Longitudinal Modeling of Rank-based Global Outcome.

Area of Science:

• Biostatistics
• Clinical Research Methodology
• Chronic Disease Epidemiology

Background:

• Chronic diseases present complex symptoms, necessitating integrated outcomes for comprehensive assessment.
• Global outcomes, like the global rank-sum, are used to combine multiple individual disease indicators.
• Existing methods may not fully capture the dynamic nature of disease burden over time.

Purpose of the Study:

• To develop a novel global percentile outcome for longitudinal data, reflecting time-varying global disease burden.
• To establish robust regression strategies for analyzing this new outcome within a flexible modeling framework.
• To extend the methodology to handle missing data common in clinical studies.

Main Methods:

• Development of a global percentile outcome for longitudinal disease burden.
• Application of a monotonic index model with a maximum rank correlation estimator.
• Extension of methods to address missing at random (MAR) dropout scenarios.
• Proposal of efficient estimation and variance calculation procedures.

Main Results:

• The proposed maximum rank correlation estimator demonstrates desirable asymptotic properties.
• The methods are computationally stable and efficient for parameter and variance estimation.
• Numerical studies confirm the method's good performance in realistic settings.
• The approach was successfully applied to Parkinson's disease trial data.

Conclusions:

• The global percentile outcome provides a valuable tool for characterizing time-varying global disease burden in longitudinal studies.
• The developed regression strategies offer a flexible and robust approach to analyzing complex chronic disease data.
• The methodology effectively handles missing data and identifies risk factors for disease progression, as demonstrated in a Parkinson's disease cohort.