Derivation of real metrics of long term patient and analytical variation of three hemoglobin A1c assays demonstrates both borderline and highly acceptable analytical performance

Background: Previously, we demonstrated that sequential intra-patient data can yield realistic measures of short term patient biologic and analytic variation. We now determine the long term (LT) combination of biologic and analytic variation associated with three hemoglobin A1c (HbA1c) assays. Methods: Three sets of patient HbA1c data were analyzed: 2 years of Sebia Capillarys 2 Flex Piercing® HbA1c from a Quebec hospital laboratory, 3 years of Roche Tinia Quant Gen II HbA1c assay measured with Cobas 8000, c502 and Cobas 6000, c501 analyzers by a New Hampshire laboratory and 3.5 years of Siemens Vista HbA1c from an Ontario hospital. We generated graphs of the LT combination of biologic and analytic variation for the three methods and four patient subpopulations: (I) virtually the entire population, (II) high HbA1c, (III) low to normal HbA1c, and (IV) inner 50 percentile (P) HbA1c. Results: The Sebia and the Roche combinations of LT biologic and analytic variations were consistently lower than those of the Vista. The Vista yielded the highest y intercept and the highest LT biologic/analytic combination variations at 26 weeks. The elevated Vista variations were most obvious for the low-normal and midlevel HbA1c groups. Conclusions: The Vista assay demonstrated excess analytic variation. HbA1c assays with LT imprecisions exceeding 3% will result in artefactually elevated glycemic variability. The Roche and Sebia assays are fit for purpose. We recommend that post-market evaluations of LT intra-patient HbA1c variation be conducted as part of external quality assessment (EQA) schemes in order to provide clinically relevant perspectives of assay performance.