| الوصف: |
Neurological and neurodegenerative disorders, including Alzheimer's disease, Parkinson's disease, Huntington's disease, and amyotrophic lateral sclerosis (ALS), are characterized by heterogeneous disease progression. As a group, these complex diseases are likely influenced by an interplay between numerous genetic and environmental factors. Drug development for these diseases could benefit from a statistical framework capable of efficiently stratifying participants using information available at baseline, but current methods have largely failed in this regard. In ALS, both slowly and rapidly progressing patients have been identified as confounding analyses of clinical trials. Statistically speaking, heterogeneous disease progression rates in ALS clinical trials contribute to both trial arm misbalances and high variances of study populations, making it difficult to detect treatment effects. These issues increase the sample size, duration, and cost of clinical trials in neurological diseases and contribute to high failure rates. The PRO-ACT ALS database, composed of the clinical trial records of over 10,700 ALS patients, represents an ideal dataset for building a prototypical statistical framework for increasing the efficiency of drug development. We have used PRO-ACT to develop a dozen machine-learning based models that describe the progression of the disease. In this chapter we review our experience using the models to create drug development applications that can be used throughout clinical development, including as virtual controls in studies that lack a concurrent control (e.g., phase I, phase II, open-label extension, and phase IV studies), and in phase IIb and phase III randomized control trials to improve randomization, enrichment, covariate adjustment, and subgroup analysis. These machine-learning based applications will improve the chances of successful development of neurotherapeutics. |
