Piecewise Growth Mixture Tobit Models: Application to AIDS Studies.

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Clinical Trials

Background:

• Piecewise growth models offer flexibility in analyzing developmental phases in longitudinal data.
• Simultaneous modeling of heterogeneous growth, segmented progression, and left-censoring with skewed responses is challenging.
• Existing methods have limitations in handling complex longitudinal data structures.

Purpose of the Study:

• To present novel piecewise growth mixture Tobit models (PGMTMs).
• To describe phasic changes in individual trajectories over time.
• To address challenges in longitudinal data with mixed subpopulations and left-censoring.

Main Methods:

• Development of piecewise growth mixture Tobit models (PGMTMs).
• Application of PGMTMs to longitudinal data with left-censoring and subpopulations.
• Illustration using real-world data from an AIDS clinical study.

Main Results:

• Identified two distinct classes of viral load growth trajectories.
• Class 1: Initial viral load decline followed by rebound post-change-point.
• Class 2: Initial decline similar to Class 1, followed by a slower, sustained decrease.

Conclusions:

• PGMTMs effectively model complex longitudinal data with multiple features.
• The models provide insights into distinct viral load progression patterns in AIDS patients.
• This approach enhances understanding of treatment effects and disease dynamics.