Objective: The ALSFRS-R is limited by multidimensionality, which originates from the summation of various subscales. This prevents a direct comparison between patients with identical total scores. We aim to evaluate how multidimensionality affects the performance of the ALSFRS-R in clinical trials. Methods: We simulated clinical trial data with different treatment effects for the ALSFRS-R total score and its subscales (i.e. bulbar, fine motor, gross motor and respiratory). We considered scenarios where treatment reduced the rate of ALSFRS-R subscale decline either uniformly (i.e. all subscales respond identically to treatment) or non-uniformly (i.e. subscales respond differently to treatment). Two main analytical strategies were compared: (1) analyzing only the total score or (2) utilizing a subscale-based test (i.e. alternative strategy). For each analytical strategy, we calculated the empirical power and required sample size. Results: Both strategies are valid when there is no treatment benefit and provide adequate control of type 1 error. If all subscales respond identically to treatment, using the total score is the most powerful approach. As the differences in treatment responses between subscales increase, the more the total score becomes affected. For example, to detect a 40% reduction in the bulbar rate of decline with 80% power, the total score requires 1380 patients, whereas this is 336 when using the alternative strategy. Conclusions: Ignoring the multidimensional structure of the ALSFRS-R total score could have negative consequences for ALS clinical trials. We propose determining treatment benefit on a subscale level, prior to stating whether a treatment is generally effective.
Keywords: ALSFRS-R; clinical trials; models; multidimensionality; therapy.