Clin. Cardiol. 15 11-10-11-13 (1992)
Accuracy and Analysis of Ambulatory Blood Pressure Monitoring Data WILLIAM B.WHITE, M.D.
Section of Hypertension and Vascular Diseases, University of Connecticut School of Medicine, Farmington, Connecticut,USA
Summary: Two devices used to record blood pressure, the ambulatory blood pressure monitoring recorder and standard stethoscope and mercury column, were tested for accuracy against the direct intra-arterial blood pressure of patients at rest and during exercise. Recorders were found to be as accurate as mercury column measurement in patients at rest. A number of assessmenttechniques of ambulatory data are reviewed, including: calculation of mean or median pressures, assessment of blood pressure load, and integrationof the area under the blood pressure curve over time. These have been applied during the daytime and nighttime hours. Blood pressure load and area under the blood pressure curve, using different threshold criteria for nighttime and daytime, are recommended because of their potentially closer relation to target-organ disease of hypertension than are office blood pressure readings. Key words: ambulatory monitoring, area under curve (AUC), blood pressure load, left ventricular hypertrophy, blood pressure profile
Introduction While noninvasive ambulatory blood pressure monitoring (ABPM) has been shown to be of great value in the diagnosis of mild and borderline hypertension,' it is not used widely in the management of hypertensionin the clinical setting. Office measurement is still the standard, in
Address for reprints: William B. White,M.D. Section of Hypertension and Vascular Diseases University of Connecticut School of Medicine Farmington, CT 06030, USA
part because questions remain as to the accuracy and validity of ABPM recordings. However, ABPM has been used to verify standard office or clinic measurements in patients suspected of having office hypertension.2
Accuracy of ABPM Devices A number of monitors are now available for clinical or research use. Most recorders measure blood pressure by means of auscultation of Korotkoff sounds via a microphone sewn into the blood pressure cuff or taped over the brachial artery. Other devices measure blood pressure by the oscillometric method, which detects pressure oscillations transmitted from the brachial artery to the cuff as it is deflated. We evaluated four ambulatory blood pressure recorders in 48 hypertensive patients at rest and during exercise in a hernodynamiclab~ratory,~ where well-controlled, intra-arterial blood pressure measurements were made simultaneously. In addition, an experienced clinician using a standard mercury column measured blood pressures in ten hypertensive patients undergoing simultaneous intra-arterial measurement. The differences between the intra-merially determined values and those obtained by each of four ABPM devices were compared to the differences between the clinician-measuredblood pressures and those measured intra-arterially. Measurements were made at rest in a seated position and during bicycle exercise at an energy level of 100 W, the approximate equivalent of cycling up a 20-degree grade 5 to 7 miles an hour. During this exercise, the cuffed arm was maintained in a motionless state by keeping it at rest on pillows placed on a table. Figure 1 shows the differences between the intra-arteria1 blood pressures and the values obtained by the clinicians, and by the ABPM recorders at rest. As the figure illustrates, clinician measurement differs by 4 mmHg from both the systolic and diastolic pressures obtained by direct intra-arterial measurement. These two measurements involved contralateral arms, which could have accounted for some variability. In any event, this difference between di-
W.B. White: Accuracy of ABPM data
RG.1 Blood pressure differences at rest. SBP; DBP; *p< 0.01 vs clinicians; **p140/90 mmHg and a minimum of 50% sleep readings >120/80 mmHg. These data strongly indicate that the cut-off value for defining abnormal blood pressure during sleep should be 120/80M g , not 14/90 mmHg. We therefore proposed4 that when using ABPM, the diagnosis of sustained hypertension should be based on 240% of awake readings >140/90 mmHg, andor 240% of sleep readings 2 120/80 mmHg. The percentage of total ABPM readings that are abnormal is the ‘blood pressure load.’ In another study5 we compared casual measurement in the doctor’s office, 24-h ABPM measurements and blood
IT-12
Clin. Cardiol. Vol. 15 (Suppl. 11),October 1992
pressure load with the frequency of echocardiographicallyderived left ventricular hypertrophy in middle-aged, never previously treated, mild-to-moderatehypertensives. Figure 3 presents these systolic blood pressure determinants versus the frequency of increased left ventricular mass index. We used 140 mmHg as the cut-off for casual pressures, 135 mmHg as a mean for 24-h pressures, and 50% abnormal as a cut-off value for blood pressure load. As can be seen, a casual blood pressure greater than 140 mmHg indicated a 50%chance of left ventricular enlargement. A 24-h blood pressure mean greater than 135 mmHg raised the incidence to about 70%, suggesting that the ABPM mean value was a better discriminator than the clinic pressure measurement as to who did and who did not have cardiac indices of hypertensive disease. When the blood pressure load was used as an indicator, it proved to be superior as a predictor of possible end-organ disease than office blood pressure measurements. We found similar results for the diastolic blood pressures.
as for researchers performing diagnostic or therapeutic evaluations in clinical trials. A variety of data-handling methods are available for consideration. One technique is to analyze the mean or extreme values. The profile and shape of the blood pressure curve is becoming of increasing interest, especially when comparing different patient populations. In drug therapy studies, the drug action may be assessed by observing the peak and trough effects. These data can then be smoothed with mathematical formulas or remain unsmoothed as raw data. Different time blocks of the 24-hour period may be evaluated, such as 2-, 4-, or 6-hour averages. Finally, the blood pressure load can be observed in two ways: one is by treating values such as 140/90 mmHg or 120/80 mmHg as thresholds, above which measurements are considered abnormal; or one can focus on an integrated value (AUC) under the blood pressure curve. Criteria for Data Analysis
Methods of Data Analysis There are a number of approaches to the analysis of ABPM data. In the 24-h monitoring period some investigators measure blood pressure as many as 8 times an hour, others for as few as four times an hour. In any event, ABPM may produce between 75 and 200 blood pressure and heart rate measurements over the course of a single day. In addition, there is the normal diurnal variation associated with activity and sleep which needs to be considered when assessing ABPM data. These data present a heavy statistical task for physicians in clinical practice, as well
Certain criteria should be applied to each method of data analysis. First, the method should have physiologic and pharmacologic application. Second, statistical coherence needs to be maintained. In ABPM, raw data from some 75 to 200 readings are obtained, and these must be rationalized into a number which neither covers up or oversmooths, which can be tested statistically, and which can be shown to be reproducible from day to day and month to month. Finally, this analytical record has to be related to the hypertensive disease process so that conventional clinical blood pressure measurement can be improved. An ambulatory blood pressure profile in a drug study offers an example of one method of analysis. Figure 4 demonstrates a placebo-subtracted, 24-h diastolic blood
100 1
3
a
0 -2
5
-4
0 .-
Casual >140 mmHg
24-h >135 mmHg
24-h 435 mmHg
Load ~50%
Load
40%
FIG. 3 Frequencies of an abnormal cardiac mass for different levels of systolic blood pressure. Adapted from White et al. “Assessment of daily blood pressureload as a determinant of cardiacfuntion inpatients with mild to moderate hypertension” Am Heart J, Vol. 118, 782-785 1989. (Reproduced with permission of Am Heart J, Ref. 6.)
.-3 -6 ‘0 .-c -8 g) -10
s -12
6
0
2
4
6
8 10 12 14 16 18 20 22 24 Time since dosing (h)
FIG.4 Unsmoothed, placebo-subtracted, 24-h diastolic blood pressure curve in a patient demonstratingeffects of a long-acting antihypertensive agent. (Reproduced with permission of J Hyper, Ref. 5.)
11- 13
W.B. White: Accuracy of ABPM data
m
&
0000
0500
1200 1800 Time of day
2400
&’
0000
0500
1200 1800 Time of day
2400
FIG.5 Evaluation of antihypertensive therapy using the area under the systolic blood pressure curve with cut-off values of 140 mmHg during waking hours and 120 mmHg during sleep for (a) placebo and (b) antihypertensive therapy. (Reproduced with permission of J Hyper; Ref. 5 . )
pressure curve in a patient taking a long-acting antihypertensive agent. The net change in blood pressure elicited by the drug is compared with placebo and plotted against time since the dose. The blood pressure on placebo is subtracted from the blood pressure on the drug by hourly averages, allowing observation of the effect of the drug hour-by-hour throughout the 24-h period. When evaluating unsmoothed raw data from a single individual, the inherent variability of blood pressure associated with activity and/or psychosocial factors can obscure the point at which peak drug effect occurs. Fourier transformation, is an appropriate mathematical procedure for smoothing periodicity. Based on complex regression analyses, Fourier transformation clarifies the peak and/or trough effects of antihypertensive drugs in individual ABPM studies. Another concept, already discussed briefly, is blood pressure load. Blood pressure load may be defined in a variety of ways. Essentially, it is a percentage of values above or below an arbitrary threshold. We have found this method of evaluating high blood pressure over time to be quite useful in clinical practice as well as in studies of mild, untreated hypertensives. In addition, the concept of blood pressure load involves an elevated chronic pressure overload in an individual or a group. One takes the area under the curve for the total 24-h period and arrives at one number-a percentage-a convenient way of reducing data to a single value. One technique is to analyze the data under the blood pressure curve in terms of threshold values. Figure 5 is the curve of systolic blood pressures, with threshold values of 140 mmHg during daytime hours, and 120 mmHg during sleep and daytime hours, comparing placebo (a) with an antihypertensive agent (b) over time. This shows clearly how often, and by how much blood pressure is elevated, expressing this as blood pressure load. Data removal or smoothing is unnecessary, and the efficacy of the drug over the entire 24-h period is instantly visualized.
Conclusions In the analysis of ABPM, many approaches can be utilized in order to obtain the best overall perspective regarding an individual’s blood pressure profile. Different methods complement each other. Multiple smoothing techniques have been recommended in the literature over the last few years, and Fourier analysis has been the most widely used. This statistical approach has proven useful in some assessments of drug effects, but does not appear to offer important advantages to clinicians in diagnosing hypertension or evaluating drug efficacy. Blood pressure load, whether as a percentage of all ABPM pressures or of those in an area under the curve, is a method which has been correlated with cardiac target organ indices. It is a simple concept which practicing physicians can easily utilize when they receive a blood pressure profile from a specialist referral center.
References 1. Kleinert HD, Harshfield GA, Pickering TG, Blank S, Laragh
2.
3.
4.
5. 6.
J: What is the value of home blood pressure measurement in patients with mild hypertension? Hypertension 6 , 574-578 (1 984) Pickering TG, James GD, Boddie C, Harschfield GA, Blank S, Laragh J: How common is white coat hypertension? J A m Med ASSOC259,225-228 (1988) White WB, Lund-Johansen P, Omvik P: Assessment of four ambulatory blood pressure monitors and measurements by clinicians versus intra-arterial blood pressure at rest and during exercise. A m J Curdiol65,60-66 (1990) White WB, Morganroth J: Usefulness of ambulatory monitoring of blood pressure in assessing antihypertensive theory. Am J Curdiol63,94-98 (1989) White WB: Analysis of ambulatory blood pressure data in antihypertensive drug trials. J Hyperfens 9 (suppl I), S27-32 (1991) White WB:Assessment of daily blood pressure load as a determinant of cardiac function in patients with mild to moderate hypertension. Am Heart J 118,782-785 (1989)
Accuracy and Analysis of Ambulatory Blood Pressure Monitoring Data WILLIAM B.WHITE, M.D.
Section of Hypertension and Vascular Diseases, University of Connecticut School of Medicine, Farmington, Connecticut,USA
Summary: Two devices used to record blood pressure, the ambulatory blood pressure monitoring recorder and standard stethoscope and mercury column, were tested for accuracy against the direct intra-arterial blood pressure of patients at rest and during exercise. Recorders were found to be as accurate as mercury column measurement in patients at rest. A number of assessmenttechniques of ambulatory data are reviewed, including: calculation of mean or median pressures, assessment of blood pressure load, and integrationof the area under the blood pressure curve over time. These have been applied during the daytime and nighttime hours. Blood pressure load and area under the blood pressure curve, using different threshold criteria for nighttime and daytime, are recommended because of their potentially closer relation to target-organ disease of hypertension than are office blood pressure readings. Key words: ambulatory monitoring, area under curve (AUC), blood pressure load, left ventricular hypertrophy, blood pressure profile
Introduction While noninvasive ambulatory blood pressure monitoring (ABPM) has been shown to be of great value in the diagnosis of mild and borderline hypertension,' it is not used widely in the management of hypertensionin the clinical setting. Office measurement is still the standard, in
Address for reprints: William B. White,M.D. Section of Hypertension and Vascular Diseases University of Connecticut School of Medicine Farmington, CT 06030, USA
part because questions remain as to the accuracy and validity of ABPM recordings. However, ABPM has been used to verify standard office or clinic measurements in patients suspected of having office hypertension.2
Accuracy of ABPM Devices A number of monitors are now available for clinical or research use. Most recorders measure blood pressure by means of auscultation of Korotkoff sounds via a microphone sewn into the blood pressure cuff or taped over the brachial artery. Other devices measure blood pressure by the oscillometric method, which detects pressure oscillations transmitted from the brachial artery to the cuff as it is deflated. We evaluated four ambulatory blood pressure recorders in 48 hypertensive patients at rest and during exercise in a hernodynamiclab~ratory,~ where well-controlled, intra-arterial blood pressure measurements were made simultaneously. In addition, an experienced clinician using a standard mercury column measured blood pressures in ten hypertensive patients undergoing simultaneous intra-arterial measurement. The differences between the intra-merially determined values and those obtained by each of four ABPM devices were compared to the differences between the clinician-measuredblood pressures and those measured intra-arterially. Measurements were made at rest in a seated position and during bicycle exercise at an energy level of 100 W, the approximate equivalent of cycling up a 20-degree grade 5 to 7 miles an hour. During this exercise, the cuffed arm was maintained in a motionless state by keeping it at rest on pillows placed on a table. Figure 1 shows the differences between the intra-arteria1 blood pressures and the values obtained by the clinicians, and by the ABPM recorders at rest. As the figure illustrates, clinician measurement differs by 4 mmHg from both the systolic and diastolic pressures obtained by direct intra-arterial measurement. These two measurements involved contralateral arms, which could have accounted for some variability. In any event, this difference between di-
W.B. White: Accuracy of ABPM data
RG.1 Blood pressure differences at rest. SBP; DBP; *p< 0.01 vs clinicians; **p140/90 mmHg and a minimum of 50% sleep readings >120/80 mmHg. These data strongly indicate that the cut-off value for defining abnormal blood pressure during sleep should be 120/80M g , not 14/90 mmHg. We therefore proposed4 that when using ABPM, the diagnosis of sustained hypertension should be based on 240% of awake readings >140/90 mmHg, andor 240% of sleep readings 2 120/80 mmHg. The percentage of total ABPM readings that are abnormal is the ‘blood pressure load.’ In another study5 we compared casual measurement in the doctor’s office, 24-h ABPM measurements and blood
IT-12
Clin. Cardiol. Vol. 15 (Suppl. 11),October 1992
pressure load with the frequency of echocardiographicallyderived left ventricular hypertrophy in middle-aged, never previously treated, mild-to-moderatehypertensives. Figure 3 presents these systolic blood pressure determinants versus the frequency of increased left ventricular mass index. We used 140 mmHg as the cut-off for casual pressures, 135 mmHg as a mean for 24-h pressures, and 50% abnormal as a cut-off value for blood pressure load. As can be seen, a casual blood pressure greater than 140 mmHg indicated a 50%chance of left ventricular enlargement. A 24-h blood pressure mean greater than 135 mmHg raised the incidence to about 70%, suggesting that the ABPM mean value was a better discriminator than the clinic pressure measurement as to who did and who did not have cardiac indices of hypertensive disease. When the blood pressure load was used as an indicator, it proved to be superior as a predictor of possible end-organ disease than office blood pressure measurements. We found similar results for the diastolic blood pressures.
as for researchers performing diagnostic or therapeutic evaluations in clinical trials. A variety of data-handling methods are available for consideration. One technique is to analyze the mean or extreme values. The profile and shape of the blood pressure curve is becoming of increasing interest, especially when comparing different patient populations. In drug therapy studies, the drug action may be assessed by observing the peak and trough effects. These data can then be smoothed with mathematical formulas or remain unsmoothed as raw data. Different time blocks of the 24-hour period may be evaluated, such as 2-, 4-, or 6-hour averages. Finally, the blood pressure load can be observed in two ways: one is by treating values such as 140/90 mmHg or 120/80 mmHg as thresholds, above which measurements are considered abnormal; or one can focus on an integrated value (AUC) under the blood pressure curve. Criteria for Data Analysis
Methods of Data Analysis There are a number of approaches to the analysis of ABPM data. In the 24-h monitoring period some investigators measure blood pressure as many as 8 times an hour, others for as few as four times an hour. In any event, ABPM may produce between 75 and 200 blood pressure and heart rate measurements over the course of a single day. In addition, there is the normal diurnal variation associated with activity and sleep which needs to be considered when assessing ABPM data. These data present a heavy statistical task for physicians in clinical practice, as well
Certain criteria should be applied to each method of data analysis. First, the method should have physiologic and pharmacologic application. Second, statistical coherence needs to be maintained. In ABPM, raw data from some 75 to 200 readings are obtained, and these must be rationalized into a number which neither covers up or oversmooths, which can be tested statistically, and which can be shown to be reproducible from day to day and month to month. Finally, this analytical record has to be related to the hypertensive disease process so that conventional clinical blood pressure measurement can be improved. An ambulatory blood pressure profile in a drug study offers an example of one method of analysis. Figure 4 demonstrates a placebo-subtracted, 24-h diastolic blood
100 1
3
a
0 -2
5
-4
0 .-
Casual >140 mmHg
24-h >135 mmHg
24-h 435 mmHg
Load ~50%
Load
40%
FIG. 3 Frequencies of an abnormal cardiac mass for different levels of systolic blood pressure. Adapted from White et al. “Assessment of daily blood pressureload as a determinant of cardiacfuntion inpatients with mild to moderate hypertension” Am Heart J, Vol. 118, 782-785 1989. (Reproduced with permission of Am Heart J, Ref. 6.)
.-3 -6 ‘0 .-c -8 g) -10
s -12
6
0
2
4
6
8 10 12 14 16 18 20 22 24 Time since dosing (h)
FIG.4 Unsmoothed, placebo-subtracted, 24-h diastolic blood pressure curve in a patient demonstratingeffects of a long-acting antihypertensive agent. (Reproduced with permission of J Hyper, Ref. 5.)
11- 13
W.B. White: Accuracy of ABPM data
m
&
0000
0500
1200 1800 Time of day
2400
&’
0000
0500
1200 1800 Time of day
2400
FIG.5 Evaluation of antihypertensive therapy using the area under the systolic blood pressure curve with cut-off values of 140 mmHg during waking hours and 120 mmHg during sleep for (a) placebo and (b) antihypertensive therapy. (Reproduced with permission of J Hyper; Ref. 5 . )
pressure curve in a patient taking a long-acting antihypertensive agent. The net change in blood pressure elicited by the drug is compared with placebo and plotted against time since the dose. The blood pressure on placebo is subtracted from the blood pressure on the drug by hourly averages, allowing observation of the effect of the drug hour-by-hour throughout the 24-h period. When evaluating unsmoothed raw data from a single individual, the inherent variability of blood pressure associated with activity and/or psychosocial factors can obscure the point at which peak drug effect occurs. Fourier transformation, is an appropriate mathematical procedure for smoothing periodicity. Based on complex regression analyses, Fourier transformation clarifies the peak and/or trough effects of antihypertensive drugs in individual ABPM studies. Another concept, already discussed briefly, is blood pressure load. Blood pressure load may be defined in a variety of ways. Essentially, it is a percentage of values above or below an arbitrary threshold. We have found this method of evaluating high blood pressure over time to be quite useful in clinical practice as well as in studies of mild, untreated hypertensives. In addition, the concept of blood pressure load involves an elevated chronic pressure overload in an individual or a group. One takes the area under the curve for the total 24-h period and arrives at one number-a percentage-a convenient way of reducing data to a single value. One technique is to analyze the data under the blood pressure curve in terms of threshold values. Figure 5 is the curve of systolic blood pressures, with threshold values of 140 mmHg during daytime hours, and 120 mmHg during sleep and daytime hours, comparing placebo (a) with an antihypertensive agent (b) over time. This shows clearly how often, and by how much blood pressure is elevated, expressing this as blood pressure load. Data removal or smoothing is unnecessary, and the efficacy of the drug over the entire 24-h period is instantly visualized.
Conclusions In the analysis of ABPM, many approaches can be utilized in order to obtain the best overall perspective regarding an individual’s blood pressure profile. Different methods complement each other. Multiple smoothing techniques have been recommended in the literature over the last few years, and Fourier analysis has been the most widely used. This statistical approach has proven useful in some assessments of drug effects, but does not appear to offer important advantages to clinicians in diagnosing hypertension or evaluating drug efficacy. Blood pressure load, whether as a percentage of all ABPM pressures or of those in an area under the curve, is a method which has been correlated with cardiac target organ indices. It is a simple concept which practicing physicians can easily utilize when they receive a blood pressure profile from a specialist referral center.
References 1. Kleinert HD, Harshfield GA, Pickering TG, Blank S, Laragh
2.
3.
4.
5. 6.
J: What is the value of home blood pressure measurement in patients with mild hypertension? Hypertension 6 , 574-578 (1 984) Pickering TG, James GD, Boddie C, Harschfield GA, Blank S, Laragh J: How common is white coat hypertension? J A m Med ASSOC259,225-228 (1988) White WB, Lund-Johansen P, Omvik P: Assessment of four ambulatory blood pressure monitors and measurements by clinicians versus intra-arterial blood pressure at rest and during exercise. A m J Curdiol65,60-66 (1990) White WB, Morganroth J: Usefulness of ambulatory monitoring of blood pressure in assessing antihypertensive theory. Am J Curdiol63,94-98 (1989) White WB: Analysis of ambulatory blood pressure data in antihypertensive drug trials. J Hyperfens 9 (suppl I), S27-32 (1991) White WB:Assessment of daily blood pressure load as a determinant of cardiac function in patients with mild to moderate hypertension. Am Heart J 118,782-785 (1989)
