Relation between birth weight at term and systolic blood pressure in adolescenceBMJ 1994; 308 doi: https://doi.org/10.1136/bmj.308.6936.1074 (Published 23 April 1994) Cite this as: BMJ 1994;308:1074
- J W A Matthes,
- P A Lewis,
- D P Davies, Professor,
- J A Bethel
- Correspondence to: Professor Davies
- Accepted 25 January 1994
Objective : To examine whether birth weight is related to systolic blood pressure during adolescence.
Design : Retrospective (comparative) cohort study. The observers who traced and studied the subjects were unaware of their case-control status.
Subjects : 330 subjects were born in Cardiff in 1975-7. Cases who were low birth weight at term (<2500 g) were matched with controls of normal birth weight (3000-3800 g) at term.
Main outcome measures : Systolic blood pressure measured by random zero sphygmomanometry in the subject's right arm with the subject supine, corrected for size and age.
Results : The mean age at examination was 15.7 years. The mean systolic blood pressure of the cases was 105.8mm Hg and of the controls 107.5 mm Hg. The corrected difference (95% confidence interval) in systolic blood pressure between the cases and controls was 1 mm Hg (−3 to +1 mm Hg; two tailed probability 0.33).
Conclusion : Systolic blood pressure in adolescents of low birth weight is not significantly different from that of adolescents of normal birth weights.
An inverse relation between birth weight and cardiovascular disease in adults has been reported in retrospective cohort studies
One possible mechanism is via a link with raised systolic blood pressure
No difference was found between the systolic blood pressure of adolescents who as a group were growth retarded at term and a comparative group of normal birth weight
This study does not lend support to one suggested mechanism linking intrauterine growth retardation and later cardiovascular disease
Further studies are required on new cohorts to examine the mechanism by which intrauterine growth retardation might be linked to a risk of adult disease
Hypertension is a well accepted risk factor for ischaemic heart disease and stroke. To the known risk factors for hypertension of obesity,1,2 lifestyle,3,4 and increased salt intake5 has recently been added low growth rate in utero with its proxy of low birth weight at term.*RF 6-9* The hypothesis advanced is that programming of physiological or metabolic events caused by undernutrition10 at critical or sensitive periods early in life can amplify certain long term changes which eventually manifest in disease, in this case hypertension and ischaemic heart disease. Evidence for this hypothesis comes largely from studies of cohorts born in Britain earlier this century, most of whom were men.*RF 6-9* A criticism of these studies is that people whose growth has been affected in early development may continue to have imposed during childhood and adult life a hostile environment that leads eventually to disease.11 The confounding influence of these external environmental effects may be evident in advance of the consequences of the intrauterine environment. Also in some circumstances the extrauterine effects may be overwhelming.
We aimed to minimise the effects of different extrauterine environments by investigating the association between low birth weight at term with systolic blood pressure in schoolchildren.
Subjects and methods Study design
This was a retrospective cohort study in which the subjects were traced and assessed by observers who were unaware of whether they were cases or controls. A strict protocol was drawn up for the identification and selection of subjects, methods of clinical examination, and statistical analysis. Ethical approval was granted by the South Glamorgan Health Authority and the local family health services authority.
Tracking of systolic blood pressure is well documented throughout childhood and adult life.12 Raised blood pressure may be amplified throughout life so that any differences associated with low birth weight may become larger with advancing age.13 If very young children were studied tracing might be easier as few would have changed address, but any differences in blood pressure due to prenatal experiences might be small. Older children would be more difficult to find, but the differences in blood pressure might be larger. Children in school are a fairly easy group to locate with the availability of computerised school health records, and they are a homogeneous group whose environment is similar for part of the day so that some confounding variables (which might begin to impose themselves when the subject leaves school - for example, occupation) would be lessened.
Cases and controls were identified from the Cardiff birth survey. Every birth in Cardiff and later in South Glamorgan has been documented since 1965. Cases were defined as infants with a birth weight under 2500 g at 38 weeks' gestation or more, a group likely to have suffered retardation of intrauterine growth. Unlike low birth weight and preterm delivery there is no internationally agreed definition of fetal growth retardation. Controls were infants born at 38 weeks' gestation or more and weighing 3000-3800 g. These infants were judged to have grown adequately in utero.
Multiple pregnancies and infants with congenital abnormalities were excluded, as were deaths and medical conditions that might influence systolic blood pressure. Only births at the two large obstetric departments were used as the small numbers born elsewhere would not have sufficient well matched controls. There were 8920 relevant deliveries during the study.
The initial variables used to calculate sample size were (alpha)=0.05 and β=0.10, and the smallest difference in systolic blood pressure that could be viewed as clinically important was 4mm Hg. No estimate of the standard deviation of the differences in systolic blood pressure between cases and controls matched as described was available. The estimated standard deviation of systolic blood pressure for 15 year olds was 12 mm Hg,14 giving a standardised difference of 4/12=0.33. Without matched data a sample size of between 173 and 234 in each group at analysis would be required. The data from the birth survey showed that the academic year cohorts from 1 September 1975 to 30 August 1977 would yield 230 potential cases. At the start of the study (March 1991) all were expected to be receiving compulsory full time education. If the matching were to reduce the variance then a greater power would ensue.15
Matching of cases and controls
For each index case a list of eligible controls was drawn up with a computer. Cases and controls were matched for sex, hospital of delivery, parity of the mother, date of birth, and length of gestation. Length of gestation was calculated from the mother's last menstrual period as ultrasound scanning was not routinely available. If there was doubt regarding length of gestation a Dubowitz assessment of the neonate was performed. Prior rules for selection of controls were applied in sequence until a match was found. These rules were designed so that if any bias was introduced by matching it would result in the cases having a higher systolic blood pressure than the controls, which is in favour of the alternative hypothesis.
Data on the mother collected at delivery included social, lifestyle, and clinical variables associated with the current pregnancy and relevant medical history. Data on the infant included details of the delivery and relevant clinical variables. The following multiple sources were used to trace the subjects: computerised school records, South Glamogran family health services authority, NHS central register (including armed forces), telephone directory, South Glamorgan Social Services, hospital patient administration system, general practitioners, door to door inquiries, hospital records at the time of delivery (which occasionally contained telephone numbers of relatives), and the medical directory.
The subjects who agreed to participate were visited at home by one of two observers who had been trained in the techniques of anthropometry and measurement of blood pressure.
Weight was measured to the nearest 100 g with the subject lightly clothed and without shoes by using a portable Soehnle scale. Height was measured by using a portable Harpenden stadiometer to the nearest 1 mm. Systolic blood pressure was measured with a random zero sphygmomanometer by following recommendations of the British Hypertension Society.16
The statistical package SPSS/PC was used for analysis.17 Means were compared by using a paired t test. The effect on systolic blood pressure of sex was examined by using one way analysis of variance, and the effect of birth weight, maternal height, and gestation was determined by using the non-parametric Kruskal-Wallis test. Results were all rounded after calculation.
Of 230 potential low birthweight cases identified, eight were ineligible (six were dead and two had medical conditions not recorded at birth (pseudohypoparathyroidism and severe cerebral palsy)), leaving a cohort of 222. The records of five were marked as “confidential - not to be used for research studies”; four subjects were living too far from Cardiff to visit, 17 subjects could not be traced at all, and 19 refused to participate. A total of 177 cases were finally studied.
Of the controls for these cases, 75% (133/177) were found. When it was recognised early in the study that a control would be unavailable a subsequent control was drawn. For the 30 identified subsequent controls, 93% (28) were seen. This left 12 cases who were seen but did not have matched controls and could not therefore be used in the paired analysis. In total 74% (165/222) of the case-control pairs were found. The subjects studied were of mean age 15.7 years. In one control the weight had not been recorded.
We did not consider it appropriate to determine the exact stage of puberty during a home visit. Tables I and II show the closeness of the case-control matching. Table III gives a comparison of the cases and controls. There were important differences in age, height, and weight at examination which influence systolic blood pressure. These differences may be adjusted for by fitting a linear regression model to the data.18 No existing model was suitable so age, height, weight, sex, body mass index (weight (kg)/(height (m)2)), and case-control status were entered into a stepwise multiple regression to produce a linear model for systolic blood pressure. All the available data were used. The best fit to the data was: systolic blood pressure=24.7+0.31 weight (kg) + 4.1 age (years); r2=0.24; SE weight coefficient=0.05; SE age coefficient=0.69.
This model was then used to adjust the paired differences in systolic blood pressure. For one subject the weight was not available and so adjusted systolic blood pressure could not be calculated. The mean (SD) difference in adjusted systolic blood pressure for the 164 remaining pairs of subjects was −1.0 (13.0)mm Hg, with a median of 0.35, minimum of - 41.4, and maximum of 33.2. The standard error was 1.01 and 95% confidence interval −3.0 to 1.0, giving a two tailed P value of 0.33. Table IV gives the non-significant differences in systolic blood pressure by sex. The effect of birth weight of cases (P=0.85), maternal height for cases (P=0.60), and length of gestation (P=0.66) on systolic blood pressure were all found to be non-significant by using the Kruskal-Wallis test, there being no obvious trend in the results.
Infants who weighed less than 2500 g at 38 weeks' gestation or more (cases) are considered to have probably suffered retardation of growth in utero. These infants have a lower than average ponderal index because of fetal undernutrition from various causes.19 The control infants, on the other hand, were considered to have been adequately nourished during gestation.
This is the first reported population based comparative study to investigate the influence of low birth weight at term on later blood pressure. No link could be found between birth weight and blood pressure in adolescence once the differences in size and age between the two groups at examination were considered. Other studies have also failed to find an association. In a previous study mean systolic blood pressure of very low birthweight preterm children aged 8 years was 1.8 mm Hg lower than that of normal birth weight controls, but no correction was made for body size.20 Cook et al found no association between birth weight and blood pressure in first degree relatives (mean age 43 years) of people with non-insulin dependent diabetes mellitus.21 A study from New Zealand showed that corrected systolic blood pressure in 7 year olds was 1 mm Hg higher in those who had suffered retardation of intrauterine growth compared with those of birth weight between the 10th and 90th centiles, but at the age of 18 years the differences were not significant.22 In nearly 33 000 men and women aged 17 years Seidman et al found that systolic blood pressure was positively associated with birth weight, but the correlation coefficients were low; current body weight was significantly associated with systolic blood pressure, suggesting that an adverse extrauterine environment was a much more important determinant of blood pressure than birth weight.18
Other studies have found associations between birth weight and later blood pressure. Multiple regression analysis of factors affecting systolic blood pressure in Croatian subjects aged 18-23 years showed an inverse relation with birth weight.23 In 36 year old subjects from the Medical Research Council's 1946 birth cohort systolic blood pressure fell by 2.6 mm Hg in men and 1.8 mm Hg in women from the highest to the lowest third of the distribution of birth weight.7 A study of men and women aged 50 years from Preston found that a higher systolic blood pressure was related to increased placental weight and lower birth weight.6 Barker et al found a similar relation between low birth weight and increasing systolic blood pressure in men aged 59-70 years in Hertfordshire.8 Higher standardised mortality ratios for ischaemic heart disease in men in Sheffield were associated with low birth weight, small head circumferences, and reduced ponderal indices at birth.9 Whincup et al studied children aged 5-7 years and found a relation between low birth weight and systolic blood pressure corrected for age, sex, height, and body mass index.24 The effect was similar for prematurity and growth retardation, and one suggested explanation was that increased postnatal growth rates might be more important in determining future blood pressure than prenatal growth.
Sources of bias
Many of these studies rely on cohorts born a long time ago, when health care practices, environmental conditions, and diet were probably different.*RF 6,8-10* Also only the survivors from times when infant mortality was high have been studied. These factors could be a source of bias. Different factors may now predispose to intrauterine growth retardation - for example, maternal smoking. Several groups have questioned whether the links between intrauterine growth retardation and later disease might be explained by a continuum of social and environmental insults throughout life.*RF 25-28* Our findings of no link at the age of 15 years may be because any possible adverse life style has not yet had sufficient time to exert an effect.
The possibility of a “false negative” result can never be discounted. There are two sources of bias which may have influenced our study: case ascertainment bias and recruitment bias. Case ascertainment bias may lead to important subgroups being underrepresented in the sample. Here we failed to use 57 cases (25%). Control selection and ascertainment bias may similarly lead to underrepresentation of important subgroups. Attempts were made to control these types of potential bias by keeping the field workers blind to subjects' case-control status. Also the controls were selected by using prior rules which entailed no additional judgment. The strength of the study lies in the fact that it is based on the population of a well defined geographical location, every member of that population being recorded. While the number of cases (165) may seem small compared with other studies with much larger samples the effective population studied here is over 9000 with just the cases of particular interest selected.
Our results are inconsistent with adolescents born at term with low birth weight having a systolic blood pressure more than 1 mm Hg above or more than 3 mm Hg below that of controls with normal birth weight. This result lends no support to the hypothesis that low birth weight at term is related to a higher blood pressure during adolescence.
The hypothesis that hypertension may be initiated in fetal life is fascinating and derives additional support from studies on animals. Before it shifts from hypothesis to established fact, however, further large population studies are required to determine any association between birth weight (a proxy for changes in structure, physiology, and metabolism determined in utero) and blood pressure through the whole span of life.
This study was supported by a grant from Children Nationwide. We thank the steering committee of the Cardiff birth survey for the use of the data, and Mrs D Savory, who conducted much of the fieldwork. We are also greatful for the willing cooperation of all of the subjects.