动静脉同步换血换血量与胆红素水平关系的回归再分析

Regression Reanalysis on the Relationship between Bilirubin Volume and Exchange Transfusion Volume

目的通过对加大样本量的线性回归分析,进一步研究适用于一般临床外周动静脉同步换血应用的反映换血量和总胆红素(TBIL)下降关系的回归方程。方法对75例采用外周动静脉同步换血疗法的高胆红素血症新生儿,按换血前TBIL数值高低依序分为5组,对每换血50 ml/kg进行TBIL值检测和回归分析,并对回归方程参数的95%置信区间进行分析和检验。结果 (1)每增加50 ml/kg换血量,TBIL累积换出率差异有非常显著意义(P<0.01);经200 ml/kg换血后TBIL的累积换出率均值为61.5%。(2)换血过程中TBIL浓度(y)与换血量(x)呈线性相关(r值在0.96～0.98之间,P<0.001)。(3)分组系列回归方程的斜率(a)与换血前TBIL浓度初始值(y0)也呈高度线性相关,其(r=0.99,P<0.001)。(4)各组方程参数回归效果的方差分析检验结果均有非常显著统计学意义(P<0.01)。(5)对换血回归方程参数适用区间95%置信度的检验结果有非常显著统计学意义(P<0.001)。结论经分别对换血过程TBIL浓度值y与换血量x相关性依模型y=y0-ax以及对方程斜率参数a与换血前TBIL浓度初始值y0相关性依模型a=αy0+β进行的线性回归分析,可建立适用于临床指导换血用量的综合单自变量线性回归方程y=y0-ax=y0-(0.0035 y0-0.081)x。

Objective To further study and explore the regression equation between the decrease of serum total bilirubin （TBIL） volume and the exchange transfusion volume via peripheral vessels, through linear regression analysis in the large sample size. Methods 75 newly born infants with hyper- bilirubinaemia were divided into 5 groups according to the level of their TBIL volume before exchange transfusion. TBIL was measured after each exchange of 50 ml/kg. The 95% of the regression coefficient＇ s confidence interval was examined and analized. Results （1） The accumulative clearance rate of TBIL after each exchange of 50 ml/kg had significant difference （ P 〈 0. 01 ）. It was about 61.5% when exchanged 200 ml/kg. （2） The TBIL volume （y） showed linear relationship with the exchange transfusion volume （x） during the blood exchange transfusion process. The value of coefficients （r） was between 0. 96 and 0. 98 （P 〈0. 001）. （3） The slope rate of series regression equations （a） also showed a highlylinear correlation with the initial value of TBIL volume （Yo） before exchange transfusion. （4） The variance analysis on the regression effect of each group had significant difference （P 〈0. 01 ）. （5） The 95% of the regression coefficient＇ s confidence interval had significant difference （ P 〈 0. 001 ）. Conclusion The linear regression analysis for the correlation equation model y = Y0 - ax as well as a=αy0＋β could established a comprehensive linear regression equation y = Yo - ax = Y0 - （0. 0035y0 -0. 081 ）x with a single independent variable, which could provide an useful guidance for the clinical work.