Journal of Human Hypertension (2002) 16, 327–332
 2002 Nature Publishing Group All rights reserved 0950-9240/02 $25.00
www.nature.com/jhh
ORIGINAL ARTICLE
Assessing resting heart rate in
adolescents: determinants and correlates
F Rabbia, T Grosso, G Cat Genova, A Conterno, B De Vito, P Mulatero, L Chiandussi
and F Veglio
Department of Medicine and Experimental Oncology, University of Turin, Italy; Cattedra di Medicina
Interna, Ospedale S. Vito, 10133 Turin, Italy
The aim of this study was to evaluate the distribution of
resting heart rate and its biological and environmental
determinants in adolescents. The study was crosssectional and the population consisted of 2230 children
and adolescents, age range 12–18 years, enrolled randomly from state schools in Turin, Italy. In all participants the following parameters were evaluated: heart
rate, blood pressure (BP), weight, height, degree of sexual development, physical activity, parental socio-cultural level. Heart rate and BP were measured after 5, 10
and 15 min in a sitting position. Furthermore, to obtain
regression equations to define heart rate as a function
of the other variables available, a multiple regression
analysis was performed. In both sexes BP, but not heart
rate, declined significantly from the first to the last
determination. Heart rate was positively and significantly correlated to BP level in both sexes; heart rate
was higher in girls (3 bpm) and followed a progressive
decreasing trend with age in both sexes, that was
opposite to BP values. Age, sexual maturation, height,
physical activity and parental socio-cultural level were
independent determinants of resting heart rate. In conclusion, resting heart rate in adolescents is related to
several methodological, constitutional and environmental factors that have to be taken into account when
assessing heart rate values and constructing tables of
normal values.
Journal of Human Hypertension (2002) 16, 327–332. DOI:
10.1038/sj/jhh/1001398
Keywords: heart rate; adolescence; risk factors; blood pressure
Introduction
In recent years evidence has been accumulating that
heart rate (HR) and blood pressure (BP) are positively correlated at all ages and in both genders,1–3
even after adjustment for the usual confounders.
Fast HR also predicts future hypertension in young
subjects with normal or borderline elevated BP
values.4–6 Moreover, HR has been associated with an
increased risk of cardiovascular and non-cardiovascular deaths in adult populations.7–10 The most
important determinant of HR is parasympathetic
modulation of the heart, combined with sympathetic
efferent influences. Several studies on experimental
animals, healthy humans and patients with cardiovascular diseases have shown that sympathetic activation and inactivation are usually associated with
increases and reductions in the HR.11–13
HR is widely regarded as a simple, although
aspecific, marker of sympathetic activity that can
have clinical applications. Nonetheless, physicians
Correspondence: Dr F Rabbia, Cattedra di Medicina Interna,
Ospedale S. Vito, Strada S. Vito, 34 –10133 Torino, Italy.
E-mail: centroiperten.torino얀libero.it
Received 8 November 2001; revised and accepted 10 January
2002
tend to underestimate this evidence and the HR is
either ignored or viewed as a benign prognostic sign.
One of the most important reasons for the underestimation of the importance of HR, is that HR
behaviour is highly variable, it may be influenced
by several constitutional and environmental factors.
HR measurement has not yet been standardised. In
fact, no specific guidelines have been given for the
assessment of resting HR, consequently its value
may be affected by the different methods of
measurement,14,15 and the adherence to a detailed
protocol is necessary to obtain individual values
that can be compared with population grids. In adolescence, the assessment of resting HR may be even
more difficult due to the particular dynamic of
growth in this period,16 and at present few data
exist.17–20
This paper is aimed at the examination of the biological and environmental determinants of resting
HR in adolescents. From this analysis the factors
that should be considered when assessing HR levels
in this particular period of life are determined.
Heart rate in adolescents
F Rabbia et al
328
Materials and methods
Study population
The study population consisted of 2230 students
(1204 females and 1026 males, 92% of those
recruited), age range 12–18 years. All children were
enrolled in Turin state junior high, high and
vocational schools, choosing one kind of school for
each of the 23 town areas, at random. Two classes
from each school were randomly selected and the
students evaluated. The population studied consisted of a large percentage of Caucasian children
(96%) and 4% of African and Asian, for this reason
they were analysed together.
The study was approved by the local education
authority. Informed consent was obtained from the
students and from a parent or a legal guardian.
Measurements
Two medical teams conducted the clinical examination in the morning, during routine school activity.
Procedures were fully explained to the children,
who were studied in groups of three to five, to minimise the anxiety level caused by the medical examination.
We evaluated: HR, BP, weight, height, body mass
index (BMI), degree of sexual development, physical
activity, parental socio-cultural level. BP was
recorded according to the Update on the II Task
Force guidelines,21 ie with the child in a comfortable
sitting position with the right arm fully exposed and
resting on a supportive surface at the heart level; a
mercury sphygmomanometer was used with an
appropriately sized cuff.
At the same time, three HR measurements (radial
pulse) were taken after 5, 10 and 15 minutes of being
in a sitting position and the mean was calculated.
HR was recorded by a physician over a 1-min period.
Current weight was measured with students without clothes and bare-footed, using an electronic
scale.
Height was measured with students bare-footed
and using a right angle ruler placed upon the head,
against a tape measure secured to the wall.
Degree of sexual development was calculated,
after physical examination, by using Tanner indexes
of sexual development. For practical reasons, children were classified into three groups: class A (low
sexual development) corresponding to Tanner Stage
1; class B corresponding to Tanner 2 and 3; and class
C (complete sexual development) corresponding to
Tanner stage 4 and 5.22
Physical fitness was calculated as weekly hours
of sport or play activity, including school-time gym
classes (in Italy, usually less than 2 h per week).
Parental educational attainment was considered
as an index of the socio-cultural level choosing the
highest degree of the two parents and creating four
groups: group 1 = no educational qualification or
Journal of Human Hypertension
primary school; group II = junior high school; group
III = senior high school; group IV = graduation.
Data analysis
Statistical analysis was performed using Statistical
Analysis System Package (SAS).23
Means, standard deviations (s.d.) of descriptive
statistics were calculated. BP and HR data were
presented at first (5 min), third (15 min) and the
mean of the three measurements in order to test for
the presence of alarm reaction.
Pearson linear correlation was used to analyse the
relationship between continuous variables. Multiple
regression analysis was performed to explain HR as
a function of the biological and environmental variables evaluated. Tanner score, sex and socio-cultural
level were considered as dummy variables.
Results
Table 1 shows the general features of the population
studied. Males were significantly taller, heavier and
less sedentary than females and they had significantly higher systolic BP (SBP) values and slower
HR than females. Tables 2 and 3 show HR and SBP
values obtained at first and third measurement and
the mean of the three measurements by gender and
age (diastolic BP (DBP) values followed a similar
pattern to SBP and have not been shown). HR progressively decreased with age up to the age limit
included in the study, in both sexes. At the same
time BP followed a quite opposite trend. HR measured at 5 min (first measurement) was significantly
higher than that measured at 15 min (last
measurement) only in one age group (17 years old),
whereas BP measured at 5 min was always signifiTable 1 General characteristics of the study population, by
gender
Age (years)
Height (cm)
Weight (kg)
BMI (kg/m2)
Physical activity (hours/week)
SBP (mm Hg)
DBP (mm Hg)
HR (beats/min)
Tanner
Stage 1
Stage 2
Stage 3
Socio-cultural level
group I
group II
group III
group IV
Males
(n = 1026)
Females
(n = 1204)
14.9 ± 1.9
165.7 ± 11.7
57.1 ± 12.5
20.6 ± 3.0
5.2 ± 3.2
115.5 ± 11.3
69.8 ± 9.5
77.5 ± 8.9
15.3 ± 2.0
159.9 ± 7.4*
51.7 ± 8.3*
20.2 ± 2.7
3.8 ± 2.3*
112.7 ± 10.3*
69.8 ± 9.2
79.4 ± 9.6*
7%
31%
62%
5%
18%
77%
23%
39%
29%
9%
18%
36%
34%
12%
Values are expressed as mean ± standard deviation. *P ⬍ 0.05,
males vs females.
Heart rate in adolescents
F Rabbia et al
Table 2 Values of heart rate at first and third measurement and overall mean of three measurements in males (M) and females (F)
No. of subjects
First measurement
Third measurement
Mean ± s.d.
12 M
12 F
50
72
81.1 ± 10.4
84.9 ± 13.7
80.1 ± 8.6
81.1 ± 10.0
80.5 ± 7.6
82.7 ± 10.5
13 M
13 F
299
283
80.9 ± 10.6
83.9 ± 12.1
79.3 ± 9.7
82.3 ± 9.5
80.0 ± 9.0
82.9 ± 9.2
14 M
14 F
301
299
78.6 ± 10.1
82.4 ± 11.7
78.2 ± 9.1
81.2 ± 10.1
78.3 ± 8.4
81.7 ± 9.2
15 M
15 F
89
56
76.2 ± 10.7
80.4 ± 10.3
76.1 ± 9.3
79.4 ± 7.4
76.4 ± 8.7
80.6 ± 7.9
16 M
16 F
100
160
77.9 ± 11.8
77.3 ± 8.3
75.1 ± 10.0
76.1 ± 8.1
76.7 ± 10.6
76.7 ± 8.1
17 M
17 F
116
186
76.8 ± 8.2
77.4 ± 9.7
74.3 ± 6.8*
75.3 ± 8.7
75.7 ± 7.1
76.4 ± 8.7
18 M
18 F
71
148
75.0 ± 8.0
75.9 ± 9.7
73.8 ± 7.0
74.1 ± 7.5
74.5 ± 7.4
75.1 ± 8.3
Age (years)
329
Values are expressed as mean ± standard deviation. *P ⬍ 0.01 vs HR at 5 min.
Table 3 Values of SBP at first and third measurement and overall mean of three measurements in males (M) and females (F)
No. of subjects
First measurement
Third measurement
Mean ± s.d.
12 M
12 F
50
72
115.1 ± 12.3
114.2 ± 11.9
108.9 ± 10.8*
105.9 ± 10.2*
111.7 ± 10.6*
109.7 ± 10.1*
13 M
13 F
299
283
115.6 ± 12.1
115.1 ± 12.5
108.2 ± 11.3
107.3 ± 10.3*
111.2 ± 10.7
110.6 ± 10.2*
14 M
14 F
301
299
121.2 ± 12.9
115.6 ± 13.3
112.1 ± 11.1*
108.4 ± 10.9*
116.1 ± 11.0*
111.5 ± 10.9*
15 M
15 F
89
56
120.1 ± 11.7
117.4 ± 14.0
111.1 ± 11.3*
110.0 ± 12.2*
115.2 ± 10.8*
113.0 ± 12.9*
16 M
16 F
100
160
128.1 ± 19.3
115.8 ± 13.5
118.9 ± 10.4*
112.7 ± 11.0*
122.2 ± 12.4*
113.8 ± 11.7
17 M
17 F
116
186
124.2 ± 11.3
117.6 ± 12.2
119.6 ± 8.9*
113.9 ± 9.1*
121.6 ± 9.7*
115.4 ± 9.9*
18 M
18 F
71
148
120.1 ± 9.1
115.4 ± 10.8
118.4 ± 7.6*
112.9 ± 8.4*
119.0 ± 8.0
113.8 ± 9.2
Age (years)
Values are expressed as mean ± standard deviation. *P ⬍ 0.01 vs SBP at 5 min.
cantly higher than at 15 min in all groups and in
both sexes.
Figure 1 shows the correlation coefficients
between HR and BP by age and gender. HR is
directly and significantly related to BP at all ages
and in both sexes.
Table 4 shows the multiple regression equation
prepared to predict HR. Sex (females have 3 bpm
more than males), weight but not height (0.1 bpm/kg
of weight), age (−1.1 bpm/year of age), physical
activity (−0.4 bpm/hour of activity), Tanner score
and socio-cultural level were independently and
inversely associated to HR values.
Discussion
We conducted a cross-sectional study in order to
evaluate factors affecting resting HR in adolescent
Figure 1 Correlation coefficients between heart rate and blood
pressure by age and gender.
Journal of Human Hypertension
Heart rate in adolescents
F Rabbia et al
330
Table 4 Multiple regression analysis for heart rate (HR) dependent variable: HR. R2 = 0.1532
Independent variables
Intercepta
Age (years)
Height (cm)
Weight (kg)
Physical activity (hours/week)
Parental socio-cultural level
1
2
3
Tanner
1
2
Sex
F
Beta
Standard
error
P⬍
130.7
−1.143
−0.027
0.066
−0.406
6.605
0.324
0.039
0.031
0.086
0.0001
0.0003
0.499
0.0311
0.0001
2.487
2.655
1.053
1.127
1.049
1.050
0.027
0.011
0.316
5.111
2.155
1.028
0.592
0.0001
0.0003
3.027
0.525
0.001
Regression equation for females: HR = 130.7–1.143 age + 3.027
sex + 0.066
weight − 0.027
height − 0.0406
sport + Tanner
score + cultural level. Cultural level 4, Tanner stage 3 and male
gender were considered as reference variables.
age. The sample studied consisted of children from
puberty to the end of adolescence. Our results
showed that HR measurement is associated to a
longer white coat effect than BP. It follows an
inverse pattern throughout adolescence in comparison with, and it is positively correlated to, BP. HR
is higher in girls and is independently associated to
somatic growth indexes, physical activity and sociocultural level.
The HR is highly variable, and in adolescents this
variability may be even more increased than in adult
age. We have considered two sources of variability:
the first is the method of measurement. In particular
some methodological sources of variability may be
very important: the position of the body, the resting
time before the measurement, the duration of the
measurement, the way of recording. Since there
were no criteria for the measurement of HR, we customised the well standardised protocol for BP
measurement of the second Task Force on BP evaluation in children,21 ie children in a sitting position,
three measurements after 5 min of resting. Moreover
we considered duration of measurement of 1 min.
The sitting position is more practical than the
supine one, even though venous pooling of blood in
the lower extremities, while the patient is sitting,
causes a decrease of the sympatho-inhibition
exerted by the cardiopulmonary reflex. Thus, a
higher HR is generally expected in sitting position.
About the resting time before the measurement, in
the majority of the studies the HR was measured
after a 5 min rest, whereas a longer time may favour
a reduction in the alarm reaction.24 We tried to
assess this methodological problem, by comparing
the measurement at 5 and at 15 minutes of both HR
and BP. In comparison with BP, there was only a
little tendency of HR to follow a specific reduction
pattern over time. This pattern confirms the eviJournal of Human Hypertension
dence from Mancia et al24 who demonstrated that,
in adults, a doctor’s visit is accompanied by an
immediate rise in BP and HR, which peak during
the early part of the visit and subsequently tend to
slowly decline. In children the decline of HR is less
evident than BP. This particular pattern may allow
some considerations: either there is no significant
white coat response, or if such a response exists that
it lasts for longer than 15 minutes. However, it is
not possible to distinguish the two alternatives from
the data presented.
HR was recorded over a 1-min period in order to
obtain greater precision, even if a 30-s reading seems
also a reasonable choice.14,19 Shorter periods seem
not to be acceptable. However, HR was not measured for shorter periods in our sample and the comparison is not possible. About the way of recording,
the use of radial pulse has been found to be a valid
method when compared with ECG measurement.25
The second important source of variability are the
physiological factors, in particular: sex, chronological age, body maturation and physical activity, as
for BP levels.16,26
As have been demonstrated in adults,8,9,27 HR is
higher in girls than it is in boys. This sex-related
difference has been found both in economically
advanced countries and in undeveloped countries.
In our population it is about 3 bpm. Gender differences are independent from body structure (height,
and sexual development) and from physical activity,
as shown in multiple regression analysis.
An inverse independent association of age to HR
through adolescence is well documented,17,19,20 as in
adults.27 As it has been demonstrated for BP,16,28
weight and sexual maturation are independent
determinants of resting HR values. For this purpose,
there is much evidence suggesting that growth-promoting processes stimulate both growth and the
maturation of the cardiovascular system thereby
raising vascular resistance and arterial pressure. One
of the purposed mechanisms is mediated via growth
hormone (GH) and insulin-like growth factor
(IGF1).29–31 The rise of GH and IGF1 is stimulated at
puberty by an increase of gonadal and adrenal sex
hormones.32–35 The close inverse relationship
between physical fitness and HR has been demonstrated in previous experimental36,37 and epidemiological studies38,39 at any age and in both sexes.
Physical activity can be regarded as a useful and
physiological method for decreasing HR, and its cardioprotective action could be partially explained
through the effects on HR.40
HR was also significantly associated to the degree
of parental education. It is difficult to explain this
association; it may be related to a higher degree of
economicaland social means that can lead, therefore, to higher physical activity and fewer social
stressors.
We compared our results with data from Bogalusa19 (conducted on a biracial population), Philadelphia17 (African American population) and
Heart rate in adolescents
F Rabbia et al
Table 5 Comparison heart rate (HR) values between our data and
other studies
5
The Bogalusa Heart Study: 50th percentile of resting HR by sex
and age
6
Age (years)
White males
White females
Afro-American males
Afro-American females
12
76
86
75
81
13
76
83
71
80
14
74
84
73
80
15
73
83
71
80
16
70
84
69
78
17
73
81
66
77
The Philadelphia Blood Pressure Project: 50th percentile of
resting HR by sex and age
Age (years)
Afro-American males
Afro-American females
12
74
76
13
74
78
14
72
78
15
72
75
16
68
74
8
17
68
79
The Chicago Heart Association Paediatric Heart Screening
Project: Mean ± sd of HR
HR (mean ± s.d.)
7
9
10
White
males
AfroAmerican
males
White
females
11
77.4 ± 13.1
76.4 ± 13.8
82.1 ± 13.0
12
13
Chicago20 (biracial population, and data expressed
only as mean ± s.d.). All the studies followed a different methodology for the assessment of resting HR,
nevertheless the three studies are comparable
(Table 5); we found that our population showed faster HR values than African American boys and substantially similar mean values with respect to white
children and African American girls.
In conclusion, a systematic quantitative approach
toward ascertaining resting HR may be an additional
parameter in the study of cardiovascular risk factors
in youth as in adulthood. However, although relatively easily measured, HR in adolescents shows an
important relationship to various physiological,
environmental variables that are to be taken into
account. Moreover, adherence to a detailed protocol
is necessary to obtain individual values that can be
reliably compared.
14
15
16
17
18
19
20
References
1 Stamler J, Berkson DM, Dyer A. Relationship of multiple variables to blood pressure: findings from four
Chicago epidemiologic studies. In: Paul O (ed.) Epidemiology and Control of Hypertension. Symposia
Specialists: Miami, 1975, pp 307–352.
2 Berenson GS et al. Racial differences of parameters
associated with blood pressure levels in children: the
Bogalusa Heart Study. Methabolism 1979; 28: 1218–
1228.
3 Simpson FO, Waal-Manning HJ, Boli P, Spears GF. The
Milton Survey 2. Blood pressure and the heart. NZ
Med J 1978; 88: 1– 4.
4 Selby JV et al. Precursors of essential hypertension:
pulmonary function, heart rate, uric acid, serum chol-
21
22
23
24
25
26
esterol, and other serum chemistries. Am J Epidemiol
1990; 131: 1017–1027.
Garrison RJ, Kannel WB, Stokes J III. Incidence and
precursors of hypertension in young adults. Prev Med
1987; 16: 235–251.
Kim JR et al. Heart rate and subsequent blood pressure
in young adults: The CARDIA Study. Hypertension
1999; 33: 640–646.
Greenland P et al. Resting heart rate is a risk factor for
cardiovascular and noncardiovascular mortality: the
Chicago Heart Association Detection Project in Industry. Am J Epidemiol 1999; 149: 853–862.
Gillman MW, Kannel WB, Belanger A, D’Agostino RB.
Influence of heart rate on mortality among persons
with hypertension: the Framingham Study. Am Heart
J 1993; 125: 1148–1154.
Gillum RF, Makuc DM, Feldman JJ. Pulse rate, coronary heart disease, and death: the NHANES I epidemiologic follow-up study. Am Heart J 1991; 121: 172–177.
Kannel WB, Kannel C, Paffenbarger RS, Cupples LA.
Heart rate and cardiovascular mortality: The Framingham Study. Am Heart J 1987; 113: 1489–1494.
Shepherd JT, Mancia G. Reflex control of human cardiovascular system. Rev Physiol Biochem Pharmacol
1986; 105: 1–99.
Julius S. Changing role of the autonomic nervous system in human hypertension. J Hypertens 1990; 8
(Suppl 7): S59–S65.
Mancia G et al. Reflex cardiovascular control in congestive heart failure. Am J Cardiol 1992; 69: 17G–23G.
Palatini P. Need for a revision of the normal limits of
resting heart rate. Hypertension 1999; 33: 622–625.
Palatini P. Office versus ambulatory heart rate in the
prediction of the cardiovascular risk. Blood Press
Monit 1998; 3: 173–180.
Lever FA, Harrap SB. Essential hypertension: a disorder of growth with origins in childhood? J Hypertens
1992; 10: 101–120.
Hediger ML et al. Resting blood pressure and pulse
rate distributions in black adolescents: the Philadelphia Blood Pressure Project. Pediatrics 1984; 74:
1016–1021.
Iliff A, Lee VA. Pulse rates, respiratory rate, and body
temperature of children between two months and
eighteen years of age. 1952; Child Dev 23: 237–245.
Voors AW, Webber LS, Berenson GS. Resting heart rate
and pressure-rate product of children in a total biracial
community. The Bogalusa Heart Study. Am J Epidemiol 1982; 116: 276–286.
Miller RA, Shekelle RB. Blood pressure in tenth-grade
students. Results from the Chicago Heart Association
Pediatric Heart Screening Project. Circulation 1976;
54: 993–1000.
Report of the Second Task force on blood pressure
control in children. Pediatrics 1987; 79: 1–25.
Tanner JM. Growth at Adolescence. Blackwell Scientific Publications. Oxford, 1962.
SAS Institute Inc. SAS/STAT User’s Guide. SAS Institute: Cary (NC). 1988.
Mancia G et al. Effects of blood pressure measurement
by the doctor on patients blood pressure and heart
rate. Lancet 1983; 2: 695–698.
Erikssen J, Rodhal K. Resting HR in apparently healthy
middle aged men. Eur J Appl Physiol 1979; 42: 61–69.
Rabbia F et al. Relationship between birth weight and
blood pressure in adolescence. Prev Med 1999; 29:
455– 459.
331
Journal of Human Hypertension
Heart rate in adolescents
F Rabbia et al
332
27 Morcet JF et al. Associations between heart rate and
other risk factorsin a large French population. J Hypertens 1999; 17: 1671–1676.
28 Lauer RM, Anderson AR, Beaglehole R, Burns TL. Factors related to tracking of blood pressure in children.
Hypertension 1984; 6: 304 –314.
29 Hall K, Sara VR. Somatomedin levels in childhood,
adolescence and adult life. J Clin Endocrinol Metab
1982; 54: 254 –260.
30 Stanhope R, Preece MA, Grant DB, Brook CGD. New
concepts of the growth spurt of puberty. Acta Pediatr
Scand Suppl 1988; 347: 30–37.
31 Rechler M, Nissley SP, Roth J. Hormonal regulatuion
of human growth. N Engl J Med 1987; 316: 941–942.
32 Merimee TJ, Fineberg SE. Studies of the sex based variation of human growth hormone secretion. J Clin
Endocrinol Metab 1985; 62: 159–164.
33 Link K et al. The effects of androgens on the pulsatile
release and the 24-hour mean concentration of growth
hormone in peripubertal males. J Clin Endocrinol
Metab 1985; 62: 159–164.
34 Wiedemann E, Schwartz E, Frantz AG. Acute and
chronic estrogen effects upon serum somatomedin
Journal of Human Hypertension
35
36
37
38
39
40
activity, growth hormone, and prolactin in man. J Clin
Endocrinol Metab 1976; 42: 942–952.
Ho Ky et al. Effects of sex and age on the 24-hour profile of growth hormone secretion in man: importance
of endogenous estradiol concentrations. J Clin Endocrinol Metab 1987; 64: 51–58.
Seals DR, Hagberg JM. The effect of exercise training
on human hypertension: a review. Med Sci Sports
Exercise 1984; 16: 207–215.
Kingwell BA, Jennings GL. Effects of walking and
other exercise programs upon blood pressure in normal subjects. Med J Aust 1993; 158: 234 –238.
Blair SN, Goodyear NN, Gibbons LW, Copper KH.
Physical fitness and incidence of hypertension in healthy normotensive men and women. JAMA 1984; 252:
487– 490.
Paffenbarger RS Jr, Jung DL, Leung RW, Hyde RT.
Physical activity and hypertension: an epidemiological review. Ann Intern Med 1991; 23: 319–327.
Somers V, Conway J, Johnston J, Sleight P. Effects of
endurance training on baroreflex sensitivity and blood
pressure in borderline hypertension. Lancet 1991; 337:
1363–1368.