Clinical Chemistry 44, No. 3, 1998
of the materials, the matrices in which they are dissolved, and the addition of preservative agents , we started our work by carefully looking at the Hb pattern in the chosen materials. Particularly, we decided to investigate total Hb concentration, the Hb derivatives pattern, and Hb electrophoresis. The analysis of Hb derivatives was considered important because most of the actual glyHb methods are based on spectrophotometric measurements in the Soret region (400–430 nm). In this spectral region the absorptivities of Hb derivatives are very high and even small changes in the wavelength may cause large changes in the glyHb value if an abnormal concentration of Hb derivatives or nonnative derivatives is present in the materials .
With regard to total Hb concentration, most of the materials were found similar to patient samples, with the exception of the Boehringer controls (Precinorm N and Precipath P), which were very dilute. These materials were therefore excluded from the intermethod compari- sons, also because of the high proportion of MetHb and because of the totally unusual pattern obtained from cellulose acetate electrophoresis. To this regard, Boehr- inger declares that Precinorm controls contain a mixture of human and sheep blood, but we were unable to find any trace of human Hb A, even on our high-resolution HPLC system (data not shown here). It is unclear if the other controls, which are all of human origin, do really originate from healthy subjects and from diabetic pa- tients. Only in the case of the Bayer materials is the origin from diabetic patients stated relative to the Bayer A (abnormal) control.
The composition of other materials was more similar to
native fresh blood, with the exception of Bayer and Roche controls, which were found to contain 10% MetHb. Abbott controls too were considered nonnative from this point of view, since they were all in the cyanMetHb form. The electrophoresis patterns confirmed the results of the spectral analysis, and also showed unusual findings with regard to the Bio-Rad controls, possibly because of the high content of preservatives added by the manufacturer to protect Hb during the lyophilization process. Despite such findings, Abbott, Bayer, and Roche controls were
By taking together the results from Hb pattern analysis and intermethod comparisons, we concluded that only Menarini and our materials were best ranked to be used as calibrators. An a posteriori correction of the original data indeed showed that bias between methods was drastically reduced after correction, especially for all the
HPLC Roche) clinical results
methods. For some methods (Boehringer and the cumulative distributions were still too wide for
use (i.e., obtained
1 Hb A1c % unit) and harmonization of by these methods was not successful.
Harmonization of cially by the use of with regard to the
Abbott’s results is achievable, espe- our home-prepared materials, but only mean bias obtained after calibration.
A possible explanation for the poor success obtained by calibrating Abbott, Boehringer, and Roche methods by Menarini’s and our materials used as calibrators could be the higher imprecision of these techniques with respect to the HPLC methods. As a matter of fact, sample residuals were higher in these method comparisons with respect to the comparisons among HPLC techniques only. With regard to the results obtained with the correction of the data from the Boehringer immunochemical method, the quasiparadox result (i.e., correction with calibration in- creased the absolute bias value) could be partially ex- plained by greater imprecision of this method with re- spect to HPLC. In such a case the study would have to be repeated with a greater number of replicates, either for patient or control specimens.
The inability of Menarini’s and our materials to har-
monize data obtained by the immunochemical techniques needs further investigation. An interlaboratory trial actu- ally under development in Italy, with 160 laboratories and several techniques, will compare the ability of Mena- rini’s and of our home-prepared controls in reducing the intermethod variability for several different techniques, including the Boehringer, Roche, and Bayer immuno- chemical methods. In such a study the measurements on the different materials are being performed in triplicate, so that the impact of method-dependent imprecision will
be minimized. As a matter of demonstrated that, by using our (previous lots), we were able
fact, we have already home-prepared controls to reduce intermethod
also tested together with the remaining and ours) for commutability by the analysis of methods comparison.
controls (Menarini linear regression
variability also for procedures ().
approach was adopted from other studies ([3, 4]), where the limits of 3 normalized residuals were chosen as cutoff values for noncommutability. From this point of view, most of the materials listed in Table 2 were consid- ered commutable, with the exception of two. For one (Bio-Rad Lyp II, in the comparison between Bio-Rad Variant and Abbott IMx), the finding of noncommutabil- ity was in agreement with the manifested matrix effect found in Sweden when using Bio-Rad freeze-dried con- trols in national surveys including laboratories perform-
ing the Abbott IMx method.
In conclusion, in this study about methods and control materials for glyHb analysis, we proved that relevant differences still exist among different methods and that commercial materials are very different among each other with respect to Hb composition and commutability prop- erties. Noncommutability appears to be rather frequent, and only our specially prepared human controls and Menarini’s were commutable for the majority of method pairs. Differential sources, derivatization of Hbs, and lyophilization processes can explain the absence of com- mutability. Since human source materials, which most closely mimic patients’ samples, may be used as inter-