ACTS (AD Clinical Trial Simulation): Developing Advanced Informatics Approaches for an Alzheimer's Disease Clinical Trial Simulation System

摘要 阿尔茨海默病和相关痴呆(AD/ADRD)是最常见的神经退行性脑病 以记忆和学习能力的大量丧失为特征。AD/ADRD影响600多万美国人 并给社会上的照顾者带来了沉重的负担。然而，AD/ADRD仍缺乏有效的治疗方法。 虽然随机临床试验(RCT)可以为干预措施的有效性提供可靠的证据，但他们 也有固有的局限性，包括成本高和执行时间长。此外，RCT通常是 在特定人群和专门环境中进行，但后续行动时间有限。因此，他们 在推广到现实世界的临床实践中可能会受到限制。临床试验模拟正在成为一种 在实施之前评估可行性、调查假设和完善研究方案的有效方法 真正的审判。提高现实世界数据(RWD)的可用性和粒度，例如电子健康记录 (EHR)和医疗索赔数据以及数据科学的进步提供了尚未开发的机会 RWD用于试验模拟研究，以生成真实世界证据(RWE)。尽管如此，还是有方法论 支持基于RWD的试验模拟研究的障碍和信息学挑战，特别是AD：(1) 临床试验需要使用正式和标准的方法(即，本体论)来表示，以捕获整个 试验的范围，特别是资格标准和结果衡量标准(即有效性和安全性)； 需要使正式和标准表示与RWD标准(例如，公共数据)可互操作 模型)以确定研究队列和相关的、重要的患者特征(即通过可计算的表型 而自然语言处理[NLP]方法作为丰富的AD相关信息，如认知分数往往 存在于非结构化的临床记录中)；以及(3)需要实施综合的和可重复使用的管道 可以与现有的大规模RWD无缝对接，为AD生成高质量的分析就绪数据集 临床试验模拟研究。为了解决这些障碍，我们建议创建并试行这些法案 (阿尔茨海默病临床试验模拟)系统，利用三大RWD集合(约2000万 来自One佛罗里达网络、UT内科医生临床数据研究仓库和Optom的患者 临床信息学数据)。具体地说，我们建议开发新的信息学方法来代表整体 在考虑RWD的联系的同时进行AD试验(目标1)，使用结构化和非结构化RWD来 开发稳健的表型算法，使以前无法计算的AD研究特征可计算(AIM 2)，并开发ACTS网络应用程序，为AD研究人员提供一个集成环境 使用交互式Web界面构建虚拟AD试验并获取用于试验的分析就绪数据集 模拟研究(目标3)。

ABSTRACT Alzheimer's disease and related dementias (AD/ADRD) are the most common neurodegenerative brain disease and characterized by massive loss of memory and learning. AD/ADRD affects more than 6 million Americans and puts a heavy burden on caregivers in society. However, effective treatment of AD/ADRD is still lacking. While randomized clinical trials (RCT) can provide reliable evidence on the effectiveness of interventions, they also have inherent limitations including high cost and long execution time. In addition, RCTs usually are conducted on selected populations and in specialized environments with limited follow up time. Therefore, they could have limitations in generalizability to real-world clinical practice. Clinical trial simulation is becoming an effective approach to assess feasibility, investigate assumptions, and refine study protocols before conducting the actual trials. Increased availability and granularity of real-world data (RWD) such as electronic health record (EHR) and medical claims data along with advances in data science offer untapped opportunities to leverage RWD for trial simulation studies to generate real world evidence (RWE). Nevertheless, there are methodological barriers and informatics challenges in supporting RWD-based trial simulation studies, especially for AD: (1) clinical trials need to be represented using a formal and standard approach (i.e., ontologies) to capture the entire scope of a trial, especially eligibility criteria and outcome measures (i.e., both effectiveness and safety); (2) such formal and standard representation needs to be made interoperable with RWD standards (e.g., common data models) to identify study cohorts and relevant, important patient characteristics (i.e., via computable phenotypes and natural language processing [NLP] methods as rich AD-related information such as cognitive scores often exist in unstructured clinical notes); and (3) comprehensive and reusable pipelines need to be implemented that can seamlessly align with existing large-scale RWD for generating high-quality analytic-ready datasets for AD clinical trial simulation studies. To address these barriers, we propose create and pilot test the ACTS (Alzheimer's disease Clinical Trial Simulation) system, leveraging three large collections of RWD (~20 million patients from the OneFlorida network, UT Physician Clinical Data Research Warehouse, and the Optum’s Clinformatics data). Specifically, we propose to develop novel informatics approaches to represent the entirety of AD trials while considering the connection of RWD (Aim 1), to use both structured and unstructured RWD to develop robust phenotyping algorithms that will render previously incomputable AD study traits computable (Aim 2), and to develop the ACTS web application, which will provide an integrated environment for AD researchers to construct virtual AD trials using an interactive web interface and obtain analytic-ready datasets for trial simulation studies (Aim 3).