Sample size determination for testing equality in frequency data under an incomplete block crossover design

ABSTRACT

When there are more than two treatments under comparison, we may consider the use of the incomplete block crossover design (IBCD) to save the number of patients needed for a parallel groups design and reduce the duration of a crossover trial. We develop an asymptotic procedure for simultaneously testing equality of two treatments versus a control treatment (or placebo) in frequency data under the IBCD with two periods. We derive a sample size calculation procedure for the desired power of detecting the given treatment effects at a nominal-level and suggest a simple ad hoc adjustment procedure to improve the accuracy of the sample size determination when the resulting minimum required number of patients is not large. We employ Monte Carlo simulation to evaluate the finite-sample performance of the proposed test, the accuracy of the sample size calculation procedure, and that with the simple ad hoc adjustment suggested here. We use the data taken as a part of a crossover trial comparing the number of exacerbations between using salbutamol or salmeterol and a placebo in asthma patients to illustrate the sample size calculation procedure.

KEYWORDS:

• Sample size determination
• incomplete block crossover design
• Type I error
• power
• frequency
• weighted least-squares estimator

Acknowledgements

The author wishes to thank the two referees for many valuable and helpful comments to improve the clarity and contents of this paper.

Disclosure statement

No potential conflict of interest was reported by the authors.