Vitamin D Supplementation, Glycemic Control, and Insulin Resistance in Prediabetics

A Meta-Analysis

Naghmeh Mirhosseini; Hassanali Vatanparast; Mohsen Mazidi; Samantha M. Kimball


J Endo Soc. 2018;2(7):687-709. 

In This Article

Materials and Methods

Literature Search Strategy

We followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Guidelines.[44] The study protocol was registered with the International Prospective Register of Systematic Reviews, PROSPERO (registration no. CRD42017055326). The primary outcome of interest was to systematically review the effect of vitamin D supplementation on glycemic control and insulin resistance as measured by HbA1c, FPG, fasting HOMA-IR, and 2HPG in adult populations with prediabetes, overweight, or obesity. We also evaluated the impact of coadministration of calcium with vitamin D, obesity, serum 25(OH)D status at the beginning and the end of intervention, and the duration of vitamin D supplementation on the abovementioned glycemic measures.

Search terms included [vitamin D; vitamin D3; cholecalciferol; 25(OH)D] AND/OR [prediabetes; insulin resistance; at risk for diabetes; hyperglycemia; HbA1c; glucose] in the title and/or abstract. Studies published in English and those conducted on adults (age ≥18 years old) were selected. Calcium was permitted as a combined supplement with vitamin D or as a supplement provided to both treated and control groups.

Data Sources

We searched multiple databases including PUBMED/Medline (Medical Literature Analyses and Retrieval System Online), Cumulative Index to Nursing and Allied Health Literature (CINAHL), the Cochrane library, and Google Scholar (as the gray literature suggested by the Cochrane guideline). The reference lists of all articles that met the selection criteria and systematic reviews already published were also hand searched by two reviewers (N.M. and M.M.). Databases were searched for studies published from January 1999 until April 2017.

Study Selection

Two reviewers (N.M. and M.M.) selected the studies for the systematic review, which were approved by a third reviewer (S.M.K.). Studies were included in this systematic review and meta-analysis of RCTs if: (1) they were placebo-controlled trials; (2) the population included persons with prediabetes, impaired fasting glucose, impaired glucose tolerance, overweight, or obese; (3) were conducted in adults (≥18 years); (4) intervention was a minimum of 2 months (defined as the minimum time required to assess changes in glucose blood markers);[45] (5) supplementation was provided on a daily, biweekly, or weekly basis; and (6) reported baseline and end-of-study measures for serum 25(OH)D concentrations and HbA1c or FPG or HOMA-IR or 2HPG, or providing delta values (changes over time).

Studies were excluded if: (1) they were not placebo controlled; (2) the population included subjects with type 2 diabetes, end-stage renal disease, or gestational diabetes or were conducted in children/adolescents; (3) the duration of the trial was less than 2 months; (4) supplementation was provided less frequently than each week (e.g., monthly basis or as a single large bolus dose); (5) serum 25(OH)D concentrations were not reported at baseline and end of study; (6) they were studies in which the mean change or SD of the outcome measures was not reported; or (7) the study had an observational, case control, cross-sectional, or cohort design, or the publication was a narrative review, comment, opinion piece, methodological report, editorial, letter, or conference abstract.

Data Extraction and Management

The full-text of each publication that met the inclusion criteria was screened to determine eligibility (N.M. and M.M.). Any disagreement between the researchers regarding the eligibility of particular studies was resolved through discussion with a third reviewer (S.M.K. or H.V.). Following assessment of methodological quality of the trials by the first reviewer (N.M.), data were extracted using a data extraction form and the most important results from each study were summarized by N.M. and M.M. The extracted data were approved by the third and fourth researchers (H.V. and S.M.K.). Data extracted from each study included first author, reference, year of publication, country of study, study design, inclusion criteria, sample size, form of vitamin D, vitamin D dose and supplementation interval, control group, duration of supplementation, participants' characteristics [sex (n, % male), age, weight, body mass index (BMI)], cosupplementation with calcium, comorbidities, baseline and follow-up serum 25(OH)D concentration (nmol/L), and outcome measures. Any necessary calculations on data were conducted by the first reviewer (N.M.) and checked by the second reviewer (H.V.).

Quality Assessment

Study quality and the risk of bias in the eligible RCTs were systematically assessed using the Cochrane criteria checklist.[46] The items used for the assessment of each study were: (1) adequacy of random sequence generation; (2) allocation concealment; (3) sample size and power of study; (4) quality of blinding (both participants and personnel); (5) intention-to-treat analysis (incomplete data); (6) compatibility of groups; (7) clear definition of inclusion and exclusion criteria; and (8) the description of withdrawal and dropout. A judgment of "adequate" (√) indicated low risk of bias, whereas "inadequate" (–) indicated a high risk of bias, taking into account the recommendations of the Cochrane Handbook. We labeled uncertain or unknown risk of bias as "unclear." Quality assessment was performed by one reviewer (N.M.) and approved by the other reviewers (S.M.K. and H.V.).

Statistical Analysis and Data Synthesis

To calculate the effect size, we followed the Cochrane Handbook and used the mean change from baseline to the end of the intervention in the levels and SD of the outcome measures for both control and intervention groups.[46] A meta-analysis was conducted to combine the individual study results. Meta-analysis of RCTs was conducted using Comprehensive Meta-Analysis version 3 software (Biostat 2014, Englewood, NJ).[47] A P value <0.05 was considered to be statistically significant. We based the meta-analysis on calculating net changes from baseline to the endpoint, when the mean and SDs of the changes were reported, as: [(measure at endpoint in the treatment group – measure at baseline in the treatment group) – (measure at endpoint in the control group – measure at baseline in the control group)]. Effect sizes were expressed as the between-group weighted (standardized) mean difference and 95% CI.

Serum 25(OH)D levels were collated in nmol/L, and we used a multiplication factor of 2.496 to convert 25(OH)D levels respectively from ng/mL to nmol/L.[48] Plasma glucose levels (FPG and 2HPG) were collated in mmol/L, and a multiplication factor of 0.0555 was used to convert glucose levels respectively from mg/dL to mmol/L as appropriate.[49]

Data were analyzed using a random-effects model (DerSimonian-Laird method) and the generic inverse variance method to compensate for the heterogeneity of studies due to the broad demographic characteristics of populations being studied.[47,49,50] Heterogeneity was assessed using the I 2 index with values ≥50% determining the use of the random-effects model. The effect size was determined using weighted mean difference with a 95% CI. For treatment effect, a negative value represents a reduction in the outcome in the intervention group relative to the change in the placebo group. When more than one dose of vitamin D was examined in the same study, data from the highest dose was compared with placebo. Studies with more than two intervention groups (e.g., vitamin D alone and in combination with calcium) were used in subgroup analysis as multiple studies and both compared with the placebo.

We conducted a sensitivity analysis using the leave-one-out method (removing one study each time and repeating the analysis). This analysis allowed us to determine the impact of each study on the overall effect size.[51]

Publication Bias

We explored publication bias using a visual inspection of Begg's funnel plot asymmetry, supplemented with Egger's weighted regression tests.[49,52] Funnel plots were derived separately from changes in FPG, HbA1c, HOMA-IR, 2HPG, and 25(OH)D depicted as weighted mean difference vs its standard error (SE). This step was followed by adjusting the analysis for the effect of publication bias using the Duval and Tweedie "trim and fill" method.[53]

Subgroup Analysis

In addition to running a sensitivity analysis and using random-effect models, we addressed heterogeneity using subgroup analyses. Subgroup analyses were performed for the following: (1) population was comprised of individuals with prediabetes vs overweight or obese (prediabetes was diagnosed by having HbA1c measured in the range of 5.8% to 6.4%, and any participant with BMI ≥25 kg/m2 was considered overweight/obese); (2) whether calcium was used in combination with vitamin D; (3) age group (≤45 years vs >45 years); (4) vitamin D deficiency at baseline [serum 25(OH)D level <50 nmol/L vs ≥50 nmol/L]; (5) serum 25(OH)D level at follow-up; (6) body weight status (overweight and obese with BMI ≥25 kg/m2 vs BMI <25 kg/m2); and (7) duration of the intervention (<6 months vs ≥6 months).

For each outcome, the effect size for subgroups (two subgroups) was calculated as the weighted mean difference between treatment and placebo groups using Comprehensive Meta-Analysis software (Biostat 2014). Then a between-subgroup comparison for each outcome was conducted with a simple t test, and the P values were adjusted by Bonferroni correction. Subgroup analyses were determined a priori according to established cutoff points such as vitamin D deficiency, defined as a serum 25(OH)D level of less than 50 nmol/L, or based on the distribution of study populations such as median for age or serum 25(OH)D levels at follow-up.