Abstract Background Across four global laboratory locations in North America, Europe, and Asia a quality control system was implemented in the measurement of Hemoglobin A1c (HbA1c). The intent was to improve accuracy and reduce imprecision, both being independent contributors to total analytical performance. Methods Quality control (QC) data, created during routine clinical operations, was used for this analysis. Analysis of QC data evaluated bias and precision separately and provided the ability to predict and monitor performance. The in-process harmonization program indicated a change in instrumentation was needed to meet and maintain a more stringent analytical error profile. The instrumentation change allowed for a before and after comparison across platforms. Results Use of a harmonized quality control system allowed for the visualization of the role bias and precision play in measurements. Improved performance was confirmed using native matrix samples tested routinely from participation in the National Glycohemoglobin Standardization Program. Independently assessing bias and precision highlighted the cumulative and offsetting effects that challenge quality assessment and risk mitigation. Bias appeared to have both a transient and systematic nature. The transient bias contributed to the overall normal distributive property of the quality data. In contrast, the systematic accumulation of bias appeared to occur slowly, over multiple reagent and calibration cycles. The accumulation of bias occurred in the direction of the assay drift. While precision commonly consumes the largest amount of analytical error within an instrument, the dual (negative and positive) nature of bias similarly affects the harmonization of an instrument group. Conclusions Changing the way data is viewed to include assessments of bias improved performance and aided in the identification of opportunities to reduce analytical variances that results in inter- and intra-individual variation. Harmonization of instruments should include methods to assess and describe bias, independent of precision.
Read full abstract