Prediction of risk of disability worsening and inflammatory disease activity in MS utilizing multimodal prediction algorithms

项目总结： 多发性硬化症(MS)表现出明显的异质性和不可预测的病程，临床上 范围从一些患者的非常轻微的疾病形式(通常被称为“良性多发性硬化”)到 攻击性疾病的病程，其他人的残疾迅速积累。此外，似乎还有 对许多可用的疾病修正疗法(DMT)的反应存在显著的个体间差异。 已有多种因素被认为与多发性硬化症的病程有关，包括人口统计学， 生活方式因素、临床特征、MRI衍生测量和血清神经丝轻链升高 (NFL)，以及其他。然而，目前尚不清楚这些因素在预测方面是互补的还是多余的 价值。此外，缺乏经过验证的工具来在个人层面上准确预测未来的煽动性 疾病活动或残疾恶化。这在很大程度上是因为缺乏足够大小、宽度和 代表性。近年来，电子病历(EMR)的使用急剧增加， 能够捕获来自大量个人的各种数据测量。此外， 统计机器学习方法的发展和改进彻底改变了分析方法 这样的高维数据集。这一背景为我们提供了一个独特的机会来利用和分析 为了开发MS的临床风险预测算法和个性化医疗工具。 多发性硬化症合作伙伴先进技术和健康解决方案(MS PATS)是一个由10人组成的网络 MS中心具有标准化的临床实践要素，以实施集中健康 信息交换架构。MS PATHS是围绕学习健康系统的概念而设计的 (LHS)，通过收集标准化的临床和成像数据将研究与持续的患者护理相结合 例行的医疗检查。截至2019年8月，已有15,000名患者选择参加MS路径。因此， 多发性硬化症路径网络是一种理想的、表型深刻的、“现实世界”的、庞大的多发性硬化症患者群体，其中 临床上相关的预测算法可以被开发和验证。 本项目的目标是开发和验证临床诊断的多模式预测算法。 We女士的相关疾病结果我们假设整合了各种潜在的预测因素，包括 人口学、临床特征(包括当前/既往使用DMT)、合并症/生活方式因素、MRI- 衍生措施和实验室数据(包括血清NFL)将导致开发和验证 可以准确预测未来临床残疾恶化和炎症性疾病活动的算法。 此外，这一方法将允许评估特定预测者对 开发了预测算法，并可能有助于识别疾病严重程度的新风险因素 女士。

Project Summary: Multiple sclerosis (MS) exhibits a markedly heterogeneous and unpredictable course, with a clinical spectrum ranging from very mild forms of the disease in some patients (often termed “benign MS”) to an aggressive disease course with rapid accumulation of disability in others. Furthermore, there appears to be significant inter-individual variability in the responses to the many available disease-modifying therapies (DMT). A variety of factors have been proposed to be associated with the disease course in MS, including demographics, lifestyle factors, clinical characteristics, MRI-derived measures, and elevated serum neurofilament light chain (NfL), among others. It remains unclear though if these factors are complementary or redundant in their predictive value. Moreover, there is a lack of validated tools to accurately predict, at an individual level, future inflammatory disease activity or disability worsening. This is largely due to the lack of datasets with sufficient size, breadth and representativeness. The use of electronic medical records (EMR) has dramatically increased in recent years, enabling the capture of a wide variety of data measures from large numbers of individuals. Furthermore, the development and refinement of statistical machine learning methods has revolutionized the approach to analysis of such high-dimensional datasets. This background provides a unique opportunity to leverage and analyze “big data” in order to develop clinical risk prediction algorithms and personalized medicine tools in MS. Multiple Sclerosis Partners Advancing Technology and Health Solutions (MS PATHS) is a network of 10 MS centers that have standardized elements of their clinical practice to implement a centralized health information exchange architecture. MS PATHS was designed around the concept of a learning health system (LHS), merging research with ongoing patient care by collecting standardized clinical and imaging data during routine medical visits. As of August, 2019, >15,000 patients have opted to participate in MS PATHS. Thus, the MS PATHS network is an ideal, deeply phenotyped, “real-world”, large population of people with MS, in which clinically relevant predictive algorithms may be developed and validated. The goal of the current project is to develop and validate multi-modal predictive algorithms of clinically relevant disease outcomes in MS. We hypothesize that integrating a wide variety of potential predictors, including demographics, clinical characteristics (including current/historical DMT use), comorbidities/lifestyle factors, MRI- derived measures and laboratory data (including serum NfL) will lead to the development and validation of algorithms that may accurately predict future clinical disability worsening and inflammatory disease activity. Furthermore, this approach will allow the assessment of the individual contribution of specific predictors to the developed predictive algorithms, and may aid with the identification of novel risk factors of disease severity in MS.