Point-of-Care Assays for Hemoglobin A1c: Is Performance Adequate?

To the Editor: The routine measurement of hemoglobin A1c (Hb A1c)1 has become an essential component of the standard of care for patients with diabetes and has been recommended by major clinical diabetes organizations. A small amount of evidence shows that having the Hb A1c result at the time of the doctor's visit is beneficial (1). Recently, there has been discussion about whether the quality of point-of-care (POC) testing for Hb A1c is sufficient to meet clinical needs (2). The National Glycohemoglobin Standardization Program (NGSP) certifies methods annually at the manufacturer level, with only 1 lot of reagents being used at any point in time. Although College of American Pathologists proficiency testing provides an excellent snapshot of the performance of each method in the clinical laboratory, POC methods are CLIA waived; thus, users are not required to participate in proficiency testing. Inadequate performance of some of these methods in the hands of experienced users (2) has raised concerns about the ability of these methods to perform well enough for diabetes monitoring. The manufacturer of the A1cNow device (Bayer HealthCare) noted (3) that the use of EDTA-containing blood in the previous study (2) was not in accordance with manufacturer recommendations. The manufacturers of Afinion (Axis-Shield) and in2it (Bio-Rad Laboratories) purportedly have made improvements to these methods since the original evaluation; therefore, we reexamined the Afinion, A1cNow (using heparinized blood), and in2it methods in this study. Two lots of A1cNow and Afinion reagents were shipped to each of 2 NGSP secondary reference laboratories (SRLs), one in the US (laboratory A) and the other in Europe (laboratory B) for a total of 4 reagent lots tested for each method. Two lots of the in2it reagents were shipped to one of the laboratories (laboratory B). For each method, each lot was evaluated according to CLSI EP-5 guidelines for imprecision with whole blood and/or manufacturer QC material. All evaluations for the A1cNow device were performed with heparin-anticoagulated whole blood. EDTA-anticoagulated whole blood was used for both the Afinion and in2it evaluations. Each lot was compared with an NGSP SRL method, as would be done for NGSP method certification, with Bland–Altman assessment of agreement with current NGSP manufacturer certification limits (4, 5). For the A1cNow and Afinion methods, a likelihood ratio test was used to evaluate for statistically significant differences in results among reagent lots between and within laboratories. For the in2it method, 2 reagent lots were evaluated in a single laboratory; therefore, an overall test of coincidence of the least squares regression lines was used to test for a statistically significant difference between the lots. For all tests, a P value <0.05 was considered to indicate statistical significance. Total CVs were between 3.4% and 5.1% for the A1cNow method (Table 1), whereas CVs were lower for the Afinion method (between 1.2% and 2.7%) and the in2it method (between 2.4% and 3.0%). For the Afinion method, imprecision was slightly better for the QC material than for whole blood. The whole blood estimate may better reflect the variation in patient whole blood results for the Afinion method. Data are presented as the Hb A1c percentage; a boldface number indicates a failed result. WB1, whole blood sample 1; QC1, QC material 1. Data collected over a period of <20 days. Data are presented as the Hb A1c percentage; a boldface number indicates a failed result. WB1, whole blood sample 1; QC1, QC material 1. Data collected over a period of <20 days. The 95% confidence limits for the differences between the methods and the NGSP SRLs are also shown in Table 1. In laboratory A, both A1cNow lots passed the NGSP certification criteria, whereas both lots failed in laboratory B. For the Afinion method, one of the lots passed in laboratory A, and the second lot failed. In laboratory B, both lots passed. Both in2it lots passed NGSP certification. For the A1cNow method, there was no significant difference between pairs of reagent lots within each laboratory (P = 0.72, and P = 0.49 in laboratories A and B, respectively). The results for the 2 lots in laboratory A, however, were significantly different from those for the 2 lots in laboratory B (P < 0.0001). The Afinion method showed a significant difference between lots in laboratory A (P = 0.0002), a marginal but insignificant difference between lots in laboratory B (P = 0.05), and a significant difference in lots between the 2 laboratories (P < 0.0001). For the in2it method, there was a very small but significant difference (P = 0.02) between the 2 lots of reagent in a single laboratory. Although the ideal analytical goal for the imprecision of Hb A1c methods is ≤2%, CVs of ≤3% are certainly reasonable. In the present study, the Afinion imprecision was similar to previous results, with CVs of <3%. The A1cNow CVs were considerably greater than 3% and were therefore considered unacceptable. The total CVs for the in2it method were improved over those of the original evaluation and were acceptable in the current evaluation. Lot-to-lot variation for the A1cNow and Afinion methods is a concern. Two of 4 A1cNow lots and 1 Afinion lot did not pass NGSP certification. It is important to consider clinical needs when selecting Hb A1c assay methods, including POC methods. Clinicians must recognize that although POC Hb A1c measurement offers convenience in some clinical settings, the performance of some POC methods may not be sufficient to meet clinical needs. hemoglobin A1c point-of-care National Glycohemoglobin Standardization Program secondary reference laboratory The authors acknowledge the donation of reagents from Bayer, Axis-Shield, and Bio-Rad Laboratories for this study and their continued interest in improving Hb A1c measurements.