ABSTRACT
Randomization ensures that, on average, both the known and unknown covariates are well balanced between the treatment groups. However, randomization does not guarantee such balance, particularly for a moderate-sized or small trial. Any such imbalance can give an unfair advantage to one treatment group over another if not accounted for in the analysis. The Cox proportional hazards regression model is usually used to adjust for covariates, such as age, gender, or disease stage, that may be associated with the survival outcome, thus confounding the treatment effect. This chapter derives a sample size formula for the score test under the contiguous alternative. The score test statistic based on the partial likelihood for the Cox regression model can be used to test the treatment effect. Schoenfeld derived a sample size formula for the score test for testing the treatment effect. Hsieh and Lavori extended Schoenfeld's result to the case of a non-binary covariate.
