Comparison of glyburide and insulin in the management of gestational diabetes: A meta-analysis

Objective The aim of this meta-analysis was to determine the efficacy and safety of glyburide as a treatment for gestational diabetes mellitus (GDM) compared to insulin. Methods A meta-analysis was conducted to compare the management of gestational diabetes with glyburide and insulin. Studies fulfilling all of the following inclusion criteria were included in this meta-analysis: subjects were women with gestational diabetes requiring drug treatment; the comparison treatment included glyburide vs insulin; one or more outcomes (maternal or neonatal) should be provided in the individual study; the study design should be a randomized control trial. Exclusion criteria: non-RCT studies; non-human data. PubMed, Embase and CENTRAL databases were searched from inception until 10 October 2016. Results Ten randomized control trials involving 1194 participants met the inclusion criteria and were included. 13 primary outcomes (6 maternal, 7 neonatal) and 26 secondary outcomes (9 maternal, 17 neonatal) were detected and analyzed in this study. Glyburide significantly increased the risk of any neonatal hypoglycemia [risk ratio (RR), 1.89; 95% confidence interval (95%CI), 1.26 to 2.82; p = 0.002]. Sensitivity analysis confirmed robustness of this result [RR, 2.29; 95%CI, 1.49 to 3.54; p = 0.0002]. No differences were observed between the two groups with respect to birth weights [mean difference (MD), 79; 95%CI, -64 to 221.99; p = 0.28] and the risk of macrosomia [RR, 1.69; 95%CI, 0.57 to 5.08; p = 0.35]. Conclusion For women with gestational diabetes, no differences in maternal short term outcomes were observed in those treated with glyburide or insulin. However, the incidence of neonatal hypoglycemia was higher in the glyburide group compared to the insulin group.


Results
Ten randomized control trials involving 1194 participants met the inclusion criteria and were included. 13 primary outcomes (6 maternal, 7 neonatal) and 26 secondary outcomes (9 maternal, 17 neonatal) were detected and analyzed in this study. Glyburide significantly increased the risk of any neonatal hypoglycemia [risk ratio (RR), 1

Conclusion
For women with gestational diabetes, no differences in maternal short term outcomes were observed in those treated with glyburide or insulin. However, the incidence of neonatal hypoglycemia was higher in the glyburide group compared to the insulin group. PLOS

Introduction
Gestational diabetes mellitus (GDM) was originally defined as any degree of glucose intolerance that was first recognized during pregnancy [1]. More recently, the American Diabetes Association has recommended that women diagnosed with diabetes in the first trimester be classified as having type 2 diabetes, whereas GDM should be defined as diabetes diagnosed in the second or third trimester of pregnancy that is clearly neither type 1 nor type 2 diabetes [2]. According to the "International Association of Diabetes and Pregnancy Study Group (IADPSG) diagnostic criteria" [3], the prevalence of GDM has been reported to be 17.6%, 4.2%, 11.8%, 9.5%, 23.3%, 8.6%, and 45% in Singapore [4], Greenland [5], Switzerland [6], South Korea [7], Sri Lanka [8], Sub-Saharan Africa [9], and the United Arab Emirates [10], respectively. In recent years, the prevalence of GDM has increased, which has been attributed to the higher incidence of obesity in the general population and the increase in the number of pregnancies in older women [11]. GDM carries risks for both the mother and the infant [12].
In mothers, it is associated with a higher risk of pregnancy-induced hypertension, pre-eclampsia, cesarean delivery, and an increased risk of developing diabetes later in life. Infants of women with GDM are at a higher risk of neonatal hypoglycemia, macrosomia, respiratory distress syndrome, neonatal death, and stillbirth [13]. Therefore, GDM is associated with significant transgenerational maternal and neonatal morbidity [14]. Glucose levels can be managed by lifestyle changes alone in most patients with GDM. Over the past few decades, insulin therapy has been the first-line agent recommended for the treatment of GDM in patients that have failed to achieve desired glycemic goals through lifestyle changes [15,16]. However, there are several disadvantages to this approach, including hypoglycemia, weight gain, the requirement for multiple daily subcutaneous injections, the need to train patients in the required technique, and an increased medical cost burden [17,18]. In recent years, a growing body of research has suggested that oral hypoglycemic agents, such as glyburide and metformin, could be used for the treatment of GDM [15,[19][20][21][22][23][24]. Metformin has been increasingly recognized as an alternative to insulin therapy for GDM [16,25] and there is strong evidence for its effectiveness and safety [26,27]. Currently, treatment with metformin is preferable to insulin for maternal health if it sufficiently controls hyperglycemia [2,11].
Glyburide belongs to the class of longer-acting sulfonylureas. Data regarding its use in GDM are conflicting [18,24,[26][27][28][29][30][31]. Two new randomized control trials comparing treatment with glyburide and insulin in GDM have recently been published [29,32]. A meta-analysis of the updated data, including previously available data and the recently published trials, might provide stronger evidence with respect to the effectiveness and safety of glyburide. Therefore, the aim of our study was to reassess the efficacy and safety of glyburide compared to those of insulin in the management of GDM based on all available data. These findings will provide valuable evidence regarding the use of glyburide in the treatment of GDM.

Inclusion and exclusion criteria
Inclusion criteria: study subjects were GDM patients who were not well controlled with lifestyle adjustment and needed drug treatment to control their glycemic levels; the treatment schedule in the control and interventional group were insulin and glyburide, respectively; the study design was randomized control trial; primary or secondary outcomes reported in the trials included maternal weight gain during pregnancy, type of delivery, neonatal hypoglycemia, birth weight, and macrosomia; the language used for the individual studies was not limited; only studies with full text available online were included in the study. Studies meeting all the above criteria were included in the meta-analysis.
Exclusion criteria: retrospective cohort study; case report; reviews; animal experiment.

Literature search and study selection
Databases searched included PubMed, EMBASE, and CENTRAL (Cochrane Cental Register of Controlled Trials), from inception through October 2016. The following keywords were used in the searches: glyburide or glibenclamide; gestational diabetes, gestational diabetes mellitus; randomized controlled trials. S1 Table shows the detailed search strategy. Additionally, literature including unpublished studies, data from academic conference and dissertations were manually identified in searches from Google Scholar and other sources. Two investigators performed the search processes and study selection independently according to the above criteria. Discrepancies were resolved by discussion or by involving a third assessor.

Assessment of risk of bias and data collection
Two review authors independently assessed the quality of each included study by using the tool in the Cochrane Handbook for Systematic Reviews of Intervention. The major biases including random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other biases. Low risk, high risk or unclear risk was used in the assessment. The overall risk of bias was presented in six domains. Data were collected for baseline information and outcomes according to the above criteria. Any disagreement was resolved by discussion.

Statistical analysis
Review Manager software (version 5.3) was used to perform the statistical analysis. Data were pooled according to the type of outcomes. Dichotomous outcomes were calculated using risk ratio (RR) with 95% confidence intervals (CI). Continuous outcomes were calculated using mean difference (MD) with 95% confidence intervals (CI). The Inverse Variance method was used for continuous outcomes and the Mantel Haenszel method for dichotomous outcomes. Heterogeneity was measured by I 2 statistics (I 2 >50% was regarded as heterogeneity) and Q statistics (P<0.1 was considered heterogeneity). A fixed-effects model was used when I 2 <50% (homogeneity), otherwise a random-effects model was used. P values less than 0.05 were considered to be statistically significant. Potential publication bias was assessed by Funnel plots.

Results
Literature search and study characteristics S1 Table provides the detailed search strategy. A total of 174 articles were found through the searches of three databases (PubMed, EMBASE, and CENTRAL), with 137 records left after removing duplicates. After reading the titles or abstracts, 118 were excluded leaving 19 studies for further review. Fig 1 shows the process of the literature search and selection. A final set of ten randomized control trials with a total of 1194 participants meeting the inclusion criteria were included in this meta-analysis (575 on glyburide; 619 on insulin) [19,24,28,29,[32][33][34][35][36][37].
Countries included in these studies were USA, Brazil, India and Iran. The two larger studies involved more than 100 subjects each [19,32], while the other eight trials involved fewer than 100 participants each. Characteristics and baseline data are shown in Tables 1-3, respectively. Baseline data were collected and compared for maternal age, prepregancy BMI (body mess index), gestational age at entry, fasting plasma glucose at OGTT (oral glucose tolerance test), 2h postprandial glucose at OGTT, and HbA1c (glycated haemoglobin) level at entry (see Table 4), and the results indicated that there was no difference observed between the groups for these baseline characteristics.

Assessment of risk of bias
We performed quality assessments of the 10 included studies. The studies by Mirzamoradi [29] and Behrashi [32] are the new RCTs included in our metaanalysis. Six of the studies provided specific sequence generation methods [19,28,29,32,36,37], while the remaining four studies did not offer information on this domain [24,[33][34][35]. Allocation concealment was provided in five studies [19,29,33,36,37], whereas this information was unclear in the other five studies [24,28,32,34,35]. For blinding of participants and personnel, blinding of outcomes assessment, we found low risk of bias in all the included studies. We observed attrition bias in two studies [35][36][37], while the other eight studies provided relatively complete outcomes data. Only one study showed unclear risk of bias for selective reporting [35], while for the rest, no selective reporting was noticed. No other risk of bias was observed in all the included studies. Most of the data in this meta-analysis was from studies at low risk of bias.

Primary outcomes
None of the primary maternal outcomes presented a significant difference between the glyburide and insulin groups as shown in Fig 4. No significant differences were found with regard to   Fig 5. All the primary outcomes were reported as shown in Table 5.

Secondary outcomes
Nine secondary maternal outcomes were reported as shown in  Other neonatal outcomes were shown in Table 6. No study reported the incidence of 1 minute Apgar score <7 or 5 minute Apgar score <7. Five studies reported severe neonatal hypoglycemia. Severe neonatal hypoglycaemia was defined as 'required intravenous therapy or  [24,34,36]; however, the definition was not available in another 2 studies [29,32]. Treatment with glyburide increased the incidence of severe neonatal hypoglycemia compared to treatment with insulin, but there was no statistically significant difference [RR,4.

Heterogeneity test and sensitivity analysis
Sources of heterogeneity were detected as statistically significant heterogeneity existing in several primary outcomes, such as gestational age at delivery, birth weight, macrosomia and any neonatal hypoglycemia. Funnel plots indicated that the newest study, published in 2016 [32] was the major source for heterogeneity. Sensitivity analysis was performed for the above primary outcomes after excluding the newest study. As shown in S1 Fig, no difference was observed for gestational age at delivery [MD, 0.06; 95%CI, -0.16 to 0.28], which is consistent with the previous result, but I 2 decreased from 77% to 0%. After excluding the newest study, birth weight appeared higher in patients receiving glyburide than in those receiving insulin [MD, 109.16; 95%Cl, 42.59 to 175.72; p = 0.001], with I 2 decreasing from 83% to 0%. Risk of macrosomia appeared higher in patients receiving glyburide than in those receiving insulin

Discussion
Recently, oral hypoglycemic agents have been identified as alternatives to insulin in the management of GDM. An alternative was sought owing to high insulin costs, inconvenience, and the probability of a higher risk of hypoglycemia. This updated meta-analysis was conducted to evaluate the efficacy and safety of glyburide in patients with GDM and to compare it with insulin therapy. Several maternal and neonatal outcomes were assessed. Maternal glycemic control was not significantly different between the two treatment groups in our meta-analysis, which indicated that glyburide and insulin were equally effective for the treatment of GDM. This was consistent with previous studies [26][27][28]30]. However, with respect to the oral treatment of GDM, metformin may be preferred over glyburide as first-line therapy, according to a recent RCT conducted by Nachum Z et al [38]. In addition, similar to the results reported by Balsells et al [27], there were no cases of severe maternal hypoglycemia reported in the 10 trials included in this meta-analysis. Furthermore, there were no differences in the other primary indicators of maternal outcome, including weight gain, pre-eclampsia, caesarean section, and gestational age at delivery, between the two groups, which was also consistent with the findings of previous studies [27,39]. However, Malek et al [40] found that the risk of pre-eclampsia was higher in the group treated with glyburide than in the group treated with insulin. With respect to secondary maternal outcomes, such as maternal trauma, Cord C peptide, and Cord insulin, our results were consistent with those of a previous review [27], which suggested that glyburide did not harm the mother compared to insulin.
Neonatal hypoglycemia frequently occurs in infants of women with GDM. All trials included in this meta-analysis reported the incidence of neonatal hypoglycemia and the results indicated a higher risk of any neonatal hypoglycemia after maternal treatment with glyburide compared to treatment with insulin (p = 0.002). This was consistent with several previous meta-analyses [26,27,39,40] and a sensitivity analysis confirmed the robustness of this result. Like any neonatal hypoglycemia, treatment with glyburide resulted in an increase in the incidence of severe neonatal hypoglycemia compared to treatment with insulin; however, the difference was not statistically significant.
All trials included in this meta-analysis indicated that maternal treatment with glyburide resulted in an increased birth weight compared to treatment with insulin, except the newest large-scale RCT carried out by Behrashi et al [32]. Our results suggested that there was no difference between the two groups with regard to birth weight, which was different from a previous conclusion [27]. Macrosomia was consistently defined as birth weight no less than 4000 g in the 6 evaluated studies. Like birth weight, no differences were observed between the two groups with respect to the risk of macrosomia in our meta-analysis. Similarly, several studies showed no significant difference between the insulin and glyburide groups in the prevalence of macrosomia [1,41,42]. However, several other studies [27,30,40,43] indicated that glyburide was associated with a higher incidence of macrosomia. Conversely, Behrashi et al [32] found that the incidence of macrosomia in the glyburide group was significantly lower than that in the group that received insulin. This new trial [32] was the largest trial in recent years. However, the time when birth weight was measured was not standard and the starting insulin dose was 0.2 U/kg, which was lower than that used in other trials. It is possible that these two factors accounted for the different results in this trial. A heterogeneity test was conducted by excluding the newest study and the findings indicated that birth weight was significantly higher when the mothers received glyburide than when the mothers received insulin. In addition, macrosomia occurred significantly more often when the mothers received glyburide than when the mothers received insulin.
In this meta-analysis, there was no statistically significant difference between the treatment groups with respect to the incidence of neonatal hyperbilirubinemia and phototherapy. Balsells et al [27] also found no difference in the incidence of neonatal jaundice between the two treatment groups. Other secondary neonatal outcomes, including neonatal respiratory distress syndrome, stillbirth, and NICU admission, were not significantly different between the two groups. This was also similar to the conclusion of Balsells et al [27]. However, the results of Malek et al [40] indicated that the risk of NICU admission was higher in the group treated with glyburide than in the group treated with insulin. In addition, our results indicated that there were no significant differences in the prevalence of hypocalcemia, polycythemia, birth trauma, shoulder dystocia, or neonatal circumference between the two groups. These indicators were all evaluated for the first time in our study when compared to other studies [27,40].
In summary, the use of glyburide in pregnancy for women with GDM appears to be as effective as the use of insulin, but neonatal hypoglycemia should be monitored. Additionally, the potential risk of glyburide to the fetus is unclear, especially over the long-term, and should be reassessed in the future because the evidence that indicates glyburide is noticeable in the fetal circulation [44].

Limitations in current evidence
There were several limitations in our present meta-analysis that deserve comment. First, only one original study that was not written in English was included in this meta-analysis, which could have resulted in bias or limited our ability to draw substantial conclusions. Second, some outcomes were reported in only one study or no cases were reported in some of the trials included in the meta-analysis, which limited the analysis of some of the outcomes of interest. Third, none of these studies evaluated long-term maternal and neonatal outcomes. In the future, additional long-term data on maternal and neonatal outcomes should be evaluated to confirm the safety of glyburide use in women with GDM. Moreover, the quality assessment of the included trials indicated that not all of the studies were of high quality.

Conclusions
Our study indicated that glyburide was safe and effective for use in GDM, provided neonates are monitored for hypoglycemia.
Supporting information S1