Sample size determination in stratified trials to establish the equivalence of two treatments.

When designing a trial to establish that a new treatment is as effective as a standard one, the conventional test procedure and sample size based on a null hypothesis of no difference between two treatments is inappropriate. Several authors have investigated test statistics and corresponding sample sizes based on the null hypothesis that the standard treatment is more effective than the new by at least some specific value for a single 2 x 2 table. This paper considers a trial that involves several 2 x 2 tables and presents an approximate formula for the sample size required to obtain a given power of a one-tailed score test for a null hypothesis of a specific common non-zero difference between two treatments across strata. I show that the sample size for a trial based on an unstratified test is always larger than that based on a stratified test when the design is balanced.

[1]  W. G. Cochran.  Some Methods for Strengthening the Common χ 2 Tests , 1954 .

[2]  M. Carcangiu.  Re: Endometrial cancer in tamoxifen-treated breast cancer patients: findings from the National Surgical Adjuvant Breast and Bowel Project (NSABP) B-14. , 1994, Journal of the National Cancer Institute.

[3]  Calyampudi R. Rao.  Large sample tests of statistical hypotheses concerning several parameters with applications to problems of estimation , 1948, Mathematical Proceedings of the Cambridge Philosophical Society.

[4]  J J Gart,et al.  Approximate interval estimation of the difference in binomial parameters: correction for skewness and extension to multiple tables. , 1990, Biometrics.

[5]  W C Blackwelder,et al.  "Proving the null hypothesis" in clinical trials. , 1981, Controlled clinical trials.

[6]  C. Dunnett,et al.  Significance testing to establish equivalence between treatments, with special reference to data in the form of 2X2 tables. , 1977, Biometrics.

[7]  A. D. Schryver,et al.  Systemic treatment of early breast-cancer by hormonal, cytotoxic, or immune therapy: 133 randomized trials involving 31000 recurrences and 24000 deaths among 75000 women: 1 , 1992 .

[8]  M. Bartlett,et al.  APPROXIMATE CONFIDENCE INTERVALS , 1953 .

[9]  C. Farrington,et al.  Test statistics and sample size formulae for comparative binomial trials with null hypothesis of non-zero risk difference or non-unity relative risk. , 1990, Statistics in medicine.

[10]  C. Comnougue,et al.  How to establish equivalence between treatments: a one-sided clinical trial in paediatric oncology. , 1989, Statistics in medicine.

[11]  P. A. R. Koopman,et al.  Confidence intervals for the ratio of two binomial proportions , 1984 .

[12]  D. Sackett,et al.  The Burlington randomized trial of the nurse practitioner. , 1974, The New England journal of medicine.

[13]  Anthony Howell,et al.  Systemic treatment of early breast cancer by hormonal, cytotoxic, or immune therapy. 133 randomised trials involving 31,000 recurrences and 24,000 deaths among 75,000 women. Early Breast Cancer Trialists' Collaborative Group. , 1992 .

[14]  T. A. Brown,et al.  Theory of Equations. , 1950, The Mathematical Gazette.

[15]  Homogeneity score tests with nuisance parameters , 1988 .

[16]  M Nurminen,et al.  Comparative analysis of two rates. , 1985, Statistics in medicine.

[17]  M. Bartlett,et al.  APPROXIMATE CONFIDENCE INTERVALSMORE THAN ONE UNKNOWN PARAMETER , 1953 .

[18]  S. Radhakrishna,et al.  Combination of results from several 2 X 2 contingency tables , 1965 .