CC r-

HE
Go ..-.#{149}
-p70-os
p 7001
CLINICALCHEMISTRY,Vol. 28, No. 12, 1982 2405
OLIN. CHEM. 28/12, 2405-2407 (1982)
Four Methods for Glucose Assay Compared for Various Glucose
Concentrations and under Different Clinical Conditions
Ottavio Giampietro, Alessandro P110,Giuseppe Buzzigoli, Claudlo Boni, and Renzo Navalesi
Glucose was measured by the ferricyanide, the Beckman
glucose oxidase, and the hexokinase procedures in 228
plasma samples taken during standard oral glucose-tol-
erance tests in 17 normal subjects and in 21 chemical
diabetics. The neocuproine method was also used to
measure glucose concentration in 156 samples (78 before
and 78 after dialysis) collected from six diabetic and uremic
patients who were on maintenance hemodialysis. Fern-
cyanide in all conditions and neocuproine in uremic pa-
tients overestimated glucose concentrations over the
entire experimental range as compared with either enzy-
mic method. This bias or systematic error of the reducing
vs the enzymic procedures, due to nonglucose reducing
substances (saccharoids), becomes considerably
greater when their concentration is increased as in chronic
uremia. Also, the inverse relation between glucose con-
centration and overestimation of glucose by the reducing
methods has been detected. With respect to the hexoki-
nase method, a mild but significant underestimate of glu-
cose oxidase readings has been observed for higher glu-
cose concentrations. We find neocuproine to be the most
imprecise of these procedures.
Addftlonal Keyphrases: glucose oxidase Beckman Glucose
Analyzer ferricyanide hexokinase neocuproine
uremia #{149} diabetesmellitus
Methods forglucosemeasurement based on the reducing
propertiesof glucose,because of interference by so-called
saccharoids (nonglucose reducing substances) (1), sys-
tematically overestimate the true plasma glucose concentra-
tion, as compared with enzymic procedures (2). This sys-
tematic error, or bias, is especially relevant in chronic uremia,
in which these interfering substances are present in concen-
trations much higher than normal (1, 3). Here we report re-
sults of a comparison between two reducing (ferricyanide,
neocuproine) and two enzymic (glucose oxidase, hexokinase)
procedures for glucose analysis under various clinical condi-
tions.
Materials and Methods
Glucosewas measured in228 plasmasamplestakenduring
standard oral glucose-tolerance tests in 17 normal subjects and
in 21 with diabetes mellitus. Each sample was assayed by the
ferricyanide procedure, automated on a Technicon AutoAn-
alyzer(4);the hexokinase procedure, automated on a LKB
8600 Reaction Rate Analyzer (5); and the glucose oxidase
procedure, with a Beckman Glucose Analyzer (6).
Clinica Medica II and Cattedra di Malattie del Ricambio dellU-
niversit#{224} di Pisa; Istituto di Fisiologia Clinica del C.N.R. di Pisa,
Italy.
Address correspondence to O.G., at the Istituto di Fisiologia Clinics
delC.N.R.,Via Savi 8, 56100 Pisa, Italy.
Received Sept. 21, 1981; accepted July 13, 1982.
Next, we measured glucose by these same methods and also
by the neocuproine method, automated on a Technicon Au-
toAnalyzer (7), in 156 samples (78 before and 78 after dialysis)
collected from six diabetic and uremic patients who were on
maintenance hemodialysis; the dialysate was free of glucose.
The venous blood samples were all collected during one
month, a sample being withdrawn before dialysis, and the
othersample from thesame patientafter4 to5 h of dialysis.
Each blood sample was collected, separated, and stored as
reported elsewhere (8). Creatinine and nonprotein nitrogen
(azotemia) were measured in uremic sera by usual methods.
Results
Because results by the various analytical procedures for
normal individuals and the diabetics were superimposable,
we show in Figure 1 only those for the diabetics, for whom the
range of glucose concentrations is broader. Ferricyanide sig-
nificantly overestimates glucose concentrations over the entire
experimental range as compared with either enzymic method.
Values by the latter methods tend mostly to concordance,
though glucose oxidase gives generally higher readings than
>0
0
>5
2>0.
90.
l70.
50.
30.
>0.
090.
070.
050.
BAS 30 60 90
I I ME (minUtes)
ito
Fig. 1. Mean (and SE) plasma glucose concentrations during an
oral glucose-tolerance test in subjects with diabetes, assayed
with the glucose-oxidase (Gox), ferricyanide (FC), and hexoki-
nase (HK) methods
Comparison by Students paired t-test
450
4.00.
350
3.00.
2.00
.50,
I.00.
4.00
3.50.
3.00.
2.50
4,
0
U
>5 2.00
0
55.
I.00
NC .----o
CC D-O
00, #{149}-.
HE
#{149} = p 000
P0.0i
I t t I I I
0=12 fl23 fl16 014 fl-S fl 3
FC 1-,
NC ___.
001 #{149}-#{149}
HI>
#{149} - p =0.05
p 0.0,
0 00 4 ISO 4200 4250
0 L U C 0 8 B 0/i.
300 =35Q b
1.00
I I I I t
0=5 n_li n=2o n-So fl#{149}14 n-S
Fig. 2. Mean (and SE) plasma glucose concentrations measured in six diabetic uremics before(a)and after(b)dialysis, by the FC
neocuproine (NC), Gox, and HK methods
Each series of 78 specimens was divided on the basis of the glucose concentration Into six groups: <1 g/L, ito 1.5 g/L, 1.5 to 2 g/L. 2 to 2.5 g/L, 2.5 to 3 g/L
3 to 3.5 g/L; mean values of each glucose group, obtained with the four methods, were compared both before and after dialysis. n - number of specimens In eact
group of glucose concentrations. Comparison by Students paired f-test. ns, not significant
41.50 42.00 =250 4300 35(
GLUCOSE 0/L
2406 CLINICAL CHEMISTRY, Vol. 28, No. 12, 1982
hexokinaseexceptformean glucosevalues<1.00 g/L, when
hexokinase gives higher values.
As for uremics: (a) Before dialysis (Figure 2a) both of the
reducing methods significantly overestimate-ferricyanide
more so-the glucose concentrations over the whole range
explored, as compared with the enzymic procedures, of which
glucose oxidase provides the higher values, except at mean
glucose values <1.00 g/L; (b) After dialysis (Figure 2b) the
blood creatinine and azotemiaof uremicswere changed re-
spectively from mean valuesof160 mg/L to 70 mg/L and from
1900 mg/L to 850 mg/L. The two reducing methods gave sig-
nificantly higher values than the two enzymic assays, but,
under these conditions, the difference between the reducing
and enzymic methods was smaller. Both ferricyanide and
glucose oxidase gave consistently higher readings than neo-
cuproine and, respectively, than hexokinase, although this last
method gave higher values than did glucose oxidase near 1.00
g/L.
Table 1 reportsthe regression lines and the comparisons
foreachmethod,undertheseclinical conditions, vstheglucose
oxidase method. As may be seen both from the slope and the
intercept of the regression line and from the values in Table
1, ferricyanide gave higher glucose values (p <0.01) than did
glucose oxidase, by about 0.08 g/L allthrough the experi-
mental range of glucose concentrations in the diabetics (and
the normals). In uremics, before dialysis, ferricyanide over-
estimated glucose concentration as compared with glucose
oxidase, by about 0.28 g/L; this overestimate is of approxi-
mately constant magnitude (bias or systematic error) for
glucoseconcentrationbetween 1.00 and 3.00 g/L; after dial-
ysis, the bias of ferricyanide is decreased to about 0.16-0.20
gIL. As for the behavior of hexokinase, in chemical diabetics
(as in normals) and in uremics, a mild but significant under-
estimate of glucose-oxidase readings may be observed for
higherglucoseconcentrations (Table1).With respecttothe
neocuproinemethod,the dispersion about the regression line
considerablyexceeds that for the other methods (see the
standarderrorsofestimateinTable 1),which means thatthis
is the most imprecise method (9). The neocuproine results
exhibit a similar but smaller positive bias than do the fern-
cyanide vs glucose oxidase readings.
Discussion
We used the glucoseoxidasemethod as the comparison
method because of its better precision and practicability in
our laboratory. Moreover,itreportedly(3, 9, 10) is a stable
method, specific for glucose, uninfluenced by the interfering
substances thatareespecially presentin the uremic samples.
Our data confirm that ferricyanide gives higherglucosevalues
than enzymicmethods do,both under normal conditions and
when therearemany circulating interfering substances. The
fernicyanide-assay error due to nonglucosereducingsub-
stances becomes considerably greater when their concentra-
tion is increased as in chronic uremia. Clearly, the positive bias
shown by the ferricyanide method is much higher in uremic
patients before dialysis, decreases after dialysis, and is much
less in chemical diabetics. Limited to uremic patients, similar
conclusions can be drawn forneocuproine, which gives higher
(+7.8%)
HK 0.99
(-0.5%)
Uremicsbeforedialysis (B)
FC 1.298 2.298
(+28.6%)
HK 0.98
(-1.8%)
NC 1.138
(+ 13.4%)
(+ 14.8%)
1.958
(-2.4%)
(+ 10.2%)
2.928
(-2.6%)
(+2 9%) a narrow range of glucose concentrations (most of them were
1 89a around 0.80-1.20g/L) and so theirconclusionsmay not be justified forhigherbloodglucoseconcentrations. In the other
(5.3%) paper (3), from the experimentaldata (see Table 3 in the
paper of Powell and Djuh) it is clear that the percent bias
3.31 a shown by fernicyanide over the o-toluidine readingsisin-
verselyrelatedto glucoseconcentration, althoughthe bias
increasesas an absolutevalue(g/L) with increase in glucose
(3).
. Resultsby thehexokinasemethod aregenerallyverynear
to those of glucose oxidase, but tend to be a little lowerinall
(+ 11.2%) (+ 10.5%) clinical conditions-and become significantly loweraround
1.8-2 g/L. However, hexokinase, like glucose oxidase, mea-
3 15a sures no uremic interferences, as may be seen from the nearly
(+8.8%) (+5.2%) absolute agreement between glucose measurements before
1.948 2.848 and after dialysis.
Uremicsafterdialysis (C)
FC 1.198 2.17a
(+19.5%)
HK 1.04
NC
(+3.8%)
1.08
(+8.3%)
(-3.1%)
2.10
(+5.2%)
(-5.4%) 3 12 b In conclusion, our data confirm that reducing procedures
forglucoseassayshouldbe discontinued(3, 9, 10), because
(+4.1 #{176}) they are strongly affected by interfering substancesand (to
a lesser degree) also by glucose concentration excursions. Thus
these methods may involveerrorsin the glucosereadings
mainly in conditionsof borderlinecarbohydrateabnor-
mality.
CLINICAL CHEMISTRY, Vol. 28, No. 12, 1982 2407
Table 1. Mean Plasma Glucose Concentrations as
Measured by the Ferricyanide (FC), Hexokinase
(HK), and Neocuproine (NC) Methods, Compared
with the Glucose Oxldase (Gox) Method
1.00 2.00 3.00
Glucose, g/L.
Normals and diabetics (A)
FC 1.08 2.06a
These data have been computed from the regression lines for each method vs
glucose oxidase values in correspondence to the concentrations of 1.00, 2.00,
and 3.00 g/L. The percent differences from the glucose oxidase readings are
also reported. The equations of the regression lines, the respective correlation
coefficients (,), and the standard errors of estimate (SEE) are as follows:
A. FC = 0.098 + 0.98 Gox; r = 0.99, SEE = 0.051.
HK = 0.096 + 0.90 Gox; r = 0.99, SEE = 0.064.
B. FC = 0.276 + 1.01 Gox; r = 0.99, SEE = 0.084.
HK = 0.012 + 0.97 Gox; r = 0.99, SEE = 0.115.
NC = 0.044 + 1.09 Gox r = 0.96, SEE = 0.257.
C. FC =0.215 + 0.98 Gox; r = 0.98, SEE = 0.127.
Ilk = 0.137 + 0.90 Gox; r = 0.98, SEE = 0.131.
NC 0.063 + 1.02 Gox; r 0.95, SEE = 0.283.
The significance of the differences of each value from the glucose oxidase
value shown in the heading of the columns has been computed from the conf I-
dence limits of the respective regression lines. >0p<0.01; #{176}p 0.05.
readings than glucose oxidase both before and after dialysis,
although less so than fernicyanide. Finally, the positive bias
of reducing vs enzymic procedures remains higher in dialysed
uremics than the biasin normals,as would be expectedif
circulating interferences in chronic uremia are not completely
removed by thesingledialysis.
The relationbetween glucoseconcentration and overesti-
mation of glucose by the reducing methods should be noted.
Reportedly (9) the positive percent bias of neocuproine is
related to the glucose concentration, that forglucosecon-
centrations of 0.90-1.00 g/L being about twice that observed
for concentrations of 2.80-3.00 g/L (9). These data, which were
obtained on control sera (9),are confirmedby our data on
uremic samples. Moreover, fernicyanide gives higher glucose
readings than does glucoseoxidase,the difference beingap-
proximately constant in grams per liter, both in normal con-
ditions and in the presence of many interfering substances
(before and after dialysis); evidently the percent error by the
ferricyanide method is higher at lower glucose concentrations
(8, 29, and 20%, respectively, in chemical diabetics, in uremics
before dialysis, and in uremics after dialysis, at 1.00 g of glu-
cose per liter) and becomes lower at higher glucose concen-
trations (3, 15, and 9% in the three groups at 2.00 g of glucose
per liter). These data do not accord with some others (3,11).
Those authors concluded that the higherthe glucosecon-
centration, the higher the error or positive bias exhibited by
fernicyanide, respectively, overglucoseoxidase(11) or over
o-toluidine (3), a procedurevery similartoglucoseoxidase(1,
3). But Eschwege etal.(11) studied patients specimenswith
References
1. Pileggi, V. J., and Szustkiewicz, C. P., Carbohydrates. In Clinical
Chemistry: Principles and Technics, R. J. Henry, D. C. Cannon, and
J. K. Winkelman, Eds., Harper & Row Publishers, Inc., Hagerstown,
MD, 1974, pp 1265-1325.
2. Mager, M., and Farese, G., What is true blood glucose? A com-
parison of three procedures. Am. J. Clin. Pathol. 44, 104-108
(1965).
3. Powell, J. B., and Djuh, Y.-Y., A comparison of automated methods
for glucose analysis in patients with uremia before and after dialysis.
Am. J. Clin. Pathol. 56, 8-16 (1971).
4. Hoffmann, W. S., A rapid method for the determination of glucose
in blood and urine. J. Biol. Chem. 120, 51-55 (1937).
5. Schmidt, F. H., Die enzymatische Bestimmung von Glucose und
Fructose Nebeneinander. KIm. Wochenschr. 39, 1244-1247 (1961).
6. Kadish, A. H., Little, R. L., and Sternberg, J. C., A new and rapid
method for the determination of glucose by measurement of rate of
oxygen consumption. Clin. Chem. 14, 116-118(1969).
7. Bittner, D. L., and Manning, J., Automated neocuproine glucose
method: Critical factors and normal values. In Automation in Ana-
lytical Chemistry, Technicon Symposia 1966, 1, N. B. Scova et al.,
Eds., Mediad, White Plains, NY, 1967, pp 33-36.
8. Giampietro, 0., Navalesi, R., Buzzigoli, G., et al., Decrease in
plasma glucose concentration during storage at -20 #{176}C. Clin. Chem.
26, 1710-1712 (1980).
9. Sheiko, M. C., Burkhardt, R. T., and Batsakis, J. G., Glucose
measurements: A 1977 CAP survey analysis. Am. J. Clin. Pathol. 72,
337-340 (1979).
10. Passey, R. B., Gillum, R. L., Fuller, J. B., et al., Evaluation and
comparison of 10 glucose methods and the reference method recom-
mended in the Proposed Product Class Standard (1974). Clin. Chem.
23, 131-139 (1977).
11. Eschwege, E., Warnet, J. M., Patois, E., et al., Comparaison des
glyc#{233}mies mesur#{233}es par Ia m#{233}thode automatique r#{233}ductim#{233}trique
et a Ia glucose oxydase chez 211 sujets a jeun et 2 heures apr#{233}s charge
orale glucos#{233}e. Pat ho!. Rio!. 20,3-4, 187-190 (1972).