Maternal Glycemic Control and Hypoglycemia in Type 1 Diabetic Pregnancy: A randomized trial of insulin aspart versus human insulin in 322 pregnant women

A total of 417 women with type 1 diabetes participated in this open-label, randomized, parallel-group study conducted at 63 sites in 18 countries, mainly within Europe. The study was performed in accordance with the Declaration of Helsinki and was approved by respective ethics committees and health authorities according to local regulations. Written informed consent was obtained from subjects before study start.

Eligible subjects were aged ≥18 years with insulin-treated type 1 diabetes for ≥12 months and were either pregnant with a singleton pregnancy (gestational age ≤10 weeks) or planning to become pregnant. A1C was ≤8% at confirmation of pregnancy. Subjects not pregnant at screening were withdrawn if not pregnant ≤12 months after randomization. Subjects with multiple pregnancy, fertility treatment, clinically significant gynecological conditions, diabetic nephropathy; or medical problems; a previous child born with major congenital malformations; multiple miscarriage; or stillbirths (more than two) were excluded.

Subjects were randomized (1:1) to IAsp (100 units/ml; Novo Nordisk, Bagsvaerd, Denmark) or HI (100 IU/ml; Novo Nordisk) in combination with NPH insulin (Novo Nordisk) one to four times per day. Subjects were allocated to the lowest available treatment number at each center.

IAsp was injected immediately before meals and HI within 30 min before the meal. All insulins were injected subcutaneously using the NovoPen 3.0 (Novo Nordisk). Because study insulin injection timing varied, an open-label approach was used. The starting dose for both study insulins was 100% of dose at study entry. Insulin doses were titrated to optimal levels throughout the study based on self-monitored plasma glucose levels and the targets for blood glucose control: preprandial 4.1–6.1 mmol/l, 1 h postprandial <8.6 mmol/l, 2 h postprandial <7.5 mmol/l (according to American Diabetes Association guidelines), and A1C <6.5%.

Subjects pregnant at screening attended a first-pregnancy assessment/randomization (P1) visit (<2 weeks after screening); clinic visits at the end of the first, second, and third trimester (P2–P4) (12, 24, and 36 weeks’ gestation); delivery/termination (T); and follow-up visit 6 weeks postdelivery. Subjects not pregnant at randomization attended 3-monthly clinic visits until pregnant. On pregnancy confirmation, they attended visit P1. Thereafter, clinic visits were as described for subjects pregnant at screening. Study duration and number of visits varied between subjects depending on time of conception relative to screening. Maximum duration of participation was 22 months.

Primary study end point was major (requiring third-party assistance with plasma glucose <3.1 mmol/l or reversal of symptoms after food, glucagon, or intravenous glucose) hypoglycemia during pregnancy. Minor (plasma glucose <3.1 mmol/l with or without symptoms) and symptoms-only (no plasma glucose measurement or plasma glucose >3.0 mmol/l) hypoglycemia were also recorded by subjects in their diaries. Nocturnal hypoglycemia was taken as episodes between midnight and 0600 h.

Efficacy end points were A1C and self-measured 8-point plasma glucose profile. Subjects were asked to perform an 8-point plasma glucose profile during the week before randomization and clinic visits P1–P4 using a Medisense (Maidenhead, U.K.) blood glucose meter. Other safety assessments included maternal adverse events, obstetric complications, diabetes complications, pregnancy outcomes, and delivery details.

Treatment satisfaction was assessed using the Diabetes Treatment Satisfaction Questionnaire (30) at randomization and at follow-up visits for subjects pregnant at screening, visit P1, and at follow-up for subjects pregnant after screening. Subjects ranked eight items on a 7-point Likert scale to measure overall treatment satisfaction (satisfaction with treatment, flexibility, diabetes understanding, convenience, and willingness to continue treatment and recommend treatment) and perception of hyper- and hypoglycemia. Items were scored on a 0–6 scale then transformed to a 0–100 scale (higher scale = greater treatment satisfaction).

Laboratory analyses (A1C, hematology, biochemistry, and urinalysis) were performed by MDS Pharma Services Central Lab (Hamburg, Germany). A1C was analyzed using a National Glycohemoglobin Standardization Program–certified method (Diabetes Control and Complications Trial standard).

Assuming an incidence of one major hypoglycemic episode during pregnancy with 7 months of insulin treatment (11), 305 subjects were required to be randomized and to complete the study to detect a treatment difference of 40% with a power of 80% (5% significance level). Planned recruitment was 380 pregnant women with 100 enrolled before pregnancy.

Results presented are based on the intention-to-treat (ITT) analysis set (all treated/exposed subjects confirmed pregnant during the study even if they did not complete all study visits, ITT_pregnant n = 322, with 264 completers).

Risk of major maternal hypoglycemia was assessed from its incidence during pregnancy. Episodes were analyzed as recurrent events using a γ frailty model with treatment as covariate. This Cox regression model appropriately handles the recurrent aspects of episodes (30,31). Delayed entry was used for those pregnant at screening, to account for the different observation periods. The number of minor hypoglycemic episodes was analyzed using the same model.

As supportive analyses to the primary safety end point, a noninferiority criterion (<0.4% difference in A1C) was tested at visits P3 and P4 using a linear mixed model, with treatment and pregnancy status at screening as fixed effects and country as a random effect using a one-sided t test with a 2.5% significance level. Average plasma glucose increments (average postprandial values minus preprandial values), average 24-h plasma glucose, and each of the 8-point plasma glucose values at visits P2–P4 were analyzed based on the model described above.

Treatment differences in quality-of-life assessments at the follow-up visit were analyzed using the Wilcoxon rank-sum test. A significance level of 5% was used for statistical analyses, which were generated using SAS version 8.2 (SAS Institute, Cary, NC) on a UNIX platform or S-PLUS version 6.0 Professional for Windows (Microsoft, Seattle, WA).

Subjects were recruited between September 2002 and August 2004; the last follow-up visit was in April 2005. In total, 412 subjects were randomized and treated. Of these, 322 (IAsp, 157; HI, 165) were pregnant during the study (ITT_pregnant cohort). In total, 223 (IAsp, 113; HI, 110) were pregnant at screening, and 99 (IAsp, 44; HI, 55) became pregnant after screening. Withdrawal patterns in pregnant women were similar between treatments. Overall, 190 (IAsp, 102; HI, 88) of those pregnant and 74 (IAsp, 31; HI, 43) of those not pregnant at screening completed pregnancy and the trial intervention. Of 58 noncompleters (IAsp, 24; HI, 34), 31 were withdrawn due to adverse events (IAsp, 14; HI, 17) and 27 for other reasons (IAsp, 10; HI, 17). Age, A1C, BMI, duration of diabetes, and mean daily insulin requirements at baseline were similar between treatment groups (Table 1).

Observed rates of major maternal hypoglycemia were lower with IAsp than HI (Table 2). A 28% lower risk for major hypoglycemia (IAsp/HI; relative risk [RR] 0.720 [95% CI 0.36–1.46]) and a 52% lower risk for major nocturnal hypoglycemia (0.48 [0.20–1.14]) was estimated for the IAsp versus HI groups, respectively, although this did not reach statistical significance. Risks for major daytime (0.85 [0.40–1.78]), all (0.97 [0.66–1.44]), and minor (0.97 [0.66–1.43]) hypoglycemia were similar between treatments (Table 2). The estimated risk of any nocturnal hypoglycemia was 24% lower on IAsp (0.76 [0.57–1.03]). Risk of all daytime hypoglycemia was similar between treatments (1.08 [0.71–1.63]).

Treatment with IAsp was noninferior to treatment with HI as assessed by A1C at the end of the second and third trimesters (mean difference [95% CI]) (IAsp minus HI: −0.04% [−0.18 to 0.11], P = NS; −0.08% [−0.23 to 0.06], P = NS). In both treatment groups, A1C decreased during the first two trimesters then increased toward delivery and follow-up (Fig. 1A). A1C was ≤6.5% for most subjects in both treatment groups during the second and third trimesters (P3: IAsp, 83%; HI, 79%; P4: IAsp, 78%; HI, 73%).

Mean 8-point plasma glucose profile at P2 is shown in Fig. 1B. Overall profiles were similar at P3 and P4 (data not shown), although estimated mean values of 24-h plasma glucose increased from visit P2 to P3, then decreased at visit P4 (IAsp: 6.82, 6.96, and 6.23; HI: 6.82, 7.10, and 6.48 mmol/l, respectively). Postprandial plasma glucose levels were consistently lower with IAsp after breakfast (B90), with statistically significant between-treatment differences at P2 (P = 0.044) and P4 (P = 0.0007) but not at P3 (P = 0.153). Preprandial (prebreakfast, prelunch, and predinner) plasma glucose levels were comparable between treatments at all visits.

Mean prandial plasma glucose increments (mean of difference in pre- and postprandial plasma glucose at breakfast, lunch, and dinner) during pregnancy were lower with IAsp than HI. Between-treatment differences were statistically significant at visits P2 and P4 (IAsp minus HI [in mmol/l] P2: −0.75 [95% CI −1.25 to −0.25], P = 0.003; P4: −0.40 [−0.80 to −0.01], P = 0.044).

Mean total daily insulin doses were similar between treatments. Doses increased during pregnancy and were lower than prepregnancy doses after delivery. At visit P4, total insulin dose (mean ± SD) was 1.08 ± 0.38 and 1.15 ± 0.44 units/kg in the IAsp and HI groups, respectively. Mean daily requirement of bolus insulin during the study was lower in the IAsp group than in the HI group (IAsp: 0.60 ± 0.29; HI: 0.70 ± 0.38 units/kg at visit P4). At visit P4, mean mealtime doses were 0.19 ± 0.10 units/kg with IAsp and 0.23 ± 0.12 units/kg with HI. At visit P4, mean daily dose of NPH was 0.48 ± 0.21 units/kg in the IAsp group compared with 0.45 ± 0.25 units/kg in the HI group. In the third trimester, mean daily insulin doses in the IAsp group were similar between subjects achieving A1C <6.5% and those not meeting target (1.07 vs. 1.12 units/kg). In the HI group, doses tended to be higher in those achieving target (1.16 vs. 1.09 units/kg) due to a higher bolus dose (0.74 vs. 0.60 units/kg). During pregnancy, most (59–75%) pregnant subjects in both treatment groups used at least two basal insulin injections per day. At visit P4, 50% of subjects on IAsp were on two daily injections of NPH and 23.1% were on three or four daily NPH injections. In the NPH group, comparable proportions were 42.7 and 23.1%, respectively.

No maternal deaths were reported. Both insulins were well tolerated, and the adverse event profiles were similar. Most events were mild or moderate and considered unlikely to be related to study products. Eighteen serious adverse events (IAsp, 6; HI, 12) were considered to have a possible relation to study medication (cesarean section [IAsp, 1; HI, 0], abortion [IAsp, 2; HI, 0], hypoglycemic coma [IAsp, 2; HI, 5], investigator-defined inadequate glycemic control [IAsp, 2; HI, 4], and hyperglycemia [IAsp, 0; HI, 3]). Two further events of hypoglycemic coma occurring before and after pregnancy were considered to be possibly related to treatment.

The frequency and profile of obstetric complications were similar between treatments. The most frequent complications were preeclampsia (IAsp, 13; HI, 11), threatened preterm labor (IAsp, 6; HI, 7), prolonged labor (IAsp, 5; HI, 7), and unplanned cesarean section (IAsp, 20; HI, 19).

Thirty-one subjects left the study because of adverse events (IAsp, 14; HI, 17). These were due to fetal loss (induced/spontaneous abortion or stillbirth) in 13 IAsp-treated and 14 HI-treated subjects. Most (IAsp, 79%; HI, 60%) fetal losses were spontaneous, occurring in the first 12 gestational weeks. Other withdrawals were due to hypoglycemia (IAsp, 1) and congenital malformation (HI, 3).

No clinically significant difference in deterioration in fundoscopy was reported in either treatment group. Treatment groups were not different with respect to changes in vital signs, physical examination parameters, electrocardiograms, or clinical laboratory findings.

Comparison of pregnancy outcomes showed no significant between-treatment difference in live births (IAsp, 87.3%; HI, 79.4%), fetal losses (IAsp, 8.9%; HI, 12.1%), and congenital malformations (IAsp, 4.3%; HI, 6.6%). Pregnancy outcome was unknown for 20 subjects in the IAsp group and 6 subjects in the HI group. Additional data are reported in a separate study (M. Hod, P. Damm, R. Kaaja, F. Dunne, I. Demidova, A.-S.P. Mansen, H. Mersebach, unpublished data).

At follow-up, the IAsp group reported a significantly greater (P = 0.031) overall treatment satisfaction (87.6 ± 12.0) than the HI group (83.4 ± 15.3). Between-treatment differences were largely due to more IAsp-treated subjects reporting satisfaction with flexible treatment (Diabetes Treatment Satisfaction Questionnaire scores) (IAsp, 85.9 ± 15.0; HI, 75.8 ± 23.8) and willingness to continue on present treatment (IAsp, 90.1 ± 16.2; HI, 81.9 ± 25.2). No statistical testing was performed on subitems.

This is the largest, randomized, controlled study to date of a rapid-acting insulin analog in pregnant women with type 1 diabetes. Although glycemic control, assessed by A1C, was similar with IAsp and HI, postprandial hyperglycemic excursions were lower with IAsp than with HI, especially after breakfast, with no difference in preprandial glucose control and no increase in major hypoglycemia. Indeed, the observed rate of major episodes was lower for IAsp-treated than HI-treated subjects.

Hypoglycemia, especially nocturnal episodes, is more frequent during pregnancy (16–18), especially during intensified insulin treatment (16,32). In the present study, rates of major hypoglycemia (all and nocturnal episodes) tended to be lower with IAsp than with HI treatment. Similarly, Garg et al. (28) reported relatively few major hypoglycemic episodes in 62 insulin lispro-treated pregnant women with type 1 diabetes. The lack of statistical significance in the risk estimates may be because rates of major hypoglycemia in the present study were lower than expected, reducing its statistical power. Furthermore, the mean duration of exposure during pregnancy was slightly less than the planned 7 months (IAsp, 6.5 months; HI, 6.2 months), and fewer pregnant subjects than planned (264 vs. 305) completed the study.

Postprandial glucose excursions were generally lower with IAsp than with HI treatment, particularly during the first and third trimesters, with significantly lower glycemic excursions after breakfast. Preprandial plasma glucose values were similar between treatment groups. The improvement in postprandial glycemic control with IAsp in this study was similar to that reported in nonpregnant subjects (22,33–36). Lower postprandial glucose levels during pregnancy have been linked to decreased neonatal risks and perinatal complications (37).

Mean A1C levels during pregnancy were not different between treatments, and similar proportions of subjects achieved A1C <6.5% with a trend toward a lower incidence of major hypoglycemia in IAsp-treated subjects. Although recent data suggest that this target should be reevaluated as A1C levels are as low as 4.4–5.6% in the healthy pregnancy (36,38), this has to be balanced against the risk of hypoglycemia.

In the current study a minor deterioration in glycemic control with trend toward increasing A1C in the last trimester occurred in both treatment groups. Total daily insulin doses during this trimester were similar between IAsp-treated subjects achieving and not achieving target A1C <6.5% levels, but HI-treated subjects not meeting target had lower bolus insulin doses, suggesting that they could have benefited from further dose increments.

Throughout pregnancy, total daily insulin doses were similar between treatment groups, although bolus insulin doses were consistently lower for patients receiving IAsp than HI. Despite increases in bolus insulin doses toward the end of pregnancy, dose titration may have been insufficient to maintain or optimize postprandial glycemic control during pregnancy due to changes in insulin sensitivity, body weight, food consumption, and reduced exercise. By the end of the third trimester, doses of IAsp and HI were at their highest, and it was at this point that IAsp was again superior to HI with regard to control of postprandial hyperglycemia. The apparently lower incidence of major hypoglycemia with IAsp may allow more aggressive dose titration late in pregnancy to optimize glycemic control.

The greater treatment satisfaction score seen with IAsp compared with HI has been described previously in trials of rapid-acting analogs and may reflect the differences in the timing of injection relative to eating (34,39). Safety profiles of IAsp and HI were comparable. No maternal deaths were reported and pregnancy outcome was comparable between treatments.

In conclusion, treatment with IAsp resulted in superior postprandial glycemic control to HI with a nonsignificantly lower incidence of major hypoglycemia at comparable levels of A1C, which were mostly ≤6.5%. Maternal safety profiles were similar between treatments, and patients showed greater treatment satisfaction with IAsp. These data suggest that IAsp is at least as safe and effective as HI when used as mealtime insulin in a basal-bolus regimen with NPH insulin in pregnant women with type 1 diabetes and has the potential to offer some clinical benefits in terms of postprandial glucose control.

This trial was sponsored by Novo Nordisk.

We thank the Insulin Aspart Pregnancy Study Group: M.H., Peter Damm, Risto Kaaja, Gerard H.A. Visser, Fidelma Dunne, Irina Demidova, Anne-Sofie Pade-Hansen, Henriette Mersebach, E.R.M., B.K., S.A.A., S.H., D.M., S.D., S.B., A.R., Linda Elise Jensen, Julie Marie Fiona Clyde, Mark White, Anne Romstad, Karin Kanc Hanzel, Julius Anthony Vaz, and Benedikte Jade Lertoft. We also thank all investigators who participated in this trial (online appendix 1 [available at http://dx.doi.org/10.2337/dc06-1887]).