Clinical Chemistry 44, No. 3, 1998
Table 2. Methods comparison with Bio-Rad Variant system: regression parameters calculated from the analyses on blood samples (top) and normalized residuals relative to control materials (bottom).
Tosoh Alc 2.2
1.010 0.027 2.10 0.22 0.981 0.36 50
1.027 0.056 0.24 0.43 0.924 0.76 49
1.031 0.038 0.04 0.30 0.966 0.51 48
r Sy x n Control materials
acetate electrophoresis pattern) and the ability of mini- mizing the normalized residuals, we then selected Mena- rini’s and our home-prepared controls as optimal materi- als to be used as calibrators. Therefore, to examine the potential utility of these control materials in standardiz- ing glyHb results obtained by different techniques, the original data relative to the differences between pairs of methods on patient samples were recalculated after cali- bration of the dependent method with pairs of materials. The cumulative distributions of such differences (in the original units of Hb A1c, % of total Hb) are presented in Fig. 2. For the purpose of clarity, in all plots of Fig. 2 we reported the distribution centiles 5th, 50th, and 95th as dotted lines, and the no-bias value with a vertical contin- uous line.
As can be seen, the distribution plots of the original
data (Fig. 2A) Variant HPLC,
were variously biased against Bio-Rad in agreement with the intercept values
Globally, glyHb results obtained by the different methods ranged from 2.12% Hb A1c units from the Bio-Rad Variant result (low limit of the Roche immunochemical
method) up to Hitachi HPLC)
3.1% units (upper limit of the Merck-
After recalculation, patient results calibrated with Menarini controls (Fig. 2B) were found to have a greater overlapping. Particularly, Menarini and Kontron results were almost superimposable to the Bio-Rad Variant data, and Table 3 reports such deviations in Hb A1c units. Calibration with Menarini controls was not so effective in reducing the bias of the other two HPLC methods (Tosoh and Merck-Hitachi). Moreover, calibration with the home- made controls Sam 1 and Sam 3 (Fig. 2C) produced a more
marked reduction in of 200 Hb A1c results
bias between (a total of n
methods, since only 13 7 from Merck-Hitachi,
3 from Tosoh, n
2 from Kontron, and n
0.5% (Hb A1c
units) off the value mea-
Unimate and Boehringer Tina-quant data.
Discussion The various methods currently available to measure glyHb in blood are known to be well correlated, but frequently biased among each other [15–17]. To this regard, our results were quite consistent with those from others, except that in most cases the Bio-Rad Diamat HPLC was used as the x-method in method comparisons trials. However, glyHb results obtained by Bio-Rad Dia- mat and Bio-Rad Variant HPLC were very similar and closely correlated [18, 11], so our comparison among Bio-Rad Variant and other methods can be compared with other studies in which the Diamat system was used (). In the case of Abbott IMx, the standardized Hb A1c Abbott data were correlated to Bio-Rad Variant results similarly to that found previously . However, sample residuals found in this study were slightly higher than those reported by Wilson et al. (Sy x 0.76% vs 0.57%, respective- ly). Also for the immunoturbidimetric Roche method, we found results well correlated to the Bio-Rad HPLC, with sample residuals slightly higher than those reported by Grey et al. (Sy x 0.52% vs 0.39%, respectively; ). A somewhat better agreement between the standardized immunoturbidimetric Boehringer method and Bio-Rad Variant HPLC was found, as already reported  with sample residuals from both immunoturbidimetric assays and Bio-Rad Variant, very similar in this and the previous study (0.51% and 0.52%). However, also taking into