生存率置信区间的五种估计方法

Five Methods for the Estimation of Confidence Intervals on Survival Rate

生存率是医学随访研究资料分析中常用的指标，例如适用于小样本资料的Ｋａｐｌａｎ－Ｍｅｉｅｒ乘积限估计和大样本资料的寿命表法生存率估计。本文对生存率置信区间的估计方法进行了讨论。主要介绍了五种置信区间的估计方法：经典法（基于Ｇｒｅｅｎｗｏｏｄ方差公式）、校正法、反正旋转换法、ｌｏｇ（－ｌｏｇ）转换法及ｌｏｇｉｔ转换法。文中给出了两个实例，并就生存率９５％置信区间的估计做了详细介绍，还进一步讨论了它们在小样本和大样本随访资料分析中的特点和应用。

Descriptive analysis on medical follow-up data often involves estimation of survival rates, such as Kaplan-Meier product limit estimator or life table survival rate for small or large scale data sets,respectively.Five methods for estimating the confidence intervals(CI)of survival rates were described.These methods for estimation of intervals were the classic(based on Greenwood′s for-mula),correctness,arcsine transfomation,log(-log)transfomation and logit transfomation.Two examples were described in detailed for the calculation of 95%confidence intervals.