“Changing Patterns in Medication Prescription for Gestational Diabetes During a Time of Guideline Change in the USA: A Cross-sectional Study”

Abstract

OBJECTIVE:

To define patterns of prescription and factors associated with choice of pharmacotherapy for gestational diabetes mellitus (GDM), namely metformin, glyburide, and insulin, during a time period of evolving professional guidelines.

DESING:

Cross-sectional study.

SETTING:

U.S. commercial insurance beneficiaries from Market-Scan (late 2015 to 2018)

STUDY DESIGN:

We included women with GDM, singleton gestations, 15–51 years of age on pharmacotherapy. The exposure was pharmacy claims for metformin, glyburide, and insulin.

MAIN OUTCOMES:

Pharmacotherapy for GDM with either oral agent, metformin or glyburide, compared to insulin as the reference, and secondarily, consequent treatment modification (addition and/or change) to metformin, glyburide, or insulin.

RESULTS:

Among 37,762 women with GDM, we analyzed data from 10,407 (28%) with pharmacotherapy, 21% with metformin (n=2,147), 48% glyburide (n=4,984), and 31% insulin (n=3,276). From late 2015 to 2018, metformin use increased from 17% to 29%, as did insulin use from 26% to 44%, while glyburide use decreased from 58% to 27%. By 2018, insulin was the most common pharmacotherapy for GDM; and metformin was more likely to be prescribed by 9% compared to late 2015–2016, while glyburide was less likely by 45%. Treatment modification occurred in 20% of women prescribed metformin compared to 2% with insulin, and 8% with glyburide.

CONCLUSIONS:

Insulin followed by metformin has replaced glyburide as the most common pharmacotherapy for GDM among a privately insured U.S. population during a time period of evolving professional guidelines. Further evaluation of the relative effectiveness and safety of metformin compared with insulin is needed.

TWEETABLE ABSTRACT:

Insulin followed by metformin has replaced glyburide as the most common pharmacotherapy for gestational diabetes mellitus in the United States.

INTRODUCTION

The prevalence of gestational diabetes mellitus (GDM) has continued to increase over the last two decades affecting as many as 1 in 5 U.S. pregnant women.^{1, 2} Nearly 1 in 3 women with GDM fail to achieve euglycemia with diet and exercise alone and require pharmacotherapy.^3–5 Compared with injectable insulin, the oral agents metformin and glyburide have been commonly prescribed because of their efficacy to achieve maternal glycemic control and prevent short-term adverse neonatal outcomes, such as macrosomia, large for gestational age at birth, and neonatal hypoglycemia,^{6, 7} as well as their lower cost,⁸ convenience, and ease of administration.⁴ Metformin is a biguanide that lowers basal and postprandial glucose,⁹ and is the most common drug used to treat type 2 diabetes outside of pregnancy.¹⁰ Glyburide is a second-generation sulphonylurea, and enhances insulin secretion.¹¹

Since 2008, clinical trials have demonstrated comparable efficacy of metformin and glyburide relative to insulin for the treatment of GDM.^12–14 Meta-analyses have confirmed that glyburide may be equivalent and metformin superior to insulin,^{6, 7, 15, 16} though recent data have suggested that glyburide may be inferior, including a higher risk of large for gestational age and hypoglycemia.^{13, 17} Despite concerns about demonstrated placental transfer from mother to fetus of both oral agents and lack of adequate long-term safety data,¹⁸ a recent meta-analysis did not definitively find that fetal exposure to metformin increased childhood obesity and metabolic dysfunction compared with insulin.¹⁹

Recommendations regarding the choice of pharmacotherapy for GDM have rapidly evolved between 2016 to 2018, the time period of the current analysis, in response to clinical trials,¹³ observational data,¹⁷ and meta-analyses.^{6, 7} These professional society guidelines generally reinforce insulin as the first-line option with an increasing preference for metformin over glyburide when oral agents are considered.^{18, 20} The American Diabetes Association (January 2016) recommends insulin as the preferred first-line regimen, but states that oral agents, including metformin and glyburide, may be preferable to insulin for some women with GDM due to cost, comprehension, or cultural influences.²¹ The American College of Obstetricians and Gynecologists (July 2017, February 2018) recommends insulin over metformin, stating that metformin is a second-line option for women who either decline insulin injections or whose obstetric providers deem unable to safely administer insulin.²² In contrast, the Society for Maternal Fetal Medicine (May 2018) states that both insulin and metformin are reasonable first-line options.²³

Data estimating medication choice for GDM in the U.S. are primarily from 2000 to 2015, when glyburide use had overtaken insulin.^{17, 24} But data primarily from Europe through 2016 suggest that insulin is the most common agent and metformin use has rapidly increased.²⁵ And after initiation of pharmacotherapy, understanding later treatment modification, particularly for metformin, remains poorly understood.²⁶ We hypothesized that in response to recent professional guidelines,^{18, 20} glyburide use would decrease, while metformin and insulin use would increase.

In the current study, we examined current patterns and factors associated with choice of pharmacotherapy for GDM, namely metformin, glyburide, and insulin during a time period of evolving professional society guidelines.

METHODS

Study setting:

We conducted a cross-sectional study among commercially insured U.S. women with GDM in the IBM Watson Health’s MarketScan® Commercial Claims and Encounters database from October 1, 2015 to December 31, 2018. Briefly, MarketScan® is the largest database of administrative health claims from patients with employer-sponsored health insurance in the U.S.; it captures patient-level data on inpatient, outpatient, and prescription drug claims from approximately 100 large employers and health plans.^{27, 28} Our methods are consistent with previous analyses conducted by our group using the MarketScan® database to describe medications use in pregnancy,²⁹ including for GDM.²⁴ The start date of this study is consistent with the transition to International Classification of Diseases, 10^th Revision (ICD-10) in the U.S., as well as in this database. This analysis using a de-identified dataset was deemed exempt for human subject research review by the Office of Responsible Research Practices at The Ohio State University. Patients were not involved in the development of the research.

Participants:

We identified pregnant women with a delivery code for a liveborn singleton infant through ICD-10-CM delivery codes as well as Current Procedural Terminology (CPT) codes (Table S1), where the date for the earliest claim was defined as the delivery date. To identify women with GDM who were treated pharmacologically, we used outpatient pharmacy-dispensing claims to identify prescriptions of oral agents (glyburide and metformin) and insulin. Using National Drug Codes, we classified women as being prescribed metformin, glyburide, or insulin, based on the drug class of their earliest claim prior to delivery date, which was defined as initiation of pharmacotherapy (i.e., “index date”).

Consistent with an earlier analysis by our group,²⁴ we included women with the following criteria: 1) continuous enrollment without interruption in insurance coverage for healthcare and pharmacy claims during the year prior to and 3 months after the delivery date, 2) diagnosis code for GDM (ICD-10 codes: O24, O24.4, 024.4X, O24.9X, P70.0) up to 1 year or 365 days prior to delivery, AND 3) earliest pharmacy claim for metformin, glyburide, or insulin within 150 days prior to delivery (as women may have been prescribed these agents for GDM in a prior pregnancy or pregestational diabetes or metformin for another clinical indication), hereafter referred to as the “index date” (Figure 1 and Supplemental Table S2). This cohort was restricted to the first eligible pregnancy with GDM for a given woman during the study time period.

Figure 1.

Flowchart of women with gestational diabetes initiating pharmacotherapy with glyburide, metformin, and insulin, late 2015–2018

As shown in Figure 1, we excluded women if they were <15 years or > 51 years at the time of delivery, had a multiple gestation, and had any diagnosis codes for pre-gestational diabetes (i.e., type 1 or 2 diabetes) prior to the delivery date.²⁴ After preliminarily identifying women with GDM treated pharmacologically, we then excluded women who were prescribed glyburide or insulin for >150 days duration before pregnancy onset for >300 days before the delivery date suggesting a diagnosis of pregestational diabetes rather than treatment for prior GDM in a previous pregnancy; were prescribed glyburide or insulin between the index date and 365 days prior to the delivery date suggesting diagnosis of pregestational diabetes; were prescribed any of the following antidiabetic medications used for pregestational diabetes anytime earlier than the delivery date (pramlintide, acarbose, miglitol, alogliptin, linagliptin, sitagliptin, albiglutide, dulaglutide, exenatide, liraglutide, semaglutide, nateglinide, repaglinide, dapagliflozin, canagliflozin, empagliflozin, ertugliflozin, glimepiride, gliclazide, glipizide, chlorpropamide, tolazamide, tolbutamide, rosiglitazone, pioglitazone) (Table S2); or were prescribed thiazolidinediones, including rosiglitazone, pioglitazone and pioglitazone, for GDM. Because metformin is used off-label for infertility and to reduce the risk of miscarriage, we did not exclude women with prior use of metformin early in pregnancy, defined as a prescription claim in early pregnancy between the index date to one year or 365 days before the delivery date.³⁰

Exposures, outcomes, and covariates:

Maternal demographic characteristics available using insurance claims data included age at delivery, delivery year (late 2015–2016, 2017, and 2018), number of other dependent children ≤18 years old covered by the same insurance policy on the delivery date (0, 1, >1) an approximate measure of parity, which was not available using insurance claims data, and U.S. geographic region (Northeast, South, Midwest, West).²⁸ We identified the following maternal comorbid endocrine conditions using ICD-10-CM diagnosis or CPT-4 codes assessed before the index date, for which women could have had one or more diagnoses or treatments: infertility diagnosis or treatment (at least one claim for clomiphene, urofollitropin, follitropin, menotropin, ganirelix, cetrorelix), obesity, hypothyroidism, hyperandrogenism (inclusive of alopecia, hirsutism, or acne), metabolic syndrome, and polycystic ovarian syndrome (Table S3). We also identified women with metformin use early in pregnancy as described above.

The primary outcome was initial treatment for GDM with either oral agent, metformin or glyburide, based on the index date compared to insulin as the reference; and secondarily, later treatment modification from the initial therapy after the index date but before the delivery date. This included either addition of a new drug class (glyburide, metformin or insulin) onto the initial base regimen, or changing the base regimen to a new drug class (glyburide, metformin or insulin). While few women modified their regimen more than once, we assessed the pattern of their initial or first treatment modification. We defined and analyzed these outcomes consistent with clinical guidelines for GDM pharmacotherapy in which initial treatment was with one agent sequentially followed by treatment modification (which could include either addition or substitution). This study did not include a core outcome set.

Statistical analysis:

Prevalence in the use of glyburide, metformin, and insulin during the study period was estimated using as the denominator the total number of women with GDM treated pharmacologically with these three agents. We determined the association between baseline patient characteristics (i.e., age, region, endocrine comorbidities, calendar time) and initial GDM pharmacotherapy with metformin versus insulin and glyburide versus insulin. We then determined the risk of consequent treatment modification to one of the other three agents for metformin versus insulin and glyburide versus insulin. Finally, because factors associated with treatment modification for oral agents remain poorly understood, we determined the association between the above patient characteristics and treatment modification for those initially treated with metformin and glyburide versus insulin, respectively. Binomial regression was used to calculate unadjusted and adjusted prevalence ratios (when the outcome was the frequency of initial GDM pharmacotherapy) and risk ratios (when the outcome was the likelihood of treatment modification), with 95% confidence intervals (CI). A prevalence ratio is analogous to a risk ratio of a cohort study, and is preferable when the period for being at risk of developing the outcome extends over a period of time (months to years).³¹ All of the above models adjusted for the following covariates, calendar year (late 2015–2016, 2017, 2018), U.S. census region (Northwest, Midwest, South, West, unknown), maternal age (continuous), number of dependent children (0, 1, >1), and an endocrine comorbidity as listed above (yes/no for each modeled separately); the treatment modification model also adjusted for metformin use in early pregnancy. Given this was an exploratory and not causal study, covariates were selected for inclusion both a priori as well as based on statistical significance (p<0.10). In sensitivity analysis, we re-performed the primary analysis excluding women who were prescribed metformin early in pregnancy (N=959). All analyses were conducted using SAS 9.4 (SAS Institute, Cary, NC).

RESULTS

Out of 1,179,677 women with a delivery code, 471,729 had continuous enrollment without interruption in insurance coverage during pregnancy, of which 456,505 were singleton gestations between 15–51 years of age. A total of 37,762/456,505 (8%) women initially met our eligibility criteria for GDM, and after further exclusion of women with likely pre-gestational diabetes, 10,407 (28%) were treated with pharmacotherapy (Figure 1). The current analysis is restricted to these 10,407 women with GDM, of whom 21% (n=2,147) initially had a pharmacy claim for metformin, 48% for glyburide (n=4,984), and 31% for insulin (n=3,276).

Among women treated for GDM with pharmacotherapy, the mean age was 34 years (standard deviation, SD: 4.8) (Table 1 as column percentages and Table S4 as row parentages by pharmacotherapy type, respectively). A total of 42% of women delivered in late 2015 (after 10/01) and 2016, 33% in 2017, and 25% in 2018. Half (51%) of women who were prescribed glyburide were in the Midwest, as were 43% of those prescribed metformin. Both metformin and glyburide were more commonly prescribed compared with insulin across the U.S., except for the Northeast (13% and 11% vs. 22%). Most women had a diagnosis code for an endocrine comorbidity, and the frequency was higher for those prescribed metformin (66%) and insulin (66%), and less so with glyburide (60%). The most common endocrine conditions (not mutually exclusive) were obesity (47%), hypothyroidism (14%), polycystic ovarian syndrome (11%), and infertility (8%). Nearly a tenth of women had been prescribed metformin early in pregnancy (i.e., before the index date), which was more frequent among those later prescribed metformin for GDM (20%) compared to glyburide and insulin (6% and 7%).

Table 1.

Characteristics of women with GDM initiating pharmacotherapy with glyburide, metformin, and insulin, late 2015–2018

Characteristics (column percentage)	Total	Glyburide	Metformin	Insulin
	N=10,407	N=4,984	N=2,147	N=3,276
	N (%)	N (%)	N (%)	N (%)
U.S. census region 1
Northeast	1,564 (15.0)	567 (11.4)	285 (13.3)	712 (21.7)
South	2,192 (21.1)	987 (19.8)	452 (21.1)	753 (23.0)
Midwest	4,494 (43.2)	2,532 (50.8)	917 (42.7)	1,045 (31.9)
West	2,139 (20.6)	892 (17.9)	488 (22.7)	759 (23.2)
Age group, years
15–24	402 (3.9)	206 (4.1)	101 (4.7)	95 (2.9)
25–29	1,608 (15.5)	817 (16.4)	377 (17.6)	414 (12.6)
30–34	3,889 (37.4)	1,867 (37.5)	792 (36.9)	1,230 (37.5)
35–39	3,414 (32.8)	1,622 (32.5)	657 (30.6)	1,135 (34.6)
40 or greater	1,094 (10.5)	472 (9.5)	220 (10.2)	402 (12.3)
Mean age, years (SD)	33.6 (4.8)	33.3 (4.8)	33.2 (4.9)	34.1 (4.7)
Number of dependent children
>1	2,246 (21.6)	1,143 (22.9)	428 (19.9)	675 (20.6)
1	3,579 (34.4)	1,754 (35.2)	710 (33.1)	1,115 (34.0)
0	4,582 (44.0)	2,087 (41.9)	1,009 (47.0)	1,486 (45.4)
Endocrine comorbidity 4
Infertility treatment	779 (7.5)	281 (5.6)	180 (8.4)	318 (9.7)
Hypothyroidism	1,503 (14.4)	634 (12.7)	347 (16.2)	522 (15.9)
Polycystic ovarian syndrome	1,098 (10.6)	409 (8.2)	358 (16.7)	331 (10.1)
Hyperprolactinemia	66 (0.6)	28 (0.6)	11 (0.5)	27 (0.8)
Hyperandrogenism	908 (8.7)	404 (8.1)	215 (10.0)	289 (8.8)
Metabolic syndrome	266 (2.6)	95 (1.9)	95 (4.4)	76 (2.3)
Obesity	4,898 (47.1)	2,316 (46.5)	992 (46.2)	1,590 (48.5)
No comorbidity	3,847 (37.0)	1,986 (39.8)	734 (34.2)	1,127 (34.4)
Metformin use early in pregnancy 2	959 (9.2)	300 (6.0)	421 (19.6)	238 (7.3)

¹Missing data: U.S. region (n=34)

²Metformin use early in pregnancy was defined as a prescription claim in early pregnancy (>150 days to 365 days before delivery).

³Abbreviations: standard deviation (SD).

⁴Endocrine comorbidities are not mutually exclusive (i.e., each woman could have more than one).

From late 2015 to 2018, metformin use increased from 17% to 29%, as did insulin from 26% to 44%, while glyburide declined from 58% to 27% (Figure 2). By the end of the study period, insulin followed by metformin were the most commonly prescribed agents for GDM. By 2018, women were more likely to be prescribed metformin compared to insulin by 9% compared to late 2015–2016 (adjusted prevalence ratio: 1.09; 95% CI: 1.01 to 1.18) (Table 2). And by 2018, glyburide was less likely to be prescribed compared to insulin by 45% compared to late 2015–2016 (adjusted prevalence ratio: 0.55; 95% CI: 0.52 to 0.59).

Figure 2.

Frequency of pharmacotherapy with glyburide, metformin, and insulin for gestational diabetes by year, 2015–2018

N=4,437 for 2015-2016, 3,437 for 2017, and 2,623 for 2018. The numeric frequencies are provided in Appendix Table S7.

Table 2.

Characteristics of women with gestational diabetes with prescription for metformin compared to insulin, and glyburide compared to insulin

	Metformin compared to insulin		Glyburide compared to insulin
	Unadjusted risk ratio (95% CI)	Adjusted risk ratio (95% CI)1	Unadjusted risk ratio (95% CI)	Adjusted risk ratio (95% CI)1
Calendar year of delivery
Late 2015/2016	1.00	1.00	1.00	1.00
2017	1.00 (0.92, 1.09)	1.06 (0.98, 1.15)	0.91 (0.88, 0.95)*	0.92 (0.89, 0.96)*
2018	1.01 (0.94, 1.10)	1.09 (1.01, 1.18)*	0.53 (0.50, 0.57)*	0.55 (0.52, 0.59)*
U.S. region
Northeast	1.00	1.00	1.00	1.00
Midwest	1.25 (1.14, 1.38)*	1.18 (1.07, 1.30)*	1.19 (1.14, 1.24)*	1.15 (1.10, 1.21)*
South	1.57 (1.44, 1.71)*	1.46 (1.34, 1.59)*	1.52 (1.45, 1.59)*	1.47 (1.41, 1.54)*
West	1.30 (1.19, 1.43)*	1.26 (1.14, 1.38)*	1.15 (1.09, 1.20)*	1.12 (1.07, 1.17)*
Age group, years
15–24	1.00	1.00	1.00	1.00
25–29	0.91 (0.75, 1.10)	0.90 (0.74, 1.08)	0.96 (0.86, 1.08)	0.98 (0.88, 1.08)
30–34	0.74 (0.62, 0.88)*	0.74 (0.62, 0.88)*	0.87 (0.78, 0.96)*	0.90 (0.81, 0.99)*
35–39	0.69 (0.57, 0.83)*	0.70 (0.58, 0.84)*	0.84 (0.76, 0.94)*	0.89 (0.80, 0.98)*
40 or greater	0.65 (0.52, 0.81)*	0.70 (0.57, 0.88)*	0.76 (0.66, 0.87)*	0.84 (0.74, 0.95)*
Number of dependent children
0	1.00	1.00	1.00	1.00
1	0.98 (0.91, 1.05)	1.01 (0.94, 1.09)	1.05 (1.01, 1.09)*	1.04 (1.00, 1.08)
>1	0.98 (0.90, 1.06)	1.03 (0.95, 1.13)	1.07 (1.03, 1.12)*	1.07 (1.02, 1.11)*
Endocrine comorbidity
No comorbidity	1.00	1.00	1.00	1.00
Infertility treatment	0.89 (0.78, 1.00)	0.86 (0.75, 0.98)*	0.74 (0.67, 0.81)*	0.93 (0.86, 1.00)
Hypothyroidism	1.01 (0.92, 1.12)	0.98 (0.89, 1.09)	0.87 (0.82, 0.93)*	0.96 (0.90, 1.02)
Polycystic ovarian syndrome	1.34 (1.22, 1.47)*	1.00 (0.88, 1.14)	0.87 (0.80, 0.94)*	1.03 (0.96, 1.11)
Hyperprolactinemia	0.76 (0.46, 1.28)	0.67 (0.39, 1.16)	0.73 (0.51, 1.03)	0.95 (0.74, 1.23)
Hyperandrogenism	1.08 (0.94, 1.24)	1.05 (0.92, 1.21)	0.83 (0.74, 0.92)*	0.98 (0.90, 1.07)
Metabolic syndrome	1.58 (1.34, 1.87)*	1.26 (1.01, 1.58)*	0.86 (0.69, 1.08)	0.94 (0.77, 1.15)
Obesity	0.93 (0.87, 1.00)	0.90 (0.84, 0.97)*	0.92 (0.88, 0.96)*	0.97 (0.93, 1.00)
Metformin use early in pregnancy 2	1.76 (1.65, 1.89)*	1.74 (1.61, 1.88)*	0.92 (0.85, 0.99)*	0.92 (0.85, 1.00)

¹Adjusted for calendar year, U.S. region, maternal age, number of dependent children, and endocrine comorbidity.

²Metformin use early in pregnancy was defined as a prescription claim in early pregnancy (>150 days to 365 days before delivery).

^*p-value <0.05.

Compared with insulin, women who were prescribed metformin were more likely to reside in the Midwest, South, and West, be <30 years of age, and have a diagnosis of metabolic syndrome, and less likely to have a diagnosis of obesity or infertility (Table 2). Women who had early pregnancy metformin use were more likely to be later prescribed metformin rather than insulin for GDM.

Similar to women prescribed metformin, those prescribed glyburide over insulin were more likely to reside in the Midwest, South, and West and be <30 years of age. They were also more likely to have dependent children (>1). Comorbid endocrine conditions were not associated with glyburide prescription in adjusted analyses. The above associations with metformin and glyburide remained after excluding women who were prescribed metformin early in pregnancy (Table S5).

With regards to later treatment modification before delivery, 20% (n=423) of women initially treated with metformin for GDM had a later medication switch or addition, most frequently insulin (14%) followed by glyburide (5%) (Table 3). Among women initially treated with glyburide, 8% (n=402) had a modification, and 6% with insulin and 2% with metformin. And among those initiating insulin, 2% (n=67) had a modification, and 2% with metformin and 1% with glyburide. Women initiating metformin were ten times more likely to have a treatment modification compared to those on insulin (adjusted risk ratio: 11.67; 95% CI: 9.50 to 14.30), and women initiating glyburide were four times more likely to have a treatment modification compared to those on insulin (adjusted risk ratio: 4.40; 95% CI: 3.44 to 5.66).

Table 3.

Frequency and association between initial pharmacotherapy for GDM and treatment modification

	Frequency of substitution (row percentage) n (%)				Association between initial pharmacotherapy and modification
	Overall	Glyburide	Metformin	Insulin	Metformin compared to insulin	Glyburide compared to insulin
Initial pharmacotherapy					Adjusted risk ratio (95% CI)	Adjusted risk ratio (95% CI) 1
Glyburide (N=4,984)	402 (8.1)	--	82 (1.6)	320 (6.4)	--	4.40 (3.44, 5.66)*
Metformin (N=2,147)	423 (19.7)	117 (5.4)	--	306 (14.3)	11.67 (9.50, 14.30)*	--
Insulin (N=3,276)	67 (2.1)	19 (0.6)	48 (1.5)	--	1.00	1.00

¹Adjusted for calendar year, U.S. region, maternal age, number of dependent children, and endocrine comorbidity, and metformin use in early pregnancy.

^*p-value <0.05.

²Treatment modification may include either addition or replacement of the prior regimen.

Among women treated with metformin, factors associated with a higher likelihood of treatment modification included later year of delivery (2018 and 2017), and generally increasing maternal age at delivery (Table S6). Among women treated with glyburide, later year of delivery and an obesity diagnosis was associated with a higher likelihood of treatment modification, and less so increasing maternal age.

DISCUSSION

Main Findings:

In a commercially insured U.S. cohort from late 2015 through 2018, we found an increase in the use of insulin as well as metformin and a decrease in the use of glyburide for women with GDM requiring pharmacotherapy.

The time period of the current study was concurrent with rapidly evolving professional guidelines from the American Diabetes Association,²¹ American College of Obstetricians and Gynecologists,²² Society for Maternal Fetal Medicine,²³ and expert opinion,^{18, 20} as well as results from meta-analyses,^{6, 7, 16} observational data,¹⁷ and a randomized controlled trial of glyburide versus insulin,^{13, 32} which generally concluded that insulin and metformin were preferable to glyburide. Our findings are consistent with professional guidelines that recommend insulin as the first-line pharmacotherapy over oral agents, and then when oral agents are considered due to patient preference, metformin as the agent of choice.

Strengths and limitations:

There are several study limitations to note. First, similar to prior studies assessing pharmacotherapy for GDM using health claims data,^{17, 24} we could not assess other factors that may influence GDM management and choice of pharmacotherapy, including glucose tolerance test results, glycemic patterns, gestational age, body mass index, and race or ethnicity. This study focused on choice of pharmacotherapy and associated patient characteristics, and not perinatal outcomes. Future analyses will need to carefully analyze such outcome data accounting for confounding by indication (i.e., pharmacotherapy choice) with appropriate analytical techniques, such as propensity score matching. Second, it is likely some maternal characteristics were underreported, for example obesity, by using ICD-10-CM diagnosis codes rather than having access to body mass index. We suspect that variable misclassification was likely non-differential in relation to pharmacotherapy choice, and likely biased our findings toward the null. Third, our definition of GDM relied on diagnostic codes as well as excluding women with pharmacy claims for some of the drugs of interest (glyburide and insulin) early in pregnancy as well as pre-pregnancy claims for anti-diabetic medications to yield a cohort of women with “true” GDM.³³ In the case of metformin, it is possible that some women who were prescribed it before or early in pregnancy for an endocrine diagnosis may have not stopped taking it and may have continued metformin following a diagnosis of GDM. Fourth, similar to any analysis done using large insurance claims data, our estimates reflect medications dispensed by an outpatient pharmacy and not whether the medication was actually taken. Additionally, we could not assess medication dosage, particularly in the setting of treatment modification. Fifth, consistent with prior analyses using the MarketScan^® extract,^{24, 29} this cohort was limited to women with continuous enrollment through pregnancy, and hence selection bias is a possibility. Despite careful exclusion of women with likely pre-gestational diabetes based on diagnosis codes as well as history and pattern of prior pharmacotherapy, it is possible some women classified as GDM may have had pregestational diabetes. This may have increased the proportion of women prescribed insulin (both initially and later modification) as insulin is the first-line therapy for pregestational diabetes. Finally, our analysis was limited to a population with employer-provided insurance, and results may not be generalizable to other populations of pregnant women with GDM, including those covered by Medicaid or who are uninsured. The frequency of GDM in the current commercially insured population was lower than national estimates, which likely reflects the higher frequency of GDM among women of lower socio-economic status receiving public insurance. While prior population-based U.S. samples suggest that most women with GDM have commercial insurance,³⁴ the prevalence of GDM has increased the most among women with public insurance,³⁵ and this analysis will need to be conducted in these patient populations.

Strengths of this study include an analysis of pharmacotherapy for GDM across a sample of U.S. women during an approximately three-year period of evolving professional guidelines. Using commercial claims data, this study included a large sample size of women with GDM initiating pharmacotherapy. Given this analysis included women with more significant GDM who were treated with pharmacotherapy, the likelihood of under-ascertainment of GDM was likely lower than prior analyses using claims data in which the frequency of under-ascertainment was higher with less severe cases not requiring pharmacotherapy.³⁶ Recent validation studies support that ICD-10 codes accurately identify women with GDM.³⁷ Additionally, by restricting our cohort to women who were continuously enrolled in the year before delivery, we were able to approximate timing of medication use during pregnancy. This study also identified patterns of treatment modification, particularly when initially started on metformin, and further data are needed to better understand this subpopulation.

Interpretation:

Our results are different from a previous analysis from the MarketScan^® database from 2000–2011.²⁴ At that time, glyburide had overtaken insulin as the primary agent for GDM pharmacotherapy, comprising over two-thirds of prescriptions after 2007. Notably that analysis did not include metformin prescription. A more recent international collaborative using both health registries and insurance claims data found that through 2016, insulin was the most common agent, with increasing metformin use.²⁵ During the time period of the current analysis, a French randomized controlled trial comparing glyburide to insulin found that glyburide was not non-inferior to insulin for the prevention of composite neonatal adverse outcome.^{13, 32} Additionally, two large meta-analyses suggested that while metformin may have favorable neonatal outcomes relative to insulin, particularly with regard to reducing fetal overgrowth, glyburide was inferior to insulin.^{6, 7} Our results demonstrate that metformin use has increased significantly, and this underscores the need to better understand the impact of known placental transfer and potential fetal programming from metformin exposure on long-term outcomes among exposed children.^{26, 38, 39}

In the current study, we noted regional variations in prescription patterns, in which insulin was favored in the Northeast, and oral agents, particularly glyburide, in the Midwest. We found that women with an infertility or obesity diagnosis code were less likely to be treated with metformin, and the opposite with metabolic syndrome. This may reflect prior use of metformin for treating these conditions earlier or possible interaction between these conditions and GDM with metformin.⁴⁰ Women with an obesity diagnosis code were more likely to receive insulin over metformin, which may reflect an increased concern for pregestational diabetes with obesity. We did not identify an association between glyburide and endocrine comorbidities.²⁴

We found that nearly 20% of women with GDM who started on metformin later required a treatment modification, compared with less than half as many who started on glyburide. Most of these modifications, inclusive of either a change or addition, were with insulin. Few women who started insulin required either oral agent. It is possible that a reason many women who later required treatment modification in the metformin group was because of a high proportion of this subgroup had been on metformin in early pregnancy. Our results are consistent with data that suggest >20% of women with GDM who initiate metformin require supplemental insulin,⁶ and the rate of modification in the current analysis was less than 50% in the now over decade old Metformin in Gestational Diabetes (MiG) Trial.¹² In the recent French trial of glyburide versus insulin, nearly 20% of women on glyburide required additional insulin.¹³ There remains a need to better understand patterns and factors associated with treatment modifications,⁴¹ particularly with metformin, and the contribution of higher fasting glucose levels.⁴²

As a result of the uptake of metformin and insulin, robust evaluation of metformin compared with insulin in regards to efficacy and long-term safety among exposed children is warranted to inform treatment decisions for women with GDM.

CONCLUSIONS

In conclusion, insulin followed by metformin has replaced glyburide as the most common pharmacotherapy for GDM. These findings stand in contrast to 10 years ago when glyburide was the most common choice.²⁴ These results reflect the impact of changing professional guidelines during this time period that have recommended insulin and to a lesser extent metformin over glyburide.