Portal de Periódicos da CAPES

The effect of stratified randomization on size and power of statistical tests in clinical trials

1978; Pergamon Press; Volume: 31; Issue: 6-7 Linguagem: Inglês

10.1016/0021-9681(78)90008-5

1878-0679

Sylvan B. Green, David P. Byar,

Optimal Experimental Design Methods

Stratified randomization is often used to ensure comparability between treatment groups in a clinical trial. An analytic formulation of the effect of such stratification on the probability of Type I and Type II error is presented, based on calculation of exact binomial probabilities. Normal approximations which can be quickly and easily calculated give quite satisfactory results. Numerical results for investigation of Type I error compare favorably with the computer simulations of Feinstein and Landis. These results are extended to power calculations. It is shown how the Feinstein-Landis model must be modified to correct the variance for the effect of stratified randomization. The advantages of the analytic approach include greater insight into the problem, and easy calculation of probability of Type I and Type II error for any choice of sample size, strata, and alternative hypotheses.