- Email: [email protected]
Effect of exercise training on heart rate variability in healthy older adults
Effect of exercise training on heart rate variability in healthy older adults Phyllis K. Stein, PhD,a Ali A. Ehsani, MD,c Peter P. Domitrovich, PhD,b Robert E. Kleiger, MD,a and Jeffrey N. Rottman, MDd St Louis, Mo, and Nashville, Tenn
Objective To determine the effect of exercise training on cardiac autonomic modulation in normal older adults by using analysis of heart rate variability.
Subjects The exercise group consisted of 7 men and 9 women aged 66 ± 4 years. The comparison group consisted of 7 men and 9 women also aged 66 ± 4 years. Method Heart rate variability was determined from 24-hour Holter recordings before and after 12 months of supervised exercise, which consisted of 3 months of stretching and 9 months of 5 hours/week aerobic exercise at approximately 70% of maximal oxygen uptake. Heart rate variability was measured at baseline and 12 months later in the comparison group, who had not changed their usual activity level.
Results In the exercise group maximal oxygen consumption increased from 1.8 ± 0.5 L/min to 2.2 ± 0.7 L/min (P < .05). The standard deviation of normal interbeat intervals increased from 126 ± 21 ms to 142 ± 25 ms. Mean nighttime heart rate decreased from 67 ± 6 beats/min to 63 ± 5 beats/min. Increased fitness level had little effect on indexes of heart rate variability, which reflect parasympathetic or mixed sympathetic/parasympathetic modulation of heart rate. There was no change in heart rate or heart rate variability in the comparison group. Conclusions Exercise training increases total heart rate variability in normal older adults. The most marked alterations are in nocturnal heart rate. Heart rate variability is stable over a 1-year period in older adults who do not alter their activity level. (Am Heart J 1999;138:567-76.)
There is a slow, progressive decline in cardiovascular function with advancing age that is significantly accelerated by a sedentary lifestyle. This decline is characterized by diminished inotropic and chronotropic sensitivity to catecholamines, impairment of left ventricular function during exercise, increased arterial stiffness, and decreased heart rate variability.1-3 Changes in indexes of heart rate variability reflect changes in autonomic modulation of the heart both within a single recording and over time.4 Twenty-four-hour measurements of heart rate variability are reproducible over a period of weeks to months.5 Decreased heart rate variability is associated with advancing age and with an increased risk of cardiac From the aDivision of Cardiology, Washington University School of Medicine, bBarnes-Jewish Hospital, and the cDivision of Cardiology and Division of Geriatrics and Gerontology, Washington University School of Medicine, St. Louis; and the dDepartments of Medicine (Cardiology) and Pharmacology, Vanderbilt University School of Medicine, Nashville. Supported in part by the Claude D. Pepper Older American Independence Center NIH Grant AG 13629 and NIH Individual National Research Service Award F32 HL-08538. Submitted September 29, 1998; accepted January 15, 1999. Reprint requests: Phyllis K. Stein, PhD, Division of Cardiology, Barnes-Jewish Hospital, 216 S Kingshighway Blvd, St. Louis, MO 63110. E-mail: [email protected] Copyright © 1999 by Mosby, Inc. 0002-8703/99/$8.00 + 0 4/1/97177
events in clinically disease-free patients, even after adjusting for known risk factors.6 Similarly, decreased heart rate variability is associated with increased mortality rate in patients after myocardial infarction and in many other patient populations.7-9 Conversely, in a variety of circumstances increased heart rate variability is associated with lower mortality rate, for example, in β-blockade in patients after myocardial infarction,10,11 angiotensin-converting enzyme inhibition in congestive heart failure,12,13 and smoking cessation.14 Thus increases in heart rate variability might reflect improved cardiac autonomic modulation and prognosis. Because trained athletes have higher heart rate variability than sedentary individuals,15 it has been suggested that exercise training can increase heart rate variability in normal populations. There are no data on the effect of endurance exercise training on 24-hour indexes of heart rate variability in previously sedentary older men or women. The hypothesis of this study was that improvement in aerobic exercise capacity resulting from exercise training in older men and women would be associated with increased heart rate variability. A secondary and related hypothesis was that heart rate variability would be unchanged after 1 year in healthy older adults who did not change their activity patterns.
American Heart Journal September 1999
568 Stein et al
Figure 1
Baseline and posttest SDNN for comparison and training groups. Error bars indicate means ± 1 SD.
Methods Experimental patients Sixteen adults (7 men and 9 women) aged 66.2 ± 4.2 years completed the exercise part of this study. All patients were participating in a larger project evaluating the cardiovascular effects of exercise in older adults. Subjects were nonsmokers, without evidence of cardiovascular disease by history, cardiac examination, 12-lead electrocardiogram, and treadmill exercise testing. None were taking cardioactive medications and none had engaged in regular exercise for at least 2 years before beginning the program. To compare the change in indexes of heart rate variability over a similar period of 1 year among those who did not participate in a program of supervised activity, a separate group of age- and sex–matched comparison subjects, 7 men and 9 women aged 65.7 ± 4.1 years, was recruited from healthy older adults participating in a cross-sectional study of heart rate variability at the same institution. Subjects in this study were nonsmokers, free of cardiovascular disease by history and physical examination, including blood pressure measurement and 12-lead electrocardiogram, and had normal Holter recordings. None were taking cardioactive medications. These subjects were asked to return approximately 1 year later for repeat Holter monitoring, at which time it was determined that they had not changed their usual level of exercise. Of the 16 comparison subjects, 9 reported engaging in no regular exercise, 2 reported engaging in regular light exercise, 3 reported regular walking and 2 reported regular endurance exercise.
Measurement of maximal oxygen consumption Each subject in the exercise study underwent an initial treadmill test16 to evaluate heart rate, blood pressure, and
electrocardiographic response to exercise and to select the appropriate protocol for determination of maximal oxygen consumption. One to 2 weeks later, another treadmill exercise test was performed to determine maximal oxygen consumption as previously described.17 In brief, an initial treadmill speed was chosen that elicited 60% to 70% of maximal heart rate during the 5-minute warmup. Thereafter, treadmill grade was increased by 2% every 2 minutes. Oxygen uptake was measured continuously by open-circuit spirometry and averaged every 30 seconds with the use of an automated on-line system previously validated against the Douglas bag technique.16,17 Inspiratory volume was measured with a Parkinson-Cowan CD-4 dry gas meter. Fractional concentrations of oxygen and carbon dioxide were sampled from a mixing chamber and quantified with electronic oxygen (Applied Electrochemistry S3-A) and carbon dioxide (Beckman LB-2) analyzers. Maximal oxygen consumption was defined as (1) attainment of a plateau of oxygen consumption with increasing intensity, (2) heart rate within 10 beats/min of age-predicted maximal heart rate, and (3) respiratory exchange ratio exceeding 1.10.
Training program The training program consisted of 2 phases: a 3-month flexibility program and a 9-month endurance exercise program. During the endurance exercise program, control subjects underwent supervised exercise for 45 to 60 minutes/day for 5 days/week. Endurance exercise consisted of walking (including uphill treadmill walking), jogging, cycling, and rowing. Initial intensity of exercise was set at 60% to 70% of maximal oxygen consumption based on maximal oxygen consumption determined during a maximal exercise treadmill test and was gradually increased to 70% to 85% of maxi-
American Heart Journal Volume 138, Number 3, Part 1
Stein et al 569
Table I. Results of exercise training (n = 16) Baseline Maximal oxygen consumption (mL/kg/min) Maximal oxygen consumption (L/min) Maximal heart rate (beats/min) Resting heart rate (beats/min)
23.6 ± 3.8* 1.8 ± 0.5* 163 ± 8 71 ± 8†
After training 30.8 ± 5.2* 2.2 ± 0.7* 163 ± 9 66 ± 9†
*Statictically significant, P < .001. †P = .082.
mal oxygen consumption. The exercise prescription was adjusted on a weekly basis.
Measurement of heart rate variability Twenty-four-hour ambulatory electrocardiographic recordings were obtained for each patient with Marquette Series 8500 Holter recorders at baseline (during the stretching phase) and after completion of the exercise training program.
Data collection and analysis Holter recordings were analyzed with a Marquette SXP Laser Holter scanner (software version 5.8). All subjects were determined to be in sinus rhythm and without episodes of sustained atrial arrhythmias such as atrial fibrillation or multifocal atrial tachycardia. Beat-stream files, representing the time and classification of each QRS complex, were transferred to a Sun Sparcstation computer, where they underwent standard secondary editing and heart rate variability analysis procedures using previously validated techniques.18,19 Time domain analysis of successive normal-to-normal RR intervals. Time domain indexes of heart rate variability are
derived from statistical calculations performed on the set of normal-normal (N-N) interbeat intervals. Average heart rate in beats per minute and the following commonly reported time domain measures of heart rate variability were calculated for daytime (08:00 to 20:00), nighttime (00:00 to 0:600), and 24hour periods: AVGNN (average heart period in ms), SDNN (the standard deviation of N-Ns in milliseconds), SDANN (the standard deviation of 5-minute mean values of N-Ns for each 5-minute interval in milliseconds), SDNNIDX (the average of standard deviations of N-Ns for each 5-minute interval in milliseconds), rMSSD (the root mean square successive difference of N-Ns in milliseconds), and pNN50 (the percent of successive N-N differences >50 ms for each 5-minute interval). RMSSD and pNN50 primarily reflect parasympathetically mediated changes in heart rate.20 The other time domain variables reflect a mixture of parasympathetic, sympathetic, and other physiologic influences.20 Power spectral analysis of NN intervals. Power spectral analysis partitions the variance in the heart rate or heart period signal into its underlying components. The following standard frequency domain components of heart rate vari-
ability, measured in milliseconds squared, were computed in this analysis: (1) high-frequency power, the component of the total variance in heart period between 0.15 and 0.40 Hz, which primarily reflects vagal modulation of heart rate, 21 (2) low-frequency power, the component between 0.04 and 0.15 Hz, which reflects both sympathetic and parasympathetic modulation of heart rate,21 (3) very-low-frequency power, the component between 0.0033 to 0.04 Hz, which may represent the influence of the thermoregulatory 22 or renin-angiotensin23 systems, and (4) ultra-low-frequency power, the component between 1.15X10–5 and 0.00335 Hz, which reflects circadian variations. Total power (1.15X10 –5 to 0.40 Hz) for the total 24-hour cycle was also determined. The ratio of low- to high-frequency power, which has been suggested as a marker of sympathetic/parasympathetic balance,24 was also calculated. The methods used for spectral analysis have been previously described.19,25 In brief, the sequence of normal-to-normal intervals was resampled and filtered to provide a uniformly spaced time series. Missing or noisy segments were replaced by linear interpolation from the surrounding signal. The average of the sampled series was subtracted from the time series and a fast Fourier transforms performed to determine the frequency components underlying the cyclic activity in the sampled time series. Measurement of ultra-low and very-low-frequency power was based on en bloc analysis of the entire 24-hour recording.25 Other power indexes reported here reflect the average of 5-minute segments in which ≥80% of the beats are normal. Heart rate tachograms. To permit visual inspection of heart rate patterns at night when periodic heart rate rhythms associated with sleep-disordered breathing might affect indexes of heart rate variability, heart rate tachograms were generated from the interbeat interval files. Each interbeat interval was converted to its associated heart rate and the resultant heart rate time series plotted.
Statistical analysis Heart rate variability indexes were tested for normality and log-transformed when necessary to permit parametric statistical analysis. (All frequency domain indexes, rMSSD, and pNN50 required log transformation.) Paired t tests were performed to compare baseline and posttraining maximal oxygen consumption and maximal and resting heart rates for the training group. Student t tests compared baseline
American Heart Journal September 1999
570 Stein et al
Figure 2
Baseline and posttest SDNNIDX for comparison and training groups. Error bars indicate means ± 1 SD.
Table II. 24-hour time domain heart rate variability determined approximately 1 year apart in older adults continuing usual activities and older adults undergoing intensive exercise training Comparison group
AVGNN (ms) SDNN (ms) SDANN (ms) SDNNIDX (ms) rMSSD (ms) pNN50 (%)
Training group
Baseline
After test
P value
Baseline
After test
821 ± 106 135 ± 32 126 ± 32 48 ± 12 26 ± 10 6±5
832 ± 97 134 ± 32 123 ± 32 50 ± 12 27 ± 9 6±5
.35 .75 .55 .30 .61 .85
787 ± 45 126 ± 21 116 ± 21 46 ± 11 26 ± 16 5±7
823 ± 68 142 ± 25 129 ± 23 50 ± 15 27 ± 14 6±6
P value .04 .02 .04 .08 .43 .65
P values for before-to-after comparisons within groups. All values mean ± SD.
values of heart rate and heart rate variability between the exercise and the comparison groups to test for comparability at the start of the study. t Tests on differences between initial and final measurements of heart rates, heart rate variability, and respiratory rates were used to determine if changes over a 1-year period were different between the training and comparison groups. Each patient therefore acted as his or her own control. Correlational analysis was used to determined if changes in maximal oxygen consumption were associated with changes in indexes of heart rate variability. Statistical significance was set at P < .05. SPSSPC software (SPSS, Inc) was used for all statistical analyses.
Results Baseline measurement Initial indexes of heart rate and heart rate variability were compared between the training and comparison groups (no change in activity pattern). No statistically significant differences were observed.
Effect of exercise training on maximum oxygen consumption, maximum heart rate, and ectopy Exercise training resulted in a marked increase in maximal oxygen consumption, expressed in either liters per minute (22.2%) or mL/kg/min (30.5%) (P < .001). This increase occurred in all patients. Maximal heart rate was not altered (Table I). A trend to a lower supine resting heart rate as measured before each treadmill test did not attain statistical significance (P = .08, Table I). There was no difference in the number of extrasystoles before and after exercise training (P = .994).
Effect of exercise training on time domain indexes of heart rate variability As shown in Table II, exercise training resulted in a significant (12.6%) mean increase in SDNN (Figure 1) and a significant (11.2%) mean increase in SDANN (P
American Heart Journal Volume 138, Number 3, Part 1
Stein et al 571
Figure 3
Baseline and posttest AVGNN for comparison and training groups. Error bars indicate means ± 1 SD.
Table III. Daytime and nighttime indexes of heart rate and heart rate variability determined approximately 1 year apart in older adults continuing usual activities and older adults undergoing intensive exercise training Comparison group
Daytime (08:00-22:00) AVG NN (ms) SDNN (ms) SDANN (ms) SDNNIDX (ms) rMSSD (ms) pNN50 (%) Nighttime (00:00-06:00) AVG NN (ms) SDNN (ms) SDANN (ms) SDNNIDX (ms) rMSSD (ms) pNN50 (%)
Training group
Baseline
After test
P value
Baseline
After test
P value
752 ± 105 80 ± 24 77 ± 26 44 ± 11 22 ± 9 4±5
763 ± 100 95 ± 28 81 ± 27 44 ± 10 22 ± 9 4±6
.49 .24 .40 .62 .93 .89
726 ± 41 99 ± 20 87 ± 21 43 ± 8 21 ± 11 4±6
752 ± 75 118 ± 34 103 ± 33 48 ± 14 23 ± 9 4±4
.12 .07 .14 .11 .43 .91
968 ± 113 91 ± 30 65 ± 26 55 ± 16 32 ± 11 9±7
973 ± 99 92 ± 38 61 ± 28 61 ± 22 35 ± 13 11 ± 8
.47 .85 .49 .05 .05 .08
906 ± 81 87 ± 27 62 ± 22 52 ± 22 34 ± 32 9 ± 15
961 ± 75 88 ± 25 60 ± 19 55 ± 20 34 ± 23 10 ± 11
.01 .90 .68 .31 .85 .56
P values for before-to-after comparisons within groups. Values given as mean ± SD.
≤ .04), indexes that primarily reflect circadian rhythms. Although mean values for heart rate decreased for the exercise group (P = .04), when before-to-after changes in the exercise and comparison groups were compared, changes were significantly different only for SDNN and SDANN. There was no significant exercise-related change in any other 24-hour time domain index of heart rate variability. Figure 2 illustrates the similar changes over a 1-year period in 24-hour SDNNIDX for the training and comparison groups. When time domain indexes of heart rate variability were separately analyzed for daytime (08:00 to 22:00)
and nighttime (00:00 to 06:00) periods, exercise training tended to decrease mean daytime heart rates (P = .12) and increase mean SDNN (P = .07), but the only significant change in before-to-after values was a 6% increase in mean nighttime heart periods (ie, decrease in nighttime heart rates) among exercisers (P = .01, Table III and Figure 3).
Effect of exercise training on frequency domain indexes of heart rate variability When these data were analyzed with frequency domain indexes of heart rate variability, consistent with
American Heart Journal September 1999
572 Stein et al
Figure 4
Baseline and posttest natural logarithm high-frequency power for comparison and training groups. Error bars indicate means ± 1 SD.
Table IV. 24-hour frequency domain heart rate variability determined approximately 1 year apart in older adults continuing usual activities and older adults undergoing intensive exercise training Comparison group Baseline ln TP ln ULF ln VLF ln LF ln HF
9.8 ± 0.4 9.7 ± 0.4 7.1 ± 0.5 6.3 ± 0.6 5.0 ± 0.7
After test 9.8 ± 0.5 9.6 ± 0.5 7.2 ± 0.5 6.4 ± 0.7 5.1 ± 0.7
Training group P value .708 .600 .229 .115 .327
Baseline
After test
P value
9.6 ± 0.4 9.5 ± 0.4 7.0 ± 0.4 6.1 ± 0.8 4.9 ± 1.0
9.9 ± 0.4 9.7 ± 0.4 7.2 ± 0.5 6.3 ± 1.0 5.1 ± 1.1
.027 .041 .133 .151 .340
P values for before-to-after comparisons within groups. TP, total power; ULF, ultra-low frequency; VLF, very-low frequency; HF, high frequency; LF, low frequency.
findings in the time domain, significant increases were seen in natural logarithm total power and natural logarithm ultra-low-frequency power in the exercise group compared with the comparison group. Figure 4 illustrates the similar before-to-after changes of natural logarithm high-frequency power in both the exercise and comparison groups. When results were analyzed separately for daytime and nighttime frequency domain measures (data not shown), there were no differences between groups in the before-to-after change in heart rate variability, despite the significant before-to-after increases in natural logarithm very-low-frequency power and natural logarithm low-frequency power at night for the comparison group. When the change in the ratio of low- to high-frequency power, which has been suggested as an index of sympathetic activation,24 was com-
pared, there were no differences between groups and no significant change for either group. Similarly, the change in the coefficient of variance (SDNN/AVGNN), another index of autonomic balance, before-to-after was not significantly different between groups.
Relation between changes in heart rate variability and changes in maximal oxygen consumption When the correlations between changes in heart rate variability and changes in maximal oxygen consumption were determined, although exercise training increased 24-hour SDNN, there was no relation between the change in maximal oxygen consumption and the change in SDNN (Figure 5). A modest relation was observed between the change in maximal oxygen consumption in milliliters per kilogram
American Heart Journal Volume 138, Number 3, Part 1
Stein et al 573
Figure 5
Relation between change in maximal oxygen consumption and change in 24-hour SDNN.
per minute and the change in SDNNIDX at night (r = 0.45, P = .04), low-frequency power at night (r = 0.52, P = .02), high-frequency power at night (r = 0.45, P = .04), and 24-hour very-low-frequency power (r = 0.49, P = .03). Of these, the changes in SDNNIDX at night, low-frequency power at night, and very-lowfrequency power were highly correlated (correlation coefficient between 0.77 and 0.88, P ≤ .001). There were no significant relations between changes in heart rate variability and changes in the numbers of extrasystoles. There was also no sex effect. Changes in heart rate variability after 1 year in the comparison group. In general, there were no
significant differences between initial and final values of heart rate or heart rate variability in the comparison group, indicating the reproducibility of heart rate variability over a 1-year period in this older population (Tables II through IV, Figures 1 to 4). However, although there was no difference between groups in before-to-after changes, significant increases were found for nighttime values of the highly correlated indexes of natural logarithm very-low-frequency power, natural logarithm low-frequency power, and SDNNIDX. In particular, 5 patients in the training group and 5 in the comparison group had increases of ≥10 ms in SDNNIDX at night, and 4 in the training
group and 1 in the comparison group had decreases in SDNNIDX at night of ≥10 ms. These increases were associated with an increase in the percentage of time in which periodic heart rate rhythms, as seen on the heart rate tachogram, were present (Figure 6). In some cases, these patterns were similar to those associated with periodic limb movements or sleep disordered breathing such as sleep apnea. In others they were less periodic and seemed to reflect more frequent arousals. The decrease in SDNNIDX at night in one of the exercise subjects was associated with a marked decrease in periodic heart rate rhythms after exercise training.
Discussion The results of this study suggest that in healthy older adults the expected adaptive increase in maximal oxygen consumption in response to exercise training is associated with an increase in longer term indexes of heart rate variability. The clearest change is a decrease in nocturnal mean heart rate. Mean SDNN and SDANN, which reflect the longest term heart rate variability trends, are consequently increased. Low values for SDNN and SDANN have been associated with an increased risk for death after myocardial infarction, and low values for these indexes reflect
American Heart Journal September 1999
574 Stein et al
Figure 6
Ten-minute tachograms showing (top) quiet sleep, (middle) sleep apnea–like pattern in a different subject, and (bottom) large changes in heart rate, possibly from frequent arousals in same subject as (top).
the failure of heart rate to decrease substantially at night.26 The physiologic mechanism underlying these changes is not defined by this study. There were no significant changes in index means reflecting primarily vagal tone or in those reflecting short-term cardiac autonomic modulation in the exercise group. However, although group means in these variables were unaffected, changes in individual maximal oxygen consumption were correlated with individual changes in the intermediate term indexes nighttime SDNNIDX, nighttime low-frequency power, and 24hour very-low-frequency power, and the short-term index nighttime high-frequency power. Aerobic fitness has also been associated with a better quality of sleep, as reflected in shorter sleep onset latencies, less wake time after onset, fewer discrete sleep episodes, fewer sleep stage shifts during the initial portions of the night, a higher sleep efficiency, and more total slow waves in older men.27 It is possible that the decrease in heart rates at night after aerobic exercise training reflects improved sleep attributable to increased fitness.
It has been generally assumed that exercise training increases heart rate variability. However, the effect of exercise training on heart rate variability in this older adult population was modest (12.6% for mean SDNN) when compared to the substantial (more than 22%) gain in mean maximal oxygen consumption. Moreover, maximal oxygen consumption increased even among those exercise subjects in whom SDNN decreased (Figure 1). In fact, there is little evidence from prospective studies that an exercise training program will dramatically increase heart rate variability in older adults. The presumed relation between exercise training and increased heart rate variability is largely based on cross-sectional investigations comparing athletes and sedentary controls. Interestingly, the comparison group of this study suggests a difference between athletes and sedentary subjects: in the 2 athletic patients who engaged in regular endurance training, SDNN (Figure 1) and AVGNN at night (Figure 3) were noticeably higher, whereas baseline values for the remaining members of the comparison group were similar to those in the training group. In twin studies, maximal oxygen consumption improve-
American Heart Journal Volume 138, Number 3, Part 1
ment as a result of exercise training is strongly influenced by genetic endowment.28 It is possible that the high heart rate variability levels seen in cross-sectional studies in endurance athletes29 may in part reflect genetics and selection. Indexes of heart rate variability were remarkably similar over the 1-year period in the group, which did not change their usual activities. This confirms the reproducibility of heart rate variability indexes over intermediate periods of time5 and suggests that in healthy older adults the age-mediated decline in heart rate variability is relatively slow. A limitation of this study is that the comparison group did not consist entirely of sedentary individuals and that maximal oxygen consumption was not measured in the comparison group. However, the statistical design allowed each participant in the study to serve as his or her own control. Our assumption of little or no change in maximal oxygen consumption over a 1-year follow-up period among the comparison group that did not change their usual activities appears reasonable. The lack of change in nighttime heart rates and 24-hour heart rate variability in the comparison group strongly suggests that the changes seen in the exercise group can be attributed to aerobic exercise training. Our study suggests a relation between beneficial effects of exercise training on aerobic fitness and beneficial effects on 24-hour indexes of heart rate variability in normal older adults. Other measures of autonomic function, such as baroreceptor reflex sensitivity or response to orthostatic challenge, were not determined in this study. Because some of these measures convey prognostic information in addition to that provided by heart rate variability, they generate reasonable hypotheses for future investigation. This study, together with the compelling associations between heart rate variability and outcome, emphasizes the intriguing connections among modulation of heart rate, sleep, exercise training, and prognosis.
References 1. Fleg JL. Alterations in cardiovascular structure and function with advancing age. Am J Cardiol 1986;57:33C-44C. 2. Ehsani AA. Aging and the heart. West J Med 1993;159:501-2. 3. Bigger JT, Fleiss LF, Steiman RC, Rolnitzsky LM, Schneider WJ, Stein PK. RR variability in healthy, middle-age persons compared with patiems with chronic coronary heart disease or recent acute myocardial infarction. Circulation 1995;91:1936-43. 4. Stein PK, Bosner MS, Kleiger RE, Conger BM. Heart rate variability: a measure of cardiac autonomic tone. Am Heart J 1993; 127:1376-81. 5. Kleiger RE, Bigger JT, Bosner MS, Chung MK, Cook JR, Rolnitzsky LM, et al. Stability over time of variables measuring heart rate variability in normal subjects. Am J Cardiol 1991;68:626-30.
Stein et al 575
6. Tsuji H, Lawrence MG, Venditti FJ Jr, Manders ES, Evans JC, Feldman CL, et al. Impact of reduced heart rate variability on risk for cardiac events. The Framingham Heart Study. Circulation 1996; 94:2850-5. 7. Kleiger RE, Miller JP, Bigger JT Jr, Moss AJ, and the MPIP Research Group. Decreased heart rate variability and its association with increased mortality after acute myocardial infarction. Am J Cardiol 1987;59:256-62. 8. Odemuyiwa O, Farrell T, Malik M, Bashir Y, Poloniecki J, Ward DE, et al. The effect of age on the electrophysiological and autonomic correlates of sudden death after acute myocardial infarction. PACE 1991;14:2049-55. 9. Molgaard H, Sorensen KE, Bjerregaard P. Attenuated 24-h heart rate variability in apparently healthy subjects subsequently suffering sudden cardiac death. Clin Auto Rsch 1991;1:233-7. 10. Sandrone G, Mortara A, Torzillo D, La Rovere MT, Malliani A, Lombardi, F. Effects of beta blockers (atenolol or metoprolol) on heart rate variability after acute myocardial infarction. Am J Cardiol 1994;74:340-5. 11. Yusuf S, Peto R, Lewis J, Collins R, Sleight P. Beta blockade during and after myocardial infarction: an overview of the randomized trails. Prog Cardiovasc Dis 1985;XXVII:335-71. 12. The SOLVD Investigators. Effect ofenalapfil on survival in patients with reduced left ventricular ejection function and congestive heart failure. N Engl J Med 1991;325:293-302. 13. Binkley PF, Haas GJ, Starling RC, Nunziata E, Hatton PA, Leier CV, et al. Sustained augmentation of parasympathetic tone with angiotensinconverting enzyme inhibition in patients with congestive heart failure. J Am Col Cardiol 1993;21:655-61. 14. Stein PK, Rottman JN, Kleiger RE. Effect of 21 mg transdermal nicotine patches and smoking cessation on heart rate variability. Am J Cardiol 1996;77:701-5. 15. Goldsmith RL, Bigger JT Jr, Steinman RC, Fleiss JL. Comparison of 24-hour parasympathetic activity in endurance-trained and untrained younger men. J Am Coll Cardiol 1992;20:552-8. 16. Bruce RA. Exercise testing of patients with coronary heart disease. Ann Clin Res 1971;3:323-9. 17. Khort WM, Malley MT, Coggan AR, Spina R J, Ogawa T, Ehsani AA, et al. Effects of gender, age, and fitness level on the response of VO2 max to training in 60- to 71-yr-olds. J Appl Physiol 1991; 71:2004-11. 18. Malik M, Farrell T, Cripps T, Camm AJ. Heart rate variability in relation to prognosis after myocardial infarctions: selection of optimal processing techniques. Eur Heart J 1989;10:1060-74. 19. Berger, RD, Akselrod S, Gordon D, Cohen R. An efficient algorithm for spectral analysis of heart rate variability. IEEE Trans Biomed Eng 1986;9:900-4. 20. Kleiger RE, Stein PK, Bosner MS, Rottman JN. Time domain measurements of heart rate variability. Cardiol Clin 1992;10:487-98. 21. Akselrod S, Gordon D, Madwed JB, Snidman NC, Shannon DC, Cohen RI. Hemodynamic regulation investigated by spectral analysis. Am J Physiol (Heart Circ Physiol 18) 1985;249:H867-H875. 22. Fleisher LA, Frank SM, Sessler DI, Cheng C, Matsukawa T, Vannier CA. Thermoregulation and heart rate variability. Clin Sci 1996; 90:97-103. 23. Akselrod S, Gordon D, Ubel FA, Shannon DC, Barger AC, Cohen RI. Power spectrum analysis of heart rate fluctuation: a quantitative probe of beat-to-beat cardiovascular control. Science 1981;213:220-2. 24. Malliani A, Pagani M, Lombardi F, Cerutti S. Cardiovascular neural
576 Stein et al
regulation explored in the frequency domain. Circulation 1991; 84:482-92. 25. Rottman JN, Steinman RC, Albrecht P, Bigger JT, Rolnitzky LM, Fleiss JL. Efficient estimation of the heart period power spectrum suitable for physiologic or pharmacologic studies. Am J Cardiol 1990;66:1522-4. 26. Kleiger RE, Miller JP, Krone RJ, Bigger JT Jr, and the MPIP Research Group. The independence of cycle length variability and exercise testing on predicting mortality of patients surviving acute myocardial infarction. Am J Cardiol 1990;65:408-11.
American Heart Journal September 1999
27. King AC, Oman RF, Brassington GS, Bliwise DL, Haskell WL. Moderate-intensity exercise and self-rate quality of sleep in older adults. A randomized controlled trial. JAMA 1997;277:32-7. 28. Prud’homme D, Bouchard C, Leblanc C, Landry F, Fontaine E. Sensitivity of maximal aerobic power to training is genotype-dependent. Med Sci Sports Exerc 1984;16:489-93. 29. Sacknoff DM, Gleim GW, Stachenfeld N, Coplan NL. Effect of athletic training on heart rate variability. Am Heart J 1994; 127:1275-78.
Objective To determine the effect of exercise training on cardiac autonomic modulation in normal older adults by using analysis of heart rate variability.
Subjects The exercise group consisted of 7 men and 9 women aged 66 ± 4 years. The comparison group consisted of 7 men and 9 women also aged 66 ± 4 years. Method Heart rate variability was determined from 24-hour Holter recordings before and after 12 months of supervised exercise, which consisted of 3 months of stretching and 9 months of 5 hours/week aerobic exercise at approximately 70% of maximal oxygen uptake. Heart rate variability was measured at baseline and 12 months later in the comparison group, who had not changed their usual activity level.
Results In the exercise group maximal oxygen consumption increased from 1.8 ± 0.5 L/min to 2.2 ± 0.7 L/min (P < .05). The standard deviation of normal interbeat intervals increased from 126 ± 21 ms to 142 ± 25 ms. Mean nighttime heart rate decreased from 67 ± 6 beats/min to 63 ± 5 beats/min. Increased fitness level had little effect on indexes of heart rate variability, which reflect parasympathetic or mixed sympathetic/parasympathetic modulation of heart rate. There was no change in heart rate or heart rate variability in the comparison group. Conclusions Exercise training increases total heart rate variability in normal older adults. The most marked alterations are in nocturnal heart rate. Heart rate variability is stable over a 1-year period in older adults who do not alter their activity level. (Am Heart J 1999;138:567-76.)
There is a slow, progressive decline in cardiovascular function with advancing age that is significantly accelerated by a sedentary lifestyle. This decline is characterized by diminished inotropic and chronotropic sensitivity to catecholamines, impairment of left ventricular function during exercise, increased arterial stiffness, and decreased heart rate variability.1-3 Changes in indexes of heart rate variability reflect changes in autonomic modulation of the heart both within a single recording and over time.4 Twenty-four-hour measurements of heart rate variability are reproducible over a period of weeks to months.5 Decreased heart rate variability is associated with advancing age and with an increased risk of cardiac From the aDivision of Cardiology, Washington University School of Medicine, bBarnes-Jewish Hospital, and the cDivision of Cardiology and Division of Geriatrics and Gerontology, Washington University School of Medicine, St. Louis; and the dDepartments of Medicine (Cardiology) and Pharmacology, Vanderbilt University School of Medicine, Nashville. Supported in part by the Claude D. Pepper Older American Independence Center NIH Grant AG 13629 and NIH Individual National Research Service Award F32 HL-08538. Submitted September 29, 1998; accepted January 15, 1999. Reprint requests: Phyllis K. Stein, PhD, Division of Cardiology, Barnes-Jewish Hospital, 216 S Kingshighway Blvd, St. Louis, MO 63110. E-mail: [email protected] Copyright © 1999 by Mosby, Inc. 0002-8703/99/$8.00 + 0 4/1/97177
events in clinically disease-free patients, even after adjusting for known risk factors.6 Similarly, decreased heart rate variability is associated with increased mortality rate in patients after myocardial infarction and in many other patient populations.7-9 Conversely, in a variety of circumstances increased heart rate variability is associated with lower mortality rate, for example, in β-blockade in patients after myocardial infarction,10,11 angiotensin-converting enzyme inhibition in congestive heart failure,12,13 and smoking cessation.14 Thus increases in heart rate variability might reflect improved cardiac autonomic modulation and prognosis. Because trained athletes have higher heart rate variability than sedentary individuals,15 it has been suggested that exercise training can increase heart rate variability in normal populations. There are no data on the effect of endurance exercise training on 24-hour indexes of heart rate variability in previously sedentary older men or women. The hypothesis of this study was that improvement in aerobic exercise capacity resulting from exercise training in older men and women would be associated with increased heart rate variability. A secondary and related hypothesis was that heart rate variability would be unchanged after 1 year in healthy older adults who did not change their activity patterns.
American Heart Journal September 1999
568 Stein et al
Figure 1
Baseline and posttest SDNN for comparison and training groups. Error bars indicate means ± 1 SD.
Methods Experimental patients Sixteen adults (7 men and 9 women) aged 66.2 ± 4.2 years completed the exercise part of this study. All patients were participating in a larger project evaluating the cardiovascular effects of exercise in older adults. Subjects were nonsmokers, without evidence of cardiovascular disease by history, cardiac examination, 12-lead electrocardiogram, and treadmill exercise testing. None were taking cardioactive medications and none had engaged in regular exercise for at least 2 years before beginning the program. To compare the change in indexes of heart rate variability over a similar period of 1 year among those who did not participate in a program of supervised activity, a separate group of age- and sex–matched comparison subjects, 7 men and 9 women aged 65.7 ± 4.1 years, was recruited from healthy older adults participating in a cross-sectional study of heart rate variability at the same institution. Subjects in this study were nonsmokers, free of cardiovascular disease by history and physical examination, including blood pressure measurement and 12-lead electrocardiogram, and had normal Holter recordings. None were taking cardioactive medications. These subjects were asked to return approximately 1 year later for repeat Holter monitoring, at which time it was determined that they had not changed their usual level of exercise. Of the 16 comparison subjects, 9 reported engaging in no regular exercise, 2 reported engaging in regular light exercise, 3 reported regular walking and 2 reported regular endurance exercise.
Measurement of maximal oxygen consumption Each subject in the exercise study underwent an initial treadmill test16 to evaluate heart rate, blood pressure, and
electrocardiographic response to exercise and to select the appropriate protocol for determination of maximal oxygen consumption. One to 2 weeks later, another treadmill exercise test was performed to determine maximal oxygen consumption as previously described.17 In brief, an initial treadmill speed was chosen that elicited 60% to 70% of maximal heart rate during the 5-minute warmup. Thereafter, treadmill grade was increased by 2% every 2 minutes. Oxygen uptake was measured continuously by open-circuit spirometry and averaged every 30 seconds with the use of an automated on-line system previously validated against the Douglas bag technique.16,17 Inspiratory volume was measured with a Parkinson-Cowan CD-4 dry gas meter. Fractional concentrations of oxygen and carbon dioxide were sampled from a mixing chamber and quantified with electronic oxygen (Applied Electrochemistry S3-A) and carbon dioxide (Beckman LB-2) analyzers. Maximal oxygen consumption was defined as (1) attainment of a plateau of oxygen consumption with increasing intensity, (2) heart rate within 10 beats/min of age-predicted maximal heart rate, and (3) respiratory exchange ratio exceeding 1.10.
Training program The training program consisted of 2 phases: a 3-month flexibility program and a 9-month endurance exercise program. During the endurance exercise program, control subjects underwent supervised exercise for 45 to 60 minutes/day for 5 days/week. Endurance exercise consisted of walking (including uphill treadmill walking), jogging, cycling, and rowing. Initial intensity of exercise was set at 60% to 70% of maximal oxygen consumption based on maximal oxygen consumption determined during a maximal exercise treadmill test and was gradually increased to 70% to 85% of maxi-
American Heart Journal Volume 138, Number 3, Part 1
Stein et al 569
Table I. Results of exercise training (n = 16) Baseline Maximal oxygen consumption (mL/kg/min) Maximal oxygen consumption (L/min) Maximal heart rate (beats/min) Resting heart rate (beats/min)
23.6 ± 3.8* 1.8 ± 0.5* 163 ± 8 71 ± 8†
After training 30.8 ± 5.2* 2.2 ± 0.7* 163 ± 9 66 ± 9†
*Statictically significant, P < .001. †P = .082.
mal oxygen consumption. The exercise prescription was adjusted on a weekly basis.
Measurement of heart rate variability Twenty-four-hour ambulatory electrocardiographic recordings were obtained for each patient with Marquette Series 8500 Holter recorders at baseline (during the stretching phase) and after completion of the exercise training program.
Data collection and analysis Holter recordings were analyzed with a Marquette SXP Laser Holter scanner (software version 5.8). All subjects were determined to be in sinus rhythm and without episodes of sustained atrial arrhythmias such as atrial fibrillation or multifocal atrial tachycardia. Beat-stream files, representing the time and classification of each QRS complex, were transferred to a Sun Sparcstation computer, where they underwent standard secondary editing and heart rate variability analysis procedures using previously validated techniques.18,19 Time domain analysis of successive normal-to-normal RR intervals. Time domain indexes of heart rate variability are
derived from statistical calculations performed on the set of normal-normal (N-N) interbeat intervals. Average heart rate in beats per minute and the following commonly reported time domain measures of heart rate variability were calculated for daytime (08:00 to 20:00), nighttime (00:00 to 0:600), and 24hour periods: AVGNN (average heart period in ms), SDNN (the standard deviation of N-Ns in milliseconds), SDANN (the standard deviation of 5-minute mean values of N-Ns for each 5-minute interval in milliseconds), SDNNIDX (the average of standard deviations of N-Ns for each 5-minute interval in milliseconds), rMSSD (the root mean square successive difference of N-Ns in milliseconds), and pNN50 (the percent of successive N-N differences >50 ms for each 5-minute interval). RMSSD and pNN50 primarily reflect parasympathetically mediated changes in heart rate.20 The other time domain variables reflect a mixture of parasympathetic, sympathetic, and other physiologic influences.20 Power spectral analysis of NN intervals. Power spectral analysis partitions the variance in the heart rate or heart period signal into its underlying components. The following standard frequency domain components of heart rate vari-
ability, measured in milliseconds squared, were computed in this analysis: (1) high-frequency power, the component of the total variance in heart period between 0.15 and 0.40 Hz, which primarily reflects vagal modulation of heart rate, 21 (2) low-frequency power, the component between 0.04 and 0.15 Hz, which reflects both sympathetic and parasympathetic modulation of heart rate,21 (3) very-low-frequency power, the component between 0.0033 to 0.04 Hz, which may represent the influence of the thermoregulatory 22 or renin-angiotensin23 systems, and (4) ultra-low-frequency power, the component between 1.15X10–5 and 0.00335 Hz, which reflects circadian variations. Total power (1.15X10 –5 to 0.40 Hz) for the total 24-hour cycle was also determined. The ratio of low- to high-frequency power, which has been suggested as a marker of sympathetic/parasympathetic balance,24 was also calculated. The methods used for spectral analysis have been previously described.19,25 In brief, the sequence of normal-to-normal intervals was resampled and filtered to provide a uniformly spaced time series. Missing or noisy segments were replaced by linear interpolation from the surrounding signal. The average of the sampled series was subtracted from the time series and a fast Fourier transforms performed to determine the frequency components underlying the cyclic activity in the sampled time series. Measurement of ultra-low and very-low-frequency power was based on en bloc analysis of the entire 24-hour recording.25 Other power indexes reported here reflect the average of 5-minute segments in which ≥80% of the beats are normal. Heart rate tachograms. To permit visual inspection of heart rate patterns at night when periodic heart rate rhythms associated with sleep-disordered breathing might affect indexes of heart rate variability, heart rate tachograms were generated from the interbeat interval files. Each interbeat interval was converted to its associated heart rate and the resultant heart rate time series plotted.
Statistical analysis Heart rate variability indexes were tested for normality and log-transformed when necessary to permit parametric statistical analysis. (All frequency domain indexes, rMSSD, and pNN50 required log transformation.) Paired t tests were performed to compare baseline and posttraining maximal oxygen consumption and maximal and resting heart rates for the training group. Student t tests compared baseline
American Heart Journal September 1999
570 Stein et al
Figure 2
Baseline and posttest SDNNIDX for comparison and training groups. Error bars indicate means ± 1 SD.
Table II. 24-hour time domain heart rate variability determined approximately 1 year apart in older adults continuing usual activities and older adults undergoing intensive exercise training Comparison group
AVGNN (ms) SDNN (ms) SDANN (ms) SDNNIDX (ms) rMSSD (ms) pNN50 (%)
Training group
Baseline
After test
P value
Baseline
After test
821 ± 106 135 ± 32 126 ± 32 48 ± 12 26 ± 10 6±5
832 ± 97 134 ± 32 123 ± 32 50 ± 12 27 ± 9 6±5
.35 .75 .55 .30 .61 .85
787 ± 45 126 ± 21 116 ± 21 46 ± 11 26 ± 16 5±7
823 ± 68 142 ± 25 129 ± 23 50 ± 15 27 ± 14 6±6
P value .04 .02 .04 .08 .43 .65
P values for before-to-after comparisons within groups. All values mean ± SD.
values of heart rate and heart rate variability between the exercise and the comparison groups to test for comparability at the start of the study. t Tests on differences between initial and final measurements of heart rates, heart rate variability, and respiratory rates were used to determine if changes over a 1-year period were different between the training and comparison groups. Each patient therefore acted as his or her own control. Correlational analysis was used to determined if changes in maximal oxygen consumption were associated with changes in indexes of heart rate variability. Statistical significance was set at P < .05. SPSSPC software (SPSS, Inc) was used for all statistical analyses.
Results Baseline measurement Initial indexes of heart rate and heart rate variability were compared between the training and comparison groups (no change in activity pattern). No statistically significant differences were observed.
Effect of exercise training on maximum oxygen consumption, maximum heart rate, and ectopy Exercise training resulted in a marked increase in maximal oxygen consumption, expressed in either liters per minute (22.2%) or mL/kg/min (30.5%) (P < .001). This increase occurred in all patients. Maximal heart rate was not altered (Table I). A trend to a lower supine resting heart rate as measured before each treadmill test did not attain statistical significance (P = .08, Table I). There was no difference in the number of extrasystoles before and after exercise training (P = .994).
Effect of exercise training on time domain indexes of heart rate variability As shown in Table II, exercise training resulted in a significant (12.6%) mean increase in SDNN (Figure 1) and a significant (11.2%) mean increase in SDANN (P
American Heart Journal Volume 138, Number 3, Part 1
Stein et al 571
Figure 3
Baseline and posttest AVGNN for comparison and training groups. Error bars indicate means ± 1 SD.
Table III. Daytime and nighttime indexes of heart rate and heart rate variability determined approximately 1 year apart in older adults continuing usual activities and older adults undergoing intensive exercise training Comparison group
Daytime (08:00-22:00) AVG NN (ms) SDNN (ms) SDANN (ms) SDNNIDX (ms) rMSSD (ms) pNN50 (%) Nighttime (00:00-06:00) AVG NN (ms) SDNN (ms) SDANN (ms) SDNNIDX (ms) rMSSD (ms) pNN50 (%)
Training group
Baseline
After test
P value
Baseline
After test
P value
752 ± 105 80 ± 24 77 ± 26 44 ± 11 22 ± 9 4±5
763 ± 100 95 ± 28 81 ± 27 44 ± 10 22 ± 9 4±6
.49 .24 .40 .62 .93 .89
726 ± 41 99 ± 20 87 ± 21 43 ± 8 21 ± 11 4±6
752 ± 75 118 ± 34 103 ± 33 48 ± 14 23 ± 9 4±4
.12 .07 .14 .11 .43 .91
968 ± 113 91 ± 30 65 ± 26 55 ± 16 32 ± 11 9±7
973 ± 99 92 ± 38 61 ± 28 61 ± 22 35 ± 13 11 ± 8
.47 .85 .49 .05 .05 .08
906 ± 81 87 ± 27 62 ± 22 52 ± 22 34 ± 32 9 ± 15
961 ± 75 88 ± 25 60 ± 19 55 ± 20 34 ± 23 10 ± 11
.01 .90 .68 .31 .85 .56
P values for before-to-after comparisons within groups. Values given as mean ± SD.
≤ .04), indexes that primarily reflect circadian rhythms. Although mean values for heart rate decreased for the exercise group (P = .04), when before-to-after changes in the exercise and comparison groups were compared, changes were significantly different only for SDNN and SDANN. There was no significant exercise-related change in any other 24-hour time domain index of heart rate variability. Figure 2 illustrates the similar changes over a 1-year period in 24-hour SDNNIDX for the training and comparison groups. When time domain indexes of heart rate variability were separately analyzed for daytime (08:00 to 22:00)
and nighttime (00:00 to 06:00) periods, exercise training tended to decrease mean daytime heart rates (P = .12) and increase mean SDNN (P = .07), but the only significant change in before-to-after values was a 6% increase in mean nighttime heart periods (ie, decrease in nighttime heart rates) among exercisers (P = .01, Table III and Figure 3).
Effect of exercise training on frequency domain indexes of heart rate variability When these data were analyzed with frequency domain indexes of heart rate variability, consistent with
American Heart Journal September 1999
572 Stein et al
Figure 4
Baseline and posttest natural logarithm high-frequency power for comparison and training groups. Error bars indicate means ± 1 SD.
Table IV. 24-hour frequency domain heart rate variability determined approximately 1 year apart in older adults continuing usual activities and older adults undergoing intensive exercise training Comparison group Baseline ln TP ln ULF ln VLF ln LF ln HF
9.8 ± 0.4 9.7 ± 0.4 7.1 ± 0.5 6.3 ± 0.6 5.0 ± 0.7
After test 9.8 ± 0.5 9.6 ± 0.5 7.2 ± 0.5 6.4 ± 0.7 5.1 ± 0.7
Training group P value .708 .600 .229 .115 .327
Baseline
After test
P value
9.6 ± 0.4 9.5 ± 0.4 7.0 ± 0.4 6.1 ± 0.8 4.9 ± 1.0
9.9 ± 0.4 9.7 ± 0.4 7.2 ± 0.5 6.3 ± 1.0 5.1 ± 1.1
.027 .041 .133 .151 .340
P values for before-to-after comparisons within groups. TP, total power; ULF, ultra-low frequency; VLF, very-low frequency; HF, high frequency; LF, low frequency.
findings in the time domain, significant increases were seen in natural logarithm total power and natural logarithm ultra-low-frequency power in the exercise group compared with the comparison group. Figure 4 illustrates the similar before-to-after changes of natural logarithm high-frequency power in both the exercise and comparison groups. When results were analyzed separately for daytime and nighttime frequency domain measures (data not shown), there were no differences between groups in the before-to-after change in heart rate variability, despite the significant before-to-after increases in natural logarithm very-low-frequency power and natural logarithm low-frequency power at night for the comparison group. When the change in the ratio of low- to high-frequency power, which has been suggested as an index of sympathetic activation,24 was com-
pared, there were no differences between groups and no significant change for either group. Similarly, the change in the coefficient of variance (SDNN/AVGNN), another index of autonomic balance, before-to-after was not significantly different between groups.
Relation between changes in heart rate variability and changes in maximal oxygen consumption When the correlations between changes in heart rate variability and changes in maximal oxygen consumption were determined, although exercise training increased 24-hour SDNN, there was no relation between the change in maximal oxygen consumption and the change in SDNN (Figure 5). A modest relation was observed between the change in maximal oxygen consumption in milliliters per kilogram
American Heart Journal Volume 138, Number 3, Part 1
Stein et al 573
Figure 5
Relation between change in maximal oxygen consumption and change in 24-hour SDNN.
per minute and the change in SDNNIDX at night (r = 0.45, P = .04), low-frequency power at night (r = 0.52, P = .02), high-frequency power at night (r = 0.45, P = .04), and 24-hour very-low-frequency power (r = 0.49, P = .03). Of these, the changes in SDNNIDX at night, low-frequency power at night, and very-lowfrequency power were highly correlated (correlation coefficient between 0.77 and 0.88, P ≤ .001). There were no significant relations between changes in heart rate variability and changes in the numbers of extrasystoles. There was also no sex effect. Changes in heart rate variability after 1 year in the comparison group. In general, there were no
significant differences between initial and final values of heart rate or heart rate variability in the comparison group, indicating the reproducibility of heart rate variability over a 1-year period in this older population (Tables II through IV, Figures 1 to 4). However, although there was no difference between groups in before-to-after changes, significant increases were found for nighttime values of the highly correlated indexes of natural logarithm very-low-frequency power, natural logarithm low-frequency power, and SDNNIDX. In particular, 5 patients in the training group and 5 in the comparison group had increases of ≥10 ms in SDNNIDX at night, and 4 in the training
group and 1 in the comparison group had decreases in SDNNIDX at night of ≥10 ms. These increases were associated with an increase in the percentage of time in which periodic heart rate rhythms, as seen on the heart rate tachogram, were present (Figure 6). In some cases, these patterns were similar to those associated with periodic limb movements or sleep disordered breathing such as sleep apnea. In others they were less periodic and seemed to reflect more frequent arousals. The decrease in SDNNIDX at night in one of the exercise subjects was associated with a marked decrease in periodic heart rate rhythms after exercise training.
Discussion The results of this study suggest that in healthy older adults the expected adaptive increase in maximal oxygen consumption in response to exercise training is associated with an increase in longer term indexes of heart rate variability. The clearest change is a decrease in nocturnal mean heart rate. Mean SDNN and SDANN, which reflect the longest term heart rate variability trends, are consequently increased. Low values for SDNN and SDANN have been associated with an increased risk for death after myocardial infarction, and low values for these indexes reflect
American Heart Journal September 1999
574 Stein et al
Figure 6
Ten-minute tachograms showing (top) quiet sleep, (middle) sleep apnea–like pattern in a different subject, and (bottom) large changes in heart rate, possibly from frequent arousals in same subject as (top).
the failure of heart rate to decrease substantially at night.26 The physiologic mechanism underlying these changes is not defined by this study. There were no significant changes in index means reflecting primarily vagal tone or in those reflecting short-term cardiac autonomic modulation in the exercise group. However, although group means in these variables were unaffected, changes in individual maximal oxygen consumption were correlated with individual changes in the intermediate term indexes nighttime SDNNIDX, nighttime low-frequency power, and 24hour very-low-frequency power, and the short-term index nighttime high-frequency power. Aerobic fitness has also been associated with a better quality of sleep, as reflected in shorter sleep onset latencies, less wake time after onset, fewer discrete sleep episodes, fewer sleep stage shifts during the initial portions of the night, a higher sleep efficiency, and more total slow waves in older men.27 It is possible that the decrease in heart rates at night after aerobic exercise training reflects improved sleep attributable to increased fitness.
It has been generally assumed that exercise training increases heart rate variability. However, the effect of exercise training on heart rate variability in this older adult population was modest (12.6% for mean SDNN) when compared to the substantial (more than 22%) gain in mean maximal oxygen consumption. Moreover, maximal oxygen consumption increased even among those exercise subjects in whom SDNN decreased (Figure 1). In fact, there is little evidence from prospective studies that an exercise training program will dramatically increase heart rate variability in older adults. The presumed relation between exercise training and increased heart rate variability is largely based on cross-sectional investigations comparing athletes and sedentary controls. Interestingly, the comparison group of this study suggests a difference between athletes and sedentary subjects: in the 2 athletic patients who engaged in regular endurance training, SDNN (Figure 1) and AVGNN at night (Figure 3) were noticeably higher, whereas baseline values for the remaining members of the comparison group were similar to those in the training group. In twin studies, maximal oxygen consumption improve-
American Heart Journal Volume 138, Number 3, Part 1
ment as a result of exercise training is strongly influenced by genetic endowment.28 It is possible that the high heart rate variability levels seen in cross-sectional studies in endurance athletes29 may in part reflect genetics and selection. Indexes of heart rate variability were remarkably similar over the 1-year period in the group, which did not change their usual activities. This confirms the reproducibility of heart rate variability indexes over intermediate periods of time5 and suggests that in healthy older adults the age-mediated decline in heart rate variability is relatively slow. A limitation of this study is that the comparison group did not consist entirely of sedentary individuals and that maximal oxygen consumption was not measured in the comparison group. However, the statistical design allowed each participant in the study to serve as his or her own control. Our assumption of little or no change in maximal oxygen consumption over a 1-year follow-up period among the comparison group that did not change their usual activities appears reasonable. The lack of change in nighttime heart rates and 24-hour heart rate variability in the comparison group strongly suggests that the changes seen in the exercise group can be attributed to aerobic exercise training. Our study suggests a relation between beneficial effects of exercise training on aerobic fitness and beneficial effects on 24-hour indexes of heart rate variability in normal older adults. Other measures of autonomic function, such as baroreceptor reflex sensitivity or response to orthostatic challenge, were not determined in this study. Because some of these measures convey prognostic information in addition to that provided by heart rate variability, they generate reasonable hypotheses for future investigation. This study, together with the compelling associations between heart rate variability and outcome, emphasizes the intriguing connections among modulation of heart rate, sleep, exercise training, and prognosis.
References 1. Fleg JL. Alterations in cardiovascular structure and function with advancing age. Am J Cardiol 1986;57:33C-44C. 2. Ehsani AA. Aging and the heart. West J Med 1993;159:501-2. 3. Bigger JT, Fleiss LF, Steiman RC, Rolnitzsky LM, Schneider WJ, Stein PK. RR variability in healthy, middle-age persons compared with patiems with chronic coronary heart disease or recent acute myocardial infarction. Circulation 1995;91:1936-43. 4. Stein PK, Bosner MS, Kleiger RE, Conger BM. Heart rate variability: a measure of cardiac autonomic tone. Am Heart J 1993; 127:1376-81. 5. Kleiger RE, Bigger JT, Bosner MS, Chung MK, Cook JR, Rolnitzsky LM, et al. Stability over time of variables measuring heart rate variability in normal subjects. Am J Cardiol 1991;68:626-30.
Stein et al 575
6. Tsuji H, Lawrence MG, Venditti FJ Jr, Manders ES, Evans JC, Feldman CL, et al. Impact of reduced heart rate variability on risk for cardiac events. The Framingham Heart Study. Circulation 1996; 94:2850-5. 7. Kleiger RE, Miller JP, Bigger JT Jr, Moss AJ, and the MPIP Research Group. Decreased heart rate variability and its association with increased mortality after acute myocardial infarction. Am J Cardiol 1987;59:256-62. 8. Odemuyiwa O, Farrell T, Malik M, Bashir Y, Poloniecki J, Ward DE, et al. The effect of age on the electrophysiological and autonomic correlates of sudden death after acute myocardial infarction. PACE 1991;14:2049-55. 9. Molgaard H, Sorensen KE, Bjerregaard P. Attenuated 24-h heart rate variability in apparently healthy subjects subsequently suffering sudden cardiac death. Clin Auto Rsch 1991;1:233-7. 10. Sandrone G, Mortara A, Torzillo D, La Rovere MT, Malliani A, Lombardi, F. Effects of beta blockers (atenolol or metoprolol) on heart rate variability after acute myocardial infarction. Am J Cardiol 1994;74:340-5. 11. Yusuf S, Peto R, Lewis J, Collins R, Sleight P. Beta blockade during and after myocardial infarction: an overview of the randomized trails. Prog Cardiovasc Dis 1985;XXVII:335-71. 12. The SOLVD Investigators. Effect ofenalapfil on survival in patients with reduced left ventricular ejection function and congestive heart failure. N Engl J Med 1991;325:293-302. 13. Binkley PF, Haas GJ, Starling RC, Nunziata E, Hatton PA, Leier CV, et al. Sustained augmentation of parasympathetic tone with angiotensinconverting enzyme inhibition in patients with congestive heart failure. J Am Col Cardiol 1993;21:655-61. 14. Stein PK, Rottman JN, Kleiger RE. Effect of 21 mg transdermal nicotine patches and smoking cessation on heart rate variability. Am J Cardiol 1996;77:701-5. 15. Goldsmith RL, Bigger JT Jr, Steinman RC, Fleiss JL. Comparison of 24-hour parasympathetic activity in endurance-trained and untrained younger men. J Am Coll Cardiol 1992;20:552-8. 16. Bruce RA. Exercise testing of patients with coronary heart disease. Ann Clin Res 1971;3:323-9. 17. Khort WM, Malley MT, Coggan AR, Spina R J, Ogawa T, Ehsani AA, et al. Effects of gender, age, and fitness level on the response of VO2 max to training in 60- to 71-yr-olds. J Appl Physiol 1991; 71:2004-11. 18. Malik M, Farrell T, Cripps T, Camm AJ. Heart rate variability in relation to prognosis after myocardial infarctions: selection of optimal processing techniques. Eur Heart J 1989;10:1060-74. 19. Berger, RD, Akselrod S, Gordon D, Cohen R. An efficient algorithm for spectral analysis of heart rate variability. IEEE Trans Biomed Eng 1986;9:900-4. 20. Kleiger RE, Stein PK, Bosner MS, Rottman JN. Time domain measurements of heart rate variability. Cardiol Clin 1992;10:487-98. 21. Akselrod S, Gordon D, Madwed JB, Snidman NC, Shannon DC, Cohen RI. Hemodynamic regulation investigated by spectral analysis. Am J Physiol (Heart Circ Physiol 18) 1985;249:H867-H875. 22. Fleisher LA, Frank SM, Sessler DI, Cheng C, Matsukawa T, Vannier CA. Thermoregulation and heart rate variability. Clin Sci 1996; 90:97-103. 23. Akselrod S, Gordon D, Ubel FA, Shannon DC, Barger AC, Cohen RI. Power spectrum analysis of heart rate fluctuation: a quantitative probe of beat-to-beat cardiovascular control. Science 1981;213:220-2. 24. Malliani A, Pagani M, Lombardi F, Cerutti S. Cardiovascular neural
576 Stein et al
regulation explored in the frequency domain. Circulation 1991; 84:482-92. 25. Rottman JN, Steinman RC, Albrecht P, Bigger JT, Rolnitzky LM, Fleiss JL. Efficient estimation of the heart period power spectrum suitable for physiologic or pharmacologic studies. Am J Cardiol 1990;66:1522-4. 26. Kleiger RE, Miller JP, Krone RJ, Bigger JT Jr, and the MPIP Research Group. The independence of cycle length variability and exercise testing on predicting mortality of patients surviving acute myocardial infarction. Am J Cardiol 1990;65:408-11.
American Heart Journal September 1999
27. King AC, Oman RF, Brassington GS, Bliwise DL, Haskell WL. Moderate-intensity exercise and self-rate quality of sleep in older adults. A randomized controlled trial. JAMA 1997;277:32-7. 28. Prud’homme D, Bouchard C, Leblanc C, Landry F, Fontaine E. Sensitivity of maximal aerobic power to training is genotype-dependent. Med Sci Sports Exerc 1984;16:489-93. 29. Sacknoff DM, Gleim GW, Stachenfeld N, Coplan NL. Effect of athletic training on heart rate variability. Am Heart J 1994; 127:1275-78.