Published by Oxford University Press on behalf of the International Epidemiological Association
© The Author 2005; all rights reserved. Advance Access publication 15 April 2005
International Journal of Epidemiology 2005;34:905–913
doi:10.1093/ije/dyi071
Influence of short stature on the change in
pulse pressure, systolic and diastolic blood
pressure from age 36 to 53 years: an analysis
using multilevel models
Claudia Langenberg,1,3* Rebecca Hardy,2 Elizabeth Breeze,1 Diana Kuh2 and Michael EJ Wadsworth2
Accepted
8 March 2005
Background Previous cross-sectional analyses of this cohort have shown that short height and
leg length are associated with higher pulse pressure and systolic blood pressure
in middle age. It is unclear how these adult measures of childhood growth
influence the change in blood pressure as it increases with age.
Methods
Multilevel models were fitted to investigate associations between components
of height and the change in blood pressure between 36, 43, and 53 years in a
prospective national cohort of 1472 men and 1563 women followed-up since
birth in 1946.
Results
Shorter height and leg length, but not trunk length, were associated with higher
blood pressure, similarly in men and women. Longitudinal analyses showed that
the effects of both height and leg length on pulse pressure and systolic blood
pressure became significantly stronger with age. For example, the change in
systolic blood pressure was found to be 0.021 mm Hg (95% confidence interval
0.029 to 0.013) per year lower for every centimetre increase in leg length
(P 0.001). In other words, the increase in systolic blood pressure over a 10 year
period of a participant whose legs were 10 centimetres shorter was 2.1 mm Hg
higher (P 0.001), compared with a taller participant. Associations were
independent of a number of potential confounders.
Conclusions These results support the hypothesis that short people may be more susceptible
to the effects of ageing on the arterial tree. Childhood growth may contribute to
the tracking of cardiovascular risk throughout life.
Keywords
Body height, growth, blood pressure, pulse pressure, cohort study
Atherosclerotic changes of the arterial wall begin at an early
age, even in apparently healthy children and adolescents.1
Levels of blood pressure and other cardiovascular risk factors
in childhood persist over time and cluster both in childhood
and adulthood, influencing subsequent subclinical and clinical
1 Department of Epidemiology and Public Health, University College London
Medical School, 1-19 Torrington Place, London WC1E 6BT, UK.
2 MRC National Survey of Health and Development, Department of
Epidemiology and Public Health, University College London Medical
School, 1-19 Torrington Place, London WC1E 6BT, UK.
3 Department of Family and Preventive Medicine, University of California San
Diego, School of Medicine, 9500 Gilman Drive, La Jolla, CA 92093, USA.
* Corresponding author. Department of Epidemiology and Public Health,
University College London Medical School, 1-19 Torrington Place, London
WCIE 6BT, UK. E-mail: [email protected]
cardiovascular disease.2 This suggests early acquired risk tracks
into adulthood, but the origins of such risk remain unclear.
People of shorter height have an increased risk of cardiovascular
disease.3–5 The growth of the two main components of height,
leg and trunk length differs in timing and magnitude. Leg
length represents the growth of the long bones in the first years
of life and may be the component of height responsible for
the association between shortness and cardiovascular risk.6–10
Leg length is particularly sensitive to post-natal environmental
influences on growth.11,12 It has previously been suggested
that prepubertal growth rate is associated with the formation of
mechanisms controlling blood pressure in later life,13 and we
have previously shown strong, inverse associations between leg
length and both systolic blood pressure and pulse pressure, a
measure of arterial stiffness, in men and women of this cohort
905
906
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
at age 53 years.14 Arterial stiffness, pulse pressure, and blood
pressure increase with age and detrimental influences on
vascular structure and function in the first years of life may
increase vulnerability to the effects of ageing on the arterial
tree. If poor early growth contributed to the tracking of pulse
pressure and blood pressure, then those with shorter height
and shorter legs may experience a steeper increase in these
measures throughout life. This may differ between the genders,
due to the influence of hormonal levels on vascular function.
Earlier findings in this cohort were based on cross-sectional
analyses,14 which are unable to investigate the amplification of
the effect of short height on the change in blood pressure with
age. Longitudinal analyses may contribute to the understanding
of how exposures in early life interact with age to influence
blood pressure over the life course.
Several explanations have been suggested for the associations
between shortness, cardiovascular disease, and associated risk
factors, including poor prenatal growth, socioeconomic
disadvantage at different ages, and adverse health behaviour.
All these factors may contribute to the association between
short stature and increased blood pressure.
No previous study has investigated the influence of poor
childhood growth on the change in blood pressure during
middle age. Using adult leg and trunk length as markers of
growth at different phases of development, we compare
associations between components of height and change in pulse
pressure, and systolic and diastolic blood pressure from age 36
to 53 years in a prospective birth cohort study.
artery of the upper left arm after 5 min rest, with the participant
in the sitting position. At age 53 years blood pressure was
measured with the validated Omron HEM-705 (Omron Corp.,
Tokyo, Japan) automated digital oscillometric sphygmomanometer, and at 43 and 36 years the Hawksley random zero
sphygmomanometer was used. Second blood pressure readings
were used for analysis except if only the first was available.
Pulse pressure was calculated as the difference between systolic
and diastolic pressure at each age.
Methods
Birthweight
Birthweight was extracted, to the nearest quarter of a pound,
from medical records by health visitors within a few weeks of
delivery, and converted into kilograms.
Participants
The Medical Research Council’s National Survey of Health &
Development (NSHD) is a prospective birth cohort study of a
class stratified sample (5362 births; 2547 women, 2815 men)
of all births that occurred in the first week of March 1946 in
England, Scotland, and Wales. Follow-up included 20 contacts
with the whole cohort between birth and 53 years of age,
when 3035 participants (1472 men and 1563 women) provided
information. The majority of participants (2989) were then
interviewed and measured at home by trained research nurses
using a standardized protocol. Those not visited at home
completed a postal questionnaire (46). The participation rate
was 70.4% among survivors still resident in England, Wales, or
Scotland, and 89.6% for whom contact was attempted. Contact
was not attempted for those previously refusing to take part
(640), living abroad at time of interview (119), emigrated (461),
or those who had already died (469). The data collection
received MREC approval, and respondents gave informed
consent to each set of questions and measures. The sample is
reasonably representative of the national population of the
same or similar age.15 Similar data collections occurred at ages
36 (N = 3322) and 43 years (N = 3262).16
Anthropometric variables
At 53 and 43 years measures of weight (kilograms (kg)), height
(centimetres (cm)), and trunk length (cm) were obtained.
Weight was measured to the nearest 0.1 kg with participants
wearing light indoor clothing and no shoes. Height was
measured to the nearest 0.5 cm, using a portable stadiometer
with participants standing without shoes and with heels against
the wall as tall as possible with the head in the Frankfort plane.
Sitting height, used to represent trunk length, was measured to
the nearest 0.5 cm. Participants were asked to sit upright, with
their back against the vertical stand of the stadiometer, on the
base plate located on a hard, flat seat, with the head in the
Frankfort plane and their feet on the floor. Leg length was
calculated as the difference between standing and sitting
heights. Height and weight at 36 years were measured
according to the same standard protocol. Body mass index
was defined as weight/height2 (kg/m2). Trunk length was not
measured at 36 years and so leg length was unavailable at this
age. Information on height, leg and trunk length at 53 years was
therefore used or, if unavailable, at 43 years (474 participants).
Social class
Social class (manual/non-manual) was based on occupation
according to the Registrar General’s Classification. Childhood
social class was based on father’s occupation when survey
members were 4 years old, or if unavailable, when survey
members were aged 11 years (n = 125) or 15 years (n = 48).
Adult social class was based on survey member’s own
occupation at 53 years or if unavailable on occupation at 43
years (n = 513) or 36 years (n = 185).
Education
Highest educational or training qualifications achieved by 26
years (Department of Education and Science, 1972), were
grouped into either less than advanced secondary education
(‘A’-levels usually attained at 18 years, and their training
equivalents) or advanced secondary or higher.
Measurements
Smoking and exercise
At 53 years participants reported smoking status, and ‘current’
smokers were distinguished from ‘previous’ and ‘never’
smokers. Information on physical exercise was based on
reported participation in sports or vigorous leisure time
activities during spare time in the last 4 weeks.
Blood pressure
Blood pressure was measured by trained research nurses at
36, 43, and 53 years, according to a standardized protocol.
Peripheral blood pressure was measured twice in the brachial
Medication
Nurses recorded participants’ current medication at 53 years,
which was coded according to the British National Formula
(BNF) Number 40 (2000). Current use of antihypertensive
INFLUENCE OF SHORT STATURE ON PULSE PRESSURE AND BLOOD PRESSURE
medication (BNF sections 2.2—Diuretics, 2.4—Beta blockers,
2.5—Drugs affecting the renin–angiotensin system and some
other antihypertensive drugs, 2.6.2—Calcium-channel blockers)
was used in this analysis. At ages 43 and 36 years participants
were asked whether they had taken any prescribed medicines
or tablets for high blood pressure in the last year.
Statistical analysis
Linear regression analysis was used to estimate cross-sectional
relationships between the explanatory variables and blood
pressure measures at ages 36, 43, and 53 years. Models were
fitted including both men and women and an interaction term
between sex and either height, leg, or trunk length was used
to investigate whether the effect of anthropometry on blood
pressure differed significantly between the sexes. These analyses
were performed using Stata 7.0 software.17
Multilevel models18 were then used, with blood pressure
as a repeated outcome measure, using the package MLwiN.19
These models take account of the correlation between repeated
measures on the same individual and allow for incomplete
outcome data as long as a missing at random process can be
assumed.18,20,21 First, the change in blood pressure with age
was modelled. The intercept (mean blood pressure at 36 years)
and linear and quadratic terms for age were used to model the
non-linear change in pulse pressure and systolic blood pressure
over time. The change in diastolic blood pressure with age was
found to be linear and the quadratic increase observed for pulse
pressure and systolic blood pressure was non-significant and
omitted from models for diastolic blood pressure. In all models,
the variance of blood pressure was allowed to change with
age (level 1 random variation), and both intercept and linear
changes with age (slope) were allowed to vary between
individual cohort members (level 2 random variation). In all
analyses, separate curves were modelled for men and women
by including a sex variable in the model and also interactions
between sex and the linear effect of age and sex and the
quadratic effect of age (Appendix, Equation 1).
The intercept was then allowed to vary according to height,
leg or trunk length. To test whether associations between
components of height and blood pressure changed with
increasing age, interactions between the anthropometric
measures and age were added to each model. Initially,
interactions with the linear as well as the quadratic effect of age
were considered; however, the latter was non-significant in all
models and was therefore omitted from further analyses
(Appendix, Equation 2). χ2-tests based on (ˆ /se(ˆ ))2, where ˆ
is the regression coefficient, were carried out to assess levels of
significance for the fixed effect parameters as suggested by
Goldstein.18
Analyses were first performed with the maximum number of
observations (9086) and repeated including only observations
from participants with information on all covariates (7304),
required for adjusted analyses.
Adjustments were performed introducing potential confounders (or groups of confounders representing similar underlying mechanisms) one at a time and all together, to investigate
their separate and joint effects on the associations of interest.
First, antihypertensive treatment status for each time point was
considered. Interactions between components of height and
treatment status were also added to assess whether associations
907
differed according to treatment status. Second, body mass index
at each time point was added to the model, influencing both
intercept and slope, as a time varying covariate. Third, the
influence of the other adult factors, social class, smoking,
exercise, and educational attainment was considered. Fourth,
the impact of early life factors was investigated by introducing
birthweight and father’s social class into the model. Finally, a
model was fitted including all variables.
Analyses were repeated using sex-standardized measures
of blood pressure (z-scores) to assess whether the increase in
the variance of blood pressure with age, potentially due to the
different measurement instruments used, had an impact on the
findings.
Results
A total of 3414 participants (1721 men and 1693 women) had
at least one measure of blood pressure (both systolic and
diastolic blood pressure) and corresponding height and leg
length measures. Mean pulse pressure, and systolic and diastolic
blood pressure increased with age in both men and women
between 36 and 53 years (Table 1). The increase in pulse
pressure and systolic blood pressure was greater between 43
and 53 years, compared with their earlier change; their
variation was also found to increase with age.
Pulse pressure, and systolic and diastolic blood pressure at 53
and 43 years decreased significantly with increasing height and
leg length, but not trunk length (Table 2). Regression estimates
for all three measures of blood pressure were smaller at 43 years
for both height and leg length, compared with those at age 53
years. Taller height was also significantly associated with lower
systolic and diastolic blood pressure at 36 years, as was shorter
leg length with lower diastolic blood pressure. Although
regression coefficients for pulse pressure at this age were in the
same direction, effects were smaller and not statistically
significant.
Table 1 Mean blood pressure (standard deviation (SD)) in men and
women at ages 36, 43, and 53 years
n
Pulse
pressure
Mean (SD)
Systolic
blood
pressure
Mean (SD)
Diastolic
blood
pressure
Mean (SD)
36 years
Overall
3030
43.2 (11.3)
120.1 (15.1)
76.9 (12.4)
Men
1514
44.2 (11.7)
122.8 (14.8)
78.6 (12.5)
Women
1516
42.1 (10.8)
117.3 (14.8)
75.2 (12.0)
43 years
Overall
3153
43.7 (11.6)
123.3 (16.0)
79.6 (12.4)
Men
1585
43.1 (11.5)
125.1 (15.6)
82.1 (12.1)
Women
1568
44.2 (11.7)
121.4 (16.2)
77.2 (12.2)
53 years
Overall
2903
51.7 (13.5)
136.1 (20.0)
84.4 (12.2)
Men
1440
52.7 (13.4)
140.0 (19.8)
87.3 (12.3)
Women
1463
50.6 (13.6)
132.2 (19.5)
81.6 (11.5)
908
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Table 2 Cross-sectional associations between components of height and blood pressure measures at each age (regression coefficients (95% CI))
representing the change in blood pressure (mm Hg) for each centimetre increase in height, leg, or trunk length)
Pulse pressure
Systolic blood pressure
Diastolic blood pressure
nb
Coeff.
95% CI
Coeff.
95% CI
Coeff.
95% CI
Overalla
3034
0.029
0.093–0.035
0.120
0.20 to 0.034
0.089
0.16 to 0.020
Men
1516
0.013
0.077–0.10
0.038
0.15–0.076
0.051
0.15–0.046
Women
1518
0.078
0.077–0.012
0.210
0.34 to 0.088
0.130
0.23 to 0.034
36 years
Height
Test for interactionc
P = 0.16
P = 0.044
P = 0.24
Leg length
Overalla
3030
0.008
0.088–0.072
0.100
0.21–0.0041
0.930
Men
1514
0.014
0.13–0.10
0.067
0.21–0.080
0.052
0.18–0.071
Women
1516
0.002
0.11–0.11
0.140
0.29–0.013
0.140
0.26 to 0.014
Test for interactionc
P = 0.88
P = 0.50
0.18 to 0.0056
P = 0.34
Trunk length
Overalla
3032
0.048
0.14–0.047
0.110
0.24–0.013
0.065
Men
1514
0.058
0.080–0.20
0.014
0.16–0.19
0.044
0.19–0.10
Women
1518
0.160
0.29 to 0.030
0.250
0.43 to 0.067
0.087
0.23–0.059
Test for interactionc
P = 0.024
P = 0.042
0.17–0.039
P = 0.69
43 years
Height
Overalla
3154
0.065
0.13 to 0.0004
0.150
0.23 to 0.057
0.080
Men
1586
0.071
0.16–0.015
0.092
0.21–0.025
0.021
0.11–0.070
Women
1568
0.057
0.15–0.039
0.210
0.34 to 0.076
0.150
0.25 to 0.05
Test for interactionc
P = 0.83
P = 0.20
0.15 to 0.013
P = 0.059
Leg length
Overalla
3153
0.095
0.18 to 0.013
0.190
0.30 to 0.078
0.096
Men
1585
0.10
0.22–0.0087
0.150
0.30–0.0051
0.044
0.16–0.075
Women
1568
0.085
0.21–0.035
0.240
0.40 to 0.072
0.150
0.28 to 0.028
Test for interactionc
P = 0.82
P = 0.43
0.18 to 0.0098
P = 0.21
Trunk length
Overalla
3155
0.016
0.11–0.082
0.069
0.20–0.065
0.054
Men
1585
0.027
0.16–0.11
0.017
0.20–0.17
0.009
0.13–0.15
Women
1570
0.004
0.15–0.14
0.120
0.32–0.072
0.120
0.27–0.029
Test for interactionc
P = 0.82
P = 0.44
0.16–0.049
P = 0.22
53 years
Height
Overalla
2905
0.250
0.33 to 0.17
0.360
0.47 to 0.25
0.110
Men
1441
0.230
0.33 to 0.12
0.310
0.46 to 0.15
0.080
0.18–0.017
Women
1464
0.280
0.40 to 0.16
0.420
0.59 to 0.26
0.140
0.24 to 0.045
Test for interactionc
P = 0.51
P = 0.32
0.18 to 0.04
P = 0.37
Leg length
Overalla
2903
0.360
0.46 to 0.26
0.480
0.62 to 0.34
0.120
Men
1440
0.380
0.51 to 0.25
0.500
0.70 to 0.31
0.120
0.24–0.0017
Women
1463
0.340
0.48 to 0.19
0.450
0.66 to 0.25
0.120
0.24 to 0.0043
Test for interactionc
P = 0.65
P = 0.74
0.21 to 0.03
P = 0.96
Trunk length
Overalla
2904
0.057
0.17–0.059
0.130
0.30–0.041
0.071
Men
1440
0.010
0.17–0.15
0.016
0.25–0.22
0.003
0.15–0.14
Women
1464
0.110
0.28–0.06
0.250
0.49 to 0.007
0.140
0.29 to 0.0003
Test for interactionc
P = 0.43
P = 0.18
P = 0.18
a Adjusted for sex.
b Numbers are based on the model including pulse pressure and may vary slightly in analysis of systolic or diastolic blood pressure.
c Test for interaction is for sex by component of height in a model including both men and women.
0.17–0.032
INFLUENCE OF SHORT STATURE ON PULSE PRESSURE AND BLOOD PRESSURE
Of 27 tests for interactions that were carried out to assess
whether the effect of components of height differed between
the genders, only three were statistically significant (P 0.05)
(Table 2). At 53 and 43 years, none of the tests for interaction
reached conventional levels of statistical significance. We
therefore present estimates for men and women combined,
adjusted for sex, in all further analyses.
Multilevel modelling of the associations between
components of height and repeated measures of
blood pressure at 36, 43, and 53 years
Separate inclusion of height, leg, or trunk length and interactions
between each of the components of height and the linear effect
of age showed that the effects of both height and leg length on
pulse pressure became significantly stronger with age (Table 3).
The linear change in pulse pressure was found to be 0.020 mm Hg
[95% Confidence Interval (CI)] (0.026 to 0.014) per year
lower for every centimetre increase in leg length. The change in
systolic blood pressure was found to be 0.021 mm Hg (0.029
to 0.013) per year lower for every centimetre increase in leg
length. The model leads to an estimated effect of leg length
on systolic blood pressure at 36 years of 0.086 mm Hg
(0.18–0.012) per centimetre increase in leg length and one
of 0.44 mm Hg per centimetre increase in leg length at 53
years. These estimates are of similar magnitude to those
observed in the cross-sectional analysis (Table 2). According to
this model, the estimated systolic blood pressure of a participant
whose legs were 10 centimetres shorter compared with a taller
participant was just under 1 mm Hg higher at age 36 years and
increased by 3.6 mm Hg more over the 17 years of follow-up (or
0.21 mm Hg per year), resulting in a difference of 4.4 mm Hg at
age 53 years.
Diastolic blood pressure at 36 years (intercept) was
significantly influenced by height and leg length, however,
these effects did not appear to get stronger over time, as
indicated by the non-significant results for the slope (P 0.5 in
both cases). Trunk length was not related to any of the blood
pressure measures at 36 years, and none of these associations
909
changed with age. Trunk length was therefore omitted from all
further analyses.
Results from analyses in the restricted sample with complete
information (7304 observations) showed that estimates for the
amplification of the effect of height and leg length on pulse
pressure and systolic blood pressure with age were of similar
magnitude compared with previous analyses. Levels of
significance remained identical (Table 4).
Adjusted analyses
For pulse pressure, individual adjustment for confounders or
groups of confounders had only a small effect on the estimates,
but coefficients were reduced slightly further when all variables
were included simultaneously in the same model (Table 4).
However, levels of significance remained unchanged (P 0.001
in all cases). For systolic blood pressure, body mass index and
the significant change of its effect on blood pressure with age
had the largest impact on the amplification of the effect of
height or leg length. Full adjustment including all variables
under consideration did not reduce the estimates much further.
Again, levels of significance remained high (P 0.01 in all
cases). In addition, no evidence was found that the associations
between either height or leg length and blood pressure were
modified by treatment status (results not shown).
Sensitivity analysis
Replacing the blood pressure measures by their internally
derived standard deviation scores did not alter associations
between the anthropometric measures and pulse pressure or
systolic blood pressure. A significant increasing effect of height
and leg length with age remained for both pulse pressure and
systolic blood pressure using standardized outcomes (results not
shown).
Discussion
Main findings and their interpretation
In this prospective birth cohort study we found strong evidence
that the inverse associations between both height and leg
Table 3 The effects of anthropometry (cm) on blood pressure between 36 and 53 years. Regression coefficients (95% CI) for the effect on blood
pressure at 36 years (intercept) and on the linear change (slope) between 36 and 53 years from a multilevel model including 9086 observations
Intercept (mm Hg)
Slope (mm Hg/year)
Coefficient
95% CI
P-valuea
Coefficient
95% CI
P-valuea
0.005
0.064–0.054
0.87
0.014
0.020 to 0.0081
0.001
Pulse pressure
Height
0.010
0.063–0.083
0.79
0.020
0.026 to 0.014
0.001
0.040
0.13–0.046
0.36
0.001
0.0088–0.067
0.80
Height
0.091
0.17 to 0.013
0.023
0.013
0.021 to 0.0052
0.001
Leg length
0.086
0.18–0.012
0.085
0.021
0.029 to 0.013
0.001
Trunk length
0.095
0.21–0.021
0.11
0.0001
0.0098–0.0098
0.99
0.083
0.15 to 0.020
0.010
0.001
0.0049–0.0029
0.62
Leg length
0.088
0.17 to 0.0076
0.032
0.002
0.0079–0.0039
0.51
Trunk length
0.053
0.15 to 0.041
0.27
0.001
0.0088–0.0068
0.80
Leg length
Trunk length
Systolic blood
Diastolic blood pressureb
Height
a χ2-test for linear trend.
b Model for diastolic blood pressure differs slightly from others, as it does not include a quadratic term for age.
910
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Table 4 The effects of height and leg length (cm) on blood pressure between 36 and 53 years. Regression coefficients (95% CI) for the effect on
the linear change of blood pressure (slope) between 36 and 53 years from multilevel models including 7110 observations before and after
adjustments
Height
Leg length
Coefficient
95% CI
P-valuea
Coefficient
95% CI
P-valuea
Unadjusted
0.013
0.019 to 0.0071
0.001
0.021
0.029 to 0.013
0.001
Adjustment for antihypertensive treatment
0.012
0.018 to 0.0061
0.001
0.020
0.028 to 0.012
0.001
Adjustment for body mass indexb
0.012
0.018 to 0.0061
0.001
0.019
0.027 to 0.011
0.001
Adjustment for adult factorsc
0.013
0.019 to 0.0071
0.001
0.021
0.029 to 0.013
0.001
Adjustment for early life factorsd
0.014
0.020 to 0.0081
0.001
0.021
0.029 to 0.013
0.001
Adjustment for all variables mentioned above
0.011
0.017 to 0.0051
0.001
0.018
0.026 to 0.010
0.001
0.015
0.023 to 0.0072
0.001
0.022
0.032 to 0.012
0.001
0.001
Pulse pressure (mm Hg/year)
Systolic blood pressure (mm Hg/year)
Unadjusted
Adjustment for antihypertensive treatment
0.013
0.021 to 0.0052
0.001
0.021
0.031 to 0.011
Adjustment for body mass indexb
0.012
0.020 to 0.0042
0.003
0.015
0.025 to 0.0052
0.003
Adjustment for adult factorsc
0.014
0.022 to 0.0062
0.001
0.022
0.032 to 0.012
0.001
Adjustment for early life factorsd
0.014
0.022 to 0.0062
0.001
0.022
0.032 to 0.012
0.001
Adjustment for all variables mentioned above
0.010
0.018 to 0.0022
0.012
0.015
0.025 to 0.0052
0.003
a χ2-test for linear trend.
b Adjustment includes body mass index as a time varying covariate and an interaction term for BMI and age.
c Adjustment for adult life factors: educational attainment, smoking, exercise, and adult social class.
d Adjustment for early life factors: birthweight and childhood social class.
length, and pulse pressure and systolic blood pressure were
amplified with age, independently of potential confounders.
Early endocrine control (hormonal levels and receptor
expression) may simultaneously influence growth spurts of the
long bones (of the leg) and arterial growth, leading to changes
in the structure and function of the developing vasculature, and
altering the susceptibility for arterial stiffness and hypertension
in later life.22 We show that short leg length is the key
component of height associated with larger than average
increases in pulse pressure and systolic blood pressure up to
middle age. Systolic blood pressure and arterial stiffness increase
with age. If poor early growth contributed to the tracking of
these measures, its detrimental influence on vascular structure
and function in the first years of life may amplify the
vulnerability of those with shorter legs to the effects of ageing
on the arterial tree. Our evidence supports this hypothesis by
showing an amplification of the effect of leg length, as a marker
of early growth, on pulse pressure and systolic blood pressure
between 36 and 53 years.
One previous small study suggested that poor growth in
childhood is an important determinant of systolic blood
pressure and pulse pressure, but not diastolic blood pressure,
using a single measure of blood pressure taken in early old
age.13 Interestingly, as in our study, associations were observed
with systolic blood pressure and pulse pressure, but not diastolic
blood pressure. In industrialized countries, systolic blood
pressure increases progressively throughout adult life, while
diastolic blood pressure increases less steeply and ceases to rise
or even falls around 55 years,23 resulting in a rise in pulse
pressure throughout adult life. Accordingly, in this study of
blood pressure in the middle years of life, pulse pressure and
systolic blood pressure show a greater rise with age, compared
with diastolic blood pressure. If growth limiting factors
influence the age–blood pressure relationship, then stronger
associations may therefore be expected with those measures of
blood pressure whose increase is more closely linked with ageing.
The majority of cardiovascular risk associated with hypertension is due to blood pressure gradually increasing with age.
The age specific rise, and therefore hypertension, is essentially
absent in certain rural communities, and studies have shown
that this protection is partly lost through migration to
industrialized communities.23 This suggests that protective
environmental factors are necessary for maintaining low
pressures into later life, and hence supporting the hypothesis
that environmental influences on growth modify the age–blood
pressure relationship.
Several determinants of growth may contribute to the
association of blood pressure with its age-related rise. The
quality of early nutrition, rather than simply energy intake, has
a critical influence on growth.24 Dietary interventions may affect
the individual age–blood pressure relationship and therefore
impact on the population burden of hypertension and associated
cardiovascular disease. Growth limiting factors such as impaired
fetal development and disadvantageous socioeconomic
conditions have been associated with high blood pressure and
cardiovascular risk and were considered as alternative
explanations in previous investigations.14 In this study,
associations remained unchanged after adjustment for
birthweight; however, as birthweight is only a crude marker of
growth and development before birth, this does not preclude the
possibility of prenatal factors simultaneously influencing growth
and later blood pressure. Adjustment for childhood social class,
educational attainment, adult social class and potential
confounding factors in adulthood (obesity, smoking, and lack of
INFLUENCE OF SHORT STATURE ON PULSE PRESSURE AND BLOOD PRESSURE
exercise) slightly reduced the estimates, but did not alter the
levels of significance. Some of these factors, particularly obesity,
showed independent effects on blood pressure, and their
importance for the control of cardiovascular risk should not
be understated. Furthermore, a great variety of influences on
growth are socially distributed, and these contribute to the
shorter height, and particularly leg length, of children growing up in disadvantageous social conditions.11,12,25,26 These
influences include prenatal development, prematurity, maternal
health, behaviour and care for the child, early nutrition, living
conditions, infections, and age at puberty.26–30 Adjustment for
childhood social conditions is unlikely to fully account for the
diverse influences of these factors on early and later blood
pressure.
Short height and blood pressure regulation might be jointly
genetically determined, but this cannot be investigated in the
NSHD. However, the evidence of a genetically determined
association between height and cardiovascular risk factors to
date is weak.31
Strengths and limitations of the study
This is the oldest ongoing birth cohort internationally, and no
other birth cohort study is available with earlier measures of
blood pressure and a follow-up till middle age. Potentially
avoidable loss of participants in this study and the impact of
survivor bias have previously been discussed in detail14,15,32 and
would result in an underestimation of the observed associations.
This study is restricted to an investigation of changes in blood
pressure measured at three time points during the middle years
of life. It may be that the observed associations will become
stronger as the cohort ages and arterial stiffness and blood
pressure increase further due to the continuing effects of ageing
on the arterial tree. This will be investigated at future data
collections. The availability of just three measures of blood
pressure taken at fairly distant time points allows for only
relatively simplistic modelling of changes in blood pressure,
which may be unable to account for short-term changes or
more complex variations of blood pressure over time.
We used adult leg and trunk length as markers of growth
at different phases, rather than height measured during
childhood, for two reasons. First, this approach is comparable
with recent studies suggesting that leg length, as opposed to
trunk length, is the component of height most closely linked
to cardiovascular risk in adult life. Second, the importance of
children’s height at a given age for determining their final
adult height depends on rate (growth tempo) and timing of
maturation. The comparison of leg to trunk length distinguishes
the rapid growth of long bones occurring in childhood from
the slower and later growth of the trunk.24 Adult leg and trunk
length were used, as components of height in childhood were
not available in this cohort.
Using adult indicators of childhood growth has some
disadvantages. Although leg length is regarded as a marker of
environmental exposures during the first years of life, pubertal
growth and timing of puberty as well as genetic factors will
also influence attained leg length. Even if leg length is predominantly a marker of the prepubertal growth phase, it
remains unclear whether it is early post-natal or later
prepubertal growth that is most important. Shrinkage may
introduce error in measures of adult height. However, previous
911
studies have shown that loss of stature at these early ages
(before 53 years) is only small.33 In this study, only total and
sitting height (trunk length) were measured, and leg length
calculated from them, resulting in greater measurement error in
leg length (non-differential misclassification), which may bias
results towards the null. Nevertheless, associations with blood
pressure were observed for leg length, not trunk length, and the
strength of the association may therefore be even greater if leg
length was measured more accurately.
Leg and trunk length were only measured at 43 and 53 years.
Participants measured at 36 years, therefore, needed to
participate at 43 or 53 years to be included in the analysis,
possibly resulting in selection bias.
The increase in mean levels of blood pressure between 43
and 53 years may be influenced systematically by differences in
the sphygmomanometers used. However, readings between
instruments are not likely to vary systematically by components
of height. The variation in blood pressure reading might also
vary between instruments or increase with age, as observed for
pulse pressure and systolic blood pressure in this study. Using a
standardized outcome measure, which accounts for the increase
in the variation in blood pressure with age, a significant
increasing effect of height and leg length with age was observed
for pulse pressure and systolic blood pressure, suggesting that
the amplification of the effects of height and leg length were
not simply due to increasing variance with age or change in
measurement instrument.
Conclusions
In this middle-aged cohort pulse pressure and systolic blood
pressure increased linearly with decreasing height and leg
length, but not trunk length, suggesting the role of poor
childhood growth for the development of high blood pressure
in later life. Importantly, these associations were shown to be
significantly amplified with age. This is evidence for the
hypothesis that people with restricted growth in the first years
of life may be more susceptible to the effects of ageing on the
arterial tree. Poor early growth may therefore contribute to
the tracking of cardiovascular risk throughout life and indicate
the need for early prevention of increasing blood pressure
during mid-life. Early growth restricting factors may be potential
mediators. While infant diet, childhood infections, and
psychosocial deprivation limit early growth,34,35 previous
studies have not investigated how these factors may link
short height and cardiovascular risk. Future research on these
childhood factors could therefore provide further insights into
the aetiology of hypertension and arterial stiffness. This may
help to identify modifiable childhood factors, which can be
targeted during specific periods of development, increasing the
practicality of a public health intervention.
Acknowledgements
The study has been supported by grants from the Medical
Research Council. C.L. is funded by a Medical Research Council
and Department of Health Research Training Fellowship. R.H.,
D.K., and M.W. are supported by the Medical Research Council.
The views expressed in this publication are those of the authors
and not necessarily those of the Department of Health or MRC.
912
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
KEY MESSAGES
•
Shorter height and leg length, but not trunk length, are associated with higher pulse pressure and systolic blood pressure,
similarly in men and women.
•
Longitudinal analyses of repeated measures of blood pressure show that these associations are significantly amplified with age,
suggesting that people with short stature may be more susceptible to the effects of ageing on the arterial tree.
•
Poor growth may contribute to the tracking of cardiovascular risk throughout life and indicate the need for early prevention of
increasing blood pressure during mid-life.
References
1 Berenson GS, Srinivasan SR, Bao W Newman WP III, Tracy RE,
Wattigney WA. Association between multiple cardiovascular risk
factors and atherosclerosis in children and young adults. The Bogalusa
Heart Study. N Engl J Med 1998;338:1650–56.
2 Bao W, Srinivasan SR, Wattigney WA, Berenson GS. Persistence of
multiple cardiovascular risk clustering related to syndrome X from
childhood to young adulthood. The Bogalusa Heart Study. Arch Intern
Med 1994;154:1842–47.
3 Marmot MG, Shipley MJ, Rose G. Inequalities in death—specific
explanations of a general pattern? Lancet 1984;1:1003–06.
4 McCarron P, Okasha M, McEwen J, Davey Smith G. Height in young
adulthood and risk of death from cardiorespiratory disease: a
prospective study of male former students of Glasgow University,
Scotland. Am J Epidemiol 2002;155:683–87.
5 Song YM, Davey Smith G, Sung J. Adult height and cause-specific
mortality: a large prospective study of South Korean men. Am J
Epidemiol 2003;158:479–85.
14 Langenberg C, Hardy R, Kuh D, Wadsworth ME. Influence of height,
leg and trunk length on pulse pressure, systolic and diastolic blood
pressure. J Hypertens 2003;21:537–43.
15 Wadsworth ME, Butterworth SL, Hardy RJ et al. The life course
prospective design: an example of benefits and problems associated
with study longevity. Soc Sci Med 2003;57:2193–205.
16 Hardy R, Kuh D, Langenberg C, Wadsworth ME. Birthweight,
childhood social class, and change in adult blood pressure in the 1946
British birth cohort. Lancet 2003;362:1178–83.
17 Stata Statistical Software: Release 7.0 College Station. Texas: Stata
Corporation, 2002.
18 Goldstein H. Multilevel Statistical Models. 2nd edn. London: Edward
Arnold, 1995.
19 Goldstein H, Rasbash J, Plewis I et al. A User’s Guide to MLwiN. Multilevel
Models Project. 2nd edn. London: Institute of Education, University of
London, 1998.
20 Raudenbush SW, Bryck AS. Hierarchical Linear Models. 2nd edn.
Thousand Oaks, CA: Sage Publications, 2002.
6 Gunnell DJ, Davey Smith G, Frankel S et al. Childhood leg length and
21 Little RJ, Raghunathan T. On summary measures analysis of the
adult mortality: follow up of the Carnegie (Boyd Orr) survey of diet
and health in pre-war britain. J Epidemiol Community Health
1998;52:142–52.
linear mixed effects model for repeated measures when data are not
missing completely at random. Stat Med 1999;18:2465–78.
7 Davey Smith G, Greenwood R, Gunnell D, Sweetnam P, Yarnell J,
Elwood P. Leg length, insulin resistance, and coronary heart disease
risk: the Caerphilly Study. J Epidemiol Community Health 2001;55:
867–72.
8 Lawlor DA, Ebrahim S, Davey Smith G. The association between
components of adult height and Type II diabetes and insulin
resistance: British Women’s Heart and Health Study. Diabetologia
2002;45:1097–1106.
9 Gunnell D, Whitley E, Upton MN, McConnachie A, Davey Smith G,
Watt GC. Associations of height, leg length, and lung function with
cardiovascular risk factors in the Midspan Family Study. J Epidemiol
Community Health 2003;57:141–46.
10 Lawlor DA, Taylor M, Davey Smith G, Gunnell D, Ebrahim S.
Associations of components of adult height with coronary heart
disease in postmenopausal women: the British women’s heart and
health study. Heart 2004;90:745–49.
11 Gunnell DJ, Davey Smith G, Frankel SJ, Kemp M, Peters TJ. Socio-
economic and dietary influences on leg length and trunk length in
childhood: a reanalysis of the Carnegie (Boyd Orr) survey of diet and
health in prewar Britain (1937–39). Paediatr Perinat Epidemiol 1998;
12 (Suppl. 1):96–113.
12 Wadsworth MEJ, Hardy RJ, Paul AA, Marshall SF, Cole TJ. Leg and
trunk length at 43 years in relation to childhood health, diet and
family circumstances; evidence from the 1946 national birth cohort.
Int J Epidemiol 2002;31:383–90.
13 Montgomery SM, Berney LR, Blane D. Prepubertal stature and blood
pressure in early old age. Arch Dis Child 2000;82:358–63.
22 Leeson CP, Kattenhorn M, Deanfield JE, Lucas A. Duration of breast
feeding and arterial distensibility in early adult life: population based
study. BMJ 2001;322:643–47.
23 Rose G. Hypertension in the community. In: Bulpitt C (ed).
Epidemiology of Hypertension. Amsterdam: Elsevier, 1985, pp. 1–14.
24 Cole TJ. Secular trends in growth. Proc Nutr Soc 2000;59:317–24.
25 Tanner JM. A History of the Study of Human Growth. Cambridge:
Cambridge University Press, 1981.
26 Kuh D, Wadsworth M. Parental height: childhood environment and
subsequent adult height in a national birth cohort. Int J Epidemiol
1989;18:663–68.
27 Goldstein H. Factors influencing the height of seven year old
children—results from the National Child Development Study. Hum
Biol 1971;43:92–111.
28 Rona RJ, Swan AV, Altman DG. Social factors and height of primary
schoolchildren in England and Scotland. J Epidemiol Community Health
1978;32:147–54.
29 Smith AM, Chinn S, Rona RJ. Social factors and height gain of primary
schoolchildren in England and Scotland. Ann Hum Biol 1980;7:115–24.
30 Michaelsen KF, Larsen PS, Thomsen BL, Samuelson G. The
Copenhagen cohort study on infant nutrition and growth: duration of
breast feeding and influencing factors. Acta Paediatr 1994;83:565–71.
31 Langenberg
C, Marmot M. Commentary: disentangling the
association between short height and cardiovascular risk-genes or
environment? Int J Epidemiol 2003;32:614–16.
32 Kuh D, Hardy R, Langenberg C, Richards M, Wadsworth ME.
Mortality in adults aged 26–54 years related to socioeconomic
INFLUENCE OF SHORT STATURE ON PULSE PRESSURE AND BLOOD PRESSURE
conditions in childhood and adulthood: post war birth cohort study.
BMJ 2002;325:1076–80.
33 Cline MG, Meredith KE, Boyer JT, Burrows B. Decline of height with
age in adults in a general population sample: estimating maximum
height and distinguishing birth cohort effects from actual loss of
stature with aging. Hum Biol 1989;61:415–25.
34 Nystrom Peck AM, Lundberg O. Short stature as an effect of economic
and social conditions in childhood. Soc Sci Med 1995;41:733–38.
35 Zimet GD, Owens R, Dahms W, Cutler M, Litvene M, Cuttler L.
Psychosocial outcome of children evaluated for short stature. Arch
Pediatr Adolesc Med 1997;151:1017–23.
Appendix
Parameters and estimates of the basic and
extended multilevel models
In the following equations, yij denotes the blood pressure (pulse
pressure, systolic or diastolic blood pressure) of subject j
( j = 1, . . .,N) on measurement occasion i (i = 1, 2, 3) and ageij,
the age at which that measurement was taken (36, 43, and
53 years). The fixed parameter β0 represents the mean
intercept, in this example, the overall mean blood pressure at
age 36 years. The fixed parameter β1 represents the mean slope
or equivalently the linear change in blood pressure for each
yearly increase in age. β2–β5 denote the fixed effects of age2,
sex, and the interaction terms age by sex and age2 by sex,
respectively. The basic multilevel model for repeated measures
of blood pressure at age 36, 43, and 53 years is then written as
Yij = β0ij + β1ij ageij + β2 age2ij + β3 sexj + β4 ageij sexj + β5 age2ij
sexj + β6 legj + β7 legj ageij
(2)
β1ij = β1 + u1j + e1ij
sexj + β5age2
ij sexj
β0ij = β0 + u0j + e0ij
sex were therefore omitted. The parameters u0j and u1j are the
random (between-individual) effects, which allow each
individual to have their own intercept and slope, respectively,
and indicate the deviation of each individual’s intercept and
slope from the mean intercept and slope. These ‘level 2’ random
effects parameters are assumed to be bivariate normal with
mean 0 and variance defined by the variance–covariance matrix,
having entries given by the variance of u0j (var(u0j) = 2u0), the
variance of u1j (var(u1j) = 2u1), and the covariance between u0j
and u1j (cov(u0j,u1j) = u01). The random effect of the quadratic
effect of age was considered, but found to be very small and
statistically non-significant and was therefore not included in the
final models. The within-individual (‘level 1’) variation was
allowed to increase with age and is represented by the terms
e0ij and e1ij. These ‘level 1’ random effects are assumed to be
bivariate normally distributed, with mean 0 and variance
defined by the variance–covariance matrix, having entries
given by the variance of e0ij (var(e0ij) = 2e0), the variance of
e1ij (var(e1ij) = 2e1), and the covariance between e0ij and
e1ij (cov(e0ij,e1ij) = e01). The variance 2e1 was set to 0, so that
the level 1 variance increased linearly with age.
The model given in Equation 1 was then extended to assess
how anthropometric measures influenced both the intercept
and the increase in blood pressure with age
β0ij = β0 + u0j + e0ij
yij = β0ij + β1ijageij + β2age2ij + β3sexj + β4ageij
913
(1)
β1ij = β1 + u1j + e1ij
For diastolic blood pressure, the quadratic increase of blood
pressure was non-significant and age2 and its interaction with
The age2 by anthropometric term was not significant in any
of the models and was thus omitted from our final models
(see Equation 2 with leg length as the anthropometric
measure). As outlined above, age2 and its interaction with sex
were omitted for diastolic blood pressure.