Related Experiment Videos
Detecting treatment-by-centre interaction in multi-centre clinical trials
R F Potthoff1, B L Peterson, S L George
1Cancer Center Biostatistics, Duke University Medical Center, P.O. Box 3958, Durham, North Carolina 27710, USA. dpotthoff@ccstat.mc.duke.edu
Statistics in Medicine
|February 13, 2001
Summary
This study evaluated permutation tests for treatment-by-centre interaction in clinical trials. Results show varying test performance, with generally low power to detect interactions, impacting reliability.
Area of Science:
- Biostatistics
- Clinical Trials Methodology
- Survival Analysis
Background:
- Multi-centre clinical trials are crucial for evaluating treatments.
- Detecting treatment-by-centre interaction is vital for understanding treatment generalizability.
- Survival data with censoring presents unique analytical challenges.
Purpose of the Study:
- To evaluate and compare the power of several permutation tests for treatment-by-centre interaction.
- To identify the most effective permutation tests under various simulation conditions.
- To address methodological challenges in power simulations for permutation tests.
Main Methods:
- Simulation of survival times and censoring times.
- Application and comparison of existing and novel permutation tests.
- Specialized simulation methodology to accommodate permutation test properties.
Main Results:
- Different simulation conditions identified different optimal interaction tests.
- One permutation test demonstrated consistently good power across most conditions.
- Overall power for detecting interaction was generally low across tested scenarios.
Conclusions:
- The choice of the best permutation test for treatment-by-centre interaction depends on specific trial conditions.
- While one test showed robust performance, others may also be valuable.
- Low power limits the reliable detection of treatment-by-centre interaction in many clinical trial settings.