Objective
To evaluate glycemic control in women receiving metformin or insulin for gestational diabetes, and to identify factors predicting the need for supplemental insulin in women initially treated with metformin.
Study Design
Women with gestational diabetes who failed to achieve glycemic control with diet and exercise were randomized to receive metformin (n = 47) or insulin (n = 47). Criteria for inclusion were singleton pregnancy, diet, and exercise for a minimum period of 1 week without satisfactory glycemic control, absence of risk factors for lactic acidosis, and absence of anatomic and/or chromosome anomalies of the conceptus. Patients who were lost to prenatal follow-up were excluded.
Results
Comparison of mean pretreatment glucose levels showed no significant difference between groups ( P = .790). After introduction of the drug, lower mean glucose levels were observed in the metformin group ( P = .020), mainly because of lower levels after dinner ( P = .042). Women using metformin presented less weight gain ( P = .002) and a lower frequency of neonatal hypoglycemia ( P = .032). Twelve women in the metformin group (26.08%) required supplemental insulin for glycemic control. Early gestational age at diagnosis (odds ratio, 0.71; 95% confidence interval, 0.52–0.97; P = .032) and mean pretreatment glucose level (odds ratio, 1.061; 95% confidence interval, 1.001–1.124; P = .046) were identified as predictors of the need for insulin.
Conclusion
Metformin was found to provide adequate glycemic control with lower mean glucose levels throughout the day, less weight gain and a lower frequency of neonatal hypoglycemia. Logistic regression analysis showed that gestational age at diagnosis and mean pretreatment glucose level were predictors of the need for supplemental insulin therapy in women initially treated with metformin.
Gestational diabetes (GDM) is observed in 7% to 18% of pregnancies. Possible complications include a higher frequency of fetal macrosomia, preeclampsia, cesarean section, and neonatal hypoglycemia, all of which are substantially reduced when glucose levels are controlled, either by diet and exercise primarily or through medication when the first approach proves insufficient. Standard treatment for achieving adequate glucose levels is insulin therapy. However, this therapy requires multiple daily injections of insulin, which may reduce patient adherence. Furthermore, its high cost may preclude treatment for some patients.
Oral antidiabetic agents have been investigated as an alternative to insulin therapy because of their ease of use and lower cost. Metformin is a biguanide hypoglycemic agent that reduces hepatic gluconeogenesis and increases peripheral insulin sensitivity. Over the past few years, a number of studies have investigated the use of metformin for the treatment of GDM and 2 randomized trials have shown similar neonatal results, concluding that metformin seems to be an effective alternative for the treatment of GDM. However, response to treatment in patients with gestational diabetes is highly dependent on patient characteristics.
Considering that neonatal outcomes hinge on the adequacy of metabolic control, the primary objective of the present study was to compare glycemic control in women who received metformin or insulin for the treatment of GDM in our population. Because some patients receiving metformin do not achieve adequate glucose control, consequently having to receive insulin, a second objective was to identify factors that could predict the need for supplemental insulin in women initially treated with metformin.
Materials and Methods
Ninety-two women diagnosed with GDM, who received prenatal care at the Obstetrics Clinic of Hospital das Clinicas, Sao Paulo University School of Medicine (HC-FMUSP), Sao Paulo, Brazil, were studied prospectively between Nov. 1, 2007, and Jan 31, 2010. Criteria for inclusion were singleton pregnancy, use of diet and exercise for a minimum period of 1 week without satisfactory glycemic control, absence of risk factors for lactic acidosis (renal failure, heart failure, chronic liver disease, severe chronic pulmonary disease, coronary insufficiency, history of thromboembolic phenomena), and absence of anatomic and/or chromosome anomalies of the conceptus detected by ultrasonography. Patients were included after they had signed the free informed consent form. The study was approved by the Ethics Committee of the Clinical Board of the HC-FMUSP (no. 0530/2007). Patients lost to prenatal follow-up were excluded.
A diagnosis of GDM at the time of the study was established when the patient presented 2 or more altered results on the oral glucose tolerance test with 100 or 75 g glucose, according to the criteria of the American Diabetes Association (ADA). Values of glucose considered normal were as follows: 95 mg/dL at 0 minutes, 180 mg/dL at 60 minutes, 155 mg/dL at 120 minutes, and 140 mg/dL at 180 minutes. After the diagnosis of GDM, the women were referred for follow-up to an outpatient clinic specialized in diabetes and pregnancy. On the first prenatal visit, information on demographic was obtained and clinical data collected. For nutritional counseling, a calorie intake of 25 to 35 kcal/kg weight per day was recommended depending on the classification of the pregestational body mass index (BMI) of the patient. In addition, the calories were divided to comprise 55% carbohydrates, 15% proteins, and 30% lipids. A 30-minute walk, 3 times a week, was recommended as exercise. For glucose monitoring, a fingertip capillary blood sample was obtained 4 times per day (at fasting, 2 hours after breakfast, 2 hours after lunch, and 2 hours after dinner). The patients were asked to record all capillary glucose measurements and the memory of the glucose meters was tested weekly. The glucose reference values recommended by the ADA were used: fasting (≤95 mg/dL) and 2 hours after a meal (≤120 mg/dL). Patients were excluded if they had performed less than 85%b of the tests.
Unsatisfactory glycemic control was defined among patients who presented more than 30% of capillary glycemia results above the reference values 1 week after commencing diet therapy combined with physical activity. Medication-based treatment was then initiated. At this time, patients who met all inclusion criteria were randomized to receive either metformin (study group) or insulin (control group) according to an electronic randomization list. Data were collected weekly during return visits.
The control group received human NPH insulin. The starting dose was 0.4 units per kg body weight per day, with half the dose being administered in the morning (before breakfast), ¼ of the total dose before lunch, and ¼ at 22:00 hours. This group was asked to monitor glucose 7 times per day (at fasting, 2 hours after breakfast, 1hour before lunch, 2 hours after lunch, 1 hour before dinner, 2 hours after dinner and at 3 in the morning). The doses were adjusted weekly to achieve adequate glycemic control. If preprandial glucose levels were normal and postprandial glucose levels were high, regular insulin was added half an hour before that meal in addition to NPH insulin.
The metformin group received an initial metformin dose of 1700 mg/d (850 mg 3 times a day) and if adequate control was not achieved the dose was raised the next week to 2550 mg/d (850 mg twice a day). Patients who did not achieve satisfactory glycemic control with metformin received supplemental insulin. The data were analyzed on an intention to treat basis. Adequate control was defined for both groups as having less than 30% of capillary glycemia results above the reference values during 1 week of treatment.
Compliance was defined as having performed at least 85% of the scheduled capilar glicemias.
The women were followed up until delivery and the occurrence of preeclampsia, prematurity (birth before 37 weeks of gestation), and neonatal outcomes. Preeclampia was defined by the appearance of hypertension after 20 weeks of gestation associated to edema and proteinuria. Superimposed preeclampsia was defined as worsening of tension levels associated to edema and appearance or worsening of proteinuria in women with chronic hypertension predating pregnancy. Neontal outcomes were hypoglycemia (capillar glicemia inferior to 40 mg/dL in the first 48 hours of life), macrosomia (neonatal weight equal or superior to 4000 g), hyperbilirubinemia, and SIRS.
All neonates had capilar glicemia tested at 1.5 hours of life, 3 hours, 6 hours, 12 hours, 24 hours, 48 hours of life, and if they had symptoms.
Patients using metformin were actively asked about side effects symptoms: nausea, vomiting, and frequency of bowel movements.
The sample size was calculated assuming a difference in mean glucose level of 15 mg/dL between groups, a standard deviation of 20 mg/dL, level of significance of 5%, and power of the test of 90%. The estimated number of patients, assuming a 10% loss of subjects in each group, was 82, comprising 41 patients per group.
Mean glucose levels obtained during the whole follow-up period, from introduction of medication to birth, were compared at each time of the day by analysis of variance with repeated measures for the comparison of treatments. When a significant difference was detected, Tukey’s multiple comparison test was applied to verify if there was a difference between groups at any specific period.
Numerical variables were compared by the Student t test or Mann-Whitney test. The χ 2 test, Fisher exact test or likelihood ratio tests were used to compare categorical variables. In addition, logistic regression analysis was performed to predict the need for supplemental insulin in women initially treated with metformin. Individual variables that showed a significant difference between metformin and ‘metformin + insulin’ groups were selected for entry to the multiple logistic regression model. A level of significance of 5% was adopted for all tests.
Results
The initial sample consisted of 94 women who were randomized into the metformin group (group 1, n = 47) or the insulin group (group 2, n = 47). Two women were excluded for discontinuing prenatal care (1 from each group). This gave a final sample of 92 women (46 in the metformin group and 46 in the insulin group) ( Figure 1 ). One of the women who presented metformin intolerance wished to discontinue the medication and human NPH was therefore introduced. For the purposes of data analysis, this patient belonged to group 1. No patients were excluded because of lack of compliance.
Demographic and baseline clinical data
No significant differences between groups were observed in terms of maternal age, parity, gestational age at diagnosis of GDM, gestational age at the beginning of medication-based treatment, pregestational BMI, BMI at diagnosis, BMI at the beginning of medication-based treatment, or glycated hemoglobin (HbA1C) at diagnosis ( Table 1 ). However, there was a significant difference in the number of pregnancies between groups, with a median number of 2 pregnancies in group 1 vs 3 pregnancies in group 2 ( P = .035). No significant difference between groups was observed regarding the proportion of nulliparous women (10/46 [21.7%] in group 1 and 17/46 [37%] in group 2; P = .109), frequency of chronic hypertension predating pregnancy (14/46 [30.4%] in group 1 and 12/46 [26.1%] in group 2; P = .643), or frequency of smoking (2/46 [4.3%] in group 1 and 5/46 [10.9%] in group 2; P = .434). There was also no difference in the number of patients having previous cesarean-sections. In the group insulin 20 patients (43%) had at least 1 previous cesarean-section and in the group metformin 17 patients (36.9%) had at least 1 previous c-section ( P = .524).
| Variable | Metformin group | Insulin group | P value |
|---|---|---|---|
| Age, y | 31.93 ± 6.02 | 32.76 ± 4.66 | .464 |
| Number of pregnancies | 2 (1-8) | 3 (1-8) | .035 a |
| Parity | 1 (0-5) | 1 (0-6) | .072 |
| GA at the time of diagnosis of gestational diabetes | 30.40 ± 3.71 | 30.63 ± 3.35 | .756 |
| GA at the beginning of treatment | 32.18 ± 3.70 | 32.05 ± 3.50 | .864 |
| Pregestational BMI | 28.68 ± 5.60 | 27.99 ± 5.86 | .563 |
| BMI at diagnosis | 31.97 ± 4.71 | 31.31 ± 5.80 | .549 |
| BMI at the beginning of treatment | 31.96 ± 4.75 | 31.39 ± 5.71 | .603 |
| HbA1C at diagnosis | 5.90 ± 0.75 | 5.93 ± 0.80 | .863 |
Weight gain
With respect to weight gain, less gain was observed in the metformin group between the diagnosis of GDM and delivery (group 1: 0.53 ± 2.52 kg vs group 2: 2.3 ± 2.77 kg; P = .002) and between the beginning of medication-based treatment and delivery (group 1: 0.43 ± 1.99 kg vs group 2: 2.07 ± 2.39 kg; P = .001).
Pregnancy outcome
The 2 groups did not differ in terms of the frequency of preeclampsia (10/46 [21.7%] in group 1 and 7/46 [15.2%] in group 2; P = .420), Of those, pre-eclampsia was superimposed to chronic hypertension in 5 patients that had chronic hypertension in group 1 (5/14, 35.7%) and in 3 patients (3/12, 25%) in group 2 ( P = .683). The 2 groups did not differ in terms of prematurity (5/46 [10.9%] in group 1 and 5/46 [10.9%] in group 2; P > .99), and cesarian section (33/46 [71.7%] in group 1 and 30/46 [65.2%] in group 2; P = .774).
Neonatal outcomes
No significant differences between the 2 groups were observed regarding the following neonatal outcomes: gestational age at birth (group 1: 38.33 ± 1.45 weeks vs group 2: 38.24 ± 1.53 weeks; P = .776), 1-minute Apgar score (group 1: 9 [0-10] vs group 2: 9 [4-10]; P = .980), 5-minute Apgar score (group 1: 10 [0-10] vs group 2: 10 [0-10]; P = .188), umbilical artery pH at birth (group 1: 7.22 ± 0.07 vs group 2: 7.22 ± 0.08; P = .824), or newborn weight (group 1: 3143.7 ± 446.6 g vs group 2: 3237.6 ± 586.8 g; P = .390) ( Table 2 ). There were no fetuses with macrosomia in the group metformin vs 3 (6.5%) cases in the insulin group ( P = .242). In the group receiving metformin 6 neonates (13%) were small for gestational age (SGA), 38 (82.6%) were adequate for gestational age (AGA) and 2 (4.3%) were large for gestational age (LGA). In the group receiving insulin 4 neonates (8.7%) were SGA, 39 (84.8%) were AGA, and 3 (6.5%) were LGA ( P = .735).
| Variable | Metformin group | Insulin group | P value |
|---|---|---|---|
| Gestational age at birth | 38.33 ± 1.45 | 38.24 ± 1.53 | .776 |
| 1-min Apgar score | 9 (4-10) | 9 (0-10) | .980 |
| 5-min Apgar score | 10 (0-10) | 10 (0-10) | .188 |
| Umbilical artery pH at birth | 7.22 ± 0.07 | 7.22 ± 0.08 | .824 |
| Newborn weight | 3143.7 ± 446.6 | 3237.6 ± 586.8 | .390 |
A higher frequency of neonatal hypoglycemia was observed in cases treated with insulin (10/46, 22.2%) compared with newborns from the metformin group (3/46, 6.5%) ( P = .032) ( Table 3 ).
| Variable | Metformin group | Insulin group | P value | ||
|---|---|---|---|---|---|
| n | % | n | % | ||
| Macrossomia | 0 | 0 | 3 | 6.5 | .242 |
| Neonatal hypoglicemia | 6.5 | 13 | 10 | 22.2 | .032 a |
| Respiratory distress syndrome | 7 | 15.2 | 7 | 15.2 | .964 |
| Hyperbilirrubinemia | 4 | 8.7 | 5 | 11.1 | .739 |
Metformin side effects
Twenty-one (45.65%) of the 46 women who received metformin reported some side effect. The most frequent symptom was nausea, reported by 8 (38%) women. Seven (33%) patients reported an increased frequency of bowel movements and 3 (14.28%) reported the triad of nausea, vomiting, and increased frequency of bowel movements. Only one woman had epigastralgia and constipation and preferred to stop metformin and to start taking insulin.
Glycemic control
With respect to glycemic control, no significant difference in mean pretreatment glucose levels was observed between groups ( P = .790). However, after introduction of the drugs, higher mean glucose levels were observed in the insulin group ( P = .020), mainly because of higher levels observed after dinner ( P = .042) ( Table 4 ).
