This decision analytic model showed that treating GDM was cost-effective as long as the cost to treat GDM was less than $3555 or at the baseline cost to treat GDM of $1786, when the efficacy of treatment met at least 49% of its expected goal. Above a cost of $3555 or below to treat GDM, an efficacy of 49% at baseline costs ($1786), treating GDM was no longer cost-effective in terms of reducing maternal and neonatal adverse outcomes, including preeclampsia, shoulder dystocia, maternal death, macrosomia, permanent and transient brachial plexus injury, neonatal hypoglycemia, neonatal hyperbilirubinemia, and NICU admissions.
Macrosomia, shoulder dystocia, and brachial plexus injury are among the well-described adverse maternal and neonatal outcomes associated with GDM.1
Another perspective on the benefits of treating mild GDM is a number-needed-to-treat (NNT) analysis to calculate the number of women who would need to be treated to decrease the incidence of a complication by 1. NNT analysis of this model showed an NNT of 12 for macrosomia, an NNT of 75 for shoulder dystocia, and an NNT of 320 for transient brachial plexus injury. Furthermore, only 14 women would need to be treated to reduce 1 case of a cesarean section.
The cost-effectiveness results are similar to a cost-consequence analysis of the ACHOIS trial, performed in Australia, which showed that treating mild GDM was more expensive vs routine care but cost-effective at $2186/QALY, based on an incremental cost to treat GDM of $247/patient (converted from 2002 Australian dollars to 2009 US dollars).24
Just as the costs to treat GDM in the United States are an order of magnitude higher than in Australia ($1786 in the United States vs $247 in Australia), the cost per QALY is also similarly higher ($20,412 in the United States vs $2186 in Australia).
Several costs were not included in this model because the clinical trials did not measure the outcomes associated with those costs or the data were not available. First, although antenatal admissions can add to the overall cost of managing women with GDM, they were not included in the costs because this outcome was not measured, and it was unclear how treating GDM would affect this parameter. For example, the cost-consequence analysis of the ACHOIS trial showed that the difference in antenatal admissions between the treatment and routine-care groups was not significant (health services use for antenatal inpatient admissions: 135 for treatment vs 133 for routine-care, adjusted treatment effect 1.11 (95% confidence interval, 0.91–1.36; P
Thus, including antenatal admissions is unlikely to significantly affect the cost-effectiveness analysis. Similarly, emergency room visits were not taken into account because again the trials did not measure this outcome, and it is unclear how treating GDM would affect this factor.
Second, indirect costs from GDM such as time off from work were not included because these data were not available. Again, the cost-consequence analysis of the ACHOIS trial surveyed 108 of the 1000 women who participated in the RCT to determine the mean charges to women and their family from randomization to birth. These charges included paid child care, travel to/from appointments, food substitution, time off from work for both the mother and partner, and blood glucose monitoring equipment and consumables. There was no difference between the intervention and routine care group (Australian dollars 2002: $367 for the intervention group vs $302 for the routine-care group, P
= .34). If the cost of the blood glucose monitoring equipment and consumables was removed because they were already accounted for in the cost to treat GDM, then the difference was even smaller ($314 for the intervention group vs $285 for the routine-care group because some in the routine-care group were diagnosed with GDM after randomization).24
Third, although the costs of the various complications associated with GDM were included in this model, the indirect costs associated with these complications, for example, the economic losses from a child with a permanent brachial plexus injury, were not included in this analysis. These costs are likely to be higher in the nontreatment group; thus, this analysis errs on the conservative side on this point.
One limitation of this model is the lack of recent data on the incremental cost to treat GDM (costs above routine care costs) in the United States. Although there are more recent studies that report this cost in other countries that range from $247 to $458,24,25
the most recent study in the United States is from a review published in 2000 in which the incremental costs ranged from $1786 to $3352.16
This analysis showed, however, that as long as the cost to treat GDM was below $3555, a value above the upper range of these cost estimates, treating GDM was cost-effective. More recent data are likely to show that the cost to treat GDM is lower than these 2000 estimates because GDM has become more prevalent and treatment practices more standardized, leading to more efficiently administered perinatal clinics that could absorb the additional costs over a larger patient population, thus lowering the per-patient cost.
In addition, the costs associated with various GDM outcomes represented in this analysis are taken from a wide range of published studies. These studies use varying methodologies to determine said costs and as such may not be valid when combined into a unifying analysis. However, univariate sensitivity analyses were performed over a wide range on all costs, and only the cost to treat GDM crossed the cost-effectiveness threshold.
Another limitation is that the probabilities of the effect of GDM treatment on maternal and neonatal outcomes come from 1 trial. However, this was a well-powered, multicenter, RCT.6
Furthermore, the results of this trial were very similar to the results ACHOIS trial.5
Because the threshold analysis showed the model to be robust over a wide range of probabilities, costs, and utilities, this illustrates that there is a wide margin in which treatment is cost-effective as long as it decreases maternal and/or neonatal outcomes.
Finally, a decision analytic model has its inherent limits in simulating reality. For example, some neonatal outcomes are modeled as discrete, mutually exclusive outcomes; in other words, in this model a neonate cannot have both hypoglycemia and hyperbilirubinemia, even though both outcomes can occur in reality. Adding permutations for every possible combination of outcomes quickly makes decision analytic models unwieldy. However variables that were known to have strong associations, such as the relationship between macrosomia, shoulder dystocia, and brachial plexus injury, were taken into account in this model. For example, a neonate could have shoulder dystocia both in the presence and absence of macrosomia; similarly, shoulder dystocia and macrosomia were both factors that affected the probability brachial plexus injury. Moreover, we were unable to model the long-term downstream effects on the offspring from treatment of GDM. Within increasing evidence from the field of fetal programming, it does appear that there are effects from maternal diet and hyperglycemia that may lead to obesity and type 2 diabetes mellitus in the offspring. However, the incorporation of these results would only have made our findings more robust.
Whereas treating mild GDM is more expensive than not treating, it is also more effective with cost-effectiveness within usual cost-effective thresholds. Because the costs of treating GDM are expected to decrease with increased efficiency in management, technology advances, and lower cost of supplies and pharmaceuticals, it is likely to be both less expensive and more effective in the future. These results in combination with the results of the Hyperglycemia and Adverse Pregnancy Outcome study, which showed a continuous relationship between maternal glucose levels below what is considered overt GDM and various outcomes including birthweight above the 90th percentile and primary cesarean delivery,27
suggest that lowering the thresholds for diagnosing GDM may be cost-effective. Such policy efforts are currently underway,28
and more clinical research is warranted to clearly identify specific cut-off values.
Based on this analysis, the expected value of perfect information29
of such research is expected to be high because there is a small difference in cost-effectiveness between treating and not treating, there is some uncertainty about cost-effective estimates, the consequences of GDM can be serious, and the disease is increasing in prevalence.30