Ichihara K, Kawai T
Department of Clinical Pathology, Kawasaki Medical School, Kurashiki, Japan.
J Clin Lab Anal. 1996;10(2):110-7. doi: 10.1002/(SICI)1098-2825(1996)10:2<110::AID-JCLA9>3.0.CO;2-G.
By using a new international reference preparation for the plasma proteins (CRM470) prepared by IFCC, we launched a project to determine reference intervals for 13 plasma proteins (immunoglobulins, compliment components, transferrin, alpha 1antitrypsin, alpha 1acid glycoprotein, haptoglobin, alpha 2macroglobulin, ceruloplasmin, albumin, transthyretin) in accordance with the guideline proposed by NCCLS. As reference individuals, 999 subjects were first selected through medical examination and a health-check questionnaire. However, there were many who had "abnormal" values in some of the 25 screening tests results measured simultaneously (serum enzymes, lipids, blood counts, etc.). Therefore, we adopted a criterion that those individuals who had "abnormal" values (beyond 99% confidence intervals) in the 25 tests would be excluded. The exclusion resulted in appreciable narrowing of the distributions in all 13 plasma proteins. This implied that "apparent" normality of major screening tests results may be used as a criterion to select truly relevant reference individuals. A parametric approach using a modified Box-Cox power transformation formula and the maximal likelihood estimation led to almost perfect normalization of the reference distributions on all the items tested. Since the nonparametric method proposed by the NCCLS guideline gave less reproducible results in our simulation study, the parametric method appears to be a method of choice for the calculation.
通过使用国际临床化学和检验医学联合会(IFCC)制备的新型血浆蛋白国际参考制剂(CRM470),我们启动了一个项目,按照美国国家临床实验室标准委员会(NCCLS)提出的指南,确定13种血浆蛋白(免疫球蛋白、补体成分、转铁蛋白、α1抗胰蛋白酶、α1酸性糖蛋白、触珠蛋白、α2巨球蛋白、铜蓝蛋白、白蛋白、甲状腺素转运蛋白)的参考区间。作为参考个体,首先通过医学检查和健康调查问卷筛选出999名受试者。然而,在同时进行的25项筛查检测结果(血清酶、血脂、血细胞计数等)中,许多人的某些检测结果呈现“异常”。因此,我们采用了一项标准,即排除那些在这25项检测中出现“异常”值(超出99%置信区间)的个体。这种排除使得所有13种血浆蛋白的分布明显变窄。这意味着主要筛查检测结果的“表观”正常性可作为选择真正相关参考个体的标准。使用修正的Box-Cox幂变换公式和最大似然估计的参数方法,使所有检测项目的参考分布几乎完全正态化。由于在我们的模拟研究中,NCCLS指南提出的非参数方法给出的结果重复性较差,因此参数方法似乎是计算的首选方法。