Original Article

Glyburide Compared to Insulin for the Treatment of Gestational Diabetes Mellitus: A Cost Analysis

Department of Obstetrics and Gynecology, Baylor College of Medicine, Houston, TX, USA

Correspondence to: Laura Goetzl, MD, MPH, Department of Obstetrics and Gynecology, Baylor College of Medicine, 6550 Fannin Street, Suite 801, Houston, TX 77030, USA

Abstract

OBJECTIVE: To compare the costs associated with glyburide compared to insulin for the treatment of gestational diabetes unresponsive to dietary therapy.

STUDY DESIGN: A cost model was designed. The model excluded costs that were identical for both treatment arms, such as the cost of monitoring glucose control. Insulin treatment costs included average wholesale drug costs, wholesale delivery costs (syringes, alcohol pads), and costs of office staff educating patients. Glyburide costs were based on average wholesale drug costs. Downstream costs of potential inpatient evaluation for hypoglycemia were included in the model.

RESULTS: In our baseline model, glyburide was significantly less costly than insulin for the treatment of gestational diabetes. The average cost saving per patient based on wholesale drug costs and hospital costs was US$165.84. Actual retail drug savings and hospital charge savings are potentially considerably greater. The strongest determinant of cost savings was medication cost. The model was less sensitive to the one-time costs of inpatient treatment and patient education.

CONCLUSION: Glyburide is less costly than insulin for the treatment of gestational diabetes. Cost models can be useful to physicians deciding between two equally efficacious medications, allowing them to incorporate information about their individual practice styles with a complex balance of cost implications. Journal of Perinatology (2002) 22, 403-406 doi:10.1038/sj.jp.7210759

INTRODUCTION

Gestational diabetes is the most common medical complication of pregnancy, affecting up to 3% of pregnant women. Insulin is the traditional treatment for gestational diabetes mellitus unresponsive to dietary interventions^1,2 among physicians who choose medical treatment. Until recently, oral therapy with sulfonylurea drugs had been contraindicated due to concerns regarding teratogenicity^3,4 and the possibility of neonatal hypoglycemia.⁵ However, a recent report by Langer et al.⁶ demonstrates the effectiveness and safety of glyburide in the treatment of gestational diabetes. Treatment with glyburide resulted in similar glycemic control with significantly less risk of maternal hypoglycemia. It is the underlying assumption of this manuscript, based on reports from the literature, that glyburide and insulin provide similar glycemic control when both insulin and glyburide are used as reported.⁶ This assumption precludes a cost-effectiveness analysis, a type of analysis that is more commonly seen in the literature. Therefore, in our analysis, we specifically do not address issues of effectiveness. Instead, we performed a cost analysis to compare the economic impact of the cost of the two regimens combined with downstream medical costs.

MATERIALS AND METHODS

A cost model was designed to compare the costs of treatment between insulin and glyburide therapy in the setting of gestational diabetes (Data 3.5, TreeAge Software, Williamstown, MA). Costs that were expected to be identical between the treatment arms, such as the cost of monitoring glucose control, were not included in the model. As the literature suggests that patients treated with insulin are significantly more likely to experience hypoglycemia,⁶ we included hypoglycemia in our model to determine the potential increased cost associated with this outcome. The basic model structure is illustrated in Figure 1. The variables used in the model are defined in Tables 1 and 2, along with references for the sources of our baseline values and ranges. All costs are reported in year 2000 US dollars. Wholesale costs and hospital costs were used to avoid regional differences in charges, a standard practice in cost analysis.

The cost of insulin treatment included the costs of the medication itself as well as the cost of the materials needed to administer the insulin. Further, we included the costs incurred in teaching patients how to draw up the insulin dose and how to self-inject, based on a 15-minute session with a licensed vocational nurse or medical assistant. We modeled this cost based on the average hourly rate at our institution, not including fringe benefits. For the sensitivity analysis, we modeled instructional sessions between 6 minutes and 1hour. The cost of insulin was based on the average wholesale cost per unit of insulin multiplied by the weekly dose of insulin expected to obtain glycemic control. In addition, the wholesale cost of two insulin syringes and two alcohol pads per day were included. To estimate the range of cost, we calculated the cost of insulin dosed at the upper and lower 95% confidence intervals of insulin doses reported in the literature.⁶

The cost of glyburide was based on the average wholesale cost per milligram of glyburide multiplied by the weekly dose of glyburide expected to obtain glycemic control. The lowest cost was based on the lowest wholesale cost in conjunction with a glyburide regimen that represented the lower 95% confidence interval of glyburide dose reported in the literature.⁶ The highest cost was based on the highest wholesale cost in conjunction with a glyburide regimen at the maximum dose of 20 mg/day. The cost of glyburide failure necessitating subsequent insulin treatment was modeled based on a 5-week trial of glyburide as described in Langer et al. Patients switching to insulin incurred the educational costs of an insulin new start.

The cost of inpatient evaluation for hypoglycemia was based on the charges at Saint Luke's Episcopal Hospital. Included in this cost isa 6-hour evaluation on our labor and delivery unit followed by a 24- to 48-hour admission. We assumed that the initial period of evaluation would include a nonstress test, serial blood glucose measurements, serum electrolytes and urinalysis. These charges were then adjusted with a cost-to-charge ratio based on the hospital's internal cost estimates.

It was assumed that all patients could achieve glycemic control on adequate doses of insulin. The point estimates for the probability of glycemic control on glyburide and the probability of hypoglycemia on each regimen were taken from the literature.⁶ The ranges represent the 95% confidence intervals for each rate based on sample size. The probability of inpatient treatment for hypoglycemia in the setting of gestational diabetes was not available in the literature. Further, the decision to admit a patient for hypoglycemia represents a practice pattern and will vary by physician. Therefore, expert opinion was substituted regarding a wide reasonable range of potential values as noted. The effect of these assumptions was then rigorously examined in subsequent sensitivity analysis.

RESULTS

In our baseline model, glyburide was significantly less costly than insulin for the treatment of gestational diabetes. The average cost saving per patient was US$165.84. The strongest determinant of this cost saving was medication costs. The model was less sensitive to the one-time costs of inpatient treatment for hypoglycemia and patient education.

Rigorous sensitivity analyses were performed to examine the impact of estimates from the literature and model assumptions based on expert opinion. All of the model variables were tested and the most relevant results presented. One-way sensitivity analysis of the cost of a week of glyburide reveals that glyburide is cost-saving when compared to insulin over a broad range of values (Figure 2). A threshold cost (where insulin becomes less costly) of US$18.24 per week of glyburide therapy is only achieved by assuming the highest range of wholesale cost for glyburide in conjunction with an average daily dose of 18.9 mg.

Physician practice has a strong impact on the relative cost savings of glyburide. While glyburide is less costly than insulin for all durations of treatment, physicians who diagnose gestational diabetes earlier in pregnancy will recognize a greater cost savings (Figure 3). Similarly, while glyburide is less costly than insulin regardless of the rate of hospitalization for hypoglycemia (0-10%), more conservative physicians who favor an inpatient evaluation following hypoglycemia will save more with glyburide (Figure 4).

DISCUSSION

Most physicians consider costs in the health care decisions that they make. Cost becomes a more important factor when physicians select among medications that are similar in their effectiveness, such as two antibiotics that offer identical antimicrobial coverage. However, often, physicians cannot easily factor in all elements of cost. Cost analyses are useful to physicians in that they can incorporate not only initial drug cost, but also other secondary (downstream) costs that are often not considered. In addition, the sensitivity analyses presented in this manuscript enable physicians to visualize how different parameters affect costs in their individual practice.

Our model suggests that glyburide is considerable less costly than insulin for the treatment of gestational diabetes that is unresponsive to dietary intervention. We estimate that the cost saving per pregnant women treated with glyburide in lieu of insulin is US$165.84. While at face value this figure is not impressive, it is important to note that due to standard cost analysis techniques, this figure represents costs, not charges. The actual saving in retail medication expenditures and hospital charges is potentially considerably higher and will vary regionally. In addition, we chose to minimize the cost of the education required for a new insulin start by calculating these costs based on 15-minute time for a physician assistant, without additional physician time. In practices where the teaching process is more costly, the savings with glyburide are higher. For example, when the educational costs are set to US$40.00, the saving with glyburide increases to US$200.47.

While our assumptions are based on the best available information, precise estimates are not always possible. Because of the inherent inaccuracies of estimating cost, we have performed extensive sensitivity analyses to examine the effect of less-than-perfect information. This technique allows us to examine the outcome of our cost model over a reasonable range of alternative values for a given variable. For example, if the model is sensitive to a given variable, then insulin becomes less costly than glyburide at a “threshold value.” If the model is insensitive to a given variable, glyburide remains less costly than insulin over the full range of reasonable assumptions.

We performed one-way sensitivity analyses for each variable in our model. The majority of the variables did not have a marked effect on cost. The sensitivity analysis of three variables ¾ the cost of glyburide, the probability of hospitalization for hypoglycemia and the weeks of treatment ¾ are presented. While glyburide remained cost-saving regardless of weeks of treatment, the saving was less for shorter lengths of treatment. Given that the actual drug cost of glyburide is less than the cost of insulin, it is not surprising that each week of treatment results in an incremental cost saving. Therefore, physicians who screen for gestational diabetes early in the third trimester and treat aggressively realize greater cost savings with glyburide compared to insulin. Glyburide will still be cost-saving for physicians who screen later in the third trimester and treat for a shorter period of time; however, the cost savings will be of a lesser degree.

When the wholesale cost of glyburide was examined, glyburide is less costly than insulin until the wholesale cost of a week of glyburide exceeds US$18.24. This cost corresponds to the highest wholesale drug cost, in conjunction with a mean daily dose of 18.9 mg. However, a dose of 18.9 mg is more than 1 1/2 SD higher than the mean dose of glyburide that has been reported for adequate glycemic control.⁶ For glyburide to be more costly than insulin, patients requiring 18.9 mg of glyburide must be controlled on mean daily dose of 85 U of insulin (baseline model assumption). The possibility remains, however, that individual patients who are resistant to glyburide, but controlled on average doses of insulin, are more cost-effectively treated with insulin. For the clinician, it would be difficult to know a priori which patients fall into this group unless an individual patient has had prior experience with both insulin and glyburide. Without a priori knowledge, the average patient will be treated at lower cost with glyburide.

Finally, our model cannot address some of the additional benefits of glyburide over insulin. These would include patient preference for oral medication over parenteral, with the attendant potential for increased compliance. In addition, the dosing of glyburide is more standardized than that of insulin, potentially increasing the comfort of the generalist in treating gestational diabetics who are unresponsive to dietary intervention. Lastly, the increased rate of hypoglycemic events with insulin has other potential health effects, unrelated to cost, which may make glyburide preferable.

Overall, glyburide is a less costly alternative agent than insulin for the treatment of gestational diabetes. Cost models can be useful to physicians deciding between two equally efficacious medications, allowing them to incorporate information about their individual practice styles with a complex balance of cost implications.

Acknowledgements

This material is the result of work supported with resources and the use of facilities at the Houston Center for Quality of Care and Utilization Studies, Department of Veterans Affairs Health Services Research and Development Center of Excellence, Houston, TX. Special thanks to Glenn Ruggles at the St. Luke's Episcopal Hospital Financial Accounting and Reporting Department for his assistance in providing hospital cost data.

References

1 Gestational Diabetes. ACOG Practice Bulletin No. 30. American College of Obstetricians and Gynecologists. Obstet Gynecol 2001; 98: 525–38.

2 American Diabetes Association. Gestational diabetes mellitus. Diabetes Care 1998; 21: (Suppl 1) S60–1.

3 Picaquadio K, Hollingsworth DR, Murphy H. Effects of in-utero exposure to oral hypoglycaemic drugs. Lancet 1991; 338: 866–9. MEDLINE

4 Schaefer-Graf UM, Buchanan TA, Xiang A, Songster G, Montoro M, Kjos SL. Patterns of congenital anomalies and relationship to initial maternal fasting glucose levels in pregnancies complicated by type 2 and gestational diabetes. Am J Obstet Gynecol 2000; 182: 313–20. MEDLINE

5 Zucker P, Simon G. Prolonged symptomatic neonatal hypoglycemia associated with maternal chlorpropamide therapy. Pediatrics 1968; 42: 237–45.

6 Langer O, Conway DL, Berkus MD, Xenakis EM-J, Gonzales O. A comparison of glyburide and insulin in women with gestational diabetes mellitus. N Engl J Med 2000; 343: 1134–8. MEDLINE

7 2000 Drug Topic Red Book Medical Economics, 2000.

Figures

Figure 1 Cost model. Legend: (□) decision node; (○) chance node; (▹) terminal node. pprobglycontrolglyburide=probability of glycemic control on glyburide therapy; pprobhypoinsulin=probability of hypoglycemia on glyburide therapy; pprobhypoglyburide=probability of hypoglycemia on insulin therapy; pprobinpatientRx=probability of hospitalization for hypoglycemia; #=1 minus probability from paired arm.

Figure 2 Sensitivity analysis on cost of glyburide medication. Legend: (•) insulin treatment; (♦) glyburide treatment. Threshold value=US$18.24.

Figure 3 Sensitivity analysis on treatment duration. Legend: (•) insulin treatment; (♦) glyburide treatment.

Figure 4 Sensitivity analysis on probability of inpatient treatment for hypoglycemia. Legend: (•) insulin treatment; (♦) glyburide treatment.

Tables

Table 1 Definition of Variables

Table 2 Definitions of Cost Variables

July/August 2002, Volume 22, Number 5, Pages 403-406

Table of contents    Previous  Article  Next    [PDF]