|
Activity Number:
|
199
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, July 30, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #309378 |
|
Title:
|
Controlling for Time-Dependent Confounding Using Marginal Structural Models in the Case of a Continuous Treatment Covariate
|
|
Author(s):
|
Ouhong Wang*+ and Trevor McMullan
|
|
Companies:
|
Amgen Inc. and Validant Consulting
|
|
Address:
|
One Amgen Center Drive, Thousand Oaks, CA, 91320,
|
|
Keywords:
|
Time Dependent Confounding ; Marginal Structural Models ; Inverse Probability of Treatment Weighting ; Counterfactuals
|
|
Abstract:
|
Marginal Structural Models, whose parameters are estimated by Inverse Probability of Treatment Weighting (IPTW), can be used as an alterative to g-estimation to adjust for time dependent confounding, eliminating the need to construct the full counterfactural dataset. When the treatment variable is binary, the number of counterfacturals that need to be generated are 2**t per patient where t is the number of time points. When the treatment variable is continuous the number of counterfacturals needed is no longer finite. In these situations the performance of IPTW is not well characterized. Here we model time dependent confounding using IPTW and a continuous treatment variable with both observed data and bootstrapped simulated data. These results are compared to standard time dependent Cox proportional hazards models, as well as IPTW results with ordinal treatment.
|
