Abstract
One of the most important issues in the area of clinical trial research is the determination of the sample size required to insure a specified power in detecting a real or clinically relevant difference of a stated magnitude. Increasingly, medical journals are requiring authors to provide information on the sample size needed to detect a given difference. We restrict our attention to the designs far comparirng two survival distributions. These are concerned with the survival time which is defined as the interval from a baseline(e.g. randomization) to failure (e.g. death, recurrence of disease). Survival times axe right censored when patients have not foiled by the time of analysis or have been loss to follow-up during the trial. For different types of clinical trials for comparing survival distributions, there have been marry research in sample size determination. We review the existing literature concerning commonly used sample size formulae in the design of randomized clinical trials, and compare the assumption, the power and the sample size calculation of these methods. We also compare by simulation the expected power and observed power of each method under various circumstances. As a result, guidelines in terms of practical usage are provided.
임상시험연구의 설계에서 적절한 표본수의 결정은 매우 중요한 문제 중의 하나이다. 본 논문에서는 생존분포를 비교하기 위한 여러 가지 방법들을 소개하고, 각 방법에서의 가정들을 고찰하였다. 또한 다양한 상황에서의 표본수와 검정력 등을 비교제시하고 모의실험을 통해 각 방법들의 이론상의 검정력과 실제 검정력을 알아보았다 그 결과로서 의학연구자들이 처한 여러 상황에 적합한 표본수의 결정방법을 제시하였다.