Analysis of multivariate failure-time data from HIV clinical trials

Area of Science:

• Biostatistics
• Epidemiology
• Clinical Trials

Background:

• Analysis of multivariate failure-time data is complex, especially with composite endpoints like AIDS or death in HIV infection.
• Composite endpoints often include numerous events of varying severity, necessitating specialized analytical approaches.
• Existing trials may lack the power to detect treatment effects on individual events within a composite endpoint.

Purpose of the Study:

• To illustrate the application of marginal methods for analyzing multivariate failure-time data in HIV clinical trials.
• To investigate if combination antiretroviral therapy delays the development of new AIDS events.
• To present treatment effects on multiple disease processes in an easily understandable manner.

Main Methods:

• Utilized marginal methods for multivariate failure-time data analysis.
• Employed two approaches: grouping events and using only the first event, or fitting separate baseline hazards for all events.
• Constructed model-based or minimum-variance estimates of overall treatment effects and used covariance matrices for multiple testing.

Main Results:

• Analysis of the Delta trial data suggests combination antiretroviral therapy (ART) with AZT plus ddI or ddC may delay progression to more severe AIDS events compared to AZT monotherapy.
• Identified that these late-stage AIDS events are generally untreatable and lack available prophylaxis.
• Acknowledged that trial analyses are exploratory due to limitations in detecting effects on individual events.

Conclusions:

• Marginal multivariate models offer a practical approach for modeling covariate effects on multiple disease processes.
• These models facilitate the presentation of treatment effects in an easily interpretable format.
• Marginal models are versatile for exploring diverse treatment effect patterns and testing multiple hypotheses across various composite endpoints.