Circadian Variation and Influence of Risk Factors on Heart Rate Variability in Healthy Subjects Henning Mglgaard, MD, Keld E. Sgrensen, MD, and Preben Bjerregaard, MD, PhD
Quantification of variations in instantaneous heart rate (HR) can be used to evaluate cardiac autonomic function. A 24-hour standard deviation of all normal RR intervals 6%-an index of parasympathetic activity. The 240hour standard deviation varied between 66 and 261 ms (median 139). Range for index of parasympathetic activity was 0.1 to 29.6% (median 4.4). Twenty percent of the interindividual variation in HR variability was explained by impact of risk factors. Standard deviation was uninfluenced by age, whereas parasympathetic activity decreased by increasing age. High physical training level was independently associated with significantly higher standard deviation (and parasympathetic activity) values during both day and night. Hourly figures of standard deviation decreased during the night, whereas parasympathetic activity increased and peaked early morning. Standard deviation values as low as those reported in high-risk patients were not observed, but comparable low values for, and lack of diurnal variation in, parasympathetic activity were seen in healthy subjects also. In conclusion, risk factors and, in particular, the physical training level have impact on 24hour HR variability in healthy subjects. This may prove valuable for modification of cardiac autonomic activity in patients. (Am J Cardiol 1991;68:777-794)
From the University Department of Cardiology, Skejby Sygehus, Arhus N, Denmark. Manuscript received February 5,199l; revised manuscript received and accepted May 29, 1991. Address for reprints: Henning Mplgaard, MD, University Department of Cardiology, Skejby Sygehus, DK-8200 Arhus N, Denmark.
nstantaneous heart rate (HR) is mainly determined by activity of the cardiac sympathetic and parasympathetic system. From basic physiologic experiments, correspondencebetween cardiac nervous activity and immediate RR interval changes has been found.1-4This is the principle for exploration of cardiac autonomic activity by means of HR variability. Measurement of HR variability in 24-hour Holter recordings gives access to study autonomic function both day and night,5-7 and during ambulatory conditions. This is important because arrhythmic cardiac eventsdependenton the autonomic system8,gappear to accumulate at certain times of the day.‘OJl In survivors of myocardial infarction, low 24-hour HR variability is independently associatedwith a high mortality.‘j Decreasedvariability has also beenobserved in victims of sudden cardiac death.‘* In prospective data from apparently healthy subjects, prognostic significance of HR for sudden cardiac death has been found.13 Individual HR level is partly determined by conventional risk factors.l4 Values for and factors determining HR variability in healthy subjectshave so far not been reported. This study describes data for 24-hour HR variability in healthy subjects, during ambulatory conditions, and evaluatesimpact of conventional risk factors on parameters of HR variability.
I
METHODS Material: One 24-hour Holter monitoring was recorded in 140 apparently healthy subjects, who were continuously without symptoms and signs of cardiopulmonary disease at an &year follow-up investigation. Criteria for healthy subjectswere: (1) that they considered themselveshealthy, (2) were free of cardiopulmonary symptoms, (3) had a normal clinical examination including blood pressure,(4) a normal 12-lead electrocardiogram, and (5) a normal chest x-ray. None were taking cardioactive drugs. The population consistedof 89 men and 51 women, mean age 53 years, range 40 to 77. Forty-seven subjects were aged 40 to 49 years, 60 were 50 to 59 years, and 33 were 60 to 77 years. Subjects were classified according to their participation in leisure-time physical activity as (1) active in competitive sports (active), (2) participants in leisure-time physical exercise (intermediate), and (3) not participants in such activities (passive).According to smoking HEART RATE VARIABILITY 777
BMDP statistical software was used for statistical analysis. Except for mean RR, distributions of values of the other HR variability parameters were right skewed and were therefore log transformed. For description of the population, median value and 2.5 and 97.5 percentiles are given for all variables. Paired comparisons were analyzed by t tests. Correlation between variability parameters was evaluated by Pearsoncorrelations and least-squaresregression lines of Y on X. Influence of risk factors on HR variability (log-transformed values) was evaluated by multiple linear regression (2-tailed, significance level 5%). Statistics:
TABLE I Heart Rate Variability in 140 Healthy Subjects Aged 40 to 77 Years (measured as the standard deviation and as percentage of successive RR interval differences > 6% and > 50 ms)* Percentiles Median T Mean
RR
D
N RRD-RRN SD
97.5
645 611 741 -43 78 68 41 0.3 0.3 0.2 0.4 0.2 0.2
1,089 1,019 1,375 -397 231 191 151 19.5 21.1 28.4 27.2 24.7 42.7
603-l 154 701-1445 -7--433
139 117
7
58-239
N
79
A
27-188
1 4
0.1-28.7 0.0-33.1 0.1-30.2 0.1-28.5 0.0-46.6
D
L
2.5
637-1253
D
rd T+tJ %DIF50
770 7 956 A -200
Range
T
T %DIF6%
812
LD
N
68-261
‘4.4
4.0 5.0 4.7 3.8 7.4
0.1-29.6
‘t A
*Values are shown for 24 hours, day and mghttime. Mean RR, standard deviation and day-night difference in mean RR, in ms. tp 6%/>50 ms; N = night; RRD-RRN = day-night difference in mean RR; SD = standard deviation; T = 24 hours.
habits the subjects were grouped as (1) nonsmokers, (2) moderate smokers (< 10 cigarettes/day or other forms of tobacco, or a combination), and (3) heavy smokers(>lO cigarettes/day). Sixty percent were nonsmokers and 62% were physically passive. Holter method: Twenty-four-hour ambulatory electrocardiograms were recorded by an Oxford 2-channel tape recorder during daily routine activities. Analysis was performed by the Reynolds Pathfinder II system and our own software, executed on a personal computer. The performance of this systemwith regard to QRS detection15and HR variability analysis has been previously reported5 Only cycles in which beats had normal morphologic characteristicswere used for HR variability analysis. All RR intervals > 1.5 secondswere validated, and all intervals >2.3 secondswere discharged, since no true RR interval in the 140 subjectswas >2.3 seconds.To correct for speederrors that may have oc&red during recording or replay, the timing track and speedsurveyor were used.5 Heart rate variability cahulations: Each 24-hour recording was split into daytime and nighttime. Because no diaries were available, nighttime was approximated according to a HR curve (beats/min). The typical pattern of the transition zone from awake to sleepand vice versa, present in most recordings, was included in daytime calculations. Nighttime therefore consistedof pure sleeping time. HR variability was calculated as (1) mean RR (mean of all RR intervals), (2) standard deviation of all RR intervals around mean RR, and (3) percentage of successiveRR interval differences >6%. These counts are also reported for numeric (>50 ms) changesin successiveRR intervals. 778
THE AMERICAN JOURNAL OF CARDIOLOGY VOLUME 68
RESULTS Median value and range for HR variability parameters are listed in Table I. Values for standard deviation were highest during the daytime. The lowest observed 24-hour standard deviation was 68 ms. Increasing day-night difference in mean RR was closely (r = -0.86, p 50 ms were significantly higher during the nighttime, whereas no such difference was present using a threshold of >6% (Table I). Twenty-four-hour figures for the percentageof successiveRR interval differences>6% were much less dependent on day-night differences in mean RR (Figure 1B), and values decreasedsignificantly by increasing age during the day and night (Figure 3, A and B). Physically passivesubjects had profoundly lower standard deviations and parasympathetic index values during both day and night (Table II). Smokers had lower HR variability values, but only during the day. During nighttime, standard deviation figures were higher in men. The circadian variation of mean hourly values of the 2 HR variability parameters are shown in Figure 4. After midnight standard deviation values initially decreased,whereasparasympathetic index figures showed a constant increase, reaching a peak in early morning. Figure 5 shows diurnal variation for 10% of the subjects with the highest and lowest 24-hour standard deviation. A similar, but substantially enhanced,pattern is seenfor subjects with high HR variability. DISCUSSION The clinical significance of HR variability has been related to the 24-hour standard deviation and percentage of successiveRR interval differences>50 ms, measured during ambulatory conditions.6y7J2The exact
SEPTEMBER 15. 1991
contribution of sympathetic and parasympathetic activity to the RR variability measuredby the standard deviation is not clear. Parasympathetic activity is associated with immediate alterations in length of the succeedingRR interval,1,16and can be estimated separately in different ways.7~17Estimation of pure sympathetic activity in ambulating subjects has not yet been possible. By calculating the spectral content and the Mean
RR day-Mean
density of the RR fluctuations, a better separation of sympathetic activity may be possible.17,18 Parasympathetic activity index: Ewing et al’ introduced threshold counts as a way to monitor cardiac parasympathetic damage in diabetic patients. By measuring the incidence of changes >50 ms in successive RR intervals, they clearly separated healthy and diabetic subjectswith clinical evidenceof parasympathetic
RR night
(ms)
5Oi
-150-200. -250-
-350. -4oo-
. .*
FIGURE 1. Relation between day-night
-450 -
dlfkrenee in mean RR (Mean RR dayMean RR night). A, twenty-four-hour standard devlatlon (sD-Totaltlnle) (r = -0.66, p 6% (% DIF 6% Totakime) (r = -0.34, p 50 ms and >6% were closely correlated (r = 0.97). Using 50 ms as a threshold increasednighttime values, whereasa threshold of 6% increased daytime counts. We have chosento emphasizevalues for relative changesbecause animal studies have shown that efferent vagal activity, both alone and during concomitant sympathetic activity, is related to proportional changesin RR intervals.‘T4
damage, and additionally found extremely low counts, and lack of day-night difference in counts, in patients with proven parasympathetic damage or surgically denervated hearts. Count values were unaltered by p blockade. These findings strongly suggestthat threshold counts represent vagus activity. In this study, the decreasein counts with increasing age and peak values in the early morning are in accordancewith this concept and with findings using other indexes for parasympaSD-day (ms)
316
200
100
25I
6
A 40
45
50.
I
1
I
55
60
65
70
I
c
75
81
AGE (years)
SD-night (ms)
L-L .:
,oo l **. 1:
’
.
50-
0:s
l -0
l /:: l
l
.
l
.
.
.-;i
.
-
.
-.
.’
.;.
.
*
l .
I--.
:..I
:
. .
::*:-y
0
l
.
l
.
l -
.
. . . 1
. .
B 40
45
I
I
I
50
55
60
I
!
65
70
I
75
81
AGE (years)
780
THE AMERICAN JOURNAL OF CARDIOLOGY VOLUME 68
SEPTEMBER 15, 1991
FIGURE 2. Heart rate vadbility in 140 healthy subjects, measured as the standard devhtion (SD), during the day (A) and nighttime (B), and shown in relation to age. Far age groups 40 to 49, SO to 59 and 60 to SO years, mean values (open limits are squares) and 95% cont&ne shown. Regression he: Y = 2.07 - 1. X (r = -0.02, p = 0.85) for day: Y = 2.12 0.004 - X (r = -024, p = 0.004) for nightthe.
high training level not only increasedmean values during both day and night, but also enhancedthe circadian variation. Twelve of the 14 subjects with high 24-hour standard deviations were physically trained (Figure 5). The significant decreasein percentageof successiveRR interval differences >6% that occurred with increasing age were likely due to decreasingefferent vagus activity, but could also be explained by lower responsiveness of the sinus node to autonomic activity in the elderly.21 However, parasympathetic activity values were also sig-
In addition, we found that there were more counts due to decelerations than to accelerations in instantaneous HR, for threshold >6% compared to >50 ms, which also favors relative changesas an index of vagus activity, and are in accordancewith the conceptof vagus as a brake on HR. Our results show that in 95% of the beat-tobeat variabilities during a 24-hour period, variability is 6% day (A) and ni&ttime (B), and shown in datiom to age. For age 6roups4Oto49,6OtoSSand6Oto6O yearq mean vabe (open squares) and 96% contidence limits are shown. Line of regresdon:Y=1.62-02=X(r=-O.38, p 40. C&f. = coefficients; %DIFG%D/N = percentageof successive RR interval differences >6%, daytime/nighttime; SDD = standard deviation, daytime; SDN = standard deviation, nighttime.
nificantly higher in older physically trained subjects. Hourly values of parasympathetic index show a clear increase during the nighttime, with a peak value at 5 A.M. (Figure 4). This parallels the increase in separate parasympathetic activity, as estimated by power spectral analysis, observed during the night in young subjects.*8 Standard deviation: The standard deviation is the averagedeviation of all intervals from the mean RR for a given time period. The length of that time period influences the interpretation of the autonomic contribution to the observedHR variability. In the experimental setting, without HR level shifts, figures for standard deviation are positively associatedwith cardiac vagal traffic.’ Owing to simple mathematics, standard deviation figures are, theoretically, greatly dependent on trends and mean HR level shifts for the time period considered. Huikuri et a122calculated standard deviation per 5-minute segments and found lower mean hourly values and, in contrast to this study, highest val-
ues during the night. By using 5-minute values, some of the variability is initially averaged out. Also, HR is significantly more stationary in 5-minute than l-hour periods, and therefore standard deviation values are much lessinfluenced by general HR shifts that occur often during the daytime. Power spectral data from 28 healthy subjects showed that the most prominent day/night shift was the decreasein power of the low frequency peak, assumedto reflect sympathetic activity.18The lower mean and hourly values of the standard deviation during the night must therefore partly be caused by withdrawal of sympathetic activity. Nighttime standard deviation values probably reflect paraDiF 6% (%)
SD (ms)
130 100 140,
SD (ms)
DiF 6% (%)
t 10
P
9
140
10
120
a
130
9
110 -
7
I
120110-
90
ioo-
a0
6
1
t5 4
Low HRV
90-
70
3
90-
60
2
70-
50
1 2i
24
04
08
Time
FIGURE 4. Circadiin variation in hourly vabes of the standard deviation (SD) (closed &c/es) and perceMqe of successive RR interval crifkrences >6% (DiF 6%) (open &c/es) in 140 healthy subjects. Mean value and etandard error of the mean for each hour are shown.
782
THE AMERICAN JOURNAL OF CARDIOLOGY VOLUME 68
FIGURE 5. Diurnal variatii in houlry values of the standard &via6011 (SD) (closed circles) and pementage of succedve RR intewal diffemms >G%(DiF 6%) (open circles) in healthy subjscts wtth hii and low heart rate variability (HRV). Mean heudy values for the 10% (14 subjects) with the higheet 24hew standard dovbtbn (High HRV) are shown on llpper 2 aavos, and flgures for the 16% with bwest 24-r standard dovlation (Low HRV) are shown on lower 2 curves.
SEPTEMBER 15, 1991
sympathetic activity more directly as also shown by the decreaseby increasing age (Figure 2B). The 24-hour standard deviation was very intimately correlated to the day/night difference in mean RR. Standard deviation values increased very closely by increasing difference in mean RR day/night (Figure 1A). The day-night ratio in mean RR was not different for young or old subjects. This association probably representsthe effect of increasedbasic parasympathetic activity. Physically trained subjectshave a generally increased parasympathetic activity,23 and lower HRs, in particular, at rest and sleep.In the multivariate analysis physically trained subjects had higher standard deviations during both the day and night, and this effect of physical training was still present after accounting for the mean RR level. This strongly indicates that 24-hour standard deviation values increase by increasing basic parasympathetic activity. Twenty percent of the interindividual variation could be explained by differencesin risk factors. Smokers have lower standard deviation figures, but only during the daytime. Power spectral data indicate that this is due to reduced vagal activity.24 A consistent finding in studies on HR is that men have a lower HR during the night than women.l4 This finding and our observation that the standard deviation was higher during the night in men may be attributed to higher parasympathetic activity in men. The gender differences are not measurable in pharmacologically denervated healthy subjects.25 The intraindividual coefficient of variation for 24hour standard deviation figures between repeated recordings has been shown to be below 7%.22Similarly, minimal HR has been found to be a stable individual parameter over long periods of time.14 Heart rate variability and heart disease: A 24-hour standard deviation
Quantification of variations in instantaneous heart rate (HR) can be used to evaluate cardiac autonomic function. A 24-hour standard deviation of all normal RR intervals 6%-an index of parasympathetic activity. The 240hour standard deviation varied between 66 and 261 ms (median 139). Range for index of parasympathetic activity was 0.1 to 29.6% (median 4.4). Twenty percent of the interindividual variation in HR variability was explained by impact of risk factors. Standard deviation was uninfluenced by age, whereas parasympathetic activity decreased by increasing age. High physical training level was independently associated with significantly higher standard deviation (and parasympathetic activity) values during both day and night. Hourly figures of standard deviation decreased during the night, whereas parasympathetic activity increased and peaked early morning. Standard deviation values as low as those reported in high-risk patients were not observed, but comparable low values for, and lack of diurnal variation in, parasympathetic activity were seen in healthy subjects also. In conclusion, risk factors and, in particular, the physical training level have impact on 24hour HR variability in healthy subjects. This may prove valuable for modification of cardiac autonomic activity in patients. (Am J Cardiol 1991;68:777-794)
From the University Department of Cardiology, Skejby Sygehus, Arhus N, Denmark. Manuscript received February 5,199l; revised manuscript received and accepted May 29, 1991. Address for reprints: Henning Mplgaard, MD, University Department of Cardiology, Skejby Sygehus, DK-8200 Arhus N, Denmark.
nstantaneous heart rate (HR) is mainly determined by activity of the cardiac sympathetic and parasympathetic system. From basic physiologic experiments, correspondencebetween cardiac nervous activity and immediate RR interval changes has been found.1-4This is the principle for exploration of cardiac autonomic activity by means of HR variability. Measurement of HR variability in 24-hour Holter recordings gives access to study autonomic function both day and night,5-7 and during ambulatory conditions. This is important because arrhythmic cardiac eventsdependenton the autonomic system8,gappear to accumulate at certain times of the day.‘OJl In survivors of myocardial infarction, low 24-hour HR variability is independently associatedwith a high mortality.‘j Decreasedvariability has also beenobserved in victims of sudden cardiac death.‘* In prospective data from apparently healthy subjects, prognostic significance of HR for sudden cardiac death has been found.13 Individual HR level is partly determined by conventional risk factors.l4 Values for and factors determining HR variability in healthy subjectshave so far not been reported. This study describes data for 24-hour HR variability in healthy subjects, during ambulatory conditions, and evaluatesimpact of conventional risk factors on parameters of HR variability.
I
METHODS Material: One 24-hour Holter monitoring was recorded in 140 apparently healthy subjects, who were continuously without symptoms and signs of cardiopulmonary disease at an &year follow-up investigation. Criteria for healthy subjectswere: (1) that they considered themselveshealthy, (2) were free of cardiopulmonary symptoms, (3) had a normal clinical examination including blood pressure,(4) a normal 12-lead electrocardiogram, and (5) a normal chest x-ray. None were taking cardioactive drugs. The population consistedof 89 men and 51 women, mean age 53 years, range 40 to 77. Forty-seven subjects were aged 40 to 49 years, 60 were 50 to 59 years, and 33 were 60 to 77 years. Subjects were classified according to their participation in leisure-time physical activity as (1) active in competitive sports (active), (2) participants in leisure-time physical exercise (intermediate), and (3) not participants in such activities (passive).According to smoking HEART RATE VARIABILITY 777
BMDP statistical software was used for statistical analysis. Except for mean RR, distributions of values of the other HR variability parameters were right skewed and were therefore log transformed. For description of the population, median value and 2.5 and 97.5 percentiles are given for all variables. Paired comparisons were analyzed by t tests. Correlation between variability parameters was evaluated by Pearsoncorrelations and least-squaresregression lines of Y on X. Influence of risk factors on HR variability (log-transformed values) was evaluated by multiple linear regression (2-tailed, significance level 5%). Statistics:
TABLE I Heart Rate Variability in 140 Healthy Subjects Aged 40 to 77 Years (measured as the standard deviation and as percentage of successive RR interval differences > 6% and > 50 ms)* Percentiles Median T Mean
RR
D
N RRD-RRN SD
97.5
645 611 741 -43 78 68 41 0.3 0.3 0.2 0.4 0.2 0.2
1,089 1,019 1,375 -397 231 191 151 19.5 21.1 28.4 27.2 24.7 42.7
603-l 154 701-1445 -7--433
139 117
7
58-239
N
79
A
27-188
1 4
0.1-28.7 0.0-33.1 0.1-30.2 0.1-28.5 0.0-46.6
D
L
2.5
637-1253
D
rd T+tJ %DIF50
770 7 956 A -200
Range
T
T %DIF6%
812
LD
N
68-261
‘4.4
4.0 5.0 4.7 3.8 7.4
0.1-29.6
‘t A
*Values are shown for 24 hours, day and mghttime. Mean RR, standard deviation and day-night difference in mean RR, in ms. tp 6%/>50 ms; N = night; RRD-RRN = day-night difference in mean RR; SD = standard deviation; T = 24 hours.
habits the subjects were grouped as (1) nonsmokers, (2) moderate smokers (< 10 cigarettes/day or other forms of tobacco, or a combination), and (3) heavy smokers(>lO cigarettes/day). Sixty percent were nonsmokers and 62% were physically passive. Holter method: Twenty-four-hour ambulatory electrocardiograms were recorded by an Oxford 2-channel tape recorder during daily routine activities. Analysis was performed by the Reynolds Pathfinder II system and our own software, executed on a personal computer. The performance of this systemwith regard to QRS detection15and HR variability analysis has been previously reported5 Only cycles in which beats had normal morphologic characteristicswere used for HR variability analysis. All RR intervals > 1.5 secondswere validated, and all intervals >2.3 secondswere discharged, since no true RR interval in the 140 subjectswas >2.3 seconds.To correct for speederrors that may have oc&red during recording or replay, the timing track and speedsurveyor were used.5 Heart rate variability cahulations: Each 24-hour recording was split into daytime and nighttime. Because no diaries were available, nighttime was approximated according to a HR curve (beats/min). The typical pattern of the transition zone from awake to sleepand vice versa, present in most recordings, was included in daytime calculations. Nighttime therefore consistedof pure sleeping time. HR variability was calculated as (1) mean RR (mean of all RR intervals), (2) standard deviation of all RR intervals around mean RR, and (3) percentage of successiveRR interval differences >6%. These counts are also reported for numeric (>50 ms) changesin successiveRR intervals. 778
THE AMERICAN JOURNAL OF CARDIOLOGY VOLUME 68
RESULTS Median value and range for HR variability parameters are listed in Table I. Values for standard deviation were highest during the daytime. The lowest observed 24-hour standard deviation was 68 ms. Increasing day-night difference in mean RR was closely (r = -0.86, p 50 ms were significantly higher during the nighttime, whereas no such difference was present using a threshold of >6% (Table I). Twenty-four-hour figures for the percentageof successiveRR interval differences>6% were much less dependent on day-night differences in mean RR (Figure 1B), and values decreasedsignificantly by increasing age during the day and night (Figure 3, A and B). Physically passivesubjects had profoundly lower standard deviations and parasympathetic index values during both day and night (Table II). Smokers had lower HR variability values, but only during the day. During nighttime, standard deviation figures were higher in men. The circadian variation of mean hourly values of the 2 HR variability parameters are shown in Figure 4. After midnight standard deviation values initially decreased,whereasparasympathetic index figures showed a constant increase, reaching a peak in early morning. Figure 5 shows diurnal variation for 10% of the subjects with the highest and lowest 24-hour standard deviation. A similar, but substantially enhanced,pattern is seenfor subjects with high HR variability. DISCUSSION The clinical significance of HR variability has been related to the 24-hour standard deviation and percentage of successiveRR interval differences>50 ms, measured during ambulatory conditions.6y7J2The exact
SEPTEMBER 15. 1991
contribution of sympathetic and parasympathetic activity to the RR variability measuredby the standard deviation is not clear. Parasympathetic activity is associated with immediate alterations in length of the succeedingRR interval,1,16and can be estimated separately in different ways.7~17Estimation of pure sympathetic activity in ambulating subjects has not yet been possible. By calculating the spectral content and the Mean
RR day-Mean
density of the RR fluctuations, a better separation of sympathetic activity may be possible.17,18 Parasympathetic activity index: Ewing et al’ introduced threshold counts as a way to monitor cardiac parasympathetic damage in diabetic patients. By measuring the incidence of changes >50 ms in successive RR intervals, they clearly separated healthy and diabetic subjectswith clinical evidenceof parasympathetic
RR night
(ms)
5Oi
-150-200. -250-
-350. -4oo-
. .*
FIGURE 1. Relation between day-night
-450 -
dlfkrenee in mean RR (Mean RR dayMean RR night). A, twenty-four-hour standard devlatlon (sD-Totaltlnle) (r = -0.66, p 6% (% DIF 6% Totakime) (r = -0.34, p 50 ms and >6% were closely correlated (r = 0.97). Using 50 ms as a threshold increasednighttime values, whereasa threshold of 6% increased daytime counts. We have chosento emphasizevalues for relative changesbecause animal studies have shown that efferent vagal activity, both alone and during concomitant sympathetic activity, is related to proportional changesin RR intervals.‘T4
damage, and additionally found extremely low counts, and lack of day-night difference in counts, in patients with proven parasympathetic damage or surgically denervated hearts. Count values were unaltered by p blockade. These findings strongly suggestthat threshold counts represent vagus activity. In this study, the decreasein counts with increasing age and peak values in the early morning are in accordancewith this concept and with findings using other indexes for parasympaSD-day (ms)
316
200
100
25I
6
A 40
45
50.
I
1
I
55
60
65
70
I
c
75
81
AGE (years)
SD-night (ms)
L-L .:
,oo l **. 1:
’
.
50-
0:s
l -0
l /:: l
l
.
l
.
.
.-;i
.
-
.
-.
.’
.;.
.
*
l .
I--.
:..I
:
. .
::*:-y
0
l
.
l
.
l -
.
. . . 1
. .
B 40
45
I
I
I
50
55
60
I
!
65
70
I
75
81
AGE (years)
780
THE AMERICAN JOURNAL OF CARDIOLOGY VOLUME 68
SEPTEMBER 15, 1991
FIGURE 2. Heart rate vadbility in 140 healthy subjects, measured as the standard devhtion (SD), during the day (A) and nighttime (B), and shown in relation to age. Far age groups 40 to 49, SO to 59 and 60 to SO years, mean values (open limits are squares) and 95% cont&ne shown. Regression he: Y = 2.07 - 1. X (r = -0.02, p = 0.85) for day: Y = 2.12 0.004 - X (r = -024, p = 0.004) for nightthe.
high training level not only increasedmean values during both day and night, but also enhancedthe circadian variation. Twelve of the 14 subjects with high 24-hour standard deviations were physically trained (Figure 5). The significant decreasein percentageof successiveRR interval differences >6% that occurred with increasing age were likely due to decreasingefferent vagus activity, but could also be explained by lower responsiveness of the sinus node to autonomic activity in the elderly.21 However, parasympathetic activity values were also sig-
In addition, we found that there were more counts due to decelerations than to accelerations in instantaneous HR, for threshold >6% compared to >50 ms, which also favors relative changesas an index of vagus activity, and are in accordancewith the conceptof vagus as a brake on HR. Our results show that in 95% of the beat-tobeat variabilities during a 24-hour period, variability is 6% day (A) and ni&ttime (B), and shown in datiom to age. For age 6roups4Oto49,6OtoSSand6Oto6O yearq mean vabe (open squares) and 96% contidence limits are shown. Line of regresdon:Y=1.62-02=X(r=-O.38, p 40. C&f. = coefficients; %DIFG%D/N = percentageof successive RR interval differences >6%, daytime/nighttime; SDD = standard deviation, daytime; SDN = standard deviation, nighttime.
nificantly higher in older physically trained subjects. Hourly values of parasympathetic index show a clear increase during the nighttime, with a peak value at 5 A.M. (Figure 4). This parallels the increase in separate parasympathetic activity, as estimated by power spectral analysis, observed during the night in young subjects.*8 Standard deviation: The standard deviation is the averagedeviation of all intervals from the mean RR for a given time period. The length of that time period influences the interpretation of the autonomic contribution to the observedHR variability. In the experimental setting, without HR level shifts, figures for standard deviation are positively associatedwith cardiac vagal traffic.’ Owing to simple mathematics, standard deviation figures are, theoretically, greatly dependent on trends and mean HR level shifts for the time period considered. Huikuri et a122calculated standard deviation per 5-minute segments and found lower mean hourly values and, in contrast to this study, highest val-
ues during the night. By using 5-minute values, some of the variability is initially averaged out. Also, HR is significantly more stationary in 5-minute than l-hour periods, and therefore standard deviation values are much lessinfluenced by general HR shifts that occur often during the daytime. Power spectral data from 28 healthy subjects showed that the most prominent day/night shift was the decreasein power of the low frequency peak, assumedto reflect sympathetic activity.18The lower mean and hourly values of the standard deviation during the night must therefore partly be caused by withdrawal of sympathetic activity. Nighttime standard deviation values probably reflect paraDiF 6% (%)
SD (ms)
130 100 140,
SD (ms)
DiF 6% (%)
t 10
P
9
140
10
120
a
130
9
110 -
7
I
120110-
90
ioo-
a0
6
1
t5 4
Low HRV
90-
70
3
90-
60
2
70-
50
1 2i
24
04
08
Time
FIGURE 4. Circadiin variation in hourly vabes of the standard deviation (SD) (closed &c/es) and perceMqe of successive RR interval crifkrences >6% (DiF 6%) (open &c/es) in 140 healthy subjects. Mean value and etandard error of the mean for each hour are shown.
782
THE AMERICAN JOURNAL OF CARDIOLOGY VOLUME 68
FIGURE 5. Diurnal variatii in houlry values of the standard &via6011 (SD) (closed circles) and pementage of succedve RR intewal diffemms >G%(DiF 6%) (open circles) in healthy subjscts wtth hii and low heart rate variability (HRV). Mean heudy values for the 10% (14 subjects) with the higheet 24hew standard dovbtbn (High HRV) are shown on llpper 2 aavos, and flgures for the 16% with bwest 24-r standard dovlation (Low HRV) are shown on lower 2 curves.
SEPTEMBER 15, 1991
sympathetic activity more directly as also shown by the decreaseby increasing age (Figure 2B). The 24-hour standard deviation was very intimately correlated to the day/night difference in mean RR. Standard deviation values increased very closely by increasing difference in mean RR day/night (Figure 1A). The day-night ratio in mean RR was not different for young or old subjects. This association probably representsthe effect of increasedbasic parasympathetic activity. Physically trained subjectshave a generally increased parasympathetic activity,23 and lower HRs, in particular, at rest and sleep.In the multivariate analysis physically trained subjects had higher standard deviations during both the day and night, and this effect of physical training was still present after accounting for the mean RR level. This strongly indicates that 24-hour standard deviation values increase by increasing basic parasympathetic activity. Twenty percent of the interindividual variation could be explained by differencesin risk factors. Smokers have lower standard deviation figures, but only during the daytime. Power spectral data indicate that this is due to reduced vagal activity.24 A consistent finding in studies on HR is that men have a lower HR during the night than women.l4 This finding and our observation that the standard deviation was higher during the night in men may be attributed to higher parasympathetic activity in men. The gender differences are not measurable in pharmacologically denervated healthy subjects.25 The intraindividual coefficient of variation for 24hour standard deviation figures between repeated recordings has been shown to be below 7%.22Similarly, minimal HR has been found to be a stable individual parameter over long periods of time.14 Heart rate variability and heart disease: A 24-hour standard deviation
