Objective
Type A1 gestational diabetes mellitus (A1GDM), also known as diet-controlled gestational diabetes, is associated with an increase in adverse perinatal outcomes such as macrosomia and Erb palsy. However, it remains unclear when to deliver these women because optimal timing of delivery requires balancing neonatal morbidities from early term delivery against the risk of intrauterine fetal demise (IUFD). We sought to determine the optimal gestational age (GA) for women with A1GDM to deliver.
Study Design
A decision-analytic model was built to compare the outcomes of delivery at 37-41 weeks in a theoretical cohort of 100,000 women with A1GDM. Strategies involving expectant management until a later GA accounted for probabilities of spontaneous delivery, indicated delivery, and IUFD during each week. GA-associated risks of neonatal complications included cerebral palsy, infant death, and Erb palsy. Probabilities were derived from the literature, and total quality-adjusted life years were calculated. Sensitivity analyses were used to investigate the robustness of the baseline assumptions.
Results
Our model showed that induction at 38 weeks maximized quality-adjusted life years. Within our cohort, delivery at 38 weeks would prevent 48 stillbirths but lead to 12 more infant deaths compared to 39 weeks. Sensitivity analysis revealed that 38 weeks remains the optimal timing of delivery until IUFD rates fall <0.3-fold of our baseline assumption, at which point expectant management until 39 weeks is optimal.
Conclusion
By weighing the risks of IUFD against infant deaths and neonatal morbidities from early term delivery, we determined that the ideal GA for women with A1GDM to deliver is 38 weeks.
The prevalence of gestational diabetes mellitus (GDM) in the United States is now at approximately 6-7% of the population. GDM is on the rise in the United States in concert with the obesity epidemic, and this is concerning because pregnancies complicated by GDM have an increased risk of adverse perinatal outcomes. Studies have shown that women with GDM are more prone to preeclampsia, operative deliveries, and subsequent type 2 diabetes mellitus. Furthermore, neonates born to mothers with GDM have an increased incidence of shoulder dystocia, macrosomia, hypoglycemia, hyperbilirubinemia, subsequent obesity, and impairment of glucose tolerance. Consequently, there is a higher prevalence of adverse newborn outcomes such as major neurodevelopmental disabilities, Erb palsy, intrauterine fetal demise (IUFD), and neonatal death.
Women with GDM undergo glycemic management to decrease the rates of these complications. While some women are successfully treated with diet and exercise (type A1 GDM [A1GDM]), others require medical therapy (type A2 GDM). In addition to interventions to achieve normal glucose levels and antenatal testing, women with type A2 GDM are generally delivered by 39 weeks’ gestation. However, women with A1GDM have much less consistent guidance regarding timing of delivery.
Numerous guidelines have been established on when to deliver women with various conditions or complications such as chronic hypertension, oligohydramnios, and placenta previa. However, it remains unclear what the ideal gestational age (GA) is for women with A1GDM to deliver to minimize adverse outcomes for both the mother and the newborn. For example, the most recent recommendations from the Eunice Kennedy Shriver National Institute of Child Health and Human Development and the American Congress of Obstetricians and Gynecologists do not recommend a specific GA other than to discourage delivery <39 weeks’ gestation. Therefore, the goal of our study was to perform a decision analysis balancing the tradeoffs of delivering at various GAs at term to determine the optimal GA for women with A1GDM to deliver.
Materials and Methods
A decision-analytic model was created using TreeAge software (TreeAge Pro 2013; TreeAge Software, Inc., Williamstown, MA) to compare the outcomes of planning to deliver at 37-41 weeks in a theoretical cohort of 100,000 women with A1GDM ( Figure 1 ). Strategies involving expectant management until a later GA accounted for probabilities of spontaneous delivery, indicated delivery, and IUFD during each successive week. GA-associated risks of neonatal complications included cerebral palsy, infant death, IUFD, and Erb palsy. Maternal outcomes in the model included maternal death and mode of delivery. Probabilities were derived from the literature, and total quality-adjusted life years (QALYs) were calculated using both utilities from the maternal and neonatal perspective from the literature. Utilities are measures of quality of life in various health states that range from 0 for death to 1 for optimal health. For baseline reference in this model, the maternal utility for an uncomplicated vaginal delivery was set at 1. Sensitivity analyses were used to investigate the robustness of the baseline assumptions.
Probabilities
All probability inputs in the model were derived from the literature ( Table 1 ). The baseline probabilities for cesarean deliveries with expectant management and cesarean deliveries after induction at various GAs from 37-41 weeks were derived from a 2006 retrospective cohort study comparing the outcomes of women who were induced and those who were expectantly managed. Baseline probabilities for maternal deaths from cesarean and vaginal deliveries were derived from a 2003 case-control study on pregnancy-related deaths. Regarding neonatal outcomes, the GA-specific probabilities for macrosomia (defined as a birthweight of >4000 g) were derived from a 2008 retrospective cohort study on perinatal outcomes in low-risk term pregnancies. Since women with GDM are more likely to have cesarean deliveries and macrosomic infants, these were each accounted for with an odds ratio of 2.2 from a 2009 retrospective cohort study on perinatal mortality among women with A1GDM. In our model, macrosomia was considered as an intermediate outcome in the decision-analytic model–one that would impact downstream outcomes such as cesarean deliveries and brachial plexus injuries. Therefore, we did not consider macrosomia alone as an outcome as it did not impact maternal or neonatal utilities, and thus would not affect calculation of QALYs. Since literature for A1GDM-specific women was limited, probabilities for infant deaths (defined as infants who die within 1 year of birth) in women with GDM were derived from a 2012 retrospective cohort study on the risks of stillbirth and infant death in women with GDM. IUFD rates were calculated using values from the literature and then extrapolated to women with A1GDM with an odds ratio of 1.08 from a 1997 retrospective cohort study comparing the outcomes of women with GDM to those of the general obstetric population. Cerebral palsy probabilities and Erb palsy probabilities were both derived from the literature.
| Outcome | Probabilities | Utilities (maternal, neonatal perspective) | References |
|---|---|---|---|
| Maternal outcomes | |||
| Cesarean delivery in women with GDM | 0.996, 1 | ||
| Expectant management (wks of GA: 37, 38, 39, 40) | 0.252, 0.252, 0.280, 0.436 | ||
| After induction (wks of GA: 37, 38, 39, 40, 41) | 0.229, 0.229, 0.269, 0.361, 0.414 | ||
| Maternal deaths | 0 | ||
| Cesarean | 0.000359 | ||
| Vaginal | 0.000092 | ||
| Neonatal outcomes | |||
| Macrosomia (birthweight >4000 g) in women with GDM (wks of GA: 37, 38, 39, 40, 41) | 0.0425, 0.095, 0.158, 0.241, 0.344 | ||
| Shoulder dystocia | |||
| In macrosomic infants | 0.06 | ||
| In nonmacrosomic infants | 0.009 | ||
| Infant deaths from women with GDM (wks of GA: 37, 38, 39, 40, 41) | 0.00140, 0.00106, 0.00087, 0.00095, 0.00115 | 0.92, 0 | |
| IUFD in women with GDM (wks of GA: 37, 38, 39, 40) | 0.000462, 0.000529, 0.000562, 0.000632 | 0.92, 0 | |
| Cerebral palsy (wks of GA: 37, 38, 39, 40, 41) | 0.0023, 0.0012, 0.0009, 0.0010, 0.0010 | 0.733, 0.612 | |
| Erb palsy in setting of shoulder dystocia | 0.78, 0.87 | ||
| In macrosomic infants | 0.0612 | ||
| In nonmacrosomic infants | 0.0295 |
Utilities
Utilities from the maternal and neonatal perspective were all derived from the literature ( Table 1 ). The maternal utility for a vaginal delivery was set at 1, while the utility of a cesarean delivery was set at 0.996 based on literature concerning maternal preferences in mode of delivery. Maternal death utility was assumed to be 0 from the maternal perspective. The utility of maternal death from the neonatal perspective was not accounted for because of controversy over how to determine this value. The utility for cerebral palsy from the maternal perspective was derived from a 2002 study on maternal preferences. The maternal utilities for IUFD and infant death were derived from a 2000 cross-sectional study of 534 patients in California. There were no maternal or neonatal utility values found for Erb palsy, so we assumed Erb palsy to be similar to mild cerebral palsy. Utility values were applied over the lifetime of either the mother (life expectancy 56.75 years, assuming delivery at age 25 years) or the infant (life expectancy 78.5 years). When the infant had cerebral palsy the life expectancy for the infant was changed to 66.562 years.
Analysis
Baseline analysis used literature-derived probabilities of adverse perinatal maternal and neonatal outcomes at each gestational week to compare women planning to deliver with A1GDM at 37, 38, 39, 40, and 41 weeks’ gestation. QALYs for each management strategy were then calculated.
Sensitivity analysis was used to evaluate the robustness of this model. With sensitivity analysis, we were able to vary variables individually to determine thresholds in which the optimal strategy would differ from our original result.
Monte Carlo analysis was also performed to further evaluate the robustness of this model. In this analysis, distributions of the probabilities, rather than the means of the probabilities, were used to run a theoretical cohort of 100,000 women through the model. This sort of analysis provides us with proportions of when each outcome was optimal. By sampling distributions we incorporated the uncertainty underlying the model inputs to produce confidence estimates regarding the outcomes examined.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
