单因素方差解析及其在药品检测实验室中的应用

One-way Analysis of Variance and Its Application in Drug Testing Laboratory

目的:利用单因素随机效应模型的实验设计,形成一个通用模板,为读者快速理解和计算方法验证中的精密度、实验室的能力评定标准差、标准物质的协作标定赋值、以及样品均匀性判断中所需的参数提供帮助。方法:建立一个通用的单因素随机效应模型,通过方差解析获得组间均方(MS_A)与组内均方(MS_E),由这两个参数进一步推导出所需参数的运算公式并进行实例分析。结果:确定分析方法精密度的中间精密度或再现性标准差和重复性标准差的计算公式分别为σ_(IP)、σ_(Rep)=(1/rMS_A+((r-1)/r)MS_E)^(1/2)、σ_r=S_r=MS_E^(1/2);在进行实验室能力评定或标准物质标定过程中,样本间标准差是用于判断均匀性的指标,其计算公式为S_L=((MS_A-MS_E)/r)^(1/2);实验室能力评定标准差为σ_(PT)=((MS_A)/r)^(1/2);标准物质协作标定中的标准不确定度μ()为μ()=((MS_A)/cr)^(1/2),置信区间为(=x)±t_(1-ɑ:c-1)μ()=(=x)±t_(1-ɑ:c-1)·((MS_A)/cr)^(1/2)。结论:掌握单因素方差分析的原理、设计和分析方法,对药品质控实验室评估所用方法的变异、评价产品均匀性和标准物质的不确定度、以及科学评价分析实验室内部人员的检测能力等方面都必不可少。

Objective： A general template was created by using the experimental design of one factor randomeffects model, which would be helpful for readers to quickly understand and calculate the precision for the method validation, the standard deviation of proficiency assessment of the laboratory, the assignment of standard materials in the collaborative calibration, and to help to determine the uniformity of the sample. Methods： A general one factor random-effects model was established. The mean square between groups（MS_A） and the mean square within group（MS_E） were obtained from the analysis of variance. The two parameters were further deduced to obtain the function of the parameters needed and the related examples were demonstrated. Results：The formula for determining the intermediate precision or reproducibility standard deviation and the standard deviation of repeatability of the precision of the analytical method was calculated as σ_（IP） or σ_（Rep）=（1/rMS_A＋（（r-1）/r）MS_E）^（1/2）,σ_r=S_r=MS_E^（1/2）; in the course of laboratory proficiency assessment or standard substance calibration, the standard deviation between samples that was the statistic for evaluating the homogeneity was calculated as S_L=（（MS_A-MS_E）/r）^（1/2）; the standard deviation of proficiency testing was σ_（PT）=（（MS_A）/r）^（1/2）; the standard uncertainty μ（） in the collaborative calibration of standard materials was μ（）=（（MS_A）/cr）^（1/2）, confidence interval was （=x）±t_（1-ɑ：c-1）μ（）=（=x）±t_（1-ɑ：c-1）·（（MS_A）/cr）^（1/2）. Conclusion： It was essential to master the principle, design and analysis methods of one-way analysis of variance for assessing the variation of testing methods, evaluation of homogeneity of product and the uncertainty of the reference materials, as well as scientific evaluation and analysis of testing abilities of personnel and other aspects in drug testing laboratory.