|
Activity Number:
|
77
- Contributed Poster Presentations: Biopharmaceutical Section
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2020 : 10:00 AM to 2:00 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
|
Abstract #314054
|
|
|
Title:
|
On the Use of the Bootstrap to Test Functions of Correlated Variances in Pre-Post Clinical Trials
|
|
Author(s):
|
Navneet Hakhu* and Dan Gillen
|
|
Companies:
|
University of California, Irvine and University of California, Irvine
|
|
Keywords:
|
Clinical trials;
correlated variances;
testing;
heterogeneity;
bootstrap
|
|
Abstract:
|
In clinical trials we seek to minimize bias and variance to maintain trial integrity and reliably answer the pre-specified primary question of interest. Randomization serves to minimize bias, on average. Furthermore, prevention of missing data (e.g. drop out) will also reduce bias and reduce variance. Increased variability in response measures is well known to decrease precision of intervention estimates and lead to less efficient designs. We hypothesized that heterogeneity arises in Alzheimer’s disease (AD) trials from a participant’s enrolled study partner type (e.g. spouse vs. non-spouse) who completes assessments evaluating the participant’s cognition and function during the course of the trial (at minimum two correlated assessments: pre-randomization and post-randomization). We present empirical results for two estimators of the difference between a function of correlated study partner specific variances (i.e. difference in differences and difference in ratios) via the bootstrap to quantify finite sample properties and to compare against asymptotic results. Our methods are empirically evaluated via simulation and applied to data from two multi-center international AD trials.
|
Authors who are presenting talks have a * after their name.
