Glycohemoglobin Results in Samples with Hemoglobin C or S Trait: A Comparison of Four Test Systems

A wide variety of commercial methods are available to measure glycohemoglobin (gHb). These methods measure various species of gHb, e.g., total gHb, hemoglobin A1 (Hb A1), or Hb A1c. Some methods are based on charge differences between glycated and nonglycated hemoglobins (e.g., cation-exchange chromatography, electrophoresis, and isoelectric focusing), whereas boronate affinity methods depend upon the binding of the sugar groups on the hemoglobin molecule (1). Immunoassay of Hb A1c depends on the presence of an epitope that includes glucose and N-terminal amino acids of the β chain of hemoglobin. These methods usually measure Hb A1c as a percentage of total hemoglobin. Some immunoassay methods also measure Hb S1c or Hb C1c, whereas others do not (2). The National Glycohemoglobin Standardization Program (NGSP) was established to certify the various commercial methods so that they can be related to the candidate reference method used in the Diabetes Control and Complications Trial (DCCT) (3)(4). Recently, the effects of Hb AC and Hb AS on one gHb immunoassay method were assessed (5). Samples containing Hb C trait showed modestly higher results than the HPLC method used. Although hemoglobin variants are relatively rare in Caucasians of northern European descent, the African-American population has an 8% prevalence of Hb AS and a 3% prevalence of Hb AC (6)(7). In this study, we describe the influence of Hb AS and Hb AC on four DCCT-traceable gHb methods. A total of 129 samples were collected in evacuated tubes containing EDTA as an anticoagulant. Samples with hemoglobin variants were identified by comparison of retention times on the Diamat system (Bio-Rad Clinical Laboratories) to known retention times for Hb S and Hb C. Of the 129 samples collected, 47 were homozygous for Hb A, 39 were heterozygous for Hb C, and 43 were heterozygous for Hb S. Samples were shipped on dry ice and stored at −70 °C until analysis. Four commercial systems were used to measure gHb according to the manufacturers’ instructions: the Diamat, the DCA 2000 (Bayer Diagnostics), the Unimate performed on a Cobas Mira L (Roche Diagnostics Systems), and the A1c 2.2 Plus (Tosoh Medics). Results for the Diamat and DCA 2000 methods using these samples have been published(5). The purpose of this study was to investigate the effects of variant hemoglobins on the A1c 2.2 Plus and Unimate methods. Data analysis including Deming regression was performed using EP Evaluator release 3 software (David G. Rhoads). The mean percentage of the Hb S in samples with Hb AS was 38% (range, 25–41%), and the mean percentage of Hb C in samples with Hb AC was 38% (range, 29–41%). Samples from patients homozygous for Hb A served as controls for detecting calibration differences between methods. The Diamat was chosen as the comparison method, and Deming regression analysis was performed (Fig. 1). Comparison of Hb A1c results obtained by the Diamat, A1c 2.2 Plus, DCA 2000, and Unimate methods. Results were compared using Deming regression analysis (——–); an ideal comparison with slope = 1.00 and intercept = 0.0 is indicated by (- - - - - - -). Results for samples containing only Hb A are shown in panels A (n = 47; slope = 0.88 ± 0.03; intercept = 1.46 ± 0.26; r = 0.975; Sy|x = 0.51), B (n = 45; slope = 0.96 ± 0.03; intercept = 0.51 ± 0.25; r = 0.982; Sy|x = 0.46), and C (n = 47; slope = 0.96 ± 0.02; intercept = −0.12 ± 0.18; r = 0.990; Sy|x = 0.35). Results for samples containing Hb C trait are shown in panels D (n = 36; slope = 0.90 ± 0.02; intercept = 0.95 ± 0.43; r = 0.949; Sy|x = 0.61), E (n = 36; slope = 1.21 ± 0.06; intercept = −1.07 ± 0.51; r = 0.957; Sy|x = 0.67), and F (n = 39; slope = 1.66 ± 0.10; intercept = −3.30 ± 0.89; r = 0.925; Sy|x = 1.26). Results for samples containing Hb S trait are shown in panels G (n = 37; slope = 0.87 ± 0.05; intercept = 1.34 ± 0.43; r = 0.946; Sy|x = 0.63), H (n = 40; slope = 1.11 ± 0.04; intercept = −0.61 ± 0.35; r = 0.975; Sy|x = 0.49), and I (n = 43; slope = 1.34 ± 0.05; intercept = −1.89 ± 0.50; r = 0.965; Sy|x = 0.79). The Diamat and A1c 2.2 Plus HPLC methods showed similar agreements for all three sample types (Fig. 1 , A, D, and G). The DCA 2000 immunoassay and the Diamat agreed well for samples containing only Hb A (Fig. 1B), but for samples containing Hb S, and especially Hb C, slopes of the regression line were >1 (Fig. 1 , E and H). The Unimate immunoassay method agreed with the Diamat on samples containing only Hb A (Fig. 1C), but the results were higher than the Diamat results for samples with Hb S (Fig. 1I) and even higher for samples with Hb C (Fig. 1F). We calculated the mean difference attributable to the presence of a variant hemoglobin from Diamat results for each method at Diamat Hb A1c values of 6% and 9% (Table 1). The results obtained with the A1c 2.2 Plus method were modestly lower than the results obtained with the Diamat for samples with Hb C trait at Hb A1c concentrations of both 6% and 9% and minimally lower for samples with Hb S trait. The results obtained with the DCA 2000 method were higher than the results obtained with the Diamat at 9% Hb A1c for both Hb AS and Hb AC. The results obtained with the Unimate were higher than the results obtained with the Diamat at 6% and 9% Hb A1c for both Hb AS and Hb AC samples. Average differences from Diamat results for samples with Hb C or Hb S traits.1 Deming regression analysis was performed using the Diamat as the reference method. The average difference of each of the other three methods at Diamat medical decision levels of 6% and 9% was calculated for each hemoglobin trait. To correct for intermethod calibration differences, the mean difference for homozygous Hb A samples was subtracted from that calculated for samples containing the Hb C or Hb S trait. Average differences from Diamat results for samples with Hb C or Hb S traits.1 Deming regression analysis was performed using the Diamat as the reference method. The average difference of each of the other three methods at Diamat medical decision levels of 6% and 9% was calculated for each hemoglobin trait. To correct for intermethod calibration differences, the mean difference for homozygous Hb A samples was subtracted from that calculated for samples containing the Hb C or Hb S trait. Certification of gHb methods has aided in the reduction of differences between methods, as demonstrated in a recent College of American Pathologists gHb survey (8). All four methods in this study have been certified as traceable to the DCCT by the NGSP. Both the DCA 2000 and Unimate methods yielded data consistent with certification by the NGSP when Hb A samples were analyzed. When samples containing Hb C trait were analyzed with the DCA 2000 method, the mean result was substantially higher than that obtained with the Diamat at 9% Hb A1c. When samples containing Hb C and Hb S traits were analyzed with the Unimate method, the mean result was substantially higher than that obtained with the Diamat at 9% Hb A1c. When the A1c 2.2 Plus method was designated the reference method, the DCA 2000 results again were substantially higher for samples containing Hb C, as were the Unimate results for samples containing both Hb C and Hb S (data not shown). These apparent overestimations of Hb A1c by the Unimate method in particular would appear to be clinically significant. It is noteworthy that both immunoassays showed significant differences from the Diamat HPLC method only at Hb A1c values above the upper limits of the reference interval. Previous reports that indicated that the DCA 2000 method is not affected by hemoglobin variants used only samples from nondiabetic patients(2)(9). These observations confirm the necessity of the NGSP guidelines requiring bias estimates at 6% and 9% Hb A1c (3). The package insert for the Unimate method indicates that specimens containing Hb C and Hb S variants may yield higher than expected Hb A1c results, but the magnitude of this problem is not quantified. There are several possible explanations for the higher values seen with the immunoassay methods compared with the HPLC methods for samples containing Hb C trait. One explanation is that the HPLC methods are underestimating the true Hb A1c value. Both HPLC methods express Hb A1c as a percentage of total Hb A, as previously recommended (10), whereas the immunoassay methods express results as a percentage of total hemoglobin. Inspection of a Diamat chromatogram reveals the presence of a peak eluting after Hb A but before Hb C that probably is Hb C1c. This peak is adequately resolved from Hb A when the Diamat is operated using an extended program. Another possible explanation is that the immunoassay methods overestimate the true Hb A1c concentration. The monoclonal antibody used in the DCA 2000 and Unimate assays is identical (11). Hb S and Hb C differ from Hb A by a single amino acid substitution at the sixth position of the β chain (Glu to Val or Glu to Lys, respectively). These amino acid substitutions are very close to the glycated amino terminus that is recognized by the monoclonal antibody in the immunoassays. It is possible that the glycated variant hemoglobins Hb C1c and Hb S1c have a higher affinity than Hb A1c for the mouse monoclonal antibody-latex particles used in the assay. The difference between the DCA 2000 and Unimate immunoassays might arise from the proteolytic digestion that occurs with the Unimate assay. Pepsin, the protease used in the Unimate method, can cleave the β chain of Hb A, Hb C, and Hb S between amino acids 3 and 4, 7 and 8, 11 and 12, and 14 and 15 (12). The sequence of the tripeptide generated is not altered; however, the sequences of the peptides containing 7, 11, and 14 amino acids are altered at the sixth position in Hb C and Hb S. We hypothesize that these peptides and intact Hb C1c have an increased affinity for the anti-Hb A1c antibody. The difference between the DCA 2000 and Unimate assays thus may be attributable to different steric effects at the antigen recognition site produced by differences in the latex particles to which the common monoclonal antibody is attached. The magnitude of differences between methods observed in this study was dependent on the type of hemoglobin present, the percentage of Hb A1c in the sample, and the method used. The presence of a hemoglobin variant such as Hb S or Hb C may adversely impact the accuracy of a gHb result. Physicians and laboratorians should be able to refer to the manufacturer’s technical literature to judge whether the presence of a common hemoglobin variant has a clinically significant effect on the test result. More information needs to be collected on the effect of various hemoglobin variants on specific Hb A1c test systems. One advantage of an HPLC method for gHb analysis is that the analyst may be alerted to the presence of a variant hemoglobin in many cases, whereas no such warning is generated by immunoassay methods. No data relating mean blood glucose and Hb A1c in patients with Hb C or S trait have been published to our knowledge. The red cell survival in patients with Hb S trait is comparable to that seen in patients who are homozygous for Hb A (13). We presume the same is true for Hb C trait, but no published study could be found. We hypothesize that Hb A1c results measured by the Diamat or A1c 2.2 Plus methods in patients with Hb C or Hb S trait correspond to the same mean blood glucose values as they would in patients homozygous for Hb A. fax 312-738-5808, e-mail koryw@ascp.org