Verification of the hemolysis index measurement: imprecision, accuracy, measuring range, reference interval and impact of implementing analytically and clinically derived sample rejection criteria

Abstract

Automated spectrophotometric measurement of hemolysis index (HI) allows rapid and cost-effective assessment of hemolysis interference. We evaluated the analytical performance of HI on two different platforms. Further, the impact of implementing analytically and clinically derived sample rejection criteria was investigated. Precision profiles were established, and analytical error was assessed by comparison with the reference method for hemoglobin measurement on Architect c8000/c16000 (Abbott) and Cobas c702 (Roche Diagnostics) instruments. The impact of a more analyte-specific cut off based on analytical and biologival variation was examined for five hemolysis-sensitive plasma analyses according to European recommendations. Lastly, a reference interval was established for the HI on Cobas, using the CLSI C28A3c nonparametric method. Imprecision for HI of 0.6-3.0 % for Architect and 1.5-4.5 % for Cobas was considered acceptable, which also applied for trueness in the measuring tange > 2 g/L. If cutoffs based on analytical and biological variation were used to manage results from hemolyzed samples, more potassium, LDH, conjugated bilirubin and phosphate results would be suppressed. Considering RCV only LDS remained sensitive to hemolysis. The Cobas-based HI reference interval was established to 0.01–0.16 g/L. Thorough verification of the HI on two different clinical chemistry analyzers reveals acceptable analytical performance. HI cutoffs suggested by manufacturers may be optimized by clinical laboratories using analytical and/or biological variation. A reference interval for the HI analysis is relevant as the analysis has been suggested as a diagnostic tool in the assessment of in vivo hemolysis.

Keywords:

• Verification
• pre-analytical phase
• hemolysis index
• sample quality

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was part of a PhD fellowship for Charlotte Gils supported by the University of Southern Denmark [SDUSF-2016-203-(882)] and the Region of Southern Denmark [17/15099].