Evaluation of measurement uncertainty for HbA1c by four approaches in clinical laboratory / 中华检验医学杂志

ABSTRACT

Objective:To compare results of four glycosylated hemoglobin A1c (HbA1c) detection methods and to evaluate the uncertainty of HbA1C results in clinical laboratory, and to provide method for clinical laboratory on the evaluation of uncertainty.Methods:According to the four uncertainty evaluation methods, which were recommended by "CNAS-TRL-001, the evaluation and expression of measurement uncertainty in medical laboratory", the relative and absolute uncertainty of low, medium and high HbA1c in 33 clinical laboratories measured in 2019 and 35 clinical laboratories measured in 2020 was evaluated by more than 6 months of internal quality control (IQC) data, trueness verification and external quality assessment (EQA) data. The four uncertainty evaluation methods were: IQC data and trueness verification data (method 1), only trueness verification data (method 2), IQC and EQA data (method 3) and only EQA data (method 4). The related statistical methods used in this analysis were Friedman and Wilcoxon signed rank test.Results:For method 1, the median range of relative and absolute uncertainty of low, medium and high HbA1c detection in 2019 and 2020 ranged from 4.21% to 9.24% and from 0.27% to 0.64%, respectively. Compared to method 1, the relative and absolute uncertainties obtained by method 2 were smaller, and the differences were statistically significant ( P<0.016 7, P<0.05). Compared to method 1, the relative uncertainties obtained by method 3 and method 4 were smaller, except for the high concentration of HbA1c level in 2020. Among the 6 pairs of comparisons (low, medium and high HbA1c in 2019 and 2020), there were 3 pairs (high HbA1c in 2019, low and medium HbA1c in 2020) and 2 pairs (low and high HbA1c in 2020) of differences with statistical significance (all P<0.016 7). Conclusion:The uncertainty evaluation of HbA1c detection in clinical laboratory should be evaluated based on IQC and trueness verification data.