Comparative Safety of Non-Insulin Agents in Pregnant Women with Pregestational Diabetes

在一般人群中，非胰岛素药物已被广泛接受用于治疗2型糖尿病， 考虑到其与皮下胰岛素注射相比的功效和耐受性。最常见的一 妊娠合并症是妊娠前糖尿病，目前的管理重点是 通过饮食、锻炼和胰岛素治疗（如果需要）实现血糖控制。美国糖尿病 协会和妇产科学院建议计划怀孕的妇女和 对于妊娠期患者，在获得进一步数据之前，应使用胰岛素治疗替代口服降糖药 关于安全性的问题已经得到解决。然而，孕妇通常暴露于这些制剂，因为 1)指南并没有得到普遍遵循，2）一些妇女拒绝使用胰岛素，3）50%的孕妇 计划外的，如果一个女人已经使用口服药物，没有时间转换到胰岛素之前， 器官发生此外，在怀孕期间，口服药物对患者具有潜在益处 可接受性和依从性，从而改善血糖控制和妊娠结局。即使对那些 对于单独口服降糖药不足以实现血糖控制的患者， 减少胰岛素剂量。由于妊娠期随机对照试验具有足够的样本量来定义 这些代理的安全性是不现实的，我们需要及时的信息， 观察性研究在此之前，缺乏信息将继续是使用这些技术的一个关键障碍， 怀孕我们的主要目的是量化与下列因素相关的母体和胎儿不良事件的风险： 与单独使用胰岛素和单独使用二甲双胍相比，妊娠期间特定的非胰岛素抗糖尿病疗法。 我们建立了一个与婴儿相关的300万例妊娠队列， 两个基于人群的医疗保健数据库中的处方和临床状况：医疗补助分析 eXtract（MAX）和Truven Health MarketScan（Truven）。本研究将确定超过18，000（最大）的队列 和15，000名（Truven）在2000-2018年分娩的患有孕前2型糖尿病的女性。药物暴露 将根据药房配药记录确定，结局将基于住院和门诊 诊断和程序，使用经过验证的定义。为了控制糖尿病的严重程度和其他混杂因素， 我们将（i）将人群限制在2型糖尿病女性;（ii）使用积极的参考组;（iii）使用 倾向评分分层以平衡糖尿病严重程度的方面;以及（iv）使用新的倾向评分 校准方法，通过将来自子样本的血糖控制数据与 实验室记录或链接的电子医疗记录。将使用具有国家代表性的调查数据 以确保对美国人口的普遍性。广义线性模型将估计相对风险及其 95%置信区间。将进行敏感性分析，以检验结果的稳健性。先前工作 研究人员和试点数据支持拟议研究的可行性。

In the general population, non-insulin agents have gained wide acceptance for the treatment of type 2 diabetes, given their efficacy and tolerability when compared to subcutaneous insulin injections. One of the most common medical complications in pregnancy is pre-gestational diabetes and its management currently focuses on achieving glucose control with diet, exercise and, if needed, insulin treatment. Both the American Diabetes Association and the College of Obstetrics and Gynecology recommend to women planning pregnancy and for those who are pregnant that oral antidiabetic agents be substituted with insulin therapy until further data regarding safety become available. However, pregnant women are commonly exposed to these agents because 1) guidelines are not universally followed, 2) some women refuse to use insulin, and 3) 50% of pregnancies are unplanned and if a woman is already using an oral agent there is no time to switch to insulin before organogenesis. Furthermore, in pregnancy, oral agents have potential benefits with respect to patient acceptability and adherence, and therefore improved glycemic control and pregnancy outcomes. Even for those patients for whom oral antidiabetic agents alone are inadequate to achieve glycemic control, they can be used to reduce insulin dose. Since randomized controlled trials in pregnancy with sufficient sample size to define the safety of these agents robustly are not realistic, we need timely information based on carefully conducted observational studies. Until then, the lack of information will continue to be a critical barrier to their use in pregnancy. Our primary objective is to quantify the risk of maternal and fetal adverse events associated with specific non-insulin antidiabetic therapies during pregnancy in comparison to insulin alone and metformin alone. We have established a cohort of 3 million pregnancies linked to infants with longitudinal information on prescriptions and clinical conditions within two population-based healthcare databases: the Medicaid Analytic eXtract (MAX) and Truven Health MarketScan (Truven). This study will identify cohorts of over 18,000 (MAX) and 15,000 (Truven) women with pre-gestational type 2 diabetes who delivered in 2000-2018. Drug exposure will be determined based on pharmacy dispensing records, and outcomes will be based on in- and outpatient diagnoses and procedures, using validated definitions. To control for diabetes severity and other confounders, we will (i) restrict the population to women with type 2 diabetes; (ii) use an active reference group; (iii) use propensity score stratification to balance aspects of diabetes severity; and (iv) use a novel propensity score calibration approach to further adjust by incorporating data on glycemic control from subsamples with either laboratory records or linked electronic medical records. Data from nationally-representative surveys will be used to ensure generalizability to the US population. Generalized linear models will estimate relative risks and their 95% confidence intervals. Sensitivity analyses will be conducted to test the robustness of the findings. Prior work by the investigators and pilot data support the feasibility of the proposed study.