Vitamin D levels in an Australian and New Zealand cohort and the association with pregnancy outcome

This study utilised data collected from the multi-centre, prospective cohort Screening for Pregnancy Endpoints (SCOPE) study [28]. Nulliparous women carrying a singleton pregnancy were recruited between November 2004 and September 2008 in Adelaide (Australia) and Auckland (New Zealand). Ethics approval was obtained from Central Northern Adelaide Health Service Ethics of Human Research Committee on 2 September 2005 (ethics number REC 1714/5/2008) and Northern Region Ethics Committee, in Auckland on 23 April 2003 (ethics number AKX/02/00/364). All participants provided written informed consent. At 15 ± 1 weeks’ gestation, women were interviewed by a research midwife and asked questions on maternal demographics and lifestyle and had physical measurements taken, including height and weight. Records included ethnicity, age, body mass index (BMI), socioeconomic index (SEI) and multivitamin use [29]. BMI was calculated as weight (kg)/height² (m²). Obesity was defined as BMI ≥30 kg/m², overweight as BMI ≥25 and < 30 kg/m², normal weight as BMI > 20 and < 25 kg/m² and underweight as BMI ≤20 kg/m². SEI was calculated using the New Zealand SEI of occupational status, deriving a number between 10 and 90 based on the woman’s occupation; a higher number indicates higher socioeconomic status [30]. This population of women was predominantly Caucasian and thus maternal ethnicity was categorised into 2 main groups; Caucasian and non-Caucasian. Uncomplicated pregnancies were defined as those without any pregnancy disorder who delivered an appropriate weight for gestational age infant at term (≥37 weeks gestation) [31]. Pregnancy complications studied included PE, gestational hypertension (GH), GDM, sPTB and SGA and have been previously defined [28, 29, 31, 32].

Non-fasting whole peripheral blood samples were collected into non-heparinised tubes at 15±1 weeks’ gestation. Serum was processed within 4 h of collection and stored at − 80 °C until required. Unlike the previously published data on vitamin D in the Auckland cohort [20], serum 25(OH)D was measured using the IDS-iSYS chemiluminescent-based assay (Abacus, ALS) as per the manufacturer’s instructions. Average intra- and inter-assay coefficient of variation (CV) were 5.4 and 8.8%, respectively. Both serum 25(OH)D₃ and 25(OH)D₂ were measured independently and combined to provide a total 25(OH)D in nmol/L [27].

Statistical analyses were performed in R (v3.1.1) [33]. Data were checked for normality using a Shapiro-Wilks test and differences between women recruited in Adelaide compared to Auckland were tabulated and compared using a Welch’s t-test (continuous variables) or Fisher’s exact test (categorical variables).

Given that the month in which the serum was sampled heavily influences vitamin D status, “standardised” serum 25(OH)D concentrations were calculated as previously described [34] in order to normalise against seasonal variation. Briefly, vitamin D concentrations were standardized by taking the difference of 25(OH)D concentration to the average 25(OH)D concentration of the corresponding month. This 25(OH)D concentration was then added to the overall population mean 25(OH)D concentration: 68.09 nmol/L. Quartiles, based on the distribution amongst the population of women studied, in the standardised concentrations were then used to create cut-points and designated ‘low’ (< 44 nmol/L), ‘low-moderate’ (44–63 nmol/L), ‘moderate-high’ (63–81 nmol/L) and ‘high’ (> 81 nmol/L) categories of serum 25(OH)D.

Generalised linear models (Poisson with log link and robust variance estimates) were used to calculate the risk ratios for pregnancy complications by standardised serum 25(OH)D concentrations calculated among the women who had an uncomplicated pregnancy. Potential confounders for vitamin D status were assessed using linear modelling. Maternal age, BMI, SEI, alcohol consumption at 15 ± 1 weeks’ gestation (never/former vs. current), recreational walking (1–3 times/week and ≥ 4 times/week vs. never), ethnicity (Caucasian vs. non-Caucasian) and recruitment site (Adelaide vs. Auckland) were significantly associated with serum 25(OH)D and along with smoking status at 15 ± 1 weeks’ gestation (never/former vs. current) included as main effects within the generalised linear models. These analyses were also repeated but using the current definitions of vitamin D deficiency: < 25, 25–50, 50–75 and > 75 nmol/L in the non-standardised data [7]. We also stratified based on fetal sex to evaluate possible effect measure modifications according to whether the mother was carry a male or female fetus.

Of 3229 women recruited as part of SCOPE at the Adelaide and Auckland centres, serum samples at 15 ± 1 weeks’ gestation to measure 25(OH)D were available for 2800 (87%) women of whom, 1156 (41%) were recruited in Adelaide and 1644 (59%) were recruited in Auckland. Maternal characteristics are shown in Table 1. Compared to women who were recruited at the Adelaide site, women recruited in Auckland were older, had a lower BMI, less likely to smoke or drink alcohol during pregnancy and more likely to eat fruit and undertake recreational walks. Mean ± SD SEI of the Auckland women was also higher compared to that in women recruited in Adelaide (Auckland: 48.0 ± 14.8 vs. Adelaide: 27.7 ± 10.5, P < 0.001). Given that all these factors significantly influenced serum 25(OH)D in the linear regression model, it was unsurprising that vitamin D status of the women was significantly different between the two recruitment sites. However, after adjusting for maternal age, BMI, SEI, smoking status and alcohol consumption at 15 ± 1 weeks’ gestation, ethnicity, recreational walking and season, the women recruited in Adelaide still had significantly lower serum 25(OH)D when compared with those in Auckland (Adelaide: 58.4 ± 50.3 vs. Auckland: 70.2 ± 54.5 nmol/L, P < 0.001) indicating the influence of other confounders not measured as part of the study on vitamin D status.

As expected, there was a seasonal influence on serum 25(OH)D and thus, “standardised” vitamin D concentrations were calculated to account for the month of serum collection. Average hours of sunlight per day in Adelaide from September 2005 to September 2008 were obtained from the Australian Government Bureau of Meteorology and compared against serum vitamin D measured in the women recruited in Adelaide (Fig. 1a; black line/left axis serum 25(OH)D and grey line/right axis average hours of sunlight). The means for serum 25(OH)D by month of sampling were also separated for women recruited in Adelaide and women recruited in Auckland (Fig. 1b).

Of the 2800 women present in this study, 1217 (43%) women went on to develop a pregnancy complication and included 5.8% (n = 161) who developed PE, 7.6% (n = 213) who developed GH, 3.3% (n = 92) who developed GDM, 5% (n = 139) who spontaneously delivered preterm and 10.6% (n = 298) who delivered an SGA infant.

The association between “standardised” 25(OH)D with pregnancy complications is presented in Table 2. Using moderate-high as a reference (standardised vitamin D 63–81 nmol/L), there was no appreciable effect of having a low, low-moderate or high vitamin D with the risk of developing any pregnancy complication after adjusting for confounders. However, when each pregnancy complication was analysed separately, a 53% decreased risk for GDM was observed with high vitamin D status when compared to moderate-high status (aRR: 0.47; 95% CI: 0.23, 0.96). When women were categorised based on clinical definitions of vitamin D deficiency and serum 25(OH)D between 50 and 75 nmol/L used as the reference, no significant relationship between vitamin D status and adverse pregnancy outcome was found (Table 2). Although, the point estimates indicated a marginal increased risk for developing any pregnancy complication with severe (< 25 nmol/L) vitamin D deficiency compared to those who had levels between 50 and 75 nmol/L (aRR: 1.10; 95% CI: 0.89, 1.36).

As there is evidence to suggest that vitamin D metabolism within the placenta may differ with respect to fetal sex [35] thus, we assessed the effect of “standardised” vitamin D status on pregnancy outcome stratified by fetal sex (Table 3). Although not statistically significant, point estimates for women carrying a male fetus indicated decreased risk of having any pregnancy complication with high serum 25(OH)D compared to moderately-high (aRR: 0.86; 9%% CI: 0.71, 1.04). This was largely driven by a 55% decreased risk of GDM in women with a male fetus with high standardised serum 25(OH)D when compared to those carrying a male fetus with moderate-high serum 25(OH)D (aRR: 0.45; 95% CI: 0.15, 1.38). Conversely, if a woman was carrying a female fetus, there was generally no effect of having low, low-moderate or high standardised serum 25(OH)D on developing any pregnancy complication compared to moderate-high levels (Table 3). Although, a 56% decreased risk for GDM was observed in those with low-moderate levels of standardised vitamin D when compared to moderate-high levels for women carrying a female fetus (aRR: 0.44; 95% CI: 0.20, 0.97). Similar results were observed when vitamin D status was based on clinical definitions of deficiency (Additional file 1: Table S1).