Comparison and bias estimation of three methods in determination of glycated hemoglobin A1c

Abstract

To evaluate the bias of three different methods in determination of glycated hemoglobin A1c (HbA1c). According to Clinical and Laboratory Standards Institute (CLSI)EP9-A3 document, 40 serum specimens among linear range were tested by immunoturbidimetry assay (Roche Tina-quant), capillary electrophoresis (Sebia Minicap FP), high performance liquid chromatography system (HPLC, Trinity Biotech Premier Hb 9210™). RESULTS were analyzed by pairwise comparision. The difference plots and scatter plots were analyzed. The results of outliers were examined by extreme studentized deviate (ESD) method. The best regression model was chosen to fit the regression equation and the biases were calculated at the level of medical decision and judged the comparability according to the half of Tea acceptance standard. Scatter plots and ESD results showed good correlation among test results of 40 samples, no abnormal values were found. Based on proportional difference deviation plots, proportional difference rank order deviation plots, numerical deviation plots and ranked order difference plots, weighted least squares (WLS) (Roche(X) vs sebia(Y1)), ordinary linear regression (OLR) (Roche(X) vs Trinity (Y2)) and Deming (Sebia(Y1) vs Trinity(Y2)) were chosen to fit regression analysis model. The regression equations were Y1 =-0.420 3+ 1.052 7X, Y2 =-0.174 0+ 1.027 0X, Y2=0.423 1+ 0.959 0Y1, respectively. Assuming the HbA1c medical decision points of 10%, 16% to regression equation, the proportional biases were 1.06% and 2.61% at Roche(X) and sebia(Y1) system, 0.96% and 1.60% at Roche(X) and Trinity (Y2) system, 0.13% and -1.47% at Sebia(Y1) and Trinity(Y2) system respectively, which were all lower than the acceptable values. The measurement values of HbA1c by capillary electrophoresis (Sebia Minicap FP), HPLC (Trinity Biotech Premier Hb 9210™) and immunoturbidimetry assay (Roche Tina-quant), are comparable.