|
Activity Number:
|
576
|
|
Type:
|
Invited
|
|
Date/Time:
|
Thursday, August 6, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #303075 |
|
Title:
|
Survival Analysis with Dynamic Marginal Structural Models
|
|
Author(s):
|
Miguel A. Hernan*+
|
|
Companies:
|
Harvard School of Public Health
|
|
Address:
|
Epidemiology, Boston, MA, 02115,
|
|
Keywords:
|
causal inference ; dynamic regimes ; marginal structural models ; inverse probability weighting
|
|
Abstract:
|
When estimating the effect of time-varying treatments, inverse probability (IP) weighting can appropriately adjust for confounding due to time-varying covariates that are affected by prior treatment. Most applications of IP weighting of marginal structural Cox models have focused on static treatment regimes. For example, several articles compared the AIDS-free survival of HIV-infected patients under the static regimes "initiate highly active antiretroviral therapy (HAART) at baseline" vs "never initiate HAART during the follow-up." However, the comparison of dynamic treatment regimes is often more interesting. IP weighting can be used to compare dynamic regimes by combining it with artificial censoring (an inefficient approach), or as described by Orellana et al (2007) for mean models. This talk describes an application of a dynamic marginal structural model to the failure time setting.
|
