Efficiency of multivariate tests in trials in progressive supranuclear palsy

Area of Science:

• Clinical Trial Statistics
• Neurological Disease Progression Measurement
• Statistical Modeling in Clinical Research

Background:

• Measuring disease progression in multifaceted conditions like progressive supranuclear palsy (PSP) presents significant challenges for clinical trials.
• The Progressive Supranuclear Palsy Rating Scale (PSPRS) is a common tool, but its optimal statistical analysis for detecting treatment effects is debated.

Purpose of the Study:

• To evaluate and compare various statistical approaches for analyzing PSPRS outcomes in clinical trials.
• To introduce and assess novel Item Response Theory (IRT) based methods for measuring disease status and treatment effects in PSP.

Main Methods:

• Simulation study assessing the performance of sum scores, modified PSPRS scales, multivariate tests, multiple comparison approaches, and two new IRT models.
• Re-analysis of the ABBV-8E12 clinical trial data to illustrate the practical application of the evaluated statistical methods.

Main Results:

• Classical PSPRS sum scores show moderate to high power when treatment effects are uniform across items.
• IRT models demonstrate the highest statistical power when data align with their underlying assumptions.
• Multiple testing approaches are more powerful when treatment effects are localized to specific PSPRS domains or items.

Conclusions:

• No single statistical method is universally optimal for analyzing PSPRS data; the best approach depends on the pattern of treatment effects.
• IRT models and multiple testing strategies offer advantages over traditional sum scores in specific clinical trial scenarios for PSP.
• Findings have implications for optimizing the design of future clinical trials for PSP and similarly complex neurological diseases.