This Article Abstract Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Gillman, M. W. Articles by Cook, N. R. Search for Related Content PubMed PubMed Citation Articles by Gillman, M. W. Articles by Cook, N. R.

(Circulation. 1995;92:1049-1057.)
© 1995 American Heart Association, Inc.

Articles

Blood Pressure Measurement in Childhood Epidemiological Studies

Matthew W. Gillman, MD, SM; Nancy R. Cook, ScD

From the Department of Ambulatory Care and Prevention, Harvard Medical School and Harvard Community Health Plan (M.W.G., N.R.C.), and the Division of Preventive Medicine, Department of Medicine, Brigham and Women's Hospital (N.R.C.), Boston, Mass.

	Abstract

Top Abstract Introduction Factors That Affect Blood... Blood Pressure Instruments for... How Many Measurements to... References

 
Background Accurate measurement of blood pressure in childhood epidemiological studies requires standardized conditions, valid instruments, and multiple measurements.

Methods and Results We used published literature and our own data to make recommendations on the reduction of biases caused by various technical factors, to discuss the advantages and disadvantages of selected measurement devices, and to evaluate the optimal number of visits and measurements per visit for accurate estimation of a child's blood pressure level. The conditions under which blood pressure is measured should be standardized. This includes training and certification to minimize observer biases; equipment factors such as use of an appropriate cuff bladder size, subject factors such as minimizing activities before and during the reading, environmental factors such as accounting for the time of day and ambient temperature, and technique factors such as recording both the fourth and fifth Korotkoff sounds. The choice of instrument for measuring blood pressure depends on the goals of the study and the resources available to the investigators.

Conclusions Although relatively economical and easy to use, the standard mercury sphygmomanometer is subject to the bias resulting from knowledge of earlier readings. The random-zero sphygmomanometer overcomes this bias, but it is more expensive and difficult to use and may underestimate blood pressure levels. In contrast to auscultatory devices, automated oscillometric devices are not subject to observer biases. They are gaining wider use and may be especially appropriate for younger children. However, they are expensive, and each model requires validation before use in epidemiological studies. Ambulatory blood pressure monitoring represents a potentially useful technique for future epidemiological studies. Multiple measurements are vital in estimating a child's blood pressure, and the number of visits, days or weeks apart, is at least as important as the number of measurements per visit.

Key Words: blood pressure • epidemiology

	Introduction

Top Abstract Introduction Factors That Affect Blood... Blood Pressure Instruments for... How Many Measurements to... References

 
Among all known childhood predictors of adult blood pressure level, the level of blood pressure in childhood is by far the strongest.^{1 2 3} Thus, the potential to prevent adult hypertension starting in childhood depends on knowledge of the determinants of childhood blood pressure and aspects of its short- and long-term variability.⁴ As a result, accurate determination of blood pressure is of paramount importance in epidemiological studies of blood pressure in children.

The purposes of this article are (1) to review factors that affect blood pressure measurement in children, including equipment, subject, environmental, and observer or technique factors; (2) to discuss the strengths and weaknesses of available instruments with a focus on mercury sphygmomanometers and oscillometric automated recording devices; and (3) using data we obtained on school-age children in two studies, to consider the optimal number of visits and measurements per visit to be used with the different instruments. Throughout, we highlight areas in which measurement issues in children may differ from those in adults.

	Factors That Affect Blood Pressure Measurement

Top Abstract Introduction Factors That Affect Blood... Blood Pressure Instruments for... How Many Measurements to... References

 
Equipment
Cuff Bladder Size
It is well accepted that a cuff with too narrow a bladder will overestimate blood pressure levels in children and adults. Most studies in children also found that too large a cuff will underestimate blood pressure level, although the magnitude of the error may be less with a too-large cuff than with a small cuff.^{5 6 7 8} Although some authors recommended using cuff bladders whose width exceeds two thirds of the length of the upper arm, keying cuff bladder width to arm circumference rather than arm length leads to more valid results.⁹ Most authorities now agree that a cuff bladder width that is 40% of the arm circumference most closely approximates intra-arterial readings.⁵ Prineas and Elkwiry¹⁰ recommended a bladder width that is 38±5% of the arm circumference. In addition, bladder length seems to make a difference more in children than in adults. Whereas a length of at least 80% of the arm circumference seems adequate in adults, at least 90%, if not 100% (ie, overlapping) is necessary to avoid overestimation of blood pressure in children.¹¹

A less well-known cuff size phenomenon is due to the relation of arm circumference to blood pressure level demonstrated in Fig 1. These data demonstrate that the blood pressure differences between cuffs are largely independent of arm circumference.⁶ Thus, at the point where one changes from infant to child cuff or from child to adult cuff, there is a "step" function. This finding has two implications for epidemiological studies. The first is that any association of blood pressure with a variable related to arm circumference, such as body mass or fatness, will be biased toward the null.¹² This bias results from the misclassification of blood pressure around the "step" points, even when recommended cuff bladder sizes are used. For instance, blood pressure measured with the child cuff in children with an arm circumference of 22 cm will be (falsely) higher on average than blood pressure measured with the adult cuff in children with an arm circumference of 23 cm.

View larger version (11K): [in this window] [in a new window]   Figure 1. Plot of mean systolic blood pressures (SBP) measured with infant, child, and adult cuffs in relation to arm circumference. Effect on blood pressure measurement of the use of the guidelines of the American Heart Association (ie, the change to a larger cuff is made when the cuff width to arm circumference ratio falls below 40%) is shown. Reproduced with permission from Whincup et al.6

The second implication of these findings concerns longitudinal studies of blood pressure. In such studies, as a child grows, a progressively larger cuff size is used. When one changes to a larger cuff, measured blood pressure is lower than prior readings. This phenomenon could lead to anomalous inverse relations of change in age or height with blood pressure level.

The current recommendations for cuff bladder size are based on associations of indirectly measured blood pressure with intra-arterial readings. Unless accuracy compared with intra-arterial readings is of paramount concern, epidemiologists studying body size and blood pressure or conducting longitudinal studies should consider using a single cuff size or, if using several cuff sizes, standardizing the results to a single cuff size.⁶

Stethoscope Bell Versus Diaphragm
In adults, it is clear that use of the bell is preferred because Korotkoff sounds can be heard clearly.¹³ In children, who have small arms, a tight seal with the bell may be hard to achieve. In addition, excessive pressure on the arm with the bell, ie, too tight a seal, will cause underestimation of diastolic pressure; Korotkoff sounds may be heard even to 0 mm Hg. Thus, some authors recommended using the diaphragm for small children.¹⁴ Nevertheless, proper use of the bell, with smaller bells for smaller arms and care to achieve a seal with the lightest possible pressure, should achieve the best results.¹⁵

Subject Factors
Activities Before and During the Measurement
If undertaken within 30 minutes before blood pressure measurement, such stimuli as food, alcohol, caffeine, nicotine, and exercise may affect the reading.⁵ Crossing one's legs may raise measured blood pressure, as may talking or performing mental tasks during the measurement.⁵ Unless the goal is to quantify such activities, as in ambulatory blood pressure monitoring, children should avoid eating and exercise for 30 minutes before a reading, and they should rest quietly with legs uncrossed for 5 minutes before and throughout blood pressure measurement.

Arm and body positions are also important for accuracy of readings. The standard position of the arm requires the forearm to be at heart level. If the arm is too high, blood pressure will be underestimated; if the arm is too low, which may occur with the subject simply in the supine position, blood pressure will be overestimated.¹⁶ For infants and very small children, the sitting position is impractical, and one should measure blood pressure with patients in the supine position. For children older than about 3 years, the sitting position is standard.¹⁷

Age and height
It has been well recognized for decades that blood pressure rises with age throughout childhood in developed countries. The Second Task Force on Blood Pressure Control in Children consolidated data from >70 000 children to arrive at age-specific means for boys and girls.¹⁸ More recently, however, Rosner et al¹⁹ have modified these data to take account of the additional predictive information provided by height. With height controlled for, age is still a substantial, but weaker, predictor of childhood blood pressure level. Kahn and colleagues^{20 21} suggested that the distance from the heart to the vertex of the head is more important than overall height. In epidemiological studies of blood pressure in children, investigators should obtain data on age and height in addition to measures of body mass and fatness.

Environmental Factors
Time of Day
It is clear from studies in children who have used ambulatory blood pressure monitoring that blood pressure is lowest during sleep.^{22 23} Some data suggest two peaks of blood pressure during the day, in the late morning and in midafternoon.²⁴ These patterns are similar to those in adults but may be altered by activities such as school attendance or exercise. It is prudent to record the time of day when blood pressure is measured to control for it or to examine its effects in data analysis.

Surroundings
The "white-coat" effect exists in adults.²⁵ That is, blood pressure of many persons is substantially higher in the office than at home, especially in the presence of a physician. Whether this phenomenon occurs in children is not clear.^{24 26} As discussed above, however, quiet surroundings and minimized extraneous stimuli are important in estimating basal blood pressure level.

Season and Temperature
Few studies have examined the effect of season and ambient temperature on blood pressure in children. Prineas et al¹⁵ and Jenner et al²⁷ agreed that systolic pressure rises substantially with decreasing ambient temperature, but their findings on diastolic pressure were divergent. As with time of day, recording of ambient temperature is necessary should an investigator wish to control for this factor in analyses of data.

Observer or Technique Factors
Standardization and Training
Observer bias has been identified as an important factor in many large epidemiological studies involving blood pressure measurement.^{28 29} Although the details for training and certification are beyond the scope of this article, it is important to note that standardization, restandardization (eg, at 6-month intervals), and certification of observers are vital to achieving valid results in any study of blood pressure. In this section, we address two specific issues, common observer biases and measurement of diastolic pressure.

Observer Biases
Digit preference is the tendency for observers to record round numbers as the terminal digit of any reading. The most common terminal digit is zero in studies that use the standard mercury sphygmomanometer.²⁸ The effect of digit preference can be seen most readily in studies involving classification into categories. For example, in a study in which subjects are to be included if diastolic pressure is at least 90 mm Hg, rounding up readings of 88 mm Hg to 90 mm Hg will misclassify some individuals into the group to be included for study.

More important for studies in children, which usually do not involve categorization, is the bias resulting from knowledge of earlier readings. When taking multiple readings over several minutes, observers tend to record readings that resemble the previous readings. This bias tends to falsely reduce intraperson variability and therefore may actually obscure associations of blood pressure with other variables.²⁸

Although standardized training may minimize these biases, use of devices other than the standard mercury sphygmomanometer may also be helpful (see below).

Measurement of Diastolic Pressure
For at least 50 years, there has been a controversy over whether muffling (Korotkoff phase 4 [K4]) or the disappearance of sounds (K5) should be used for measurement of diastolic pressure. Although apparently settled for adolescents and adults, in whom K5 is generally accepted, the controversy continues for children under the age of 13 years. Some favor K4 in this age group because it more closely approximates intra-arterial pressure and because K5 may be heard as low as 0 mm Hg, especially in the presence of undue pressure on the stethoscope head.¹⁴ Indeed, the Second Task Force on Blood Pressure Control in Children report advocates using K4.¹⁸ Others favor K5 because of the difficulty of hearing the muffling of sounds corresponding to K4 and the comparability of diastolic pressure before and after one's 13th birthday.³⁰ In fact, some authors who participated in the Second Task Force now recommended use of K5.¹⁹ In any case, Sinaiko et al³¹ showed that in 50% of cases, K4 and K5 are not equal, and in approximately 15% of subjects, they differ by more than 10 mm Hg. Although current evidence seems to favor the use of K5, it is prudent to record both K4 and K5 in epidemiological studies.

	Blood Pressure Instruments for Children

Top Abstract Introduction Factors That Affect Blood... Blood Pressure Instruments for... How Many Measurements to... References

 
This section focuses on three instruments: the standard mercury sphygmomanometer, which has been the standard for epidemiological studies for 40 to 50 years; the random-zero sphygmomanometer, which has been in use for about 20 years; and automated oscillometric devices, specifically the Dinamap devices (Critikon, Inc), which have become more prominent in the last decade. We also briefly address the use of ambulatory blood pressure monitors but do not comment on Doppler systems, which are very useful in infants, especially premature infants.

Standard Mercury Sphygmomanometer
Strengths
The standard mercury sphygmomanometer has been in widespread use in epidemiological studies for several decades. Much of what we know about the relations of blood pressure with stroke, coronary heart disease, and other outcomes is built on studies, such as the Framingham Study, that use this device.³² Because standard mercury readings are the main basis for blood pressure–disease associations, one might go so far as to say that they are the epidemiological gold standard for measurement. By extension, studies relating childhood with adulthood blood pressure or examining blood pressure within childhood might benefit from the use of this instrument. Because this is the instrument generally used in doctors' offices, findings from epidemiological studies with standard mercury readings are directly applicable to clinical practice. Furthermore, the standard mercury sphygmomanometer is relatively inexpensive, easy to transport, and easy to maintain, and compared with the random-zero instrument, study personnel can be trained easily to use it.

Weaknesses
The standard mercury sphygmomanometer is subject to digit preference and observer bias resulting from knowledge of previous readings. The latter, especially, may severely compromise results in childhood studies because of the importance of accurately characterizing intraperson variability in children.^{33 34} Although easier to use than a random-zero sphygmomanometer, the standard mercury device requires detailed training, standardization, and certification. Additionally, because of the difficulty of hearing Korotkoff sounds in young children, using this device is problematic in this age group.³⁵

Random-Zero Sphygmomanometer
The random-zero sphygmomanometer was devised in the late 1960's and early 1970's to overcome the observer biases of the standard mercury device.³⁶ The device incorporates a reservoir that contains an amount of mercury determined by, but unknown to, the observer before each reading. This amount is visible at the end of the reading; the observer then subtracts this "random-zero" quantity from the observed reading to obtain a corrected reading. Thus, the device blinds the observer to the actual blood pressure level until after the measurement is complete.

Strengths
The random-zero sphygmomanometer has been used in many epidemiological studies in both children and adults during the past two decades.^{28 29} Some now consider it the gold standard for epidemiological blood pressure measurement.

It is clear that the random-zero sphygmomanometer minimizes but does not eliminate digit preference. Furthermore, there is evidence that it reduces or eliminates the bias resulting from knowledge of earlier readings.²⁸ The Table gives our data from 162 children 8 to 12 years of age from East Boston, Mass. We measured each subject's blood pressure at four visits 1 week apart. At each visit, we measured blood pressure three times at 1-minute intervals using a random-zero sphygmomanometer and then three times using a standard mercury sphygmomanometer. The Table shows that for both systolic and diastolic pressure, the within-visit variance is almost twice as high with the random-zero sphygmomanometer compared with the standard mercury device. Although alternative explanations, such as order effect, may explain some of this finding, the likelihood is that much of the difference can be explained by a false reduction in within-person variability with the standard mercury device owing to knowledge of previous readings.

View this table: [in this window] [in a new window]   Table 1. Childhood BP Variance Estimates From Three Measurement Devices

Weaknesses
Some of the weaknesses of the random-zero sphygmomanometer are practical in nature. It is approximately 10 times as expensive as a standard mercury device. Because of its complexity, it is more difficult to maintain and requires more intensive training. Because it is bulky, it is relatively difficult to transport. And like the standard mercury sphygmomanometer, it is difficult to use with small children.

Recently, several authors questioned the accuracy of the random-zero sphygmomanometer.^{37 38} Although the final assessment is not available, a few points are clear. The random-zero instrument underestimates both systolic and diastolic pressures compared with the standard mercury instrument. In adults, the underestimation appears to be on the order of 1 to 3 mm Hg for both systolic and diastolic pressures, although one study found a 7.5 mm Hg difference in the diastolic reading.³⁹ Fewer data are available in children, but the underestimation appears to be of the same magnitude.⁴⁰ In adults, the bias seems to be accentuated at higher blood pressure levels, although one study of children showed more differences between the two instruments at the lowest levels.⁴¹ The underestimation with the random-zero device has been ascribed to observer, technique, and instrument factors^{42 43 44} ; the true cause is unknown.

This weakness of the random-zero sphygmomanometer is most serious in blood pressure studies designed to apply to directly to clinical practice, such as clinical trials with disease outcomes. In such cases, cutoff points used in the research setting may not apply to those in the practice settings. It is less serious in the types of studies usually seen in children but still is an important source of bias.

Automated Oscillometric Devices
Several automated devices use the oscillometric principle of measuring blood pressure. The most widely used of these devices are those manufactured under the name Dinamap. Dinamap has developed several models; the measurement algorithm is updated with each new model. A Dinamap device measures blood pressure by first inflating the cuff rapidly above systolic pressure and then deflating the cuff in a stepwise fashion. When blood starts flowing through the artery, oscillations are detected by the surrounding cuff. The point of maximal oscillations corresponds to mean arterial pressure. Systolic and diastolic pressures are calculated as functions of the mean and are calibrated to be equivalent to corresponding intra-aortic pressures.⁴⁵ The digital readout of the Dinamap face plate displays systolic and diastolic pressures, heart rate, and either temperature or mean arterial pressure.

Strengths
The device requires relatively little training, involving attention only to subject and environmental factors, application of the appropriate cuff, and correct use of the machine controls. Because there is no observer, bias caused by digit preference or knowledge of previous readings is eliminated. It is relatively easy to use with small children because there is no need for auscultation. In some studies, the device can be used in an interactive fashion that is fun for children and helpful to investigators. For instance, in a recent intervention trial that we conducted in the fifth grade of a Boston public elementary school, an automated device was connected through an interface to an Apple IIE computer. The computer screen instructed the subject to enter his or her identification number; the program then asked for a double check of the subject's name and instructed the Dinamap device to take the subject's blood pressure and heart rate four times at 1-minute intervals. All blood pressure data were automatically recorded on a floppy disk; investigators and subjects were blinded to these data until the end of the study.⁴⁶

Another strength of the Dinamap is its strong correlation with intra-arterial readings. Fig 2 shows data from Park and Menard³⁵ that demonstrate the very close association between the two methods.

View larger version (19K): [in this window] [in a new window]   Figure 2. Scatterplot showing the relation between systolic pressures measured by a Dinamap oscillometric device and radial artery catheter. The broken line is calculated regression line; solid line is the line of identity. Reproduced with permission from Park and Menard.35

Weaknesses
First, a Dinamap device is very expensive—perhaps 50 to 100 times the cost of a standard mercury sphygmomanometer—but this expense must be weighed against the possible personnel savings of a system that requires minimal training and no direct observation of measurements. Second, as opposed to the mercury devices, a Dinamap device has not been used extensively in epidemiological studies. This may be changing, however. Some large studies of children, notably the Child and Adolescent Trial for Cardiovascular Health,⁴⁷ are using a Dinamap device. Third, motion of the child's arm during measurement may cause artifactual readings. If a child moves his or her arm during an auscultatory reading, the observer may repeat the measurement. With an automated device, there is no standard for dealing with such events. Fourth, readings correspond with intra-aortic, not the epidemiological standard of auscultatory, measurements.⁴⁸ This consideration may make it difficult to compare Dinamap data with data obtained from other methods.

Fifth, we and others have found a "first-reading effect" with a Dinamap device.¹⁷ That is, the first of several readings obtained within a few minutes of each other tends to be higher, by about 3 to 5 mm Hg, than subsequent readings. This phenomenon, which may be caused by subject and instrument factors, is of more concern during attempts to estimate blood pressure levels accurately than in examinations of change of blood pressure over time in an individual. Finally, different models of the Dinamap devices have different measurement algorithms and thus measure different quantities. An older model, the 845, tended to estimate systolic pressure very accurately compared with auscultatory methods, but the newer model 1846 tends to overestimate systolic pressure (see Fig 3).⁴⁹ One needs validation data for each model contemplated for use in epidemiological studies.

View larger version (15K): [in this window] [in a new window]   Figure 3. Plot showing a review of published comparisons of blood pressure measurement between Dinamap oscillometric devices and sphygmomanometers. Shown are the name of the first author, the number of the Dinamap model, the type of sphygmomanometer, and the mean difference between the instruments along with 95% CI. SS indicates standard sphygmomanometer; RZ, random-zero sphygmomanometer; and MM, mercury manometer. Reproduced with permission from Whincup et al.49

Ambulatory Blood Pressure Monitors
Small, portable blood pressure monitors are now available that children can wear continuously for 12 to 24 hours.^{22 23 24 50 51} They can be programmed to take blood pressure readings at various intervals, often every 15 to 30 minutes during the day and every 30 to 60 minutes at night. Such ambulatory blood pressure monitors (ABPMs) work by either the auscultatory or oscillometric principle.

Strengths
One of the main purposes of the use of ABPMs in adults has been to quantify the effects on blood pressure of time of day and various activities such as sleep, work, mental stress, and physical activity. Likewise, they can be used in children for examination of effects of sleep and mental and physical activities. In addition, they provide multiple readings of blood pressure over a relatively short period of time, which can be helpful in characterizing within-person variability in the face of normal daily activity. Similarly, the multiple readings may help to reduce the nuisance effect of within-person variability in clinical trials or longitudinal studies. In adults, these multiple readings and the availability of readings away from the doctor's office (to reduce the white-coat effect) can make risk classification more accurate.⁵² In addition, treatment monitoring and adherence may be aided by the use of ABPMs.

Weaknesses
Even in adults, the lack of standardized validation criteria hinders assessment of ABPM data.⁵³ In children, there are far fewer data from which to draw conclusions. It is not clear which measures of blood pressure correspond with clinical outcomes of interest. Some authors recommend using a 24-hour average; some advocate the average during sleep; and others recommend using the time above a specified cutoff point.⁵⁴ As implied by some studies of adults, it may also turn out to be useful to estimate blood pressure response to certain stimuli such as exercise or mental stress, but this is not currently known.⁵⁵ It is also not known how many days of recording are necessary to characterize accurately a person's (adult's or child's) blood pressure response to any of these stimuli. There are also practical problems associated with ABPM use such as the high cost; necessity of moderating a child's activity while he or she is wearing the cuff; startle reaction to cuff inflation at night, which seems more of a problem in children than adults; and motion artifact and cuff or transducer slippage, which tend to invalidate readings.⁵⁰

Blood Pressure Instruments: Summary and Recommendations
The standard mercury sphygmomanometer, although used for several decades, has the important drawbacks of two observer biases, digit preference, and knowledge of previous readings. It is most useful in situations in which blood pressure is not the primary variable of interest, cost is an overriding concern, or one is comparing results with other studies that have used the same instrument.

The random-zero sphygmomanometer, by contrast, minimizes the biases of the standard mercury instrument. Thus, it is useful in studies in which blood pressure is the primary variable of interest or eligibility is determined by blood pressure level. Investigators should also consider its use in long-term follow-up studies of childhood blood pressure in which they may ultimately wish to compare adult levels of blood pressure with those from other studies. One should be aware, however, of recent data suggesting that the random-zero instrument underestimates blood pressure level compared with the standard mercury sphygmomanometer. The causes and implications of these findings remain to be clarified.

Automated oscillometric devices represent a new technology for blood pressure measurement. There is now enough experience with them in childhood epidemiological studies for investigators to consider their use in studies in which accurate blood pressure measurement is necessary, certainly in short-term intervention studies and perhaps also in long-term follow-up studies. They are particularly suitable for studies of young children. These devices may continue to gain wider use, but one should be aware of the need to assess the validity of each model of each device before using it in an epidemiological study.

At this time, ABPMs should not be considered for routine use in epidemiological studies because of the lack of standardized validation criteria, ambiguities of analyzing information from the devices, lack of data correlating adult ABPM results with morbidity, and practical difficulties of using the machines. However, ABPMs represent a promising technique for examining sources of blood pressure variability and may turn out to be the measurement method of choice in the future for tracking and intervention studies.

	How Many Measurements to Take

Top Abstract Introduction Factors That Affect Blood... Blood Pressure Instruments for... How Many Measurements to... References

 
In any study of blood pressure, the goal is to characterize each subject's blood pressure as accurately as possible. In theory, then, one would like to measure a person's "true" blood pressure, the average of an infinite number of measurements over a relatively short period of time. This theory applies to studies of stimuli-associated blood pressure, as measured with ABPMs, and to basal blood pressure. For example, if one were concerned with characterizing blood pressure response to exercise, one would theoretically want an infinite number of exercise-associated blood pressure readings.

In practice, one is limited to a finite number of visits, days or weeks apart, and to a finite number of measurements per visit to characterize basal blood pressure. Knowledge of blood pressure variability allows us to estimate the point where increasing the number of visits or measurements per visit leads to diminishing returns.

Blood Pressure Variability
Variance of blood pressure can be divided into the between-person component, which is generally the quantity of interest in an epidemiological study, and the within-person component, a nuisance variable to be taken into account. Within-person variability is similar to measurement error in that it can bias estimates of effects and correlations.⁵⁶ In children compared with adults, the within-person component represents a higher fraction of total variance.⁵⁷ This finding makes accounting or correcting for within-person variability a high priority in studies of blood pressure in children; thus, the number of visits and measurements per visit is a vital consideration.

When blood pressure is measured multiple times at each of several discrete visits, one can divide the within-person variance into between-visit and within-visit (between-reading) components. We have performed such calculations in two studies. The first involved measurement of blood pressure over four weekly visits, three measurements per visit, with both a random-zero and a standard mercury sphygmomanometer among 162 children 8 to 12 years of age. The second study used an automated oscillometric device. During the baseline period of an intervention trial, we measured blood pressure over three weekly visits, four measurements per visit, using a Dinamap model 845XT, among 106 children 9 to 13 years of age.

Optimal Number of Measurements
To examine the benefit of increasing numbers of visits and measurements per visit, we first calculated the between-person and the between- and within-visit components of within-person variance. Using a nested ANOVA model, we fit the data with PROC MIXED of SAS Institute Inc,⁵⁸ including terms for age, sex, race, and measurement number. The Table shows the values for the variance components. Between-visit variances for both the random-zero and standard instruments were higher than the within-visit variance; the opposite was true for the Dinamap device.

We then used these estimates to calculate the reliability (R), or the ratio of between-person to total variance (optimal ratio, 1) for specified values of numbers of visits and measurements per visit, as follows:

where _p² is the between-person variance, _a² is the between-visit variance with n visits, and _e² is the within-visit variance with n visits and k measurements per visit.

The results for systolic pressure for the random-zero and standard sphygmomanometers and a Dinamap device are shown in Figs 4 through 6, respectively. For the random-zero device (Fig 4), the slope of each "visit" line is almost level after about two or three measurements per visit, but the distance between visit lines is relatively large until about three or four visits. The implication is that the number of visits is quite important, with less importance placed on the number of measurements per visit after, eg, three. Although the absolute values of variance components are different for the standard instrument, this same relation of number of visits and number of measurements holds, as Fig 5 shows.

View larger version (23K): [in this window] [in a new window]   Figure 4. Plot showing reliability (ratio of between-person to total variance) for various combinations of number of visits and number of measurements per visit with a random-zero sphygmomanometer. Data from 162 children 8 to 12 years of age from East Boston, Mass.

View larger version (23K): [in this window] [in a new window]   Figure 5. Plot showing reliability (ratio of between-person to total variance) for various combinations of number of visits and number of measurements per visit with a standard mercury sphygmomanometer. Data from 162 children 8 to 12 years of age from East Boston, Mass.

View larger version (24K): [in this window] [in a new window]   Figure 6. Plot showing the reliability (ratio of between-person to total variance) for various combinations of number of visits and number of measurements per visit with an automated oscillometric device (Dinamap 845XT). Data from 106 children 9 to 13 years of age from Boston, Mass.

In contrast, for the Dinamap device, increasing the number of measurements per visit past three continues to improve reliability. This can be seen in Fig 6, where the slope of each visit line is steeper than for the auscultatory devices, leveling off only after four to five measurements per visit. Perhaps counterintuitively, there is still a relatively large jump from one visit to two or even three with the Dinamap device. This result suggests that, even with a relatively small between-person variance, the number of visits is still important. The reason is evident in the equation above in which _e² is divided by n and k, but _a² is divided only by n. That is, the contribution of _a² to the total variance is reduced only by a factor of n, not by nk.

Given that the first of several Dinamap measurements at a particular visit is generally higher than those following, the question arises as to whether to drop the first reading in analyses. If the primary goal, as in many studies of children, is to reduce the effects of within-person variability, then our data suggest that retaining the first value is the best strategy. For example, after dropping the first of four measurements per visit at each of the three visits in our study with the Dinamap device, the between-person variance was 35.9, and the between-visit and within-visit components were 11.5 and 30.3, respectively. The resulting reliability is 0.83. This value is less than that found by using all four measurements per visit over the three visits, 0.86.

It should be noted that estimates of variance components may differ by age or other variables in addition to the type of instrument. For instance, Lombardi et al⁵⁹ used the Dinamap model 845-A to assess variability among adolescent girls and boys 13 to 17 years of age. In contrast to the above results, they found the between-visit component to exceed the within-visit component. Therefore, although the general conclusion holds that the number of visits is at least as important as the number of measurements per visit, precise estimates of variance components and thus reliability estimates may vary from study to study and by instrument.

The conclusions from these data are that in children (1) the number of visits is at least as important as the number of measurements per visit in characterizing a subject's blood pressure; (2) the shape of the reliability curves, and therefore the optimal number of measurements, may differ by device, age, or other variables; and therefore (3) investigators should regard the quantitative data presented here as only rough guides to study design. In addition, exactly how many visits and measurements per visit to use in any study is a function of the costs and benefits associated with each point on such curves. However, it is possible to put forward some general guidelines for the instruments considered here. Compared with obtaining data at multiple visits, taking replicate readings at a visit is usually easier and less costly. From the data we have presented, about three readings per visit would be useful with the auscultatory devices; four or five readings would be useful with the Dinamap devices. The importance of the number of visits, however, mandates that at least two, and preferably three, visits should be obtained with any of the devices.

Conclusions
It is important to measure blood pressure accurately in epidemiological studies of children. To reduce observer bias, standardization and certification of observers are critical. Reduction of biases resulting from equipment, subject, environment, and technique factors is desirable. The choice of instrument depends on the goals of the study and the resources available to the investigators, and this choice may have an important influence on the precision of blood pressure estimates. Multiple measurements are vital in estimating childhood blood pressure, and obtaining data from several visits, days or weeks apart, is generally more important than increasing the number of measurements per visit beyond a few.

	Acknowledgments

 
This research was supported by NIH grants R29-HL-48236 and HL-48289 and by Harvard Medical School and the Harvard Community Health Plan Foundation. Dr Gillman is a Merck/Society for Epidemiologic Research Clinical Epidemiology Fellow. We thank Terry Field, ScD, and Walter Gamble, MD, for their analytic contributions to this article.

	Footnotes

 
Reprint requests to Dr M.W. Gillman, Department of Ambulatory Care and Prevention, HMS/HCHP, 126 Brookline Ave, Ste 200, Boston, MA 02215.

Dr Gillman presented this material in part as the Richard D. Remington Methodology Lecture at the 34th Annual Meeting of the Council on Epidemiology and Prevention, American Heart Association, Tampa, Fla, March 1994.

Received October 13, 1994; revision received February 2, 1995; accepted February 19, 1995.

	References

Top Abstract Introduction Factors That Affect Blood... Blood Pressure Instruments for... How Many Measurements to... References

 
1. Gillman MW, Cook NR, Rosner B, Evans DA, Keough ME, Taylor JO, Hennekens CH. Identifying children at high risk for the development of essential hypertension. J Pediatr. 1993;122:837-846. [Medline] [Order article via Infotrieve]

2. Lauer RM, Clarke WR. Childhood risk factors for high adult blood pressure: the Muscatine Study. Pediatrics. 1989;79:1-25. [Abstract/Free Full Text]

3. Berenson GS, Srinivasan SR, Webber LS, Nicklas TA, Hunter SM, Johnson CC, Arbeit ML, Dalferes ER, Wattigney WA, Lawrence MD. Cardiovascular risk in early life: the Bogalusa Heart Study. In: Current Concepts. Kalamazoo, Mich: Upjohn Co; 1991.

4. Gillman MW, Ellison RC. Childhood prevention of essential hypertension. Pediatr Clin North Am. 1993;40:179-194. [Medline] [Order article via Infotrieve]

5. Perloff D, Grim C, Flack J, Frohlich ED, Hill M, McDonald M, Morgenstern BZ. Human blood pressure determination by sphygmomanometry. Circulation. 1993;88:2460-2470. AHA Medical/Scientific Statement. [Free Full Text]

6. Whincup PH, Cook DG, Shaper AG. Blood pressure measurement in children: the importance of cuff bladder size. J Hypertens. 1989;7:845-850. [Medline] [Order article via Infotrieve]

7. Lum LG, Jones JMD. The effect of cuff width on systolic blood pressure measurements in children. J Pediatr. 1977;91:963-966. [Medline] [Order article via Infotrieve]

8. Steinfeld L, Dimich I, Reder R, Cohen M, Alexander H. Sphygmomanometry in the pediatric patient. J Pediatr. 1978;92:934-938. [Medline] [Order article via Infotrieve]

9. Gomez O, Prineas RJ, Rastam L. Cuff bladder width and blood pressure measurement in children and adolescents. J Hypertens. 1992;10:1235-1241. [Medline] [Order article via Infotrieve]

10. Prineas RJ, Elkwiry ZM. Epidemiology and measurement of high blood pressure in children and adolescents. In: Loggie JM, ed. Pediatric and Adolescent Hypertension. Boston, Mass: Blackwell Scientific Publications; 1992 ;91-103.

11. Vyse TJ. Sphygmomanometer bladder length and measurements of blood pressure in children. Lancet. 1987;1:561-562.

12. Jenner DA, Vandongen R, Beilin LJ. Blood pressure and body composition in children: importance of allowing for cuff size. J Hum Hypertens. 1992;6:367-374. [Medline] [Order article via Infotrieve]

13. Prineas RJ, Jacobs DR. Quality of Korotkoff sounds: bell vs. diaphragm, cubital fossa vs brachial artery. Prev Med. 1983;12:715-719. [Medline] [Order article via Infotrieve]

14. Londe S. Blood pressure measurement. Pediatrics. 1987;80:967-968. Letter. [Abstract/Free Full Text]

15. Prineas RJ, Gillum RF, Horibe H, Hannan PJ. The Minneapolis Children's Blood Pressure Study, I: standards of measurement for children's blood pressure. Hypertension. 1980;2(suppl 1):I-18-I-24.

16. Mitchell PL, Parlin RW, Blackburn H. Effect of vertical displacement of the arm on indirect blood-pressure measurement. N Engl J Med. 1964;271:72-74.

17. Park MK, Menard SM. Normative oscillometric blood pressure values in the first 5 years in an office setting. AJDC. 1989;143:860-864.

18. Task Force on Blood Pressure Control in Children. Report of the Second Task Force on Blood Pressure Control in Children, 1987. Pediatrics. 1987;79:1-25.

19. Rosner B, Prineas RJ, Loggie JM, Daniels SR. Blood pressure nomograms for children and adolescents, by height and age, in the United States. J Pediatr. 1993;123:871-876. [Medline] [Order article via Infotrieve]

20. Kahn HS, Bain RP, Pullen-Smith B. Interpretation of children's blood pressure using a physiologic height correction. J Chronic Dis. 1986;39:521-531. [Medline] [Order article via Infotrieve]

21. Kahn HS, Bain RP. Vertex-corrected blood pressure in black girls: relation to obesity, glucose, and cations. Hypertension. 1987;9:390-397. [Abstract/Free Full Text]

22. Portman RJ, Yetman RJ, West MS. Efficacy of 24-hour ambulatory blood pressure in monitoring in children. J Pediatr. 1991;118:842-849. [Medline] [Order article via Infotrieve]

23. Harshfield GA, Pulliam DA, Alpert BS, Stapleton B, Willey ES, Somes GW. Ambulatory blood pressure patterns in children and adolescents: influence of renin-sodium profiles. Pediatrics. 1991;87:94-100. [Abstract/Free Full Text]

24. Krull F, Buck T, Offner G, Brodehl J. Twenty-four hour blood pressure monitoring in healthy children. Eur J Pediatr. 1993;152:555-558. [Medline] [Order article via Infotrieve]

25. White WB, Schulman P, McCabe EJ, Dey HM. Average daily blood pressure, not office blood pressure, determines cardiac function in patients with hypertension. JAMA. 1989;261:873-877. [Abstract/Free Full Text]

26. Hornsby JL, Mongan PF, Taylor T, Treiber FA. `White coat' hypertension in children. J Fam Pract. 1991;33:617-623. [Medline] [Order article via Infotrieve]

27. Jenner DA, English DR, Vandongen R, Beilin LJ, Armstrong BK, Dunbar D. Environmental temperature and blood pressure in 9-year-old Australian children. J Hypertens. 1987;5:683-686. [Medline] [Order article via Infotrieve]

28. Canner PL, Borhani NO, Oberman A, Cutler J, Prineas RJ, Langford H, Hooper FJ. The Hypertension Prevention Trial: assessment of the quality of blood pressure measurements. Am J Epidemiol. 1991;134:379-392. [Abstract/Free Full Text]

29. Dischinger P, DuChene AG. Quality control aspects of blood pressure measurements in the Multiple Risk Factor Intervention Trial. Control Clin Trials. 1986;7:137S-157S. [Medline] [Order article via Infotrieve]

30. Uhari M, Nuutinen M, Turtinen J, Pokka T. Pulse sounds and measurement of diastolic blood pressure in children. Lancet. 1991;338:159-161. [Medline] [Order article via Infotrieve]

31. Sinaiko AR, Gomez-Marin O, Prineas RJ. Diastolic fourth and fifth phase blood pressure in 10-15-year-old children: the Children and Adolescent Blood Pressure Program. Am J Epidemiol. 1990;132:647-655. [Abstract/Free Full Text]

32. Kannel WB. Impact of single and multiple risk factors on coronary disease and its treatment: focus on hypertension. Clin Ther. 1987;9:2-19.

33. Gillman MW, Cook NR, Rosner B, Beckett LA, Evans DA, Keough ME, Taylor JO, Hennekens CH. Childhood blood pressure tracking correlations corrected for within-person variability. Stat Med. 1992;11:1187-1194. [Medline] [Order article via Infotrieve]

34. Gillman MW, Rosner B, Evans DA, Keough ME, Taylor JO, Smith LA, Hennekens CH. Blood pressures at multiple visits increase tracking correlations in childhood. Pediatrics. 1991;87:708-711. [Abstract/Free Full Text]

35. Park MK, Menard SM. Accuracy of blood pressure measurement by the Dinamap monitor in infants and children. Pediatrics. 1986;79:907-914. [Abstract/Free Full Text]

36. Wright BM, Dore CF. A random-zero sphygmomanometer. Lancet. 1970;1:337-338. [Medline] [Order article via Infotrieve]

37. O'Brien E, Mee F, Atkins N, O'Malley K. Inaccuracy of the Hawksley random zero sphygmomanometer. Lancet. 1990;336:1465-1468. [Medline] [Order article via Infotrieve]

38. Conroy RM, O'Brien E, O'Malley K, Atkins N. Measurement error in the Hawksley random zero sphygmomanometer: what damage has been done and what can we learn? Br Med J. 1993;306:1319-1322.

39. Churchill D, Beevers M, Beevers DG. The Hawksley random zero sphygmomanometer: don't condemn it without proper evidence. Br Med J. 1993;307:123-124. Letter.

40. Uhari M, Nuutinen M, Turtinen J, Pokka T, Kuusela V, Akerblom HK, Dahl M, Kaprio EA, Pesonen E, Pietikainen M, Salo MK, Viikari J. Blood pressure in children, adolescents and young adults. Ann Med. 1991;23:47-51. [Medline] [Order article via Infotrieve]

41. Nuutinen M, Turtinen J, Uhari M. Random-zero sphygmomanometer, Rose's tape, and the accuracy of the blood pressure measurements in children. Pediatr Res. 1992;32:243-247. [Medline] [Order article via Infotrieve]

42. The Hawksley random zero sphygmomanometer. Br Med J. 1993;307:123-125. Correspondence.

43. Kronmal RA, Rutan GH, Borhani NO, Manolio TA, Furberg CD. Potential problems with random zero sphygmomanometer. Lancet. 1990;335:360. Letter.

44. The Hawksley random zero sphygmomanometer. Br Med J. 1993;307:562-563. Correspondence. [Free Full Text]

45. Ramsey M III. Blood pressure monitoring: automated oscillometric devices. J Clin Monit. 1991;7:56-67. [Medline] [Order article via Infotrieve]

46. Gillman MW, Hood MY, Moore LL, Singer MR, Nguyen U, Andon MB. Effect of calcium supplementation on blood pressure in inner-city children. J Pediatr. 1995;127:186-192. [Medline] [Order article via Infotrieve]

47. Perry CL, Stone EJ, Parcel GS, Ellison RC, Nader PR, Webber LS, Luepker RV. School-based cardiovascular health promotion: the Child and Adolescent Trial for Cardiovascular Health (CATCH). J Sch Health. 1990;60:406-413. [Medline] [Order article via Infotrieve]

48. Weaver MG, Park MK, Lee D. Differences in blood pressure levels obtained by auscultatory and oscillometric methods. AJDC. 1990;144:911-914.

49. Whincup PH, Bruce NG, Cook DG, Shaper AG. The Dinamap 1846SX automated blood pressure recorder: comparison with the Hawksley random zero sphygmomanometer under field conditions. J Epidemiol Community Health. 1992;46:164-169. [Abstract/Free Full Text]

50. Jacoby AC, Fixler DE, Torres EJ. Limitations of an oscillometric ambulatory blood pressure monitor in physically active children. J Pediatr. 1993;122:231-236. [Medline] [Order article via Infotrieve]

51. Harshfield GA, Alpert BS, Pulliam DA, Somes GW, Wilson DK. Ambulatory blood pressure recordings in children and adolescents. Pediatrics. 1994;94:180-184. [Abstract/Free Full Text]

52. Appel LJ, Stason WB. Ambulatory blood pressure monitoring and blood pressure self-measurement in the diagnosis and management of hypertension. Ann Intern Med. 1993;118:867-882. [Abstract/Free Full Text]

53. O'Brien E, O'Malley K, Sheridan J. The need for a standardized protocol for validating non-invasive ambulatory blood-pressure measuring devices. J Hypertens. 1989;7:S19-S20.

54. American College of Physicians. Automated ambulatory blood pressure and self-measured blood pressure monitoring devices: their role in the diagnosis and management of hypertension. Ann Intern Med. 1993;118:889-892. [Abstract/Free Full Text]

55. Staessen J, Celis H, De Cort P, Fagard R, Thijs L, Amery A. An inventory of studies on ambulatory blood pressure in large groups of subjects. J Hypertens. 1991;9:S48-S50.

56. Liu K. Measurement error and its impact on partial correlation and multiple linear regression analyses. Am J Epidemiol. 1987;127:864-874. [Abstract/Free Full Text]

57. Rosner B, Cook NR, Evans DA, Keough ME, Taylor JO, Polk BF, Hennekens CH. Reproducibility and predictive values of routine blood pressure measurements in children. Am J Epidemiol. 1987;126:1115-1125. [Abstract/Free Full Text]

58. SAS Institute Inc. SAS/STATUser's Guide, Version 6. 4th ed. Cary, NC: SAS Institute Inc; 1993.

59. Lombardi DA, Gillman MW, Ellison RC, Hosmer DW, Puleo EM. Estimates of blood pressure variability in adolescents using an automatic blood pressure measurement device. J Hum Hypertens. 1994;8:441-443.[Medline] [Order article via Infotrieve]

This article has been cited by other articles:

				R. Bakker, S. L. Rifas-Shiman, K. P. Kleinman, S. E. Lipshultz, and M. W. Gillman Maternal Calcium Intake During Pregnancy and Blood Pressure in the Offspring at Age 3 Years: A Follow-up Analysis of the Project Viva Cohort Am. J. Epidemiol., December 15, 2008; 168(12): 1374 - 1380. [Abstract] [Full Text] [PDF]

				S. Kinra, K V Rameshwar Sarma, Ghafoorunissa, V. V. R. Mendu, R. Ravikumar, V. Mohan, I. B Wilkinson, J. R Cockcroft, G. Davey Smith, and Y. Ben-Shlomo Effect of integration of supplemental nutrition with public health programmes in pregnancy and early childhood on cardiovascular risk in rural Indian adolescents: long term follow-up of Hyderabad nutrition trial BMJ, July 25, 2008; 337(jul25_1): a605 - a605. [Abstract] [Full Text] [PDF]

				X. Chen and Y. Wang Tracking of Blood Pressure From Childhood to Adulthood: A Systematic Review and Meta-Regression Analysis Circulation, June 24, 2008; 117(25): 3171 - 3180. [Abstract] [Full Text] [PDF]

				X. Chen, Y. Wang, L. J. Appel, and J. Mi Impacts of Measurement Protocols on Blood Pressure Tracking From Childhood Into Adulthood: A Metaregression Analysis Hypertension, March 1, 2008; 51(3): 642 - 649. [Abstract] [Full Text] [PDF]

				L. K. Washburn, P. A. Nixon, and T. M. O'Shea Follow-up of a Randomized, Placebo-Controlled Trial of Postnatal Dexamethasone: Blood Pressure and Anthropometric Measurements at School Age Pediatrics, October 1, 2006; 118(4): 1592 - 1599. [Abstract] [Full Text] [PDF]

				A. H Hemachandra, M. A Klebanoff, A. K Duggan, J. B Hardy, and S. L Furth The association between intrauterine growth restriction in the full-term infant and high blood pressure at age 7 years: results from the Collaborative Perinatal Project Int. J. Epidemiol., August 1, 2006; 35(4): 871 - 877. [Abstract] [Full Text] [PDF]

				S. Y Huh, S. L Rifas-Shiman, K. P Kleinman, J. W Rich-Edwards, S. E Lipshultz, and M. W Gillman Maternal protein intake is not associated with infant blood pressure Int. J. Epidemiol., April 1, 2005; 34(2): 378 - 384. [Abstract] [Full Text] [PDF]

				M. K. Park Blood Pressure Tables Pediatrics, March 1, 2005; 115(3): 826 - 827. [Full Text] [PDF]

				M. W. Gillman, S. L. Rifas-Shiman, K. P. Kleinman, J. W. Rich-Edwards, and S. E. Lipshultz Maternal Calcium Intake and Offspring Blood Pressure Circulation, October 5, 2004; 110(14): 1990 - 1995. [Abstract] [Full Text] [PDF]

				J. A. Clark, M. W. Lieh-Lai, A. Sarnaik, and T. K. Mattoo Discrepancies Between Direct and Indirect Blood Pressure Measurements Using Various Recommendations for Arm Cuff Selection Pediatrics, November 1, 2002; 110(5): 920 - 923. [Abstract] [Full Text] [PDF]

				A. M. Simckes, T. Srivastava, and U. S. Alon Ambulatory Blood Pressure Monitoring in Children and Adolescents Clinical Pediatrics, October 1, 2002; 41(8): 549 - 564. [Abstract] [PDF]

				M. L. Cruz, T. T-K. Huang, M. S. Johnson, B. A. Gower, and M. I. Goran Insulin Sensitivity and Blood Pressure in Black and White Children Hypertension, July 1, 2002; 40(1): 18 - 22. [Abstract] [Full Text] [PDF]

				M. K. Park, S. W. Menard, and C. Yuan Comparison of Auscultatory and Oscillometric Blood Pressures Arch Pediatr Adolesc Med, January 1, 2001; 155(1): 50 - 53. [Abstract] [Full Text] [PDF]

				M. Arafat and T. K. Mattoo Measurement of Blood Pressure in Children: Recommendations and Perceptions on Cuff Selection Pediatrics, September 1, 1999; 104 (3): e30 - e30. [Abstract] [Full Text] [PDF]

				J J O'Sullivan, G Derrick, P Griggs, R Foxall, M Aitkin, and C Wren Ambulatory blood pressure in schoolchildren Arch. Dis. Child., June 1, 1999; 80(6): 529 - 532. [Abstract] [Full Text]

This Article Abstract Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Gillman, M. W. Articles by Cook, N. R. Search for Related Content PubMed PubMed Citation Articles by Gillman, M. W. Articles by Cook, N. R.