Efficacy of vitamin D supplementation in gestational diabetes mellitus: Systematic review and meta-analysis of randomized trials

Background Trials have examined on the benefits of vitamin D supplementation in pregnant women. Objective This review aimed to evaluate whether oral vitamin D supplements, when given to pregnant women with gestational diabetes mellitus (GDM), would improve maternal and neonatal outcomes, compared with no treatment or placebo. Method We performed a systematic review following Cochrane methodology, and randomized trials were included where pregnant women with GDM received vitamin D supplementation versus placebo/no treatment or vitamin D and calcium versus placebo/no treatment. Primary outcomes were preeclampsia, preterm birth, cesarean delivery, gestational hypertension, and adverse events related to vitamin D supplementation. The search strategies were applied to the following databases: MEDLINE, Embase, LILACS, and CENTRAL. Similar outcomes in at least two trials were plotted using Review Manager 5.3 software. The quality of evidence was generated according to the Grading of Recommendations Assessment, Development, and Evaluation (GRADE). Results The total of 1224 references were identified, eleven trials were potentially eligible, and six were included in this review (totaling 456 women). The meta-analysis of frequency of cesarean deliveries did not show significant differences between groups, none of the trials evaluated the remaining primary outcomes. For secondary outcomes, our results suggest that vitamin D supplementation in pregnant women with GDM may reduce newborn complications such as hyperbilirubinemia, polyhydramnios (RR: 0.40, 95% CI: 0.23 to 0.68; RR: 0.17, 95% CI: 0.03 to 0.89; respectively), and the need for maternal or infant hospitalization (RR: 0.13; 95% CI: 0.02 to 0.98; RR: 0.40, 95% CI: 0.23 to 0.69). However, the evidence was of low or very low quality. Conclusion We did not find moderate or high quality evidence indicating that vitamin D supplementation, when compared with placebo, improves glucose metabolism, adverse maternal and neonatal outcomes related to GDM in pregnant women.


Introduction
GDM is associated with maternal and neonatal risks [1]. A large, multinational cohort study, Hyperglycemia and Adverse Pregnancy Outcome study [2], demonstrated that GDM or obesity alone, compared with normoglycemic pregnant controls, had significantly greater odds of low birth weight, newborn percent body fat, primary cesarean delivery, and preeclampsia. In addition, the risk of adverse maternal, fetal, and neonatal outcomes continuously increased as a function of maternal glycemia at 24-28 weeks, even within ranges previously considered normal [2]. In addition, a systematic review showed a positive association between maternal hyperglycemia and caesarean section, induction of labor, large for gestational age, macrosomia, and shoulder dystocia [3].
Vitamin D is a group of fat-soluble secosteroids predominantly found in fish-liver oils, fatty fish, mushrooms, egg yolks, and liver. Furthermore, vitamin D can also be produced in the body in the presence of sunlight [4]. Its two physiological active forms are vitamin D 3 (also known as cholecalciferol) and vitamin D 2 (ergocalciferol). In response to parathyroid hormone, both forms are first hydroxylated in the liver to 25-hydroxyvitamin D (25(OH)D or calcidiol), and sequentially converted to 1,25-dihydroxyvitamin D (calcitriol) in the kidneys. Vitamin D 3 is three times more effective than vitamin D 2 in increasing vitamin D concentrations and maintaining those levels for a longer period of time [5].
The most common function of vitamin D is the maintenance of calcium homeostasis and bone integrity. However, besides bone and parathyroid glands, there are other sites of vitamin D action, which includes the skin, intestines, immune system, and pancreas. It is known that vitamin D has immunomodulatory and anti-inflammatory effects [6], and these effects aids glucose metabolism by regulating the release of insulin according to levels of glucose [7][8][9][10].
This effect of vitamin D explains the association between maternal vitamin D deficiency in early pregnancy and the elevated risk for GDM, and it has been demonstrated in some observational studies [11]. A case-control study involving 1280 women with GDM and 3438 controls evaluated the association of 25(OH)D concentrations with risk of GDM. After adjusting for confounding factors, women with low concentrations (< 50.0 nmol/L) of 25(OH)D displayed a significantly increased risk of GDM and adverse pregnancy outcomes (e.g., anemia, macrosomia, abnormal amniotic fluid, and stillbirth or miscarriage) [12].
A cohort study conducted in Brazil evaluated the effect of vitamin D deficiency on neonatal outcomes of pregnant women with GDM. The authors identified that newborns of women with vitamin D deficiency had a significantly higher incidence of hospitalization in critical care units, hypoglycemia, and small size for gestational age. The incidence of prematurity, jaundice, and dystocia of the shoulder were not statistically significant between groups [13].
As vitamin D supplementation can be easily administered without apparent serious adverse events, and vitamin D deficiency is frequent among pregnant women, studies have examined on vitamin D supplementation in pregnant women with GDM [13,14]. Li et al. [14] evaluated the effect of vitamin D 3 -supplemented yogurt on glucose metabolism and lipid concentrations in pregnant women with GDM. After 16 weeks of intervention, both fasting plasma glucose and lipid levels were markedly lower in the supplementation group than control participants.
Thus, under the hypothesis that vitamin D supplementation during pregnancy protects pregnant women against adverse outcomes related to GDM, this review aimed to evaluate whether oral supplements with vitamin D given to pregnant women with GDM would improve maternal and neonatal outcomes, compared to no treatment or placebo.

Protocol and registration
We performed a systematic review following Cochrane methodology [15] and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines [16]. Its protocol was registered in the International Prospective Register of Systematic Reviews database, accessible under the protocol number CRD42016021971.

Eligibility criteria
We included randomized controlled trials (RCT), which adopted the Patient Intervention Comparison Outcome structure.

Patients
We were exclusively interested in pregnant women diagnosed with GDM. We considered GDM as diabetes diagnosed in the second or third trimester of pregnancy by fasting plasma glucose levels or oral glucose tolerance test (50 and 100 g) at 24-28th week of gestation, and according to either the American Diabetes Association, World Health Organization, National Institute of Health, or International Association of the Diabetes and Pregnancy Study Groups.

Exclusion criteria
We excluded trials of pregnant women without GDM, with calcium metabolic disorder, known diabetes prior to pregnancy, and diabetes diagnosed in the first trimester of gestation.

Search strategy
The following electronic databases were consulted: MEDLINE (via PubMed; 1968 to Septem- The basic research strategy was developed for PubMed and modified as required for other databases (S1 Appendix). We used the health descriptors available in Descriptors in Health Sciences and Medical Subject Heading. The basic research strategy included "Diabetes, Gestational," "Diabetes Mellitus, Type 2", "Pregnancy", "Vitamin D", "Cholecalciferol", and "Vitamin D Deficiency". There was no language restriction. References to selected articles, including relevant review articles, were reviewed to identify all relevant studies. A manual search of references of RCTs was carried out in relevant journals and congresses in the area.

Selection of studies
For this review, two researchers (MRKR and SAML) independently reviewed the eligibility of the titles and abstracts. The studies potentially eligible for inclusion were selected for full reading and subsequently assessed for compliance with the Patient Intervention Comparison Outcome structure. Disagreements regarding the selection of articles were resolved by discussion with a third researcher (VSNN).

Data extraction
Two researchers (MRKR and SAML) independently extracted relevant data (participants, specific vitamin D intervention, and outcome characteristics) from each full-text article, using a standardized form based on the Cochrane Handbook [15]. The selection was compared for accuracy, and any discrepancies were resolved by consensus or discussion with another researcher (VSNN). If necessary, the corresponding authors of the original studies were contacted to obtain missing information. In cases of duplicate publications or multiple reports from the primary study, data extraction was optimized using the best information available for all items.

Evaluation of risk of bias
Two investigators (MRKR and SAML) independently assessed the risk of bias of each eligible RCT. Any discrepancies were resolved by consensus or discussion with another investigator (VSNN). The Cochrane Collaboration tool for risk assessment of bias in RCTs was used, which includes seven criteria described in the Cochrane Reviewers' Manual [15].

Data synthesis and analysis
Similar outcomes in at least two trials were plotted using Review Manager 5.3 software. Continuous data were expressed as mean difference (MD) and standard deviations, and the difference of means with 95% confidence interval (CI) was used as an estimate of the intervention effect. For dichotomous data, the relative risk (RR) was also calculated with 95% CI. The random effect model was used, while the inverse variance method was used to weigh the effect estimates between included trials. The inconsistency of variance between the results of the included trials was ascertained by the Higgins inconsistency test (I 2 ), [15] where I 2 < 25% indicated low probability, I 2 = 50% moderate probability and, I 2 > 75% indicated a high probability of heterogeneity. If vitamin D levels were given in ng/mL, the values were converted using the following formula: 1 ng/mL = 2.5 nmol/L.

Quality of evidence
The quality of evidence of the intervention effect estimates for the outcomes that could be plotted was generated according to GRADE (Grading of Recommendations Assessment, Development, and Evaluation) [17].

Selected articles
After searching the electronic health databases, 1224 references were identified (Fig 1). Eleven articles were potentially eligible for inclusion in this review and were therefore read in full. Vitamin D supplementation in gestational diabetes mellitus After reading in full, six studies met the inclusion criteria and were included [18][19][20][21][22]. One trial demonstrated two outcomes of interest for this systematic review (fasting glucose and calcium concentration), but the data could not be plotted in the meta-analysis, as they were not expressed in MD and standard deviation, even after contacting the authors [22].
Five studies were excluded. One study was excluded because it was uncontrolled [23]; in one study vitamin D supplementation was applied at 6 to 48 months postpartum [24]; in one study the authors randomized the patients to different doses of vitamin D (i.e., there was no control group with placebo or no treatment) [25]; one study was excluded because the selected outcomes (glycated hemoglobin, serum concentration of vitamin D and serum calcium concentration) were evaluated only postpartum [26]; and the last exclusion occurred because the patients were randomized before GDM diagnosis [27].

Description of the studies
Six RCTs were included, totaling 456 pregnant women with GDM diagnosed in the second or third trimester of pregnancy. GDM diagnosis was according to criteria by the American Diabetes Association. The trials were published between 2013 and 2016; four were developed in Iran [18][19][20][21] and two in China [14,22]. Tables 1 and 2 show the baseline characteristics and eligibility criteria of the included trials. In all trials, intervention and controls groups were similar in maternal ages, maternal height, weight at the beginning of the intervention, BMI, and concentration of 25(OH)D. In all trials, the gestational age at onset of intervention was the third trimester of pregnancy, and the status of 25(OH)D at baseline visit was not different between groups, with a mean level of approximately 20 ng/mL.
In the four Iranian trials, the period of treatment was six weeks, Li et al. [14] was 16 weeks, and Zhang et al. [22] was until delivery.  In the four Iranian trials, [18][19][20][21] the intervention participants were prescribed 50,000 IU of vitamin D 3 every 2 weeks. In Zhang et al. (2016), [22] there was more than one intervention group, and the maximum dose of vitamin D 3 was 50,000 IU every 2 weeks. In Li et al. (2016), [14] the vitamin D supplementation occurred using a vitamin D 3 -supplemented yogurt drink, providing 1000 UI daily. In all trials, the control groups were given a placebo. Two of the Iranian trials also administered calcium carbonate (1000 mg daily) to the intervention groups [19,21].

Risk of bias
The risk of bias assessment is shown in Fig 2. Regarding the randomization process, all trials used a random numerical list generated by a computer system, therefore being classified as low risk.
Regarding allocation concealment, in the four Iranian trials the allocations were concealed from the researcher and participants until the main analyses were completed (low risk of bias).    )), all outcomes were reported as intended, and we considered low risk of bias for selective reporting. For the remaining included trials, there was not enough information to fully assess the potential for selective reporting bias, therefore they were judged as being an unclear risk.

Meta-analysis
Maternal outcomes. Regarding our primary outcomes, none of the trials evaluated maternal adverse events (e.g., hypercalcemia, kidney stone, among others), neonatal death (within Vitamin D supplementation in gestational diabetes mellitus 28 days of delivery), stillbirth, low birth weight (< 2,500 g), or neonatal infection (respiratory infections within 28 days after delivery).
Two trials evaluated the frequency of preeclampsia and preterm newborns. [20,21] Due to scarce number of events, effect estimates were imprecise and consequently differences between groups were not statistically different ( Table 3).
The meta-analysis of frequency of cesarean deliveries did not show significant differences between groups [20,21], but did describe that cesarean deliveries tended to be lower in the interventional group (Table 3).
The decrease in fasting blood glucose at the end of the trial was reported by Asemi et al.  (Fig 3). In the trial by Zhang at al. (2016), there was no difference between groups for this outcome; however, data could not be plotted due missing standard deviation. Note: To determine a GRADE quality of the evidence, the GRADE approach begins by assigning findings to one of two starting levels of quality depending on the study design. Randomized trials are high quality, while observational studies are low quality. Additionally, two other levels exist; moderate and very low. This gives four levels:  The outcome of need for maternal hospitalization, as measured by Karamali et al. (2015) and Asemi et al. (2015), showed a significant difference between groups, favoring vitamin D supplementation (RR: 0.13; 95% CI: 0.02 to 0.98, Table 3), but with an important imprecision due to wide confidence interval. Asemi Fig 4).
Neonatal outcomes. The 5 minutes Apgar outcome was evaluated by Asemi et al. (2015) and Karamali et al. (2015). In Asemi et al. (2015), the Apgar score was higher in the supplementation group, and in Karamali et al. (2015), there was no difference between groups. The meta-analysis was not performed due to the presence of this heterogeneity between the two comparisons. Asemi et al. (2015) and Karamali et al. (2015) evaluated the frequency of hyperbilirubinemia. The frequency of newborns with this outcome was significantly lower in the intervention groups (RR: 0.40, 95% CI: 0.23 to 0.68; Fig 5).
The meta-analysis of frequency of hospitalized newborns, measured by Asemi et al. (2015) and Karamali et al. (2015), showed a significant difference between groups, favoring the vitamin D supplementation group (RR: 0.40, 95% CI: 0.23 to 0.69) (Fig 6). Asemi et al. (2015) and Karamali et al. (2015) reported the frequency of macrosomic newborns. There was no difference between the supplemented and control groups (RR: 0.20, 95% CI: 0.04 to 1.13).  (Table 3) The quality of evidence assessment was performed for the outcomes that were thought to be most important from a patients' point of view and were plotted in the meta-analysis. The quality of evidence was very low for the frequency of preeclampsia, prematurity, caesarian deliveries, and maternal hospitalization. For these outcomes, it was necessary to rate them down one level for risk of bias and two levels for imprecision. For newborn hospitalization the quality evidence was low; it was necessary to rate it down one level of evidence for risk of bias and imprecision. As less than ten RCTs were included in this review, we could not analyze the presence of publication bias.

Discussion
During pregnancy, maternal vitamin D deficiency has been associated with adverse maternal and fetal outcomes, such as increased incidence of preeclampsia, insulin resistance, diabetes mellitus, and increased frequency of cesarean delivery [12,13,28,29].
As vitamin D deficiency has been a common and often underdiagnosed health problem worldwide, and considering the benefits of vitamin D supplementation in GDM control [14] and on maternal and neonatal outcomes in normoglycemic pregnant women [30], we performed this systematic review and meta-analysis to evaluate the efficacy of vitamin D supplementation in pregnant women with GDM. Our hypothesis was that this supplementation would protect pregnant women against maternal and neonatal complications related to GDM.
Six trials fulfilled our eligibility criteria and were included in this review [14,[18][19][20][21][22]. Four hundred and fifty-four pregnant women with GDM were randomized to supplementation  Vitamin D supplementation in gestational diabetes mellitus with vitamin D 3 or placebo. The data of five trials could be plotted in the meta-analysis. Regarding maternal risk in GDM, the most important outcomes evaluated were frequency of preeclampsia, hospitalization and cesarean delivery. None of these comparisons showed significant differences between groups; however, due to the small number of events and trials, the 95% CI was very wide, resulting in a very low quality of evidence according to GRADE approach.
The most important neonatal outcomes that could be plotted on meta-analysis were frequency of prematurity, need for hospitalization of newborn. For prematurity, there was no difference between the groups, but with a very wide CI of 95%. The meta-analysis showed that offspring of women with GDM and without vitamin D supplementation were more likely to require hospitalization, presenting a risk difference in favor of the intervention of 30%. However, the quality of evidence was low, resulting in a lack of confidence in the effect estimate and that the true value may be substantially different.
Our systematic review had some limitations, with the main one being related to the small number of trials and patients included. All were single center trials, which tend to provide larger treatment effects than do multicenter RCTs, and hence, they should be carefully used in decision making [31]. Additionally, no trial evaluated adverse events of vitamin D supplementation. Another important limitation is we combined data from two trials testing supplementation with vitamin D alone and two with vitamin D plus calcium. Although there was no heterogeneity in the four outcomes which differences between groups favored vitamin D supplementation (Figs 3 to 6), this may introduce confounding effects since the interaction between the two supplements and any independent effects are not clear.
In the literature, this is the first published systematic review evaluating the effects of vitamin D supplementation focused on pregnant women with GDM. In the Cochrane Database of Systematic Reviews, there is a review published in 2016 that evaluated vitamin D supplementation in normoglycemic pregnant women. [30] Data from two RCTs (219 pregnant women) showed that women with supplementation of vitamin D alone, compared to placebo, had a lower risk of preeclampsia, similar risk of GDM; however, these had low and very low qualities of evidence, respectively. Regarding infant outcomes, supplementation compared to no intervention or placebo (three RCTs with 477 women) suggested that supplementation reduces the risk of preterm births (8.9% versus 15.5%; RR 0.36, 95% CI: 0.14-0.93) and a birthweight below 2500 g (RR 0.40, 95% CI: 0.24-0.67). Despite a moderate quality of evidence for these two last outcomes, the authors concluded that is unclear whether vitamin D supplementation in normoglycemic pregnancy improves maternal and neonatal outcomes. The most recent systematic review including pregnant women with vitamin D supplementation is from Roth et al. [32] published in 2017. From 43 eligible trials (8406 participants), maternal clinical outcomes were rarely reported, and available data did not provide evidence of benefit. Overall, vitamin D increased mean birth weight of 58.33 g (95% CI: 18.88 g to 97.78 g) and reduced the risk of small for gestational age births (RR 0.60, 95% Cl: 0.40 to 0.90), but findings were not robust in sensitivity and subgroup analyses.

Implications for practice
We did not find moderate or high quality evidence from RCTs suggesting that in pregnant women with GDM, vitamin D supplementation, when compared with placebo, improves adverse maternal and neonatal outcomes related to GDM.

Implications for research
Further large, well done, and multicenter RCTs are required to determine the efficacy and safety of vitamin D supplementation in pregnant women with GDM, with the aim to improve the maternal and neonatal adverse outcomes related to GDM.