This study was performed at Ewha Womans University Hospital, a representative hospital for women in Korea. The Ewha Birth & Growth Cohort was constructed. From August 2001 to December 2004, a total of 800 apparently healthy women receiving antenatal care between 24 and 28 weeks of gestation at the hospital agreed to participate. Exclusion criteria were deliveries at other hospitals, pregnancy-induced hypertension and gestational diabetes, and no biological sample from the mothers. Among a total of 593 pregnant women, antioxidant vitamins and oxidative stress levels were measured and a baseline interview was conducted to obtain parental sociodemographic and anthropometrical characteristics on antenatal visits.

The cohort was constructed for 593 infants to whom enrolled pregnant women gave birth. Excluding twin births, preterm births, and congenital malformations, a total of 383 followed-up infants were analyzed.

This study was approved by the Institutional Review Board of Ewha Womans University Hospital, and informed consent was obtained from each woman at enrollment.

The primary exposure variables were levels of maternal antioxidant vitamins and oxidative stress status in serum and urinary samples between 24 and 28 weeks of gestation. All fasting venous blood and urine samples obtained from the subjects were aliquoted and stored at −70°C until analysis. Plasma vitamin A and E levels were measured by high-performance liquid chromatography (HPLC) using a reversed-phase column and an ultraviolet detector. For vitamin C analysis, plasma was mixed with 5% metaphosphoric acid, vortexed, and centrifuged for 10 min at 3,000×g. The supernatant was filtered using a 0.45 µm syringe filter, and 10 µL samples were subjected to HPLC analysis. During the same period, the oxidative stress status was assessed by measuring the spontaneous urinary concentrations of 8-OHdG (8-hydroxydeoxyguanine) and MDA (malondialdehyde). The amount of urinary 8-OHdG was determined using a competitive in vitro ELISA (JICA, Fukuroi, Japan) for quantitative measurement of the oxidative DNA adducts. MDA was determined by HPLC for the adduct obtained with thiobarbituric acid (TBA).

For vitamins A and E, the intra- and interassay coefficients of variation were 3 and 6%, and those for vitamin C were 5 and 9%, respectively. The urinary levels of 8-OHdG and MDA are expressed as µg/g creatinine and µmol/g creatinine, respectively.

After delivery, trained nurses in the delivery room routinely measured and recorded the birth weight, height, head circumference, and gestational age to compile the offspring data, and obtained the maternal age, weight, and height. Gestational age was assessed by the mother's estimated date of last menstrual period or by ultrasound measurement. Birth weight was measured to the nearest 50 g, and the crown–heel length and head circumference were measured to the nearest 0.1 cm. Twin births, maternal hypertension, gestational diabetes, and major congenital malformations were also recorded.

For measurements after birth, infants were followed up at 6, 12, 18, 24, and 36 months through outpatient clinic visits. When measurement at the clinic was not possible, a questionnaire asking for information on weight, height, head circumference, and general characteristics of the infant was mailed.

A Bland–Altman assessment (8) was used to assess the degree of agreement between these two measurement methods. The range of agreement was defined as the mean bias ±2 SD. The 95% limits of agreement between the clinical measurements and self-reported measurements ranged from −3.2 to 5.3 in height and from −1.2 to 1.5 in weight. The fact that the frequencies exceeding the difference between two measurements were 5.9% for height and 4.7% for weight indicated that the self-reported measurements could be used as a substitute for direct measurements. Weight was measured to the nearest 0.01 kg, and height and head circumference were measured to the nearest 0.1 cm. Moreover, age- and sex-specific percentiles for weight, height, and head circumference based on the Korean National Growth Curves (9) were included using the LMS method (Lambda, mu, sigma) (10). The world health organization (WHO) Child Growth Standards were used for comparison with the international standards of weight and height (11).

By mailed questionnaire, the following information on factors that may affect infant growth and maternal vitamin and oxidative stress levels was collected: first weaning month, breastfeeding, mother's employment status, secondhand smoke exposure, caregiver (mother or other individual), infant dietary supplementation, hospital admission history, parental education period, and household income.

Levels of maternal vitamins and oxidative stress were dichotomized into low/high groups at the 25th or 75th percentile as determined by the distribution of serum or urinary concentrations among the entire cohort. The rationale for using different percentiles is that for vitamin A as the 25th percentile, a deficiency effect was important; for vitamins C and E as the 75th percentile, a supplementing effect was important; and for 8-OHdG and MDA as the 75th percentile, an excess effect was important.

Between the two low/high groups, the mean infant weight, height, head circumference, weight percentile, height percentile, and head circumference percentile at birth and at 6, 12, 18, 24, and 36 months were assessed using the PROC Mixed procedure in the Statistical Analysis Systems statistical software package version 9.1 for Windows (SAS Institute, Cary, NC, USA).

The Student's t test and chi-squared test were conducted to select confounding factors. Finally, we selected the five potential confounding variables that provided valid and precise results.

A total of 383 infants whose mothers’ vitamin and oxidative stress levels were measured were followed up at least once between birth and 36 months (follow-up rate of 64.5%). Table 1 lists the number of infants followed at birth and at 6, 12, 18, 24, and 36 months as well as their mean weight, height, and head circumference. Figure 1 shows that the weight, height, and head circumference from birth to 36 months were all higher in male than in female infants. All measurements were within the normal ranges based on comparison with the Korean National Growth Curves (Fig. 1) and WHO Child Growth Standards (9, 11). The levels of maternal antioxidant vitamins and oxidative stress and the general characteristics of the mothers, fathers, and infants, are shown in Table 2.

A total of 383 infants were followed up, and 124 infants dropped out during follow-up. The differences among these infants were in the maternal vitamin A level (p=0.01) and caregiver (mother or other individual) (p<0.0001) (Table 3).

Because gestational age and infant sex were thought to be the primary confounding variables for the association between maternal vitamin levels and infant growth, these two variables were adjusted first (Table 4). A high maternal concentration of vitamin C was significantly associated with increased infant weight (p=0.02) and head circumference (p=0.03) during the first 3 years of life. Conversely, a high maternal MDA was associated with decreased infant weight (p=0.01) and height (p=0.01). Similarly, infants with a high maternal concentration of vitamin E showed a lower weight and height, but only the association with height was statistically significant (p=0.02). Next, other potential confounders including gestational age, infant sex, breastfeeding, period of breastfeeding, household income, and dietary supplementation were adjusted in the final model. After adjustment for these variables, the group with low maternal vitamin A presented a lower head circumference percentile than the group with high maternal vitamin A (p<0.01). Likewise, the vitamin C group showed a similar pattern in head circumference percentiles, but there was no statistical significance (p=0.34); they also weighed less than subjects in the high vitamin C group (p=0.01) (Table 5). In the mixed model, there was a significant interaction between maternal vitamin A intake and age in terms of head circumference when adjusting for confounding variables (Fig. 2). This denotes that these indices increased to a greater extent in the high vitamin A and C groups than in the low group as individuals became older. In particular, the effect of age on increased weight was modified by maternal antioxidative vitamin levels in that a stronger effect was seen in the group with high vitamin C intake than in the group with low vitamin C intake. The anthropometric measurements of oxidative stress in the high oxidative stress group were lower than those in the low oxidative stress group, but the differences were not statistically significant.