Effects of Hemoglobin C and S Traits on Glycohemoglobin Measurements by Eleven Methods

Patients with diabetes mellitus routinely have glycohemoglobin (GHb) testing performed to monitor glycemic control and assess risk for developing complications of their disease (1). The accuracy of several GHb methods can be adversely affected by the presence of hemoglobin (Hb) C or S trait (2)(3)(4)(5)(6). It has been estimated that there are at least 200 000 Americans with diabetes mellitus who also have either Hb C or S trait (6). We have recently shown that the presence of Hb C or S trait does not affect the accuracy of GHb measurements made by the CLC 330 boronate affinity HPLC method (7). We therefore evaluated the effects of Hb C and S traits on 11 commercial GHb methods, using the CLC 330 assay as the comparison method. Whole blood samples from individuals homozygous for Hb A (n = 73) and heterozygous for Hb C or S (n = 46 and 76, respectively) were collected in EDTA-containing tubes. After routine clinical testing had been completed, Hb variants were identified by inspection of chromatograms obtained with a VARIANT analyzer (Bio-Rad Laboratories) and the Beta Thal Short program run according to the manufacturer’s instructions. Aliquots of these samples that had 4–14% Hb A1c were stored at 2–8 °C and analyzed within 10 days of collection except for aliquots for the HA8160 and HA8160 Beta Thal (BT) methods, which were shipped on dry ice and stored frozen until analysis. Not all samples were analyzed by each analytic method. This study was approved by the Institutional Review Board of the University of Utah. Samples were analyzed by the following instruments/methods: A1c 2.2 Plus and G7 (Tosoh); A1cNow (Metrika); D-10, DiaSTAT, and VARIANT II (Bio-Rad Laboratories); Dimension RxL (Dade Behring); HA8160 HbA1c and HA8160 BT (Menarini Diagnostics); and PDQ (Primus). All of these methods were used according to the manufacturers’ instructions and have been certified by the National Glycohemoglobin Standardization Program (NGSP). The CLC 330 GHb analyzer (Primus) was used as the comparison method in an NGSP Network Laboratory with in-house calibrator materials and assigned values. Results for all methods are reported as NGSP Hb A1c equivalents. For each test method, results obtained for each type of sample (homozygous Hb A, heterozygous Hb C, and heterozygous Hb S) were compared with those obtained by the CLC 330 comparison method. An overall test of coincidence of two least-squares linear regression lines was performed with SAS software (SAS Institute) to determine whether the presence of Hb C or S trait caused a statistically significant difference (P <0.01) in results relative to the comparison method. Deming regression analysis was performed to determine whether the presence of Hb C or S trait produced a clinically significant effect on GHb results. Given recommendations by the American Diabetes Association of an upper reference limit of 6% and an action limit of 8%, we chose Hb A1c evaluation limits of 6% and 9%. After correcting for possible calibration bias by comparing results from the homozygous Hb A sample group, we evaluated method bias attributable to the presence of Hb C or S trait, with a clinical significant difference being >10% (i.e., 0.6% at 6% Hb A1c and 0.9% at 9% Hb A1c). The presence of Hb C trait produced statistically significant differences (P <0.01) for all methods tested except for the DiaSTAT, HA8160, and PDQ methods. The presence of Hb S trait produced statistically significant differences for all methods except for the D-10 and PDQ methods. Box-plots for each combination of sample type and method are shown in Fig. 11 . We observed no clinically significant interference attributable to Hb C or S trait with the A1c 2.2 Plus, Dimension RxL, G7, HA 8160, HA 8160 BT, and PDQ methods (Table 11 ). The presence of both Hb C and S traits produced clinically significant positive biases for the A1cNow and VARIANT II methods at 6% and 9% Hb A1c, respectively. Hb C trait produced a clinically significant negative bias for the D-10 method at 9% Hb A1c. Hb S trait produced a clinically significant positive bias at 6% Hb A1c for both the DiaSTAT and DS5 methods. Several methods, including the D-10, DiaSTAT, Dimension RxL, DS5, HA8160, and VARIANT II, showed increased scatter when samples containing Hb C or S trait were tested compared with that seen for samples homozygous for Hb A. Our results have several similarities to previous studies. For example, one immunoassay method, the Dimension RxL, exhibited no clinically significant effect with either Hb C or S trait, but another immunoassay method, the A1cNow, exhibited a clinically significant positive bias with samples containing both Hb C and S traits. The Unimate and Cobas Integra immunoassay methods have been shown to exhibit a positive bias with both Hb C and S traits, whereas the DCA 2000, Tina-quant, and SYNCHRON CX 7 methods exhibit no clinically significant bias with either Hb C or S trait (3)(4)(6). Likewise, several ion-exchange methods, including the A1c 2.2 Plus, G7, and HA8160, were not affected by either Hb C or S traits. The DiaSTAT, DS5, and VARIANT II all exhibited a clinically significant positive bias with samples containing Hb S trait. It has previously been shown that some ion-exchange methods, including the Diamat, HA8140, and VARIANT, have a positive bias with samples containing Hb S trait (4)(6). It is noteworthy that two ion-exchange methods in the present study exhibited a clinically significant bias with samples containing Hb C trait. The D-10 method exhibited a negative bias, whereas the VARIANT II exhibited a positive bias. If a method demonstrates a positive bias attributable to Hb variants, then overly rigorous blood glucose control may be instituted with a concomitant increase in hypoglycemic episodes. If a method demonstrates a negative bias, then more rigorous blood glucose control may not be instituted, and the patient may have poorer glycemic control than is optimal. We have previously examined the effects of Hb C and S traits on the A1c 2.2 Plus and VARIANT II ion-exchange methods (4)(6). Neither method exhibited a clinically significant effect with either variant Hb in these earlier reports. In our present study, the A1c 2.2 Plus method exhibited no clinically significant effects with either variant Hb, but the VARIANT II method exhibited clinically significant effects for both Hb C and S traits. We have previously noted that another ion-exchange GHb method may intermittently show effects with Hb S trait samples, and we speculated that it may be attributable to variability from lot to lot in mobile phase or column packing material (4). It appears that intermittent interferences from Hb C and S traits with some, but not all, ion-exchange GHb methods continues to be an issue. In summary, some current GHb methods show clinically significant interferences with samples containing Hb C or S trait. These interferences are not necessarily consistent within method types, and with ion-exchange methods may vary over time with changes in column or reagent lots. Box-plots summarizing the absolute differences between each test method and the comparison method for each Hb phenotype. The horizontal line inside each box is the median difference between the test and comparison methods. The upper and lower limits of each box correspond to the 25th and 75th percentiles of the differences, respectively. The upper and lower error bars represent the maximum and minimum differences between the test and comparison methods. Differences from Hb AA that are statistically significant (P <0.01) are indicated (#), as are clinically significant differences (∗). Mean differences between the test methods and the comparison method for samples containing either Hb C or S trait.1 Deming regression analysis was performed with CLC 330 as the comparison method. The mean differences (%) between the comparison method and each of the other methods at clinical decision cutoffs of 6% and 9% were calculated for each Hb trait. To correct for intermethod calibration differences, the mean difference for homozygous Hb A samples was subtracted from that calculated for samples containing Hb C or Hb S trait. Difference was clinically significant (>0.6% or >0.9% Hb A1c at 6% and 9% Hb A1c, respectively). Mean differences between the test methods and the comparison method for samples containing either Hb C or S trait.1 Deming regression analysis was performed with CLC 330 as the comparison method. The mean differences (%) between the comparison method and each of the other methods at clinical decision cutoffs of 6% and 9% were calculated for each Hb trait. To correct for intermethod calibration differences, the mean difference for homozygous Hb A samples was subtracted from that calculated for samples containing Hb C or Hb S trait. Difference was clinically significant (>0.6% or >0.9% Hb A1c at 6% and 9% Hb A1c, respectively). This work was supported by the ARUP Institute for Clinical & Experimental Pathology.