Incorporation of multilevel ontologies of adverse events and vaccines for vaccine safety surveillance

纳入不良事件和疫苗的多级本体 疫苗安全监测 项目摘要 疫苗面临着比大多数药品更严格的安全标准，因为它们是由 给健康的人，通常是孩子。对疫苗接种后不良事件进行有效和严格的分析 (AEs)对确保疫苗的安全至关重要。疫苗不良事件报告系统 （VAERS）是一项国家疫苗安全性监测计划，其中包含来自 1990年至今。统计方法已用于VAERS提取重要信号 隐藏在这个庞大而复杂的数据库中，并提供了一个无假设的安全特性视图， 底层数据。然而，由于建模，现有方法可能会错过检测严重AE 在不同类型AE之间独立性的错误假设下。 作为对FOA PA-18-873的回应，本提案提出了以下具体目标： “建立/评价用于分析疫苗安全性数据的统计方法， 从现有的数据源，如被动报告系统或医疗保健数据库。 我们建议制定一系列疫苗安全性监测方法， 不良事件本体以及疫苗本体。具体来说，我们将使用医学词典 用于监管活动（MedDRA）和疫苗本体（VO），以形成我们的模型基础， 系统地挖掘和监测安全信号。据我们所知，这是第一个 尝试在信号检测方法中直接结合AE和疫苗本体。多起ae 可能个别罕见到不被发现，但如果他们是相关的，他们可以借用力量， 以增加被标记的可能性此外，借贷强度导致 减少相关AE，从而也减少了引人注目的假阳性。此外，本发明还 多起AE可能共同指向潜在的不良原因，结合其他专家 从疫苗本体的知识，如疫苗的组成部分，我们将能够了解 不同类型AE的根本原因。 1

Incorporation of multilevel ontologies of adverse events and vaccines for vaccine safety surveillance PROJECT SUMMARY Vaccines face tougher safety standards than most pharmaceutical products because they are given to healthy people, often children. Effective and rigorous analyses of post-vaccination adverse events (AEs) is critical to ensure the safety of vaccines. The Vaccine Adverse Event Reporting System (VAERS) is a national vaccine safety surveillance program which contains spontaneous reports from 1990 to present. Statistical approaches have been used on VAERS to extract important signals hidden in this large, complex database and offer a hypothesis-free view of the safety characteristics in the underlying data. However, existing methods may miss detecting serious AEs due to modeling under the false assumption of independence between different types of AEs. In response to the FOA, PA-18-873, this proposal addresses the specific objective: “creation/evaluation of statistical methodologies for analyzing data on vaccine safety, including data available from existing data sources such as passive reporting systems or healthcare databases.” We propose to develop a series of methods for vaccine safety surveillance while incorporating adverse event ontology as well as vaccine ontology. Specifically, we will use the Medical Dictionary for Regulatory Activities (MedDRA) and the vaccine ontology (VO) to form the basis of our models for systematically mining and monitoring safety signals. To the best of our knowledge, this is the first attempt to directly incorporate AE and vaccine ontologies in the signal detection method. Multiple AEs may individually be rare enough to go undetected, but if they are related, they can borrow strength from each other to increase the chance of being flagged. Furthermore, borrowing strength induces shrinkage of related AEs, thereby also reducing headline-grabbing false positives. Additionally, multiple AEs may collectively point to an underlying adverse cause, combined with additional expert knowledge from the vaccine ontology, such as vaccine components, we will be able to understand the root cause of different types of AEs. 1