Marek Malik, U-ID, MD, Thomas Farrell, MD, and A. John Camm, MD
This study examined heart rate (HR) variability in patients surviving acute myocardial infarction (AMI) to find the optimum time and duration of cording of the ambulatory electrocardiogram for the prediction of the risk of sudden cardiac death, or serious arrhythmic events, or both. Twenty patients (group I) who initially survived an AMI but later experienced serious events (death or symptomatic sustained ventricular tachycardia) during a 6-month follow-up were compared with 20 patients (group II) who remained free of complications for >6 months after discharge. Groups I and II were matched with regard to age, gender, infarct site, ejection fraction, and p-blocker treatment. HR variability was assessed in the 24-hour electrocardiograms recorded during the first 2 weeks after an AMI and in various portions of the complete 24hour recording, with both the beginning and the length of the analyzed portion varied by 20 minutes (a total of 5,113 possibilities). The maximum reduction of HR variability in group I patients was systematically found when assessing HR variabil in recordings starting approximately at 6 A.M. an fasting for approximately 6 hours. the low-risk patient, the diurnai bility is more marked than in the and the long-term components of HR variability due to the diurnal variation must be included in
From the Department of Cardiological Sciences, St. George’s Hospital Medical School, London, England. This study was supported in part by the British Heart Foundation, London, England. Manuscript received April 19, 1990; revised manuscript received and accepted June 21, 1990. Address for reprints: Marek Malik, PhD, MD, Department of Cardiological Sciences, St. George’s Hospital Medical School, Cranmer Terrace, London SW1 7 ORE, England.
he importance of heart rate (FIR) variability for the stratification of patients who are at high risk after acute myocardial infarction (AMI) has well documented.im3Although 1 of the first studies reported used short-term electrocardiograms,4the majority of studies assessedHR variability in long-term electrocardiograms. In this study, HR variability was measuredfrom different portions of 24-hour electrocardiographic recordings of patients after an AMI. The aim of the study was to investigate the circadian rhythm of HR variability and its influence on the predictive value of HR variability in patients after an AMI. atients: During the 19 months of the study, 320 patients who were admitted to the hospital with AMI survived to discharge. AMI was diagnosedas the presence of 2 of the 3 following criteria: chest pain of ischemit type lasting 120 minutes; a sequential elevation and fall in the plasma concentrations of aspartate transaminase, ,&hydroxybutyric dehydrogenase or creatine phosphokinase,or a combination, with a peak concentration at least twice the upper limit of the reference range of our laboratory; development of new pathologic Q waves or persistent ST-T changes suggestiveof nonQ-wave infarction. Patients were not included if they had noncardiac diseaselikely to influence mortality, important nonischemic cardiac disease,a history of cardiac surgery or permanent pacemaker insertion, and if they refused or were unable to be followed-up. Patients with atria1 fibrillation were excluded becausethis interferes with the HR variability analysis. Patients with bundle branch block or ventricular preexcitation were excluded becausethe wide electrocardiographic patterns disturb the computer analysis of the 24-hour recordings. During the first 6 months of follow-up, 12 patients died from cardiac arrhythmia: 11 died suddenly (during the first hour from the onset of symptoms or durin sleep) and 1 died of heart failure after documented episodesof clinical ventricular tachycardia. Eight patients dev~lQpeddocumented,symptomatic sustainedventricular tachycardia. These 20 patients comprised group I. Of the remaining 300 patients, 20 were selected for a case-control study to assessthe different spectral components of the HR variability in the high-risk patients. The control group was selected to match with group I with regard to gender, age, infarct site, ejection fraction and treatment with ,f3blockers during and after hospitalization. Complete correspondencewas required with re-
THE AMERICAN JOURNAL OF CARDIOLOGY
NOVEMBER 1, 1990
spect to gender (15 male pairs, 5 female pairs), infarct site (8 inferior with Q waves,7 anterior with Q waves, 3 anterior and non-Q, 2 inferior and non-Q) and @blocker treatment (3 pairs of patients). An absolute correspondencein respect to the numerical values of the 2 remaining parameters was not always possible and patients with near values were selected. Statistically (paired t test), there were no differences between the groups regarding age (64 f 6 years [mean f standard deviation] for both groups) and ejection fraction (49 f 13 for group I and 49 f 14% for group II). Holter recording technique: Two-channel 24-hour electrocardiographic recordings (modified lead III and CM5) were obtained using a Tracker recorder (Reynolds Medical Ltd., United Kingdom). In all patients, the recordings were made at a median of 7 days (range 5 to 9) after hospital admission. There was no significant difference between the groups regarding the interval between admission and recording time. A commercially available system for long-term electrocardiographic analysis (Pathfinder III, Mk 2, Reynolds Medical Ltd.) was used to obtain the sequenceof durations of intervals betweenadjacent QRS complexesof normal “supraventricular” morphology for each patient. The exact time of each QRS complex was also derived from the recordings. Measurement of heart rate variability: For each patient, HR variability was measuredin different portions of the 24-hour time interval, with the beginning and duration of the analyzed portion varied. The onset of the analyzed portions varied by 20 minutes (72 values ranging from 0O:OOto 23:40) and the duration of the portions analyzed also varied by 20 minutes (72 values ranging from 20 minutes to 24 hours). In total, 5,184, including 72 identical 24-hour intervals (5,113 different portions of each record) were analyzed. Where the analyzed interval involved the physical end of the recording, the 24-hour electrocardiogram was recorded as a continuous loop-i.e., the end of the recording was extended with the properly timed beginning of the recording from the previous day.
1250
For each time interval analyzed, the frequency distribution of durations of normal-to-normal RR intervals was constructed. HR variability was expressedas the baseline width of this distribution curve measured by the method of minimum square difference interpolation.3 For each interval analyzed, the mean duration of RR intervals was also computed. The more usual method expressingHR variability by the standard deviation of RR interval durations’ was not used becausethe computer analysisof the recording system that we used is known to be affected by recording noise and misrecognition artifact.3 The method of triangular interpolation of the frequency distribution of RR interval duration that we usedis not affectedby low levels of noise and artifact.3*5 Normalized heart rate variability: Previous studies2 have shown that HR and HR variability constitute independentprognostic factors in patients after myocardial infarction. However, these parametersare likely to be correlated. A pathologic processthat shortens the RR intervals is also likely to reduce the absolutevalue of the differences between individual RR intervals, To overcome the interdependence between HR and HR variability, we used the concept of normalized HR variability, which excludes the basic mathematic influence of the mean HR. For each analyzed portion of the recording, the normalized HR variability was obtained by dividing its nonnormalized value by the mean duration of RR intervals within the same portion of the record. Statistical analysis: Paired t testswere used to compare the values found in the correspondingportions of the electrocardiographic recordings obtained in groups I and II. The tests were performed for values of HR. HR variability and normalized HR variability in all analyzed portions of the recordings and the study involved more than 15,000individual t tests.However, the values measured in the individually analyzed portions of the electrocardiographic recordings were not independent (mainly because of frequent overlaps of the portions). Therefore, the problems resulting from the multiplicity of statistical tests were reduced. HRV (ms)
RR (ms) 280
1125
240 200
1000
160 875 120
0
2
4
6
8
10
12
14
16
18
20
22
Time of the day
0
2
4
6
8
10
12
14
16
18
20
22
Time of the day
FIGURE 1. Mean cycle length and short-term heart rate variability measured in separate 40-minute intervals In group I (Positive) and group II (Negative) electrocardiograms. Leff, mean ms duration of normal-to-normal intervals (RR). Rig/H, heart rate variability as the baseline width (in ms) of the sample density distributions of RR interval durations. HRV = mean heart rate values (k standard errors).
1050
THE AMERICAN JOURNAL OF CARDIOLOGY VOLUME 66
RESULTS Circadian
rhythm
of heart
length and short-term HR variability were measured in separate 40-minute portions of 24-hour recordings. In group II recordings, the physiologic diurnal variation of the mean cycle length was preserved, with the slowest HR during the night and the fastest during the afternoon. In contrast, the mean cycle length of the
rate and its variability:
Figure 1 shows circadian variations of mean cycle length and short-term HR variability found in both groups of patients. Mean values f standard errors are shown for both groups; in each patient, the mean cycle
FIGURE 2. Probability levels of the significance at which heart rate variability was reduced in group I (paired I test) compared with that in group II for all analyzed intervals. The vertical axis of the plot corresponds to the beginning of the analyzed interval (in hours, starting at midnight); the horilontal axis corresponds to the duration of the interval (in hours). For each analyzed Interval, the graphic symbols express the significance level between both groups: spaces = no significant difference; dots = p
This study examined heart rate (HR) variability in patients surviving acute myocardial infarction (AMI) to find the optimum time and duration of cording of the ambulatory electrocardiogram for the prediction of the risk of sudden cardiac death, or serious arrhythmic events, or both. Twenty patients (group I) who initially survived an AMI but later experienced serious events (death or symptomatic sustained ventricular tachycardia) during a 6-month follow-up were compared with 20 patients (group II) who remained free of complications for >6 months after discharge. Groups I and II were matched with regard to age, gender, infarct site, ejection fraction, and p-blocker treatment. HR variability was assessed in the 24-hour electrocardiograms recorded during the first 2 weeks after an AMI and in various portions of the complete 24hour recording, with both the beginning and the length of the analyzed portion varied by 20 minutes (a total of 5,113 possibilities). The maximum reduction of HR variability in group I patients was systematically found when assessing HR variabil in recordings starting approximately at 6 A.M. an fasting for approximately 6 hours. the low-risk patient, the diurnai bility is more marked than in the and the long-term components of HR variability due to the diurnal variation must be included in
From the Department of Cardiological Sciences, St. George’s Hospital Medical School, London, England. This study was supported in part by the British Heart Foundation, London, England. Manuscript received April 19, 1990; revised manuscript received and accepted June 21, 1990. Address for reprints: Marek Malik, PhD, MD, Department of Cardiological Sciences, St. George’s Hospital Medical School, Cranmer Terrace, London SW1 7 ORE, England.
he importance of heart rate (FIR) variability for the stratification of patients who are at high risk after acute myocardial infarction (AMI) has well documented.im3Although 1 of the first studies reported used short-term electrocardiograms,4the majority of studies assessedHR variability in long-term electrocardiograms. In this study, HR variability was measuredfrom different portions of 24-hour electrocardiographic recordings of patients after an AMI. The aim of the study was to investigate the circadian rhythm of HR variability and its influence on the predictive value of HR variability in patients after an AMI. atients: During the 19 months of the study, 320 patients who were admitted to the hospital with AMI survived to discharge. AMI was diagnosedas the presence of 2 of the 3 following criteria: chest pain of ischemit type lasting 120 minutes; a sequential elevation and fall in the plasma concentrations of aspartate transaminase, ,&hydroxybutyric dehydrogenase or creatine phosphokinase,or a combination, with a peak concentration at least twice the upper limit of the reference range of our laboratory; development of new pathologic Q waves or persistent ST-T changes suggestiveof nonQ-wave infarction. Patients were not included if they had noncardiac diseaselikely to influence mortality, important nonischemic cardiac disease,a history of cardiac surgery or permanent pacemaker insertion, and if they refused or were unable to be followed-up. Patients with atria1 fibrillation were excluded becausethis interferes with the HR variability analysis. Patients with bundle branch block or ventricular preexcitation were excluded becausethe wide electrocardiographic patterns disturb the computer analysis of the 24-hour recordings. During the first 6 months of follow-up, 12 patients died from cardiac arrhythmia: 11 died suddenly (during the first hour from the onset of symptoms or durin sleep) and 1 died of heart failure after documented episodesof clinical ventricular tachycardia. Eight patients dev~lQpeddocumented,symptomatic sustainedventricular tachycardia. These 20 patients comprised group I. Of the remaining 300 patients, 20 were selected for a case-control study to assessthe different spectral components of the HR variability in the high-risk patients. The control group was selected to match with group I with regard to gender, age, infarct site, ejection fraction and treatment with ,f3blockers during and after hospitalization. Complete correspondencewas required with re-
THE AMERICAN JOURNAL OF CARDIOLOGY
NOVEMBER 1, 1990
spect to gender (15 male pairs, 5 female pairs), infarct site (8 inferior with Q waves,7 anterior with Q waves, 3 anterior and non-Q, 2 inferior and non-Q) and @blocker treatment (3 pairs of patients). An absolute correspondencein respect to the numerical values of the 2 remaining parameters was not always possible and patients with near values were selected. Statistically (paired t test), there were no differences between the groups regarding age (64 f 6 years [mean f standard deviation] for both groups) and ejection fraction (49 f 13 for group I and 49 f 14% for group II). Holter recording technique: Two-channel 24-hour electrocardiographic recordings (modified lead III and CM5) were obtained using a Tracker recorder (Reynolds Medical Ltd., United Kingdom). In all patients, the recordings were made at a median of 7 days (range 5 to 9) after hospital admission. There was no significant difference between the groups regarding the interval between admission and recording time. A commercially available system for long-term electrocardiographic analysis (Pathfinder III, Mk 2, Reynolds Medical Ltd.) was used to obtain the sequenceof durations of intervals betweenadjacent QRS complexesof normal “supraventricular” morphology for each patient. The exact time of each QRS complex was also derived from the recordings. Measurement of heart rate variability: For each patient, HR variability was measuredin different portions of the 24-hour time interval, with the beginning and duration of the analyzed portion varied. The onset of the analyzed portions varied by 20 minutes (72 values ranging from 0O:OOto 23:40) and the duration of the portions analyzed also varied by 20 minutes (72 values ranging from 20 minutes to 24 hours). In total, 5,184, including 72 identical 24-hour intervals (5,113 different portions of each record) were analyzed. Where the analyzed interval involved the physical end of the recording, the 24-hour electrocardiogram was recorded as a continuous loop-i.e., the end of the recording was extended with the properly timed beginning of the recording from the previous day.
1250
For each time interval analyzed, the frequency distribution of durations of normal-to-normal RR intervals was constructed. HR variability was expressedas the baseline width of this distribution curve measured by the method of minimum square difference interpolation.3 For each interval analyzed, the mean duration of RR intervals was also computed. The more usual method expressingHR variability by the standard deviation of RR interval durations’ was not used becausethe computer analysisof the recording system that we used is known to be affected by recording noise and misrecognition artifact.3 The method of triangular interpolation of the frequency distribution of RR interval duration that we usedis not affectedby low levels of noise and artifact.3*5 Normalized heart rate variability: Previous studies2 have shown that HR and HR variability constitute independentprognostic factors in patients after myocardial infarction. However, these parametersare likely to be correlated. A pathologic processthat shortens the RR intervals is also likely to reduce the absolutevalue of the differences between individual RR intervals, To overcome the interdependence between HR and HR variability, we used the concept of normalized HR variability, which excludes the basic mathematic influence of the mean HR. For each analyzed portion of the recording, the normalized HR variability was obtained by dividing its nonnormalized value by the mean duration of RR intervals within the same portion of the record. Statistical analysis: Paired t testswere used to compare the values found in the correspondingportions of the electrocardiographic recordings obtained in groups I and II. The tests were performed for values of HR. HR variability and normalized HR variability in all analyzed portions of the recordings and the study involved more than 15,000individual t tests.However, the values measured in the individually analyzed portions of the electrocardiographic recordings were not independent (mainly because of frequent overlaps of the portions). Therefore, the problems resulting from the multiplicity of statistical tests were reduced. HRV (ms)
RR (ms) 280
1125
240 200
1000
160 875 120
0
2
4
6
8
10
12
14
16
18
20
22
Time of the day
0
2
4
6
8
10
12
14
16
18
20
22
Time of the day
FIGURE 1. Mean cycle length and short-term heart rate variability measured in separate 40-minute intervals In group I (Positive) and group II (Negative) electrocardiograms. Leff, mean ms duration of normal-to-normal intervals (RR). Rig/H, heart rate variability as the baseline width (in ms) of the sample density distributions of RR interval durations. HRV = mean heart rate values (k standard errors).
1050
THE AMERICAN JOURNAL OF CARDIOLOGY VOLUME 66
RESULTS Circadian
rhythm
of heart
length and short-term HR variability were measured in separate 40-minute portions of 24-hour recordings. In group II recordings, the physiologic diurnal variation of the mean cycle length was preserved, with the slowest HR during the night and the fastest during the afternoon. In contrast, the mean cycle length of the
rate and its variability:
Figure 1 shows circadian variations of mean cycle length and short-term HR variability found in both groups of patients. Mean values f standard errors are shown for both groups; in each patient, the mean cycle
FIGURE 2. Probability levels of the significance at which heart rate variability was reduced in group I (paired I test) compared with that in group II for all analyzed intervals. The vertical axis of the plot corresponds to the beginning of the analyzed interval (in hours, starting at midnight); the horilontal axis corresponds to the duration of the interval (in hours). For each analyzed Interval, the graphic symbols express the significance level between both groups: spaces = no significant difference; dots = p
