Vitamin D supplementation in pregnancy: A systematic review

Photosynthesis and metabolism of vitamin D

Vitamin D is a secosteroid which is synthesised in the skin by the action of sunlight. It plays a crucial role in bone metabolism and skeletal growth. 42 Around 95% is acquired via photosynthesis in the skin, with the minority from the diet. 43 There are two dietary forms: D₂, from plants, and D₃, from animals (the latter mainly found in oily fish and fortified margarines and breakfast cereals). 43 Vitamin D is synthesised from the action of sunlight (wavelengths 290–315 nm) on cutaneous 7-dehydrocholesterol, converting it to pre-vitamin D₃. 10,42 Once formed, pre-vitamin D₃ undergoes membrane-enhanced temperature-dependent isomerisation to vitamin D₃,42 which is translocated into the circulation, where it binds to vitamin D-binding protein (DBP). 10 The main determinant of vitamin D synthesis in the skin is the level of sun exposure. The total amount of energy accrued from sunlight is dependent on duration and extent of skin exposure, but also on latitude and season. Thus, pigmented skin and covering, particularly relevant to the dark-skinned, and potentially covered, ethnic minority groups in the UK, reduce synthesis; using sun block with a factor higher than 8 almost completely prevents formation of vitamin D. 43 At latitudes of 48.5° (Paris, France), the skin is unable to form vitamin D between the months of October through to March. 42 In northern latitudes this results in a seasonal variation in levels of vitamin D, with a peak over the summer months and a trough in the winter. 10 Use of sunscreen during the summer may prevent adequate synthesis of vitamin D and subsequent storage in fat for the winter months, thus leading to deficiency; greater adiposity is also associated with reduced levels. 10 Circulating vitamin D is converted in the liver to 25(OH)D (calcidiol), which is the main circulating store. This step, which involves the cytochrome P450 system, is not tightly regulated, and thus an increase in photosynthesis of vitamin D in the skin will lead to an increase in 25(OH)D in the circulation,10,44 bound to DBP. Excess 25(OH)D is converted to 24,25(OH)D, which is thought be relatively metabolically inactive. 10 The 25(OH)D–DBP complex enters renal tubule cells by membrane-bound megalin transport, where the enzyme 1-α-hydroxylase converts it to 1,25(OH)₂-vitamin D (calcitriol), which is the active compound. 44 Although the kidney is the primary site for conversion of circulating 25(OH)D, many cells and tissues, such as macrophages, osteoblasts, keratinocytes, prostate, colon and breast, express the 1-α-hydroxylase enzyme. 42,45,46 As anephric patients have very low levels of 1,25(OH)₂-vitamin D in the blood, it seems likely that these extrarenal sites function at the paracrine level, and do not play a major role in calcium homeostasis. 43

Food sources, recommended intakes and dose response

Few foods contain significant amounts of vitamin D. The most effective sources are oily fish (e.g. salmon, mackerel) and fortified foods such as margarine and breakfast cereal. The amount of vitamin D derived from fish is modest: wild salmon contains around 400 IU per 3.5 oz (100 g). 10 There is much controversy over the recommended daily intake of vitamin D. Older guidance has suggested 200 IU per day for children and adults aged ≤ 50 years and 400–600 IU for older adults. 47 However, humans have evolved to synthesise much higher levels of vitamin D in the skin: 30 minutes exposure at mid-day in the summer sun at a southerly latitude in a bathing suit will release around 50,000 IU into the circulation within 24 hours in white persons. 48 Previous guidelines were not based on any rigorous assessment of the effects of levels and more recent dosing studies have shown that supplementation with 200–400 IU per day is unlikely to maintain levels of 25(OH)D over winter months, let alone replenish stores in somebody who is frankly vitamin D deficient. 49 Thus, a daily maintenance dose of around 1000 IU per day may be more appropriate in people without adequate sunshine exposure, with higher initial dosing required to reverse frank deficiency. 50

Physiology of vitamin D in pregnancy

During pregnancy there is an increase in 1,25(OH)₂-vitamin D, which may be largely due to an increase in DBP. 51 This rise is associated with an increase in intestinal calcium absorption (to around 80% intake), and an absorptive hypercalciuria. 51 There does not seem to be a rise in maternal PTH or 25(OH)D during pregnancy, suggesting that the rise in 1,25(OH)₂-vitamin D may be due to another factor, such as PTH-related peptide, which may be secreted by the placenta. 52 Studies of maternal bone mass in pregnancy have been conflicting, but most suggest a probable decrease, with a possibly greater decrease in lactation. 53–57 The vitamin D receptor (VDR) appears to develop after birth in the infant intestine, and thus calcium absorption is a passive process immediately after birth. 58 The role of vitamin D in utero is uncertain, although 25(OH)D does cross the placenta. 59 In a mouse model, lack of VDR did not significantly affect placental calcium transport or skeletal mineralisation;58 conversely, in the rat, 1,25(OH)₂-vitamin D did seem to influence placental calcium flux. 60 Additionally, chondrocytes are an extrarenal source of 1-α-hydroxylase activity [and so conversion of 25(OH)D to 1,25(OH)₂-vitamin D]. 61 This observation therefore suggests a possible mechanism by which maternal 25(OH)D status might influence bone size in the fetus. Further evidence to support this notion comes from mouse models in which the gene for 1-α-hydroxylase (Cyp27b1) was either knocked out or overexpressed in chondrocytes, leading to altered growth plate morphology. 62 Few data exist in humans at the level of cell biology. Some suggestions have come from recent epidemiological work described above, in which maternal 25(OH)D concentrations positively predicted offspring bone mass at birth,40 and at 9 years old,1,41 with umbilical cord calcium concentrations and placental calcium transporters63 implicated in the mechanisms.

Normal range and measurement of vitamin D

Circulating 25(OH)D is the major store of vitamin D and is the most appropriate for measurement. 1,25(OH)₂-vitamin D is an adaptive hormone, and therefore its level will reflect prevailing conditions such as calcium intake, and thus defining a normal level may not be meaningful. 43 The concept of what is the normal range for 25(OH)D is highly controversial at the moment. One view is that, given that humans seem to have evolved to require much higher levels of vitamin D than are observed in the UK currently, the process of measuring levels in a population and defining a lower cut-off of the distribution as deficient is likely not to be valid. Historically in the UK, serum levels have been classed as ‘replete’ (> 50 nmol/l), insufficient (25–50 nmol/l) or deficient (< 25 nmol/l). (Older studies often use ng/ml as the unit of measurement: 1 ng/ml = 2.5 nmol/l.) The Institute of Medicine in the USA has recently reiterated the 50 nmol/l threshold as the desirable level of circulating 25(OH)D. 64 The distinction between replete and insufficient/deficient has been made on the basis of whether or not there is a secondary rise in PTH. Other approaches to definition have been based on fractional calcium absorption and bone turnover markers. However, a recent review of the available studies relating 25(OH)D concentration to PTH concentration found, across the 70 studies, that a continuous relationship was observed in eight studies, no relationship in three and a thresholded relationship in the remaining 59. 65 Where a threshold was detected, this varied between 25 and 125 nmol/l. Studies of fractional calcium absorption are similarly heterogeneous. 66 Furthermore, in an autopsy-based study of 675 cadavers,67 although bone mineralisation defects (osteomalacia) were not observed in any individual with 25(OH)D > 75 nmol/l, in those with levels < 25 nmol/l, a substantial proportion were found to have normal bone histology. Taken with the range of attempts to define cut-offs for deficiency, these results clearly make the point that extrapolation from 25(OH)D concentration alone to disease is difficult at the level of the individual.

There are several different methods available to measure 25(OH)D. The gold standard is seen to be gas chromatography–mass spectrometry, but this technique is slow, expensive and time-consuming. Most labs use commercial kit assays, which are usually radioimmunometric assays [e.g. Immunodiagnostic Systems (IDS), DiaSorin, Nicholls], although a chemiluminescence assay also exists (e.g. LIAISON^®, DiaSorin, Stillwater, MN, USA). The assays tend to be less accurate than gas chromatography–mass spectrometry and high-performance liquid chromatography (HPLC), and also discriminate less well between the D₂ and D₃ forms. 68 Comparison of the DiaSorin RIA kits with HPLC showed good correlation for D₃, but D₂ tended to be slightly underestimated. 69 A national system now exists to standardise measurement of 25(OH)D across laboratories in the UK [Vitamin D External Quality Assessment Scheme (DEQAS)],70 and the US National Institutes of Health are leading a global programme aimed at standardisation of 25(OH)D assays across both platform and laboratory. 71

Infant postnatal vitamin D intake

Infant feeding, supplementation and sunlight exposure are strong determinants of postnatal infant 25(OH)D levels and bone health. 72 Concentrations of 25(OH)D in breast milk depend on the mother’s blood levels and so, if the mother is deficient in vitamin D during pregnancy, she is likely to continue to be deficient through lactation, yielding a double insult to the child in the absence of adequate sun exposure. Clearly, postnatal vitamin D supplementation of either the mother (during breastfeeding) or the infant directly, together with maternal or childhood sun exposure, could confound any early outcomes attributed to maternal vitamin D status in pregnancy.

Osteomalacia: definition

Osteomalacia is a bone disease caused by inadequate mineralisation of the bone protein matrix, most often, in the UK, as a result of low levels of vitamin D. 73 Inadequate calcium and phosphate are other potential causes, seen more frequently in developing countries, or as a result of genetic abnormalities leading to phosphate loss. Although osteomalacia is therefore a histological term, it is used to describe the finding of low vitamin D status in a patient with bone/muscle pain, weakness, waddling gait, skeletal fragility and appropriate biochemical abnormalities (e.g. hypocalcaemia). 73 Very few studies have examined osteomalacia in pregnancy, although, anecdotally, the incidence of the clinical syndrome is rising in dark-skinned ethnic minorities in the UK. Clearly the definition of osteomalacia used in studies considered for this review will be critical as the symptoms of osteomalacia overlap considerably with those of chronic pain syndromes such as fibromyalgia. Bone biopsy is the only way to diagnose osteomalacia histologically, but the interventional nature of this procedure means that it is unsuitable for large-scale population studies. One recent study of 675 human subjects at autopsy has demonstrated that there is no threshold in circulating 25(OH)D level below which osteomalacic changes on bone biopsy are always seen. 67

Sample studied

This must include pregnant women or pregnant women and their offspring.

Exposure

This must include either assessment of vitamin D status [dietary intake, sunlight exposure, circulating 25(OH)D concentration] or supplementation of participants with vitamin D or food containing vitamin D (e.g. oily fish).

Outcomes

Study type and setting

Studies which reported data on individuals were included. Ecological and animal studies were excluded. Examples of eligible study designs, together with associated level of resulting evidence quality (Centre for Evidence-Based Medicine)78 are shown below:

• level 1a: systematic review (with homogeneity) of randomised controlled trials (RCTs)

• level 1b: individual RCT [with narrow confidence interval (CI)]

• level 2a: systematic review (with homogeneity) of cohort studies

• level 2b: individual cohort study

• level 3a: systematic reviews (with homogeneity) of case–control studies

• level 3b: individual case–control study.

All studies which contributed relevant information were included, regardless of the setting. However, the setting was noted as part of data abstraction and was used in narrative synthesis. Studies were not excluded on the basis of publication date.

Screening of abstracts

When applying selection criteria, all abstracts and potentially relevant papers were independently assessed by two reviewers (CH, and PC or RM) and decisions shown to be reproducible. Disagreements over inclusion were resolved through consensus and, where necessary, following discussion with a third member of the review team (NCH).

Data extraction

Data extraction was carried out by two reviewers. Disagreements were resolved in the same way as for screening of abstracts. Separate forms were used to mark or correct errors or disagreements and a database kept for potential future methodological work.

Data were abstracted onto an electronic form. This contained the following items: general information (e.g. date of data extraction, reviewer ID); study characteristics (e.g. study design, inclusion/exclusion criteria); study population characteristics; method of assessment of vitamin D status; baseline data (e.g. age, sex, ethnicity, measures of vitamin D status/supplementation); quality criteria; outcomes (what they were and how they were ascertained); confounding factors; analysis (statistical techniques, sample size based on power calculation, adjustment for confounding, losses to follow-up); results (direction of relationship, size of the effect and measure of precision of effect estimate such as 95% CI or standard error). The data extraction forms for different study types are included in Appendix 2 .

Effect modifiers/confounders

The effect modifiers and confounding factors considered included ethnicity, skin covering, season, sunlight exposure, alcohol intake, smoking, dietary calcium, physical activity, comorbidity (e.g. diabetes mellitus), current medication, maternal body mass index (BMI), infant feeding, infant supplementation and maternal postnatal supplementation if breastfeeding. Inclusion of these factors was recorded for each study and used as a marker of quality. Where meta-analysis was performed to generate a pooled effect size, inclusion and adjustment for these factors in individual studies was again recorded and used in quality assessment.

Study quality assessment

Quality assessment of studies took place (1) during data extraction and (2) in the analysis of review findings. The quality of included studies was assessed by the two reviewers, using a checklist of questions. The questions used, although based initially on CRD guidelines, were refined through piloting and agreement with the advisory group. Aspects of quality assessed included appropriateness of study design, ascertainment of exposure and outcome, consideration of the effects of important confounding factors, rigour of analysis, sample size and response rates. Quality assessment also incorporated specific issues related to vitamin D. Quality criteria are summarised in Appendix 3 . Quality data were used in narrative descriptions of study quality, and to produce composite validity scores with which to assign a quality level to each study such that studies could be stratified during synthesis of evidence. Quality assessment tools were agreed by the advisory group and refined during piloting. Each study was allocated a score for each quality criterion to estimate the overall risk of bias: +1 indicated a low risk of bias, 0 a medium risk of bias and −1 a high risk of bias. These scores were then added to give a composite score, indicating bias in relation to the review question for each study. This score was between −16 and +16 for intervention and case–control studies; cohort and cross-sectional studies were allocated a score of between −13 and +13. A total composite score < 0 indicated a high risk of bias, a score between 0 and 4 indicated a medium risk of bias and scores of ≥ 5 indicated a low risk of bias. Vitamin D-specific issues are summarised below:

• How is ‘vitamin D’ assessed (dietary intake, supplement use, blood levels of 25(OH)D, blood levels of 1,25(OH)D, PTH concentration)?

• Are season and sunlight exposures including sunscreen use and skin covering considered?

• Are ethnicity and skin pigmentation considered?

• How is 25(OH)D blood level assessed?

• What assay is used?

• Are D₂ and D₃ forms adequately measured and are quality data (e.g. DEQAS) given?

• What definition of ‘normal range’ for 25(OH)D is used?

• Is the concentration treated as categorical (e.g. deficient, insufficient, replete) or continuous?

• Has infant postnatal vitamin D intake (breast, bottle feeding, supplementation) and sunlight exposure been considered?

• Has maternal compliance with supplementation been assessed?

Synthesis of extracted evidence

The aim of this part of the review was to investigate whether or not effects were consistent across studies and to explore reasons for apparent differences. We used both descriptive (qualitative) and quantitative synthesis; our capacity for the latter was determined by the evidence available. Where meta-analysis was possible, we used standard analytical procedures. 79 Only independent studies were meta-analysed. Thus, where a study contained two treatment arms, these were not included in the same analysis. It was, therefore, not possible to include all treatment arms from all RCTs in the same analysis. Two main approaches were employed: first, a meta-analysis of low-dose studies (total dose < 120,000 IU vitamin D), including relevant single treatment arm studies, and the low-dose and placebo arms of studies with more than one treatment arm; and, second, a similar approach but including those studies/study arms with high dose (total > 120,000 IU). Inevitably, the observed estimates of the effects reported in the studies included in the meta-analysis varied. Some of this variation is due to chance alone, as no study can be large enough to completely remove the random error. However, the reported effects may also vary due not only to chance but also to methodological differences between studies. This variation between studies defines statistical heterogeneity. Statistical analysis was performed using Stata v12.1 (StataCorp LP, College Station TX, USA). Between-study statistical heterogeneity was assessed by Q-statistic and quantified by I ² test;80,81 values of I ² index of 25%, 50% and 75% indicated the presence of low, moderate and high between-trials heterogeneity, respectively, while a p-value of < 0.10 was considered to denote statistical significance of heterogeneity. Differences in mean birthweight and serum calcium between supplemented and unsupplemented groups in RCTs were analysed using weighted mean difference and 95% CIs. Results from observational studies were also synthesised. Pooled regression coefficients and odds ratios (ORs) and the 95% CIs were calculated for continuous and dichotomous outcomes respectively. For all analyses performed, if no significant heterogeneity was noted, fixed-effect model analysis using the Mantel–Haenszel method was presented; otherwise, results of the random-effects model (REM) analysis using the DerSimonian–Laird method were presented. 82

Observational studies (see Appendix 6 , Table 8 )

Nineteen observational studies24,40,41,83–98 linking maternal vitamin D status to offspring birthweight were identified. These were all of either cross-sectional (n = 5) or cohort (n = 14) design. Maternal vitamin D status was assessed by maternal serum 25(OH)D concentration in 14 studies, dietary intake in four studies and ambient UVB radiation during the last trimester of pregnancy in one. Sample sizes ranged from 84 to 13,904. Few studies considered all confounding factors of relevance to the review question. Composite bias scores ranged from −2 to +8, with 7 of the 19 studies scored as having a low risk of bias. Of the 14 studies relating maternal serum 25(OH)D concentration to offspring birthweight, only three studies83–85 demonstrated a significant positive association; one study89 found a significant negative association. In contrast, three86–88 of the four studies assessing the influence of maternal vitamin D intake during pregnancy on offspring birthweight found a significant positive association. One study41 found no significant association between ambient UVB exposure in pregnancy and offspring birthweight.

Armirlak et al. 83 (composite bias score 2, medium risk) found a positive association between maternal 25(OH)D at delivery and offspring birthweight in a cross-sectional study of 84 healthy Arab and South Asian women with uncomplicated deliveries. Maternal 25(OH)D was generally low, with a mean of 18.5 nmol/l. A large Australian study (Bowyer et al. ,84 composite bias score 4, medium risk) of 971 pregnant women found that offspring birthweight was significantly lower in those women with 25(OH)D deficiency (< 25 nmol/l) even after adjusting for gestational age, maternal age and overseas maternal birthplace. Similarly, the Amsterdam Born Children and their Development (ABCD) study (Leffelaar et al. ,85 composite bias score 4, medium risk) incorporated 3730 pregnant women and found that early pregnancy maternal 25(OH)D < 30 nmol/l was significantly associated with a lower offspring birthweight, even after adjusting for multiple confounding factors. However, when serum 25(OH)D was analysed as a continuous variable a significant association with birthweight was no longer seen. Mannion et al. 86 (Canada, composite bias score 1, medium risk), Scholl and Chen87 (USA, composite bias score 2, medium risk) and Watson and McDonald88 (New Zealand, composite bias score 3, medium risk) attempted to assess maternal vitamin D intake during pregnancy via Food Frequency Questionnaires (FFQs) at various stages of gestation. Mannion et al. 86 and Scholl and Chen87 found that maternal vitamin D intake was positively associated with offspring birthweight. Similar findings were made by Watson and McDonald88 assessing maternal vitamin D intake at 4 months; however, a relationship was no longer observed when maternal vitamin D intake was measured again at 7 months.

Only one study found a negative association between offspring birthweight and maternal 25(OH)D. Weiler et al. 89 (composite bias score 3, medium risk) found that offspring birthweight was significantly lower in women with adequate vitamin D status [defined by the study group as 25(OH)D ≥ 37.5 nmol/l]. However, the number of participants in this study was low overall and only 18 women had 25(OH)D < 37.5 nmol/l. In addition, of those women with serum 25(OH)D concentration < 37.5 nmol/l, a significantly higher percentage were of non-white race (67%) compared with those with an adequate concentration of 25(OH)D (25%).

Twelve observational studies reported no significant association between maternal vitamin D status andoffspring birthweight. Four of these studies were from Asia (Ardawi et al. ,90 Sabour et al. ,91 Maghbooli et al. 92 and Farrant et al. 93), three from the UK (Gale et al. ,24 Harvey et al. 40 and Sayers and Tobias41), two from Australia (Morley et al. 94 and Clifton-Bligh et al. 95), one from the USA (Dror et al. 96), one from Finland (Viljakainen et al. 97) and one from Africa (Prentice et al. 98). Ten studies24,40,90,92–97 had measured maternal 25(OH)D during pregnancy or at delivery, one91 had assessed vitamin D intake during pregnancy and the largest study41 of 13,904 pregnant women had assessed maternal UV sun exposure in the last trimester as a proxy measure of vitamin D status.

Evidence synthesis

Results from studies that analysed log-transformed vitamin D were synthesised separately from results of studies that analysed vitamin D in its original units. The studies included in the first meta-analytic model were Harvey et al. ,40 Gale et al. 24 and Farrant et al. ,93 using log-transformed units. The combined estimate of the unadjusted regression coefficients for changes in birthweight (grams) per 10% increase in vitamin D was positive but did not reach statistical significance (pooled regression coefficient 0.47, 95% CI −3.12 to 4.05; see Appendix 7 , Figure 2 ). In contrast, when adjusted estimates were synthesised (with adjustments being gestational age, maternal age, maternal BMI, ethnicity and parity where possible), there were significant differences in birthweight (grams) for each 10% increase in vitamin D (pooled regression coefficient 5.63, 95% CI 1.11 to 10.16; see Appendix 7 , Figure 3 ). Amirlak et al. ,83 Prentice et al. ,98 Leffelaar et al. 85 and Dror et al. 96 analysed vitamin D in its original units. All four studies provided adjusted estimates, and all but Amirlak also provided unadjusted regression coefficients. No significant differences in birthweight (grams) per 25 nmol/l increase in vitamin D were found in either combined unadjusted associations (pooled regression coefficient 0.47, 95% CI −1.14 to 2.09; see Appendix 7 , Figure 4 ) or combined adjusted (as per paper) associations (pooled regression coefficient 0.12, 95% CI −1.84 to 2.08; see Appendix 7 , Figure 5 ).

Intervention studies (see Appendix 6 , Table 9 )

Nine intervention trials3–7,21,99–101 were identified, only two99,100 of which were carried out in the last 20 years and the earliest of which was from 1980. 3 Sample sizes ranged from 40 to 350. Seven of these studies were rated as having a high chance of bias on the composite score (−2 to −9); only the most recent studies by Yu et al. 99 and Hollis et al. 100 were assessed as having a low risk of bias (composite bias score 5 and 10 respectively). Eight studies3–7,99–101 reported randomisation, although only one study (Brooke et al. 3) was of a double-blind design and this was also the only study that was placebo-controlled. In eight3–7,21,99,101 of the studies, intervention took place in the last trimester of pregnancy; one study101 intervened in months 6 and 7 of pregnancy and one study100 supplemented from weeks 12–16 onwards. Interventions were highly variable, including 1000 IU daily of ergocalciferol, two doses of 60,000 IU cholecalciferol, two doses of 600,000 IU cholecalciferol, a single oral dose of 200,000 IU and 1200 IU cholecalciferol in combination with 375 mg calcium daily. Change in maternal serum 25(OH)D concentration before and after supplementation was given in three studies only. Three4,5,101 of the eight studies (all from India) demonstrated a statistically significantly greater birthweight in offspring of supplemented than unsupplemented mothers. The remainder showed no difference in infant birthweight regardless of supplementation. 3,6,7,21,99,100

Two Indian studies, both by Marya et al. 4,5 (composite bias scores −6 and −2, respectively, high risk), demonstrated significantly higher birthweights in infants born to women supplemented with high-dose cholecalciferol (given as two doses of 600,000 IU in months 7 and 8 of gestation). The earlier of these studies also had a third arm of women supplemented with 1200 IU vitamin D plus 375 mg calcium throughout the third trimester of pregnancy. Birthweights of infants in this group were also significantly higher than in the unsupplemented group, but not by as much as in the high-dose supplement group. The third study reporting a positive association between maternal vitamin D supplementation and offspring birthweight was also from India (Kaur et al. ,101 composite bias score −7, high risk). Again, significantly higher infant birthweight was found in the supplemented group (two doses of 60,000 IU cholecalciferol in months 6 and 7) than in the unsupplemented group, although the number of participants in this study was low (n = 25 in each arm). Of note, none of the three studies measured maternal 25(OH)D at any point during pregnancy, and all were assessed to have a high risk of bias.

Three UK studies had investigated the effect on offspring birthweight of maternal vitamin D supplementation in the third trimester of pregnancy. Brooke et al. 3 (composite bias score −2, high risk) and Congdon et al. 21 (composite bias score −9, high risk) recruited only Asian women residing in the UK, whereas Yu et al. 99 (composite bias score 5, low risk) included equal numbers of four ethnic groups (black, Caucasian, Asian, Middle Eastern). None of the studies reported a significant difference in offspring birthweight between the supplemented and unsupplemented groups, even despite Brooke et al. 3 demonstrating significantly higher maternal 25(OH)D concentrations in the supplemented group at term. Two studies, both from France (Delvin et al. ,6 composite bias score −2, high risk; Mallet et al. ,7 composite bias score −3, high risk), also failed to demonstrate a significant difference in offspring birthweight with maternal vitamin D supplementation. The most recent, and largest, study (Hollis et al. ,100 composite bias score 10, low bias risk) randomised 350 pregnant women residing in the USA to either 400 IU per day, 2000 IU per day or 4000 IU per day of oral vitamin D₃ from 12 to 16 weeks’ gestation until delivery. Although maternal serum 25(OH)D at delivery was higher in those women receiving the higher dose supplement regimes, there was no significant difference in offspring birthweight among the three groups.

Evidence synthesis

Two meta-analyses were performed to combine the published evidence of an effect of vitamin D supplementation on birthweight. The first included Brooke et al. ,3 Marya et al. 4 (low dose of vitamin D), Congdon et al. ,21 Mallet et al. 7 (low dose of vitamin D) and Kaur et al. 101 (see Appendix 7 , Figure 6 ). Owing to statistically significant heterogeneity in the results (I ² = 86.3%, p < 0.001), a REM was fitted. The combined estimate showed a non-significant difference in birthweight between the unsupplemented and supplemented groups (mean weighted difference 116.23 g, 95% CI −57.0 g to 289.5 g). The second meta-analytical model included Brooke et al. ,3 Marya et al. 4 (high dose of vitamin D), Congdon et al. ,21 Mallet et al. 7 (high dose of vitamin D), Marya et al. 5 and Kaur et al. 101 (see Appendix 7 , Figure 7 ). Again, here, owing to statistically significant heterogeneity (I ² = 96%, p < 0.001), a REM was fitted and the combined results did not show a significant difference in birthweight between the supplemented and the non-supplemented groups (mean weighted difference 147.3 g, 95% CI −112.5 g to 407.15 g).

Discussion

The results of the included studies were conflicting, with some demonstrating positive associations between 25(OH)D concentration and birthweight and some no relationship. The observation studies were, on the whole, of greater quality than the intervention studies, with almost all of the latter assessed as having a high risk of bias. Meta-analysis revealed weak positive associations across three observational studies, after adjustment for potential confounders, between log-transformed 25(OH)D concentrations and offspring birthweight. However, confounding factors considered varied across the studies, and the potential for residual confounding is large. Despite these caveats, the relationships were generally positive, albeit not statistically significant, across the majority of identified studies, suggesting that further exploration in a well-designed, randomised, placebo-controlled, double-blind trial might be appropriate.

Observational studies (see Appendix 6 , Table 10 )

Thirteen observational studies24,41,85,86,90–98 including maternal vitamin D status and offspring birth length were identified; nine of the these were cohort in design, with the remaining three being cross-sectional studies. The number of participants in each study ranged from 120 to 10,584. Maternal vitamin D status was assessed by serum 25(OH)D concentration in 10 studies24,85,90,92,93–98 and by dietary intake in two;86,91 in the remaining study41 maternal ambient UVB exposure during late pregnancy was used as a surrogate marker of vitamin D status. One study91 was assessed as having a high risk of bias (composite score −2, high risk), with the others demonstrating composite scores between +1 and +8. Consideration of potential confounding factors was variable. Two studies41,91 identified a positive relationship between maternal vitamin D status and offspring birth length, neither of which directly measured maternal 25(OH)D. The remaining 10 studies24,85,90,92–98 showed no relationship. We did not identify any studies that demonstrated an inverse relationship between maternal vitamin D status in pregnancy and offspring birth length.

Sabour et al. 91 (composite bias score −2, high risk), in a cross-sectional study of 449 pregnant women in Iran, found that offspring birth length was significantly higher in mothers with adequate vitamin D intake (defined by the authors as > 200 IU vitamin D per day). This study was assessed to have a high risk of bias and maternal serum 25(OH)D was not measured, as vitamin D status was estimated from a FFQ of dietary intake. The second study showing a positive relationship came from Sayers and Tobias41 (composite bias score 3, medium risk) using data from the large UK cohort (ALSPAC). In this study, again, maternal serum 25(OH)D was not directly measured but estimated using maternal UVB exposure in the last 98 days before birth as a surrogate. Maternal UVB exposure in late pregnancy was positively associated with offspring birth length. Additionally, Leffelaar et al. 85 measured offspring length at 1 month and found that infants born to mothers with 25(OH)D < 30 nmol/l (the threshold used by the authors for vitamin D deficiency) had a significantly lower length at 1 month even after adjusting for multiple confounders (including gestational age, season of blood sample, maternal height, maternal age, smoking pre-pregnancy, smoking in pregnancy, educational level, ethnicity and parity).

The remaining 10 studies24,86,90,92–98 found no significant relationship between maternal vitamin D status and offspring birth length. Of these studies, nine used maternal 25(OH)D as the predictor and six were assessed to have a low risk of bias. Two studies were from the Middle East (Ardawi et al. ,90 composite bias score 5, low risk; Maghbooli et al. ,92 composite bias score 1, medium risk), two from Australia (Morley et al. ,94 composite bias score 8, low risk; Clifton-Bligh et al. ,95 composite bias score 6, low risk), two from North America (Mannion et al. ,86 composite bias score 1, medium risk; Dror et al. ,96 composite bias score 7, low risk) and the remainder from the UK (Gale et al. ,24 composite bias score 4, medium risk), Finland (Viljakainen et al. ,97 composite bias score 3, medium risk), India (Farrant et al. ,93 composite bias score 5, low risk) and Africa (Prentice et al. ,98 composite bias score 5, low risk).

Intervention studies (see Appendix 6 , Table 11 )

Two RCTs of vitamin D supplementation in pregnancy included birth length as an outcome; both were assessed to have a high risk of bias (composite bias score of both −2, high risk). A double-blind placebo-controlled trial (Brooke et al. 3) found no significant difference in offspring birth length in UK Asian women supplemented with 1000 IU ergocalciferol per day in the last trimester compared with the control group. In contrast, a larger Indian study by Marya et al. 5 found that birth length was significantly higher in women supplemented with a much higher dose of vitamin D (two doses of 600,000 IU cholecalciferol in the seventh and eighth month of gestation) than in unsupplemented women.

Discussion

Again, the majority of the observational studies suggested no relationship between maternal 25(OH)D status and offspring birth length. One41 of the studies which showed a significant association was large and prospective, but used ambient UVB radiation rather than a direct measure of vitamin D status. Of the two randomised trials3,5 to investigate birth length, one found a statistically significant relationship and the other did not. Thus, the results are mixed but do not support the use of maternal vitamin D supplementation to reduce the risk of low birth length.

Observational studies (see Appendix 6 , Table 12 )

Eleven observational studies24,86,90–98 assessed the relationship between maternal vitamin D status in pregnancy and offspring head circumference. Eight24,86,90,93–95,97,98 of the studies were of cohort design, with the remaining three91,92,96 being cross-sectional studies. Participant numbers ranged from 120 to 559. Maternal vitamin D status was assessed by serum 25(OH)D concentration in nine studies;24,90,92–98 the remainder used dietary intake (Sabour et al. 91 and Mannion et al. 86). Composite bias scores ranged from −2 to +8, with six studies90,93–96,98 having a low risk of bias. Of those relating maternal serum 25(OH)D to offspring head circumference at birth, no study found a statistically significant relationship, regardless of when during pregnancy 25(OH)D was measured.

Three studies were from the Middle East: Ardawi et al. 90 and Maghbooli et al. 92 found no association with offspring head circumference at birth and maternal 25(OH)D measured at delivery. Likewise, Sabour et al. 91 observed no difference in offspring head circumference in women taking < 200 IU vitamin D per day compared with those taking > 200 IU vitamin D per day. Two Australian studies (Morley et al. 94 and Clifton-Bligh et al. 95) measured maternal vitamin 25(OH)D in the third trimester of pregnancy and also found no significant association between maternal 25(OH)D concentration and offspring head circumference. Morley et al. 94 also measured 25(OH)D in early pregnancy and again a relationship was not demonstrated. Similar findings were made by Mannion et al. 86 (a Canadian study using estimated dietary intake of vitamin D in pregnancy as the predictor), Gale et al. 24 [UK, 25(OH)D measured in the third trimester], Farrant et al. 93 [India, 25(OH)D measured in the third trimester], Prentice et al. 98 [The Gambia, Africa, 25(OH)D measured in the second and third trimesters], Viljakainen et al. 97 [Finland, mean of early pregnancy and postpartum 25(OH)D concentration used] and Dror et al. 96 (USA, measured perinatally).

Intervention studies (see Appendix 6 , Table 13 )

Offspring head circumference at birth was an outcome in two RCTs3,5 of vitamin D supplementation in pregnancy, both of which were assessed as having a high risk of bias (composite bias score −2 in both). Brooke et al. 3 included 126 Asian patients and randomised in a double-blind fashion to either placebo or 1000 IU daily ergocalciferol in the last trimester. Head circumference did not differ between the treatment and placebo groups. In contrast, Marya et al. 5 randomised 200 Indian women to either no supplement or to two doses of 600,000 IU cholecalciferol in the last trimester and found that head circumference at birth was significantly higher in the supplemented group than in the unsupplemented group.

Discussion

Thus, the majority of the observational studies demonstrated no association between maternal 25(OH)D status in pregnancy and offspring head circumference at birth. One5 of the intervention studies found a positive relationship between supplement use and head circumference. It should be noted that this study generally found statistically significant relationships for most of the measured outcomes and was considered to be of high risk of bias. The evidence base is insufficient to recommend vitamin D supplementation for the optimisation of, or prevention of, low head circumference.

Observational studies (see Appendix 6 , Table 14 )

Eight observational studies1,41,89,96–98,102,103 that included offspring bone mass outcomes were identified. Five of these were cohort studies, with the remaining three being cross-sectional in design. All studies were assessed as being of medium to low risk of bias, with composite bias scores ranging from 3 to 7. The age at which offspring were assessed ranged from within 24 hours of birth to 9.9 years. Bone outcome measures also varied across the studies and included whole-body, lumbar spine, radial midshaft, tibial and femoral BMC, whole-body and lumbar spine BA, whole-body and tibial bone mineral density (BMD), tibial cross-sectional area (CSA) and whole-body BMC adjusted for BA (areal bone mineral density; aBMD). Most studies (six1,41,89,96,98,103 of eight) used DEXA to assess bone mass; two studies97,102 used peripheral quantitative computed tomography (pQCT) and one study98 used SPA in addition to DEXA. Seven studies1,89,96–98,102,103 measured maternal 25(OH)D during pregnancy or at delivery, one study41 used UVB exposure in the third trimester of pregnancy as a measure of maternal vitamin D status. Five studies1,41,89,97,102 demonstrated a positive relationship between maternal vitamin D status and offspring bone health; three studies96,98,103 showed no relationship.

Weiler et al. 89 (composite bias score 3, medium risk, n = 50) found that neonates born to mothers with adequate maternal 25(OH)D at delivery (defined by the authors as > 37.5 nmol/l) had significantly higher whole-body and femoral BMC per unit body weight than neonates born to mothers with insufficient maternal vitamin D concentration (< 37.5 nmol/l), even after adjustment for multiple confounders. However, there was no significant difference in infant lumbar spine, femoral or whole-body BMC between the two groups. Viljakainen et al. 97 (composite bias score 3, medium risk) also measured neonatal bone mass in a Finnish cohort of 125 primiparous Caucasian women. Tibial bone mass was assessed by pQCT and those with maternal 25(OH)D above the median (42.6 nmol/l) had significantly higher tibial BMC and CSA than those with maternal 25(OH)D below the median, even after adjusting for confounders including maternal height and birthweight. No relationship was seen between maternal 25(OH)D and tibial BMD. A subsample of 55 children was also assessed again at 14 months (Viljakainen et al. 102) and tibial BMC was no longer significantly different by maternal 25(OH)D status. Tibial CSA, however, remained significantly lower in those with maternal 25(OH)D below the median. Two cohort studies from the UK also demonstrated significant associations between maternal vitamin D status and offspring bone mass measured later in childhood. Javaid et al. 1 measured maternal 25(OH)D in late pregnancy and offspring bone mass by DEXA at mean 8.9 years in a cohort of 198 pregnant women. Positive associations were observed between maternal 25(OH)D and offspring whole-body and lumbar spine BMC, lumbar spine BA and whole-body and lumbar spine BMD after adjustments were made for offspring gestational age at delivery and offspring age at DEXA. Sayers and Tobias41 found that maternal UVB exposure in late pregnancy was positively associated with offspring BMC, BA and BMD in 6955 children at mean age 9.9 years. No relationship was seen between aBMD and maternal UVB exposure.

Three studies found no associations between maternal 25(OH)D and offspring bone mass. Two studies (Akcakus et al. 103 and Dror et al. 96), both cross-sectional in design, and with a similar number of participants, measured maternal 25(OH)D at delivery and used DEXA to assess offspring bone mass up to the first month of life. A third study (Prentice et al. 98) measured mid- and late-pregnancy 25(OH)D in a cohort of 125 pregnant Gambian women taking part in a larger clinical trial of vitamin supplementation. Offspring underwent assessment of BMC and BA using SPA of the midshaft radius; a subset also underwent whole-body DEXA at ages 2, 13 and 52 weeks. Again, no statistically significant relationship between maternal 24(OH)D and offspring BMC at any time point was observed. It should be noted that mean maternal 25(OH)D levels in this cohort were much higher than any other study with an average of 103 nmol/l for mid-pregnancy and 111 nmol/l for late pregnancy, and none of the women in the study was considered vitamin D deficient.

Intervention studies (see Appendix 6 , Table 15 )

One clinical trial of maternal vitamin D supplementation and its effect on offspring bone mass was identified. Congdon et al. 21 randomised 64 Asian women in the UK to either no supplement or 1000 IU vitamin D plus calcium daily in the third trimester. Forearm BMC was measured in offspring within 5 days of birth, although the type of equipment used to measure this was not recorded. No difference in offspring radial BMC was observed between the two groups. This study was assessed to have a high riskof bias (composite bias score −9) and maternal serum vitamin D concentration in pregnancy was not recorded at any time point.

Discussion

Five1,41,89,97,102 of the eight observational studies relating maternal 25(OH)D status to offspring bone outcomes demonstrated positive associations. The one small intervention study21 identified did not, but the methodology is unclear and a statistically significant result is unlikely based on the sample size. Thus, observational studies suggest that maternal 25(OH)D status may influence offspring bone development, but do not allow public health recommendations to be made. Further high-quality intervention studies are required here, such as the ongoing Maternal Vitamin D Osteoporosis Study (MAVIDOS). 104

Observational studies (see Appendix 6 , Table 16 )

Six observational studies24,41,89,94,105,106 (five cohort and one cross-sectional) have examined the relationships between maternal vitamin D status and a variety of anthropometric measures in the offspring. Composite bias scores ranged from 3 to 8, indicating a medium to low risk of bias. Five studies24,89,94,105,106 had measured maternal serum 25(OH)D in pregnancy (four in the third trimester and one at delivery); one study41 used maternal UVB exposure during the last trimester of pregnancy as a surrogate estimate of maternal vitamin D status. Anthropometric measurements of the offspring ranged across the studies and included skinfold thickness, limb circumference and muscle area. Five studies24,41,89,105,106 used DEXA to measure offspring fat and/or lean mass. Four studies41,94,105,106 demonstrated a significant relationship between offspring anthropometry and maternal 25(OH)D; the remaining two24,89 showed no relationship.

Morley et al. 94 measured offspring subscapular, triceps and suprailiac skinfold thickness using Harpenden callipers (British Indicators, Burgess Hill, UK), along with mid-upper-arm and calf circumferences using measuring tape in 374 Australian neonates. Although there no was significant association between maternal 25(OH)D at 11 weeks’ gestation and any of the neonatal outcome measures, a weak inverse association was observed between maternal 25(OH)D measured at 28–32 weeks and neonatal subscapular and triceps skinfold thickness. This association was weakened further but still remained statistically significant after adjustments were made for offspring sex, maternal height, whether or not the offspring was a first child, maternal smoking and season of blood sample. No significant association with maternal 25(OH)D was found with the other offspring anthropometric outcomes assessed. Krishnaveni et al. 105 also assessed offspring subscapular and triceps skinfolds, using callipers, in addition to arm muscle area, waist circumference, fat mass, per cent body fat, fat-free mass and per cent fat-free mass, using a combination of measuring tape and bioimpedence, in an older cohort of Indian children aged 5 years (n = 506) and again at age 9.5 years (n = 469). Children born to mothers with late-pregnancy vitamin D deficiency [25(OH)D concentration < 50 nmol/l] had significantly reduced arm–muscle area in comparison with children born to mothers with adequate levels. No significant relationship was observed with the other anthropometric measurements at either time point.

Of the four studies using DEXA to measure offspring fat and/or lean mass, two reported no relationship with maternal vitamin D status. Weiler et al. 89 used DEXA to measure whole-body fat in a group of 50 neonates in Canada. No significant difference was observed between those born to mothers with 25(OH)D concentration < 37.5 nmol/l at delivery and those born to mothers with 25(OH)D > 37.5 nmol/l. Gale et al. 24 found no significant association between maternal 25(OH)D in late pregnancy and offspring fat mass or lean mass in 178 UK children aged 9 years. Fat and lean mass tended to be lower in children born to mothers in the lowest quarter of 25(OH)D distribution, but this did not achieve significance. In contrast, Sayers and Tobias41 using maternal UVB exposure in late pregnancy as a surrogate measure for vitamin D status found that offspring lean mass at mean age 9.9 years was positively associated with maternal UVB exposure. However, no significant association was seen with fat mass. In contrast, Crozier et al. 106 (composite bias score 8, low risk) found that maternal serum 25(OH)D in late pregnancy was positively associated with offspring fat mass at birth, measured by DEXA, after adjusting for confounders. Interestingly, no significant relationship was seen between maternal 25(OH)D and offspring fat mass at 4 years, and a negative relationship was seen at 6 years of age. No significant relationship was observed between maternal 25(OH)D and offspring’s fat-free mass at any time point.

Intervention studies (see Appendix 6 , Table 17 )

Two intervention studies were identified and have been described earlier. Both studies were assessed to have a high risk of bias (composite bias score −2 for both). Brooke et al. 3 found no difference in neonatal triceps skinfold thickness or forearm length between those born to supplemented mothers and placebo group mothers. Marya et al. 5 found significantly greater mid-upper-arm circumference and triceps and subscapular skinfold thicknesses in neonates of supplemented than in those born to unsupplemented mothers (all p < 0.01).

Discussion

The identified observational studies demonstrated a variety of modest relationships between maternal 25(OH)D status and offspring anthropometric measures, with some finding positive relationships between maternal 25(OH)D status and measures of offspring muscle and fat mass. Consistent with other anthropometric outcomes in their study, Marya et al. found greater skinfold thicknesses in the supplemented group than in the unsupplemented group. The evidence base is therefore insufficient to warrant recommendation of maternal vitamin D supplementation to optimise childhood anthropometric measures.

Observational studies (see Appendix 6 , Table 18 )

Ten studies24,26,107–114 were identified that examined the relationships between maternal vitamin D intake during pregnancy, maternal serum 25(OH)D level in pregnancy, or cord blood 25(OH)D concentration and markers of atopy in the offspring. These were all observational cohort studies, ranging in size from 178 to 1724 mother–child pairs. Eight studies24,26,107–112 reported the outcome wheeze or asthma as determined by parental questionnaires at between 16 months and 9 years of age.

Four of these seven studies used maternal vitamin D intake during pregnancy as the exposure and had composite bias scores of between −1 and 2 (Erkkola et al. ,107 Devereux et al. ,26 Miyake et al. 108 and Camargo et al. 109). These four studies all reported a lower risk of wheeze in offspring of mothers with higher vitamin D intakes during pregnancy, although the definitions used for wheeze varied between studies. Miyake et al. 108 included 763 mother–offspring pairs in a prospective cohort study in Osaka, Japan (bias score −1, high risk). Vitamin D intake was measured by FFQs between 5 and 39 weeks of pregnancy and the children followed up between 16 and 24 months of age using the International Study of Asthma and Allergy in Childhood (ISAAC) questionnaire. In this study, consumption of ≥ 172 IU per day vitamin D was associated with a reduced risk of both wheeze and eczema. Camargo et al. 109 reported in a prospective cohort study in Massachusetts, USA, which included 1194 mother–offspring pairs, that children born to mothers in vitamin D intake quarters 2 (446–562 IU/day), 3 (563–658 IU/day) and 4 (659–1145 IU/day) had a reduced risk of recurrent wheeze (two or more episodes of wheeze in children with a personal diagnosis of eczema or parental history of asthma) at 3 years compared with those born to mothers in the lowest quarter of vitamin D intake, but, in contrast to Miyake et al. ,108 there was no difference in the incidence of eczema. Erkkola et al. 107 found a lower risk of persistent asthma (physician diagnosis and a requirement for asthma medication in the preceding 12 months) at 5 years in children born to mothers with higher vitamin D intake, but, similarly to Camargo et al. ,109 there was no reduced risk of atopic eczema. However, this Finnish study included only children who had human leucocyte antigen (HLA) HLA-DQB1-conferred susceptibility to type 1 diabetes mellitus. The composite bias score was −1, indicating a high risk of bias. Finally, Devereux et al. 26 also reported a lowered risk of reported wheeze in the preceding year in 5-year-old children born to mothers with the highest quintile of vitamin D intake at 32 weeks’ gestation (189–751 IU/day) compared with the lowest quintile (46–92 IU/day). There was no statistically significant reduction in the OR for wheeze when quintiles 2, 3 and 4 were compared with quintile 1 but a significant overall trend (p = 0.009).

Two studies assessed the associations between cord blood 25(OH)D and parental report of wheeze and/or asthma. These studies had composite bias scores of 2 and 3 (medium risk of bias). Camargo et al. 110 found that in 823 children in New Zealand the OR for wheeze at 5 years of age decreased across categories of cord 25(OH)D, but there was no association with incident asthma. Similarly, Rothers et al. 111 found no association between cord 25(OH)D and asthma (physician diagnosed and medication requirement in preceding year) at 5 years. Two studies, by Gale et al. 24 and Morales et al. ,112 assessed the association between maternal 25(OH)D measured in pregnancy and parental-reported wheeze or diagnosis of asthma. Gale et al. 24 (composite bias score 4, medium bias risk) assessed the association between maternal 25(OH)D in late pregnancy and parental report of asthma in 178 children. Exposure to the highest quarter of maternal concentrations of 25(OH)D was associated with an increased risk of reported asthma at age 9 years compared with children whose maternal 25(OH)D concentration had been in the lowest quarter of the distribution. In addition, the risk of offspring eczema at 9 months (assessed by either physical examination or parental report) was also higher in children in the highest quarter of maternal 25(OH)D distribution than in those in the bottom quarter. By 9 years of age, however, although offspring in the highest quarter of maternal 25(OH)D still tended to have a higher risk of reported eczema than those in the lowest quarter, the difference was no longer significant. In this study, the number of cases of asthma or eczema per maternal 25(OH)D quarter was low, ranging from 2 to 15. Conversely, Morales et al. 112 (composite bias score 3, medium bias risk) found no significant association between maternal 25(OH)D measured at mean (standard deviation; SD) 12.6 (2.5) weeks and parent-reported offspring wheeze at 1 year or 4 years, or asthma (defined as parental report of doctor diagnosis of asthma or receiving treatment for asthma) at age 4–6 years.

Four studies26,111,113,114 utilised other outcome markers of asthma and/or atopic disease; these studies were subject to less potential bias (composite bias scores −1 to 3). Two studies26,113 measured offspring spirometry: Cremers et al. 113 (bias score 3, medium risk) found no associations between maternal plasma 25(OH)D at 36 weeks’ gestation and offspring forced expiratory volume in 1 second (FEV₁) (p = 0.99) or forced vital capacity (FVC) (p = 0.59) at 6–7 years in 415 mother–offspring pairs. Similarly, Devereux et al. 26 (bias score −1, high risk) did not identify any differences in lung function at 5 years of age across quintiles of maternal vitamin D intake at 32 weeks’ gestation. Two studies also undertook skin prick testing as a measure of atopic sensitisation. Devereux et al. 26 found that maternal vitamin D intake at 32 weeks’ gestation was not associated with differences in atopic sensitisation to cat, timothy grass, egg or house dust mite at 5 years of age. Conversely, Rothers et al. 111 (bias score 2, medium risk) found that children with cord blood 25(OH)D ≥ 100 nmol/l, when compared with those with cord 25(OH)D 50–74.9 nmol/l, had a greater risk of a positive response to a skin prick testing battery that included 17 aeroallergens common to the geographical area. Finally, two studies included offspring IgE concentration as a measure of atopy. Rothers et al. 111 reported a non–linear relationship between cord 25(OH)D and total and allergen-specific IgE for six inhalant allergens. The highest levels of IgE were identified in children with cord 25(OH)D concentration < 50 nmol/l and ≥ 100 nmol/l. Conversely, Nwaru et al. 114 found increasing maternal vitamin D intake determined by FFQ was inversely associated with sensitisation (IgE > 0.35 ku/l) to food allergens (IgE > 0.35 ku/l) but not inhaled allergens at 5 years of age.

Intervention studies

No intervention studies examining the influence of vitamin D supplementation in pregnancy on offspring risk of asthma or atopy were identified.

Discussion

The studies on asthma were all observational; no intervention studies were identified. The investigations were marked by substantial heterogeneity in terms of study design, outcome definition and exposure definition, and gave a variety of conflicting results. It is difficult to conclude any definitive relationship between maternal 25(OH)D status and offspring asthma and no recommendation can be made. Further high-quality intervention studies are required here, such as the ongoing Vitamin D Antenatal Asthma Reduction Trial [VDAART; International Standard Randomised Controlled Trial Number (ISRCTN) NCT00920621] and Vitamin D Supplementation During Pregnancy for Prevention of Asthma in Childhood trial (ABCVitamin D; ISRCTN NCT00856947).

Observational studies (see Appendix 6 , Table 19 )

Seven observational studies85,103,115–119 assessing the relationship between maternal 25(OH)D and the risk of offspring being born SGA were identified. Of these, two were case–control studies,115,116 one was cross-sectional103 and four were cohort studies. 85,117–119 All achieved a composite bias score of between +1 and +7, indicating a medium–low risk of bias. Five studies85,103,115,118,119 defined SGA as birthweight below the 10th percentile according to nomograms based on sex and gestational age. Three studies reported how gestational age was assessed (known dates of last menstrual period (LMP) and/or fetal ultrasound in early pregnancy), with the remainder giving no explanation. All studies measured serum maternal 25(OH)D concentration. The time of sampling ranged from 11 weeks’ gestation to delivery. One study117 defined SGA as birthweight below the third percentile. Three studies85,115,116 (two nested case–control and one cohort study) reported a significant association between maternal 25(OH)D and risk of SGA; the remaining four studies103,117–119 did not demonstrate a significant relationship.

Leffelaar et al. 85 measured maternal 25(OH)D concentration in women at 11–13 weeks’ gestation taking part in the large ABCD study. Of the 3730 women in the cohort, 9.2% delivered SGA infants. Women with a serum 25(OH)D concentration < 30 nmol/l had a significantly higher risk of giving birth to SGA infants than women with 25(OH)D concentrations > 50 nmol/l; this relationship remained even after adjusting for gestational age, season of blood collection, sex of infant and maternal parity, age, smoking, pre-pregnancy BMI, educational level and ethnicity. No significant risk was observed, however, in women with 25(OH)D concentration between 30.0 and 49.9 nmol/l. Bodnar et al. 115 (composite bias score 7, low risk) found that the relationship between maternal 25(OH)D and SGA varied according to race. In this nested case–control study from an overall cohort of 1198 nulliparous women, 111 cases were identified and compared with 301 randomly selected controls; all had 25(OH)D measured before 22 weeks’ gestation. Among black mothers, no relationship between SGA risk and maternal 25(OH)D concentration was observed. However, in white women, a U-shaped relationship was observed between the odds of delivering a SGA infant and maternal 25(OH)D concentration. Significantly higher odds for SGA were observed in those with 25(OH)D concentrations < 37.5 and > 75 nmol/l, with the lowest odds of SGA in women with 25(OH)D concentrations of 60–80 nmol/l. These relationships remained significant even after adjusting for pre-pregnancy BMI, smoking, socioeconomic score, season, maternal age, gestational age at blood sample, marital status, insurance status, conceptual multivitamin use and preconception physical activity. Finally, Robinson et al. 116 (composite bias score 0; medium risk), in a case–control study of pregnant women, all of whom had early-onset severe pre-eclampsia (as defined by the American Congress of Obstetrics and Gynaecology), found that maternal serum vitamin D was significantly lower in cases with SGA infants than with controls. This study did not present an OR or define SGA, and it was not clear at what stage of gestation maternal vitamin D was measured.

A cross-sectional Turkish study of 100 pregnant women (Akcakus et al. ,103 composite bias score 4, medium risk), 30 of whom gave birth to SGA infants, found no difference in maternal mean 25(OH)D at delivery in cases of SGA [maternal 25(OH)D concentration 21.8 nmol/l] compared with mothers of infants born at a size appropriate for gestational age [maternal 25(OH)D concentration 21.5 nmol/l]. Average maternal concentrations of 25(OH)D in this study were low, a reflection of the fact that most women in the study were veiled. A similar finding was observed by Mehta et al. 119 (composite bias score 3, medium risk) in the African cohort study of 1078 women all infected with human immunodeficiency virus (HIV). Seventy-four SGA infants were identified. Again, no difference in mean maternal 25(OH)D concentration measured in mid-pregnancy was observed between cases and normal deliveries. Shand et al. 117 observed similar findings in a cohort study of Canadian women with biochemical or clinical risk factors for pre-eclampsia. No significantly increased odds of SGA were observed in women with 25(OH)D concentrations < 75 nmol/l compared with concentrations > 75 nmol/l. In this study, cases of SGA were low (n = 13). Finally, a Spanish cohort study from Fernandez-Alonso et al. 118 (composite bias score 3, medium risk) identified 46 cases of SGA out of a cohort of 466. No significant relationship between maternal 25(OH)D and SGA infants was observed. Neither mean 25(OH)D concentrations nor an OR were reported.

Intervention studies (see Appendix 6 , Table 20 )

Two clinical trials of maternal vitamin D supplementation evaluated the relationship between maternal 25(OH)D and risk of SGA infants. Both defined SGA as birthweight below the 10th percentile, although neither reported how gestational age was assessed. Neither observed a significant relationship. Brooke et al. ,3 in a double-blind, placebo-controlled randomised trial, allocated 67 pregnant women to either placebo (n = 67) or vitamin D₂ 1000 IU per day in the last trimester of pregnancy (n = 59). Both groups were similar in terms of maternal age, height, parity, offspring sex and length of gestation. In this British study, all participants were Asian, with the majority of Indian ethnicity. Although the mean maternal 25(OH)vitamin D concentration was significantly higher in the supplemented group at delivery than in the unsupplemented group, the percentage of SGA infants did not differ significantly between groups (19 in the placebo group vs. 9 in the supplemented group). The composite bias score of this study was −2 indicating a high risk of bias. Yu et al. 99 (composite bias score 5, low risk) reported similar findings in a more recent British clinical trial. Pregnant women were randomised to one of three arms: no supplement (n = 59); oral vitamin D₂ 800 IU per day from 27 weeks onwards (n = 60); or a single bolus dose of 200,000 IU vitamin D₂ at 27 weeks’ gestation (n = 60). Each group contained equal numbers of four ethnic groups (black, Caucasian, Asian, Middle Eastern). No significant difference in the incidence of SGA was observed across the three groups.

Discussion

There was substantial variation in the methodology, exposure and outcome definitions for studies investigating the relationship between maternal 25(OH)D status and risk of offspring being SGA. Outcomes were conflicting. The two intervention studies3,99 which included this outcome, the more recent of which was deemed of reasonable quality, found that supplementation with vitamin D during pregnancy was not associated with reduced risk. There appears to be no evidence base with which to recommend maternal vitamin be supplemented for the prevention of offspring being SGA neonatal.

Observational studies (see Appendix 6 , Table 21 )

Six observational studies117–122 relating maternal 25(OH)D to the risk of premature birth were identified (three cohort, one cross-sectional, two case–control). One further cross-sectional study123 assessing the risk of threatened premature birth was also included. Two studies were case–control,120,121 three cohort117–119 and two cross-sectional. 122,123 There was some disparity in the definition of preterm birth between studies. Most studies117–119,122 defined preterm birth as spontaneous delivery before 37 weeks’ gestation; one study121 used a threshold of < 35 weeks. Only three studies reported how gestational age was measured: two studies used a combination of LMP and/or fetal ultrasound and one used the scoring system of Dubowitz et al. 124 (based on examination of the neonate and scored on neurological and physical examination features). All studies measured maternal serum 25(OH)D at some point during pregnancy or at delivery. Only one study123 found a significant relationship between maternal 25(OH)D and risk of premature delivery.

Shibata123 (composite bias score 4, medium risk), in a cross-sectional study of 93 Japanese pregnant women attending hospital for a routine medical check-up in Toyoake, Japan, found that maternal 25(OH)D measured after 30 weeks’ gestation was significantly lower in the 14 cases of threatened premature delivery [mean 25(OH)D concentration 30.0 nmol/l] than in normal pregnancies [mean 25(OH)D concentration 37.9 nmol/l]. Threatened premature delivery was defined as progressive shortening of cervical length (< 20 mm) as detected by transvaginal ultrasound before the 34th week of gestation and/or elevation of granulocyte elastase level in the cervical mucus before 32 weeks’ gestation plus two or more uterine contractions every 30 minutes (before the 32nd week of gestation).

In contrast, six studies117,118,119–122 did not demonstrate a significant relationship between maternal 25(OH)D and premature delivery. A small case–control study by Delmas et al. 120 found no difference in mean maternal 25(OH)D concentration measured at delivery in the 10 cases of preterm birth [mean maternal 25(OH)D concentration 44.9 nmol/l] compared with the nine controls [mean maternal 25(OH)D concentration 47.4 nmol/l]. This study achieved a low composite bias score of −4, suggesting a high risk of bias. No adjustment or considerations for potential confounders were made. Similarly, a prospective cohort study from Tanzania of 1078 pregnant African women infected with HIV and taking part in a clinical trial of vitamin use (Mehta et al. ,119 composite bias score 2, medium risk) found no increased relative risk of preterm or severe preterm birth (defined as spontaneous delivery before 34 weeks’ gestation) in women with a serum 25(OH)D concentration measured at 12–27 weeks’ gestation < 80 nmol/l compared with those with levels > 80 nmol/l. A nested case–control study in North Carolina, USA (Baker et al. ,121 composite bias score 5, low risk), identified 40 cases and 120 controls matched by race/ethnicity in a 1 : 3 ratio and compared maternal 25(OH)D measured at 11–14 weeks’ gestation. Again, no significant difference in the OR for preterm birth was found in women with 25(OH)D < 75 nmol/l compared with those with 25(OH)D concentration > 75 nmol/l. Shand et al. 117 in a cohort study of 221 pregnant women in Vancouver, Canada, with either clinical or biochemical risk factors for pre-eclampsia found no significant relationship between maternal 25(OH)D, measured between 10 weeks’ and 20 weeks 6 days’ gestation, and risk of preterm birth using three different thresholds of maternal 25(OH)D (< 37.5 nmol/l, < 50 nmol/l, < 75 nmol/l) after adjustment for maternal age, BMI, season, multivitamin use and smoking. The risk factors for pre-eclampsia included an obstetric history of early-onset or severe pre-eclampsia, unexplained elevated α-fetoprotein ≥ 2.5 multiples of the median (MoMs), unexplained elevated human chorionic gonadotropin, or low pregnancy-associated plasma protein A (≤ 0.6 MoM). Hossain et al. ,122 in a cross-sectional study of 75 pregnant women in Pakistan (composite bias score 4, medium risk), found that mean maternal 25(OH)D₃ at delivery tended to be higher in those who delivered preterm [mean 25(OH)D₃ concentration 42.2 nmol/l] than in those with full term deliveries [mean 25(OH)D₃ concentration 32.9 nmol/l], but this did not achieve statistical significance and no adjustments for confounders were made. Finally, in a Spanish cohort study (Fernandez-Alfonso et al. ,118 composite bias score 3, medium risk) there was no significant difference in mean maternal 25(OH)D concentration measured at 11–14 weeks in those who delivered preterm (n = 33) and those who delivered at term (n = 433); again, no consideration for confounding factors was made.

Intervention studies

No intervention studies were identified.

Discussion

The data relating maternal 25(OH)D status to risk of offspring preterm birth are all observational. Theresults of the studies are varied but do not support the use of maternal supplementation to prevent this obstetric outcome.

Observational studies (see Appendix 6 , Table 22 )

Three observational studies (two case–control and one cohort), all from Scandinavia, were identified, relating maternal 25(OH)D status to the risk of type 1 diabetes mellitus in the offspring. 125–127 Only one ofthese studies used 25(OH)D concentration; the other two attempted to estimate vitamin D intake. Sorensen et al. 125 (composite bias score 8, low risk) performed a case–control study of 109 children with type I diabetes mellitus (mean age 9 years) and 219 controls within a cohort of 29,072 individuals. 25(OH)D concentration had been measured at a median of 37 weeks’ gestation. The mean 25(OH)D concentration in the mothers of cases was 65.8 nmol/l and in the mothers of controls was 73.1 nmol/l. Compared with children of mothers whose levels were > 89 nmol/l, children of mothers whose 25(OH)D concentrations in late pregnancy were ≤ 54 nmol/l were at increased risk of developing type 1 diabetes mellitus. Stene and Joner126 (composite bias score 2, medium risk) performed a case–control study comparing 545 children with type 1 diabetes mellitus (mean age 10.9 years) with 1668 matched controls. Maternal use of vitamin D supplementation during pregnancy was assessed retrospectively by questionnaire and no association was found between maternal vitamin D supplementation in pregnancy and risk of offspring type 1 diabetes mellitus. Marjamaki et al. 127 (composite bias score 6, low risk) studied a prospective cohort of 3723 children who were at an increased genetic risk of developing diabetes mellitus. Among this cohort 74 children developed type 1 diabetes mellitus over the mean observation period of 4.3 years. Maternal vitamin D intake was assessed retrospectively from a FFQ completed 1–3 months after delivery and which was focused on food and supplements taken in the eighth month of pregnancy. There was no statistically significant relationship observed between maternal vitamin D intake either from food or supplements, and risk of offspring type 1 diabetes mellitus.

A further study by Krishnaveni et al. 105 (composite bias score 4, medium risk), using a cohort of 506 Indian children aged 5 years (469 of whom were also followed up to 9.5 years), did not measure rates of type 1 diabetes mellitus per se, but measured fasting glucose, fasting insulin, insulin resistance and insulin increment 30 minutes after a glucose tolerance test in the children. No significant association was found between any of these offspring measurements at age 5 years and maternal 25(OH)D concentration, measured at 28–32 weeks’ gestation. At age 9 years, however, a significant inverse relationship was observed between maternal 25(OH)D concentration and offspring fasting insulin and insulin resistance after adjustment for child sex and age, maternal BMI, gestational diabetes mellitus, socioeconomic score, parity and religion.

Intervention studies

No intervention studies were identified.

Discussion

The three observational studies125–127 relating maternal serum 25(OH)D status to risk of offspring type 1 diabetes mellitus were assessed to be of moderate to low risk of bias and were generally consistent in suggesting an inverse relationship. However, one127 used vitamin D dietary intake and there were no intervention studies. Thus, maternal vitamin D supplementation to prevent offspring type 1 diabetes mellitus cannot be recommended; however, high-quality intervention studies are warranted.

Observational studies (see Appendix 6 , Table 23 )

Three observational studies91,92,119 (two cross-sectional studies and one cohort study) examining the relationship between low-birthweight infants and maternal 25(OH)D concentration were identified. All studies were from the developing world (Iran and Tanzania) and composite bias scores ranged from −2 to 3, indicating a high–medium risk of bias. The definition of low birthweight (< 2500 g) was consistent across all three studies. Two studies92,119 directly measured maternal serum 25(OH)D and reported no association with low-birthweight infants. In one study, by Sabour et al. ,91 maternal vitamin D intake during pregnancy was estimated from FFQs completed by 449 Iranian pregnant women at delivery. The incidence of low birthweight was lower in the offspring of women with adequate intake of calcium and vitamin D (100 mg calcium, 200 IU vitamin D per day) than in the offspring of those with inadequate intake (numbers not given). This study achieved the lowest composite bias score (composite bias score −2) of these studies, indicating the highest risk of bias; no consideration for potential confounders was made.

Two studies reported no significant relationship between maternal 25(OH)D and risk of offspring low birthweight. Maghbooli et al. 92 (composite bias score 1, medium risk), in a second cross-sectional study from Iran, measured maternal 25(OH)D at delivery in 552 Iranian women. The study reported that 5.4% (approximately n = 30) of the cohort had low-birthweight offspring. No significant difference in mean 25(OH)D was observed between mothers of low-birthweight offspring and mothers of normal-weight offspring [mean 25(OH)D concentration in each group not given]. Similarly, Mehta et al. 119 (composite bias score 3, medium risk), in a cohort study of 1078 HIV-infected women taking part in a vitamin supplement trial, found no significantly increased odds of low-birthweight infants (n = 80) in mothers with a 25(OH)D concentration < 80 nmol/l compared with those with a concentration > 80 nmol/l. In this study a threshold of 80 nmol/l was used to divide maternal 25(OH)D concentration into adequate or low. Adjusting the analysis for maternal multivitamin supplementation, age at baseline, cluster differentiation 4 (CD4) count at baseline and HIV disease stage did not alter the findings.

Intervention studies

No intervention studies were identified.

Discussion

Of the three observational studies relating maternal 25(OH)D status to risk of low birthweight in the offspring, only one91 demonstrated a positive result, suggesting that low birthweight was less likely where women took at least 100 mg of calcium and 200 IU vitamin D daily. However, this study was judged to be at high risk of bias; the remaining two studies92,119 demonstrated no relationship and, therefore, maternal vitamin D supplementation cannot be recommended to prevent low birthweight. Larger prospective observational studies in several different populations would be sensible before moving to an intervention study.

Observational studies (see Appendix 6 , Table 24 )

One observational study examining the relationship between maternal vitamin D status and offspring serum calcium concentration was identified. In a cross-sectional study of 264 women in Saudi Arabia, Ardawi et al. 90 found no significant correlation between maternal 25(OH)D measured at delivery and offspring venous umbilical cord blood calcium concentration. A relationship was still not observed even if the group was divided using a maternal 25(OH)D concentration of 20 nmol/l as a threshold. This study was assessed to have a low risk of bias (composite bias score 5); however, no adjustments were made for potential confounding factors.

Intervention studies (see Appendix 6 , Table 25 )

Seven clinical trials4–7,20,21 of maternal vitamin D supplementation were identified; all measured venous umbilical cord calcium concentration at delivery and three3,6,20 went on to measure offspring venous calcium again within the first week of life. None of the trials was within the last 20 years and all were found to have a high risk of bias (composite bias score −9 to −1). Sample sizes ranged from 40 to 1139. Five studies4–7,136 reported adequate randomisation; however, only two trials3,20 were placebo controlled and only one3 was of double-blind design. Supplementation strategies were highly variable: six trials3–7,21 supplemented pregnant women with vitamin D in the last trimester; one study20 supplemented from 12 weeks onwards. There was also much diversity with regards to the type of supplementation used, ranging from 1000 IU ergocalciferol daily (with or without calcium) in the last trimester3,6,7,21 to bolus oral dosing of 600,000 IU cholecalciferol twice in the last trimester. 4,5 Six studies3–6,20,21 reported higher offspring calcium concentrations in the supplemented group than in the unsupplemented group; one trial7 showed no difference in offspring venous calcium regardless of maternal vitamin D supplementation strategy.

Brooke et al. 3 (composite bias score −2, high risk), in a trial of ergocalciferol supplementation in the last trimester of pregnancy of Asian women living in the UK, found no difference in umbilical cord calcium concentration between groups, but neonatal serum calcium was greater in offspring of supplemented mothers than in the offspring of mothers who had received placebo at 3 and 6 days postnatally. There were five cases of symptomatic hypocalcaemia in the control group but none in the treatment group. Higher rates of breastfeeding were observed in the treatment group which in itself was positively associated with offspring venous calcium concentration and was not controlled for in analysis. Similar findings were noted in a larger (n = 1139) British study by Cockburn et al. 20 (composite bias score −1, high risk) and in a French study by Delvin et al. 6 (composite bias score −2, high risk). Neither study found a difference in venous cord calcium concentrations between the supplemented and unsupplemented groups, but both found higher infant venous calcium concentrations in the supplemented group, at days 620 and 4. 6 The third, and most recent, British study (Congdon et al. 21) found that cord calcium was significantly higher in the offspring of Asian women supplemented with daily 1000 IU vitamin D plus calcium in the last trimester than in the offspring of those who received no supplement. This study was assessed to have the highest risk of bias with a composite bias score of −9. The number of subjects in this trial was low, with only 19 receiving supplement, and no information about randomisation or whether or not blinding was implemented were reported. These findings are in agreement with two Indian studies, both by Marya et al. 4,5 (1981, composite bias score −6, high risk; 1989, composite bias score −2, high risk). Both studies found that cord calcium concentrations were significantly higher in those pregnant women supplemented with two doses of 600,000 IU cholecalciferol in months 7 and 8 of gestation than in the unsupplemented group.

In contrast, a French study (Mallet et al. 7 composite bias score −3, high risk) found no effect of maternal vitamin D supplementation in the third trimester on cord calcium concentration, regardless of whether supplementation was provided at 1000 IU per day for 3 months or as a single high dose of 200,000 IU in the seventh month of gestation.

Evidence synthesis

The available published results were combined in two separate models. The first meta-analysis included the studies of Cockburn et al. ,20 Brooke et al. ,3 Marya et al. 4 (low dose of vitamin D), Mallet et al. 7 (low dose of vitamin D) and Delvin et al. 6 (see Appendix 7 , Figure 8 ). Owing to statistically significant heterogeneity in the results (I ² = 67.6%, p = 0.015), a REM was fitted. Serum calcium concentration in the supplemented group did not differ from that in the unsupplemented group (mean difference 0.01 mmol/l, 95% CI −0.02 mmol/l to 0.04 mmol/l). The second meta-analytic model included the studies by Cockburn et al. ,20 Brooke et al. ,3 Marya et al. 4 (high dose of vitamin D), Mallet et al. 7 (high dose of vitamin D), Delvin et al. 6 and Marya et al. 5 (see Appendix 7 , Figure 9 ). As in the previous model, a REM was fitted owing to significant heterogeneity (I ² = 90%, p < 0.001). The combined results showed that the mean difference of serum calcium concentration between the supplemented and the unsupplemented groups was significantly different from 0 (mean difference 0.05 mmol/l, 95% CI 0.02 mmol/l to 0.05 mmol/l).

Discussion

The majority of the intervention studies and the one observational study consistently demonstrated positive relationships between maternal 25(OH)D status and offspring serum calcium concentrations measured either in venous umbilical cord serum or from postnatal venesection. Some also found a reduced risk of hypocalcaemia in the neonate. Meta-analysis of higher-dose intervention studies also suggested a positive effect. However, these intervention studies were all felt to be at high risk of bias and none of them was published within the last 20 years. Assay technology has improved dramatically over recent decades and the reliability of the relationships must be open to question. Given the known physiology of the vitamin D axis in adults, a positive association between maternal 25(OH)D and offspring calcium concentration might not be a surprising finding; however, little is known about relationships between 25(OH)D and fetal calcium concentrations in utero. Furthermore, none of the identified studies addressed postnatal factors such as mode of feeding (breast vs. formula) as potential risk modifiers. A positive relationship between maternal 25(OH)D status and offspring calcium concentrations does not justify intervention unless the increased calcium concentration brings a benefit. Symptomatic hypocalcaemia did not appear to be found in all studies and is likely to be much more common in high-risk populations. It seems reasonable, on the basis of the current evidence, to suggest that maternal vitamin D supplementation is likely to reduce the risk of neonatal hypocalcaemia, but that the dose required, duration and target group is currently unclear (e.g. by skin colour, ethnicity, or mode of infant feeding), and might usefully form the basis of further investigation.

Observational studies (see Appendix 6 , Table 26 )

Two cohort studies were identified which examined the relationship between maternal serum 25(OH)D concentration in pregnancy and offspring blood pressure. Both studies were of cohort design and measured maternal serum 25(OH)D in late pregnancy. Composite bias score was 4 for both, indicating a medium risk of bias. Gale et al. 24 measured blood pressure in 178 children aged 9 years in the Princess Anne Cohort study, UK. No association was observed between maternal 25(OH)D and offspring blood pressure. Krishnaveni et al. ,105 using a larger Indian cohort of 338 mother–offspring pairs, measured blood pressure in the offspring at two time points: age 5 and 9.5 years. Similarly, no significant difference in blood pressure was observed between those children born to mothers with vitamin D deficiency (defined by the authors as < 37.5 nmol/l) and those born to mothers without vitamin D deficiency. Adjustments for offspring sex and age, maternal BMI, gestational diabetes mellitus, socioeconomic score, parity and religion made little difference to the results.

Intervention studies

No intervention studies were identified.

Discussion

Neither of the two observational studies relating maternal 25(OH)D status to offspring blood pressure demonstrated a statistically significant relationship and therefore no treatment recommendation can be made.

Observational studies

No observational studies of maternal vitamin D status and offspring rickets were identified.

Intervention studies

No intervention studies of maternal vitamin D supplementation and offspring rickets were identified. A UK trial, by Congdon et al. ,21 found no difference in the incidence of offspring craniotabes between the supplemented group (n = 4) and the unsupplemented group (n = 3). This study was assessed to have a high risk of bias, with a composite bias score of −9.

Discussion

It is interesting that there are so few data relating maternal 25(OH)D status to offspring rickets. However, rickets does not tend to manifest until the first year of life, in contrast to neonatal hypocalcaemia, and therefore it is likely that the determinant is the child’s own sun exposure and vitamin D intake. If the child is wholly breastfed and receives little sun exposure then increased risk of rickets might be expected. However, this scenario does not fall within the remit of the current review.

Observational studies (see Appendix 6 , Table 27 )

Eleven observational studies were identified, comprising six case–control,128–133 four cohort117,118,134,135 and one cross-sectional study. 122 The case–control studies were generally of small size with the minimum number of 12 cases and maximum 55 cases and the number of control subjects ranging from 24 to 220. The definition of pre-eclampsia was similar across studies: new-onset gestational hypertension after 20 weeks [systolic blood pressure persistently (two or more occasions) ≥ 140 mmHg and/or diastolic blood pressure ≥ 85 or ≥ 90 mmHg] and proteinuria (either 300 mg protein excreted in the urine in 24 hours, or a random sample of between 1+ and 2+ protein on urine dipstick, or a protein–creatinine ratio > 0.3). Two of the case–control studies129,130 identified cases of severe pre-eclampsia only, using the American Congress of Obstetrics and Gynaecology 2002 definition [systolic blood pressure ≥ 160 mmHg and/or a diastolic blood pressure ≥ 110 mmHg on at least two occasions plus proteinuria (≥ 300 mg in a 24-hour collection or 1+ on urine dipstick), or systolic blood pressure ≥ 140 mmHg and/or diastolic blood pressure ≥ 90 mmHg plus 5 g proteinuria in a 24-hour period after 20 weeks’ gestation]. All six case–control studies, the cross-sectional study and three of the five cohort studies used serum 25(OH)D concentration as the marker of maternal vitamin D status,117,118,122,128–133 with the other two cohort studies134,135 using dietary intake. The timing of serum measurements varied across the studies with some measuring in the first trimester118,132 and others in the last,128,131 and one study133 at three time points. Composite bias scores ranged from 2 to 9, indicating that studies were considered low to medium risk of bias. Confounding factors were variably included and there was also variation in the criteria for matching to controls.

Of the included studies, three (one case–control, one cross-sectional and one cohort) reported statistically significant inverse associations between maternal vitamin D status and risk of pre-eclampsia. A further two case–control studies demonstrated a similar association between maternal 25(OH)D and risk of severe pre-eclampsia. A nested case–control study (55 cases and 220 randomly selected, unmatched controls from a cohort of 1198) from Bodnar et al. 128 (composite bias score 8, low risk) measured 25(OH)D in nulliparous pregnant women living in Pittsburgh, USA, at two time points (before 22 weeks’ gestation and pre-delivery). A significant inverse relationship was observed at both time points. At < 22 weeks’ gestation a 50 nmol/l reduction in maternal 25(OH)D was associated with an over twofold increased risk of pre-eclampsia after adjusting for maternal race, ethnicity, pre-pregnant BMI, education, season and gestational age at blood sample. A cross-sectional study from Pakistan (Hossain et al. ,122 composite bias score 4, medium risk) measured maternal 25(OH)D₃ at delivery in 75 women (76% of whom covered their face, arms, hands and head). Although the number of pre-eclampsia cases is not given, when the group was divided into thirds, a significantly increased risk of pre-eclampsia was observed for those in the lowest and middle tertile compared with the highest. The relationship between maternal 25(OH)D and pre-eclampsia was only observed in individuals with serum 25(OH)D < 50 nmol/l. In contrast to other studies, women were classified as having pre-eclampsia based on blood pressure alone (systolic blood pressure ≥ 140 mmHg and/or diastolic blood pressure ≥ 90 mmHg). The largest study to date (Haugen et al. ,134 composite bias score 2, medium risk) followed up a cohort of 23,425 pregnant women enrolled in the Norwegian Mother and Child Cohort Study. Maternal 25(OH)D was not directly measured, but estimated from a FFQ at 22 weeks. A total of 1267 cases of pre-eclampsia were identified. Lower total vitamin D intake was associated with a significantly increased risk of pre-eclampsia.

Both studies examining the relationship between severe pre-eclampsia and maternal 25(OH)D demonstrated significant inverse associations. Both were US-based case–control studies with a comparable number of cases and controls, and assessed to have a low risk of bias. Baker et al. 129 (composite bias score 9, low risk) identified 44 cases and 201 randomly selected controls matched by race/ethnicity from a cohort of 3992 women. Significantly higher odds of severe pre-eclampsia were found in those with maternal 25(OH)D < 50 nmol/l than in those with 25(OH)D > 50 nmol/l, even after adjusting for season of blood sampling, maternal age, multiparity, BMI, gestational age at blood sample. Similarly, Robinson et al. 130 (composite bias score 5, low risk), in a study of 50 cases and 100 controls matched for race and gestational age at the time of sample, found that the odds of severe pre-eclampsia significantly reduced as maternal 25(OH)D increased even after adjusting for maternal BMI, maternal age, African American race and gestational age at sample collection.

Six studies, however, found no association between maternal vitamin D status and pre-eclampsia risk. Seely et al. 131 (composite bias score 2, medium risk) observed no significant difference in late-pregnancy mean maternal 25(OH)D in 12 women with pre-eclampsia and 24 control women of similar age, gestation, height, weight, parity (primiparous or not) and ethnicity (Caucasian or not). A second US nested case–control study from Powe et al. 132 (composite bias score 4, medium risk) drew similar conclusions. In this study of 39 cases and 131 unmatched controls from an overall cohort of 9930, the odds of pre-eclampsia were not related to first-trimester maternal 25(OH)D concentration. Adjusting for maternal BMI, non-white race and summer blood collection made no difference to the results. A significant relationship was still not seen even when the analysis was restricted to mothers with a serum 25(OH)D concentration < 37.5 nmol/l. A further US nested case–control study from Azar et al. 133 (composite bias score 5, low risk) assessed pre-eclampsia risk in only white women, all with type 1 diabetes mellitus, who had serum 25(OH)D measured at three time points during their pregnancy (early, mid and late pregnancy). Twenty-three cases were identified and compared with 24 controls, matched for age, diabetes mellitus duration, glycated haemoglobin (HbA_1c) level and parity, out of a cohort of 151. Again, no statistically significant relationship between maternal 25(OH)D, measured at any time point, and pre-eclampsia risk was observed. A Canadian study of 221 pregnant women with clinical or biochemical risk factors for pre-eclampsia (Shand et al. ,117 composite bias score 6, low risk) found no significantly increased odds of pre-eclampsia in pregnant women with mid-pregnancy 25(OH)D concentrations < 37.5, < 50 or <75 nmol/l compared with those with 25(OH)D concentrations > 75 nmol/l. However, only 28 cases of pre-eclampsia were identified. The most recent study by Fernandez-Alonso et al. 118 (composite bias score 3, medium risk), again, found no difference in mean early pregnancy maternal 25(OH)D between those who developed pre-eclampsia and those with normal pregnancies. This study included the lowest number of cases (n = 7). Finally, Oken et al. 135 (composite bias score 5, low risk) identified 58 cases of pre-eclampsia from the US Project Viva Cohort Study of 1718 women. Maternal serum 25(OH)D was not measured directly, but estimated from a FFQ at mean 10.4 weeks’ gestation. No significant relationship between pre-eclampsia risk and vitamin D intake was seen.

Evidence synthesis

Usable results for meta-analysis of the risk of pre-eclampsia with increased vitamin D were available from four studies: Bodnar et al. ,128 Powe et al. ,132 Robinson et al. 130 and Azar et al. 133 (early pregnancy visit). All but Bodnar et al. 128 provided unadjusted ORs. The unadjusted estimates were synthesised in a REM owing to statistically significant heterogeneity (I ² = 78.4%, p = 0.01). The pooled estimate showed no significant risk of pre-eclampsia with increased vitamin D (pooled OR 0.78, 95% CI 0.59 to 1.05; see Appendix 7 , Figure 10 ). Synthesising the available adjusted ORs from all four studies the result was very similar; there was no statistically significant increased risk of pre-eclampsia with decreased vitamin D status (pooled OR 0.75, 95% CI 0.48 to 1.19; see Appendix 7 , Figure 11 ).

Intervention studies (see Appendix 6 , Table 28 )

One clinical trial that included maternal pre-eclampsia as an outcome measure was identified. Marya et al. 136 randomised 400 pregnant women attending an antenatal clinic in India to either a trial of vitamin D plus calcium (375 mg/day calcium plus 1200 IU vitamin D) from 20 to 24 weeks until delivery or to no supplement (n = 200 in each arm). Serum 25(OH)D concentrations were not measured during the study. There were 12 cases of pre-eclampsia in the supplemented group compared with 18 cases of pre-eclampsia in the non-supplemented group, a result which did not achieve statistical significance. Systolic and diastolic blood pressure were significantly lower in the supplemented than in the unsupplemented group at 32 and 36 weeks’ gestation, but no difference was observed at 24–28 weeks’ gestation. This study had a composite bias score of −2, indicating a high risk of bias, and clearly could not separate an effect of vitamin D from that of calcium supplementation.

Discussion

As with many other outcome measures, results of the various observational studies were conflicting, with some demonstrating an inverse association between maternal vitamin D status and risk of pre-eclampsia122,128–130,134 and others no relationship. 117,118,131–133,135 Both studies looking at the risk of severe pre-eclampsia found statistically significant inverse relationships with maternal 25(OH)D concentration. 129,130 There was, however, significant heterogeneity between studies in terms of gestational age at which maternal vitamin D status was assessed, confounding factors adjusted for and the definition of pre-eclampsia used. Most observational studies were case–control and included only small numbers of women with pre-eclampsia (n = 7118 to 55128). Only one intervention study136 was identified. This was of reasonable size; however, the study was assessed to have a high risk of bias and the supplemented group received calcium and vitamin D together, rather than vitamin D alone. No difference in the risk of pre-eclampsia was identified in the unsupplemented group. Thus, it is difficult to make any treatment recommendations based on the current evidence. Further high-quality intervention studies are needed.

Observational studies (see Appendix 6 , Table 29 )

Eight observational studies (four case–control, one cross-sectional and three prospective cohort) examined relationships between maternal 25(OH)D status and risk of gestational diabetes mellitus. 93,95,118,137–141 One study, by Maghbooli et al. ,137 found, in a cross-sectional cohort of 741 Iranian women, that mean 25(OH)D concentrations (measured at 24–28 weeks) were lower in the 52 subjects who had gestational diabetes mellitus (16.5 nmol/l) than in the 527 women who did not (23 nmol/l). There was no adjustment for confounding factors in this analysis and the overall bias score was 3, indicating a medium risk for bias. A further study from Iran, of case–control design (Soheilykhah et al. ,138 composite bias score 3, medium risk), found significantly increased odds of gestational diabetes mellitus in those with 25(OH)D concentrations < 37.5 nmol/l (measured between 24 and 28 weeks). Thus, the mean 25(OH)D concentration was 24 nmol/l in those with gestational diabetes mellitus and was 32.3 nmol/l in those without gestational diabetes mellitus. Clifton-Bligh et al. ,95 in a prospective cohort of 307 women in New South Wales, Australia, found that the mean 25(OH)D concentration (measured at a mean of 28.7 weeks) was 48.6 nmol/l in 81 women with gestational diabetes mellitus compared with 55.3 nmol/l in women without. They also found that serum 25(OH)D concentration was negatively associated with fasting glucose after adjustment for age, BMI and season. This study was found to be of low risk of bias with a score of 6. Zhang et al. 139 performed a nested case–control study within a US cohort (n = 953), containing 57 women with gestational diabetes mellitus (70% white ethnicity) and 114 controls (84% white ethnicity). Controls were frequency matched to cases by the estimated season of conception. After adjustment for maternal age, ethnicity, family history of type 2 diabetes mellitus and pre-pregnant BMI, 25(OH)D concentration < 50 nmol/l was associated with increased odds of gestational diabetes mellitus, compared with women with concentrations > 75 nmol/l. This study, again, achieved a low risk of bias, with composite score of 8.

In contrast, an Indian prospective cohort study (Farrant et al. ,93 composite bias score 5, low risk) found no difference in 25(OH)D concentrations between those with gestational diabetes mellitus [n = 34, mean 25(OH)D concentration 38.8 nmol/l] and those without [n = 525, mean 25(OH)D concentration 37.8 nmol/l] (p = 0.8). No associations were found by three further studies: Makgoba et al. 140 (composite bias score 7, low risk), in a nested case–control study of 90 women with gestational diabetes mellitus and 158 controls, within an overall cohort of 1200 women, found no difference in serum 25(OH)D concentration (47.2 nmol/l in cases vs. 47.6 nmol/l in controls, measured at 11–13 weeks’ gestation). An inverse relationship was found between the serum 25(OH)D concentration and fasting glucose, glucose concentration 2 hours after a glucose tolerance test, and HbA_1c at 28 weeks’ gestation. However, after adjustment for BMI, gestation at the time of blood sampling, smoking, ethnicity, parity, maternal age, conception status, previous gestational diabetes mellitus and season, only the relationship with 2-hour glucose concentration remained statistically significant. A nested case–control study (Baker et al. ,141 composite bias score 7, low risk), this time set within a US cohort of 4225 women in whom serum 25(OH)D concentration was assessed at 11–14 weeks’ gestation, found that among the 60 cases of gestational diabetes mellitus and 120 controls, after adjustment for maternal age, insurance status, BMI, gestational age at sample collection and season, there was no association between serum 25(OH)D concentration and gestational diabetes mellitus. Finally, in a Spanish prospective cohort of 466 women (Fernandez-Alonso et al. ,118 composite bias score 3, medium risk) in whom 25(OH)D concentrations were measured at 11–14 weeks, there was no statistically significant relationship between baseline 25(OH)D concentration and development of gestational diabetes mellitus.

Intervention studies

No intervention studies were identified.

Discussion

Several large studies, of low to moderate risk of bias, found no relationship between maternal 25(OH)D status and risk of gestation diabetes mellitus. Although two Iranian studies137,138 did find an increased risk of gestational diabetes mellitus in women with low levels of 25(OH)D, these seem at odds with the majority of investigations from elsewhere and thus there appears to be no consistent evidence on which to base a recommendation of vitamin D supplementation to prevent gestational diabetes mellitus.

Observational studies (see Appendix 6 , Table 30 )

Six observational studies90,118,142–145 were identified, one of which was case–control144 and the others cohort designs. 90,118,142,143,145 Two studies142,143 found inverse relationships between 25(OH)D status and risk of caesarean section, with the remaining studies demonstrating no statistically significant associations. 90,118,144,145 Scholl et al. 142 (composite bias score 5, low risk) studied 290 women who delivered by caesarean section, out of a cohort of 1153 pregnant women. 25(OH)D concentration was assessed at a mean of 13.7 weeks’ gestation. Compared with women who had serum 25(OH)D concentrations between 50 and 125 nmol/l in early pregnancy, those who had levels < 30 nmol/l appeared at increased risk of caesarean section, and this association persisted after adjustment for age, parity, ethnicity, gestation at entry to study, season and BMI. Merewood et al. 143 (composite bias score 6, low risk), in a cross-sectional study of US women, found increased odds of caesarean section if maternal 25(OH)D concentration was < 37.5 nmol/l in 67 cases of caesarean section compared with 277 controls, after adjustment for ethnicity, alcohol use in pregnancy, educational status, insurance status and age.

Ardawi et al. 90 (composite bias score 5, low risk) studied a cohort of 264 women in Jeddah, Saudi Arabia. Among women with serum 25(OH)D status < 20 nmol/l the frequency of caesarean section was 12.5%, compared with a frequency of 9.6% in those with serum concentrations above this level, a difference which did not achieve statistical significance. A Pakistani study (Brunvand et al. ,144 composite bias score 1, medium risk) of nulliparous Pakistani women of low social class found that the median 25(OH)D concentration in 37 women who delivered by caesarean section (measured just before delivery) was 26 nmol/l, compared with 19 nmol/l in 80 controls who delivered vaginally. This did not, however, achieve statistical significance. A UK cohort study of 1000 pregnancies yielded 199 caesarean sections (Savvidou et al. ,145 composite bias score 7, low risk) and found no relationship between 25(OH)D concentration measured between 11 and 13 weeks’ gestation and risk of caesarean section, after adjustment for maternal age, racial origin, smoking, method of conception and season. Finally, in the Spanish study of Fernandez-Alonso et al. 118 (composite bias score 3, medium risk), 105 of the cohort of 466 women underwent caesarean section. There was no relationship between 25(OH)D concentration, measured between 11 and 14 weeks’ gestation, and risk of caesarean section.

Intervention studies

No intervention studies were identified.

Discussion

The data relating to caesarean section are all observational and conflicting. Given that many other factors will influence risk of caesarean section, including physician preference, local policy and pre-existing morbidity, it seems likely that any relationships between maternal 25(OH)D concentration and caesarean section risk will be difficult to extricate from the surrounding noise. The current evidence base does not support use of vitamin D supplementation to reduce risk of caesarean section and a well-designed, prospective observational study is warranted before moving to intervention studies.

Observational studies (see Appendix 6 , Table 31 )

Three studies146–148 were identified (two cohort, one cross-sectional) which examined relationships between maternal 25(OH)D status and bacterial vaginosis. All three studies elucidated statistically significant relationships although at very different thresholds of 25(OH)D concentration. Bodnar et al. 146 (composite bias score 5, low risk) studied 469 women, all of whom were non-Hispanic and white or black. 25(OH)D concentration was measured at a mean of 9.5 weeks’ gestation. Among the 192 cases of bacterial vaginosis, median 25(OH)D concentration was 29.5 nmol/l, compared with 40.1 nmol/l in the non-diseased women. At 25(OH)D concentrations < 80 nmol/l there was an inverse association between frequency of bacterial vaginosis and early pregnancy serum 25(OH)D concentration (p < 0.0001). Above this threshold no relationship was observed. Results were adjusted for the presence of sexually transmitted diseases. Using the National Health and Nutrition Examination Survey (NHANES) cohort, Hensel et al. 147 (composite bias score 4, medium risk) found a statistically significantly increased risk of bacterial vaginosis in those women whose serum 25(OH)D concentration was < 75 nmol/l. However, it is unclear at what stage 25(OH)D concentration was measured, and the mean 25(OH)D concentrations, together with the unadjusted analyses, are not presented. Dunlop148 (composite bias score 2, medium risk) sampled 160 non-Hispanic white/non-Hispanic black women from a total of 1547 women participating in the Nashville Birth Cohort. In this cross-sectional analysis, risk of bacterial vaginosis was higher in women whose serum 25(OH)D concentration at delivery was < 30 nmol/l than in those whose levels were above this threshold, after adjustment for race, age, smoking, BMI, gestational age at delivery and health-care funding source.

Intervention studies

No intervention studies of maternal vitamin D supplementation on risk of bacterial vaginosis were identified.

Discussion

Although reasonably large, only three studies146–148 were identified that reported bacterial vaginosis as an outcome. Each study differed in methodology, using differing thresholds for low serum vitamin D, and there remains a strong possibility of residual confounding which may account for the relationships between bacterial vaginosis and maternal vitamin D. Thus, the evidence base does not currently warrant the recommendation of vitamin D supplementation to reduce the risk of bacterial vaginosis, and further high-quality prospective observational studies are required before moving to an intervention study.

Completed studies (systematic reviews)

• DARE (CRD).

• CDSR.

• HTA database (CRD).

Completed studies (other study types)

• CENTRAL.

• MEDLINE.

• EMBASE.

• BIOSIS.

• Google Scholar.

• AMED.

Ongoing studies

• National Research Register archive.

• UKCRN Portfolio.

• Current Controlled Trials.

• ClinicalTrials.gov.

Grey literature

• Conference Proceedings Citation Index-Science (1990–present).

• Zetoc conference search.

• SACN website.

• Department of Health website.

• The King’s Fund library database.

• Trip database.

• HTA website.

• HMIC database.