Within pair association between birth weight and blood pressure at age 8 in twins from a cohort study

Subjects

The Tasmanian infant health survey, a cohort study initiated to study sudden infant death syndrome, provided standard information prospectively on around one in five births in Tasmania from 1988 to 1995.13 Selection of singleton births was on the basis of a scoring system using six risk factors for sudden infant death syndrome (young maternal age, male sex, low birth weight, autumn birth, maternal intention to bottle feed, and duration of second stage labour)14; singletons with a composite score over a cut off point were eligible for inclusion. All multiple births were eligible irrespective of their scores.

Of 13 592 liveborn infants in Tasmania in 1988 and 1989, 2791 were eligible for the Tasmanian infant health survey. Of the 1498 infants born in southern Tasmania, a defined geographical region, 1440 (96.1%) were the subject of a hospital interview soon after the birth. We identified 1185 of these children from school enrolment lists in southern Tasmania, and they were invited to join a study either of childhood blood pressure in 1996 (1988 birth cohort) or of high density lipoprotein cholesterol concentration and blood pressure in 1997 (1989 cohort). All studies were approved by the University of Tasmania ethics committee, and written parental consent was obtained for each subject.

Measurements

Information on the infants had been collected from hospital obstetric records (gestational age, placental weight, birth weight, crown to heel length, head circumference) and from questionnaires and additional infant anthropometry (skinfold thicknesses) in the postnatal ward, as described previously.13

The children were seen at school when they were 8 years of age. Body weight was measured with bathroom scales, which were calibrated daily, and height was measured with a stadiometer. Skinfold thickness was measured with calipers at four sites (triceps, biceps, subscapular, suprailiac),15 and right mid upper arm circumference was measured with a tape. Blood pressure was measured three times with an adult and paediatric vital signs monitor (Dinamap, Critikon, Arlington, TX) with the cuff size appropriate for mid upper arm circumference, and the child seated with the left arm resting on a pillow and the elbow approximately at heart level.

Zygosity was assessed from parental reports at 8 years. We asked first whether any tests had been performed and, if so, about the results. If not, we accepted the maternal report of whether the multiplets were identical.

Statistical methods

We summarised the strength of associations between variables with product moment correlation coefficients and with linear regression coefficients for which the average of three systolic blood pressure values was the dependent variable. We adjusted for putative confounders—including gestational age, mode of delivery, previous pregnancies, duration of breast feeding, maternal smoking and alcohol intake (prenatal and postnatal), maternal education, and child size at age 8—if their association with blood pressure was biologically plausible, and if adjustment substantially changed (by 10% or more16) the regression coefficient. To capture the non-linear association of systolic pressure with the measure of fatness in childhood (fat mass estimated from skin folds and body weight, or body mass index), a predictor term for the measure of fatness and another for its square were entered in the same regression model.

For singletons, these statistics were routinely adjusted for four of the six selection factors (maternal age, month of birth, intention to breast feed, and duration of second stage labour) and by stratifying for another (sex). The final selection factor was the study factor, birth weight. Linear predictors were created for month of birth, intention to bottle feed, and duration of second stage labour by scoring their categories with the selection weights for each. Similar adjustments were not routinely made for multiplets because they were not selectively recruited, and there were limited degrees of freedom for those analyses. We checked, however, that artefactual differences had not arisen as a result of unlike methods of analysis. We adjusted for the child's age at blood pressure measurement (as days elapsed since seventh birthday).

In paired analyses, birth weight and blood pressure values from the second born twin were subtracted from corresponding values for the first born twin, and the differences calculated in this way became the subject of statistical inference. Each set of triplets was analysed as three sets of pairs, and within pair differences were calculated as above. To examine the within pair association between birth weight and blood pressure, we regressed the intrapair difference in blood pressure on the intrapair difference in birth weight.

Subjects

Of the 1185 children from the 1988 (572 children) and 1989 (613) Tasmanian infant health survey cohorts identified on school enrolment lists in 1996 and 1997 respectively, we were able to contact 1156 and to obtain measurements from 888 of the original 1185 (74.9%) children.

Table 1 shows the characteristics of the sample at birth and at a mean age 8.13 years (range 7.26-8.88 years), with subjects classified by sex and whether they were singletons or multiplets. Singletons comprised more boys (548 children) than girls (231) because of the weighting given to male sex in the selection criteria of the Tasmanian infant health survey. Comparatively more of the girls (143; 61.9%) than boys (296; 54.0%) had birth weights below the 50th centile for gestational age (from norms for singleton births in New South Wales17) This was because low birth weight was also a selection factor, and its comparative contribution was greater for the girls. The full range of birth weights was nevertheless represented, and 15 (6.5%) girls and 44 (8.0%) boys had birth weights exceeding the 90th centile.

From the 1988 and 1989 cohorts, 128 multiplets were identified at schools in southern Tasmania, and 109 (85.0%; 70 girls and 39 boys), including two sets of triplets, participated in this study. In five cases the cotwin was not included in the sample (three were sole survivors, one twin was not allowed to participate, and one twin was excluded on medical grounds). To maintain comparability between analyses, these five subjects were excluded. Multiplets had a lower mean birth weight than singletons (P<0.01) and a shorter mean gestation (P<0.01), but birth weight adjusted for gestational age was not statistically different: girls (−108 g; P=0.11), boys (−132 g;P=0.15). Multiplets had higher mean systolic blood pressure by around 2 mm Hg (P<0.01) than singletons.

Birth weight and childhood blood pressure

Systolic pressure was negatively associated with birth weight. Children with greater lean or fat mass at age 8 years had higher blood pressure and had been heavier at birth than smaller children at age 8 years. Adjusting for childhood fat and lean mass (the childhood size measures most strongly associated with blood pressure) removed the confounding effect of current size, thereby increasing the estimated strength of the association between systolic pressure and birth weight (figure). We adjusted only for fat mass in multiplets because of the colinearity of the fat mass and lean mass measurements for these children.

Effect of 1 kg increase in birth weight on systolic blood pressure at age 8 years, with and without adjustment for current size

• Download figure
• Open in new tab
• Download powerpoint

The association was stronger in multiplets than in singletons. For each 1 kg increase in birth weight there was an estimated change in systolic pressure of −1.94 mm Hg (95% confidence interval −2.89 to −0.98) in singletons (girls: −2.16, −4.24 to −0.08); boys: −1.73, −2.83 to −0.64) and −7.0 mm Hg (−10.1 to −3.9) in multiplets (girls: −5.9, −9.6 to −2.2; boys: −8.4, −15.6 to −1.2). Estimates for singletons, but not multiplets, were routinely adjusted for the selection criteria factors. Similar adjustment for multiplets made little difference: adjusted estimates were −5.4 mm Hg (−9.4 to −1.5) for girls and −8.9 (−16.5 to −1.2) for boys. The estimated fall in systolic pressure per 1 kg increase in birth weight was greater for multiplets than for singletons (for both sexes P<0.01, girls P=0.14, boys P=0.02). We found no significant association between birth weight and diastolic pressure.

We searched for factors that might explain the different effect size in multiplets and singletons. Gestational age did not influence systolic pressure independently of birth weight, and adjustment for it left almost unaltered the estimated effect of birth weight in multiplets They were more commonly delivered by caesarean (26 of 104 (25%) versus 127 of 778 (16.3%) singletons), and delivery of a further 15 multiplets was medically or surgically induced before full term. Removing these two groups made little difference: the birth weight effect among the 65 remaining multiple births was −8.9 mm Hg (−13.1 to −4.6) per 1 kg. Adjusting for number of previous pregnancies, breast feeding, maternal smoking and alcohol intake (prenatal and postnatal), or maternal education did not substantially change (by as much as 10%) the difference in the estimated effect of birth weight on blood pressure between multiplets and singletons.

Within pair analyses in multiplets

In multiplets the size of the birth weight effect was largely unchanged in within pair analyses in which we regressed the within pair difference in blood pressure on the within pair difference in birth weight (table 2). Estimates were generally larger for monozygotic than for dizygotic pairs, but differences were not statistically significant in this sample size. To provide greater depth to the analysis, we investigated different birth measures. The same pattern of effects and persistence of effect size after pairing was seen for crown to heel length. Associations were weaker for arm and head circumferences, body size ratios, and infant skin fold thicknesses (data not shown).

Implications of within pair analyses in multiplets

The association between birth weight and blood pressure, seen in multiplets as a group, was not diminished in within pair analyses This implies that maternal factors, including nutrient intake, cannot be directly responsible for the association between birth weight and blood pressure. The critical step in the causal pathway must concern mechanisms that lie within the fetoplacental unit, with maternal exposures operating only as antecedents or primers in the causal pathway. This is an important conclusion, which is contrary to the current paradigm in this research field. This key observation is supported by the work of Poulsen et al,18 who found an association between birth weight and type 2 diabetes in twins, which remained after within pair analysis.

Although the sample size (over 100 multiplets) provided sufficient power to determine whether the effect remained after pairing, it was insufficient to permit estimation of the difference in effect between monozygotic and dizygotic twins either before or after pairing. None the less, the strong association was comparatively unchanged in within pairs analysis in monozygotic twins, implying that the association between birth weight and blood pressure does not originate in the genes of the infant.

A possible explanation for the association is that birth size itself is a component of the principal causal pathway that leads to later high blood pressure and possibly other adverse metabolic outcomes. That the strength of the association between birth weight and blood pressure was greater in multiplets than singletons would, however, make that unlikely, unless some confounding or effect modifying factor was operating differentially in multiplets and singletons.

The different association in multiplets and singletons

The finding of a stronger association between birth weight and childhood blood pressure in multiplets than in singletons may provide a clue as to what may be responsible for the association between birth weight and blood pressure, but can we be confident of the observation? It cannot be accounted for by an unusually weak association in this study among singletons. The estimated effect of birth weight on systolic pressure in the 8 year old singletons, at around −2 mm Hg per 1 kg increase in birth weight, was within the range of effect sizes estimated in other studies of children of comparable age in Australia,19 the United Kingdom,20^–22 and Zimbabwe.23 We found slightly larger effects for singleton girls (−2.16 mm Hg per 1 kg) than singleton boys (−1.73 mm Hg per 1 kg), in common with other studies. 20 21 24 It is also unlikely to be explained by the sample selection process. Although singletons in the Tasmanian infant health survey were not representative of the total infant population, analyses were adjusted for factors used as selection criteria, and the estimated effect size in singletons was very similar to that in other populations. Multiplets were not subject to any selection bias as all born in the study period were eligible and the response at follow up was high (85.0%), as in singletons (74.9%).

The stronger association in multiplets raises the hypothesis that of the potential causes of low birth weight or later high blood pressure those that influence both were more prevalent in multiplets. What factors then within the fetoplacental unit could be associated with birth weight and blood pressure and also more prevalent in multiple pregnancies? Supply of oxygen and nutrients across the placenta can be unequally distributed between individual multiplets, and reduced oxygen or nutrient availability may be more prevalent causes of growth restriction in multiplets than in singletons and lead to permanent adverse programming of the cardiovascular system. We were unable to address these questions in any depth, but we did examine the association between birth weight and blood pressure in singletons before and after controlling for placental weight. The addition of placental weight to the regression model had little effect on the estimates, but the weight of the placenta may not reflect its function.