| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Letters |
Washington University School of Medicine,
1
Department of Medicine, Box 8046, and,
2
Department of Pathology, Box 8118, 660 South Euclid Ave., St. Louis, MO 63110
aAuthor for correspondence. Fax 314-362-4782; e-mail bnowatzk{at}im.wustl.edu.
To the Editor:
Diabetes mellitus is a serious health problem in the United States, and much of the morbidity and mortality associated with diabetes mellitus can be ameliorated by proper blood glucose control (1). Consequently, serious effort has been put forth to increase the accurate and timely monitoring of blood glucose concentrations. Hemoglobin A1c (HbA1c) is a biomarker that provides a snapshot of long-term glucose control in diabetes. The Core Laboratory for Clinical Studies at Washington University School of Medicine is a Secondary Reference Laboratory (SRL no. 4) of the National Glycohemoglobin Standardization Program (NGSP) (2). We routinely measure HbA1c percentages from whole blood samples by the HbA1c Tina-quant® II method (Roche Diagnostics Corporation) on a Roche Hitachi 917 Analyzer. The Tina-quant II HbA1c assay is an NGSP-certified method (3) that quantifies the total hemoglobin mass (g/dL) spectrophotometrically and the HbA1c mass (g/dL) by turbidimetric immunoinhibition.
During the course of a method comparison of the Tina-quant II assay
with an HPLC method (Bio-Rad VariantTM), we
observed a positive bias at HbA1c percentages below 
5.5% (data not
shown). Additionally, we observed that our Level 1 quality-control
(5.6% HbA1c) material had a long-term imprecision that was nearly
one-half as variable as our Level 2 (10.3% HbA1c) material: CV =
1.7% vs 2.6%, respectively. These observations led to dilution
linearity experiments for the individual component measurements in the
Tina-quant II assay and revealed that although the hemoglobin
measurement diluted linearly, the A1c mass component had a positive
deviation from linearity at a HbA1c concentration of 5.3% (Fig. 1A
). When we inspected the cumulative NGSP monthly survey data
and incorporated selected low-percentage HbA1c whole blood NGSP
samples, we observed a positive bias at HbA1c percentages of 5.5%
or lower compared with the NGSP HPLC-assigned value (Fig. 1B
).
|
We concluded that the A1c mass measurement bias caused the observed
positive bias at HbA1c percentages <5.5%. To correct this bias we fit
a nonlinear function that was asymptotically linear in the limit:
 | (1) |
 |
 |
by nonlinear least-squares analysis to the A1c mass value in the
dilution analysis (STATA Release 6, College Station, TX). This
correction was then applied to every Tina-quant II A1c mass value, and
the percentage of HbA1c was recalculated using the raw hemoglobin
value. The resulting HbA1c percentage was finally corrected with a
linear regression slope and intercept to agree with the NGSP. These
data were plotted against the NGSP HPLC value (Fig. 1C
), demonstrating
an improved correlation at low HbA1c percentages.
The nonlinearity of the Tina-quant II assay compared with various HPLC methods has been noted previously (4). We report here one solution to correct this deviation. Although the observed bias at HbA1c values <6% has little clinical significance [these values are in the reference range (4–6% HbA1c), which is generally considered desirable], it is undesirable for certification efforts. Additionally, such bias may be confusing during method comparisons or when changing methodologies. It is not our intention to recommend this correction for routine laboratory testing because this action may require off-line data manipulation, change the status of the assay to that of "homebrew", and have little effect on patient care. However, this correction method has improved the agreement of the Tina-quant II assay with the NGSP surveys and will allow greater likelihood of certification of candidate methods by reducing the apparent bias at low HbA1c percentages.
Acknowledgments
We wish to acknowledge the cooperation and thoughtful discussion provided by Dr. Randie Little of the University of Missouri, Columbia, MO.
References
diabetes/ngsp/index.html
(accessed August 2000)..
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
) and A1c mass (•) using the
Tina-quant II HbA1c assay. (B), percent HbA1c method
comparison of the Tina-quant II (uncorrected data) vs the NGSP HPLC
method. The inset shows expanded percent HbA1c axes to
emphasize the positive bias at low HbA1c percentages.
(C), percent HbA1c method comparison of the Tina-quant
II vs the NGSP HPLC method. The Tina-quant II data have been corrected
by application of Eq. 1