Activity Number:
|
63
- Power and Sample Size: Methods and Applications
|
Type:
|
Contributed
|
Date/Time:
|
Monday, August 3, 2020 : 10:00 AM to 2:00 PM
|
Sponsor:
|
Biopharmaceutical Section
|
Abstract #313079
|
|
Title:
|
Improved Analyses of Randomized Clinical Trials with Stratified Enrollment
|
Author(s):
|
Devan Mehrotra* and Rachel Marceau West and Julie Kobie
|
Companies:
|
Merck and Merck Research Laboratories and Merck & Company Inc
|
Keywords:
|
adaptive weights;
additive model;
estimand;
stratification;
treatment by stratum interaction
|
Abstract:
|
In a typical randomized clinical trial comparing two treatments, enrollment is often pre-stratified by one or more factors that are expected to influence the response to either treatment. Data from such trials are commonly analyzed using an additive statistical model with main effect terms for treatment and stratum. We caution that this approach is needlessly risky because a departure from the embedded assumption of no treatment by stratum interaction can result in a notably biased and/or less powerful analysis. An alternative approach is proposed in which first the treatment effect is estimated separately for each stratum, and then the stratum-level estimates are combined using a novel ‘adaptive’ weighting strategy for overall estimation and inference. The advantages of our proposed analysis over the common additive model analysis are illustrated using real datasets and simulations for binary and time-to-event endpoints.
|
Authors who are presenting talks have a * after their name.