The use of glyburide as an alternative treatment of GDM has been a more common practice since the publication of a randomized-controlled trial in which 404 women with GDM were assigned to receive either subcutaneous insulin or oral glyburide [11]. Since the study’s primary outcome was the achievement of ‘desired level of glycemic control,’ it was not statistically powered to examine the less common neonatal complications that can be associated with hyperglycemia [11]. Although the authors did not observe a statistically significant difference in perinatal outcome between the comparison groups, there was a trend toward higher frequency of neonates with birthweight >4000 g and hyperbilirubinemia in the glyburide group [11].

Following this study, many clinicians embraced glyburide as an alternative to insulin in the treatment of GDM. In a study which compared women with GDM who received glyburide to historical controls treated with insulin, the authors report no difference in birthweight, macrosomia, or cesarean delivery but did observe a statistically significantly higher incidence of preeclampsia, neonatal hyperbilirubinemia, and longer stay in the intensive care nursery with glyburide treatment compared to insulin [12]. Furthermore, another study that examined the efficacy of insulin, glyburide, and acarbose for the treatment of GDM found glyburide to be more effective than acarbose in achieving glycemic control but showed a higher rate of macrosomia and neonatal hypoglycemia compared to insulin [13].

For a large proportion of women (51%) who received oral hypoglycemic agent, treatment was coded as ‘oral hypoglycemic agent.’ However, for women in whom the specific oral hypoglycemic agent was known, 99% received glyburide, while only 1% received metformin or other oral agents. Since most clinicians who used oral hypoglycemic agents chose glyburide during the study period [15], we designated those women who received ‘oral hypoglycemic agents’ as having received glyburide. Women were categorized as receiving glyburide or insulin based on the treatment modality at the last Sweet Success visit, unless they were placed on treatment at the first Sweet Success visit, in which case they were categorized by their treatment modality at the first visit. For example, women who were initially diet-controlled but subsequently placed on oral agents were designated as the glyburide group. Similarly, women who were initially diet-controlled but then placed on insulin by their last Sweet Success visit or those who were started on glyburide but were switched to and maintained on insulin by their last Sweet Success visit were assigned as the insulin group.

We explored maternal characteristics likely to be associated with glyburide use, including maternal age, race/ethnicity, parity, prepregnancy body mass index (BMI), education level, primary language, GA at which GDM was diagnosed, and GA at delivery. Maternal outcome examined included primary cesarean delivery. Neonatal outcomes examined included preterm delivery (PTD) <37weeks and <34 weeks, birthweight >4000 g and >4500 g, large-for-gestational-age (LGA, defined as birthweight >90th centile by GA according to the neonatal weight curves established by Babson et al. [16] and Williams et al. [17]), intrauterine fetal demise (IUFD), and admission to the neonatal intensive care nursery (NICU).

Maternal and neonatal outcomes associated with GDM treatment modality were examined using chi-square test and multivariable logistic regression analyses (Table II). Designating women who were treated with subcutaneous insulin as the referent, women who took glyburide had lower odds of overall cesarean delivery (aOR 0.77, 95% CI [0.65–0.91]) but no difference in primary cesarean delivery or PTD (Table II). On the contrary, women who received glyburide were more likely to have neonates with birthweight >4000 g (aOR = 1.29 [1.03–1.64]) and higher odds of admission to the NICU (aOR = 1.46 [1.07–2.00]; Table II) among term deliveries.

To further explore differences in GDM treatment modalities, GA of GDM diagnosis was stratified by diagnosis prior to or after 24 weeks. In women who were diagnosed with GDM prior to 24 weeks (n = 2248), the median GA at diagnosis was 17.0 weeks (IQR: 13.3–20.7 weeks). Among these women who were diagnosed early in gestation, those who received glyburide had a higher odds of birthweight >4000 g (aOR = 1.57 [1.01–2.45]) or LGA > 90th centile (aOR = 1.65 [1.10–2.48]), and IUFD (aOR = 4.68 [1.02–21.5]; Table III), as compared to those receiving insulin. In women diagnosed with GDM after 24 weeks (n = 6556), the median GA at diagnosis was 28.7 weeks (IQR: 27.0–30.6 weeks). Among this subgroup, women who received glyburide were more likely to have PTD < 34 weeks (aOR = 1.75 [1.02–3.03]) and NICU admissions, even when delivered at term (aOR = 1.50 [1.01–2.23]; Table III).

To examine whether educational level would influence the allocation of GDM treatment options, the cohort was stratified by years of education into two groups: <9 years (less than high school education), and ≥9 years (high school education and above). In women who did not complete high school education, treatment with glyburide, compared to insulin injections, was associated with a lower risk of cesarean delivery but higher risks of PTD and NICU admissions in term deliveries (Table IV). In women who attended high school or beyond (≥9 years of education), the risk of cesarean delivery was lower in the glyburide group but the risk of having an IUFD was 3-fold higher with glyburide treatment compared to insulin (aOR = 3.33 [1.14–9.74]; Table IV).

Although we observed a modest increase in the risk of macrosomia and NICU admissions in women treated with glyburide, we did not have information regarding the occurrence of neonatal hypoglycemia. Besides the short-term implications of macrosomia, it has been reported that children born to women with GDM who were LGA at birth are at significantly increased risk of childhood metabolic syndrome with an onset as early as 6 years of age [19]. With respect to clinical utility, we attempt to estimate the number needed to treat from the adjusted odds ratios of our regression model for macrosomia. According to this analysis, the baseline frequency of macrosomia was 13%; if women treated with oral agents have 29% greater odds of having infants with macrosomias compared to insulin, the number needed to treat in order to avoid one case of macrosomia is 30. While this estimate may not be generalizable across all population, it is a relatively small number that may potentially have large impact on the long-term health in the children of diabetic mothers.

The American Diabetes Association (ADA) has cautioned that the use of glyburide for the treatment of GDM awaits larger studies to establish its safety [6], and the most recent ACOG practice guidelines state that subcutaneous insulin remains as the standard of care in women with GDM requiring medical therapy [1]. Yet, the utilization of oral hypoglycemic agents, particularly glyburide, has become an increasingly common practice [20]. Nearly one in five women with GDM was treated with glyburide in our cohort. Maternal characteristics associated with higher use of oral hypoglycemic agents included multiparity, Asian race/ethnicity, normal or low BMI, GA at GDM diagnosis between 20 and 32 weeks, less than high school education, and non-English as primary language. Further, in this cohort, we observed that a high proportion (39%) of women who were started on glyburide for treatment of GDM on initial visit was eventually placed on insulin therapy. While the precise reason for this crossover is uncertain, it is likely due to either intolerance of side effect or suboptimal glycemic control as previously reported by other studies [11,12]. Thus, it appeared that for the cohort of high-risk women who likely have overt hyperglycemia requiring medical therapy at time of GDM diagnosis, there is a high likelihood of need for insulin therapy.

Likely, treatment allocation is not random and clinicians may be more inclined to prescribe oral hypoglycemic agents to women considered at lower risk of insulin resistance given the reported ‘failure’ rate of glyburide of around 20% [11,15]. As our study was not a randomized trial, this difference in treatment allocation made the use of multivariable regression paramount in examining the outcomes associated with glyburide as compared to insulin.

In women who were diagnosed with GDM early in gestation (<24 weeks), glyburide use, compared to insulin, was associated with higher risk of macrosomia/LGA and IUFD. In this high-risk population, we speculate that suboptimal glycemic control and adverse outcomes are more likely with glyburide use.

We observed that the two groups of women who either did not attend high school, or do not speak English as their primary language, were more likely to receive glyburide as opposed to insulin. While it may be that these women were more likely to decline insulin therapy secondary to socio-cultural differences or perceptions regarding insulin, we also speculate that providers may have lacked the resources to adequately teach insulin administration to non-English speakers and/or were concerned about the ability of these women to self-administer insulin injections once appropriately taught. Despite these differences, when the subsets of women with lower education or who did not report English as their primary language were analyzed, those treated with glyburide had worse perinatal outcome compared to those who received insulin.

Our study has limitations. First, treatment for a large proportion of the women in this cohort was designated as ‘oral hypoglycemic agents;’ however, of the women for whom we did have such information, nearly 99% were treated with glyburide. Prior to and during the study period, most studies on oral agents focused on glyburide, as no other agents have been shown to be safe and effective [1] until a recent large study on metformin use in GDM became available in 2008 [21]. Therefore, we speculate that nearly all women in this cohort who were treated with oral hypoglycemic agents received glyburide, with very few receiving metformin or other oral hypoglycemic agents. Of note, this limitation would only appear to bias our results toward the null as the most recent randomized controlled trial of metformin versus insulin was adequately powered and demonstrated no differences in outcomes between the two groups. Another limitation of our study is that there are a number of factors and outcomes often associated with diabetes in pregnancy which we could not examine, including prior history of macrosomia, presence of coexisting medical morbidity such as chronic hypertension, details of labor/delivery, and complications often related to hyperglycemia. While preexisting medical condition would have been interesting to examine as potential effect modifiers, we believe that they would not be associated with the exposure of interest (treatment modality of GDM for our study), and thus, by definition of confounding bias, they would likely not have biased our study results. Additionally, we did not have information regarding the degree of glycemic control or whether treatment was changed during pregnancy, since treatment groups were categorized based on treatment modality at the last Sweet Success visit. Thus, a small proportion of women who were initially started on oral hypoglycemic agents but subsequently switched to insulin were analyzed as belonging to the insulin group. Although this misclassification could be bi-directional, likely much fewer women who were started on insulin would be switched to oral agents. In this setting, the misclassification likely should bias our results toward the null. Since we consistently observed that glyburide was associated with undesirable outcomes, our findings are likely valid and may underestimate the true differences.