We studied whether problems quantifying hemoglobin A2 (HbA2) could be resolved by using capillary electrophoresis. HbA2 was quantified on whole blood samples from patients with and without β-thalassemia trait and patients heterozygous for HbE, HbS, HbC, and HbD Punjab using the VARIANT II β-thalassemia (Bio-Rad, Hercules, CA) and Capillarys 2 (Sebia, Norcross, GA).
HbA2 results in patients with and without β-thalassemia trait were lower with the Capillarys 2 system. Reasonable HbA2 results were obtained for patients with HbD Punjab and HbE traits on the Capillarys 2. HbA2 results for patients with HbS, heterozygous and homozygous, were similar by both methods. Interference due to coelution for HbA2 results for patients with HbC trait was noted on the Capillarys 2.
Between-day imprecision on the VARIANT II is less than that for the Capillarys 2 system. The Capillarys 2 is superior to the VARIANT II for quantifying HbA2 in the presence of HbE and HbD Punjab traits. The Capillarys 2 offers only slight advantages over the VARIANT II for quantifying HbA2 in the presence of heterozygous and homozygous HbS. The Capillarys 2 gives inferior HbA2 results for patients with HbC trait.
High-performance liquid chromatography
In the presence of thalassemic parameters in the CBC, the accurate and precise quantification of hemoglobin (Hb)A2 (α2δ2) is essential for the diagnosis of β-thalassemia trait.1 It is necessary to have good precision in quantitative HbA2 methods because the difference in HbA2 concentrations between people with and without β-thalassemia trait is narrow. (In our laboratory, the upper limit of the reference range for HbA2 is 3.5%; HbA2 in β-thalassemia trait is usually >4.0%.)
Analytic methods to quantify HbA2 include electrophoresis at an alkaline pH, high-performance liquid chromatography (HPLC), and tandem mass spectrometry.2 Studies performed in the 1970s3–6 showed poor precision for HbA2 quantification methods based on electrophoresis. Steinberg and Adams1 concluded that although electrophoresis at an alkaline pH with densitometric tracings of electrophoretograms for quantification of HbA2 was an ideal clinical laboratory method from an ease-of-use perspective, it was inaccurate. Wild and Bain7 claim that the diagnosis of β-thalassemia trait requires approximately 10 times greater precision for HbA2 quantitation than densitometry can provide. The College of American Pathologists (CAP) has strongly recommended that electrophoretic methods not be used for quantification for HbA2 because of poor precision, and it no longer includes data from these methods in hemoglobinopathy proficiency surveys.
HPLC methods, although precise, have some limitations, including falsely decreased HbA2 levels in patients with the HbD Punjab trait due to a rising baseline,8,9 falsely increased HbA2 levels in patients with HbS (both homozygous and heterozygous),10 and coelution of various hemoglobins, including HbE, Hb Osu Christianborg, HbG Coushatta, and Hb Lepore with HbA2.9 The increase in HbA2 levels in patients with heterozygous HbS was originally thought to be due to the coelution of glycated HbS11 with HbA2 but was later shown to be due to the coelution of several HbS adducts, including carbamylated α and the βS chains with HbA2.12
Cotton et al13 described a capillary electrophoresis method for the routine determination of HbA2 and HbF and concluded that the method gave excellent precision for both. Cotton et al13 also described the measurement of HbF and HbA2 in the presence of HbS. Sebia (Norcross, GA) recently introduced a commercial capillary electrophoresis method that may resolve some of the problems in HbA2 quantification associated with HPLC while providing the excellent precision described by Cotton et al.13
The purpose of this study was to evaluate the between-day imprecision of the capillary electrophoresis method and compare the HbA2 values obtained by this method with those from a widely used HPLC method in patients with and without β-thalassemia trait, with HbD Punjab trait, and with homozygous or heterozygous HbE, HbC, or HbS.
Materials and Methods
The Bio-Rad VARIANT II β-thalassemia method (Bio-Rad, Hercules, CA) was used as directed by the manufacturer.
Capillary Electrophoresis Method
The Sebia Capillarys 2 analyzer, software version 6.10, using the HEMOGLOBIN (E) kit was used as directed by the manufacturer.
The study was performed during a 4-month period. Samples submitted to the laboratory for hemoglobinopathy/thalassemia investigation, with EDTA as the anticoagulant, were analyzed by both methods. The VARIANT II system uses a whole blood sample, but an RBC sample is required for the Capillarys 2 system. For this reason, samples were first analyzed on the VARIANT II and then analyzed by the Capillarys 2 system within 12 to 72 hours of each other. Data analysis was performed using Analyse-It for Microsoft Excel, version 2.07 (Analyse-It Software, Leeds, England).
Bio-Rad Lyphocheck HbA2 control samples (levels 1 and 2), which are whole blood control samples, were included in each of the analytic runs on the VARIANT II and Capillarys 2 analyzers for the 4-month study period. The results are given in Table 1. The excellent correlation of patient results between methods shows that good precision exists in both methods.
Equivalence of Results
The number of samples analyzed and the mean and range for HbA2 with and without β-thalassemia trait, with HbS (homozygous and heterozygous), and with heterozygous HbD Punjab, HbE, or HbC are shown in Table 2. Difference plots for these groups are shown in Figure 1.
HbA2 in Patients With and Without β-Thalassemia Trait
Deming analysis (x = VARIANT II; y = Capillarys 2) on the 207 samples with no evidence of β-thalassemia trait in the CBC with a HbA2 value within the laboratory-established reference range (up to 3.5%) or presence of hemoglobin variant showed a constant bias of 0.24 with a proportional bias of 0.93 (t statistic, 37.5; 2-tailed P < .0001) with a mean difference of 0.46. The Pearson correlation coefficient was 0.83.
Deming analysis (x = VARIANT II; y = Capillarys 2) on 91 samples with laboratory evidence of β-thalassemia trait in the CBC (mean corpuscular volume <72 μm3 [72 fL]; increased RBC count) from patients who were iron-replete (ferritin >15 μg/L [34 pmol/L]) with an HbA2 concentration higher than the in-laboratory–established VARIANT II HbA2 reference interval (upper limit, 3.5%) showed a constant bias of –0.44 with a proportional bias of 0.99 (t statistic, 16.65; 2-tailed P < .0001) with a mean difference of 0.49. The Pearson correlation coefficient was 0.94. The difference plots shown in Figure 1 for patients with and without β-thalassemia trait show a bias with the Capillarys 2 results consistently lower than the VARIANT II results.
Difference plots for hemoglobin (Hb)A2 in patients without β-thalassemia (A), with β-thalassemia (B), with HbS (C), with HbD (D), and with HbC (E). CE, capillary electrophoresis; CI, confidence interval; HPLC, high-performance liquid chromatography.
HbA2 in the Presence of HbS
The mean (range) HbA2 concentrations and range of 10 samples (from patients who had not received transfusion) with HbS concentrations less than 30% (suggestive of the presence of coinherited α-thalassemia) were 3.4% (2.4%–4.4%) on the VARIANT II and 2.8% (1.7%–3.9%) on the Capillarys 2. In 13 samples from patients with homozygous HbS (had not received transfusion), the mean (range) HbA2 on the VARIANT II was 3.7% (3.2%–4.2%) and was 3.2% (2.7%–3.7%) on the Capillarys 2. In 84 samples from patients with heterozygous HbS (HbS trait), the mean (range) HbA2 on the VARIANT II was 3.9% (3.0%–4.8%) and was 3.1% (2.4%–3.8%) on the Capillarys 2.
For the 107 samples with homozygous or heterozygous HbS, the Deming comparison (x = VARIANT II; y = Capillarys 2) between HbA2 values on the VARIANT II and Capillarys 2 showed a constant bias of –0.02 and a proportional bias of 0.81 with a Pearson correlation coefficient of 0.79, a t statistic of 13.24, and a 2-tailed P value of less than .0001.
The difference plot shows the Capillarys 2 HbA2 result to be consistently lower than the VARIANT II result in patients with heterozygous or homozygous HbS.
HbA2 in the Presence of HbD Punjab
Deming analysis (x = HPLC; y = Capillarys 2) of the results for 27 patients with heterozygous HbD Punjab showed a constant bias of –3.82 and a proportional bias of 4.45 with a mean difference of –1.29, a t statistic of –1.98, and a 2-tailed P value of less than .001. The Pearson correlation coefficient was 0.36.
The difference plot shows the HbA2 results on the Capillarys 2 to be higher than the VARIANT II results in patients with the HbD Punjab trait with a substantial degree of scatter.
HbA2 in the Presence of HbE
No comparison data were available because no HbA2 values were available from the VARIANT II owing to coelution of HbA2 and HbE. The HbA2 concentrations in 2 patients homozygous for HbE were 5.8% and 5.1%.
HbA2 in the Presence of HbC
Deming analysis (x = HPLC; y = Capillarys 2) on the results for 19 patients with heterozygous HbC showed a constant bias of –15.81 and a proportional bias of 5.76, a t statistic of 2.37, a 2-tailed P value of less than .0294, and a mean difference of 0.34. The Pearson correlation coefficient was 0.27. For 2 patients with SC disease, the HbA2 results on the VARIANT II were 3.3% and 3.6% and on the Capillarys 2 were 2.2% and 4.0%. The HbA2 results for a patient with homozygous HbC were 6.4% and 6.0% on the VARIANT II and Capillarys 2, respectively.
The difference plot shows a substantial amount of scatter between HbA2 results for patients with HbC trait on the Capillarys 2 and the VARIANT II.
The between-day imprecision (expressed as the coefficient of variation) for the Capillarys 2 quantification of HbA2 is not as good as that for the VARIANT II (6.4% vs 4.83% for the low-value control sample and 6.0% vs 2.62% for the high-value control sample). Mario et al14 stated that the capillary electrophoresis method they developed was satisfactory for precise quantifications of high HbA2 but provided no data. Based on 10 data points obtained on 10 days for each HbA2 value, Cotton et al13 obtained a coefficient of variation of 6% at HbA2 concentrations of 2.0%, 3.1%, and 5.6% and 3% at an HbA2 concentration of 2.4%. The precision for HbA2 measurements on the Sebia Capillarys 2 method is as good as that previously described by Cotton et al13 for a capillary electrophoresis method and, according to Cotton et al,13 is acceptable for establishing the diagnosis of β-thalassemia.
The HbA2 values on the Capillarys 2 for patients with and without β-thalassemia trait showed an average bias (as expressed in mean values) of approximately 0.5%. The upper limit of the HbA2 reference range (mean ± 2 SD) for the VARIANT II method was less than 3.6%, close to that previously established in this laboratory (<3.5%), and for the Capillarys 2, the upper limit of the HbA2 reference range was established as less than 3.1%. This difference in the upper end of the reference range between the VARIANT II and Capillarys 2 methods is a reflection of the lower HbA2 values obtained by capillary electrophoresis. There is good correlation between the methods, in agreement with the work by Mario et al.14 In contrast, Cotton et al9 found that HbA2 values by capillary electrophoresis were higher than those from HPLC.
Although the International Federation of Clinical Chemistry working group on the standardization of HbA2 has made substantial progress in the preparation of calibration material for HbA2, there is to date, unlike HbA1c, no reference HbA2 calibrator. Also, there is no reference method for HbA2 quantification. A review of HbA2 values on the 2008 CAP hemoglobinopathy survey program Hg-B (the first hemo-globinopathy survey to include Sebia Capillarys 2 results) showed a spread of between 0.31% and 0.65% in HbA2 values in samples without a hemoglobin variant between the Bio-Rad VARIANT II and the Sebia Capillarys 2 methods, and the difference in HbA2 values observed in this study falls within that found in the survey. HbA2 values found in the CAP survey showed a spread between 0.31% and 0.93% between methods, indicating that there is no standardization of HbA2 results between methods. Laboratories using the Capillarys 2 method would need to perform a reference range study before implementing the method and note that HbA2 values generated by their laboratory may be lower than quoted in the literature for other methods.
In patients with the HbD Punjab trait, the increase in the mean (VARIANT II, 1.5%; Capillarys 2, 2.8%) and range (VARIANT II, 1.1%–1.9%; Capillarys 2, 2.0%–3.6%) of HbA2 values between the VARIANT II and Capillarys 2 is significant (P < .0001), with the Capillarys 2 having higher values. In every other group we studied, the HbA2 values on the VARIANT II were higher than on the Capillarys 2. The observation by Dash8 and Cotton et al9 that HbA2 is decreased owing to integration error in the baseline by HPLC methods rather than a real decrease in value in patients with the HbD Punjab trait is confirmed by this study. On the Capillarys 2, the upper limit of the reference interval for HbA2 in patients with the HbD Punjab trait was not different from that found in patients without the β-thalassemia trait or a hemoglobin variant.
The increase in HbA2 means (VARIANT II, 2.95 vs 3.81, Δ 0.86; Capillarys 2, 2.49 vs 3.06, Δ 0.57; P < .0001) without and with HbS is different, with the greater difference noted with the HPLC method. This finding would suggest that any interference from HbS adducts that affects the HbA2 value in the Bio-Rad VARIANT II method and, by extension, all HPLC methods is also present, but to a lesser extent, in the Sebia Capillarys 2 capillary electrophoresis method. In contrast with the findings of Cotton et al,9 who found that there was no interference in HbA2 values generated by capillary electrophoresis in patients with HbS, we found that the HPLC HbA2 results are the same as those generated by the capillary electrophoresis method for 2 of 3 distinct groups of people with HbS and that any interference present in the Bio-Rad VARIANT II HPLC method is present in the Sebia Capillarys 2 method to a lesser extent. It is noteworthy that the HbA2 concentrations in patients with homozygous HbS, in whom any interference due to HbS adducts would be maximized, are identical. Also, the difference in HbA2 in patients with less than 30% HbS is smaller than that when the HbS is between 30% and 45% of the total hemoglobin.
The mean of HbA2 values by the Capillarys 2 in patients with the HbE trait was 3.65% (range, 2.85%–4.45%). The mean and range are higher than found in patients without the β-thalassemia trait. Bain15 commented that HbA2 in people with HbE is increased. This may be due to the decreased synthesis of the abnormal β-globin chain allowing for increased binding between the excess α and δ globin chains. Investigation of the CBC in these patients did not show thalassemia-associated parameters such as an increased RBC count. The mean corpuscular volume is decreased in patients with the HbE trait, and, therefore, it cannot be used to evaluate coinherited α-thalassemia. The mean HbE concentration on the Capillarys 2 was 23.91% (range, 20.41%–27.33%), which is significantly different from the mean HbE concentration of 30% (range, 27%–33%) found on the Bio-Rad VARIANT II system in previous reference range studies in our laboratory and stated by Bain15 to be normal. Patients with HbE concentrations less than 25% indicate the presence of coexisting α-thalassemia15 using current methods, and this value must be established for the Sebia Capillarys 2 system. One explanation for the lower HbE concentrations on the Capillarys 2 may be that the value of HbE on the VARIANT II represents the combined HbE and HbA2 concentrations owing to the coelution of HbE and HbA2, whereas on the Capillarys 2 system, HbA2 and HbE are resolved from each other, and, therefore, the value of approximately 24% probably represents the true concentration of HbE.
On the Sebia Capillarys 2, HbA2 and HbC are poorly separated. This is probably why there was poor agreement between results for HbA2 in the presence of HbC, with the Bio-Rad VARIANT II providing the better result because HbA2 is well resolved from the HbC peak.
The Sebia Capillarys 2 method shows similar analytic performance for HbA2 quantification to previously published work using capillary electrophoresis, but performance is not as good as the Bio-Rad VARIANT II method. However, the precision is acceptable for the diagnosis of β-thalassemia trait. A hematopathologist who reviewed the data expressed the opinion that if an appropriate reference range for HbA2 was quoted with all HbA2 results produced by the Capillarys 2, there was no difference in establishing the correct diagnosis of β-thalassemia trait or coinheritance of β-thalassemia with a hemoglobin variant.
The Sebia Capillarys 2 is superior to the Bio-Rad VARIANT II for the quantification of HbA2 in the presence of HbE and HbD Punjab. HbE concentrations in patients with the HbE trait obtained in the Capillarys 2 system are lower than from the VARIANT II and probably represent the true concentration of HbE. The Sebia Capillarys 2 offers minimal advantages over the Bio-Rad VARIANT II for the quantification of HbA2 in the presence of HbS and is not as good for the quantification of HbA2 in the presence of HbC.
Sebia and Somagen provided the Sebia Capillarys 2 and reagents for the study without charge.
This study was performed as part of the requirements for a degree in medical laboratory technology by Ms Mack.