Related Experiment Videos
Efficacy evaluation for monotherapies in two-by-two factorial trials
H M Hung1, G Y Chi, R T O'Neill
1Division of Biometrics, CDER/FDA, Rockville, Maryland 20857, USA.
Biometrics
|December 1, 1995
Summary
Analyzing factorial clinical trials, new statistics offer improved estimation of individual treatment effects. The maximum and two-stage statistics provide better power and bias control, especially with negative interactions.
Area of Science:
- Biostatistics
- Clinical Trial Design
Background:
- Factorial clinical trials evaluate multiple treatments simultaneously.
- Assessing individual treatment effects requires careful statistical consideration, especially when interactions are present.
Purpose of the Study:
- To compare the performance of different test statistics for estimating simple treatment effects in factorial trials.
- To identify optimal statistical methods for handling treatment interactions.
Main Methods:
- Comparison of four test statistics: always-pooled, never-pooled, two-stage, and maximum.
- Evaluation based on power, bias, and mean square error.
Main Results:
- The always-pooled statistic has correct size but unbounded bias.
- The never-pooled statistic has poor precision.
- Two-stage and maximum statistics show superiority, particularly with negative interactions; the maximum statistic is slightly favored.
Conclusions:
- The maximum and two-stage statistics offer improved estimation of simple treatment effects in factorial trials.
- The two-stage statistic is recommended when large treatment interactions are anticipated.