Determining whether disease-modifying treatments (DMTs) in early Alzheimer's disease (AD) provide clinically meaningful benefits to people living with AD is critical and has triggered much debate in the field.^1-4 AD is a slowly progressive, ultimately fatal, neurodegenerative disease, characterized by progressive loss in cognitive ability and daily function.⁵ Current DMTs aim to slow disease progression, an important treatment-related outcome for people with, or at risk for, AD and their care partners.⁶

A widely adopted approach to evaluate the clinical relevance of a treatment benefit is to compare the proportion of patients who experience a meaningful improvement or deterioration in their symptoms and/or overall condition over the course of a clinical trial. In progressive diseases where emerging DMTs aim to slow the rate of progression in symptoms and underlying disease (as opposed to the temporary improvements provided by symptomatic therapies), evaluating meaningful deterioration is arguably more appropriate. To conduct such analyses in AD, estimates of meaningful score change on the Clinical Dementia Rating–Sum of Boxes (CDR-SB) are needed to reflect clinically relevant deterioration in an individual patient's cognition and daily function. Lansdall et al.⁷ used established methods to generate meaningful within-patient change (MWPC) thresholds to define meaningful deterioration on the CDR-SB, building on existing estimates.^{8, 9} This letter aims to provide clarification on the appropriate application of MWPC thresholds to evaluate clinical trial data.

Thresholds to define meaningful individual-level change on a scale are commonly used in neurology trials, both to identify responders and progressors. Examples include confirmed disability progression on the clinician-administered Expanded Disability Status Scale in multiple sclerosis or meaningful reduction of neuropathic pain on a numeric rating scale.^{10, 11} Figure 5 of the FDA Aricept (donepezil) label, a symptomatic treatment for AD, also includes an example of how to apply MWPC thresholds to clinical trial data, using 4 and 7 points improvement on the Alzheimer's Disease Assessment Scale–Cognitive Subscale (ADAS-Cog) to report the proportion of responders across treatment and placebo arms. Although the intended use of these thresholds is to define and assess MWPC, recent publications in AD have instead applied MWPC thresholds to evaluate whether the magnitude of observed between-group differences in mean change-from-baseline (commonly the primary endpoint in AD trials) is clinically meaningful.^{3, 4} Trigg et al., highlight why MWPC thresholds cannot be applied directly to evaluate the meaningfulness of between-group differences,¹² and doing so sets an inappropriate bar for emerging DMTs in AD. For example, considering that the average observed placebo decline over 18 months in recent clinical trials of early AD of around 1.5–2.5 points on the CDR-SB^{13, 14} is approximately equal to the proposed range of MWPC thresholds, a meaningful effect could be reached only if the treatment group remained stable or a large proportion of patients achieved a marked improvement from baseline. This expectation that treatments can recover what is already lost due to AD is unrealistic for emerging DMTs aiming to slow disease progression.

An appropriate application of MWPC thresholds would be to identify individual patients who may have changed or progressed to a meaningful degree to inform “progressor” analyses of trial data, which compare, for example, (a) the likelihood of meaningful progression (i.e., deterioration) for treatment versus placebo groups (i.e., hazard ratio analyses, Figure 1A), and/or (b) the proportion of meaningful progressors across treatment arms via progressor analyses and/or annotation of empirical cumulative distribution function (eCDF) plots (Figure 1B). Examples of similar analyses have been presented at recent conferences.^{15, 16} In the context of progressive neurodegenerative disease, individuals will deteriorate to varying degrees over the course of a clinical trial. Acknowledging that not all patients will benefit from treatment and that those who benefit will do so to varying degrees, individual-level analyses leveraging MWPC thresholds to demonstrate that fewer patients meaningfully progress may offer a more suitable approach to evaluate the meaningfulness of a treatment effect.

A lack of clear and consistent terminology may have contributed to the confusion regarding the application of existing thresholds, with terms like minimal clinically important difference (MCID) often being used interchangeably to refer to both within-patient change and between-group differences. For example, Andrews et al.⁸ used methods to establish within-patient change thresholds for the CDR-SB, but use of the term MCID may have led to recent publications citing this work to evaluate the meaningfulness of between-group differences.^{3, 4} For this reason, we use MWPC to clarify the intended use of these thresholds and align with recent publications.^{1, 10, 12} When used appropriately for individual-level analyses, these thresholds provide an opportunity to build on typical primary endpoint analyses that evaluate between-group differences in change-from-baseline and add to the totality of evidence for consideration when evaluating the clinical meaningfulness of a given treatment.^{1, 2, 6} Such analyses, including event-based, dichotomous endpoints, may offer particular value for certain stakeholders and represent one important approach to evaluating the clinical meaningfulness of trial data.

确定早期阿尔茨海默病 （AD） 中的疾病缓解治疗 （DMT） 是否为 AD 患者提供具有临床意义的益处至关重要，并在该领域引发了许多争论。^1-4 AD 是一种缓慢进展、最终致命的神经退行性疾病，其特征是认知能力和日常功能进行性丧失。⁵ 目前的 DMT 旨在减缓疾病进展，这对于 AD 患者或有 AD 风险的人及其护理伙伴来说是一个重要的治疗相关结果。⁶

评估治疗益处的临床相关性的一种广泛采用的方法是比较在临床试验过程中症状和/或整体状况出现有意义改善或恶化的患者比例。在新兴 DMT 旨在减缓症状和潜在疾病进展速度（而不是对症治疗提供的暂时改善）的进展性疾病中，评估有意义的恶化可以说更合适。为了在 AD 中进行此类分析，需要估计临床痴呆评分-方框总和 （CDR-SB） 的有意义分数变化，以反映个体患者认知和日常功能的临床相关恶化。Lansdall 等人 ⁷ 使用既定方法生成有意义的患者内变化 （MWPC） 阈值，以定义 CDR-SB 的有意义恶化，建立在现有估计的基础上。^8、9本信函旨在澄清如何适当应用 MWPC 阈值来评估临床试验数据。

在神经病学试验中，通常用于定义量表上有意义的个体水平变化的阈值，用于识别反应者和进展者。示例包括在多发性硬化症中临床医生管理的扩展残疾状况量表上确认的残疾进展或在数字评定量表上显着减轻神经性疼痛。^10、11FDA Aricept（多奈哌齐）标签的图 5 是 AD 的对症治疗，还包括如何将 MWPC 阈值应用于临床试验数据的示例，使用阿尔茨海默病评估量表-认知分量表 （ADAS-Cog） 的 4 分和 7 分改进来报告治疗组和安慰剂组的反应者比例。尽管这些阈值的预期用途是定义和评估 MWPC，但最近在 AD 上的出版物反而应用 MWPC 阈值来评估观察到的相对于基线的平均变化（通常是 AD 试验中的主要终点）的组间差异的大小是否具有临床意义。^3、4Trigg 等人强调了为什么 MWPC 阈值不能直接应用于评估组间差异的意义，¹² 这样做为 AD 中出现的 DMT 设定了不合适的标准。例如，考虑到在最近的早期 AD 临床试验中观察到的安慰剂在 18 个月内的平均下降约为 1.5-2.5 分，CDR-SB^{13， 14} 大约等于拟议的 MWPC 阈值范围，只有当治疗组保持稳定或很大一部分患者从基线获得显着改善时，才能达到有意义的效果。 这种期望治疗可以恢复因 AD 而已经损失的东西，对于旨在减缓疾病进展的新兴 DMT 来说是不现实的。

MWPC 阈值的适当应用是确定可能已经改变或进展到有意义程度的个体患者，以便为试验数据的“进展者”分析提供信息，例如，比较 （a） 治疗组与安慰剂组有意义进展（即恶化）的可能性（即风险比分析，图 1A），和/或 （b） 通过进展因子分析和/或经验注释，治疗组中有意义进展者的比例累积分布函数 （eCDF） 图（图 1B）。在最近的会议上已经介绍了类似分析的例子。^15、16 元在进行性神经退行性疾病的情况下，个体在临床试验过程中会不同程度地恶化。认识到并非所有患者都会从治疗中受益，并且受益者会在不同程度上受益，利用 MWPC 阈值的个体层面分析来证明较少的患者有意义地进展，这可能提供一种更合适的方法来评估治疗效果的意义。

缺乏清晰一致的术语可能导致对现有阈值应用的混淆，像最小临床重要差异 （MCID） 这样的术语经常互换使用，以指代患者内部变化和组间差异。例如，Andrews 等人 ⁸ 使用方法建立了 CDR-SB 的患者内变化阈值，但使用术语 MCID 可能导致最近的出版物引用这项工作来评估组间差异的意义。^3、4因此，我们使用 MWPC 来阐明这些阈值的预期用途，并与最近的出版物保持一致。^1、10、12当适当地用于个体层面的分析时，这些阈值提供了一个机会，可以建立在典型的主要终点分析的基础上，这些分析评估相对于基线变化的组间差异，并增加在评估给定治疗的临床意义时可供考虑的证据总数。^1、2、6此类分析，包括基于事件的二分类终点，可能为某些利益相关者提供特殊价值，并代表评估试验数据临床意义的一种重要方法。