Structural Nested Models for Assessing the Safety and Effectiveness of Generic Drugs

项目摘要 美国食品和药物管理局(FDA)预计，批准的仿制药提供相同的 质量、安全、功效为相应品牌。尽管如此，一些临床医生和患者 由于担心疗效较差或担心毒性或副作用，不愿使用仿制药 效果。FDA试图确保患者可以自信地获得仿制药，而不合格的 产品从市场上下架。这需要适当的上市前监管和上市后监管。 监测以了解仿制药的临床效果。 我们提出了提高FDA评估仿制药安全性和有效性的方法 使用医疗保健利用数据库(索赔数据)与其品牌对应药物相关的药物，以及 电子病历(EMR)。在做出有效的因果关系推论方面存在重大挑战 使用这些二次数据来源比较仿制药和品牌药的有效性。一些人 关键挑战包括：结果分类错误，缺少混杂调整的关键变量， 以及可能因患者死亡而信息缺失的数据进行跟踪。 我们的第一个目标是开发一种严格的、最先进的因果推理方法来比较 将适用于索赔数据库的仿制药和品牌疗法的毒性和有效性。我们的 第二个目标是利用与索赔有关的电子病历数据，加强目标1中开发的方法。 第三个目标是，我们建议对仿制药办公室(OGD)的FDA科学家进行培训，以实施 我们的方法。 我们建议将我们的方法应用于乳腺癌常用药物的研究：芳香酶。 抑制剂，其仿制药是可用的。我们将使用来自Optom实验室的关联数据集，其中包括两者 理赔数据和电子病历数据。该数据库有超过1.5亿份个人患者的索赔记录和30 100万名患者申请-电子病历涵盖10年或更长时间的患者经验。 具体目标1.开发一种最先进的因果推理方法，用于比较 将适用于医疗保健利用(索赔)数据库的仿制药和品牌药。 具体目标2.证明关联索赔--电子病历数据对仿制药监测的附加值 有效性和安全性。 具体目标3.为来自OGD的FDA科学家提供培训，以实施我们的方法学 接近。

Project Summary The U.S. Food and Drug Administration (FDA) expects that approved generic products provide the same quality, safety, and efficacy as the corresponding brand. Despite this, some clinicians and patients are reluctant to use generic medications due to fears of lesser effectiveness or concerns about toxicities or side effects. The FDA seeks to ensure that patients can confidently access generic drugs, and that substandard products be removed from market. This requires appropriate pre-marketing regulation and post-marketing surveillance to understand generic drugs’ clinical effects. We propose methods to enhance the FDA’s ability to evaluate the safety and effectiveness of generic drugs relative to their branded counterparts using healthcare utilization database (claims data), and electronic medical records (EMR). There are major challenges in making valid causal inferences regarding the comparative effectiveness of generics and branded drugs using these secondary sources of data. Some of the key challenges are: misclassification of outcomes, missing key variables for confounder adjustment, and data that are potentially informatively missing due to patient losses to follow up. Our first aim is to develop a rigorous, state-of-art, causal inference approach for comparing the toxicity and efficacy of generics and branded therapeutics that will be applicable to claims databases. Our second aim is to leverage the EMR data linked to claims, to enhance the methods developed in Aim 1. In the third aim, we propose to train the FDA scientists from the Office of Generic Drugs (OGD) in implementing our methods. We propose to apply o ur approach to the study of commonly used drugs in breast cancer: aromatase inhibitors, for which generics are available. We will use a linked data set from Optum Labs which includes both claims data and EHR data. This database has in excess of 150 million individual patient records for claims and 30 million patients for claims-EMR covering 10 years or more of patient experience. Specific Aim 1. To develop a state-of-art causal inference approach for comparing the toxicity and efficacy of generics and branded drugs that will be applicable to healthcare utilization (claims) databases. Specific Aim 2. To demonstrate the added value of linked claim-EMR data for surveillance of generic drug effectiveness and safety. Specific Aim 3. To provide training to the FDA scientists from OGD in implementing our methodological approach.