Klee G G
Department of Laboratory Medicine and Pathology, Mayo Clinic, Rochester, MN 55905.
Clin Chem. 1993 Jul;39(7):1514-8.
I propose a method for defining tolerance limits for assay bias and assay imprecision, based on the effects of these tolerance limits on the clinical specificity of the assay. An analytical "error budget" is defined as the squared sums of the imprecision and bias errors. The maximum limit for this error budget is set at a value corresponding to a 50% increase in the false-positive rate for classifying healthy subjects. For gaussian distributions with +/- 2 SD used as decision limits, this error budget equates to 0.45 SD of combined within-person and between-person biological variation (SDBiol). To provide reasonable power for bias detection in an assay, I recommend that the SD of the assay be kept at less than half the bias limit. Then, for the gaussian distribution, the maximum bias limit should be < 0.36 SDBiol and the SD of the assay should be < 0.18 SDBiol. Procedures are provided for using the same principles to define tolerance limits for decision limits other than +/- 2 SD and for nongaussian distributions.
我提出了一种基于这些公差限对检测临床特异性的影响来定义检测偏差和检测不精密度公差限的方法。分析“误差预算”定义为不精密度误差和偏差误差的平方和。该误差预算的最大限值设定为与将健康受试者分类的假阳性率增加50%相对应的值。对于使用±2标准差作为决策限的高斯分布，该误差预算相当于个体内和个体间生物学变异合并标准差（SDBiol）的0.45倍标准差。为了在检测中提供合理的偏差检测能力，我建议将检测的标准差保持在偏差限的一半以下。然后，对于高斯分布，最大偏差限应<0.36 SDBiol，检测的标准差应<0.18 SDBiol。还提供了使用相同原理为除±2标准差以外的决策限以及非高斯分布定义公差限的程序。
