Abstract
Monitoring respiratory function is essential for assessing the progression of Amyotrophic Lateral Sclerosis (ALS) and planning interventions. Using spirometry data from the Radcliff Study —a fully remote, longitudinal, exploratory study with a cohort of 67 pALS—we demonstrate that flexible coaching, combined with a quality control analysis that excludes values of ‘0’ and timepoints with failed measurement trials, produces consistent remote spirometry results. Our findings indicate that home-measured Slow Vital Capacity (SVC) and Forced Vital Capacity (FVC) evolve similarly and progress linearly over the study period (7.7 ± 4.0 months). This remains true in both slow and fast progressor subpopulations. This observed linearity in respiratory trajectories supports the potential for early, accurate estimation of progression, reinforcing the feasibility of less frequent monitoring without compromising assessment precision, and reducing the burden on both pALS and the healthcare system. Furthermore, our results align with reported in-clinic pulmonary tests, validating remote monitoring as a means to promote more equitable and accessible clinical trial designs.
Introduction
Amyotrophic Lateral Sclerosis (ALS) is a progressive neurodegenerative disease that eventually leads to respiratory failure. Monitoring lung function—specifically Forced Vital Capacity (FVC) and Slow Vital Capacity (SVC)—is the gold standard for assessing disease progression and timing interventions like non-invasive ventilation. Traditionally, these tests are conducted in multidisciplinary clinics. However, for people living with ALS (pALS), traveling to a clinic is physically exhausting, logistically difficult, and expensive, especially as the disease advances. While remote monitoring offers a solution, concerns regarding data quality and the necessity of professional supervision have persisted. This study validates a flexible coaching model and an automated quality control (QC) pipeline to enable reliable home-based spirometry.
Methods
The researchers utilized data from the Radcliff Study, a fully remote, longitudinal, observational study involving 67 pALS. Participants were initially trained by proctors and subsequently allowed to choose between autonomous sessions or remote guidance via video call. A total of 1,060 sessions were recorded over a mean duration of 7.7 ± 4.0 months. To ensure data integrity, an automated QC protocol based on American Thoracic Society (ATS) standards was implemented, including back-extrapolation checks and the exclusion of erroneous trials. Statistical analysis compared linear versus quadratic models of decline and employed Mixtures of Gaussian Processes (MoGP) to identify progression phenotypes.
Results
Interchangeability and Linearity of Respiratory Metrics
A key finding of the study is the high degree of interchangeability between FVC and SVC in a remote setting. The metrics evolved almost identically, with a correlation coefficient of $\rho=0.97$ for intercepts. Furthermore, respiratory decline was found to be remarkably linear over 6–8 month windows. The linear model outperformed the quadratic model, showing a significantly lower mean Root Mean Square Error (RMSE) for FVC (5.26 vs. 5.86). This linearity allows for the early estimation of a patient's progression slope, which serves as a critical predictor of survival.
Reliability of Remote, Unproctored Monitoring
The study demonstrated that once initial training is complete, proctoring becomes optional. There was no statistically significant difference in data quality (RMSE) between proctored and unproctored sessions. This finding supports the feasibility of decentralized clinical trials where patients can perform complex breathing maneuvers correctly at home without constant professional supervision, significantly reducing the logistical burden on participants.
Progression Clusters and Trial Optimization
The MoGP model identified distinct progression clusters, including slow (< 0.4 points/month), moderate (~0.9 points/month), and fast (~2.1 points/month) progressors. Power analysis revealed that increasing study duration from 5 to 6 months for weekly sessions could reduce the required cohort size by 60% to detect a 30% change in slope. Notably, reducing session frequency from weekly to bimonthly had only a marginal impact on the required sample size, suggesting that less frequent monitoring may be sufficient for accurate tracking.
Discussion
This research confirms that home-based spirometry is a reliable, high-quality alternative to in-clinic testing for pALS. By leveraging flexible coaching and automated QC, researchers can track disease progression accurately while minimizing patient burden. While independent spirometry may become more difficult in the advanced stages of ALS, requiring caregiver assistance, the current results provide a strong foundation for more equitable and accessible clinical trial designs. Future work will explore the integration of passive monitoring technologies, such as speech acoustics, to further augment active spirometry data and provide an even lower-burden monitoring solution.
Summary updated