Impact of Vitamin D Replacement on Markers of Glucose Metabolism and Cardio-Metabolic Risk in Women with Former Gestational Diabetes—A Double-Blind, Randomized Controlled Trial

Gestational Diabetes Mellitus (GDM) and vitamin D deficiency are related to insulin resistance and impaired beta cell function, with heightened risk for future development of diabetes. We evaluated the impact of vitamin D supplementation on markers of glucose metabolism and cardio metabolic risk in Asian women with former GDM and hypovitaminosis D. In this double blind, randomized controlled trial, 26 participants were randomized to receive either daily 4000 IU vitamin D3 or placebo capsules. 75g Oral Glucose Tolerance Test (OGTT) and biochemistry profiles were performed at baseline and 6 month visits. Mathematical models, using serial glucose, insulin and C peptide measurements from OGTT, were employed to calculate insulin sensitivity and beta cell function. Thirty three (76%) women with former GDM screened had vitamin D level of <50 nmol/L at baseline. Supplementation, when compared with placebo, resulted in increased vitamin D level (+51.1 nmol/L vs 0.2 nmol/L, p<0.001) and increased fasting insulin (+20% vs 18%, p = 0.034). The vitamin D group also demonstrated a 30% improvement in disposition index and an absolute 0.2% (2 mmol/mol) reduction in HbA1c. There was no clear change in insulin sensitivity or markers of cardio metabolic risk. This study highlighted high prevalence of vitamin D deficiency among Asian women with former GDM. Six months supplementation with 4000 IU of vitamin D3 safely restored the vitamin D level, improved basal pancreatic beta-cell function and ameliorated the metabolic state. There was no effect on markers of cardio metabolic risk. Further mechanistic studies exploring the role of vitamin D supplementation on glucose homeostasis among different ethnicities may be needed to better inform future recommendations for these women with former GDM at high risk of both hypovitaminosis D and future diabetes.


Introduction
Gestational diabetes mellitus (GDM) is a state of glucose intolerance occurring during pregnancy and is related to both resistance to peripheral action of insulin and impairment of beta-cell function. Its transient presence during pregnancy alerts to a heightened risk of diabetes in the future. About 10-50% of women with GDM develop diabetes mellitus later on in life [1]. Data from Malaysia found that 50% of GDM women had developed diabetes at an interval of five to seven years post index pregnancy [2]. Vitamin D deficiency has been shown to be associated with insulin resistance and impaired pancreatic function [3,4]. Vitamin D deficiency is more prevalent in women with GDM and low vitamin D levels correlate with insulin resistance [5][6][7].
Interventional studies using vitamin D supplement in an attempt to modify glucose metabolism have yielded mixed results. This may be partly due to variable doses of supplementation used, short duration of follow up and inappropriate target group. A very short duration of less than seven days of supplementation may not be sufficient to demonstrate the potential beneficial effects [8][9][10]. Previous studies suggested that vitamin D can help with early stage of disturbance in glucose handling [11], but is unable to augment insulin secretion in subjects with chronic diabetes and exhausted pancreatic function [9]. Lack of adequate dosing may have also accounted for the failure of many previous studies to demonstrate beneficial effects of vitamin D replacement. Adequate vitamin D supplementation would ideally raise blood 25-hydroxyvitamin D (25(OH)D) levels above 80 nmol/L because diabetes risk is lowest at this vitamin D level [12]. Supplementation with 4000 IU of vitamin D3 per day for six months in a population of South Asian women with proven vitamin D deficiency safely restored the vitamin D level and improved insulin resistance [13].
Very little is known about the relationship between vitamin D status and glucose metabolism in women with former GDM. Most of the works on effect of vitamin D supplementation on GDM examined its effect during pregnancy rather than following delivery [14].
Our study aimed to evaluate the effect of adequate vitamin D supplementation on insulin sensitivity, pancreatic beta-cell function and markers of cardio-metabolic risk in Malaysian women with former GDM and vitamin D insufficiency.

Materials and Methods
The protocol for this trial and supporting CONSORT checklist are available as supporting information; see S1 CONSORT Checklist and S1 Protocol.

Study Design and Screening
This was a prospective, randomized, double-blind, placebo-controlled, parallel group trial. Women with documented GDM in the most recent pregnancy, who were between 6 to 48 months post-partum, were identified from the delivery registry of Maternity Unit, Penang General Hospital and antenatal records of community health centers. Potential participants were approached through postal and telephone contacts, pre-screened by telephone interview and invited to attend the screening visit at Clinical Research Centre at Penang Medical College, Penang, Malaysia. The screening and recruitment period was started from 30 th June 2011 through to 3 rd August 2012.
Women were excluded if they had medical conditions that might increase their risk or interfere with study assessments. These included pregnancy, breastfeeding, intolerance to vitamin D supplementation, drug or alcohol dependence, chronic renal or liver failure, hypercalcaemia, hypocalcaemia or concomitant use of calcium supplementation, anti-tuberculosis treatment or anti-epileptic medications. Women who have developed diabetes by the time of assessment were also excluded on the rationale that those with established disease might not respond as well to vitamin D supplementation and also to avoid confounding effect of hypoglycaemic medications on assessment of glucose metabolism.
After providing written informed consent and being interviewed, participants underwent anthropometric examinations (body mass index (BMI) and waist-to-hip circumference ratio (WHR)), blood sampling for 75 g oral glucose tolerance test (75g-OGTT), HbA 1c , fasting lipid profile, plasma 25(OH)D, intact parathyroid hormone (iPTH), renal profile, liver profile, bone profile, highly sensitive CRP (hsCRP) and urinary microalbumin-creatinine ratio (MCR). The 75g-OGTT was performed with serial measurements for serum insulin at time 0, 60, 90 and 120 minutes, C-peptide at time 0, 60 and 120 minutes and glucose at time 0, 30, 60, 90 and 120 minutes. Blood samples were processed on the same day for plasma and serum, and aliquots were stored frozen at -20°C until laboratory analyses were performed. Physical activity was quantified using the Paffenbarger Physical Activity Questionnaire [15].

Randomization and follow up
Women who have plasma 25(OH)D level between 15 and 50 nmol/L without overt diabetes were eligible to proceed to the randomization phase. Overt diabetes was defined as fasting plasma glucose of 7 mmol/L and/or 2-hour OGTT of 11.1 mmol/L [16]. Women with plasma 25(OH)D of less than 15 nmol/L or those with overt diabetes were excluded from the study and referred to receive active treatment. Eligible women were then randomized to receive either 4000 IU of vitamin D3 (cholecalciferol) per day (four capsules of 1000 IU each) or four capsules of matching placebo per day for six months. Both the active (vitamin D3) and placebo capsules contained 300 mg of soy oil. Offsite randomization was generated by the manufacturer of the vitamin D and placebo capsules (Blackmore Ltd) using nQuery Advisor. Randomization and allocation were fully concealed from the investigators and subjects until after completion of the study. All participants attended every second month for review and collecting new supply. Participants were advised to contact research staff immediately if they suspected a reaction to the supplements or in the event of pregnancy. Serum calcium and albumin were taken at two months to rule out hypervitaminosis. The assessments performed at the screening visit were repeated after six months of supplementation. The study was completed on 1 st February 2013 with last visit for the last participant. All the data were hosted at Penang Medical College. The data were de-identified before their use in analyses.

Insulin sensitivity and pancreatic beta-cell function
Fasting insulin sensitivity was estimated with the Quantitative Insulin Sensitivity Check Index (QUICKI), while the peripheral insulin sensitivity with the Oral Glucose Insulin Sensitivity Index (OGIS) that represents glucose clearance [17,18]. These indices have been validated toward the glucose clamp and tracer-calculated hepatic sensitivity [19].
Fasting insulin secretion was evaluated from basal insulin and C-peptide concentrations. Total (120 minutes) insulin secretion was evaluated with the area under the curve of C-peptide (AUC cp ), while the amount of insulin reaching the periphery (post-hepatic) with AUC of insulin (AUC insulin ). Beta-cell function was assessed with the insulinogenic index calculated at 60 minutes: IGI 60 = (I 60 -I basal )/ (G 60 -G basal ), and with the ratio of the incremental AUC cp (ΔAUC cp ) over the incremental AUC glucose (ΔAUC glucose ), as a global insulinogenic index [20].
A simultaneous measurement of pancreatic beta-cell function, insulin sensitivity and glucose tolerance test (BIGTT) was employed to provide an additional estimate of insulin sensitivity index (BIGTT-S) and acute insulin response (BIGTT-AIR) using serial data from OGTT; the BIGTT indices were shown to highly correlate with intravenous glucose tolerance test (IVGTT)-derived values [21]. Under normal physiology, the tight glucose regulation is maintained by a balance between insulin secretion and insulin sensitivity [22]. To describe this process, here we used the product of post-hepatic beta-cell function (ratio of AUC insulin over AUC glucose ) and insulin sensitivity (OGIS). This product, in other circumstances called disposition index (DI), interprets glucose tolerance of an individual by relating beta-cell function with whole-body insulin sensitivity [22].

Sample size and statistical analyses
By using nQuery Advisor version 6.01, based on a trial by von Hurst et al [13], a sample size of 9 will have 95% power to detect a difference in mean for Homeostatic Model Assessment of Insulin Resistance (HOMA%IR) of 0.700 (e.g. the differences between Group 1 mean, μ1, of 0.800 and a Group 2 mean, μ2, of 0.100), assuming that the common standard deviation is 0.384, with a 0.05 two-sided significance level. By allowing a 40% drop-out rate of subjects, the total sample for two arms (placebo and treatment groups) was calculated to be 26 subjects (13 patients for each arm).
All data in this study were analysed according to intention-to-treat protocol. Missing data for subjects who defaulted and withdrew from the study were analysed using listwise deletion technique. All clinical and biochemical continuous variables, except for post-partum duration, were not normally distributed. The non-normally distributed continuous variables were reported in median (25 th , 75 th percentiles) while the only normally distributed continuous variable, post-partum duration, was reported in mean ± standard deviation. Comparison between interventional and placebo groups for baseline socio-demography, clinical histories and characteristics, markers for glucose metabolism, insulin sensitivity, pancreatic beta-cell function and cardio-metabolic risk markers were performed using Mann-Whitney U tests for non-normally distributed continuous variables. Wilcoxon signed rank tests were employed to detect differences in clinical and biochemical markers pre-and post-intervention in each group, while Mann-Whitney U tests were performed to compare these markers across both groups. The two-sided statistical significance level, p-value, was set at 0.05 for all analyses in this study.

Ethical approval
The study was registered at Malaysian National Medical Research Register (NMRR) (reference number NMRR-10-679-6202) on 5 th June 2010 and ClinicalTrials.gov (NCT01992133) on 12 th November 2013. At time of publication, the authors became aware of NMRR is yet to be recognized by the WHO International Clinical Trials Registry Platform, hence the delay in registration of this study with ClinicalTrials.gov. The Medical Research and Ethics Committee (MREC), Ministry of Health Malaysia approved the study protocol by 18 th October 2010. Written informed consents were obtained from all participants. The authors confirmed that all ongoing and related trials for intervention are registered.

Randomization and adherence to vitamin D supplementation
Seventeen women screened were excluded from the randomization phase. Four women dropped out of study (one became pregnant, one developed allergic reaction to bovine capsule and two did not return for final visit due to busy family commitment). Eleven women in each group completed the study (Fig 1). The vitamin D and placebo groups did not differ at baseline with respect to socio-demographics, physical characteristics and laboratory measures, except for serum iPTH and serum total cholesterol ( Table 1). The average vitamin D capsule adherence over the 6 months follow-up period, reported by pill count, was 88.2% in the vitamin D group and 86.7% in the placebo group. The median serum 25(OH)D concentrations increased significantly in the vitamin D-supplemented group following 6 months of supplementation while the levels remained unchanged in the placebo group (Fig 2). There was no evidence of hypervitaminosis, and serum calcium after 2 months and 6 months of supplements remained normal for all subjects. One participant developed allergic rash to the bovine capsule which resolved fully on stopping the supplement. No other adverse events were reported.

Effect of vitamin D supplementation on glucose metabolism
The serial changes in median serum glucose and insulin levels during OGTT are shown in Figs 3 and 4. The median glucose in the vitamin D group decreased significantly at 2-hour but the changes at the other time points were not significant. There was no difference in the total AUCglucose within and between groups following supplementation. The HbA 1c dropped by 2 mmol/  Table 2).
Three women in the vitamin D group had IGT at baseline. At the end of the 6 months study, 1 reverted to normal while 2 others remained at the IGT level. In the placebo group, 1 woman with IGT normalized, 1 with IFG progressed to diabetes and 2 women who started with normal glucose response progressed to IGT.

Effect of vitamin D supplementation pancreatic beta-cell function and insulin sensitivity
The basal pancreatic beta-cell function, as measured by fasting insulin increased by 20% in the vitamin D group and decreased by 18% in the placebo group (p = 0.034). Fasting C-peptide increased significantly in both arms (77.8% in vitamin D and 51% in placebo). This increased basal insulin secretion in the absence of changes in basal glucose implied improved basal pancreatic beta-cell function following supplementation of vitamin D. The total insulin secretion over 2 hours following the glucose challenge, as measured by AUC insulin (total pancreatic insulin secretion) and AUC cp (total insulin reaching the peripheral circulation) did not differ between groups.
The dynamic pancreatic response following glucose challenge did not show consistent change with vitamin D supplementation ( Table 2). The BIGTT-AIR increased significantly by 52% in the vitamin D group. The IGI 60 and global insulinogenic index increased but was not affected by vitamin D supplementation. Vitamin D had no obvious effect on insulin sensitivity. No within-group changes were observed for QUICKI, OGIS and BIGTT-S. A borderline between-groups difference (p = 0.047) was noted for QUICKI but not for OGIS and BIGTT-S. The disposition index increased significantly in the vitamin D group ( Table 2).

Effect of vitamin D supplementation on markers of cardio-metabolic risk
Overall, vitamin D supplementation did not alter the markers of cardio-metabolic risk ( Table 3). There were no significant between-group differences in changes in blood pressure, BMI, WHR, hsCRP, total cholesterol, LDL-cholesterol and HDL-cholesterol. The placebo group experienced a within-group decrease in BMI, systolic and diastolic blood pressures and LDL-cholesterol, while the vitamin D group saw a rise in triglyceride and triglyceride to HDLcholesterol ratio after 6 months. These within-group differences may be due to chance findings given the small sample size.

Discussion
This study highlights a high prevalence of vitamin D insufficiency among young women of Malaysian origin with former GDM and their high risk of progression to diabetes. Supplementation with vitamin D at 4000 IU daily for six months resulted in improved 25(OH)D levels and improved basal pancreatic beta-cell function when compared with placebo. There appeared to be no similar effect on the dynamic pancreatic function. While the improved BIGTT-AIR suggests amelioration, this was not supported by the other indices of dynamic pancreatic betacell function including IGI 60 and the global insulinogenic index.
In our study, vitamin D replacement did not result in significant change in insulin sensitivity. Previous randomized controlled studies showed that vitamin D, when given in similar dose and duration, improved hepatic insulin sensitivity in insulin-resistant South Asian women [13] and obese adolescents [23]. A single injection of 300,000 IU vitamin D3 at 3-10 days after delivery reduces indices of insulin resistance in mothers with recent pregnancy complicated by GDM [24]. On the other hand, a study in obese African Americans [25], supplementation of vitamin D worsened hepatic insulin sensitivity. In observational studies, higher serum 25(OH) D and lower HOMA%IR were observed in Caucasian Americans but not in African Americans [4,26]. These seeming conflicting results suggest that the effect of vitamin D on insulin sensitivity may be different in different ethnic groups. In our study, vitamin D replacement did not result in significant change in markers of peripheral insulin sensitivity, including OGIS and BIGTT-S. While the between-group difference in QUICKI may imply that vitamin D supplementation decreased hepatic insulin sensitivity, the marginal p-value could be equally likely due to the scattering of data in a small sample size. Our study included Malaysian women of three different ethnic groups and our sample size was inadequate to differentiate effect among the different ethnic groups; however, the similar baseline characteristics seem to exclude an effect of ethnicity within Malaysian women.
Although OGIS and ratio of AUC insulin over AUC glucose did not improve significantly, their combination (the so-called disposition index) increased significantly in the vitamin D group, suggesting indeed an amelioration of the compensatory mechanism of insulin resistance by insulin release [22]. This is supported by an improvement in the glucose tolerance and a significant drop in HbA 1c of 2 mmol/mol (0.2%) in the vitamin D group, but not in the placebo arm.
The overall trend suggests that vitamin D may have beneficial effect on glucose metabolism in women with previous gestational diabetes and hypovitaminosis D, by predominantly improving the basal pancreatic beta-cell function. This observation will need to be verified in a larger trial of longer duration. Contrary to observational data, supplementation of vitamin D failed to result in significant changes in markers of cardio-metabolic risk of blood pressure, lipid profile, hsCRP and MCR in this interventional study.
The high prevalence of vitamin D insufficiency despite the tropical climate and lack of seasonal variation is alarming. Inadequate intake from diets low in vitamin D, practice of deliberate sun avoidance and foetal utilization through previous pregnancy and lactation are all possible contributing factors. In addition, darker skin pigmentation may be associated with decreased vitamin D conversion [27]. This is important in our globalized world, because people of tropical origin going-for instance-to Northern Europe will lack sun light even more, potentially worsening their hypovitaminosis. On the other hand, Powe et al recently demonstrated that racial difference in common genetic polymorphism between black and white Americans contributes to variation in the levels of vitamin D-binding protein and bioavailability of 25 (OH)D [28]. The vitamin D-binding protein was not incorporated into the assessment of vitamin D status in our multi-ethnic study population.
The strength of this study lied in the randomized placebo-controlled prospective design, in a well-defined group of individuals at high risk of progression to diabetes, the use of high dose of vitamin D over substantial duration, conduct of 2-hour OGTT which enabled use of various

Conclusions
This study has demonstrated that 6 months supplementation of vitamin D in Asian women with previous gestational diabetes and hypovitaminosis D resulted in improved basal pancreatic beta-cell function, unchanged dynamic pancreatic beta-cell function and overall amelioration of the metabolic state as shown by an improved DI and HbA 1c . There was no    Supporting Information S1 CONSORT Checklist. CONSORT Checklist.
(DOCX) S1 Study Dataset. Study data set and variable explanatory information.
(RAR) Research for permission to publish this paper. Thanks are also due to Dr Kristine Faerch from Steno Diabetes Centre, Gentofte for her critical review of this manuscript. This study has been presented as a poster at the 9 th International Diabetes Federation Western Pacific Region Congress, 4 th Scientific Meeting of the Asian Association for the Study of Diabetes at Kyoto, Japan in November 2012, and at the National Conference for Clinical Research (NCCR) in Kuala Lumpur, Malaysia in September 2013. It has also been presented as oral presentation at the Diabetes Asia Conference in Kuala Lumpur in October 2013.