Biostatistics Working Group

Spring 1999 -- Joint modeling of longintudinal and primary endpoint data

Description:

In the study of new treatments for disease and public health, data may be collected repeatedly over time on a number of variables, often referred to in this context as "markers," and on an endpoint of primary interest. For example, in the study of treatments for HIV/AIDS, observations on immunological and virological measures may be taken on each subject in a clinical trial repeatedly over the course of the trial, and the primary endpoint of interest may be the time to an event such as death or development of full-blown AIDS. Often, a key question is whether the nature of progression of such markers over time may be "associated with" the primary outcome somehow. The hope is that repeated measurement of the marker over a short term may provide information on the primary endpoint, which may take a much longer time to be observed, thus providing rationale for shorter studies where only repeated measurements on the marker are made.

Much recent interest has focused on formalizing this issue, the so-called "surrogate marker" problem, and on statistical models and methods to address this issue via joint modeling of the repeated measurement trajectories and the primary endpoint of interest. Modeling and inferential strategies for longitudinal data alone and for primary endpoints alone are well-developed; the key is unifying these methods. Research on such joint modeling is a current "hot" topic in the statistical literature, and the area is rapidly evolving. These seminars provide and introduction and overview of this problem.

Seminars and copies of slides:

• Modeling the relationship of survival to CD4 counts for patients with AIDS: Is CD4 a good surrogate marker? (1/28/99)

  ( Slides in PostScript )

• Joint Modeling of Longitudinal and Primary Endpoint Data -- An Introduction (2/25/99)

  ( Slides in PostScript )

  ( References on joint modeling in PostScript )

• Generalized Linear Models with Longitudinal Covariates (3/25/99)

  ( Slides in PostScript )

• Joint Modeling of Survival Data and Repeatedly Measured Covariates -- A Bayesian Perspective (3/25/99)

  ( Slides in PostScript )

Return to top .

Return to Biostatistics Working Group