McShane L M, Clark L C, Combs G F, Turnbull B W
National Institutes of Health, Bethesda, MD.
Am J Clin Nutr. 1991 Jun;53(6):1354-60. doi: 10.1093/ajcn/53.6.1354.
Procedures for reporting and monitoring the accuracy of biochemical measurements are presented. They are proposed as standard reporting procedures for laboratory assays for epidemiologic and clinical-nutrition studies. The recommended procedures require identification and estimation of all major sources of variability and explanations of laboratory quality control procedures employed. Variance-components techniques are used to model the total variability and calculate a maximum percent error that provides an easily understandable measure of laboratory precision accounting for all sources of variability. This avoids ambiguities encountered when reporting an SD that may taken into account only a few of the potential sources of variability. Other proposed uses of the total-variability model include estimating precision of laboratory methods for various replication schemes and developing effective quality control-checking schemes. These procedures are demonstrated with an example of the analysis of alpha-tocopherol in human plasma by using high-performance liquid chromatography.
本文介绍了生化测量准确性的报告和监测程序。这些程序被提议作为流行病学和临床营养研究实验室检测的标准报告程序。推荐的程序要求识别和估计所有主要变异来源，并解释所采用的实验室质量控制程序。方差分量技术用于对总变异进行建模，并计算最大百分比误差，该误差提供了一个易于理解的实验室精度度量，涵盖了所有变异来源。这避免了报告标准差时可能只考虑少数潜在变异来源而产生的歧义。总变异模型的其他提议用途包括估计各种重复方案的实验室方法精度，以及制定有效的质量控制检查方案。通过使用高效液相色谱法分析人体血浆中α-生育酚的实例演示了这些程序。
