Related Experiment Videos
Modelling and design of cross-over trials
1Department of Medical Statistics, School of Computing Sciences, De Montfort University, Leicester.
Statistics in Medicine
|July 15, 1996
Summary
This study introduces a computer-aided approach for creating custom crossover clinical trial designs. This method ensures designs are optimized for specific trial objectives, moving beyond generic, potentially ill-fitting, published designs.
Area of Science:
- Clinical Trials
- Biostatistics
- Experimental Design
Background:
- Crossover clinical trials compare treatments using within-subject repeated measurements.
- Traditional design selection relies on published tables or simple rules, which may not fit specific trial needs or models.
- Trial restrictions can necessitate compromises, leading to suboptimal designs.
Purpose of the Study:
- To develop a method for constructing tailor-made crossover clinical trial designs.
- To move beyond 'off-the-peg' designs that may not align with trial objectives or data models.
- To provide a flexible approach adaptable to various trial constraints and objectives.
Main Methods:
- Utilizing a computer algorithm to search for optimal crossover trial designs.
- Employing A-optimality as a criterion for design optimization, though adaptable to others.
- Selecting statistical models that suit the specific trial's characteristics and data.
Main Results:
- Demonstrated the construction of customized designs through illustrative examples.
- Showcased the application to scenarios with specific treatment sequences and subject restrictions.
- Illustrated designs for comparing two active treatments and a placebo, considering similar carry-over effects.
Conclusions:
- Tailor-made designs generated by computer algorithms offer a superior alternative to standard designs.
- This approach enhances the suitability and efficiency of crossover clinical trials.
- The methodology is versatile, accommodating diverse trial objectives and complexities, including design augmentation.