Evaluation of a new semi-automated high-performance liquid chromatography method for glycosylated haemoglobins

The measurement of glycosylated haemoglobin in blood is vital for monitoring metabolic control in diabetics [1-3]. The development of many different methods of measurement (based on cation-exchange chromatography, electrophoresis, high performance liquid chromatography [HPLC], affinity chromatography, radio and colorimetric immuno techniques) proves the importance of this assay [4]. Since the first HPLC method was described [5] a number of modifications have been reported (see the review by Ellis et al. [6]). Instruments have been specially designed and commercialized (for example by Helena Laboratories, Beaumont, Texas, USA and Kyoto Daichi, Japan).


Introduction
The measurement of glycosylated haemoglobin in blood is vital for monitoring metabolic control in diabetics [1][2][3]. The development of many different methods of measurement (based on cation-exchange chromatography, electrophoresis, high performance liquid chromatography [HPLC], affinity chromatography, radio and colorimetric immuno techniques) proves the importance of this assay [4]. Since the first HPLC method was described [5] a number of modifications have been reported (see the review by Ellis et al. [6]). Instruments have been specially designed and commercialized (for example by Helena Laboratories, Beaumont, Texas, USA and Kyoto Daichi, Japan).
The evaluation of a recently commercialized semiautomatic HPLC instrument, which performs analyses on hemolysates from the original blood samples, is reported.

Materials and methods
The instrument, commercialized by Bio-Rad Laboratories (Segrate, Milano, Italy), consists of a main unit, containing the functional parts (autosampler, buffer reservoir, pumps, column colorimeter, and a control unit). A single piston pump is used with a step-gradient valve system, by which the elution of the column is performed with three phosphate buffers ofpH 5"7, 5"8 and 5"9 (unstated concentration) and increasing ionic strength. The analytical column (4"0 x 150 mm) is filled by a new cation exchanger (TSK gel, unstated composition). The column temperature is controlled at 23 C by forced ventilation in the column compartment.
The sample (5 !1 of venous blood, with mg/ml of EDTA) is diluted automatically with ml ofhaemolysing reagent (0"1% vol/vol polyoxyethylene-ether in borate buffer); 20 1 of the haemolysate are used for analysis. Spectrophotometric detection is performed at two wavelengths (415 nm and 690nm) to ensure a stable base-line. All the operations are controlled by a micro-* All correspondence to Dr Mosca. 192 processor and a built-in integrator performs data reduction. The final results are printed out together with a chromatogram of the run. The order of elution of the haemoglobin fractions is" HbAla, HbAb, HbF, HbAlc, HbAo. Each run takes 8 min. The reference method used for the evaluation was the Bio-Rad Laboratories microchromatographic method for HbAlc performed at 23 C, as previously described [7].
The alkali denaturation method, modified according to Molden [8], was used for HbF identification.
Blood samples used were collected by the Clinical Chemistry Laboratory and the Clinical Medicine Division (H. S. Raffaele, Milano, Italy). A Coulter S-Plus IV (Kontron, Switzerland) was used to measure the haematological parameters. Some ethylene glycol stabilized haemolysates were also prepared, according to a previously reported procedure [9], and used within two months.
Analytical imprecision was measured using blood samples and a lyophylised control material (Lypocheck Hemoglobin Alc, from Bio-Rad Laboratories).
Results and discussion  [10]. In order to test whether or not the amount of haemoglobin loaded to the column would interfere with analytical accuracy, as reported for the microchromatographic methods 11 ], samples with different haematocrit values and similar glycosylated haemoglobin levels were analysed. The samples were prepared by diluting packed red blood cells with different volumes of their own plasma. The results obtained showed that, for haematocrits between 22% and 75%, no appreciable effect is produced on the results of analysis, i.e. no haematocriteffect can be detected.
The analytical precision, within and between runs, was also analysed; the results obtained are reported in table 1.
In order to evaluate the within-series precision the analyses were repeated eight times on a blood sample from a normal subject, 10 times on blood from a diabetic patient and 17 times on the control lyophylised material. o/, "  The between-run precision was tested over a period ofone month, analysing two ethylene glycol stabilized haemolysates nine times (every third day). Each daily measurement was performed in duplicate.
The accuracy in the measurement of HbAlc was evaluated by measuring this fraction in 41 blood samples with both the present method and the reference method. The data obtained are shown in figure 2.
The accuracy of HbF measurement was determined on a limited number of samples, the data being reported in figure 3. It can be seen that the accuracy is not satisfactory. Further methodological improvements must be introduced to resolve this quantitatively minor fraction which has great clinical relevance for the characterization of the [3-thalassemia syndromes.
Interference by abnormal haemoglobins was investigated by analysing haemolysates from subjects suffering from various haemoglobinopathies, which had previously been characterized. Figure 4 shows that in all the specimens analysed, the instrument produced an abnormal chromatogram; if an abnormal haemoglobin was present it showed as an anomalous peak, indicated as P4, P5 or P6.
In the case with HbS (see figure 4[b]), the HbAo peak is split into two fractions, the last one (P6) probably corresponding to the HbS present in the sample. There is good correlation between the amount of this fraction (40"7%) and the HbS concentration in the sample (38%) determined separately by cellulose acetate electrophoresis (CAE) at pH 8"4. The position of the peak suggests also that the fraction may have an electrophoretic mobility slower than that of HbAo, as is to be expected for

HbS.
A somewhat similar comparison holds for the sample showing the HbJ Paris trait (see figure 4[d]). This haemoglobin variant has greater mobility than HbA0 on CAE, and it is therefore expected to be eluted among the first fractions. The chromatogram shows an elevation in the HbAa and HbAb fractions, which are eluted together. The concentration of these fractions (13" 1%) is too high even for a sample from a diabetic person, and indicates the presence of interference. The concentration found using this analysis is significantly lower than that measured by CAE.
In the case of the Hb Lepore trait (see figure 4[e]), a variant with electrophoretic mobility close to that ofHbF, an abnormal chromatographic pattern is shown, but detection of the type of the variant from the chromatogram is very uncertain.
Characterization is also impossible in the case of the sample showing the HbE trait (see figure 4[c]). This mutant has a very low electrophoretic mobility at pH 8"4 (similar to that of HbA2) and elution is therefore expected after HbAo. In this case the HbAo peak is not split, as with the HbS trait sample, but the chromatogram is abnormal at the boundary between the HbAo peak From our experience we can therefore conclude that the instrument is more sensitive to the presence of abnormal haemoglobins in the sample than are the minicolumns which are normally used. The operator should be alerted by the appearance of anomalous peaks but further electrophoretic investigations need to be performed from full diagnosis. The estimation ofthe true HbAc levels can be attempted by a careful observation of the chromato-gram, knowing that, under standard conditions, the elution time for HbAlc is about 4.8 min.
The influence of so-called 'labile glycosylated haemoglobin' fractions on this method was also examined.
Erythrocytes were incubated separately with glucose (4"0 g/l) and with saline solution (9"0 g/1 NaC1) at 37 C, for up to 24 h. No significant increase in the HbAc levels of three blood samples (one from normal and two from diabetic subjects) after 24h of incubation with the glucose solution was detected (see figure 5[a]). Similarly, no significant decrease in the initial HbAlc levels was found in two blood samples (one normal and one diabetic) incubated with saline for 24 h (see figure 5[b]). Furthermore, in order to evaluate the presence of the labile fractions in the incubated erythrocytes, the same samples were analysed for HbAlc levels, using an haemolysing reagent without borate (the same haemolysing reagent previously commercialized Bio-Rad Laboratories forassays of HbA1). This kind of reagent does not eliminate the aldimine forms during the haemolysis step.
In fact, the results ofthese determinations (see the dashed lines in figure 5) showed an increase in the HbA levels after the glucose incubations, and a significant decrease in the HbAc concentrations after the incubations with saline.
It can therefore be concluded that the method described here is insensitive to the presence of the 'labile fractions', which are reported to interfere strongly with many chromatographic methods.
In this evaluation no experiment was performed to test the effect of temperature on the analytical resolution of the minor haemoglobin fractions. Reports in the literature indicate that this effect is evident with all the procedures based on ion-exchange chromatography and, to a smaller extent, with those using affinity chromatography [12]. It is assumed that a thermoregulated column compartment within the apparatus will eliminate any interference from temperature variations on the chromatographic pattern.
In summary, it would appear that this system could be of significant value for the measurement of glycosylated haemoglobin. It combines good precision and accuracy with practicality for speed, technical skill requirements and safety. The system can also be useful in evaluating samples from subjects suffering from some haemoglobinopathies where diagnosis may frequently be missed by conventional chromatographic techniques, normally used to measure glycosylated haemoglobin levels.