Quantitative Measurement of HbA lc by an Immunoturbidimetric Assay Compared to a Standard HPLC Method

Transcription

1 CLINICAL CHEMISTRY Original Article, Quantitative Measurement of HbA lc by an Immunoturbidimetric Assay Compared to a Standard HPLC Method AHMAD HAMWI, MD, CHRISTIAN R. SCHWEIGER, MD, MARIO VEITL, MD, AND RAINER SCHMID, PHD Determination of hemoglobin Ale (HbA, c ) is one of the most important monitoring procedures for long-term control of diabetes mellitus. Several analytical methods have been developed for the measurement of glycohemoglobin (GHb). Those most frequently used are ion-exchange chromatography for HbA lc and affinity chromatography for total GHb. In this study, a new turbidimetric immunoassay for I IbA K. (Boehringer Mannheim, Germany) was evaluated that was performed on a Hitachi 9 clinical chemistry analyzer (Boehringer Mannheim). Good linearity in the range of 5% to 15% HbA u., within-run and between-run coeffi- cients of variation ranging from 2.4% to 5.9% were obtained. Results of 179 diabetic and nondiabetic patients showed good correlation to those of a routine HPLC method (r = 0.96). In addition, HbA2, HbS, and HbF in samples from nondiabetic patients were not detected by the immunoturbimetric assay and the "labile" HbA )c fraction (Schiff base) did not interfere with the new test. (Key words: Glycohemoglobin; High performance liquid chromatography; Immunoassay; Intermethod comparison; Diabetes mellitus; Hemoglobin variants) Am J Clin Pathol 1995;104: Measurement of hemoglobin Ale (HbA ]c ) currently is accepted as a way to monitor long-term blood glucose control in patients with diabetes mellitus, because it represents a weighted average of the daily mean glucose values of the preceding 2 to 3 months. 1 The formation of stabile HbAi c is a two-stage process with initial formation of a reversible Schiff base ("labile" fraction, "labile" intermediate) resulting from the attachment of glucose to the N-terminal valine of the /3-chains of hemoglobin A. This rapid equilibrium is followed by a slow Amadori-rearrangement to form the stable ketoamine linkage of HbAi c. Influence of the "labile" fraction on the measurement of HbA, c is undesirable, because results would reflect a combination of short-and long-term glucose levels. 2 The most frequently used methods for HbAi c are ion-exchange chromatography and affinity chromatography for total glycohemoglobin (GHb). HbA lc values measured by ion-exchange chromatography on minicolumns can be falsely increased in the presence of fetal hemoglobin (HbF), whereas the results obtained by HPLC are usually not affected. 2 ' 3 Moreover, different compounds such as the Schiff base, and the carbamylated and acetylated hemoglobin, coelute with HbA lc. 2 Therefore, the "labile" fraction is eliminated by lysing blood samples in low-ph buffer 4 before analysis, whereas the carbamylated adduct may still cause an analytical problem for the chromato- From the Institute for Medical and Clinical Laboratory Diagnosis. University of Vienna, A ustria. Manuscript received September ; revision accepted February, Address reprint requests to Dr. Hamwi: Institute for Medical and Chemical Laboratory Diagnostics, University of Vienna, Vienna General Hospital, Waehringer Guertel 18-, A-1090 Vienna, Austria. graphic determination of HbAi c in uremia. 3 Furthermore, the occurrence of hemoglobin (Hb) variants, such as HbAS, HbAC, HbAE, may influence HbA lc measurement using those ion-exchange HPLC methods that cannot separate and reliably quantify these hemoglobins. 5 In contrast to ion-exchange HPLC, affinity chromatographic methods are generally not affected by Hb variants and Hb derivatives. 5 Recently, immunoassays for highly specific determination of HbA C were developed that, in contrast to the chromatographic methods, are not affected by Hb derivatives and usually are not influenced by abnormal hemoglobins. 5,6 In this study, we describe Tina-quant Hemoglobin Ale (immunoturbimetric assay), a new homogenous turbidimetric HbA, c immunoassay (Boehringer Mannheim, Mannheim, Germany). This assay employs a mixture of polyclonal antibodies raised to a single synthesized HbAi c -specific epitope, comprising the first four amino acids at the N-terminal end of the /3-chain of hemoglobin A (valine-histidine-leucine-threonine) and the glucose attached to the /J-chain N-terminal valine by the ketoamine linkage 78. Sheep were used for immunization to achieve better yields of the antibody mixture. The antibodies were isolated using ammonium sulphate precipitation and DEAE-chromatography. We performed this assay on a Hitachi 9 (Boehringer Mannheim), a recently introduced automated clinical chemistry analyzer. We compared precision, linearity, and recovery of the immunoturbimetric assay test with that of our routine Shimadzu Hemoglobin-HPLC system (Shimadzu Europe GmbH, Duisburg, Germany) for measuring HbA lc. We correlated HbA lc results of 179 diabetic and nondiabetic patients obtained by both methods. In addition, we investigated the influence of the Schiff base ("labile" fraction) and different hemoglobins (HbA2, HbS, and HbF) on both the immunoturbimetric assay and the HPLC method. 89

2 90 CLINICAL CHEMISTRY Original Article Subjects MATERIALS AND METHODS Blood samples from 179 patients that were consecutively sent to our laboratory for HbA, c determination were assayed. None of them had abnormal hemoglobins. To check interferences of hemoglobin variants, nine patients' samples from the hemoglobin screening program were also investigated. These patients had no evidence of diabetes mellitus. Selection of patients was according to the frequency in our laboratory. Two of these subjects were newborns with major levels of HbF (>70%), two other cases had high percentages of HbA2 (26% and 28%), two other patients were heterozygous for HbS (30% and 50%), whereas the rest of the samples showed moderate elevations of the percentages of HbA2 and HbF not exceeding 10% of total hemoglobin for each of both fractions. Sample Preparation Venous blood samples anticoagulated with EDTA were prepared for assay within 2 hours after arrival in the laboratory. Before the analysis by HPLC, the hemolysate was prepared by adding 10 nh of whole blood into 1 ml of hemolysing reagent (NaH2P04; phosphate buffer at ph 5; Shimadzu Europe GmbH), then the lysate was incubated for 10 minutes at 56 C (4; procedure modified) to eliminate the labile glycohemoglobin (Schiffbase). Before analysis with the immunoturbimetric assay, blood samples were lysed by adding 10 nl of specimen to 1 ml of tetradecyltrimethylammoniumbromide (TTAB, hemolysing reagent; Boehringer Mannheim). Samples of patients with hemoglobinopathies were stored at 2 C to 8 C, and assayed by both methods within 5 days after collection. To assess the influence of the "labile" intermediate of HbA lc on the immunoturbimetric assay a modification of the sample preparation was developed. Erythrocytes were prepared from 24 patient samples with various HbA, c values by washing them first in isotonic saline solution and then dividing them into two equal portions. After resuspension in isotonic saline solution, one part was immediately analyzed under the original immunoturbimetric assay procedure, the other part was first incubated for 5 hours at 37 C to remove the labile glycohemoglobin (Schiffbase) and the samples were then processed as previously described. 9 Control Materials For precision studies, we used 2 control samples purchased from Boehringer Mannheim (Precinorm HbA lc [PN] and Precipath HbA ]c [PP]), two external controls supplied by Bio-Rad Laboratories (Lyphocheck Diabetes Controls Level 1 [LI] and Level 2 [L2]), and two pooled blood samples from nondiabetic and diabetic patients, respectively. Immunoturbimetric HbA, c Assay Assay principle. In this immunoturbidimetric test, HbA lc reacts with the anti-hba 1 c antibodies to give a soluble immunocomplex. In a second step, polyhaptens with many HbA ]c -specific epitopes, which are identical to the epitope used for raising the antibody mixture, are added in excess. The free anti-hba, c antibodies form insoluble aggregates with the polyhaptens, which are measured turbidimetrically at 340 nm (reference wave length 700 nm). The measured absorbance is inversely proportional to the concentration of HbAi c in the sample. The HbA C is expressed as percentage of total hemoglobin, which forms a stabile chromophore with TTAB and is determined simultaneously at 570 nm (reference wave length 660 nm). Procedure for Immunoturbidimetric Assay The analysis was performed on a Hitachi 9 analyzer (Boehringer Mannheim), according to the manufacturer's instructions. The HbA lc -assay was calibrated by a nonlinear calibration curve composed of five standards, whereas the measurement of total hemoglobin was based on a 2-point calibration. The values of the standards provided with the kit were assigned by an HPLC method described by Bisse and Wieland. 10 Ten pl of hemolysate were mixed with 250 iil antibody solution, incubated for 5 minutes, 50 /J.L of polyhapten mixture were added, stirred and after 10 minutes photometric measurement of HbA lc (340 nm and 700 nm, respectively) was performed. For total hemoglobin til hemolysate was mixed with 230 fih of oxidizing puffer, which yields a stable chromophore with TTAB and measurement occurred at 570 nm and 660 nm, respectively. HbA ]c -results are given as percentages of total hemoglobin. If HbAi c concentration is less than 3 g/l or the Hb concentration less than 50 g/l, according to the manufacturer, the measurement of HbA C should be repeated using twice the original sample volume. Samples containing HbA C concentrations greater than that of the highest standard should be reassessed using half of the original sample volume. Within-run precision of the immunoturbimetric assay HbA lc assay was determined using the two external control specimens Lyphocheck Diabetes Controls Levels 1 and 2, and a diabetic and a nondiabetic blood sample by replicate testing (n = ) within a single calibration run. Between-run imprecision was calculated employing the two internal controls Precinorm HbA lc and Precipath HbA lc, in addition to LI and L2. Using this method, all of them were analyzed over 2 weeks using a total of separate calibration runs. HPLC Method The HPLC method for the determination of HbA C established in our laboratory is based on the procedure described by Bisse and Wieland. 10 The reagents including PolyCATA cation-exchange column and the two buffers were purchased from Shimadzu Europe. EDTA-anticoagulated blood samples were analyzed on a Shimadzu HPLC system (Shimadzu, Kyoto, Japan) according to the manufacturer's instructions. The runtime was 7 minutes. HbA lc was separated from HbAO, HbF, and HbA2 (Fig. 1A). The assay was calibrated daily with an external HbA lc reference standard (Bio-Rad Laboratories, Hercules, CA). Calibration was done through the integration software by adjusting the calculated calibration curve to the target values of the calibrator. By this method, the Shimadzu HPLC system is standardized to the Bio-Rad Diamat analyzer. The nondiabetic reference range for HbAi c assayed by this HPLC method was 4% to 6% (reference range originally established in-house for the Diamat-method; the correlation between current method and Diamat: r = 0.987) and is in agreement with other HPLC methods." A.J.C.P.-July 1995

3 HAMWI ET AL. 91 Measurement ofhba lc by an Immunoturbidimetric Assay FIG. 1. Separation of hemoglobin fractions by HPLC: HbA, c analysis with original Shimadzu HPLC method (A), determination of HbA lc with modified Shimadzu HPLC (B) and quantification of Hb A2 and HbF using Variant HPLC analyzer (C). AM JM\ JK/L. /-J vl\ kin I I I I I I I I i G To study the effect of hemoglobin variants on the determination of HbAi c w 'th this HPLC method, we modified the original Shimadzu HPLC method by changing the steepness of the buffer gradient and extending the analysis time up to 22 minutes, thus separating and quantifying HBS and HbC (Fig. 1B). Experiments were done with a set of normal (non-variant) hemoglobin samples to verify that under the changed-run condition, identical HbA, c values were obtained as in the "short" run (data not shown). Other Procedures Abnormal hemoglobins were detected by electrophoresis on agarose gel at alkaline (Paragon Hb, Beckman, Brea, CA) and acid ph (Paragon acid Hb, Beckman). HbA2 and HbF were quantified by the /^-thalassemia Short Program run on the Variant HPLC analyzer (Bio-Rad). A dual analyte calibrator for HbA2 and HbF was used. The analysis time was 6.5 minutes. A typical chromatogram is shown in Fig. 1C. Statistical Analysis The relationship between the immunoturbimetric assay hemoglobin A lc and our routine HPLC test was assessed by applying the nonparametric method of Passing and Bablock." The correlation coefficient of HbA lc percentages between both methods was determined by means of linear regression analysis. Imprecision RESULTS The within-run and between-run coefficients of variation (CV) for the immunoturbimetric assay ranged between 2.4% and 5.9%. For comparison, within-run (n = ) and betweenrun imprecision (n = ) of the HPLC method were assessed with LI and L2. The corresponding coefficient of variations ranged between 1.2% and 2.6%. Results are shown in Table 1. Linearity Blood samples from a diabetic patient and a nondiabetic sub- ject were analyzed for HbA ic by the immunoturbimetric assay and our HPLC method. Using the blood with low HbAi c, a dilution series (n = 10) of the sample drawn from a diabetic patient was prepared. Each dilution was assayed in duplicate. Both the immunoturbimetric assay and HPLC showed acceptable linearity in the HbA lc and ranged between 5% and 15%. The deviations from theoretical values did not exceed 7% (mean = 2.7%; SD = 2.6%) for immunoturbimetric assay, and 8% (mean = 3.8%; SD = 3.1%) for HPLC, respectively (Fig. 2). Recovery of Control Materials Because "true" reference material for HbA lc is currently not available, the assigned values of the two quality assurance materials LI and L2 by the manufacturer have been used to calculate the "recovery" by relating the means of the between-run measurements to the corresponding target values defined by the manufacturer of the control material, which were derived from replicate analyses using the Bio-Rad Diamat analyzer. For the immunoturbimetric assay the percent recovery of low and high control was 92% and 105.6%, respectively (mean, 98.8%). For our routine HPLC method, the corresponding percent recoveries were 97% and 94.6%, respectively (mean, 95.8%). Method Comparison Blood samples from 179 patients sent to our laboratory were assayed by immunoturbimetric assay and our routine HPLC. When HbAi c values obtained with both methods differed by more than 1.5% HbAi c, analysis was replicated. The results of repeated samples were in good accordance with the original Vol No. I

4 92 CLINICAL CHEMISTRY Original Article TABLE 1. WITHIN-RUN AND IN BETWEEN-RUN IMPRECISION FOR HBAIc BY TINA-QUANT AND BY HPLC Tina -quant HbAic /o HPLC-HbAlc /o Within Between Within Between LI* L2* Nondiabetic Blood Diabetic Blood LI* L2* PN PP LIf L2\ LIf L2f N Mean SD CV (%) PN = Precinorm HbA, c ; PP = Precipath HbA )c ; SD = standard deviation; CV = coefficient of variation; HbAic = hemoglobin Ale; HPLC = high performance liquid chromatography. * Lot no. = and 55502(LI and L2). t Lot no. = and (LI and L2). value (mean of differences: 5% HbA, c for the immunoturbidimetric assay). The comparison of the HbAi c percentages measured by the immunoturbimetric assay (y) and HPLC (x) displayed an acceptable linear correlation: y = x; r = The mean of x was 7.19% HbAi c with a standard deviation of 2.04% HbA lc, SE was 0.152; the mean of y was 6.46% HbA ic with a standard deviation of 2.24% HbA lc, standard error of the estimate (SE) was In total, 179 samples were included with HbA ]c values ranging from 3.2% to 14.2%. Following manufacturer's recommendations, the percent HbA C values measured with the immunoassay were converted using the following parameters of a regression analysis generated in a multicenter correlation study (immunoturbimetric assay to Diamat HPLC system): percent HbA C (corrected) = percent HbAi c (measured). Using this mathematical formula, the converted immunoturbimetric assay HbA ]c values showed the following relationship to the corresponding HPLC results: corrected HbAi c (immunoturbimetric assay) = HbA lc (Shimadzu HPLC) (r = 0.96, the mean of xwas7.19% HbA C with a standard deviation of 2.04% HbA, c, SE was 0.152; the mean of y was 6.93% HbA, c with a standard deviation of 2.08% HbA, c, SE was 0.156; in total, 179 samples with percent HbA, c ranging from 3.9 to 14.1). The mean difference between both methods decreased from 0.7% HbA lc (SD, 0.6) (before correction of immunoturbimetric assay values) to 0.3% HbA C (SD, 0.6) (after correction of immunoturbimetric assay results) (Fig. 3). SchiffBase Elimination of the Schiffbase by incubation of washed erythrocytes in isotonic saline solution for 5 hours at 37 C before analysis with the Immunoturbimetric assay, did not give lower values in comparison with direct measurements of the samples without previous incubation (mean without incubation, 9.31%; SD = 2.08; mean with incubation, 10.1%; SD = 2.37; n = 24). The mean of HbAi c results of seven samples obtained with the immunoturbimetric assay (mean = 10; SD = 1.2) was by 17% lower than the mean of the HPLC values assayed before elimination of the Schiffbase (mean = 12; SD = 0.8), whereas the mean of the results measured by immunoturbimetric assay was in good agreement (93% of the mean of HPLC values) with that of the HPLC values after elimination of the labile fraction (mean = 10.7; SD = 0.8). Hb Variants HbAi was c analyzed in nine nondiabetic patients' samples from our hemoglobin screening program by both methods. Using our %HbA1cby Tlna-quant FIG. 2. Linearity for HbA, c in the range of 5% to 15% measured by Immunoturbimetric assay (Left) and Shimadzu HPLC (Right) Fraction of diabetic sample (%) Fraction of diabetic sample (%) A.J.C.P.-July I995

5 HAMWI ET AL. 93 Measurement ofhba lc by an Immunoturbidimetric Assay HPlC-HbAlc % FIG. 3. Correlation of immunoturbimetric assay (corrected %HbA, c ) and Shimadzu HPLC (%HbA, c ). Regression line: y = X Unity line: x = y Corrected %HbA, c were obtained by converting measured values using the formula: %HbA lc (corrected) = %HbA lc (measured) routine HPLC method, HbA, c is separated from HbAO, HbF, and HbA2. HbS elutes incompletely after the main HbA fraction leading to erroneously calculated HbAi c results. Therefore, we assessed HbAic in the two samples with HbS using the extended HPLC procedure allowing a complete elution, and thus a correct estimation of this Hb variant (Fig. IB). The HPLC results of the samples with high levels of HbF, HbA2, and HbS were corrected by adjusting observed HbA, c values for the respective portions of these fractions of total hemoglobin 5 : corrected percent HbA )c = (peak area of HbAi c * 100)/ (peak area of total Hb peak area of Hb variant). All samples showed HbA C results, which ranged from less than to equal to the reference range by both immunoturbimetric assay test and HPLC (Table 2). DISCUSSION The immunoturbimetric HbAi c immunoassay demonstrated an overall acceptable reproducibility both within-run (CV %) and between-run (CV 3.2%-5.9%) imprecision studies (mean of all CVs 4.3%). Although immunoturbimetric assay was less precise than our Shimadzu HPLC method (showing a level of imprecision [CVs less than 2.7%], which is equal or better than most methods currently available for the determination of GHb), 613 ' 9 the reproducibility of the immunoturbimetric assay was nevertheless comparable to that of many other GHb methods. 19 The lack of standardization is assumed to be one major reason for the dispersion of GHb values obtained by the different methods that measure either total GHb, HbA lc, or HbAp Weycamp and colleagues showed that a 3- point calibration using their own calibrators with assigned HbA lc values obtained by Diamat HPLC improved the imprecision of the various GHb assays and caused their results to agree more closely. 19 Both the immunoturbimetric assay and HPLC were linear within the range from 5% to 15%, which included almost all of the results found in diabetic individuals (98%). According to the manufacturer of the immunoturbimetric assay, the measurement of HbAi c should be repeated using twice the original sample volume, if the HbA, c concentration is less than 3 g/l or the Hb concentration is less than 50 g/l. In this study, a reanalysis of HbAi c concentration using the immunoassay was only necessary in three samples containing HbF or HbS. According to the manufacturer's recommendations, no rerun of hemoglobin had to be performed in any of the studied samples with nonvariant hemoglobins, the lowest measured Hb concentration being 65 g/l. However, Tiran and colleagues found a significant bias of the immunoturbidimetric assay at the extreme ends of the hemoglobin concentration range, showing higher values in anemic patients and lower values in polycythemic patients. This finding has led to revised manufacturer's instructions raising the lower limit for total hemoglobin to 90 g/ L and halving the amount of hemolysing reagent from 1:100 to 1:50. Both the immunoturbimetric assay and HPLC showed a good "recovery" of two external HbAi c control samples (Bio- Rad Lyphocheck Diabetes Control levels 1 and 2). In the absence of primary reference standards with absolutely "true" HbAi c values, we used the target values given by the manufacturer resulting from replicate analyses using the Diamat HPLC system. We assume that it is profitable to assess the recovery of TABLE 2. MEASUREMENT OF HbAlc IN NONDIABETIC PATIENTS WITH ABNORMAL HEMOGLOBINS % HbA 1 c (Tina-quant) % HbA 1 c (HPLC) Hemoglobin Variant Measured Standardized Before Correction After Correction HbF (72%) HbF (71%) HbA2(26%), HbF (2.7%) HbA2 (28%) HbS (43%), HbA2 (4%) HbS (31%) HbA2(4.8%), HbF (10%) HbA2(3.7%), HbF (4.4%) HbF (3.9%) * 2.2* 5.2* 5.8* * 4.8* 5.4* 4.3* 4.5* t 4.5f 4.3t 5.9t 4.8f 5.4t NC NC NC * Standardized %HbAIc(to Diamat HPLC) = %HbAlc (measured). t Corrected %HbA I c = (Area of HbAlc* 100* calibration factor)/(area of total Hb-area of abnormal Hb). NC = no correction, correction was only performed when % of HbF, HbA2 or HbS exceeded 10% of total Hb. Vol. 104-No. 1

6 94 CLINICAL CHEMISTRY Original Article the immunoturbimetric assay and our routine HPLC method relative to the widely used above-mentioned system, which is calibrated on the appreciated method of Goldstein and associates. 21 The comparison of the HbA, c results assessed by the immunoturbimetric assay (y) and HPLC (x) displayed an acceptable linear correlation between both methods: y = x; r = 0.96; n = 179. Nevertheless, the significant negative intercept of the regression line reflects a bias of the immunoturbimetric assay toward lower HbA lc values. One possible reason for these lower values obtained by the immunoturbimetric assay could be a consequence of the higher specificity of this method to N-terminal glycation in comparison with HPLC, as two peaks within the HbA lc fraction were frequently observed in samples with low HbA lc percentages. 3,6 Another possible reason for this observation may be the different calibration procedures and materials. To adjust the immunoturbimetric assay results to the Diamat HPLC, the manufacturer recommends to convert percent HbA, c measured with the immunoassay (multicenter trial correction factor: percent HbA lc (corrected) = percent HbA) C (measured)). After converting the immunoturbimetric assay HbA lc values using this factor, we found the following relationship: corrected HbA lc (Immunoturbimetric assay)= HbA, c (Shimadzu HPLC); Figure 3 illustrates the closer fit of both HbA lc assays because of the calibration of our routine HPLC method to the Diamat HPLC. The mean difference between the immunoturbimetric assay and Shimadzu HPLC decreased from 0.7% HbA lc (before correction) to 0.3% HbA lc (after correction). In conclusion, the bias of the immunoturbimetric assay toward lower results resulted mainly from the use of different calibration procedures and standards. However, differences in the specificity of both assays for HbA lc seem to be an additional factor for discrepancies, especially between the results of samples with normal and moderately increased HbA lc values obtained by both methods. Thus, in the absence of an accepted reference standard and an acknowledged reference method for measuring GHb, we assume that adjusting the HbAi c results obtained by this new immunoturbimetric assay to those obtained by the widely used Diamat HPLC is a good approach toward interlaboratory standardization and harmonization of the assays for the determination of GHb. 14 ' 5 However, we have noticed that the variability between the two assays demonstrated by the scatter around the regression line was still not negligible even after using the converted immunoturbimetric assay values (Fig. 3). This may be explained by the high CV of the immunoturbimetric assay method. Although in the HPLC analysis the removal of the Schiff base is necessary for an accurate determination of HbA lc, it seems that the immunoturbimetric assay is not affected. In fact, the HbAi c results of a group of poorly controlled diabetic individuals obtained with the immunoturbimetric assay were very similar to those measured by HPLC after a complete elimination of the labile glycohemoglobin (Schiff base) through lysing blood samples in low-ph buffer. 4 Furthermore, when reanalyzing samples with the immunoturbimetric assay after incubation in saline (5 hours, 37 C) to remove the "labile" intermediate, 9 they did not show lower values than those done without incubation. Thus, the insensitivity of the immunoturbimetric assay toward the labile fraction represents an additional advantage of this method in comparison to HPLC. The study of nondiabetic blood samples containing different amounts of HbF (up to 72%), HbA2 (up to 28%) and HbS (range 31 %-43%) showed that these hemoglobins were not detected by the immunoturbimetric assay. All measured HbA lc results were under or within the reference range recommended by the manufacturer (3.9% 5.7%). Thus, we can assume that the polyclonal antibody of the immunoturbimetric assay is specific for HbA lc. However, this observation needs further confirmation by studying other common hemoglobin variants such as HbC, HbD, and HbE. Concerning the studied hemoglobins, the immunoturbimetric assay seems to behave similarly to the Dako 6 and the Bayer immunoassays 5 when measuring HbAic in nondiabetic individuals. Nevertheless, significant differences could arise in the different immunoassays while measuring HbA lc in patients with both hemoglobinopathies and diabetes mellitus. Although the Dako antibody is reported to recognize only HbA lc, the antibody of the Bayer assay is said to also recognize Hb variants with glucose moieties at the /3-chain N-terminal. 5 ' 6 Unfortunately, no diabetic patients with sickle-cell anemia or ^-thalassemia could be included in our study. However, the fact that the HbAi c results of the studied nondiabetic patients with high HbS and HbA 2 levels, obtained by immunoturbimetric assay, were not less than the reference range leads to the assumption that glycated HbS and HbA 2 are recognized by this test. In contrast to these results, the HbA,c results of the two newborns with large amounts of HbF were falsely low as its glycated form cannot be recognized by the polyclonal antibody of the assay (Table 2). Immunoturbimetric assay measuring N-terminal glycated HbS, HbA 2, and other Hb variants presenting the epitope recognized by the immunoturbimetric assay polyclonal antibody should be studied further. Similar to the immunoturbimetric assay, the HbA lc values in nondiabetic patients using our routine HPLC method were not increased, even when large amounts of HbF, HbA 2 and HbS were present, showing the specificity of HPLC for HbA) C. However, the HPLC results of these specimens were mostly below the reference range (4%-6%) because of the common variant hemoglobins. HbA 2, HbS, HbC, HbE, and HbD are more positively charged than HbA, and their glycated forms elute later than HbA lc, leading to low GHb results " 18 For subjects with heterozygous hemoglobinopathies and Hb synthesis variants, an accurate estimation of the HbA lc would only be possible if the measured HbA lc could be corrected for the percentage of the abnormal hemoglobin present in total hemoglobin. Fortunately, most HPLC methods for the analysis of HbA lc recognize hemoglobin variants. 5 In this study, the HPLC-HbAi c results of all studied nondiabetic patients with sickle-cell anemia and Hb synthesis variants were within the reference range after correcting them for the portions of abnormal hemoglobin (Table 2). The use of HPLC for measuring HbAi c allows the detection of the common hemoglobin variants, thus avoiding a false interpretation of the results (these patients have decreased erythrocyte half-life in the sequence HbAC>HbAS>HbCC> HbSC>HbSS). 18 In contrast to HPLC, the user of an immunoassay such as the immunoturbimetric assay may falsely interpret the GHb results because of a lack of information on these conditions (no chromatogram). Therefore, when using the immunoturbimetric assay in the routine, typing of the patient's hemoglobins should be performed in cases of suspected hemoglobinopathy. We believe that in patients with abnormal he- A.J.C.P.^ July 1995

7 HAMWI ET AL. 95 Measurement ofhba lc by an Immunoturbidimetric Assay moglobins, the interpretation of GHb measured with HPLC or with an immunoassay may be questionable anyway, as different factors such as decreased erythrocyte half-life, heterocellular distribution of hemoglobins, hemolytic crises, and blood transfusions can affect the results. In these relatively rare cases, the measurement of fructosamine may be a more suitable tool for long-term glucose control. 18 In principle, the immunoturbidimetric assay method can be applied to other analyzing systems such as the Cobas centrifugal analyzer (Hoffmann-LaRoche, Basel, Switzerland). The basic requirements are the handling of up to two reagents per test channel, simultaneous measurement at two different wavelengths, and a wavelength range from 340 nm to 700 nm. In conclusion, this study shows that the immunoturbimetric assay is a reliable method for measuring HbA, c, which compares well with our Shimadzu HPLC system, especially when using the correction factor recommended by the manufacturer. The application of this assay on a clinical chemistry analyzer allows a large sample throughput, reducing the need for highly qualified laboratory staff as well as costs. However, to be a final alternative for ion-exchange HPLC methods, the precision of the immunoturbimetric assay should be further improved toward the clinically desirable intralaboratory CV of <3.5%. 19 Acknowledgments. The authors thank Anneliese Nakrour for her skillful technical assistance. The authors also thank Prof. Havelec (Institute for Medical Statistics, Vienna, Austria) for her encouragement and her assistance with statistical analysis. REFERENCES 1. Mogensen CE. Management of diabetic renal involvement and disease. Lancet 1988;1: Goldstein DE, Little RR, Wiedmeyer HM, et al. Glycated hemoglobin: methodologies and clinical applications. Clin Chem 1986;32(suppl):B64-B Engbaek F, Christensen SE, Jespersen B. Enzyme immunoassay of hemoglobin Ale: Analytical characteristics and clinical performance for patients with diabetes mellitus, with and without uremia. Clin Chem 1989;35: Bisse E, Berger W, Fliickiger R. Quantitation of glycosylated hemoglobin: Elimination of labile glycohemoglobin during sample hemolysis at ph 5. Diabetes 1982;31: Weykamp CW, Penders TJ, Muskiet FAJ, Van der Slik W. Influence of hemoglobin variants and derivatives on glycohemoglobin determinations, as investigated by 102 laboratories using 16 methods. Clin Chem 1993;39: John WG, Gray MR, Bates DL, Beacham JL. Enzyme immunoassay: A new technique for estimating hemoglobin A lc. Clin Chem 1993;39: Hamwi A, Stamminger S, Gasperlmaier M, Schmid R. Quantitative measurement of HbA, c in whole blood by a turbidimetric immunoassay on a BM/Hitachi 9 (Abstr). Clin Chem 1993;39: Bieger W, Bisse E, Gillery P, et al. Multicenter evaluation of a new immunoturbidimetric assay for the determination of HbA lc on clinical chemistry analyzers. Lab Med 1993; 17: Sacks DB. Carbohydrates. In: Burtis AB, Ashwood ER, eds. Tietz Textbook of Clinical Chemistry, ed 2. Philadelphia: WB Saunders, 1994, pp Bisse E, Wieland H. High-performance liquid chromatographic separation of human hemoglobins, simultaneous quantitation of fetal and glycated hemoglobins. J Chromatogr 1988;434: Brenna S, Prencipe L, Montalbetti N, Perlangeli V, Palpon R. Assessing the automatic Diamat liquid chromatograph for glycated hemoglobins (technical brief). Clin Chem 1986; 32: Passing H, Bablock W. A new biometrical procedure for testing the equality of measurements from two different analytical methods. Eur J Clin Chem Clin Biochem 1983; 21: Torke NS, Mehta PS, Vohra RM, Patel AP, Shah PC. Comparative analysis of glycosylated hemoglobins by ion-exchange column chromatography and high-performance liquid chromatography in patients with hemoglobinopathies. Am J Clin Pathol 1988;90: Bodor GS, Little RR, Garrett N, Brown W, Goldstein DE, Nahm MH. Standardization of glycohemoglobin determinations in the clinical laboratory: three years of experience. Clin Chem 1992;38: Bruns DE. Standardization, calibration, and the care of diabetic patients (editorial). Clin Chem 1992;38: Fliickiger R, Berger W. Monitoring of metabolic control in diabetes mellitus: Methodological and clinical aspects. In: Fliickiger R, Berger W, Koppe HG, eds. Progress in Clinical Biochemistry and Medicine, vol 3. Berlin: Springer-Verlag, 1986, pp Aleyassine H. Low proportions of glycosylated hemoglobin associated with hemoglobin Sand hemoglobin C. Clin Chem 1978;25: Martina WV, Martijn EG, Van der Molen M, Schermer JG, Muskiet FAJ. B-terminal glycohemoglobins in subjects with common hemoglobinopathies: Relation with fructosamine and mean erythrocyte age. Clin Chem 1993;39: Weykamp CW, Penders TJ, Muskiet FAJ, van der Slik W. Effect of calibration on dispersion of glycohemoglobin values determined by 1 laboratories using 21 methods. Clin Chem 1994;40: Tiran A, Pieber T, Tiran B, et al. Automated determination of glycated hemoglobin: Comparative evaluation of five assay systems. J Clin Lab Anal 1994; 8: Goldstein DE, Little RR, England JD, Wiedmeyer H-M, McKenzie EM. Methods for quantitating glycosylated hemoglobins: High performance liquid chromatography and thiobarbituric acid colorimetry. In: Clarke WL, Lamer J, Pohl SL, eds. Methods of Diabetes Research, vol 2. Clinical Methods. New York: John Wiley, 1986, pp Vol. 104-No. 1