A simulation study of the behaviour of the logrank test under different levels of stratification and sample sizes.

Abstract

In clinical trials, patients are enrolled into two treatment arms. A researcher may be interested in studying the effectiveness of a new drug or the comparison of two drugs for the treatment of a disease. This survival data is later analysed using the logrank test or the Cox regression model to detect differences in survivor functions. However, the power function of the logrank test depends solely on the number of patients enrolled into the study. Because statisticians will always minimise type I and type II errors, a researcher carrying out a clinical trial must define beforehand, the number of patients to be enrolled into the clinical study. Without proper sample size and power estimation a clinical trial may fail to detect a false hypothesis of the equality of survivor functions. This study presents through simulation, a way of power and sample size estimation for clinical trials that use the logrank test for their data analysis and suggests an easy method to estimate power and sample size in such clinical studies. Findings on power analysis and sample size estimation on logrank test are applied to two real examples: one is the Veterans' Administration Lung Cancer study; and the other is the data from a placebo controlled trial of gamma interferon in chronic granulotomous disease.