Innovative precision medicine methods in subgroup identification for Alzheimer's disease

是一个主要的和迅速增长的公共卫生问题：超过3000万人 全球范围内都遭受着AD，预计到2050年将翻两番。据报道，AD是第三大 美国的死因。随着这场即将到来的全球公共卫生危机， AD进展是迫切需要的，但在最近加速批准aducenumab之前很少可用。 因此，当总体药物治疗效果不佳时，确定从药物治疗中获益最多的亚群是非常有意义的。 治疗效果最小或无临床意义。如果能够确定这些亚群， 来自阴性试验的治疗可以被证明至少帮助一部分AD人群。本提案中 我们将在重复测量的混合模型上使用基于非参数交互树（IT）的方法 （MMRM）和基于回归的方法来识别这些亚群。IT for MMRM基于评估 治疗与协变量的相互作用，并可以自动寻找其中 治疗显示出异质效应。我们还探索了一种新的，更有吸引力的融合惩罚方法， 最终的树确定而无需分组信息的任何先验知识。基于回归的方法旨在 根据其特征确定将从AD治疗中获益的亚群，这是非常灵活的 进行个体化治疗选择。最后，我们会编制和发放一套方便使用的统计数字， 软件包，使研究人员能够轻松实现这些方法。我们的扩展会更好 捕获疾病进展中的个体异质性，并促进基于证据的精准医学， 未来的研究方向和研究领域。

is a major and rapidly increasing public health concern: over 30 million individuals worldwide suffer from AD, which is projected to quadruple by 2050. AD has been reported to be the third leading cause of death in the US. With this impending global public health crisis, treatments that prevent onset or slow progression of AD are urgently needed but rarely available until the recent accelerated approval for aducenumab. Therefore, it is of great interest to identify subpopulations which benefit most from a medication when the overall treatment effect is minimum or not clinically meaningful. If such subpopulations can be identified, some of the treatments from the negative trials can be proven to at least help a portion of the AD population. In this proposal we will employ non-parametric interaction tree (IT)-based methods on mixed models for repeated measures (MMRM) and regression-based methods to identify such subpopulations. IT for MMRM builds on the assessment of the treatment-by-covariates interactions and can automatically seek subgroups of individuals in whom the treatment shows heterogeneous effects. We also explore a new and more attractive fusion penalty approach for final tree determination without any prior knowledge of grouping information. The regression-based methods aim to identify subpopulations who will benefit from AD treatment based on their characteristics, which is very flexible to make individualized treatment selection. Finally, we will develop and disseminate a user-friendly statistical software package that will enable researchers to implement these methods with ease. Our extensions will better capture individual heterogeneity in disease progression and facilitate evidence-based precision medicine in future AD studies and other research areas.