|
Activity Number:
|
62
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Sunday, July 29, 2007 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
WNAR
|
| Abstract - #309946 |
|
Title:
|
Estimating the Effect of Treatment on Rates of Unintended Events in Rheumatoid Arthritis: A Case History with Caveats
|
|
Author(s):
|
Mark Lunt*+ and Daniel H. Solomon and Deborah P.M. Symmons and Til Stürmer
|
|
Companies:
|
University of Manchester and Brigham and Women's Hospital and University of Manchester and Brigham and Women's Hospital
|
|
Address:
|
Division of PharmacoEpidemiology, Boston, MA, 02120,
|
|
Keywords:
|
Propensity Score ; effect modification ; confounding
|
|
Abstract:
|
Observational studies can provide answers to questions of clinical importance that cannot be answered through a clinical trial. However, such studies can be misleading if imbalances in potential confounders between treated and untreated subjects are not accounted for appropriately. We will present an example of estimating the effect of anti-TNF treatment on mortality in RA patients. A variety of different propensity score (PS) methods were used to balance potential confounders between treated and untreated, giving rise to widely varying estimates of the treatment effect. These different estimates were shown to be due to effect modification across the PS, with different estimators averaging the treatment effect over different distributions of the PS. The results illustrate the importance of assessing effect modification across the PS that can be real or due to unmeasured confounding.
|
