Westgard J O, Wiebe D A
Department of Pathology and Laboratory Medicine, Medical School, University of Wisconsin, Madison 53792.
Clin Chem. 1991 Nov;37(11):1938-44.
Current U.S. governmental regulations and requirements for the quality of laboratory tests do not provide a consistent form, comparable numbers, or practical specifications for the routine operation of laboratory testing processes. For cholesterol, as an example, the Health Care Financing Administration provides an analytical performance criterion for proficiency testing to enforce the Clinical Laboratory Improvement Act (CLIA), whereas the U.S. National Cholesterol Education Program (NCEP) provides clinical guidelines for test interpretation, as well as analytical goals for imprecision and inaccuracy. Routine operating process specifications for imprecision, inaccuracy, and quality control can be derived from the analytical and clinical requirements for quality. Use of an analytical "total error" model and a clinical "decision interval" model provides logically consistent and numerically comparable specifications. Studies with these coherent models indicate that a cholesterol testing process properly planned to satisfy the CLIA analytical requirement will also satisfy the NCEP clinical requirement. To provide 90% assurance of detecting systematic shifts of a magnitude that would cause the CLIA analytical requirement to be exceeded, the operational specifications for a cholesterol testing process are an allowable CV of less than or equal to 2%, an allowable bias of less than or equal to 1%, and a control procedure with two measurements per run interpreted by 1(3)s, 1(2.5)s, or 1(3)s/2(2)s/R4s control rules.
美国现行政府对实验室检测质量的法规和要求，并未为实验室检测过程的常规操作提供统一的形式、可比较的数据或实用的规范。以胆固醇检测为例，医疗保健财务管理局为能力验证提供分析性能标准，以执行《临床实验室改进法案》（CLIA），而美国国家胆固醇教育计划（NCEP）则提供检测结果解读的临床指南，以及不精密度和不准确的分析目标。不精密度、不准确和质量控制的常规操作过程规范可从质量的分析和临床要求推导得出。使用分析“总误差”模型和临床“决策区间”模型可提供逻辑一致且数值可比的规范。采用这些连贯模型的研究表明，一个经过合理规划以满足CLIA分析要求的胆固醇检测过程，也将满足NCEP临床要求。为提供90%的把握度以检测到可能导致超出CLIA分析要求的系统性偏移，胆固醇检测过程的操作规范为：允许变异系数（CV）小于或等于2%，允许偏差小于或等于1%，以及采用每次运行两次测量的控制程序，并根据1(3)s、1(2.5)s或1(3)s/2(2)s/R4s控制规则进行解读。
