Trends and associated maternal characteristics of antidiabetic medication use among pregnant women in South Korea

The prevalence of diabetes during pregnancy and the need for the treatment are increasing. We aimed to investigate antidiabetic medications (ADM) use among pregnant women and their characteristics. Using Korea’s nationwide healthcare database, we included women aged 15–49 years with births during 2004–2013. The prevalence and secular trend of ADM use were assessed in 3 periods: pre-conception period, first trimester, and second/third trimesters. To compare maternal characteristics between pregnancies with and without ADM prescription, we used the χ2 or Fisher’s exact test and Cochran-Armitage trend test. The prescription patterns analyzed by calendar year, age, insurance type, income, area, and medical institution. Of 81,559 pregnancies, 222 (0.27%) and 305 (0.37%) were exposed ADM during pre-conception and pregnancy periods, respectively. ADM prescriptions increased significantly by an 11.3-fold in second/third trimesters, while a 2.9-fold in first trimester. ADM use is more prevalent in women aged older and living in urban areas. Metformin was most used in the pre-conception period, while insulins were most during pregnancy. About 0.4% of women received ADM during pregnancy; a rate was lower than that in western countries. Non-recommended medications were more common in first trimester, which warrants pregnancy screening for women taking ADM.

). Combination products were also included in this study, with each drug's active ingredient considered individually. Among the aforementioned antidiabetic medications, our analysis focused on the three most commonly utilized drug classes (insulins, metformin, and sulfonylureas) and the remaining drug classes were categorized as other oral agents.
Exposure assessment was conducted in following periods: pre-conception period (the year before pregnancy; > 273 days), first trimester (273-184 days), and second or third trimesters (183-8 days before the delivery date) ( Figure S1). Medications prescribed within seven days before the delivery date were excluded from analysis as their prescription patterns were likely to have changed for delivery purposes. Study subjects were considered exposed to the drugs of interest if they received at least one prescription of an antidiabetic medication(s) during each of the above-mentioned periods. All medication prescription records from outpatient settings were identified from the NHIS-NSC database.
Identification of maternal characteristics. The baseline characteristics of the pregnancies were determined and stratified by whether the women were prescribed antidiabetic medications during pregnancy. Maternal age at delivery, insurance type (health insurance and medical aid), income level (low-income [0-3 deciles], middle-income [4-7 deciles] and high-income [8-10 deciles] based on income level deciles) and area (urban and rural) were extracted from the database. Nulliparity and multifetal gestation was determined by the procedure codes of delivery. The presence of chronic hypertension (HTN) (ICD-10 codes: I10-I15 and O10-O11), gestational HTN (O13), preeclampsia-eclampsia (O14-O15), pregestational DM (E10-E14 and O240-O243), GDM (O244), polycystic ovary syndrome (PCOS) (E282), and female infertility (including amenorrhea or irregular menstruation) (N91, N97) was identified by the records of appropriate the ICD-10 code diagnosed during pre- www.nature.com/scientificreports/ conception and pregnancy periods. The number of physician visits for any reason during pre-conception period was determined as a proxy for health care utilization. To understand the potential underlying causes for each class of antidiabetic prescriptions, we confirmed diagnosis codes for DM, GDM, and PCOS or female infertility using ICD-10 codes, which are corresponding to each antidiabetic prescription.
Statistical analysis. We calculated absolute standardized differences (aSD) to estimate the size of the difference in baseline characteristics between pregnancies with and without antidiabetic prescriptions during pregnancy 27 . We defined a significant difference between the two groups as aSD greater than 0.1. Descriptive statistics were presented as the numbers and prevalence of pregnancies among women prescribed antidiabetic medications in per-conception period, first trimester, and second or third trimester, separately. The prevalence was calculated as the number of pregnancies in women prescribed any antidiabetic medication, where the denominator was the total number of pregnancies with live births. The prevalence of antidiabetic medication use for each drug class was stratified by calendar year, maternal age at delivery, insurance type, income level, area and medical institution type. The medical institution types were classified according to the number of beds: primary (0-29 beds), secondary (30-99 beds), and tertiary (≥ 100 beds). To compare the maternal characteristics between pregnancies with and without prescriptions of each antidiabetic medication class, the χ2 test or Fisher's exact test was used for variables of insurance type and area, and the Cochran-Armitage trend test was used for variables of the calendar year, maternal age, and income level. p values of < 0.05 were significant. We investigated secular trends in annual drug utilization in the pre-conception period, first trimester, and second or third trimesters. The data analysis of this study was performed using SAS® software, version 9.4 (© 2002-2012 by SAS Institute Inc., Cary, NC, USA).

Results
Among the 81,559 pregnancies, 222 (0.27%) and 305 (0.37%) pregnancies were prescribed any antidiabetic medication(s) during the pre-conception and pregnancy periods, respectively ( Fig. 1). A higher proportion of pregnancies in women with antidiabetic prescriptions were to those in the 35-39 years and 40-44 years age group, compared to those without antidiabetic prescribing (30.5% vs. 15.5% at 35-39 years and 6.6% vs. 2.0% at 40-44 years) (Table 1). Moreover, a higher proportion was showed in medical aid beneficiaries (1.6% of antidiabetic exposed vs. 0.3% of their counterparts; aSD = 0.139), those living in urban areas (77.7% vs.70.4%; aSD = 0.168), multifetal gestation (4.6% vs.1.4%; aSD = 0.192). Pregnancies with antidiabetic prescriptions also had more comorbid conditions and physician visits in pre-conception period (12 or more) (46.6% vs 24.9%; aSD = 0.469) compared to their counterparts. The secular patterns of antidiabetic prescriptions among pregnancies from 2004 to 2013 revealed an overall increasing trend in all periods, especially in the second or third trimesters (Fig. 2). A sharp increase was observed in the second or third trimesters, with insulins most commonly used, indicating that the prescriptions of insulins almost coincided with those of total antidiabetic medications.
Of 222 pregnancies who were prescribed any antidiabetic medication(s) in the pre-conception period, metformin (0.17%) was most commonly used, followed by other oral agents (0.08%), and sulfonylureas (0.07%) ( Table 2). The prescription of any antidiabetics increased with maternal age (p < 0.001). Insulins were prescribed at a higher proportion in tertiary care (35.3%) than in primary (14.2%) or secondary (13.7%) care. Among women who used antidiabetic medications during the pre-conception period, 56.8% had DM only, 11.3% had DM with other condition(s) of interest, and 29.3% had PCOS or infertility (Table S2).
Any antidiabetic medication(s) was prescribed to women in the first trimester for 111 pregnancies (0.14%) and in the second or third trimesters for 254 pregnancies (0.31%) (Tables 3 and 4). The most commonly used medication in the first trimester was insulins (0.06%), followed by metformin (0.05%) and sulfonylureas (0.03%), while that in the second or third trimesters was insulins (0.29%). In the first trimester, antidiabetic prescriptions increased from 0.07% in 2004 to 0.20% in 2013, a 2.9-fold increase (p = 0.001), and from 0.06 to 0.68% in the second or third trimesters, an 11.3-fold increase (p < 0.001).
For both first and second or third trimesters, antidiabetic prescriptions increased with maternal age (p < 0.001). Especially, compared to that in the pre-conception period, antidiabetic medication use in the second or third trimesters was doubled in women aged 35-39 and 40-44 years of age (0.34% to 0.62% and 0.49% to 1.03%, respectively), while there was little change in women aged 20-34 years. In both first and second or third trimesters, medical aid beneficiaries used more antidiabetic medications than health insurance subscribers (0.70% vs. 0.13% and 1.05% vs. 0.31%, respectively), but it was not statistically significant (p = 0.058 and p = 0.060, respectively). In both first and second or third trimesters, pregnant women living in urban areas used more antidiabetic medications than those living in rural areas (p = 0.024, and p = 0.008, respectively).
Among women who were prescribed antidiabetic medications in the first trimester, 76.6% had DM only, 17.1% had PCOS or infertility, and 3.6% had DM with other condition(s) of interest. Among women who used antidiabetic medications in the second or third trimesters, 59.4% had DM only and 36.6% had GDM only (Table S2).

Discussion
Of the 81,559 pregnancies, approximately 0.4% of women received antidiabetic medication(s) during pregnancy between 2004 and 2013. Overall, the antidiabetic medication exposures declined from the pre-conception period (0.27%) to the first trimester (0.14%) but increased in the second or third trimesters (0.31%). The number of women receiving antidiabetic medications during pregnancy, especially in the second or third trimester, increased considerably over the ten-year study period. Antidiabetic prescriptions are more prevalent in women who were older and those living in urban areas, with statistical significance. In the pre-conception period, metformin www.nature.com/scientificreports/ (0.17%) was most used, while insulins were most used (0.06% and 0.29%, respectively) in the first and second or third trimesters. Pregnancies in women prescribed oral antidiabetic agents other than metformin or sulfonylurea were about 0.03% of the total pregnancies in the first trimester and 0.01% in the second or third trimesters. The prevalence of antidiabetic prescriptions during pregnancy was lower than that in western countries, although careful interpretation is required as their study period, study population, or data source varies among each study. In the US, 3.24% of pregnancies were exposed to antidiabetic medications in second or third trimesters during 2001-2007 28 . In Europe, 2.0% of pregnant women prescribed antidiabetic mediations in the year before, during, or the year following pregnancy between 2004 and 2010 29 . In a recent study, prevalence of any antidiabetic medication use was 3% from under 2% (Denmark, Norway, and Sweden) to above 5% (Australia and US) between 2006 and 2016 30 .
The utilization of antidiabetic medications rose significantly by an 11.3-fold increase in the second and third trimesters, while the increase was more gradual in the first trimester, resulting in a 2.9-fold increase from 2004 to 2013. The increase in antidiabetic medication use might be explained by changes in maternal characteristics. For example, the maternal age at delivery is increasing in Korea, from an average age of first delivery of 27.6 years in 1993 to 31.3 years in 2010 31 . Maternal age is an important risk factor of GDM 32 , and in our study, the prevalence of antidiabetic medication prescriptions increased with maternal age at delivery, an increase more noticeable in the second or third trimesters, in which GDM treatment is initiated. Moreover, the multifetal birth rate also is increasing in Korea, from 1.13% in 1993 to 2.76% in 2010 31 . Multifetal pregnancies were considered a risk factor of GDM 33,34 . We observed that antidiabetic prescriptions in the second or third trimesters increased rapidly in 2011. In Korea, GDM screening has been performed in a two-step approach using Carpenter-Coustan or NDDG criteria; however, in 2011, the KDA additionally adopted a one-step approach using the IADPSG criteria, which might result in an increasing prevalence of GDM 11,35 .
Overall, the antidiabetic medication exposures declined from the pre-conception period (0.27%) to the first trimester (0.14%) but increased in the second or third trimesters (0.31%). The decreased medication use in the first trimester may have been because the first trimester is a crucial period for placental development   www.nature.com/scientificreports/ and organogenesis. As all pregnant women are recommended to undergo GDM screening during the second trimester, it is natural that GDM treatment increased accordingly in the second or third trimesters 36 . In the preconception period, metformin use was most prevalent (0.17%). The prevalent use of metformin before pregnancy might be explained by its clinical use. Metformin is the preferred initial pharmacologic agent for type 2 DM and is also used for clinical management of PCOS or infertility as it can induce ovulation 37,38 . A study from the US showed that only 13% of women prescribed metformin in the 4 months before pregnancy were diagnosed with DM and 67% were diagnosed with PCOS or infertility 28 . This finding is different from our results that those diagnosed with DM (42.8%) were more than those diagnosed with PCOS or infertility (30.7%) among women receiving metformin before pregnancy.
In our study, we could not find a significant relationship between socioeconomic deprivation and antidiabetic medication use during pregnancy due to small sample size of medical aid beneficiaries. Previous studies have shown that the association between socioeconomic status and odds of antidiabetic prescriptions or risk of GDM varies for each country. In a large cohort study in Ontario, Canada, the risk of GDM increased among participants with a lower income 39 . Similarly, in a large Scottish cohort, there was an inverse relationship between socioeconomic status and the risk of GDM 40 . However, in a cohort study from the UK, no association was observed between socioeconomic deprivation and the rate of GDM 41 . Chinese 42 and Saudi Arabian 43 studies also showed no association between income and GDM. Meanwhile, antidiabetic medication use was more prevalent Table 1. Comparison of characteristics between pregnancies with and without antidiabetic prescriptions during pregnancy, 2004-2013 (total pregnancies = 81,559). aSD, absolute standardized difference; Rx, medical prescriptions; PCOS, polycystic ovary syndrome. *Absolute standardized difference, whereby value > 0.10 indicates significant difference between antidiabetic medication users and non-users during pregnancy. www.nature.com/scientificreports/ in pregnant women living in urban areas than that in those living in rural areas, which is similar with the results of previous study 39 . Women living in urban areas had a higher chance of receiving perinatal examinations and were, thus, GDM was more likely to be detected. Regarding the use of antidiabetic medications before and during pregnancy, the prescription rate and the most prevalent antidiabetic medication in Korea differ from those in the US 28 . The prescription rate in the US was much higher than that in Korea (1.21% vs. 0.27% in the pre-conception period, 1.35% vs. 0.14% in the first trimester, and 3.24% vs. 0.31% in second or third trimesters). In the pre-conception period, the medications used most in the US were metformin (0.84%) and insulins (0.33%), while those in Korea were metformin (0.17%) and other oral agents (0.08%). In the second or third trimesters, more than 90% of Korean women prescribed antidiabetic medications were administered insulins, while insulins (2.45%) and sulfonylureas (0.83%) were most commonly administered in the US.
Metformin was most commonly used in the pre-conception period (0.17%), while insulins were the most prevalent in the first and second or third trimesters (0.06% and 0.29%, respectively). As insulin is the preferred medication for the treatment of hyperglycemia during pregnancy, the utilization of all oral antidiabetic agents decreased as pregnancy progressed, in line with an increase in the prescription of insulin. Nevertheless, oral antidiabetic agents were commonly used and oral agents other than metformin, whose safety have not been proven, were used during pregnancy. Of note, the use of glyburide in the first trimester may be a public health concern, as several studies showed that glyburide increase the risks of neonatal intensive care unit (NICU) admissions, macrosomia, large for gestational age and pre-eclampsia in the mother 44 . Prescriptions of these nonrecommended agents may have occurred either when healthcare providers or pregnant women were unaware of their pregnancy status. Hence, it is both necessary to confirm the evidence of drug safety and to improve the system, including alerts for the prescription or dispensing of drugs not recommended to pregnant women or checking pregnancy status beforehand.
This study has several strengths. Firstly, we described the drug utilization using a nationally representative NHIS-NSC database based on systematic stratified random sampling. Secondly, this is the study to explore antidiabetic medication use among pregnant women, which included all kinds of antidiabetic medications in Korea. However, this study also had some limitations. First, the medication history from claim data should not be interpreted as the actual use due to the potential for patient non-compliance. Second, drug exposures in study periods can be misclassified due to a lack of data on gestational age. Hence, we used a previously validated algorithm to estimate the gestational age at birth 26,45 . Third, we included only pregnancies with live births, which can result in an under-estimation of the prevalence of antidiabetic prescriptions by excluding the cases with drug-induced stillbirths or abortions. Fourth, as the data used in this study was only available as of 2013, we could not assess the latest pattern of antidiabetic medication use during pregnancy but we investigated the 10-year longitudinal trends of antidiabetic prescriptions among pregnant women.
In summary, about 0.4% of women in South Korea received antidiabetic medications during pregnancy between 2004 and 2013; a rate was substantially lower than those in western countries. Antidiabetic medications use during pregnancy rose dramatically, especially in the second and third trimesters. Oral antidiabetic agents which were not recommended to pregnant women were more commonly prescribed in the first trimester than in the second or third trimesters. Considering the possible teratogenicity of the medications during early pregnancy, this finding warrants the necessity of pregnancy screening for women of childbearing age who are taking antidiabetic medications. Since the safety information of many antidiabetic medications is unclear, these medications should be used carefully in this population until evidence of their safety is established.  www.nature.com/scientificreports/

Ethical approval
This study was approved by the Institutional Review Board of Sungkyunkwan University (SKKU 2018-03-014).
As this study are based on claim data and do not contact with patients, no informed consent was required from the board. All research was conducted in accordance with guidelines and regulations of the institutional and national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Table 2. Utilization of antidiabetic medications in the pre-conception period by drug class* and maternal characteristics, 2004-2013 (N = 81,559). *When a pregnant woman was prescribed two or more drug's active ingredient, each ingredient was separated individually. † All percentages are row percentage with pregnancies corresponding to each row. ‡ The p-value s denote comparison between pregnancies with and without prescriptions of each drug class. The χ 2 test or Fisher's exact test was used for variables of insurance type and area, and the Cochran-Armitage trend test was used for variables of the calendar year, maternal age, and income level. p values of < 0.05 were significant. § In the case of medical institution type, we used the number of prescriptions, not the number of pregnancies. www.nature.com/scientificreports/  www.nature.com/scientificreports/