A-179 A Multi-Centre Assessment of Comparability of Hemolysis Indices on Automated Analyzers

Abstract Background Hemolysis is a prevalent preanalytical issue that compromises quality and patient safety during laboratory testing. It can interfere with sample analysis, leading to inaccurate results, sample rejection, and delayed patient treatment. The International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) has identified hemolysis as a high-priority quality indicator, defined as the rate of samples with hemoglobin (Hgb) above 50 mg/dL (>0.5 g/L). Automated methods have improved hemolysis detection in serum/plasma samples. However, the internal calibration of automated methods for hemoglobin estimation (i.e. Hemolysis index) is usually not verified for accuracy against external standards by the end-user nor standardized across methods. This study aims to address this gap by comparing hemoglobin quantification using the serum indices application on automated chemistry analyzers versus a conventional spectrophotometric method that is validated for plasma free hemoglobin measurement. Methods Plasma samples (n=20) were analyzed at nine laboratories across the Greater Toronto Area in Canada. The plasma free Hgb concentrations in patient samples ranging from 5 to 450 mg/dL (0.05 to 4.50 g/L) were determined with the method of Khan SE et al. (Ann Clin Lab Sci. 1981;11(2):126-31) on Agilent Cary 60 UV/Vis spectrophotometer as the reference. In contrast, estimated Hgb were measured on seven automation systems: Abbott Alinity c, Roche cobas pro (n=2), Roche cobas 8000, Roche cobas 6000, Ortho Vitros 5600 (n=2), Siemens Atellica CH and Beckman Coulter AU680. Method comparison analyses were performed using Passing-Bablok linear regression. Bias was assessed at 50 mg/dL, the IFCC threshold for the hemolysis quality indicator. Subsequently, the percent bias for each automated application was calculated as [(automated Hgb - 50 mg/dL) / 50 mg/dL] x 100. Results All methods’ coefficient of variation (CV) of the hemolysis values for each sample revealed a close agreement among the automated analyzers with a combined CV of 24% and median CV of 8% (range 3% to 73%). However, the biases for Hgb estimates at the IFCC cut-off ranged from 1.4% to 23.0% across the entire range of Hgb levels (5 to 450 mg/dL). The specific biases observed when applying the linear regression for samples with Hgb concentrations <100 mg/dL were: 1.4-10.4% for Roche cobas analyzers, 20.2-21.4% for Ortho Vitros 5600 analyzers, 16.2% for Siemens Atellica CH, 8.0% for Abbott Alinity c and 18.2% for Beckman Coulter AU680, highlighting variations in Hgb estimates among automated analyzers. All automated analyzers demonstrated excellent correlation with the Khan method. The Pearson’s R values ranged from 0.997-0.999 for Hgb concentrations between 5-450 mg/dL, and 0.981-0.999 for Hgb concentrations <100 mg/dL. Conclusion In conclusion, our study revealed excellent correlation among automated analyzers in measuring hemolysis indices, supporting the reliability of these methods. Assessment of hemolysis at the sample level supports the IFCC harmonized quality indicator. However, notable positive biases were observed among some analyzers, which may impact comparisons of hemolysis rates, emphasizing the need for further harmonization among manufacturers.