Location-Scale Latent Process Model for Repeated Ordinal Patient-Reported Outcomes.

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Longitudinal Data Analysis

Background:

• Patient-reported outcomes (PROs) are crucial for assessing quality of life in clinical trials.
• Traditional analysis of PROs often overlooks their longitudinal and ordinal characteristics.
• Electronic data collection enables frequent, daily PRO measurements, necessitating advanced statistical methods.

Purpose of the Study:

• To develop and validate a statistical model for analyzing frequent ordinal longitudinal PRO data.
• To investigate the dynamics of symptom scores and their variability over time.
• To evaluate the impact of treatments on symptom progression in clinical trials.

Main Methods:

• Proposed a location-scale latent process model to capture mean structure and variability of ordinal PROs.
• Incorporated random effects for individual patient trajectories and covariates for short-term variability.
• Estimated the model using maximum likelihood with Quasi-Monte Carlo approximation in R.
• Validated the method via simulation and applied it to asthma and COPD clinical trial data.

Main Results:

• The proposed model effectively analyzes the dynamics of ordinal PROs, considering both mean trends and variability.
• Demonstrated the model's ability to assess treatment effects on symptom progression and variability in clinical trials.
• Successfully applied the methodology to real-world data from asthma and COPD studies.

Conclusions:

• The location-scale latent process model provides a robust framework for analyzing frequent ordinal longitudinal PRO data.
• This approach enhances understanding of disease progression and treatment efficacy by capturing complex data dynamics.
• The validated methodology offers a valuable tool for clinical trial analysis, particularly for respiratory conditions.