AMERICAN JOUHNAI. or EPIDEMIOLOGY
Vol. 109, No. 3
Copyright © 1979 by The Johns Hopkins University School of Hygiene and Public Health
Printed in USA.
All rights reserved
TIME COURSE STUDIES OF BLOOD PRESSURE IN CHILDREN—THE BOGALUSA HEART STUDY A. WOUTER VOORS, LARRY S. WEBBER AND GERALD S. BERENSON1 Voors, A. W., l_ S. Webber and G. S. Berenson (Louisiana State U., Medical Center, 1542 Tulane Ave., New Orleans, LA 70112). Time course studies of blood pressure In children:—the Bogalusa Heart Study. Am J Epidemiol 109:320-334, 1979. Blood pressures (BPs) were taken with a mercury sphygmomanometer and an automatic recorder on 3524 children representative of an entire geographic community. For all children ages 5, 8, 11, and 14 years In the Initial examination, age-specific systolic and diastolic (4th phase) selected percentlles were assessed. Of these children, 1101 were reexamlned after one year. Observations from a group of 35 fifth-graders examined monthly for eight months were pooled to observe Intra-child BP variability. This estimate was used to reduce to zero In a statistical adjustment the regression toward the mean of the BPs for the after-one-year reexamlned children. It was found that those children Initially In the top ten percentiles had, upon reexamlnatlon, on the average only 3 mmHg lower systolic and 1 mmHg lower diastolic levels. In a multiple regression analysis, the previous year's BP contributed a partial correlation coefficient of 0.6-0.7 for each age cohort to the variability of the BP, controlling for other determinants. These findings, based on reliable, basal-llke measurements, point to a degree of persistence which Is quite high. The higher the degree of tracking the more likely that primary hypertension begins early in life. blood pressure; hypertension; pediatrics; school health
The variations of blood pressure (BP) levels in children over time need to be elucidated. The nature of such changes is not known, but it is possibly related to the development of essential hypertension, Although it has been stated (1, 2) that essential hypertension is rare in children, there is evolving evidence (3, 4) that essential hypertension begins in childhood,
It may also be assumed that prevention of hypertension would be more successful if the disease process were aborted early, Hence, for early diagnosis we need to know the time course of BP levels and whether "tracking" occurs (i.e., to what extent a child with relatively high BP levels will persist with relatively high levels), especially since the BP level dur-
Received for publication February 2, 1978, and in final form July 5, 1978. Abbreviations: BP, blood pressure; H, height; Hg, mercury; N, sample size; SE, standard error of the mean; W, weight. 1 From the Department of Public Health and Preventive Medicine and the Department of Medicine, Louisiana State University Medical Center, New Orleans, LA 70112. Address reprint requests to: Dr. G. S. Berenson, Department of Medicine, Louisiana State University Medical Center, 1542 Tulane Ave., New Or-
leans, LA 70112. Supported by funds from the National Heart, Lung, and Blood Institute of the USPHS, Specialized Center of Research-Arteriosclerosis (SCOR-A), HL15103. The authors are indebted to Ms. Imogene W. Talley for her work as a community coordinator; Rita B. Clayton, R.N., who read all of the physiometrics discs; the entire Bogalusa Heart Study staff (espedally the blood pressure observers) for their measurement assistance; and Ms. Leanne Warren for computations.
320
TIME COURSE STUDIES OF BLOOD PRESSURE IN CHILDREN
ing adolescence and young adulthood has been found to be predictive of later hypertension (5-9). Conventional impressions (10-14) emphasize the lability of BPs at the early phases of hypertension. However, these early phases of hypertension are often defined by casual BP measurements, which are less predictive of future hypeiicnsion than basal pressures (5, 6). Lability here may merely mean that the measured pressures of individuals exceed defined abnormal cutpoint levels more often, either because their mean is close to this cutpoint or because of increased variability of these BPs. In the first case they are not truly labile, and in the second they are not truly hypertensive. To help clarify these points, data are needed from an entire geographic population of free-living children rather than from hospital-based patients and nonrepresentative selected individuals; data obtained by standardized methodology are also required. Pertinent information is now available from the community-wide examination of school aged children as performed in the Bogalusa Heart Study. Little is known about the persistence as well as determinants of high ranking BP levels in children. If a child has a BP level which ranks high, how likely will this ranking persist after an hour, a month, a year? What causes basal BPs to persist or to change over the short term? Answers to these questions will help begin our understanding of the early onset of essential hypertension. MATERIALS AND METHODS
Population. The eligible population consisted of all children, ages five through 14 years, residing in Ward 4 of Washington Parish (County), Louisiana, including the town of Bogalusa. Ninetythree per cent participated in the study, a total of 3524 children. In one of the participating schools, 69 children in the fifth grade were reexamined for BP once a month for eight
321
consecutive months. Thirty-five of these children attended all eight sessions. Children 5, 8, 11 and 14 years of age during the first examination in the study were reexamined after one year. A total of 1101 children (83 per cent of the 1326 eligible) were examined twice. General examination procedures. Because the objective of the screening procedures was to assess risk factor variables for coronary artery disease and hypertension, the following variables were measured: blood pressure, height, weight, triceps skinfold thickness, serum lipids (total cholesterol, triglycerides, lipoprotein fractions), and hemoglobin. Each child also received a physical examination which included external maturation measurements (secondary sex characteristics as described by Tanner (15)). All measurements followed methods outlined in a detailed protocol (16—19). BPs were recorded as the last procedure of the screening process which lasted from one and a half to two hours. Each morning at the end of the examination, a random sample of four children was reexamined by the same examiners. The selection of this sample was unknown to the examiners prior to reexamination. Blood pressure instruments. Indirect BPs were obtained by the mercury sphygmomanometer (Baumanometer", W. A. Baum Co., Inc., Copiague, NY) and by the Physiometrics" automatic blood pressure recorder (Sphygmetrics, Inc., Woodland Hills, CA). The sphygmomanometer selection of the cuff bladder sizes was based on criteria for arm measurements (18), which generally designated maximal bladder size while leaving the elbow skin crease free for the stethoscope. The PhysiometricsR instrument amplifies infrasonic sounds detected by an electronic transducer and records on a paper disc rotated by an aneroid manometer in open communication with an over-sized rubber cuff bladder entirely encircling the upper arm. The transducer
322
VOORS, WEBBER AND BERENSON
is built into the cuff bladder for maximal sound amplitude. Amplification covers the 10-80 hertz frequencies and maximizes Korotkoff phase discrimination. Blood pressure measurements. The children were encouraged to urinate before BPs were measured, and local volunteers guided them through a designated random sequence for examination. For each child the BP was determined in a randomized sequence by three trained observers; each observer independently measured three BPs per child (figure 1). The total BP assessment of a child during this examination did not last more than 30 minutes. Two observers each used a well-illuminated mercury sphygmomanometer placed at eye level and the
Children Enter in Random Sequence
third a PhysiometricsR recorder; each instrument was assigned randomly on a daily basis, as were the observers to the instruments. Each day before use the PhysiometricsR instrument was calibrated with the mercury column of a sphygmomanometer. Under quiet conditions, in a screened-off corner, observers recorded the first and fourth Korotkoff phases to determine the right arm BP while the child was seated. Care was taken to cover the brachial artery with the center of the rubber bladder. Physiometrics" discs were read blindly by the same individual throughout the entire study period. After lengthy protocol development and pretesting, all BP observers (registered and licensed practical nurses) were
Random: -Instrument Sequence (Child) -Observer Allocation Among 6 Instruments (Daily)
10% Random* Sample Rescreened Daily
Finish
1
Randomized 1 Allocation I
FIGURE 1. Pattern of flow for examining children for BP in the Bogalusa Heart Study, 1973-1975. Hg = mercury gphygmomanometer, P" = Physiometrica" automatic BP recorder.
323
TIME COURSE STUDIES OF BLOOD PRESSURE IN CHILDREN TABLE 1
Measured variables entered into the multiple regression equations, Bogatusa Heart Study, 1973-1974 Description Dependent variables Intra-child standard deviation of 6 systolic pressures by mercury sphygmomanometer Intra-child standard deviation of 6 diastolic (4th phase) pressures by mercury sphygmomanometer Independent Variables Blood Pressure Mean of 3 systolic pressures by Physiometrics" Mean of 3 diastolic pressures by Physiometrics" Blood pressure examiners nos. 1-8* Demographic Age (exact calendar age) Race* Sex* Athropometric Body height Ponderosity indext = weight/height3 Upper-arm circumferencet Triceps skinfold thicknesst External maturation index (pubic hair score plus breast development or male genitalia score) Laboratory Serum total cholesterolt Serum triglyceridest Serum prebeta lipoproteint Serum beta lipoproteint (Serum prebeta lipoprotein) + (Serum beta lipoprotein)t (Serum prebeta lipoprotein) / (Serum prebeta + beta liproprotein)t Blood hemoglobin
Units of measurement
mmHg mmHg
mmHg mmHg yes = 1, no = 2 years white = 1, black = 2 male = 1, female = 2 cm kg/cm3 cm mm Score range 2-10 mg/dl mg/dl mg/dl mg/dl mg/dl
g/dl
* Entered as dummy variables. t Logarithmic transformation.
trained, tested and retrained at regular variation still rises. This confounding, if intervals (four times annually) during the it occurs, will, therefore, tend to create a spuriously high variability at both ends of entire period of investigation. the observed (not of the true) BP range in Analysis the population (20). In our analysis, we Relation between variability and level of used the six sphygmomanometer meaBP. When the BP of a child is measured surements to compute the variability, the mean of the three Physseveral times, any gross deviation in whereas 11 iometrics measurements, taken inmeasurement from the true level will independent of the sphygmomanometer fluence both the mean and the variation of the child's measured BP. For instance, readings, were used to assess the child's if the variant measurement overstates mean level of BP. The latter instrument the true pressure, both the mean and the may have produced BP levels more valid variation will rise, while a variant mea- than the sphygmomanometer (18), since surement which understates the true larger cuff bladders were used (21), inpressure will lower the mean while the frasonic amplification was applied (22),
324
VOORS, WEBBER AND BERENSON
and permanent records were produced that could be read without observer bias (23). Determinants of BP variability. Individual variability of BP (expressed as the standard deviation about the six sphygmomanometer readings) was entered as the dependent variable into a multiple regression analysis, and all other meaTABLE 2
Intra-child correlation coefficient of blood pressure (mercury sphygmomanometer) over monthly time intervals, Bogalusa Heart Study, 1973-1976* Time interval (months)
Systolic
Diastolic «th phaae)
1 2 3 4 5 6 7
0.81 0.84 0.84 0.80 0.83 0.78 0.80
0.65 0.50 0.47 0.58 0.60 0.47 0.44
* Thirty-five children in the fifth-grade of an elementary school.
sured characteristics, including individual mean BP levels, were entered as independent variables. Logarithmic transformation was applied to the independent variables (except age, height, maturation, hemoglobin, and the nominal characteristics) in order to approach Gaussian distributions. A ponderosity index (weightyheight3) was chosen as a measure of obesity for theoretic and morphologic reasons (19). This ponderosity index was used instead of the body weight as an independent variable in order to avoid artifacts resulting from the highly positive correlation between height and weight (table 1). Regression of initially high BP toward the mean. In order to compute the amount of regression toward the population mean BP as experienced by a child above the 90th percentile for his age, we used the statistical model of Gardner and Heady (24), which requires knowledge of the inter-child and intra-child variability.
TABLB 3
Intra-child correlation coefficient of blood pressure (mercury sphygmomanometer) over a one-year interval, by age-cohort, race, and sex, Bogalusa Heart Study, 1973-1975 Phase and age (years) at first examination
Systolic 5 8 11 14 Diastolic (4th phase) 5 8 11 14 ' Sample size in parentheses.
Girls
Boya White
Black
White
Black
0.64 (81)* 0.73 (90) 0.61 (92) 0.62 (81)
0.61 (43) 0.76 (60) 0.68 (72) 0.61 (53)
0.61 (74) 0.63 (88) 0.68 (100) 0.62 (60)
0.70 (35) 0.80 (52) 0.71 (69) 0.72 (51)
0.32 (81) 0.61 (90) 0.44 (92) 0.48 (81)
0.46 (43) 0.48 (60) 0.42 (72) 0.48 (53)
0.40 (74) 0.60 (88) 0.55 (100) 0.47 (60)
0.60 (35) 0.15 (52) 0.55 (69) 0.61 (51)
TIME COURSE STUDIES OF BLOOD PRESSURE IN CHILDREN
325
Relation between variability and level The inter-child variability was obtained from the same population for which we of BP at initial examination developed percentile charts (25). The For each child the intra-child standard intra-child variability (26) was computed deviation of the six systolic and diastolic by pooling the intra-child variances for sphygmomanometer pressures was comthe 35 fifth-graders examined monthly. puted. Mean (± 2 SE) intra-child stanTracking ofBP in children originally in dard deviations of systolic sphygmothe extreme deciles. In the study popula- manometer pressures are presented by tion with two complete examinations one intervals of 5 mmHg Physiometric8R presyear apart, children in the upper and sure level in figure 2. There was no statislower decile of BP (mean of six sphyg- tically significant relationship between momanometer measurements) for the ini- these two variables (test for linear regrestial examination were identified and their sion of standard deviation on level by BPs (mean of six sphygmomanometer analysis of variance procedure) at the measurements) were observed after one 0.05 level. Mean (± 2 SE) intra-child year. After reducing the regression to- standard deviations of diastolic sphygward the mean to zero in a statistical ad- momanometer pressures are also prejustment, the average reexamination sented in figure 2. Here, the association is pressures were compared to the original negative (test for linear regression of upper and lower decile pressures in order standard deviation on level by analysis of to obtain quantitative insight into the variance procedure, p < 0.0001). amount of tracking of the BP. Determinants of BP level. The deterDeterminants of BP variability minants of systolic and diastolic BP level at initial examination upon reexamination were ascertained by Time order in which measurements stepwise multiple regression analysis in were taken. Means for the first three of the which BP was the dependent variable. Innine measurements taken from each child dependent variables were all currently measured characteristics plus the BP are presented in table 4 by instrument. For systolic readings, the difference belevel obtained one year earlier. tween first and second measurements was larger than between the following meaRESULTS surements. However, the total of differences was not much more than 2.0 mmHg Bivariate correlations systolic and 0.3 mmHg diastolic for the Bivariate Pearson coefficients of corre- sphygmomanometer; and, respectively, lation between mean sphygmomanometer not much more than 2.8 and 1.0 mmHg pressures for various time intervals in the for the Physiometrics" instrument. The monthly reexamination data are shown in smallness of these differences perhaps table 2. The correlation coefficients ap- supports the notion that basal pressure proximated 0.80 for systolic and 0.50 for levels were approached. diastolic pressures. For the mercury sphygmomanometer, The annual reexamination data are differences between mean measurements broken down by age, race and sex in table by first and second station (instrument) 3. The correlation coefficients are approx- are distributed normally (figure 3). imated at 0.70 for systolic and 0.50 for Profile of determinants of BP variabildiastolic pressures. No consistent differ- ity. In order to obtain a profile of deterences in the correlation coefficients by minants of BP variability, we analyzed race and sex were observed. systolic and diastolic variability sepa-
326
VOORS, WEBBER AND BERENSON
SYSTOUC
I f
70
I
80 90 100 110 120 130 140 150 >I55 LEVEL of Blood Pressure, (mmHg) DIASTOLIC (4th phase)
§
30
40 50 60 70 80 90 >95 LEVEL of Blood Pressure (mmHg)
FIGURE 2. Variability of BP within one half hour. Intra-child standard deviation of systolic and diastolic (4th phase) BP (mean ± 2 SE) by level of systolic BP. The standard deviation was ascertained for each child by six mercury sphygmomanometric measurements. The level was ascertained for each child by three Physiometrics" measurements. Children, ages 5-14 years, the Bogalusa Heart Study, 1973-1974.
327
TIME COURSE STUDIES OF BLOOD PRESSURE IN CHILDREN
rately in a multiple stepwise regression model, entering BP variability as the dependent variable and all other measured characteristics as independent variables (table 1). The results are given in table 5. The systolic variability was found to be I20CK SYSTOLIC 1000
-30 -20 -10 0 +10 +20 +30 First Minus Second Station, mmHg FIGURE 3. Frequency distribution of difference in mean BP between first and second mercury sphygmomanometer station. Children, ages 5-14 years, the Bogalusa Heart Study 1973-1974.
negatively associated with age and positively with the index obesity (W/H3) but not with systolic level itself. The diastolic variability was found to be negatively associated with body height and weakly positively associated with maturation, but not with the diastolic level itself. All measured characteristics combined could explain only one per cent of the variation in systolic pressure variability and two per cent of the variation in diastolic pressure variability. Regression of initially high BP toward the mean The pooled intra-child standard deviation as obtained from the monthly reexamined fifth-graders was 4.1 mmHg systolic and 5.1 mmHg diastolic. Applying these results to the cross-sectional standard deviation for 11-year-olds as obtained from the initial examination of the entire population (9.2 mmHg systolic and 7.6 mmHg diastolic), the inter-child standard deviation can be computed by taking the square root of the difference in variance. The resulting inter-child standard deviations for 11-year-olds in the Bogalusa Heart Study are 8.3 mmHg systolic and 5.6 mmHg diastolic. Assuming that the BP values have Gaussian distributions, the model of Gardner and Heady (24) can be used to
TABLE 4
Mean blood pressures (mm Hg) of first, second, and third measurement in children age 5—14 years, by instrument, Bogalusa Heart Study, 1973-1974
Measurement
First Second Third All other
Mercury spbygmomanometer (JV - 2362*) Diastolic Systolic (4th phase)
Physiome tries" (W - 1164*) Diastolic Systolic (4th phase)
101.7t lOO.St 100.2t 99.7§
103.lt 101.3+ 100.3+ 100.3"
62.5 62.7 62.6 62.4§
62 6+ 62.1tt 61.8t 61.6"
* Eight missing values excluded from above totals. t Difference between indexed figures in same column is significant, p < 0.001. t Difference is significant, p < 0.05. § Not tested; averages of three mercury sphygmomanometric and three Physiometrics" measurements. 1 Not tested; averages of six mercury sphygmomanometric measurements.
328
VOORS, WEBBER AND BERENSON TABLE 5
Stepwise multiple regression of intra-child blood pressure variability, Bogalusa Heart Study, 1973-1974 Systolic Regression coefficient
Intercept Age W/H»t Observer 4 Observer 3
13.85 -0.05t 1.94§ OJ221
035'
Diastolic (4th phase) Standardized partial regression coefficient
-0.076 0.063 0.052 0.048
Regression coefficient
Intercept Height Observer 2 Observer 6 Observer 8 Maturation
7.36 -0.02$ 0.67$ 0.62t 0.43 § 0.081
Standardized partial regression coefficient
-0.132 0.111 0.102 0.072 0.073
N.B. Multiple correlation coefficient squared is 0.01 for systolic pressure and 0.02 for diastolic pressure. * 197 out of 3524 children not analyzed due to missing data; for variables entered, see table 1. tLog,* tp < 0.0001. §p < 0.001. »p < 0.01.
obtain the average decrease in pressure upon reexamination as experienced by a child ranking above the 90th percentile during the first examination. This decrease is 2.3 mmHg systolic and 4.4 mmHg diastolic (Appendix). These results are based on means of six sphygmomanometer measurements on a child per examination. Tracking of blood pressure for children originally in the extreme deciles In the study population with two complete examinations one year apart, children in the upper and lower decile of BPs (sphygmomanometer) for the first examination were identified and their BPs were observed during the second examination. Also, adjusted levels from the upper and lower deciles of the original examination were calculated after taking regression toward the mean into account. The latter are values statistically adjusted by reducing the regression towards the mean to zero (table 6). After adjustment for the annual difference in mean systolic and diastolic pressure, it was found that those children initially in the top 10 percentiles would have had average reexamination pressures corresponding to a decrease of 3.4 mmHg
systolic and 0.6 mmHg diastolic pressure, whereas those initially in the lower 10 percentiles would have had corresponding increases of 3.2 mmHg systolic and 2.0 mmHg diastolic pressure (figure 4). These mean shifts of BP levels for those children in the extreme deciles are perhaps within a five percentile range for systolic, and within a two percentile range for diastolic levels. Determinants of BP Level
The strongest determinant of BP level was the level observed one year previously (table 7). In total, all measured variables could account for 39-55 per cent of the systolic, and 32-38 per cent of the diastolic mercury sphygmomanometric pressure variation. A measure of tracking is the partial correlation coefficient of first-year with second-year BP, controlling for all other significant independent variables. This coefficient was 0.61-0.66 for systolic and 0.36—0.52 for diastolic pressure. In the regression analysis, the ponderosity index W/H3 correlates only for the two younger age groups, and the external maturation index only for the 12-year age group. Results for the Physiometrics" instrument were quite similar.
329
TIME COURSE STUDIES OF BLOOD PRESSURE IN CHILDREN
DISCUSSION
+1 +1 +1 +1 a>
a CO
ab
TOT
87. 89. 95.
00
o o
CO
The measurement of valid and replicable indirect BP in children is fraught with pitfalls, but it is apparently obtainable when attention is paid to methodology. In our study, we attempted to arrange for nearly optimal conditions under which to measure the children's BPs. By aiming for such conditions, we increased the predictive value of our measurements (5, 6) by obtaining values which approach known resting basal values for children in this age group (18). The measurement replicability was increased by measuring the BP three times by each of three different observers, and by using the 4th rather than the 5th Korotkoff phase for
1—1
CO CO CN CO
+1 +1 +1
•H
CN CD CO
51. 52. 59.
CO CN CO CD CN CN CN CN
c- o rto *
t>
CN
r-H
+1
+1
00 f-t (N CO
i-l CN
i-H
r>
CN
CN
CO
(M +1 +1
CO
105 110 114 119
CO
a>
CN CO
+1 +1 +1
c-
CD
f—I
i—i
1-1
»H
r-
i-H O
i-H
o
I 1
i—i
CN .-1 ao 00 I N i-l CD on
no O5 KX>
it
5
c o
90
80
CN
+1 +1 +1 +1 +1+1 on CO
115
+1 1—1
tI>
SYSTOLIC
120
.2
'
T3
7O-
r-t--t-t-
65
60-
Z2
cn
CN
-f
CN
CO CN
55-
CO lO CN • * o5 CN co (N
5O9
3
COCn
Vol. 109, No. 3
Copyright © 1979 by The Johns Hopkins University School of Hygiene and Public Health
Printed in USA.
All rights reserved
TIME COURSE STUDIES OF BLOOD PRESSURE IN CHILDREN—THE BOGALUSA HEART STUDY A. WOUTER VOORS, LARRY S. WEBBER AND GERALD S. BERENSON1 Voors, A. W., l_ S. Webber and G. S. Berenson (Louisiana State U., Medical Center, 1542 Tulane Ave., New Orleans, LA 70112). Time course studies of blood pressure In children:—the Bogalusa Heart Study. Am J Epidemiol 109:320-334, 1979. Blood pressures (BPs) were taken with a mercury sphygmomanometer and an automatic recorder on 3524 children representative of an entire geographic community. For all children ages 5, 8, 11, and 14 years In the Initial examination, age-specific systolic and diastolic (4th phase) selected percentlles were assessed. Of these children, 1101 were reexamlned after one year. Observations from a group of 35 fifth-graders examined monthly for eight months were pooled to observe Intra-child BP variability. This estimate was used to reduce to zero In a statistical adjustment the regression toward the mean of the BPs for the after-one-year reexamlned children. It was found that those children Initially In the top ten percentiles had, upon reexamlnatlon, on the average only 3 mmHg lower systolic and 1 mmHg lower diastolic levels. In a multiple regression analysis, the previous year's BP contributed a partial correlation coefficient of 0.6-0.7 for each age cohort to the variability of the BP, controlling for other determinants. These findings, based on reliable, basal-llke measurements, point to a degree of persistence which Is quite high. The higher the degree of tracking the more likely that primary hypertension begins early in life. blood pressure; hypertension; pediatrics; school health
The variations of blood pressure (BP) levels in children over time need to be elucidated. The nature of such changes is not known, but it is possibly related to the development of essential hypertension, Although it has been stated (1, 2) that essential hypertension is rare in children, there is evolving evidence (3, 4) that essential hypertension begins in childhood,
It may also be assumed that prevention of hypertension would be more successful if the disease process were aborted early, Hence, for early diagnosis we need to know the time course of BP levels and whether "tracking" occurs (i.e., to what extent a child with relatively high BP levels will persist with relatively high levels), especially since the BP level dur-
Received for publication February 2, 1978, and in final form July 5, 1978. Abbreviations: BP, blood pressure; H, height; Hg, mercury; N, sample size; SE, standard error of the mean; W, weight. 1 From the Department of Public Health and Preventive Medicine and the Department of Medicine, Louisiana State University Medical Center, New Orleans, LA 70112. Address reprint requests to: Dr. G. S. Berenson, Department of Medicine, Louisiana State University Medical Center, 1542 Tulane Ave., New Or-
leans, LA 70112. Supported by funds from the National Heart, Lung, and Blood Institute of the USPHS, Specialized Center of Research-Arteriosclerosis (SCOR-A), HL15103. The authors are indebted to Ms. Imogene W. Talley for her work as a community coordinator; Rita B. Clayton, R.N., who read all of the physiometrics discs; the entire Bogalusa Heart Study staff (espedally the blood pressure observers) for their measurement assistance; and Ms. Leanne Warren for computations.
320
TIME COURSE STUDIES OF BLOOD PRESSURE IN CHILDREN
ing adolescence and young adulthood has been found to be predictive of later hypertension (5-9). Conventional impressions (10-14) emphasize the lability of BPs at the early phases of hypertension. However, these early phases of hypertension are often defined by casual BP measurements, which are less predictive of future hypeiicnsion than basal pressures (5, 6). Lability here may merely mean that the measured pressures of individuals exceed defined abnormal cutpoint levels more often, either because their mean is close to this cutpoint or because of increased variability of these BPs. In the first case they are not truly labile, and in the second they are not truly hypertensive. To help clarify these points, data are needed from an entire geographic population of free-living children rather than from hospital-based patients and nonrepresentative selected individuals; data obtained by standardized methodology are also required. Pertinent information is now available from the community-wide examination of school aged children as performed in the Bogalusa Heart Study. Little is known about the persistence as well as determinants of high ranking BP levels in children. If a child has a BP level which ranks high, how likely will this ranking persist after an hour, a month, a year? What causes basal BPs to persist or to change over the short term? Answers to these questions will help begin our understanding of the early onset of essential hypertension. MATERIALS AND METHODS
Population. The eligible population consisted of all children, ages five through 14 years, residing in Ward 4 of Washington Parish (County), Louisiana, including the town of Bogalusa. Ninetythree per cent participated in the study, a total of 3524 children. In one of the participating schools, 69 children in the fifth grade were reexamined for BP once a month for eight
321
consecutive months. Thirty-five of these children attended all eight sessions. Children 5, 8, 11 and 14 years of age during the first examination in the study were reexamined after one year. A total of 1101 children (83 per cent of the 1326 eligible) were examined twice. General examination procedures. Because the objective of the screening procedures was to assess risk factor variables for coronary artery disease and hypertension, the following variables were measured: blood pressure, height, weight, triceps skinfold thickness, serum lipids (total cholesterol, triglycerides, lipoprotein fractions), and hemoglobin. Each child also received a physical examination which included external maturation measurements (secondary sex characteristics as described by Tanner (15)). All measurements followed methods outlined in a detailed protocol (16—19). BPs were recorded as the last procedure of the screening process which lasted from one and a half to two hours. Each morning at the end of the examination, a random sample of four children was reexamined by the same examiners. The selection of this sample was unknown to the examiners prior to reexamination. Blood pressure instruments. Indirect BPs were obtained by the mercury sphygmomanometer (Baumanometer", W. A. Baum Co., Inc., Copiague, NY) and by the Physiometrics" automatic blood pressure recorder (Sphygmetrics, Inc., Woodland Hills, CA). The sphygmomanometer selection of the cuff bladder sizes was based on criteria for arm measurements (18), which generally designated maximal bladder size while leaving the elbow skin crease free for the stethoscope. The PhysiometricsR instrument amplifies infrasonic sounds detected by an electronic transducer and records on a paper disc rotated by an aneroid manometer in open communication with an over-sized rubber cuff bladder entirely encircling the upper arm. The transducer
322
VOORS, WEBBER AND BERENSON
is built into the cuff bladder for maximal sound amplitude. Amplification covers the 10-80 hertz frequencies and maximizes Korotkoff phase discrimination. Blood pressure measurements. The children were encouraged to urinate before BPs were measured, and local volunteers guided them through a designated random sequence for examination. For each child the BP was determined in a randomized sequence by three trained observers; each observer independently measured three BPs per child (figure 1). The total BP assessment of a child during this examination did not last more than 30 minutes. Two observers each used a well-illuminated mercury sphygmomanometer placed at eye level and the
Children Enter in Random Sequence
third a PhysiometricsR recorder; each instrument was assigned randomly on a daily basis, as were the observers to the instruments. Each day before use the PhysiometricsR instrument was calibrated with the mercury column of a sphygmomanometer. Under quiet conditions, in a screened-off corner, observers recorded the first and fourth Korotkoff phases to determine the right arm BP while the child was seated. Care was taken to cover the brachial artery with the center of the rubber bladder. Physiometrics" discs were read blindly by the same individual throughout the entire study period. After lengthy protocol development and pretesting, all BP observers (registered and licensed practical nurses) were
Random: -Instrument Sequence (Child) -Observer Allocation Among 6 Instruments (Daily)
10% Random* Sample Rescreened Daily
Finish
1
Randomized 1 Allocation I
FIGURE 1. Pattern of flow for examining children for BP in the Bogalusa Heart Study, 1973-1975. Hg = mercury gphygmomanometer, P" = Physiometrica" automatic BP recorder.
323
TIME COURSE STUDIES OF BLOOD PRESSURE IN CHILDREN TABLE 1
Measured variables entered into the multiple regression equations, Bogatusa Heart Study, 1973-1974 Description Dependent variables Intra-child standard deviation of 6 systolic pressures by mercury sphygmomanometer Intra-child standard deviation of 6 diastolic (4th phase) pressures by mercury sphygmomanometer Independent Variables Blood Pressure Mean of 3 systolic pressures by Physiometrics" Mean of 3 diastolic pressures by Physiometrics" Blood pressure examiners nos. 1-8* Demographic Age (exact calendar age) Race* Sex* Athropometric Body height Ponderosity indext = weight/height3 Upper-arm circumferencet Triceps skinfold thicknesst External maturation index (pubic hair score plus breast development or male genitalia score) Laboratory Serum total cholesterolt Serum triglyceridest Serum prebeta lipoproteint Serum beta lipoproteint (Serum prebeta lipoprotein) + (Serum beta lipoprotein)t (Serum prebeta lipoprotein) / (Serum prebeta + beta liproprotein)t Blood hemoglobin
Units of measurement
mmHg mmHg
mmHg mmHg yes = 1, no = 2 years white = 1, black = 2 male = 1, female = 2 cm kg/cm3 cm mm Score range 2-10 mg/dl mg/dl mg/dl mg/dl mg/dl
g/dl
* Entered as dummy variables. t Logarithmic transformation.
trained, tested and retrained at regular variation still rises. This confounding, if intervals (four times annually) during the it occurs, will, therefore, tend to create a spuriously high variability at both ends of entire period of investigation. the observed (not of the true) BP range in Analysis the population (20). In our analysis, we Relation between variability and level of used the six sphygmomanometer meaBP. When the BP of a child is measured surements to compute the variability, the mean of the three Physseveral times, any gross deviation in whereas 11 iometrics measurements, taken inmeasurement from the true level will independent of the sphygmomanometer fluence both the mean and the variation of the child's measured BP. For instance, readings, were used to assess the child's if the variant measurement overstates mean level of BP. The latter instrument the true pressure, both the mean and the may have produced BP levels more valid variation will rise, while a variant mea- than the sphygmomanometer (18), since surement which understates the true larger cuff bladders were used (21), inpressure will lower the mean while the frasonic amplification was applied (22),
324
VOORS, WEBBER AND BERENSON
and permanent records were produced that could be read without observer bias (23). Determinants of BP variability. Individual variability of BP (expressed as the standard deviation about the six sphygmomanometer readings) was entered as the dependent variable into a multiple regression analysis, and all other meaTABLE 2
Intra-child correlation coefficient of blood pressure (mercury sphygmomanometer) over monthly time intervals, Bogalusa Heart Study, 1973-1976* Time interval (months)
Systolic
Diastolic «th phaae)
1 2 3 4 5 6 7
0.81 0.84 0.84 0.80 0.83 0.78 0.80
0.65 0.50 0.47 0.58 0.60 0.47 0.44
* Thirty-five children in the fifth-grade of an elementary school.
sured characteristics, including individual mean BP levels, were entered as independent variables. Logarithmic transformation was applied to the independent variables (except age, height, maturation, hemoglobin, and the nominal characteristics) in order to approach Gaussian distributions. A ponderosity index (weightyheight3) was chosen as a measure of obesity for theoretic and morphologic reasons (19). This ponderosity index was used instead of the body weight as an independent variable in order to avoid artifacts resulting from the highly positive correlation between height and weight (table 1). Regression of initially high BP toward the mean. In order to compute the amount of regression toward the population mean BP as experienced by a child above the 90th percentile for his age, we used the statistical model of Gardner and Heady (24), which requires knowledge of the inter-child and intra-child variability.
TABLB 3
Intra-child correlation coefficient of blood pressure (mercury sphygmomanometer) over a one-year interval, by age-cohort, race, and sex, Bogalusa Heart Study, 1973-1975 Phase and age (years) at first examination
Systolic 5 8 11 14 Diastolic (4th phase) 5 8 11 14 ' Sample size in parentheses.
Girls
Boya White
Black
White
Black
0.64 (81)* 0.73 (90) 0.61 (92) 0.62 (81)
0.61 (43) 0.76 (60) 0.68 (72) 0.61 (53)
0.61 (74) 0.63 (88) 0.68 (100) 0.62 (60)
0.70 (35) 0.80 (52) 0.71 (69) 0.72 (51)
0.32 (81) 0.61 (90) 0.44 (92) 0.48 (81)
0.46 (43) 0.48 (60) 0.42 (72) 0.48 (53)
0.40 (74) 0.60 (88) 0.55 (100) 0.47 (60)
0.60 (35) 0.15 (52) 0.55 (69) 0.61 (51)
TIME COURSE STUDIES OF BLOOD PRESSURE IN CHILDREN
325
Relation between variability and level The inter-child variability was obtained from the same population for which we of BP at initial examination developed percentile charts (25). The For each child the intra-child standard intra-child variability (26) was computed deviation of the six systolic and diastolic by pooling the intra-child variances for sphygmomanometer pressures was comthe 35 fifth-graders examined monthly. puted. Mean (± 2 SE) intra-child stanTracking ofBP in children originally in dard deviations of systolic sphygmothe extreme deciles. In the study popula- manometer pressures are presented by tion with two complete examinations one intervals of 5 mmHg Physiometric8R presyear apart, children in the upper and sure level in figure 2. There was no statislower decile of BP (mean of six sphyg- tically significant relationship between momanometer measurements) for the ini- these two variables (test for linear regrestial examination were identified and their sion of standard deviation on level by BPs (mean of six sphygmomanometer analysis of variance procedure) at the measurements) were observed after one 0.05 level. Mean (± 2 SE) intra-child year. After reducing the regression to- standard deviations of diastolic sphygward the mean to zero in a statistical ad- momanometer pressures are also prejustment, the average reexamination sented in figure 2. Here, the association is pressures were compared to the original negative (test for linear regression of upper and lower decile pressures in order standard deviation on level by analysis of to obtain quantitative insight into the variance procedure, p < 0.0001). amount of tracking of the BP. Determinants of BP level. The deterDeterminants of BP variability minants of systolic and diastolic BP level at initial examination upon reexamination were ascertained by Time order in which measurements stepwise multiple regression analysis in were taken. Means for the first three of the which BP was the dependent variable. Innine measurements taken from each child dependent variables were all currently measured characteristics plus the BP are presented in table 4 by instrument. For systolic readings, the difference belevel obtained one year earlier. tween first and second measurements was larger than between the following meaRESULTS surements. However, the total of differences was not much more than 2.0 mmHg Bivariate correlations systolic and 0.3 mmHg diastolic for the Bivariate Pearson coefficients of corre- sphygmomanometer; and, respectively, lation between mean sphygmomanometer not much more than 2.8 and 1.0 mmHg pressures for various time intervals in the for the Physiometrics" instrument. The monthly reexamination data are shown in smallness of these differences perhaps table 2. The correlation coefficients ap- supports the notion that basal pressure proximated 0.80 for systolic and 0.50 for levels were approached. diastolic pressures. For the mercury sphygmomanometer, The annual reexamination data are differences between mean measurements broken down by age, race and sex in table by first and second station (instrument) 3. The correlation coefficients are approx- are distributed normally (figure 3). imated at 0.70 for systolic and 0.50 for Profile of determinants of BP variabildiastolic pressures. No consistent differ- ity. In order to obtain a profile of deterences in the correlation coefficients by minants of BP variability, we analyzed race and sex were observed. systolic and diastolic variability sepa-
326
VOORS, WEBBER AND BERENSON
SYSTOUC
I f
70
I
80 90 100 110 120 130 140 150 >I55 LEVEL of Blood Pressure, (mmHg) DIASTOLIC (4th phase)
§
30
40 50 60 70 80 90 >95 LEVEL of Blood Pressure (mmHg)
FIGURE 2. Variability of BP within one half hour. Intra-child standard deviation of systolic and diastolic (4th phase) BP (mean ± 2 SE) by level of systolic BP. The standard deviation was ascertained for each child by six mercury sphygmomanometric measurements. The level was ascertained for each child by three Physiometrics" measurements. Children, ages 5-14 years, the Bogalusa Heart Study, 1973-1974.
327
TIME COURSE STUDIES OF BLOOD PRESSURE IN CHILDREN
rately in a multiple stepwise regression model, entering BP variability as the dependent variable and all other measured characteristics as independent variables (table 1). The results are given in table 5. The systolic variability was found to be I20CK SYSTOLIC 1000
-30 -20 -10 0 +10 +20 +30 First Minus Second Station, mmHg FIGURE 3. Frequency distribution of difference in mean BP between first and second mercury sphygmomanometer station. Children, ages 5-14 years, the Bogalusa Heart Study 1973-1974.
negatively associated with age and positively with the index obesity (W/H3) but not with systolic level itself. The diastolic variability was found to be negatively associated with body height and weakly positively associated with maturation, but not with the diastolic level itself. All measured characteristics combined could explain only one per cent of the variation in systolic pressure variability and two per cent of the variation in diastolic pressure variability. Regression of initially high BP toward the mean The pooled intra-child standard deviation as obtained from the monthly reexamined fifth-graders was 4.1 mmHg systolic and 5.1 mmHg diastolic. Applying these results to the cross-sectional standard deviation for 11-year-olds as obtained from the initial examination of the entire population (9.2 mmHg systolic and 7.6 mmHg diastolic), the inter-child standard deviation can be computed by taking the square root of the difference in variance. The resulting inter-child standard deviations for 11-year-olds in the Bogalusa Heart Study are 8.3 mmHg systolic and 5.6 mmHg diastolic. Assuming that the BP values have Gaussian distributions, the model of Gardner and Heady (24) can be used to
TABLE 4
Mean blood pressures (mm Hg) of first, second, and third measurement in children age 5—14 years, by instrument, Bogalusa Heart Study, 1973-1974
Measurement
First Second Third All other
Mercury spbygmomanometer (JV - 2362*) Diastolic Systolic (4th phase)
Physiome tries" (W - 1164*) Diastolic Systolic (4th phase)
101.7t lOO.St 100.2t 99.7§
103.lt 101.3+ 100.3+ 100.3"
62.5 62.7 62.6 62.4§
62 6+ 62.1tt 61.8t 61.6"
* Eight missing values excluded from above totals. t Difference between indexed figures in same column is significant, p < 0.001. t Difference is significant, p < 0.05. § Not tested; averages of three mercury sphygmomanometric and three Physiometrics" measurements. 1 Not tested; averages of six mercury sphygmomanometric measurements.
328
VOORS, WEBBER AND BERENSON TABLE 5
Stepwise multiple regression of intra-child blood pressure variability, Bogalusa Heart Study, 1973-1974 Systolic Regression coefficient
Intercept Age W/H»t Observer 4 Observer 3
13.85 -0.05t 1.94§ OJ221
035'
Diastolic (4th phase) Standardized partial regression coefficient
-0.076 0.063 0.052 0.048
Regression coefficient
Intercept Height Observer 2 Observer 6 Observer 8 Maturation
7.36 -0.02$ 0.67$ 0.62t 0.43 § 0.081
Standardized partial regression coefficient
-0.132 0.111 0.102 0.072 0.073
N.B. Multiple correlation coefficient squared is 0.01 for systolic pressure and 0.02 for diastolic pressure. * 197 out of 3524 children not analyzed due to missing data; for variables entered, see table 1. tLog,* tp < 0.0001. §p < 0.001. »p < 0.01.
obtain the average decrease in pressure upon reexamination as experienced by a child ranking above the 90th percentile during the first examination. This decrease is 2.3 mmHg systolic and 4.4 mmHg diastolic (Appendix). These results are based on means of six sphygmomanometer measurements on a child per examination. Tracking of blood pressure for children originally in the extreme deciles In the study population with two complete examinations one year apart, children in the upper and lower decile of BPs (sphygmomanometer) for the first examination were identified and their BPs were observed during the second examination. Also, adjusted levels from the upper and lower deciles of the original examination were calculated after taking regression toward the mean into account. The latter are values statistically adjusted by reducing the regression towards the mean to zero (table 6). After adjustment for the annual difference in mean systolic and diastolic pressure, it was found that those children initially in the top 10 percentiles would have had average reexamination pressures corresponding to a decrease of 3.4 mmHg
systolic and 0.6 mmHg diastolic pressure, whereas those initially in the lower 10 percentiles would have had corresponding increases of 3.2 mmHg systolic and 2.0 mmHg diastolic pressure (figure 4). These mean shifts of BP levels for those children in the extreme deciles are perhaps within a five percentile range for systolic, and within a two percentile range for diastolic levels. Determinants of BP Level
The strongest determinant of BP level was the level observed one year previously (table 7). In total, all measured variables could account for 39-55 per cent of the systolic, and 32-38 per cent of the diastolic mercury sphygmomanometric pressure variation. A measure of tracking is the partial correlation coefficient of first-year with second-year BP, controlling for all other significant independent variables. This coefficient was 0.61-0.66 for systolic and 0.36—0.52 for diastolic pressure. In the regression analysis, the ponderosity index W/H3 correlates only for the two younger age groups, and the external maturation index only for the 12-year age group. Results for the Physiometrics" instrument were quite similar.
329
TIME COURSE STUDIES OF BLOOD PRESSURE IN CHILDREN
DISCUSSION
+1 +1 +1 +1 a>
a CO
ab
TOT
87. 89. 95.
00
o o
CO
The measurement of valid and replicable indirect BP in children is fraught with pitfalls, but it is apparently obtainable when attention is paid to methodology. In our study, we attempted to arrange for nearly optimal conditions under which to measure the children's BPs. By aiming for such conditions, we increased the predictive value of our measurements (5, 6) by obtaining values which approach known resting basal values for children in this age group (18). The measurement replicability was increased by measuring the BP three times by each of three different observers, and by using the 4th rather than the 5th Korotkoff phase for
1—1
CO CO CN CO
+1 +1 +1
•H
CN CD CO
51. 52. 59.
CO CN CO CD CN CN CN CN
c- o rto *
t>
CN
r-H
+1
+1
00 f-t (N CO
i-l CN
i-H
r>
CN
CN
CO
(M +1 +1
CO
105 110 114 119
CO
a>
CN CO
+1 +1 +1
c-
CD
f—I
i—i
1-1
»H
r-
i-H O
i-H
o
I 1
i—i
CN .-1 ao 00 I N i-l CD on
no O5 KX>
it
5
c o
90
80
CN
+1 +1 +1 +1 +1+1 on CO
115
+1 1—1
tI>
SYSTOLIC
120
.2
'
T3
7O-
r-t--t-t-
65
60-
Z2
cn
CN
-f
CN
CO CN
55-
CO lO CN • * o5 CN co (N
5O9
3
COCn
