Performance of hemoglobin A1c assay methods: good enough?

Two very interesting articles on hemoglobin A1c (Hb A1c) 2 method performance appear in this issue of Clinical Chemistry. Woodworth et al. (1 ) use Hb A1c as an example of how the risk of reporting inaccurate results can be estimated to guide individualized QC strategies. They assess the risk of reporting unreliable Hb A1c results with 6 different methods in 4 academic medical centers; 5 are laboratory instruments and 1 is a point-of-care (POC) method. Patient-weighted metrics were calculated on the basis of 1–2s QC rules with an average of 1 QC for every 100 samples. The authors found considerable differences in the risk of reporting unreliable results among the methods used at these sites, and that only 1 method, the Capillarys2 Flex Piercing, achieved a patient-weighted metric 3 at a total error allowable (TEa) of 6%. This total error limit is also used for College of American Pathologists (CAP) proficiency testing and National Glycohemoglobin Standardization Program (NGSP) certification and is considered optimal for clinical care at this time. This means that for all but the aforementioned method, “maximum QC (three levels, three times per day) should be performed to achieve the necessary error detection.” Of note was that for some methods there was a substantial amount of bias (up to 5.8% relative for values within the reference interval), and CV values were higher than the recommended 2% for either the high or low QC for 3 of the methods studied (2 ). The maximum number of Hb A1c results out of 100 expected to be unreliable from an out-of-control condition at TEa of 6% ranged from 0.60 to 71.48 (!), and for 2 methods the number of results expected to be unreliable was 19 of 100 even when the method would be considered in control (mainly due to high bias). The second article on Hb A1c method performance by Lenters-Westra and Slingerland (3 ) includes an evaluation of 7 Hb A1c POC methods. This study also finds large differences in performance among methods, with CV values varying from 1% (better than most of the laboratory methods in the Woodworth study (1 )) to 3%. Precision evaluations in both studies followed the Clinical and Laboratory Standards Institute (CLSI) EP-5 protocol, so the CV values reported in the 2 studies should be directly comparable. Overall, the CV values for the POC methods overlapped those for the laboratory methods, and the CV values for the DCA Vantage (POC method) included in both studies were similar, albeit slightly above the recommended 2% for within-laboratory variability. Overall, the imprecision of all the methods evaluated in both studies varied considerably, with CV values ranging from 0.8% to 3.2% when evaluated across the measurement range. In both studies, bias was calculated on the basis of comparison to NGSP/IFCC Reference Laboratories, so the reference point in both studies is also the same. Although the way in which bias is reported is not the same (percentage bias at 2 levels vs mean absolute Hb A1c bias across the entire range), clearly, in both studies, there are large biases for some methods and almost no bias for others. All the methods evaluated in both studies were NGSP certified (on the basis of comparison data submitted by the manufacturer) at the time of each study, so one would hope that they would be able to perform at the same level in the laboratory. However, Lenters-Westra and Slingerland (3 ) found that 4 of the 7 methods evaluated in their laboratory would fail the certification criteria. It is worth noting that these evaluations were performed in a laboratory with experienced personnel; because these methods are all CLIA waived, there are no mandates that the end users be trained laboratory personnel or perform proficiency testing. Clearly, as noted in previous studies (4, 5 ), some methods that can perform well enough to pass NGSP certification when testing is performed by the manufacturer do not consistently achieve the same level of performance in the field. Also of note is that in both studies, bias rather than imprecision seemed to be the major factor when methods did not perform well; this would seem to indicate that lot-to-lot variations between reagents and/or calibrators may play a significant role. Although there were no substantial differ1 Diabetes Diagnostic Laboratory, Department of Pathology and Anatomical Sciences, University of Missouri School of Medicine, Columbia, MO. * Address correspondence to this author at: Diabetes Diagnostic Laboratory, Department of Pathology & Anatomical Sciences, University of Missouri School of Medicine, 1 Hospital Dr. M766, Columbia, MO 65212. Fax 573-884-4748; e-mail [email protected]. Received May 23, 2014; accepted May 28, 2014. Previously published online at DOI: 10.1373/clinchem.2014.225789 © 2014 American Association for Clinical Chemistry 2 Nonstandard abbreviations: Hb A1c, glycated hemoglobin; POC, point-of-care; TEa, total allowable error; CAP, College of American Pathologists; NGSP, National Glycohemoglobin Standardization Program; CLSI, Clinical and Laboratory Standards Institute. Clinical Chemistry 60:8 1031–1033 (2014) Editorials