The role of vitamin D in bone health is well established, while evidence of its role in extra-skeletal health continues to accumulate [1, 2]. Vitamin D is derived from oral intake and from skin production by UVB radiation. Vitamin D derived from both sources is then hydroxylated in the liver to make 25-hydroxyvitamin D (25OHD) and again in the kidney to the most active form 1, 25-dihydroxyvitamin D. In Ireland the main contribution to dietary vitamin D is from meat, followed to a lesser extent by fish, spreads, eggs and milk and yogurts [3]. The different sources of vitamin D increase the complexity of research in the area. Production due to sun exposure depends on multiple factors including latitude and UltraViolent B (UVB) radiation levels, skin colour, sunscreen use, covering up and practices such as purdah [4]. Similarly, oral intake depends on vitamin D levels in plant and animal foods consumed, food fortification, supplement intake, and practices such as vegetarianism and veganism [5, 6].

During pregnancy, maternal 25OHD tends to decline, but maternal production of 1, 25-dihydroxyvitamin D increases because there is up regulation of renal enzymes in order to facilitate the doubling of maternal intestinal calcium absorption during pregnancy [7–9]. Vitamin D is not required by the fetus for calcium absorption because calcium is actively transported across the placenta; vitamin D becomes essential for calcium absorption only postpartum [8]. Offspring vitamin D status at birth may be determined by measuring umbilical cord 25OHD, which correlates with maternal 25OHD [8, 10, 11]. The placenta contains 1 α-hydroxylase which converts 25OHD to the active form but its contribution to maternal calcium balance is uncertain [8, 10, 12].

There is evidence that maternal and fetal vitamin D status or season of birth may be associated with offspring adiposity [13–17]. However, some studies have reported no such associations between maternal 25OHD and offspring health [18–23]. The aim of this study was to examine the association between maternal and neonatal 25OHD levels and infant size and adiposity at birth, 6–9 months and 2–2.5 years of age in a cohort from the ROLO (Randomised cOntrol trial of LOw glycaemic index diet versus no dietary intervention to prevent recurrence of fetal macrosomia) study [24]. In a previous analysis of a subgroup of the ROLO study, late maternal 25OHD above the cohort median was associated with significantly greater neonatal length than late maternal 25OHD below the median [25].

Mother and infant pairs from the ROLO study were included in this analysis at birth (n = 292), at 6–9 months postpartum (n = 160), and at 2–2.5 years postpartum (n = 287). Of these, 159 were assessed at all three time-points (birth, 6 months and 2 years of age) and had a measurement of maternal 250HD in early pregnancy; 166 were assessed at all three time-points and had a measurement of maternal 25OHD in late pregnancy; and, 120 were assessed at all three time-points and had a measurement of neonatal 25OHD. The ROLO study was a randomised control trial of low glycaemic index diet in pregnancy to reduce macrosomia. 800 euglycaemic secundigravid women who had previously given birth to a macrosomic baby (> 4 kg), and were therefore at increased risk of delivering another macrosomic infant [26], were randomised to receive low glycaemic index (GI) dietary advice or usual antenatal care, which did not include dietary advice. Detailed methodology and results of the ROLO study have been published previously [24, 27]. In brief, the primary outcome was a reduction in birthweight and the secondary outcomes were a reduction in gestational weight gain and glucose intolerance. The primary outcome was not achieved while the secondary outcomes were both achieved. Low GI dietary advice was given at week 14 of pregnancy and demographic, well-being and lifestyle questionnaires were returned by 28 weeks gestation. 3-day food diaries were completed during each trimester of pregnancy and used to determine the glycaemic index and glycaemic load of the women’s diets, both of which were significantly reduced following the intervention [24, 28]. This study was conducted according to the guidelines laid down in the Declaration of Helsinki with institutional ethics approval from ethics committee of the National Maternity Hospital, Ireland, for the original study, and further ethical approval from the same ethics committee for the follow-up study up to 2 years of age was obtained. Informed written maternal consent was obtained at each time-point of the study. Trial Registration with Current Controlled Trials www.controlled-trials.com ISRCTN54392969. Informed written maternal consent was obtained from all participants both during pregnancy and at the 2 year follow-up appointment.

The average 25OHD in our maternal cohort, both in early and late gestation, is nearly equivalent to the IOM EAR for healthy population at 40 nmol/L [2, 36] but 30% in early pregnancy and 37% in late pregnancy were at risk of vitamin D deficiency (25OHD <30 nmol/L). These estimates of risk of vitamin D deficiency are substantially higher than prevalence estimate of 7% in healthy Irish adults in the National Adult Nutrition Survey [6]. While we have noted a marked improvement in vitamin D status in our studies over the past three decades, there are still at-risk groups in Ireland such as pregnant women especially during wintertime [25, 37–41].

We found that birthweight was negatively associated with early pregnancy and neonatal 25OHD. Birth length was negatively associated with early and late pregnancy and neonatal 25OHD; however when birthweight was controlled for in these models, this association lost significance at all time-points. Neonatal sum of all skinfolds and sum of subscapular and triceps skinfolds, both measures of overall adiposity, were negatively associated with neonatal 25OHD. There was no association between early or late pregnancy 25OHD or neonatal 25OHD and 6–9 month old anthropometry. At 2–2.5 years of age the only association observed was a negative association between weight-for-age z-score and early pregnancy 25OHD.

Our finding of negative associations between early pregnancy 25OHD and neonatal 25OHD and birthweight is contrary to the findings of several other studies at birth. A recent study by Eckhardt et al. found that maternal 25OHD <30 nmol/L compared to maternal 25OHD ≥30 nmol/L at 20 weeks gestation was associated with lower weight-for-age z-score and BMI-for-age z-score at birth [17]. A study by Leffelaar et al. did not find any association between maternal 25OHD and birthweight once confounders were controlled for, but did find that maternal 25OHD <30 nmol/L compared to 25OHD >50 nmol/L was associated with greater risk of small for gestational age [42]. While another study by Morley et al. found no association between birthweight and early or late pregnancy 25OHD [43]. An older study by Phillips and Young similarly found that birthweight was not associated with month or season of birth although 25OHD was not measured [16]. The contrary nature of our findings of a negative association between maternal 25OHD in early pregnancy and offspring birthweight to those of other studies may be due, in part, to the nature of our cohort, in which there was a high prevalence of macrosomia. Our finding that late pregnancy 25OHD was not significantly associated with birthweight is similar to the study by Gale et al. of a cohort from Southampton which found no association between maternal 25OHD in late pregnancy and birthweight, and to another study by Sayers and Tobias which found that maternal UVB exposure in late pregnancy was not significantly associated with birthweight [23, 44].

Birth length was negatively associated with early and late pregnancy 25OHD and neonatal 25OHD; however, when birthweight was controlled for in these models, to determine if this association was simply a function of birthweight, this association lost significance at all time-points. This finding is similar to a study by Gale et al. found no association between late pregnancy 25OHD and birth length [23]. A study by Morales et al. of early pregnancy 25OHD likewise found no association [15]. Previous studies have also reported positive associations between birth length and maternal 25OHD levels in pregnancy [17] and with maternal UVB exposure [43, 44]. Fewer studies have examined the longer-term associations of maternal or neonatal 25OHD and offspring length. In this cohort, no association was observed between neonatal 25OHD or maternal 25OHD in early or late pregnancy and offspring length at 6–9 months or 2–2.5 years of age which is similar to the results of other longitudinal studies examining these associations at various ages in childhood [15, 23, 45]. Leffelaar et al. [42] also examined offspring height, up to 1 year of age and found that infants of mothers who were vitamin D deficient in early pregnancy were significantly shorter than infants of mothers who were not deficient at 1 month of age. Whereas by 1 year of age this relationship had reversed and the length of infants whose mothers were vitamin D deficient was significantly greater than those of mother without deficiency.

Neonatal sum of all skinfolds and sum of subscapular and triceps skinfolds, both measures of overall adiposity, were negatively associated with neonatal 25OHD in this cohort contrary to results of a recent study by Godang et al. where neonatal 25OHD was positively associated with neonatal fat mass measured by dual electron x-ray absorptiometry (DXA) [14]. Another study, also using DXA, found that late pregnancy 25OHD was positively associated with 3 week old fat mass, which the authors classified as birth adiposity, although neonatal 25OHD was not examined [13]. Finally, a study by Eckhardt et al. found that BMI-for-age z-score at birth was positively associated with maternal 25OHD at 20 weeks gestation [17].

In terms of offspring adiposity in later childhood, at 2–2.5 years old a negative association between weight-for-age z-score and early pregnancy 25OHD was observed in this cohort. Other studies that have examined the association between maternal or neonatal 25OHD and offspring adiposity in later life have reported conflicting results. Gale et al. found no association between late pregnancy 25OHD and offspring fat mass measured using DXA at 9 years of age [23]. While Crozier et al. found that late pregnancy 25OHD was negatively associated with offspring fat mass, also measured using DXA, at 6 years but not at 4 years of age [13]. Leffelaar et al. observed significantly greater offspring weight at 6 and 9 months of age in children whose mothers were deficient rather than sufficient in vitamin D in early pregnancy, however offspring adiposity was not examined [42]. A study of early pregnancy found that 25OHD <50 nmol/L was associated with increased BMI-for-age z-score and risk of overweight at 1 year of age but not at 4 years of age [15]. Whereas, Eckhardt et al. found that BMI-for-age z-score and weight-for-age z-score at 4 months and 1 year of age were not significantly associated with maternal vitamin D deficiency at 20 weeks gestation [17]. A study of 9.9 year old offspring found that maternal UVB exposure in late pregnancy was positively associated with offspring fat mass and weight [44]. An earlier study found that obesity at age 59–73 years was positively associated with winter pregnancies in men with a trend towards the same association in women [16].

This study had some limitations. Detailed data on maternal vitamin D supplement use and maternal sun exposure was not available, but we had data on season of pregnancy and actual measured 25OHD. In addition, as mentioned, complete anthropometric measurements were not available for all neonates although we had weight, length and head circumference for the full cohort which allowed examination of the association between maternal 25OHD and basic measures of neonatal size for the full cohort, with more in depth body composition analysis possible in a subgroup. Finally, data was not available on offspring diet at 2–2.5 years, precluding adjustment for the effect of dietary intake on size and adiposity. However, we did have data on infant feeding practices and duration of breastfeeding, which was controlled for in all analyses. Some variables controlled for in the multiple regression models may have been on the causal pathway such as: season, supplement intake or maternal BMI. However, these variables equally may have been confounders in their own right: season may affect birthweight independently of sunlight exposure [46]; supplement intake may have affected birthweight through supplementation with micronutrients other than vitamin D; and, maternal adiposity may lead to low 25OHD leading to altered birthweight. Regards the latter point, we did not find an association between maternal BMI and vitamin D status. We deemed it appropriate to control for any possible confounders while recognising the limitations of this strategy. There is no uniformity about the definition of season with respect to analysis of 25OHD. Despite improvement in vitamin D in Ireland over the past 40 years, a marked seasonality in 25OHD is still observed [47]. We chose summer/winter definition in order to align with our previous ROLO study analysis [25]. The results of our study may not be generalisable other cohorts due to relatively homogenous ethnicity, high education level and high prevalence of macrosomia and these factors, particularly the risk of macrosomia, which may explain the contradictory results when compared with other studies of maternal 25OHD and offspring size. Unfortunately, selection bias is an issue with all longitudinal studies due to loss to follow-up over time and it was an issue in this study too with mothers who returned at 2–2.5 years post-partum having a higher level of education, lower BMI in early pregnancy and lower smoking rate. However, these factors were controlled for where possible and, while further research is necessary, this study adds important data to the sparse literature on the association between maternal and neonatal 25OHD and offspring size and adiposity at birth and in childhood, and is unique in examining this association using early and late pregnancy and neonatal 25OHD.5.

Maternal and neonatal 25OHD were negatively associated with offspring size and adiposity at birth, and with offspring weight-for-age at 2–2.5 years in a cohort with high prevalence of macrosomia in whom risk of vitamin D deficiency was high. No association was observed with offspring anthropometry at 6–9 months of age. Further research is necessary to determine whether an optimum 25OHD level exists in pregnancy for promotion of healthy offspring size and adiposity. Meanwhile, improvement of vitamin D status in pregnancy remains a public health concern.