Innovating Clinical Trials for Amyotrophic Lateral Sclerosis: Challenging the Established Order

Abstract

Development of effective treatments for amyotrophic lateral sclerosis (ALS) has been hampered by disease heterogeneity, a limited understanding of underlying pathophysiology, and methodologic design challenges. We have evaluated 2 major themes in the design of pivotal, phase 3 clinical trials for ALS-(1) patient selection and (2) analytical strategy-and discussed potential solutions with the European Medicines Agency. Several design considerations were assessed using data from 5 placebo-controlled clinical trials (n = 988), 4 population-based cohorts (n = 5,100), and 2,436 placebo-allocated patients from the Pooled Resource Open-Access ALS Clinical Trials (PRO-ACT) database. The validity of each proposed design modification was confirmed by means of simulation and illustrated for a hypothetical setting. Compared to classical trial design, the proposed design modifications reduce the sample size by 30.5% and placebo exposure time by 35.4%. By making use of prognostic survival models, one creates a potential to include a larger proportion of the population and maximize generalizability. We propose a flexible design framework that naturally adapts the trial duration when inaccurate assumptions are made at the design stage, such as enrollment or survival rate. In case of futility, the follow-up time is shortened and patient exposure to ineffective treatments or placebo is minimized. For diseases such as ALS, optimizing the use of resources, widening eligibility criteria, and minimizing exposure to futile treatments and placebo is critical to the development of effective treatments. Our proposed design modifications could circumvent important pitfalls and may serve as a blueprint for future clinical trials in this population.

Figure 1. Distribution of Risk Profiles in Clinical Trial and Population-Based Cohorts

For each patient, we calculated his or her risk profile (horizontal axis), a composite score of 7 prognostic variables, which ranges from approximately −12 (very long prognosis) to 0 (very short prognosis). The distribution of the risk profiles is given as density curves, with the probability density on the vertical axis. The interpretation of the figure is similar to that of a histogram. The colors represent different trial populations; the population-based cohort is in gray. The solid squares are the population medians. A clear shift is observed in trial populations towards a better prognosis compared to a population-based cohort, which reflects the underrepresentation of patients with a poor prognosis in clinical trials. Exact ranges per cohort are available from Dryad (eAppendix 1, doi.org/10.5061/dryad.fbg79cnv7); LiCALS = lithium carbonate in patients with amyotrophic lateral sclerosis; PRO-ACT = Pooled Resource Open-Access ALS Clinical Trials.

Figure 2. Expected Trial Duration Under the Null and Alternative Hypothesis

Distribution of simulated trial duration under the null hypothesis (H₀, i.e., hazard ratio [HR] 1, panel A) and alternative hypothesis (H₁, i.e., HR 0.63, panel B). The vertical axis reflects the probability density; its interpretation is similar to a histogram. Due to the lack of a treatment benefit under H₀, the required number of events occurs more rapidly and the average duration is shorter compared to H₁. The peaks reflect the average analysis time points; the last peak in panel B is centered around 30 months (i.e., when all assumptions hold). In gray is a scenario when the survival probability in the placebo arm is better than expected (60.0% instead of 56.1%). FPI = first patient first visit; PI = percentile interval of the empirical trial duration ranging from the 2.5th to the 97.5th percentile.