2种检测系统检测尿轻链K-Ig、λ-Ig的方法学比对及偏移评估  

Comparison of measurement procedure in two instruments for detecting urine K-Ig and λ-Ig light chains and bias estimation

作　　者：连丽丽[1] 王莹[1] 张天骄 许建成[1] LIAN Lili;WANG Ying;ZHANG Tianjiao;XU Jiancheng(Department of Medical Laboratory,the First Hospital of Jilin University,Changchun 130021,Jilin,China;Department of Otolaryngology Head and Neck Surgery,the First Hospital of Jilin University,Changchun 130021,Jilin,China)

机构地区：[1]吉林大学第一医院检验科,长春130021 [2]吉林大学第一医院耳鼻咽喉头颈外科,长春130021

出　　处：《临床检验杂志》2023年第1期50-55,共6页Chinese Journal of Clinical Laboratory Science

摘　　要：目的 分析2个检测系统尿轻链κ-Ig、λ-Ig结果,评价可比性、一致性及偏移。方法 分别用BNⅡ(参比方法,X)与BA400(待评方法,Y)检测162例尿κ-Ig与132例尿λ-Ig水平各1次。采用ESD法剔除离群值。散点图、偏差图及Bland-Altman图用于分析Y随X变化的大致趋势、方法间差值潜在变化特征及一致性。根据方法间差值分布状态使用平均值或中位数初步估算偏移。2组间比较采用配对样本Wilcoxon符号秩和检验。选择Deming和Passing-Bablok回归模型拟合散点图,计算医学决定水平处的偏移及可信区间,以≤±12.5%为可接受标准。结果 经ESD检验和散点图、偏差图分析筛选出146对尿κ-Ig与122对尿λ-Ig测量结果。散点图显示2种方法间测量值大致呈线性关系,偏差图显示方法间差值呈混合变化(SD和CV)。尿κ-Ig在低浓度与高浓度处估算偏移及其95%CI分别为-1.47(-1.89~-0.97)mg/L、5.98%(3.20%~8.76%);尿λ-Ig在低浓度与高浓度处估算偏移及其95%CI分别为1.70(1.26~2.15)mg/L、18.87%(14.03%~23.70%)。尿κ-Ig、λ-Ig测量结果呈非正态分布,经Wilcoxon检验方法间尿κ-Ig测量值差异无统计学意义(P>0.05),尿λ-Ig测量值差异有统计学意义(P<0.05)。一致性评价表明,2种方法间尿κ-Ig、λ-Ig的测量值差异有统计学意义(P<0.05);位于LoA范围内的点占94.52%~96.46%,但LoA及其95%CI均超出±12.5%,尚不能认为2种方法一致性良好。回归分析表明,尿κ-Ig、λ-Ig医学决定水平14.40 mg/L、7.80 mg/L处偏移的可信区间均符合可接受标准;回归方程截距和斜率95%CI表明2种方法间尿κ-Ig测定存在系统差异和比例差异,尿λ-Ig测定存在比例差异。结论 2种方法检测尿轻链存在差异,不能相互替代。一致性界限范围相差幅度在临床不可接受,尚不能认为2种方法一致性良好。Objective To analyze and evaluate the comparability,consistency and bias of measuement results for urine k-Ig andλ-Ig from two measurement procedures.Methods The levels of K-Ig in 162 cases andλ-Ig in 132 cases of urine specimens were measured once by BN II(comparative measurement procedure,X)and BA400(candidate measurement procedure,Y)respectively.ESD meth-od was used to eliminate outliers,scatter plots,difference plots.Bland-Altman plots were applied to analyze the general trend of Y val-ue changing with X value,the characteristics of underlying difference between two measurement procedures,and evaluate their consis-tency.Arithmetic mean or median were applied to estimate bias preliminary according to the distribution states of different values be-tween the two measurement procedures.The comparison between the two groups of paired samples was performed by Wilco.xon signed rank test.Deming and Passing-Bablok regression models were selected to fit scatter plots and calculate the biases and confidence inter-vals at medical decision levels.The values of less than or equal to+12.5%was considered as the acceptable criteria.Results The measured results of 146 pairs of urine k-Ig and 122 pairs of urine λ-Ig were screened respectively by ESD test,scatter plots and differ-ence plots.An approx imately linear relationship of the measured values between the two procedures was seen in scatter plots,but mixed variability(standard dev iation and coefficient of variation)appeared in difference plots.The biases with 95%CI of urine κ-g were esti-matedas-1.47(-1.89to-0.97)mg/L in low concentration and 5.98%(3.20%to 8.76%)in high concentration respectively.The biases with 95%CI of urine λ-Ig were estimated as 1.70(1.26 to2.15)mg/L in low concentration and 18.87%(14.03%to 23.70%)in high concentration respectively.The measured results of urine k-Ig andλ-Ig showed abnormal distribution.Wilcoxon test suggested that the measured results of urine K-Ig showed no significant difference between the two procedures(P>0.05),however the d

关 键 词：尿轻链 比对 医学决定水平 偏移 一致性 

分 类 号：R446[医药卫生—诊断学]