影响两类间接法建立34项生物化学分析物参考区间一致性的原因探讨

Explore the reasons affecting the consistency of reference intervals established by two types of indirect methods for 34 biochemical analytes

目的比较两类间接法建立34项生化分析物参考区间的差异,并探索影响两种方法一致性的可能因素。方法本研究为回顾性研究。基于2018年1月1日至2019年12月31日在北京协和医院进行体检的个体的白蛋白(Alb),碱性磷酸酶(ALP),丙氨酸转氨酶(ALT),载脂蛋白A1(ApoA1),载脂蛋白B(ApoB),天冬氨酸转氨酶(AST),钙(Ca),胆碱酯酶(ChE),氯(Cl),肌酐(Cr),高敏C反应蛋白(hsCRP),胱抑素C(CysC),直接胆红素(DBil),游离脂肪酸(FFA),糖化白蛋白(GA),谷氨酰基转移酶(GGT),血糖(Glu),高密度脂蛋白胆固醇(HDL-C),钾(K),乳酸脱氢酶(LD),低密度脂蛋白胆固醇(LDL-C),脂蛋白(a)[Lp(a)],钠(Na),磷(P),前白蛋白(PA),超氧化物歧化酶(SOD),总胆汁酸(TBA),总胆红素(TBil),总胆固醇(TC),总二氧化碳(TCO_(2)),甘油三酯(TG),总蛋白(TP),尿酸(UA),尿素(UR)的数据,使用Box-Cox算法改善数据分布后采用Tukey法识别异常值。建立方差成分模型并计算标准差比(SDR)以判断34项生化分析物是否需要按照年龄或性别分区建立参考区间。同时采用非参数法和kosmic算法建立34项生化分析物的参考区间及其90%的置信区间,比较两种方法参考区间界限值90%置信区间的重合情况。结果ALP(男性,女性18~59)、ALT、AST、hsCRP、DBil、GGT、Lp(a)、TBA、TBil、TG、Glu、HDL-C(男性)以及老年组CysC、GA、UR偏度系数偏离0,其峰度系数也偏离3,这些生化分析物使用两种方法建立的参考区间点估计差异大、90%的置信区间无重合。对于正态性较好的分析物Alb、ApoA1、ApoB、Ca、ChE、Cl、Cr(E)、CysC(18~59)、FFA、GA(18-59)、HDL-C(女性)、K、LDL-C、Na、P、PA、SOD、TC、TCO_(2)、TP及UR(18~59岁),2种方法参考区间点估计差异小且除钙外的分析物的界限值2种方法90%置信区间有重合,一致性好。结论数据正态性会影响不同间接法的一致性,正态性好的生化分析物2种方法的一致性好,反之2种方法的结果差异较大。

Objective To compare the differences of reference intervals(RI)established by two types of indirect methods for 34 biochemical analytes,and to explore the possible factors that affect the consistency of the two methods.Methods This was a retrospective study.Based on data of albumin(Alb),alkaline phosphatase(ALP),alanine aminotransferase(ALT),apolipoprotein A1(ApoA1),ApolipoproteinB(ApoB),aspartate aminotransferase(AST),calcium(Ca),cholinesterase(ChE),chloride(Cl),creatinine(Cr),high-sensitivity C-reactive protein(hsCRP),Cystatin(CysC),direct bilirubin(DBil),free fatty acid(FFA),glycated albumin(GA),gamma-glutamyltransferase(GGT),glucose(Glu),high density lipoprotein cholesterol(HDL-C),potassium(K),lactate dehydrogenase(LD),low density lipoprotein cholesterol(LDL-C),lipoprotein a[Lp(a)],sodium(Na),phosphorus(P),prealbumin(PA),superoxide dismutase(SOD),total bile acid(TBA),total bilirubin(TBil),total cholesterol(TC),total carbon dioxide(TCO_(2)),triglyceride(TG),total protein(TP),uric acid(UA)and urea(UR)of individuals who underwent physical examination at Peking Union Medical College Hospital from January 1,2018 to December 31,2019,Box-Cox algorithm was used to improve the data distribution and Tukey method was used to identify outliers.Variance component model was established,and standard deviation ratio(SDR)was calculated to determine whether the RIs of 34 biochemical analytes should be established according to age or sex The non-parametric method and kosmic algorithm were used to establish the RIs and 90%confidence intervals(CIs)of 34 biochemical analytes,and the coincidence of the 90%CIs of the reference limits for two methods was compared.Results The skewness coefficients of ALP(male,female18-59),ALT,AST,hsCRP,DBil,GGT,Lp(a),TBA,TBil,TG,Glu,HDL-C(male)and CysC,GA,UR in the elderly group deviated from 0,and their kurtosis coefficients also deviated from 3.For these biochemical analytes,the point estimates of the RIs established by the two methods differed greatly and the 90%CIs did not overlap.The analytes with good normality were Alb,ApoA1,ApoB,Ca,ChE,Cl,Cr(E),CysC(18-59),FFA,GA(18-59),HDL-C(female),K,LDL-C,Na,P,PA,SOD,TC,TCO_(2),TP and UR.The consistency is good.Except for Ca,90%CIs of reference limits for some analyte between the two methods coincide with each other.Conclusions The consistency of different indirect methods is affected by the normality of data.