Moussa M A
Department of Community Medicine and Behavioural Sciences, Faculty of Medicine, Kuwait University, Safat.
Stat Med. 1988 May;7(5):559-69. doi: 10.1002/sim.4780070503.
Approximate allocation designs for comparing two survival distributions are considered and compared with the exact solution. The influence of incorporating covariates by stratification, accrual and follow-up duration on trial size and power is investigated. The Bernstein and Lagakos design is suited for exponential survival due to its generality and allowance for stratification while the Freedman design is convenient for distribution-free survival in view of its simplicity. With other specifications fixed, the saving in sample size becomes minimal as the accrual period or post-accrual follow-up time units expand above three. Required sample size also increases dramatically when the number of strata exceeds three.
考虑并比较了用于比较两种生存分布的近似分配设计与精确解。研究了通过分层、累积和随访持续时间纳入协变量对试验规模和检验效能的影响。伯恩斯坦（Bernstein）设计和拉加科斯（Lagakos）设计由于其通用性和允许分层，适用于指数生存情况，而弗里德曼（Freedman）设计鉴于其简单性，适用于无分布生存情况。在其他规格固定的情况下，随着累积期或累积后随访时间单位超过三个，样本量的节省变得最小。当分层数量超过三个时，所需样本量也会急剧增加。
