Accommodating informative dropout and death: a joint modelling approach for longitudinal and semi-competing risks...

Area of Science:

• Biostatistics
• Epidemiology
• Longitudinal Data Analysis

Background:

• Longitudinal studies are crucial for tracking health outcomes over time.
• Patient dropout and death can bias observed trends by removing data.
• Accurate analysis requires methods that handle these semi-competing events effectively.

Purpose of the Study:

• To develop a statistical framework for analyzing longitudinal outcomes in the presence of informative dropout and death.
• To enable the estimation of the longitudinal outcome profile conditional on survival.
• To apply the method to CD4 count data in HIV research.

Main Methods:

• A likelihood-based approach is proposed for joint modeling.
• The model incorporates longitudinal outcomes and semi-competing event times (dropout and death).
• A key feature is the closed-form derivation of the conditional longitudinal profile for living subjects.

Main Results:

• The proposed method effectively accommodates informative dropout and death.
• It allows for the estimation of the longitudinal CD4 count profile for living patients.
• The approach provides a more accurate representation of disease progression in HIV-infected individuals.

Conclusions:

• The joint modeling approach offers a robust solution for longitudinal data with semi-competing risks.
• This method enhances the understanding of disease trajectories by focusing on the alive population.
• It is particularly valuable for long-term health studies where informative censoring is common.