Deconstructing progression of amyotrophic lateral sclerosis in stages: a Markov modeling approach

<i>Objectives:</i> Propose an empirical amyotrophic lateral sclerosis (ALS) staging approach called Fine’til 9 (FT9) based on how many of the patient’s ALS functional rating scale (ALSFRS-R) subscores are 9 or less (of normal 12). Gain insights into progression of ALS by applying Markov models to ALS stages by multiple systems (King’s, Milan–Torino system (MITOS) and FT9). <i>Methods:</i> Patients from the Pooled Resource Open-Access ALS Clinical Trials (PRO-ACT) dataset were staged using ALSFRS-R responses. Risks of progression through stages and death were estimated, as were effects of prognostic variables on these risks. <i>Results:</i> A total of 29,947 time points in 3199 patients from the PRO-ACT dataset were assigned stages. Although the three systems were moderately correlated, MITOS stages were heavily skewed toward advanced disease, whereas King’s and FT9 stages were more balanced. Non-sequential progression was observed with King’s system. Markov models adequately described transitions from stage to stage in the first year of observation, but underestimated risks beyond that point. Regardless of staging method, initial rate of ALSFRS-R decline had a powerful effect on rate of progression through sequential stages, whereas age predominantly influenced stage-specific mortality. <i>Conclusion:</i> King’s and FT9 are more sensitive to observed progression of disease in clinical trials than MITOS. FT9 can partition the course similar to King’s, and may have advantages of sequential progression and easy applicability to retrospective data. Markov transition intensity estimates may be of value for counseling, health economic studies, and research design. In particular, this framework permits estimation of multidimensional effects of variables (including treatment) on outcome.

研究目标：提出一种实证性肌萎缩侧索硬化症（amyotrophic lateral sclerosis, ALS）分期方法，命名为Fine’til 9（FT9），该方法基于患者肌萎缩侧索硬化症功能评定量表修订版（Amyotrophic Lateral Sclerosis Functional Rating Scale-Revised, ALSFRS-R）的子项得分≤9分（正常分值为12分）的数量。通过对多系统分期（King分期、米兰-都灵系统（Milan–Torino system, MITOS）及FT9）应用马尔可夫模型，深入解析ALS的疾病进展规律。研究方法：从汇集资源开放获取ALS临床试验（Pooled Resource Open-Access ALS Clinical Trials, PRO-ACT）数据集纳入研究对象，基于ALSFRS-R的应答结果对患者进行分期。估算各分期的疾病进展风险及死亡风险，并分析预后变量对上述风险的影响。研究结果：本研究共对PRO-ACT数据集中3199例患者的29947个时间点完成了分期划分。尽管三种分期系统存在中等程度的相关性，但MITOS分期显著偏向于晚期疾病，而King分期与FT9分期则更为均衡。King分期系统可观察到非序贯性疾病进展。马尔可夫模型可较好拟合随访第一年的分期间转换情况，但对该时间点之后的风险存在低估。无论采用何种分期方法，ALSFRS-R初始下降速率对序贯分期的进展速率均具有显著影响，而年龄则主要影响分期特异性死亡率。研究结论：相较于MITOS，King分期与FT9分期对临床试验中观察到的疾病进展更为敏感。FT9的疾病进程划分与King分期相似，且具备序贯性进展的优势，同时更易于应用于回顾性数据。马尔可夫转换强度估计值可为患者咨询、卫生经济学研究及研究设计提供参考价值。尤为重要的是，本研究框架可量化包括治疗在内的多种变量对结局的多维影响。