Main

Size at birth and rate of growth in infancy are important indicators of infant mortality and morbidity; smaller size being a major risk factor for mortality, particularly due to infectious disease (1, 2). More recently, early growth patterns have also been linked to adulthood metabolic and cardiovascular diseases that are exacerbated by excessive weight gain and obesity (3). Both small and large size at birth have been related to increased disease risk (4, 5), whereas childhood obesity and adult disease risk may be further programmed by rapid, or “catch-up,” infancy growth (6–8), however, the mechanisms surrounding these links remain elusive (9).

There is remarkably wide variation in rates of infancy growth (10). Its regulation is not well understood, though it is likely to involve appetite and nutrition, which may be influenced by determinants of fetal growth, as well as being influenced by choice of postnatal feeding practice (11). The effects of maternal smoking and parity on size at birth are well described (12, 13), however, their influence on postnatal growth is less clear (14–16).

We therefore examined the relationship of these factors, and also of early breast- or bottle-feeding, to size at birth and childhood growth between 0 and 5 y in a large representative birth cohort, the ALSPAC Children in Focus cohort.

METHODS

Subjects.

Recruitment of the ALSPAC cohort has recently been described (8). Briefly, ALSPAC is a geographically based cohort born between April 1991 and December 1992. The Children in Focus subcohort are a 10% random selection that includes 1335 term (gestation 37–42 wk) singletons who were measured at birth and on up on 10 further occasions until 5 y (at 4, 8, 12, 18, 25, 31, 37, 43, 49 and 61 mo). Informed consent was obtained from mothers and ethical approval was obtained from the ALSPAC ethical committee and from local ethical committees.

Data collection and auxological measurements.

Gestation was estimated using date of last menstrual period and confirmed by antenatal ultrasound reports; in cases of discrepancy, the data were reviewed by a single experienced clinician. Mother's parity and smoking during pregnancy were recorded by questionnaire completed during pregnancy. Maternal height was recorded from obstetric records and paternal height was self-reported by questionnaire completed by mother's partner during pregnancy. Birth weights were noted from hospital records and supine length was measured using a Harpenden neonatometer (Holtain Ltd., Crosswell, Dyfed, Wales, U.K.) soon after birth (median 1 d, range 1–14 d) by the ALSPAC study team. Weight (Seca 724 or 835 scales, Seca Ltd., Birmingham) and standing height (Leicester height measure, Child Growth Foundation) were measured at visits to the research clinics.

Calculations.

Sex and age independent SDS were calculated in each subject for all weight and length measurements by comparison with the U.K. 1990 growth reference (17) and also separately using internally derived means and SD for each sex at each age point: SDS = (measurement − population mean)/population SD. Postnatal catch-up growth was defined as an increase in weight SDS between birth and 2 y of >0.67 SDS (8); 0.67 SDS is equivalent to the distance between adjacent centile lines drawn on standard growth curves (i.e. 2nd, 9th, 25th, 50th, 75th, 91st, and 98th centiles); at birth, this corresponded to a difference of 320 g in weight and 1.4 cm in length. Ponderal index (weight/length3) was calculated at birth; this was an appropriate weight-for-length index at birth as this parameter was independent of length.

Analyses.

To analyze this large longitudinal data set and take into account subjects with missing measurements, we used a multilevel modeling software (MLwiN ver. 1.0 beta, Institute of Education, London, U.K.) (18). This is a hierarchical regression model, which allows consideration of individual growth curves and their summation by predefined groups. Using the whole data set from birth to 5 y, variance in weight or length/height SD scores was modeled against age. To recognize growth data for each child, two levels of hierarchy in the data were used: level 1 consisted of all measurement points, and level 2 consisted of unique codes for each child. Models with the categorical explanatory variables maternal smoking, parity, and breast-/bottle-feeding at 3 mo were built separately, and from these models graphs of the predicted values by age between birth and 5 y were drawn.

The use of separate models to analyze maternal smoking, parity, and breast-/bottle-feeding was chosen for simplicity, and was justified by the absence of association between these factors. There was no association between maternal parity and smoking (15% of primiparous mothers smoked, compared with 16% of nonprimiparous mothers, p = 0.8); nor was there a significant association between parity and breast-/bottle-feeding (44% of primiparous mothers breast-fed, compared with 48% of nonprimiparous mothers, p = 0.2). The only significant association seen was between maternal smoking and breast-/bottle-feeding (9% of mothers who breast-fed were smokers, compared with 17% of mothers who bottle-fed, p < 0.0005); therefore breast- and bottle-fed infants were compared after excluding maternal smokers. Furthermore, at each age point, cross-sectional multivariate analyses including maternal smoking, parity, and breast-/bottle-feeding together were used to examine independent differences in size and to allow further adjustments for parental heights, using generalized linear models (SPSS v.7.0, SPSS, Chicago, IL, U.S.A.). A p value < 0.05 was considered to define a statistically significant difference.

RESULTS

All 1335 children were measured at birth and also on at least one occasion during the first year; 1225 were measured at least twice in the first year. In total, 945 children were measured at 4 mo, 1231 at 8 mo, 1164 at 12 mo, 1088 at 18 mo, 1049 at 25 mo, 1053 at 31 mo, 1003 at 37 mo, 999 at 43 mo, 963 at 49 mo, and 929 at 61 mo. Multilevel modeling analysis allowed growth data from all 1335 children to be used.

Comparison with the U.K. reference.

Expressed as U.K. 1990 SDS, mean length/height remained close to the national average between birth (mean ± SD: 0.09 ± 1.01 SDS) and 5 y (0.11 ± 0.94 SDS), however weight increased progressively from birth (−0.01 ± 1.01 SDS) to 5 y (0.29 ± 1.00 SDS) (Fig. 1). This trend toward increasing weight-for-height with age in this cohort has previously been reported (19). To examine within-cohort growth determinants, we calculated internal SDS for weight and height and these have been used in all subsequent analyses.

Figure 1
figure 1

ALSPAC growth curves expressed as 1990 UK Reference weight and length/height SD scores (± 95% confidence interval).

Full size image

Maternal smoking.

One hundred ninety-nine mothers admitted to smoking during pregnancy, 1100 mothers denied smoking, and no information was available in the remainder. Compared with infants of nonsmokers, at birth, infants of smokers were lighter, shorter, and had smaller head circumferences but were no different in weight-for-height (ponderal index p = 0.7) (Table 1). Over the first year of life these infants showed catch-up growth (Fig. 2), and by 12 mo there was no difference in weight (p = 0.9), length (p = 0.2) or head circumference (p = 0.2) between the two groups. The remaining small deficit in height from 12 mo onward in children of smokers, apparent in Figure 2, did not reach statistical significance at any age on cross-sectional analyses.

Table 1 Size at birth and gestational age by maternal smoking and parity (means ± SD, adjusted for sex)
Full size table
Figure 2
figure 2

Internally derived SD scores (± 1 SEM) for weight and height grouped by maternal smoking in pregnancy.

Full size image

Parity.

Maternal parity had a large influence on size at birth. Infants of primiparous pregnancies were lighter, shorter, had smaller head circumferences and were also thinner (lower ponderal index) at birth compared with other infants (Table 1). During the first year of life, infants of primiparous mothers showed rapid catch-up growth in weight and length, overshooting other infants and they were significantly heavier and longer from 12 mo onward (Fig. 3). Their height advantage persisted after further adjustment for parental heights. Catch-up in head circumference was less rapid and there was no significant difference in head circumference by parity from 8 mo onward (data not shown).

Figure 3
figure 3

Internally derived SD scores (± 1 SEM) for weight and height grouped by maternal parity.

Full size image

Breast-/bottle-feeding.

Of the mothers who denied smoking in pregnancy, 425 reported they were still breast-feeding their infants at 3 mo of age and 440 were bottle-feeding. There was no difference in size at birth between infants who were breast- or bottle-fed. However, there was a marked divergence in growth during infancy, as breast-fed children showed slower weight and length gain than bottle-fed children (Fig. 4). The difference in weights between these groups was prolonged and was still significant at 31 mo of age (p = 0.02). Mothers who chose to breast-feed had significantly higher educational qualifications (58% had A-levels or university degrees) than mothers who bottle-fed (28%, χ2 test:p < 0.0005), however mother's educational achievement did not itself predict weight at birth (p = 0.99), 2 y (p = 0.7), or 5 y (p = 0.8). No difference in head circumference was seen at any age between breast- and bottle-fed infants.

Figure 4
figure 4

Internally derived SD scores (± 1 SEM) for weight and height grouped by breast- or bottle-feeding.

Full size image

Sex influence on catch-up growth.

The numbers of infants showing catch-up growth were similar in males (142/569, 25%) and females (139/480, 29%), however catch-up males showed a faster initial rate of gain in weight SDS (2.95 ± 0.08 SDS/y) than females (2.69 ± 0.11 SDS/y, p = 0.002).

DISCUSSION

Using longitudinal analysis of growth data from a large prospective birth cohort, we observed marked and prolonged effects of maternal smoking, parity and infant breast- or bottle-feeding on early childhood growth rates.

Multilevel modeling allows analysis of longitudinal or repeated measures data sets where individuals may have missing measurements on one or more occasions (18). A limitation in using polynomial functions in the fitting of these growth curves, particularly with the lower frequency of measurements with increasing age, was a tendency to exaggerate inflections, possibly resulting in an artificial “meandering” in the growth curves between 3 and 5 y. The robustness of the model in the first 3 y of life underlines the marked effects of antenatal and early infancy factors on growth rates in early childhood.

Our data confirm reports of the fetal growth-inhibiting effects of maternal smoking in pregnancy (12, 13). The symmetrical reduction in size at birth in infants of smoking mothers suggests that this growth inhibition was present from early fetal development. Earlier and smaller studies have debated whether any size deficit persists into childhood (14, 15, 20). In our study, infants of smoking mothers showed complete compensatory postnatal catch-up growth in weight, and there was no significant remaining height deficit compared with other infants, however, it is still possible that deficits in bone mineral density and susceptibility to respiratory disease persist (14, 20).

Infants of primiparous pregnancies were smaller and thinner at birth but showed dramatic catch-up in weight and length that overcompensated for their initial size deficit and resulted in larger childhood size, even allowing for their parents size. Fetal growth restraint with relative catch-up growth in first pregnancies, and corresponding catch-down growth in infants of high parity, have been described in rural Gambian infants, where it was suggested that availability of postnatal nutrition may be affected by family size (16). In our contemporary, relatively affluent birth cohort (21), the marked variations in postnatal growth even with small increases in parity suggest that a more intrinsic process may be involved.

The mechanisms that signal and regulate postnatal catch-up growth are unknown, however, our current data support the hypothesis that postnatal growth rates are influenced by maternal-uterine factors during antenatal life. Although it is possible that parenting behaviors may differ substantially between primiparous and nonprimiparous mothers, infant nutrition is also largely dependent on appetite. We have previously shown that catch-up growth was predicted by low leptin levels at birth (22). Leptin inhibits appetite and weight gain in humans (23, 24) and these findings support the observation that variation in infancy growth rates may be to some extent mediated through differing preset levels of satiety (11). Thus, in affluent societies where nutrition is plentiful, lower leptin levels leading to increased appetite may explain the excessive postnatal growth seen in thin, in utero restrained infants of primiparous pregnancies.

The importance of nutrition on early childhood growth rates is further emphasized by the marked differences in growth rates between infants who were breast- or bottle-fed at 3 mo. The lack of difference in size at birth excludes any initial size-based selection bias in choice of feeding. Bottle-fed infants have higher total energy and protein intake than breast-fed infants (25), and this may stimulate increased insulin secretion (26). We observed that the growth-limiting effects of breast-feeding continued well beyond infancy, and this is consistent with recent data describing protective effects of breast-feeding against obesity risk in later childhood (27). Mothers who breast-fed had higher educational achievements than mothers who bottle-fed. Although maternal education itself was not related to size at birth or postnatal growth, it is possible that decision to breast-feed was related to other maternal behaviors or choices in offspring feeding. Alternatively, early breast- or other feeding behaviors may directly influence later appetite regulation (28). This may possibly be mediated through variations in leptin sensitivity, as recently shown in rodent studies (29).

Lastly, catch-up growth was more rapid and accomplished sooner in males than in females. This could reflect the anabolic effects of early infancy sex hormone production in males (30, 31). It is also possible that nutrition-driven postnatal catch-up growth could be modified by other genetic determinants of hormone secretion or metabolism in the infant, such as polymorphisms in the estrogen receptor (32) or insulin genes (33, 34).

In conclusion, these data suggest that postnatal growth rates are strongly influenced by a drive to compensate for antenatal restraint or enhancement of fetal growth by maternal-uterine factors. The sequence of fetal growth restraint and postnatal catch-up growth may predispose to obesity risk in this contemporary population (8).