination of normal values for various time- and frequency-domain measures of RR variability as a reference for diseased groups. The committee pointed out the importance of evaluating healthy persons of age and sex similar to those of the samples of patients with cardiovascular diseases. The purpose of this investigation was to establish normal values for middle-aged persons and compare them with the values found in patients early and late after myocardial infarction. Our study was designed to clarify the extent to which age, sex, and presence or absence of coronary heart disease influence measures of RR variability in middle-aged persons. Our hypothesis was that presence or absence of coronary heart disease, age, and sex (in this order of importance) are all correlated with RR variability.
To determine normal values for RR variability in middle-aged persons so that we could compare them with values from patients who had coronary heart disease, we recruited a sample of healthy persons matched by age and sex to patients who had had a myocardial infarction and participated in the Multicenter Post Infarction Program (MPIP).2 3 Employees scheduled for periodic physical examinations in an employee health department in New York City and persons who responded to newspaper advertisements in St Louis were screened for the study. Men and women 40 to 69 years old were potentially eligible. Employees in New York (n=119) who were healthy with respect to history, physical examination, chest radiograph, 12-lead ECG, and routine blood chemistry analyses and were taking no drugs were asked to join the study. Subjects in St Louis (n=155) did not have chest radiographs or blood chemistry analyses. Thus, the designation “healthy” denotes not only the absence of coronary heart disease but also lack of any other diseases that can be detected by these tests. This study was approved by the Institutional Review Boards of the Columbia-Presbyterian Medical Center and the Jewish Hospital of St Louis. Each participant signed an informed consent form that included permission to record and use some basic medical information from the data forms and to have one 24-hour continuous ECG recording during normal daily activities. During a 4-year period (November 1988 to February 1993), 274 healthy subjects were recruited.
To determine the effect of acute myocardial infarction on RR variability, we compared measurements of RR variability from the MPIP patients with measurements made in age- and sex-matched middle-aged subjects with no history of cardiovascular disease. To determine the extent of recovery of RR variability after myocardial infarction, we compared measurements of RR variability made in the sample of healthy middle-aged persons with measurements made 1 year after myocardial infarction in patients who enrolled in the Cardiac Arrhythmia Pilot Study (CAPS).4 5
The MPIP sample was selected to be representative of the entire post–myocardial infarction population. In a previous study, we found that the CAPS and MPIP samples were comparable for measures of RR variability at baseline, 2 to 4 weeks after myocardial infarction.6 The baseline CAPS 24-hour ECG recordings were made 25±17 days after infarction, and the baseline MPIP 24-hour ECG recordings were made 11±3 days after infarction. For the present study, we analyzed the 12-month “washout” 24-hour recordings in CAPS, which were recorded 1 week or more after the discontinuation of study medication or placebo following a year of treatment, ie, after full recovery from the infarction.6
Processing of 24-Hour Holter Recordings
We processed 24-hour Holter tape or cassette recordings using recently described methods. Briefly, the 24-hour recordings were digitized by a Marquette 8000 scanner and submitted to the standard Marquette algorithms for QRS labeling and editing (version 5.8 software). Then, the data files were transferred via high-speed link from the Marquette scanner to a Sun 4/75 workstation, where a second stage of editing was done, using algorithms developed at Columbia University, to find and correct any remaining errors in QRS labeling that could adversely affect measurement of RR variability.7
Long and short RR intervals in all classes, normal to normal, normal to atrial premature complex, and normal to ventricular premature complex were reviewed iteratively until all errors were corrected.7 For a tape to be eligible for this study, we required it to have >12 hours of analyzable data and have at least half of the nighttime (midnight to 5 am) and daytime (7:30 am to 9:30 pm) periods analyzable. At least 50% of each period had to be sinus rhythm.8
Time Series Analysis of Normal RR Intervals
After the second stage of editing and review of the results by a cardiologist, the RR interval power spectrum was computed over the entire recording interval (usually 24 hours) by a method first described by Albrecht and Cohen.9 Our adaptation of the method was described by Rottman et al.10 First, a regularly spaced time series was derived from the RR intervals by sampling the irregularly spaced series defined by the succession of normal RR intervals. For each Holter ECG recording, 218 points were sampled; for recordings precisely 24 hours in duration, the sampling interval was 329 ms. A “boxcar” low-pass filter with a window twice the sampling interval was then applied. Gaps in the time series resulting from noise or ectopic beats were filled in with linear splines, which can cause a small reduction in high-frequency (HF) power but does not affect other components of the power spectrum.9 A fast Fourier transform was computed, and the resulting power spectrum was corrected for the attenuating effects of both the filtering and the sampling.9 For a recording exactly 24 hours long, the effective frequency range for this method is from 1.1574×10−5 Hz to >0.40 Hz (periods of seconds to hours). Finally, frequency-domain measures of RR variability were computed by integration over their frequency intervals. We calculated the power within four frequency bands of the 24-hour RR interval power spectrum: (1) <0.0033 Hz, ultralow-frequency (ULF) power; (2) 0.0033 to <0.04 Hz, very-low-frequency (VLF) power, which shows a relative increase in patients with congestive heart failure11 and is the lowest frequency band that can be estimated by our 5-minute method10 ; (3) 0.04 to <0.15 Hz, low-frequency (LF) power, which reflects modulation of sympathetic or parasympathetic activity by baroreflex mechanisms12 ; and (4) 0.15 to <0.40 Hz, HF power, which reflects modulation of vagal activity, primarily by breathing.13 14 In addition, we calculated total power (power in the band <0.40 Hz) and the ratio of LF to HF (LF/HF) power, a measure that has been used as an indicator of sympathovagal balance.15 High values for the ratio suggest predominance of sympathetic nervous activity. We calculated time-domain measures of RR variability using previously reported methods.8
We used frequency matching to collect a sample of healthy middle-aged persons similar in sex and age to the MPIP sample of patients each of whom recently had had myocardial infarction. We excluded the 4.6% of the MPIP sample between 21 and 39 years old. The MPIP sample between 40 and 69 years old was distributed into 12 strata based on sex and six 5-year age intervals (Table 1⇓). Our aim was to recruit a sample of 250 healthy persons with the same percentage distribution as the MPIP sample across these sex and age strata. Healthy subjects were deliberately sampled so that their distribution on age and on sex matched those of the MPIP sample. The oversampling of older healthy subjects and of women ensured good power for comparisons between subjects and patients. To take advantage of frequency matching, statistical analyses used adjustment for age and for sex.
Each time-domain and frequency-domain variable was assessed for skewness by use of the standardized third moment around the mean (skewness coefficient).8 The skewness coefficient was calculated separately for the healthy subjects, the CAPS patients, and the MPIP patients. If the skewness coefficient was >1.00 in at least two of the three populations, the variable was considered to be significantly skewed and the variable was logarithmically transformed so that meaningful averages and regression equations could be calculated and thus statistical power would be greater.8 Patients in the CAPS sample (chronic coronary heart disease) and patients in the MPIP sample (subacute coronary heart disease) were compared with healthy persons on individual measures of RR variability using t ratios that take account of the stratification on age and sex.16
The patterns of association between the measures of RR variability and age and sex—eg, linear or quadratic regression functions and parallelism between men and women—were determined for the healthy subjects with the aim of using the data from all of them to establish age- and sex-specific expected values and SDs for adults aged 40 to 69 years. Our hypothesis was that simple linear regression equations would suffice to describe the association of measures of RR variability with age. This hypothesis was tested by fitting a quadratic equation to the data and testing whether the quadratic term improved the fit.17 Lack of significant improvement indicated that linear regression would suffice. To test our hypothesis that presence or absence of disease, age, and sex (in this order of importance) are all correlated with RR variability, we used stepwise multiple regression. We used the change in the square of the multiple correlation coefficient (R2) to indicate the magnitude of improvement in fit of the multiple regression model and a value of P<.05 to indicate statistical significance.
Several statistical comparisons were often made to test a single hypothesis. The Bonferroni method was used to adjust P values for multiple comparisons. An α-level of 0.05 or 0.01 was chosen for statistical significance and then was divided by the number of comparisons.18
To take advantage of the frequency matching, we adjusted for age or for age and sex in some of the comparisons. These comparisons were stratified by the six age categories or the 12 age and sex categories, respectively.16 The mean values for each stratum were weighted by the number of persons in that stratum to calculate a weighted mean and the corresponding mean square error. This procedure effectively decreases the SEM and provides greater power to detect differences among the comparison groups.
Mean Values for Time- and Frequency-Domain Measures of RR Variability for the Healthy Subjects and for the Two Coronary Heart Disease Samples
The age and sex distributions for the three samples were similar: healthy subjects, 57±8.2 years old and 74% male; CAPS sample, 58±7.4 years old and 84% male; and MPIP sample, 58±7.5 years old and 76% male. When cross-classified by age and sex, the healthy sample was also very similar to the two coronary heart disease samples (Table 1⇑). The durations of the ECG recordings were similar for the three samples, and >98% of all RR intervals were consecutive normal complexes (Table 2⇓).
Frequency histograms show that the four mutually exclusive, all-inclusive frequency-domain measures of RR variability were markedly skewed to the right for the healthy sample (Fig 1⇓); therefore, the values for frequency-domain measures of RR variability were transformed to their natural logarithms before summarization and statistical analyses. The distributions of the time-domain measures, square root of the mean of the squared differences between adjacent normal RR intervals over the entire 24-hour recording (r-MSSD), and percent of differences between adjacent RR intervals that are >50 ms computed over the entire 24-hour ECG recording (pNN50) also were skewed to the right and were, therefore, log-transformed.
Table 2⇑ shows results of testing the hypothesis that there is a difference between the healthy subjects and one or both of the coronary heart disease groups. To adjust for multiple comparisons, we divided the α-value by two to obtain the criterion for statistical significance. The average normal RR intervals were slightly but significantly shorter for healthy subjects than the average values for the MPIP patients; there was no significant difference between the healthy subjects and the CAPS patients (Table 2⇑). The mean values for time- and frequency-domain measures of RR variability for the healthy subjects were substantially and significantly greater than mean values for patients who had recovered from myocardial infarction (CAPS sample) and for those with recent myocardial infarction (MPIP sample) (Table 2⇑).
All measures of RR variability were significantly and substantially smaller in patients with chronic or subacute coronary heart disease than in healthy subjects (Table 2⇑). The difference from normal values was much larger 2 weeks after myocardial infarction (MPIP) than 1 year after infarction (CAPS). Gaussian probability density functions were computed by use of the means and SDs of the log-transformed values of the four frequency-domain measures of RR variability for the healthy sample and for the CAPS and MPIP samples (Fig 2⇓). Not only are the mean values for RR variability lower, but also the distributions of values are substantially broader in coronary heart disease patients than in healthy subjects (Fig 2⇓). Although the distributions overlap at their upper ends, they separate at their lower ends. Although total power is substantially lower in the two samples of patients with coronary heart disease than in healthy subjects, the fractional distribution of total power into its four component bands is similar for the three groups (Fig 3⇓).
Relation Between Power Spectral Measures of RR Variability and Age for the Healthy Subjects and for the Two Coronary Heart Disease Samples
For each of the three samples, a linear regression equation was computed by the least-squares method for each of the four frequency-domain measures of RR variability versus age (Fig 4⇓). Adding a quadratic term to the regression equation did not improve the fit significantly, indicating that linear regression is sufficient for these data. The 95% confidence bands are narrow for all the lines, indicating good precision for the estimated lines. For healthy subjects, there was no significant change in ULF power with age; however, there was a significant decrease in VLF, LF, and HF power with increasing age. VLF power decreased 12% per 10-year increase in age, LF power decreased 22% per 10 years, and HF power decreased 10% per 10 years. Patients with chronic coronary heart disease (the CAPS sample) showed little relation between the four power spectral measures of RR variability and age. Patients with a recent myocardial infarction (the MPIP sample) showed a strong inverse relation between VLF, LF, and HF power and age and a weak inverse relation between ULF power and age.
Relation Between Power Spectral Measures of RR Variability and Sex for the Healthy Subjects and for the Two Coronary Heart Disease Samples
Table 3⇓ shows results of testing the hypothesis that there is a difference between men and women for RR intervals and for power spectral variables. This hypothesis was tested by comparing men and women separately in the three groups: the healthy group and the two coronary heart disease groups. Because three male-female comparisons were made to test the hypothesis, we divided the α-value by three to obtain the criterion for statistical significance. For the healthy subjects, VLF power, LF power, and the LF/HF ratio were significantly higher for men than for women, but there were no differences for total power, ULF power, or HF power (Fig 5⇓). The pattern of differences between men and women was somewhat different for the two samples of patients with coronary heart disease; ie, ULF, VLF, LF, total power, and the LF/HF ratio all were higher in men than in women. In the samples of patients with coronary heart disease, only HF power showed no difference between men and women. Regression lines and confidence bands of the four power spectral bands on age were calculated separately for healthy middle-aged men and women. The values for men were slightly but significantly higher for VLF power and LF power (Fig 5⇓). However, there were no significant differences between men and women for the slope of the line relating any power spectral variable and age; ie, the regression lines of power spectral measures on age were parallel for men and women. A similar pattern was seen in the samples of patients with coronary heart disease; ie, the slopes of the lines associating power spectral measures with age were not significantly different between men and women.
Multivariate Relations Among Power Spectral Measures of RR Variability, Age, and Sex for the Healthy Subjects and for the Two Coronary Heart Disease Samples
To determine the relative importance of group membership (healthy versus subacute or chronic coronary heart disease), age, and sex on measures of RR variability, we used stepwise multiple regression analysis. The results of these analyses are summarized in Table 4⇓. The magnitude of improvement in fit of the multiple regression model is indicated by the change in the square of the multiple correlation coefficient (R2) and the statistical significance by the P value. For all the power spectral measures, the healthy persons versus the chronic coronary heart disease sample or the healthy persons versus the recent infarct sample was selected first in the step-up multiple regression analysis. Age and sex were much less important than group membership in fitting the model. The change in R2 was much smaller for age and sex than for group membership, and the order of entry of age and sex into the multiple regression model was not consistent.
Correlations Among Time- and Frequency-Domain Measures of RR Variability in Healthy Subjects
In the healthy middle-aged sample, many of the correlation coefficients among time- and frequency-domain measures of RR variability were large, and some exceeded a value of .90 (Table 5⇓). Correlations this strong between two variables indicate that the two variables can be used interchangeably when being related to a third variable. Power in each band in the frequency domain has a corresponding variable in the time domain that correlates so strongly with it that the two variables are essentially equivalent. There are three natural clusters among 10 of the 12 measures of RR variability in Table 5⇓: (1) SD of all normal RR intervals in the 24-hour ECG recording (SDNN), SD of the average normal RR intervals for all 5-minute segments of a 24-hour ECG recording (SDANN index), total power, and ULF power; (2) VLF power, LF power, and mean of the SDs of all normal RR intervals for all 5-minute segments of a 24-hour recording (SDNN index); and (3) HF power, r-MSSD, and pNN50. The LF/HF ratio correlates moderately with the variables that measure HF fluctuations in the RR interval (HF power, r-MSSD, and pNN50) but not with LF power, the other variable used to calculate LF/HF ratio. The night-day difference correlates moderately strongly with variables that measure lower-frequency fluctuation of RR intervals—SDNN, SDANN index, total power, and ULF power. We previously reported similar patterns of correlations among the measures of RR variability in the MPIP sample.8
Separation of Patients With Chronic (CAPS) and Subacute (MPIP) Coronary Heart Disease From Healthy Middle-Aged Subjects
If values that extend 2 SD below the mean are taken as normal, then, by definition, 2.5% of the values in the normal probability density function derived from the sample of healthy middle-aged persons are “abnormal,” ie, below this limit (Fig 2⇑). Using the cutpoints established in the healthy group to classify the coronary heart disease samples, we found that the following percentages of the CAPS sample (chronic coronary heart disease) were classified as abnormal: ULF power, 26%; VLF power, 23%; LF power, 28%; and HF power, 12%. In the MPIP sample, obtained 2 weeks after acute myocardial infarction, we found that the following percentages were classified as abnormal: ULF power, 60%; VLF power, 50%; LF power, 54%; and HF power, 21%. ULF power best separates the healthy group from either of the two samples of patients with coronary heart disease (Fig 4⇑). For ULF power, there is no overlap of the 95% confidence bands among the three samples. Furthermore, since there is little or no regression of ULF power on age in any of the three groups, no account has to be taken of age when interpreting a value for an individual person. VLF and LF power also provide excellent separation between the healthy group and the two samples of patients with coronary heart disease. HF power is poorest for separating healthy subjects from patients with coronary heart disease.
RR variability is a relatively new test for predicting mortality in patients with coronary heart disease. Several studies have shown that time- or frequency-domain measures of RR variability, determined at the time of hospital discharge after acute myocardial infarction, predict mortality and nonfatal arrhythmic events over the subsequent 3 years.8 19 20 21 22 23 The predictive value of RR variability is independent of other noninvasive measures commonly used to predict death after myocardial infarction, eg, low left ventricular ejection fraction, frequent or repetitive ventricular arrhythmias, or a positive signal-averaged ECG.19 20 Like other postinfarction risk predictors, measures of RR variability have modest positive predictive accuracy when used alone. However, combining measures of RR variability with other risk predictors yields a positive predictive accuracy of about 50%.20 23
Although measurement of RR variability at the time of hospital discharge after myocardial infarction has well-established predictive value, the utility of RR variability for predicting events in patients with chronic coronary heart disease is less certain. RR variability increases substantially in the period from 2 weeks after acute myocardial infarction to 3 months and then stabilizes.6 A preliminary study suggested that recovery values for measures of RR variability are substantially below those found in healthy middle-aged individuals.6 The present study was designed to validate these preliminary results and to measure more precisely the distributions for RR variability in clinically healthy middle-aged individuals. We confirmed an earlier study that showed that, after full recovery from acute myocardial infarction, values of RR variability are reduced by about one third.6 The magnitude of the difference between a group of patients who have fully recovered from acute myocardial infarction and age- and sex-matched healthy middle-aged individuals varies for different measures of RR variability. Measures that quantify slow fluctuations in RR intervals, eg, ULF power, are reduced to a proportionately greater extent than measures that quantify fast fluctuations, eg, HF power. Moreover, in healthy persons, ULF and total power do not decrease significantly with increasing age, which simplifies the diagnostic use of these variables.
This study establishes normal values for time- and frequency-domain measures of RR variability for middle-aged healthy men and women matched on sex and age to a sample of patients with recent acute myocardial infarction. We found that healthy middle-aged persons had much higher values for measures of RR variability than did patients with recent myocardial infarction or patients who had had myocardial infarction 1 year previously. Between the ages of 40 and 70 years, age and sex had only a modest influence on the values of measures of RR variability in healthy subjects or patients with coronary heart disease.
For all four power spectral measures of RR variability, there is substantial overlap in the upper end of distributions of healthy subjects and those for CAPS and MPIP patients. However, there is good separation in the lower end of the distributions (Fig 2⇑). How would RR variability be expected to perform if used to screen the middle-aged population of the United States for high-risk patients with coronary heart disease? About 7% of the US population 40 to 69 years old has chronic coronary heart disease. We showed previously that, for 24-hour ECG recordings, ULF power was the best measure of RR variability and a value of 5000 ms2 was the best cutpoint for classifying chronic coronary heart disease patients for low or high risk of death during 2 years of follow-up.24 Only 0.7% (2 of 274) of our sample of healthy middle-aged persons had a value of ULF power <5000 ms2. Similarly, we showed that, for 5-minute segments of ECG, LF power was the best predictor of all-cause mortality and a value of 120 ms2 was the best cutpoint for classifying coronary heart disease patients for low or high risk of death during 2 years of follow-up.24 25 Only 1.1% (3 of 274) of our sample of healthy middle-aged persons had a value of LF power <120 ms2.
This study was supported in part by National Institutes of Health grants HL-41552 from the National Heart, Lung, and Blood Institute, Bethesda, Md, and RR-00645 from the Research Resources Administration and by funds from The Bugher Foundation, New York, NY, The Dover Foundation, New York, NY, and Adelaide Segerman, New York, NY. The authors gratefully acknowledge the expert technical assistance of Lynne Bartell, Bernard Glembocki, Paul Gonzalez, and Reidar Bornholdt. The authors also would like to acknowledge Dr Robert E. Kleiger and Dr Matthew S. Bosner of the Jewish Hospital of St Louis, Washington University Medical Center, for their contributions to this study.
Reprint requests to J. Thomas Bigger, Jr, MD, Columbia University, P&S 9-445, 630 W 168th St, New York, NY 10032.
- Received July 8, 1994.
- Revision received October 24, 1994.
- Accepted October 31, 1994.
- Copyright © 1995 by American Heart Association
Bigger JT Jr, Fleiss JL, Kleiger R, Miller JP, Rolnitzky LM, The Multicenter Post-Infarction Research Group. The relationships among ventricular arrhythmias, left ventricular dysfunction and mortality in the 2 years after myocardial infarction. Circulation. 1984;69:250-258.
Albrecht P, Cohen RJ. Estimation of heart rate power spectrum bands from real-world data: dealing with ectopic beats and noisy data. Comput Cardiol. 1988;15:311-314.
Saul JP, Arai Y, Berger RD, Lilly LS, Colucci WS, Cohen RJ. Assessment of autonomic regulation in chronic congestive heart failure by heart rate spectral analysis. Am J Cardiol. 1988; 61:1292-1299.
Katona PG, Jih F. Respiratory sinus arrhythmia: measure of the parasympathetic cardiac control. J Appl Physiol. 1975;39:801-805.
Fouad FM, Tarazzi RC, Ferrario CM, Fighaly S, Alicandri C. Assessment of parasympathetic control of heart rate by a noninvasive method. Am J Physiol. 1984;246:H838-H842.
Pagani M, Lombardi F, Guzzetti S, Rimoldi O, Furlan R, Pizzinelli P, Sandrone G, Malfatto G, dell’Orto S, Piccaluga E, Turiel M, Baselli G, Cerutti S, Malliani A. Power spectral analysis of heart rate and arterial pressure variabilities as a marker of sympathovagal interaction in man and conscious dog. Circ Res. 1986;59:178-193.
Fleiss JL. The Design and Analysis of Clinical Experiments. New York, NY: John Wiley & Sons; 1986:150-154.
Kleinbaum DG, Kupper LL, Muller KE. Applied Regression Analysis and Other Multivariable Methods. Boston, Mass: PWS-Kent; 1988:102-110, 228-237.
Fleiss JL. The Design and Analysis of Clinical Experiments. New York, NY: John Wiley & Sons; 1986:103-107.
Cripps TR, Malik M, Farrell TG, Camm AJ. Prognostic value of reduced heart rate variability after myocardial infarction: clinical evaluation of a new analysis method. Br Heart J. 1991;65:14-19.
Odemuyiwa O, Malik M, Farrell T, Bashir Y, Poloniecki J, Camm J. Comparison of the predictive characteristics of heart rate variability index and left ventricular ejection fraction for all-cause mortality, arrhythmic events and sudden death after acute myocardial infarction. Am J Cardiol. 1991;68:434-439.
Farrell TG, Bashir Y, Cripps T, Malik M, Poloniecki J, Bennett ED, Ward DE, Camm AJ. Risk stratification for arrhythmic events based on heart rate variability, ambulatory electrocardiographic variables and the signal-averaged electrocardiogram. J Am Coll Cardiol. 1991;18:687-697.
Bigger JT Jr, 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.
Bigger JT Jr, Fleiss J, Rolnitzky LM, Steinman RC. The ability of several short-term measures of RR variability to predict mortality after myocardial infarction. Circulation. 1993;88:927-934.