|
Activity Number:
|
260
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #305141 |
|
Title:
|
Post-Randomization Interaction Analyses in Clinical Trials with Standard Regression
|
|
Author(s):
|
Rongmei Zhang*+ and Jennifer Faerber and Thomas Ten Have
|
|
Companies:
|
University of Pennsylvania and University of Pennsylvania and University of Pennsylvania
|
|
Address:
|
503 Blcokley Hall, Philadelphia, PA, 19104,
|
|
Keywords:
|
post-randomization ; interaction ; unmeasured confounding ; sequential ignorability
|
|
Abstract:
|
We address several questions on analyzing how post-randomization factors may modify the intent-to-treat effects of randomized interventions. We investigate the assumptions underlying the standard regression model with main effects and interactions for the baseline randomized intervention and post-randomization effect modifier. The crucial assumptions are sequential ignorability and no effect of the baseline intervention on the post-randomization. We present analytic results for all terms in the model under different combinations of the assumptions. In addition, we confirm our results with simulations and further assess our results through a randomized cognitive behavioral therapy (CBT) trial example. We show that there is little bias for the interaction term and the stratified intervention effect when the above assumptions are violated.
|
