摘要
目的阐明定量方法对比研究中一致性评价Bland-Altman法进行LoA可信区间估计的必要性,探讨LoA可信区间估计的方法学性能。方法针对采用两种定量测量方法同时测量一份样本、每方法测量一次的情形,假定两种方法测量结果的差值服从标准正态分布,借助Monte-Carlo模拟,在不同的样本量下分别采用Bland-Altman法(简称B-A法)和MOVER法模拟1万次,获得95%LoA范围95%可信区间覆盖-1.96到1.96范围的比例(覆盖率),用以评价不同方法的性能,同时在临床可接受误差限值为-2.5到2.5范围的设定下判定用不同的可信区间方法达到一致的比例(达标率)加以佐证。结果直接采用LoA法评价一致性因没有考虑抽样误差影响,理论上不严谨、实用时太宽松。用B-A法估计LoA可信区间覆盖率具有样本量依赖性,而MOVER法覆盖率独立于样本量、始终紧紧围绕事先设定的可信度95%微小波动。结论直接用LoA判定定量方法对比研究中的一致性并不合适,采用MOVER法进行LoA可信区间估计用于评价一致性具有很好的可靠性,是一种值得推荐的方法。
Objective To illuminate the necessity of confidence interval estimation for the Bland - Altman limits of a- greement in quantitative method comparison studies, and to explore the performance of the confidence interval estimation for LoA. Methods This paper just focuses on the situation those two kinds of quantitative measurement methods for simultaneous measurement of a sample, and each methods measure only one time. Difference between the two measurement results is assumed to obey standard normal distribution,with the aid of the Monte - Carlo simulation in different sample sizes by Bland - Altman method （ B - A method） and MOVER method to simulate 10000 times,then obtained the coverage rates of 95% confidence in- tervals of 95% LoA range which covering the range of - 1.96 to 1.96, and evaluate the performance of different quantitative measurement methods. At the same time, corroborate with the qualified rates obtained by judging whether it reached an agreement with a different confidence interval methods under the settings of clinical acceptable error limit range of - 2.5 to 2.5. Results Directly using the LoA method to estimate the agreement for failing to consider the sampling error, in theory, is not rigorous, practical too loose. The coverage rates will change under different sample sizes by using the B - A method to estimate confidence interval for limits of agreement, while the coverage rates of the MOVER method independent of sample sizes, and closely around the small fluctuations in the credibility of 95% predefined. Conclusion It may be not appropriate to estimate the agreement by using the LoA method directly, while by using the MOVER method to estimate confidence interval for LoA has good reliability, is a recommended method.
出处
《中国卫生统计》
CSCD
北大核心
2014年第1期64-67,共4页
Chinese Journal of Health Statistics
