Low vitamin D status, as reflected by a circulating 25-hydroxyvitamin D (25(OH)D) concentration < 50 nmol/l(1), is a global public health issue that is widely prevalent among pregnant women across all WHO world regions(Reference Saraf, Morton and Camargo2,Reference van der Pligt, Willcox and Szymlek-Gay3) . In the UK, where vitamin D deficiency is defined as a 25(OH)D concentration < 25 nmol/l(4), the most recently available data from the National Diet and Nutrition Survey (years 9–11; 2016–2017 and 2018–2019) estimated 15 % of women aged 19–64 years have a vitamin D status that falls below this threshold(5); however, there is a lack of recent nationally representative data on vitamin D status among pregnant women in the UK.
Reported relationships between prenatal vitamin D status and maternal and offspring health outcomes are inconsistent, thereby challenging the concept of pregnancy-specific targeted thresholds for 25(OH)D(Reference Kiely, Wagner and Roth6). Recent and pooled data from observational studies suggest an association between low maternal 25(OH)D and increased risk of adverse outcomes including gestational diabetes mellitus (GDM)(Reference Sadeghian, Asadi and Rahmani7–Reference Agüero-Domenech, Jover and Sarrión9), preeclampsia(Reference Rouhani, Mokhtari and Lotfi10) and both preterm(Reference Zhao, Zhou and Wang11) and small-for-gestational age at birth(Reference Zhao, Zhou and Wang11). Although the benefits of routine prenatal vitamin D supplementation remain unclear(Reference Kiely, Wagner and Roth6,Reference Bialy, Fenton and Shulhan-Kilroy12) , maternal vitamin D status is a modifiable determinant of neonatal 25(OH)D(Reference Levy, O’Callaghan and Qamar13), and hence maternal vitamin D deficiency is a known risk factor for neonatal vitamin D deficiency.
Compared with a BMI within the ‘healthy’ range, a greater prevalence of vitamin D deficiency has been reported among individuals with obesity(Reference Vimaleswaran, Berry and Lu14,Reference Pereira-Santos, Costa and Assis15) , including pregnant populations(Reference McAree, Jacobs and Manickavasagar16,Reference Bodnar, Catov and Roberts17) . The inverse relationship between 25(OH)D and BMI has been attributed to both volumetric dilution and sequestration of vitamin D in adipose tissue(Reference Drincic, Armas and Van Diest18,Reference Walsh, Bowles and Evans19) , meaning a greater vitamin D intake may be required among individuals with overweight and obesity to achieve target 25(OH)D thresholds. Limited data from randomised trials suggest an inverse relationship between BMI and achieved 25(OH)D following intervention with vitamin D(Reference Moon, Harvey and Cooper20), such that greater BMI attenuates the slope of the vitamin D intake-25(OH)D response relationship during pregnancy(Reference Levy, O’Callaghan and Qamar13,Reference Alhomaid, Mulhern and Strain21) . In line with worldwide trends in overweight and obesity, the rising incidence of women who enter pregnancy with a BMI ≥ 30 kg/m2 is a global concern(Reference Devlieger, Benhalima and Damm22,Reference Chen, Xu and Yan23) . In England and Wales, 23 % of women with a recorded BMI were classified as having obesity at the first antenatal appointment in the years 2018–19(24). Earlier audit data (years 2015–2017) found variations in BMI across the main ethnic categories in the UK, with severe obesity (BMI ≥ 35 kg/m2) reported to be more common among women of White and Black ethnicity(Reference Relph25). However, the vitamin D status of pregnant women with obesity in the UK is not well characterised. Data are specifically lacking among ethnically diverse cohorts, despite previous reports from other European cohorts that clearly highlight lower vitamin D status in women of non-White ethnicity(Reference Richard, Rohrmann and Quack Lötscher26–Reference Eggemoen, Falk and Knutsen29). Understanding of the distribution of 25(OH)D and prevalence of deficiency is therefore required to inform evidenced-based guidelines for vitamin D intake in pregnant women and identify populations who would benefit most from targeted public health campaigns.
Among a large UK-based cohort, this study aimed to assess the vitamin D status of a multi-ethnic cohort of pregnant women with obesity and to examine the relationship between vitamin D status and perinatal outcomes.
Methods
Study design and setting
This study was a secondary analysis utilising biological samples and data from the UK Pregnancies Better Eating and Activity Trial (UPBEAT), a complex lifestyle intervention aiming to prevent GDM and reduce risk of large-for-gestational-age birth in 1554 pregnant women with obesity(Reference Briley, Barr and Badger30). UPBEAT was conducted in eight hospitals in inner-city settings across the UK. Ethical approval was obtained from UK IRAS (reference 09/H0802/5), and the trial was registered prospectively (ISRCTN89971375). The intervention, which encouraged improved dietary and physical activity behaviours, did not reduce risk of GDM or large-for-gestational-age birth(Reference Poston, Bell and Croker31), and for the purposes of this investigation, the trial was treated as a cohort study as there were also no differences in the 25(OH)D concentration between the intervention and standard care arms (online Supplementary Fig. 1).
Study participants
In the UPBEAT study, eligible participants were identified in antenatal clinic from general practitioner or midwife referrals. Women aged > 16 years with a BMI ≥ 30 kg/m2, singleton pregnancy and gestational age between 15+0 and 18+6 weeks were invited to participate. Women were excluded if unwilling or unable to provide informed consent, or if they had pre-existing diabetes, hypertension, renal disease, systemic lupus erythematous, antiphospholipid syndrome, sickle cell disease, thalassemia, coeliac disease, thyroid disease, current psychosis or currently prescribed metformin. Verbal and written information was provided to eligible women, and written consent was obtained(Reference Briley, Barr and Badger30). Inclusion in the present study was restricted to women for whom a baseline blood sample was available for measurement of serum 25(OH)D.
Demographic, clinical and pregnancy outcome data
Socio-demographic information was recorded at study entry, as collected through interview-administered questionnaires. Ethnicity was self-reported. Socio-economic status was assessed by Index of Multiple Deprivation (IMD), for which scores were calculated for the region of residence and presented as quintiles. The following anthropometric data were collected using standardised methods(Reference Briley, Barr and Badger30): maternal weight (kg) and height (cm); maternal hip, waist and thigh circumferences (cm) and maternal triceps (mm), biceps (mm), suprailiac (mm) and subscapular (mm) thicknesses, which were used to calculate the sum of four skinfold thickness (mm). Neonatal length and weight were measured within 72 h of birth. Customised birthweight centiles were calculated using Gestation Related Optimal Weight (GROW) software version 6·7·5·1 (Gestation Network, Perinatal Institute, Birmingham, UK; www.gestation.net), and large-for-gestational-age and small for gestational age delivery were defined as ≥ 90th and ≤ 10th percentile, respectively(Reference Briley, Barr and Badger30).
As per the UPBEAT protocol, diagnosis of GDM was defined according to the International Association of Diabetes and Pregnancy Study Groups criteria as one or more of the following: fasting capillary glucose concentrations of ≥ 5·1 mmol/l and/or 1 h venous glucose of ≥ 10·0 mmol/l and/or 2 h venous glucose of ≥ 8·5 mmol/l following an oral glucose tolerance test (OGTT). The UPBEAT protocol restricted diagnosis of GDM to participants who had an OGTT conducted between 27+0 and 28+6 weeks’ gestation; however, the present analysis pragmatically extended the timeframe to any OGTT performed between 23+2 and 30+0 weeks’ gestation in order to maximise the sample size for this outcome. Pre-eclampsia was defined as systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg or both, on at least two occasions 4 h apart, with proteinuria ≥ 300 mg/24 h or spot urine protein:creatinine ratio ≥ 30 mg/mmol or urine dipstick protein ≥ 2+.
Non-fasting venous blood samples were collected at the first study visit (second trimester), processed to serum within 2 h and stored at −80°C until analysis. The meteorological season of blood draw was assigned based on the date of the sample collection as follows: Winter – December, January, February; Spring – March, April, May; Summer – June, July, August; Autumn – September, October, November(32).
Laboratory analysis
Maternal HbA1c was measured at the University of Glasgow using a turbidimetric inhibition immunoassay on the Roche, Cobas c311 as previously described(Reference White, Lawlor and Briley33); low and high CV were 1·3 % and 1·4 %, respectively. Total 25(OH)D (sum of 25(OH)D2 and 25(OH)D3) was measured at the Institute of Cardiovascular and Medical Sciences at the University of Glasgow using an electrochemiluminescence-based automated clinical assay (Vitamin D Total assay kit #05894913 190, Roche Diagnostics, Mannheimm Germany) on a Cobas e411 analyser using the manufacturer’s standards and quality control material. This automated immunoassay method has been standardised against LC–MS/MS, which has been standardised to National Institute of Standardization and Technology (NIST) standard reference material(Reference Emmen, Wielders and Boer34,Reference Bjerg, Halgreen and Hansen35) . The assay limit of detection was 7·5 nmol/l; values below this threshold (n 27/1089; 2·5 %) were imputed at the level of the limit of detection. The low and high inter-assay CV were 11·2 % and 9·2 %, respectively.
In the present study, vitamin D deficiency and low vitamin D status were defined as a 25(OH)D concentration < 25 nmol/l and between ≥ 25 – < 50 nmol/l, respectively, to facilitate comparison with previous studies exploring maternal vitamin D status(Reference Saraf, Morton and Camargo2,Reference Alhomaid, Mulhern and Strain21) .
Statistical analysis
Data distributions were assessed using histograms and kernel density plots and summarised as mean ± sd or median (25th, 75th percentile), as appropriate. For categorical data, number (n) and percentage (%) are reported. Variables following a skewed distribution were either natural (ln)-log transformed or transformed to their base 2 log to approximate normality in regression analysis. The association between each maternal socio-demographic, clinical and anthropometric characteristic of interest and serum 25(OH)D in the second trimester was first examined using simple linear regression with log2-transformed 25(OH)D as the outcome variable given the right-skewed distribution of 25(OH)D. To create a parsimonious multivariable model, variables with P < 0·20 in the unadjusted analyses were entered into a general linear model and adjusted for IMD, ethnicity, maternal BMI at first study visit, maternal age at first study visit, season at blood sampling, gestational age at sampling and educational attainment, as appropriate to account for confounding of the exposure-vitamin D status relationship. The multivariable model included adjustment for BMI only, to avoid multicolinearity between BMI and other body size measures (i.e. weight and sum of skinfold thickness), whereby Pearson’s r ≥ 0·5 was used to define collinearity between pairs of continuous variables. Effect estimates were back-transformed and reported as mean percent differences with 95 % CI.
To facilitate meaningful comparison of effect sizes when examining the association between 25(OH)D and blood glucose measurements, both the independent (maternal 25(OH)D in second trimester) and dependent (HbA1c, fasting glucose and glucose measures at 1- and 2-h post OGTT) variables were standardised using a Fisher–Yates transformation to create a normally distributed variable with a mean of zero and an sd of one(Reference Armitage, Berry and Matthews36) before use in regression models; as such, the standardised regression can be interpreted as sd change in the outcome variables per sd change in serum 25(OH)D. Logistic regression was used for categorical outcome data and presented as OR with 95 % CI, using 25(OH)D ≥ 50 nmol/l as the reference catergory and 25(OH)D < 25 nmol/l and between ≥ 25 and < 50 nmol/l as the comparators. Multivariable models included adjustment for the intervention arm assigned at enrolment to the UPBEAT trial, as well as known demographic characteristics to be associated with 25(OH)D; ethnicity, educational attainment, maternal BMI at first visit, maternal age at first visit, season at blood sampling, gestational age at blood sampling and gestational age at delivery, as appropriate. In post hoc exploratory analysis, we created similar regression models fitted to data stratified by ethnic group to examine whether the magnitude of the effect estimates for the association between 25(OH)D and pregancy outcomes differed by ethnicity. Stratified analysis was conducted among women of White and Black ethnicity only, given the very low sample size in the other non-White ethnic groups.
Statistical analysis was conducted using Stata v17.0 (StataCorp, College Station), with significance set at P < 0·05.
Results
Population characteristics
Of 1091 participants with available 25(OH)D data in the UPBEAT cohort, two participants were excluded from the present analysis due to an early OGTT (13 weeks’ gestation) (n 1) and lack of information on the date of blood sample collection (n 1). Hence, 1089 participants were included in the present study, representing 70 % of the primary UPBEAT study cohort. The population characteristics of the full study cohort and stratified by thresholds of vitamin D status are shown in Table 1. The mean age of participants at enrolment was 30·5 ± 5·6 years and median BMI was 35·2 (32·7, 38·7) kg/m2. Over 75 % of participants were classified as having a relatively low socio-economic status based on the two highest quintiles of IMD. One-third of women identified as non-White ethnicity (33 %). The majority of participants lived in London (42 %), of which 43 % came from the most deprived areas, and just under two-thirds did not hold a university degree or equivalent (Table 1).
25(OH)D, 25-hydroxyvitamin D; GDM, gestational diabetes mellitus.
* n = 1086 due to missing data.
† Serum 25(OH)D measurements were unavailable for UPBEAT study participants recruited from community clinics or Sunderland City Hospitals Foundation Trust, and hence were not included in the present analysis.
‡ n = 1084 due to missing data.
§ n = 1080 due to missing data.
|| n = 1079 due to missing data.
¶ n = 1077 due to missing data.
** Calculated by sum of biceps, triceps, suprailiac, and subscapular skinfold thicknesses.
†† Obesity class I – BMI 30·0–34·9 kg/m2, Obesity class II – BMI 35·0–39·9 kg/m2, Obesity class III – BMI ≥ 40·0 kg/m2.
‡‡ n = 993 due to missing data.
§§ GDM diagnosis at 22+0–30+0 weeks’ gestation by oral glucose tolerance test.
There was an even distribution of blood samples drawn across all four seasons (Table 1). The median serum 25(OH)D concentration was 38·9 (24·5, 56·4) nmol/l. In total, 727 (67 %) women were classed as having low a vitamin D status (25(OH)D < 50 nmol/l) and more than one-quarter of women (26 %) had a 25(OH)D concentration < 25 nmol/l. The prevalence of vitamin D deficiency was highest among women of Asian ethnicity and lowest among women who identified as White (Table 1, Fig. 1).
Association of maternal characteristics with 25-hydroxyvitamin D concentrations
In both unadjusted and multivariable-adjusted linear regression models, maternal age was positively associated with serum 25(OH)D in the second trimester (Table 2). Participants who held a university degree had a greater 25(OH)D concentration than those without. Women living in the most deprived area had a lower 25(OH)D concentration compared with women with an IMD score < 5; however, effect estimates were attenuated in adjusted analysis such that the percent difference between IMD quintiles only remained significant for comparison between the third and fifth quintiles (Table 2). Compared with White women, 25(OH)D was lower among women of non-White ethnicity, for which the percent difference was greatest for women of Asian ethnicity (percent difference = 41 %; 95 % CI: −49, −31; P < 0·001); inferences were unchanged upon adjustment for covariates. Greater weight, BMI and skinfold thickness were associated with a lower 25(OH)D concentration in both unadjusted and adjusted models, but effect estimates were minor. Serum 25(OH)D was lower among women whose blood sample was taken in winter and spring compared with the summer months (Table 2).
* Effect estimates represents the percent difference in serum 25-hydroxyvitamin D per 1-unit increase in the predictor variable (continuous variables) or in comparison to the reference category (categorical variables).
† Adjusted for: Index of Multiple Deprivation, ethnicity, season of blood sampling, BMI, age, gestational age at blood sampling, educational attainment (university degree obtained or not). Adjustment for BMI was not included for anthropometric outcomes due to multicollinearity with weight and sum of skinfold thickness.
Association between maternal 25-hydroxyvitamin D and maternal outcomes
Compared with women with 25(OH)D concentrations ≥ 50 nmol/l in the second trimester, the occurrence of preeclampsia, preterm birth and both small for gestational age and large-for-gestational-age birth was similar in women with low vitamin D status (25(OH)D ≥ 25 – < 50 nmol/l), and vitamin D deficiency (25(OH)D < 25 nmol/l) (Table 3). In unadjusted analysis, the odds of developing GDM were greater in women with 25(OH)D < 25 nmol/l v. ≥ 50 nmol/l, but effect estimates were attenuated and rendered non-significant upon adjustment for maternal socio-demographic characteristics (Table 3). There was a minor negative association between 25(OH)D concentration and blood glucose measured at the 2-h OGTT time point that was attenuated albeit remained statistically significant in multivariable-adjusted analysis, such that each sd increase in 25(OH)D was associated with a 0·07 sd decrease in blood glucose (95 % CI: –0·14, –0·003; P = 0·042; n 931) (Table 4), which is equivalent to –0·11 mmol/l (95 % CI: –0·21, –0·005). Associations between 25(OH)D and either HbA1c or fasting glucose were not observed (Table 4).
* OR represents the probability of occurrence of the event in each category of vitamin D status compared to 25(OH)D ≥ 50 nmol/l. 25(OH)D, 25-hydroxyvitamin D; GDM, gestational diabetes mellitus; LGA, large for gestational age; OR, odds ratio; SGA, small for gestational age.
† Represents total number and percentage of women with outcome of interest.
‡ Adjusted for: ethnicity, season of blood sampling, BMI, maternal age, assigned UPBEAT intervention arm, gestational age at blood sampling and educational attainment (university degree obtained or not).
* The dependent and independent variables have been Fisher-Yates transformed to a normally distributed variables with a mean of zero and a standard deviation of one(Reference Armitage, Berry and Matthews36). The regression coefficients have been standardised and can be interpreted as sd change in the outcome variables per sd change in serum 25(OH)D. 25(OH)D, 25-hydroxyvitamin D.
† Adjusted for: ethnicity, season, BMI, maternal age, assigned UPBEAT intervention arm, gestational age at sampling and educational attainment (university degree obtained or not).
‡ Measurements taken as part of standard oral glucose tolerance test between 23+2 and 30+0 weeks’ gestation.
In subgroup analysis stratified by ethnicity, the magnitude and direction of effect estimates differed between White and Black women for most outcomes; however, the odds of having an adverse outcome at 25(OH)D < 50 nmol/l were not statistically significant in either ethnic group such that inferences from primary analysis were unchanged (online Supplementary Table 1). The relatively small sample size for Black women resulted in a low absolute number of diagnosed adverse outcomes which resulted in wide confidence intervals surrounding effect estimates and precluded estimation of the odds of preeclampsia and small for gestational age using 25(OH)D ≥ 50 nmol/l as the reference category. The negative association between 25(OH)D and blood glucose measured at the 2-h OGTT time point was not statistically significant in multivariable models in either White (mean difference = –0·07 sd/sd; 95 % CI: –0·16, 0·02; P = 0·14; n 631) or Black (mean difference = –0·13 sd/sd; 95 % CI: –0·29, 0·02; P = 0·10; n 187) women.
Discussion
Among an ethnically diverse cohort of pregnant women with obesity in the UK, our findings highlight an overall low vitamin D status, with more than two-thirds of women presenting with a 25(OH)D concentration < 50 nmol/l in the second trimester. While our findings do not support an association between maternal 25(OH)D in early-mid pregnancy and later clinical outcomes, we report a lower 25(OH)D among women of ethnic minority, highlighting a wide disparity in vitamin D status among population sub-groups within the UK. Of particular concern is the high prevalence of deficiency among women of Asian ethnicity, of whom > 50 % had a 25(OH)D concentration < 25 nmol/l, a threshold at which risk of nutritional rickets and osteomalacia is increased(1,4) . Given the resurgence in vitamin D-dependent rickets in recent decades, particularly among children of ethnic minority(Reference Goldacre, Hall and Yeates37,Reference Julies, Lynn and Pall38) , prevention of maternal vitamin D deficiency is an important consideration for ensuring adequate neonatal vitamin D status and protection against rickets in early infancy(Reference Munns, Shaw and Kiely39).
Current dietary reference values for vitamin D are not pregnancy specific but rather based on 25(OH)D targets for the maintenance of bone health, and as such, there has been little change in dietary vitamin D intake recommendations for pregnant women in the UK(4,40,41) . Furthermore, there is no specific national guidance for pregnant women with overweight or obesity. Within a cohort limited to women with obesity (BMI ≥ 30 kg/m2), we report an inverse association of 25(OH)D with BMI that remained significant after adjustment for selected maternal socio-demographic determinants of vitamin D status. The present study clearly shows a lower 25(OH)D among women at higher ends of the BMI distribution; only 23 % of women with severe obesity had a 25(OH)D concentration above the 50 nmol/l threshold. As neonatal 25(OH)D concentrations are dependent on maternal values in late gestation(Reference Levy, O’Callaghan and Qamar13), and human breastmilk is typically low in vitamin D(Reference Specker, Tsang and Hollis42,Reference Hollis, Roos and Draper43) , the high prevalence of vitamin D deficiency among pregnant women with obesity is a concern, particularly in an era where nutritional rickets remains a public health issue, both in the UK(Reference Goldacre, Hall and Yeates37,Reference Julies, Lynn and Pall38) and globally(Reference Creo, Thacher and Pettifor44).
The prevalence of vitamin D deficiency in our study population was higher in comparison to a previous UK pregnant cohort of BMI heterogenous women from Southampton(Reference Crozier, Harvey and Inskip45), yet comparable with baseline trial data among women with overweight and obesity in Northern Ireland(Reference Alhomaid, Mulhern and Strain21). Our findings suggest current national guidelines in the UK, which recommend a vitamin D intake of 400 μg/d(4), may not be adhered to, and/or may not be adequate for preventing vitamin D deficiency. While pregnancy-specific thresholds for 25(OH)D may be required, trial-derived dose–response data is needed to determine whether the nutritional requirement for vitamin D to meet existing target thresholds is the same for pregnant women at both ends of the BMI distribution. Data from Ireland(Reference O’Callaghan, Hennessy and Hull46), New Zealand(Reference Grant, Stewart and Scragg47) and Canada(Reference March, Chen and Karakochuk48) suggest a maternal 25(OH)D concentration of 50 nmol/l in late gestation is required to prevent neonatal vitamin D deficiency at the 25 nmol/l threshold. However, data from Northern Ireland has shown supplementation with 400 μg vitamin D3/d is not sufficient to raise 25(OH)D concentration > 50 nmol/l in women with overweight and obesity(Reference Alhomaid, Mulhern and Strain21). A greater vitamin D dose than currently recommended is therefore likely required in this population, particularly among women who enter pregnancy with a low vitamin D status.
In line with previously published data(Reference Richard, Rohrmann and Quack Lötscher26–Reference Eggemoen, Falk and Knutsen29), we report a greater prevalence of vitamin D deficiency among women of Black, Asian and non-White ethnicity compared with women who identify as White. As melanin hinders dermal synthesis of pre-vitamin D3 (Reference Clemens, Adams and Henderson49), darker skin pigmentation is a well-recognised risk factor for vitamin D deficiency for individuals living at northern latitudes(Reference Cashman, Dowling and Škrabáková50,Reference Martin, Gowda and Renzaho51) . While ethnic differences in vitamin D status are unlikely to be explained by variations in cutaneous production alone(Reference O’Callaghan and Kiely52), our findings reiterate the need for targeted public health messaging to prevent vitamin D deficiency among the populations who are most at risk. Given the limited sample size in the Asian and non-White ethnic groups, our post hoc analysis stratified by ethnicity was limited to White and Black women only. While 25(OH)D < 50 nmol/l was not associated with a greater odds of adverse outcomes in either ethnic group, we acknowledge imprecision of the effect estimates owing to the reduced sample size, and hence such findings should be considered as exploratory only.
We report a greater 25(OH)D among women with a university degree, yet the impact of socio-economic status on vitamin D status was less clear once additional socio-demographic factors were considered in the multivariable model. Despite the availability of funding schemes for low-income households, it was previously estimated that < 10 % of those eligible obtained free vouchers for vitamin D supplements for children and pregnant women in the UK(Reference Jessiman, Cameron and Wiggins53). Barriers such as complex ordering and reimbursement systems and limited locations from which supplements can be acquired have been reported(Reference McFadden, Green and McLeish54). Despite efforts to increase micronutrient intake among pregnant women of lower socio-economic status, such barriers to supplement use may therefore have contributed to a lower uptake and continued low vitamin D status among certain subgroups at the time blood samples were drawn for the present study.
The potential impact of vitamin D intervention on blood glucose regulation and its role in prevention of diabetes has been discussed in recent years(Reference Yaribeygi, Maleki and Sathyapalan55–Reference Mirhosseini, Vatanparast and Mazidi57). At present, pooled trial data show promising albeit conflicting evidence for an effect of vitamin D supplementation on prevention of GDM among generally healthy pregnancies(Reference Palacios, Kostiuk and Peña-Rosas58,Reference Roth, Leung and Mesfin59) . However, limited evidence from populations with obesity suggests little benefit of vitamin D supplementation. In the multicentre DALI vitamin D study(Reference Corcoy, Mendoza and Simmons60), the authors report a reduction in fasting plasma glucose in late gestation following a daily dose of 1600 μg vitamin D, but this did not translate to a reduction in GDM. The high frequency of personal micronutrient supplement use and high mean 25(OH)D concentration (> 50 nmol/l) at baseline limits generalisability of these findings to women with vitamin D deficiency(Reference Corcoy, Mendoza and Simmons60). Expression of the vitamin D receptor in pancreatic islet cells suggests a direct role for 1,25(OH)2D in glucose regulation. In vitro studies of mouse and human tissues provide mechanistic evidence linking vitamin D-mediated gene transcription to insulin secretion in response to glucose exposure. Specifically, 1,25(OH)2D upregulates the expression of voltage-gated calcium channels causing increased Ca influx to the cell, in turn stimulating insulin secretion from pancreatic β-cells(Reference Kjalarsdottir, Tersey and Vishwanath61). As with others, we did not find strong evidence for an association between 25(OH)D and glucose regulation(Reference Mendoza, Harreiter and Desoye62); compared with women with 25(OH)D ≥ 50 nmol/l, lower vitamin D status was not associated with an increased odds of developing GDM in the present study after adjusting for relevant confounders. Furthermore, while higher serum 25(OH)D was associated with a lower blood glucose concentration measured at the OGTT 2-h timepoint, the effect estimate was minor and unlikely to be clinically meaningful, and we acknowledge the possibility of type 1 errors owing to multiple testing. Given the lack of an association between 25(OH)D and fasting glucose, as well as glucose at the OGTT 1-h time point, we caution interpretation of these findings.
Strengths of this study include the large well characterised and ethnically diverse cohort of pregnant women with obesity and high levels of socio-economic deprivation, who are a high-risk group for adverse pregnancy outcomes. However, several limitations should be acknowledged. As the present study utilised data and biological samples from a previously reported intervention trial(Reference Poston, Bell and Croker31), the sample size was limited to participants with existing data, and we recognise that the study population may not be representative of the general UK population within the UK. We assessed 25(OH)D at a single time point during the second trimester and did not specifically collect data on dietary vitamin D intake or personal vitamin D supplementation use; 25(OH)D reflects habitual vitamin D intake from both cutaneous synthesis and dietary intake; however, 25(OH)D followed a skewed distribution in all ethnic groups and we therefore expect some participants to have taken supplemental vitamin D either alone or as part of a prenatal multiple micronutrient supplementation regimen. It is possible that the 25(OH)D measured at this time point is not reflective of changes in vitamin D intake later in pregnancy, which may impact pregnancy outcomes. Variability in the quality of 25(OH)D data across analytical methods has been discussed at length in the literature. We used a well-recognised automated immunoassay that was available at the time of 25(OH)D assessment. However, we acknowledge potential bias of this method due to cross-reactivity with other vitamin D metabolites, including the C3-epimer of 25(OH)D(Reference Carter, Jones and Shannon63). While only 2·5 % of samples were at or below the assay limit of detection of 7·5 nmol/l, this possible bias and relatively high limit of detection may have influenced precision of the effect estimates when modelling determinants of 25(OH)D. We used a statistical-based approach to select variables for covariate adjustment in multivariable models. While empirical methods are valid approaches for covariate selection, we acknowledge the advantages and increasing movement towards illustrative-based approaches (e.g. direct acyclic graphs) to identify confounding relationships(Reference Digitale, Martin and Glymour64). Lastly, while parathyroid hormone concentrations were not available for this cohort, the interactive effects of low 25(OH)D and elevated parathyroid hormone, representing functional vitamin D deficiency, may be a more meaningful indicator to explore rather than 25(OH)D alone(Reference Hemmingway, Kenny and Malvisi65). As the limited evidence relating functional vitamin D deficiency to an increased risk of hypertensive disorders and restricted fetal growth is mixed(Reference Hemmingway, Kenny and Malvisi65–Reference Scholl, Chen and Stein68), the possibility of a greater risk of adverse perinatal outcomes is worth further exploration in diverse cohorts.
Conclusion
In this cohort of women at high risk of pregnancy complications, our findings do not support a greater risk of GDM, pre-eclampsia, preterm birth or abnormal fetal growth among women with 25(OH)D concentrations below the conventional threshold of 50 nmol/l. However, our findings add to the increasing evidence that low vitamin D status is widespread among pregnant women with obesity in the UK, for which women of ethnic minority are most at risk. To meet current recommended thresholds for 25(OH)D, future dose–response trials are required to inform guidance for vitamin D intakes among pregnant women with obesity.
Acknowledgements
The authors thank all staff in the UPBEAT consortium and the participants in the trial for their patience, time, interest and goodwill.
This research was funded/supported by the National Institute for Health Research (NIHR) Biomedical Research Centre based at Guy’s and St Thomas’ The National Health Service Foundation Trust and King’s College London and/or the NIHR Clinical Research Facility. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health. The UPBEAT study was funded by UK’s National Institute for Health Research (RP-PG-0407-10452) and the Chief Scientist Office, Scottish Government Health Directorates (Edinburgh) (CZB/A/680), Guys and St Thomas’ Charity (1060508), Tommy’s Charity (SC039280), the NIHR Biomedical Research Centre at Guy’s and St Thomas’ The National Health Service Foundation Trust and King’s College London. Medical Research Council UK provided additional funding for the biomarker study (MR/L002477/1). The views expressed are those of the author(s) and not necessarily those of The National Health Service, the NIHR or the Department of Health.
A. C. F., S. L. W. and K. G. N. conceptualised and designed this study. K. G. N., O. F. Q., K. M. O’C. and K. V. D. analysed the data. A. C. F., K. G. N., K. M. O’C., K. V. D., L. P., J. R-F. and S. L. W. interpreted the findings. K. M. O’C., K. G. N. and A. C. F. wrote the manuscript. A. C. F. and S. L. W. have primary responsibility for final content. All authors have read and approved the manuscript.
The authors declare no conflict of interest.
The UPBEAT Scientific Advisory Committee accepts applications for use of data from this study upon request (www.medscinet.net/upbeat/).
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S0007114524000862