|
Activity Number:
|
384
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 2, 2016 : 11:35 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
|
Abstract #321728
|
|
|
Title:
|
Adjusting for Noncompliance in Randomized Clinical Trials When Noncompliance Must Be Estimated from a Biomarker
|
|
Author(s):
|
Jeffrey Boatman* and David Vock and Joseph S. Koopmeiners
|
|
Companies:
|
and University of Minnesota School of Public Health and University of Minnesota
|
|
Keywords:
|
Causal Inference ;
Randomized Clinical Trial ;
Noncompliance
|
|
Abstract:
|
Noncompliance is a pervasive problem in randomized clinical trials. Unadjusted analyses of only compliant participants give biased estimators of the causal treatment effect. There are a variety of statistical methods to adjust for noncompliance, but these methods all assume that noncompliance is known without error. This is frequently an untenable assumption as noncompliance is generally based on self-report, even with common measures of noncompliance such as the number of pills remaining in a returned bottle. A method that allows for valid inference when noncompliance is self-reported would be of great utility. We show how to use participants' biomarkers of compliance to estimate the probability of compliance and incorporate this into an inverse probability weighted estimator. This estimator provides valid causal inference provided that the biomarker is a valid measure of noncompliance. We apply our method to data from a recent clinical trial for regulatory tobacco research, and we demonstrate the small sample properties of the estimator via simulation.
|
Authors who are presenting talks have a * after their name.
