Objective
We sought to determine if insulin detemir (IDet) is noninferior to insulin neutral protamine Hagedorn (NPH) for the treatment of gestational diabetes mellitus (GDM) and type 2 diabetes mellitus (T2DM) in pregnancy.
Study Design
We conducted a randomized, controlled noninferiority trial of women with GDM and T2DM who entered our Diabetes in Pregnancy Program from March 2013 through October 2014. Exclusion criteria were type 1 diabetes, age <18 years, and insulin allergy. Women who failed to achieve good glycemic control (GC) (mean blood glucose [BG] <100 mg/dL) on diet and/or hypoglycemic agents were randomized to receive either IDet or NPH, with short-acting insulin aspart added as needed. Patients were instructed to test BG 4 times a day (fasting and 2-hour postprandial). Targets of GC were fasting BG <90 mg/dL and postprandial BG <120 mg/dL, and insulin was adjusted as needed to achieve the targets. The primary outcome was overall mean BG during insulin treatment; secondary outcomes included overall mean postprandial and fasting BG, median number of weeks to achieve GC, percent of patients with overall GC, maternal weight gain, perinatal/neonatal outcomes, and number of hypoglycemic events. Power analysis (90% power) determined that 88 patients would need to be randomized, assuming a maximal acceptable difference in overall mean BG of 7 mg/dL (SD ± 10 mg/dL). A per protocol analysis was performed.
Results
In all, 105 women were randomized. Eighteen women were excluded leaving 87 participants for analysis (45 NPH, 42 IDet). Maternal characteristics were similar in both groups. The difference in the mean BG of the groups was 2.1 mg/dL with a 1-sided upper 95% confidence limit of 5.5 mg/dL (less than the maximal acceptable difference of 7 mg/dL; P = .2937). There was no significant difference in the primary outcome when an intent-to-treat analysis was performed or when the T2DM patients were excluded. The time to achieve GC was similar in both groups. There were no differences in perinatal outcomes and maternal weight gain among the groups. There were more hypoglycemic events per patient in the NPH group.
Conclusion
IDet is noninferior to insulin NPH for the treatment of GDM and T2DM in pregnancy.
Diabetes in pregnancy, both gestational and pregestational, has adverse effects on maternal and fetal/neonatal outcomes. Prior studies have shown that enforcing strict glycemic control (GC) in the treatment of diabetes can improve pregnancy outcomes including fetal demise, birth injuries, shoulder dystocia, and congenital anomalies, and mitigate other maternal complications such as preeclampsia and weight gain in pregnancy.
Type 2 diabetes mellitus (T2DM) and gestational diabetes mellitus (GDM) have a similar pathophysiology and thereby similar management in pregnancy. Treatment options include diet modification, oral hypoglycemic agents (ie, glyburide or metformin), and insulin. Although previous studies have shown comparable GC when using either glyburide or insulin in pregnancy, insulin remains a mainstay of therapy, either as a first choice or when treatment with oral hypoglycemics fails to achieve good GC. However, the various available types of insulin have not been subject to rigorous study regarding the advantages and disadvantages of the use of one specific type over the other in pregnancy.
Insulin detemir (IDet) is a long-acting insulin analog that is increasingly being used in the treatment of diabetes. In nonpregnant patients, GC with IDet is comparable to neutral protamine Hagedorn (NPH), and there are fewer episodes of hypoglycemia. Because of the benefits seen in nonpregnant patients, recently IDet has become a favorable agent for treatment of diabetes in pregnancy.
We sought to determine if IDet is noninferior to insulin NPH in the treatment of hyperglycemia in pregnant patients with GDM and T2DM. We hypothesize that GC in women using IDet is not worse than control achieved with NPH.
Materials and Methods
This open-label, noninferiority, randomized study was conducted between March 2013 and January 2015. The study was approved by the institutional review board (#12-166) of Mount Sinai Roosevelt Hospital and was registered in ClinicalTrials.gov ( NCT01837680 ). Eligible subjects were all pregnant women with a viable singleton or twin gestation, with either preexisting T2DM or GDM at ≤34 weeks, in need of medical therapy, who received care at our institution. GDM was diagnosed based on a positive 100-g 3-hour glucose tolerance test (using ≥1 abnormal values by the Carpenter and Coustan criteria ), or a positive 75-g 2-hour glucose tolerance test using the International Association of Diabetes and Pregnancy Study Groups criteria. Exclusion criteria included patients <18 years of age, with type 1 diabetes, and with a known allergy/prior adverse reaction to NPH or IDet. Eligible patients were recruited through our Diabetes in Pregnancy Program (DIPP) from March 2013 through October 2014. Women who failed to achieve good GC (mean blood glucose [BG] <100 mg/dL) on diet and/or hypoglycemic agents and who chose to be treated with insulin were approached to participate in the study and informed written consent obtained prior to randomization.
The randomization sequence generation was obtained by nQuery Advisor (Statistical Solutions, Boston, MA) assignment by the statistician using a blocked randomization scheme (stratified by GDM and T2DM) with a varying block size (2,4,6,8) sequence to produce a computer-generated list. The allocation of assignment was concealed and prepared by placement in sequentially numbered, opaque, sealed envelopes that were kept in the DIPP office; envelopes were drawn in consecutive order. Participants, care providers, and those assessing outcomes were not blinded or masked.
The 2 interventions of NPH and IDet were administered as follows. Current weight was obtained at the visit and initial daily total insulin dose was determined based on patient weight (in kilograms) and trimester. In the first trimester, patient weight was multiplied by 0.7, in the second trimester by 0.8, and in the third trimester by 0.9 for the total daily dose of insulin (in units). The breakdown of the dosing for the insulin was based on published data and modified to incorporate the pharmacologic differences of NPH and IDet. Of the total daily insulin dose, 60% was allotted to the morning total dose of insulin, while the remaining 40% allotted to the evening total dose. Of the morning dose, two-thirds was allotted to the long-acting insulin and one-third to short-acting insulin. If the patient was taking NPH, the entire dose of short-acting insulin was administered with breakfast. However, if the patient was randomized to IDet, she took half of the short-acting dose with breakfast and the other half with lunch. The evening insulin dose (40% of the total dose) was divided in 2: half the dose taken as short-acting insulin with dinner, and the other half as long-acting insulin (either NPH or detemir) at bedtime. Doses were rounded down if decimals were present. For example, if a patient in the second trimester weighed 125 kg, her daily total insulin dose was 100 U; her total morning dose was 60 U and her total evening dose was 40 U. If she were randomized to IDet, she would get 40 U of IDet in the morning, 10 U of insulin aspart with breakfast, and 10 U of aspart with lunch. She would subsequently get 20 U of aspart with dinner and 20 U of IDet at bedtime. If she were on NPH, her long-acting units would remain the same, but her morning aspart would be 20 U at breakfast only.
Patients were managed by DIPP perinatologists according to the routine protocol as stated in our previous publication. Briefly, all patients were provided with prescriptions for insulin, a memory reflectance glucometer that designated glucose values in relation to meal times, lancets, and strips to be used with the home glucometer, which allowed for observation and monitoring of glucose values. Patients underwent teaching by a registered nurse and diabetic educator to learn how to self-administer insulin. Fingersticks were obtained 4-7 times a day (fasting, after meals, and ±before meals), every day. Patients were followed up in the DIPP every 1-3 weeks, as necessary, as determined by the achieved level of GC. At every visit, patient weight, blood pressure, urine samples to assess for ketones and protein, and fetal heart tones were obtained. Glucose values were downloaded to computer software at each visit to determine mean overall glucose, mean fasting glucose, mean postprandial glucose, number of fingersticks/d, and percentage of glucose values within, above, and below target values. Insulin doses were adjusted to maintain good GC based on targets (fasting BG <90 mg/dL, 2-hour postprandial values <120 mg/dL, overall mean glucose <100 mg/dL). Patients also had concomitant fetal surveillance starting at 32-34 weeks with serial sonographic growth scans as necessary. All patients were followed up until delivery.
The primary outcome of the study was overall mean glucose during the treatment with insulin therapy until delivery. Secondary outcomes included overall mean fasting BG, overall mean postprandial BG, time to achieve GC (sustained for 2 weeks), proportion of patients with GC overall and at specific gestational ages, maternal weight gain, and perinatal/neonatal outcomes. Unintended harms and side effects were described in the groups (allergies, number of symptomatic hypoglycemic events at any BG, and number of biochemical hypoglycemic events defined as BG <60 mg/dL regardless of symptoms).
The sample size calculation was set to achieve 90% power to detect noninferiority using a 1-sided, 2-sample t test. With a noninferiority margin of 7 mg/dL (SD ± 10) (deemed feasible based on prior retrospective data ) and a significance level of .05, a sample size of 36 patients per group is required. To account for a potential 20% incomplete/dropout rate, 44 patients per group (total of 88 patients) were targeted for recruitment. A per protocol analysis was performed using Student t test, Fisher exact test, Wilcoxon rank sum test, Poisson regression model, negative binomial regression model, and Kaplan Meir curves as appropriate. Specifically, for the primary outcome, a mixed effects model with a random intercept and unstructured correlation was used to estimate least squares means (SE) to account for the different time intervals between visits of each patient. Additionally, since the mean glucose distributions were skewed to the right, medians and interquartile ranges were computed for the primary outcome within each group and differences between group medians were estimated with the simplex algorithm via quantile regression. Lastly, for the primary outcome, an additional intention-to-treat analysis was performed to further examine the validity of the results. All statistical analyses were conducted using SAS version 9.2 (SAS Institute, Cary, NC) and hypothesis testing was conducted at the 5% level of significance.
Results
During the study period of March 2013 through October 2014, 740 women entered the DIPP. In all, 635 women were excluded, leaving 105 women consented for randomization (53 to the NPH group, and 52 to the detemir group). In the detemir group, 46 patients received the allocation for the duration of treatment and 6 patients did not as they had an allergic reaction. Two patients were lost to follow-up, and 2 discontinued the detemir due to personal preference and switched to an oral hypoglycemic regimen or diet therapy. This left 42 patients for analysis, with 10 patients excluded (6 not receiving the allocated drug, 2 lost to follow-up, 2 discontinued). In the NPH group, 51 patients received the allocated treatment, and 2 patients did not as their insurance did not cover NPH. Two patients were lost to follow-up. Four patients discontinued NPH and switched to diet or oral hypoglycemic agents. This left 45 patients for analysis, with 8 patients excluded (2 did not receive the allocation, 2 lost to follow-up, and 4 discontinued) ( Figure 1 ).
All results are reported from the per protocol analysis. An additional intention-to-treat analysis was performed with respect to the primary outcome.
Demographics and baseline characteristics in both groups were similar ( Table 1 ).
| Demographic | IDet, n = 42 | NPH, n = 45 |
|---|---|---|
| Age, y | 35 [31–38] | 35 [32–38] |
| Race | ||
| White | 11 (26%) | 17 (38%) |
| Black | 7 (17%) | 5 (11%) |
| Hispanic | 12 (29%) | 15 (33%) |
| Native American/Alaskan | 12 (29%) | 6 (13%) |
| Biracial/multiracial | 0 (0%) | 0 (0%) |
| Other | 0 (0%) | 2 (4%) |
| T2DM | 7 (17%) | 7 (16%) |
| GDM in previous pregnancies | 8 (19%) | 9 (20%) |
| Multiple gestation | 2 (5%) | 3 (7%) |
| Polycystic ovary syndrome | 5 (12%) | 12 (27%) |
| Chronic hypertension | 5 (12%) | 6 (13%) |
| Renal disease | 1 (2%) | 5 (11%) |
| Thyroid disease | 6 (14%) | 8 (18%) |
| Prepregnancy body mass index, kg/m 2 | 28.3 [24.9–33.8] | 28.6 [24.4–31.1] |
| Normal | 11 (27%) | 14 (31%) |
| Obese | 14 (34%) | 12 (27%) |
| Overweight | 12 (29%) | 16 (36%) |
| Morbidly obese | 4 (10%) | 3 (7%) |
| Gestational age diagnosed, wk | 26.1 [24.8–27.1] | 26.6 [25.4–28.2] |
| <20 | 3 (9%) | 4 (11%) |
| 20–24 | 2 (6%) | 0 (0%) |
| 24–28 | 24 (75%) | 22 (61%) |
| >28 | 3 (9%) | 10 (28%) |
| Previous management | ||
| Diet | 35 (83%) | 39 (87%) |
| Diet, metformin | 0 (0%) | 1 (2%) |
| Glyburide | 3 (7%) | 2 (4%) |
| Metformin | 3 (7%) | 2 (4%) |
| Metformin and glyburide | 1 (2%) | 0 (0%) |
| Other type of insulin | 0 (0%) | 1 (2%) |
| Gestational age at entry to DIPP, wk | 27.3 [23.3–28.5] | 28.1 [25.1–29.3] |
| Gestational age insulin started, wk | 29.6 [27.5–31.4] | 30.0 [25.1–31.5] |
| Time between visits, wk | 1.5 [1.3–2.0] | 1.5 [1.3–1.8] |
For the primary outcome, the difference between the groups’ overall mean glucose was 2.1 mg/dL, with a 1-sided upper 95% confidence limit of 5.5 (less than the maximal acceptable difference of 7 mg/dL; P = .2937) ( Table 2 ). Because the primary outcome values were right skewed, the data were additionally analyzed using nonparametric statistics. The difference between the 2 groups with respect to the median overall, fasting, and postprandial glucose values did not reach significance. Mean overall, fasting, and postprandial BG were also compared after removing the T2DM patients, and there was no statistically significant difference between the 2 groups. In addition to the per protocol analysis, the primary outcome was also calculated in an intention-to-treat analysis. In this analysis, the difference between the groups’ mean glucose was 1.3 mg/dL, with a 1-sided upper 95% confidence limit of 4.4 (less than the maximal acceptable difference of 7 mg/dL; P = .4598). There were also no differences with respect to the mean postprandial and fasting BG ( Table 3 ).
| Mean BG | IDet, n = 42 | NPH, n = 45 | Difference | P value | 1-sided upper 95% confidence limit |
|---|---|---|---|---|---|
| Overall BG | 109.5 ± 10.0 | 107.4 ± 7.1 | 2.1 | .2937 | 5.5 |
| Fasting BG | 100.7 ± 10.1 | 97.3 ± 7.4 | 3.4 | .1093 | 6.9 |
| Postprandial BG | 115.2 ± 10.2 | 112.9 ± 8.9 | 2.3 | .3204 | 6.0 |
| Median BG a | |||||
| Overall BG | 103.3 [100.0–116.3] | 103.7 [99.6–108.6] | –0.4 | .8542 | 3.0 |
| Fasting BG | 92.2 [88.5–100.3] | 91.3 [87.0–95.3] | 0.9 | .5476 | 3.3 |
| Postprandial BG | 108.8 [104.0–120.6] | 108.7 [103.0–115.4] | 0.1 | .9550 | 4.8 |
| Mean BG excluding T2DM patients | IDet, n = 35 | NPH, n = 38 | |||
| Overall BG | 107.3 ± 8.3 | 107.2 ± 7.2 | 0.1 | .9503 | 3.4 |
| Fasting BG | 97.4 ± 8.1 | 96.8 ± 7.3 | 0.6 | .7642 | 3.9 |
| Postprandial BG | 113.3 ± 9.1 | 112.1 ± 8.5 | 1.2 | .5870 | 5.1 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
