Discussion Questions

• ■
  

  What was the primary aim of the study?

  
  
• ■
  

  What was the study design?

  
  
• ■
  

  What information is in the tables?

  
  
• ■
  

  What were the study’s strengths and weaknesses?

  
  
• ■
  

  How does this research help us clinically?

  
  
• ■
  

  What are the implications for future research?

Introduction

For many women, gestational diabetes mellitus (GDM) is a precursor to enduring illness. Some 5-10% of women with GDM will be diagnosed with diabetes mellitus, mainly type 2, directly after giving birth. But even those who escape an immediate postnatal diagnosis have a 35-60% likelihood of developing diabetes in the 10-20 years following pregnancy, and thus, are advised to undergo regular monitoring. Several well-known risk factors exist, such as a family history of diabetes or a personal history of GDM or impaired glucose tolerance. A new study examined a more subtle connection; a possible link between maternal vitamin D status and GDM.

George A. Macones, MD, MSCE, Associate Editor

Introduction

For many women, gestational diabetes mellitus (GDM) is a precursor to enduring illness. Some 5-10% of women with GDM will be diagnosed with diabetes mellitus, mainly type 2, directly after giving birth. But even those who escape an immediate postnatal diagnosis have a 35-60% likelihood of developing diabetes in the 10-20 years following pregnancy, and thus, are advised to undergo regular monitoring. Several well-known risk factors exist, such as a family history of diabetes or a personal history of GDM or impaired glucose tolerance. A new study examined a more subtle connection; a possible link between maternal vitamin D status and GDM.

George A. Macones, MD, MSCE, Associate Editor

Background

Macones: What is the importance of this question? What is the plausibility of such an association?

McCartney: The incidence of GDM is approximately 14% of total pregnancies in the United States, and that rate continues to increase despite medical attention. GDM is associated with a variety of complications, including fetal macrosomia, preterm birth, newborn jaundice, and neonatal respiratory distress syndrome; mothers are at increased risk for cesarean section and future type II diabetes. If maternal vitamin D status turns out to be a modifiable risk factor in GDM, then vitamin D supplementation would be a potential low-cost preventative and therapeutic agent.

Recently, there has been notable interest in both the scientific and lay press concerning the role of vitamin D in an assortment of disease processes. In terms of women’s health, vitamin D deficiency has been linked to infertility, preterm birth, polycystic ovarian syndrome, preeclampsia, bacterial vaginosis, as well as GDM in various clinical and experimental models; higher levels have been associated with endometriosis. In addition, some reports have shown an association between maternal vitamin D levels and type I diabetes risk in offspring.

Low serum vitamin D has been found to be associated with type II diabetes and obesity in a number of clinical studies. Mechanistically, in vitro and animal models have shown that vitamin D deficiency affects insulin secretion and the expression of insulin receptors. Vitamin D is also known to have anti-inflammatory properties, and mounting evidence demonstrates that inflammation is intimately linked to the development of the metabolic syndrome and diabetes. The association between GDM and maternal vitamin D levels has been previously reported, but the data are currently confusing due to variety in study size, timing of maternal vitamin D sampling, and adjustment for potential confounding factors.

Macones: What was the study’s primary aim?

McCartney: The primary aim of the study was to determine whether there was an association between maternal 25-hydroxyvitamin D [25(OH)D] levels in the second trimester and gestational diabetes.

Study Design

Macones: What study design was used?

Hopeman: This is a case-control study “nested” within a prospective cohort study; specifically, Project Viva, a large epidemiologic prenatal prospective cohort study that is looking for links between gestational factors and offspring health. Nested case-control studies are useful when attempting to identify and analyze predictor variables associated with outcomes. In this case, the predictor or exposure is the mother’s vitamin D level at 26-28 weeks. The outcomes are GDM or impaired glucose tolerance identified at the same gestational age.

Macones: Can you briefly describe the methods, including primary outcomes? Do you have concerns about the laboratory assays used?

Hopeman: The Project Viva cohort included 2128 participants who gave birth. Women were eligible if they were fluent in English, had a singleton pregnancy, and attended their initial obstetric visit at less than 22 weeks’ gestation. From this larger cohort, participants were included in the case-control study if their second-trimester vitamin D levels were able to be measured from stored samples and if GDM status was recorded in the medical record.

The primary outcomes were development of GDM or impaired glucose tolerance, based on the results of glucose challenge testing. Each patient underwent a nonfasting, 1-hour, 50 g glucose challenge test. A result of less than 140 mg/dL was considered normal; greater than or equal to 140 mg/dL was impaired, and the participant went on to have a fasting 3-hour 100 g glucose tolerance test. If 0-1 values were abnormal, participants were classified as glucose–impaired. If 2 or more values were abnormal, the patient was classified as having gestational diabetics.

The exposure, the patient’s 25(OH)D level, was measured from maternal blood samples that had been drawn between 26-28 weeks’ gestation. Two different laboratory assays were used: an automated chemiluminescence immunoassay (CLIA) and a manual radioimmunoassay (RIA). The 2 values were then averaged because the results from the different assays were not identical.

In general, assays for measurement of the prohormone 25(OH)D are not standardized and are known to be highly variable. This makes it difficult to generalize the results of this study. Cutoff values for what constitutes low or very low vitamin D levels may be different, depending on the assay. The manual RIA is known to be more sensitive than other tests and is a well-established reference method, but it is inconvenient and therefore unlikely to be used clinically. The automated CLIA would be more convenient.

Macones: Can you describe the analytic approach?

Norman: This study used bivariate analysis to evaluate the relationship between maternal characteristics and 25(OH)D levels. Then, multivariable-adjusted multinomial logistic regression analysis was used to adjust for confounding factors. Using 25(OH)D levels of <25 nmol/L and ≥25 nmol/L in a dichotomized fashion, the odds ratios and 95% confidence intervals were calculated for gestational diabetes and impaired glucose tolerance. In addition, similar analyses were performed in order to evaluate for a relationship between vitamin D status and 1-hour glucose tolerance test results. As a result of having missing covariates in the dataset for some subjects, imputed values were generated using all 2128 subjects, but only the data from those with observed 25(OH)D were used in the analysis.

Macones: What information is contained in Table 1?

McCartney: Table 1 describes the characteristics of the study population, including the maternal second trimester 25(OH)D levels; glucose tolerance status; and a variety of demographic and lifestyle factors, including maternal age, body mass index (BMI), pregnancy weight gain, physical activity, vitamin supplementation, education, marital status, smoking status, household income, race/ethnicity, and parity. Levels of vitamin D were divided into severe deficiency (<25 nmol/L), deficiency (25-<50 nmol/L), insufficiency (50-<75 nmol/L), and sufficiency (≥75 nmol/L) groups. When compared with women who had the highest levels of 25(OH)D, those with the lowest levels were younger (30 years vs 33 years), had increased BMI (28.6 vs 23.7), increased blood glucose after the 1-hour glucose tolerance test (122 mg/dL vs 111 mg/dL), decreased vitamin D intake (373 IU/day vs 588 IU/day), decreased calcium intake (1142 mg/day vs 1455 mg/day), decreased rate of college graduation (34% vs 78%), increased rate of single status (18% vs 4%), increased rate of physical activity (9.3 hours/week vs 7.2 hours/week), increased rate of never having smoked (76% vs 68%), increased rate of smoking during pregnancy (14% vs 9%), increased rate of household income less than $70,000 (78% vs 31%), increased rate of parity >1 (71% vs 52%), decreased rate of white race (22% vs 85%), increased rate of GDM (15% vs 4%), and increased rate of impaired glucose tolerance (15% vs 9%).

Macones: What information is contained in Tables 2 and 3? Why does Table 2 contain 4 models, and what do the different models tell us?

Norman: Table 2 reflects the odds of gestational diabetes and impaired glucose tolerance for subjects with 25(OH)D levels <25 nmol/L. Model 1 shows the unadjusted odds ratios and confidence intervals. Model 2 adjusts for gestational age, season, maternal age, race/ethnicity, education, marital status, smoking, and parity. Model 3 adjusts for everything in model 2, and also adjusts for prepregnancy BMI and pregnancy weight gain. Model 4 goes further and adjusts for everything in model 3 with the addition of physical activity and dietary intake of fish and calcium. Multiple models are used to illustrate the contributions of different confounding factors, as well as to delineate the difference in the odds after adjusting for factors unique to this study. These models help reveal the impact of each successive adjustment.

In short, the researchers found that 68 women met the criteria for GDM. Compared with women who had 25(OH)D levels of 25 nmol/L or more, those with levels less than 25 nmol/L had “significantly increased odds of GDM.” This remained true in the second model after adjusting for several factors. However, Burris and colleagues noted that in the third model, adjustment for BMI weakened the association, and the confidence interval included the null value. This finding was believed to be the result of confounding. Additional adjustments for diet and exercise in Model 4 had minimal effect, leading the group to propose that vitamin D’s interaction with glucose tolerance is independent from dietary sources of the nutrient.

Table 3 shows the relationship between vitamin D status and 1-hour glucose tolerance test results in 3 different ways. Initially, the table uses 25(OH)D levels in a continuous fashion, showing the increase in blood glucose levels for each 25 nmol/L decrease in 25(OH)D levels. Then, the difference in blood glucose results are illustrated for each of the 4 categories of 25(OH)D levels; <25, 25-<50, 50-<75, ≥75. Finally, the difference in blood glucose results is shown for dichotomized 25(OH)D levels, <25 vs ≥25. As in Table 2, 4 models are used for each analysis to adjust for various confounding factors.