- Email: [email protected]
Direct reading glucose electrodes detect the molality of glucose in plasma and whole blood
Clinica Chimica Acta, 189 (1990) 33-38 Elsevier
33
CCA 04121
Direct reading glucose electrodes detect the molality of glucose in plasma and whole blood Niels Fogh-Andersen, Peter D. Wimberley, Jlargen Thode and Ole Siggaard-Andersen Department (Received
of Clinical Chemistry,
Herlev Hospital, Herlev (Denmark)
26 July 1989; revision received 1 February
1990; accepted
16 February
Key wordr: Activity; Biosensors; Enzyme electrode; Glucose oxidase; Molality vs. concentration; Water concentration
1990)
Hematocrit;
Summary
It is the activity that determines the direction of chemical processes, transport, etc. and thus provides the clinically more relevant information. Direct reading glucose electrodes consume glucose at a rate proportional to the glucose activity in the sample. The activity equals the molality (mm01 glucose per kg water), so results from direct reading glucose electrodes must differ from the conventionally measured glucose concentration. This was observed in 159 whole blood samples which gave higher results from a direct reading glucose electrode than by our conventional method (y = 1.21x - 0.37 mmol/l). However, adjustment for the different water concentration due to salt, plasma proteins, and hemoglobin occupying space, gave results equal to the concentrations (y = 1.00x - 0.28 mmol/l, r = 0.997). Furthermore, results for samples with constant glucose concentration and varying albumin concentration correlated with the albumin concentration (r = 0.989), but not after adjustment for water concentration (r = 0.037, n.s.).
Introduction
Direct reading glucose electrodes are suited for bedside analysis and may forerunners of implantable devices for insulin delivery in diabetic patients. available incorporate glucose oxidase and an amperometric electrode sensing gen peroxide. Since the enzymatic reaction depends on the glucose activity
Correspondence to: Dr. N. Fogh-Andersen, Herlev, Denmark.
0009-8981/90/$03.50
Department
0 1990 Elsevier Science Publishers
of Clinical
Chemistry,
B.V. (Biomedical
Herlev Hospital,
Division)
be the Those hydrorather
DK-2730
34
than the concentration, and the glucose electrode consumes what it is supposed measure [l], we have studied what the electrodes actually measure.
to
Material and methods We used a Stat Profile 5 glucose analyser (NOVA, Waltham, MA). The apparatus combines sensors for pH, O,, CO,, hematocrit, K+, Na+, Cl-, Ca’+, and glucose. The glucose electrode incorporates immobilized glucose oxidase, catalyzing the reaction glucose + 0, + H,O + gluconate + H++ H,O,. Hydrogen peroxide is oxidized at the platinum anode, generating a current proportional to the glucose activity in the sample (H,O, + 2H++ 0, + 2e-). For routine glucose measurements we take 20 ~1 blood from the earlobe directly into a capillary pipette with sodium heparinate and immediately transfer it to 300 ~1 of diluent (Triton X 100, 0.2%, antifoam emulsion, O.l%, NaCl, 1.00 mol/l, EDTA, 2.8 mmol/l, and phosphate, 50 mmol/l, pH 7.6) which lyses the red cells. An RA 1000 (Technicon, Terrytown, NY) measures the glucose kinetically with glucose dehydrogenase (Merck, Darmstadt, FRG) at 340 nm. We used 159 whole blood samples for comparisons between the glucose electrode and the routine method. Of these, 59 were from patients, and 100 were from a control individual. Some of the samples were equilibrated to a higher or lower poZ, some had a changed hematocrit by adding or removing plasma, and some deliberately had a lower glucose concentration during storage or a higher glucose concentration by adding glucose. All samples were prepared for routine analysis at the same time as the aspiration into the glucose electrode. Linearity was determined with aqueous standards, and between-day precision was determined with aqueous and serum based control solutions. The effect of altered molality was examined by varying the albumin concentration in solutions with constant glucose concentration (constant volume and amount of glucose). 11 solutions with different albumin concentrations were prepared by diluting a stock albumin solution (200 g/l). To each 10 ml of solution we added 100 ~1 of a glucose stock solution, 1 mol/l. The mass concentration of water in each solution was determined by the weight loss of an overnight drying at 105°C. The unmodified glucose-electrode results were adjusted to yield the substance concentrations (mmol/l) taking the different mass concentrations of water into account. First, the molality was calculated by division with the mass concentration of water in the calibrator (0.99 kg H,O/l). Secondly, the molality was changed to concentration by multiplication with the mass concentration of water in the sample. The mass concentration of water in each whole blood sample was calculated from the hematocrit assuming a mass concentration of water in red cells of 0.71 kg H,O/l [2], and a mass concentration of water in plasma of 0.93 kg H,O/l. Results The glucose results from the glucose electrode and the RA 1000 substance concentrations are shown in Fig. 1. The slope indicated that the glucose electrode
35
Fig. 1. Whole blood glucose from the glucose electrode versus whole blood glucose from RA 1000. The unmodified Stat Profile 5 glucose results in 159 whole blood samples were compared to the concentrations measured with RA 1000. Regression: y =1.21x -0.37 mmol/l, c = 0.997, SD (slope) = 0.01, SD (intercept) = 0.08 mmol/l, SD about the regression line = 0.61 mmol/l.
responded to molality of glucose (mm01 glucose/kg water) rather than concentration. We converted the results to concentrations by taking into account the different mass concentrations of water. These adjusted glucose-electrode results are plotted against RA 1000 in Fig. 2. /
I
I
I
I
I
I
I
I
40Water 3$
corrected
B-glucose.
NOVA
mmolll
3c-
25-
20-
15-
10-
5RA 1000 B-glucose.
mmolll
OI 0
I 5
, 10
, 15
I 20
I 25
30
35
40
Fig. 2. Adjusted whole blood glucose from the glucose electrode versus whole blood glucose from RA 1000. Samples were the same as in Fig. 1. Each Stat Profile 5 result was adjusted for the water concentrations of sample and calibrator. The equation was glucose concentration = electrode reading. (Hct.0.71 +(l -Hct).0.93)/0.99. Regression: y =1.00x -0.28 mmol/l. r = 0.997, SD (slope) = 0.01. SD (intercept) = 0.07 mmol/l, SD about the regression line = 0.54 mmol/l.
36
20-
NOVA glucose (mmol/l)
16-
t
16-
t
14-.A
4
12A
lo-
r =
0.969
a642Albumin. g/l
O0
I 20
I 40
I 60
I 60
I 100
1 120
I 140
I 160
I 180
I 200
Fig. 3. Glucose electrode reading versus albumin concentration. All samples had the same glucose concentration (constant volume and amount of glucose) and the indicated concentration of albumin. Albumin decreases the mass concentration of water and increases the molality of glucose. The response of the glucose electrode depends on the albumin concentration (r = 0.989, p < 0.001).
The dependence on albumin is shown in Fig. 3. With constant glucose concentration (mmol/l) higher albumin caused higher glucose molality resulting in increasing glucose-electrode results (r = 0.989), but not after the decreasing mass concentration of water had been taken into account (r = 0.037, n.s., Fig. 4).
NOVA glucose, water-corrected (mmol/l)
,6_ 161412lo-
*
.
.
. r = 0037
6642Albumm, g/l
o-
I 0
I, 20
40
I 60
I 80
I,
100 120
I
140
I
160
I
160
I
200
Fig. 4. Adjusted glucose electrode readings versus albumin concentration. Samples were the same as in Fig. 3. Each Stat Profile 5 result was adjusted for the measured mass concentration of water. The adjusted results do not depend on the albumin concentration (r = 0.037, ns.).
37
CX.mecteciNOVA minus RA loo0
0
B-glucose, mmolll
Fig. 5. Adjusted electrode minus RA 1000 values as function of hematocrit I%.Samples were the same as in Fig. 1. The overall difference between the two methods is - 0.22 mmol/l with an SD of 0.55 mmol/l. The difference depends on hematocrit. Regression: y = -0.030x + 1.16 mmol/l, r = - 0.431, p < 0.001.
Erythrocytes affected the difference between electrode and RA 1000 values. The ratio between unadjusted electrode and RA 1000 values correlated negatively to hematocrit (r = -0.298, p -c O.OOl), and adjusted electrode values minus RA 1000 values likewise depended on hematocrit (r = - 0.431, p < 0.001, Fig. 5). The glucose electrode had a linear response to 35 mmol/l, and samples with very low glucose concentration gave results consistent with those of RA 1000. The partial pressure of oxygen had no effect. The between-day CV was 2.0% from measurements on a control sernm with glucose concentration 11.6 mmol/l (n = 20). Discussion
Analyte activity is increasingly accepted as the quantity to which cells and organisms respond, since it determines the direction of the chemical processes, transport, binding of hormones to receptors, etc. Measuring and reporting the glucose molality (which equals the molal activity times the unit, mmol/kg) should provide the more clinically relevant information, with the further advantage of same values for plasma and whole blood. The activity of glucose in plasma equals that in red cells. This is so because glucose passes through the red cell membrane by facilitated diffusion requiring no energy or insulin. In human red cells the flux is about 250 times the rate of glucose utilization, making the plasma and red cell activities of glucose identical [3]. Since the typical mass concentration of water in plasma is 0.93 kg H,O/l and that in red cells is 0.71 kg H,0/1[2], whole blood with a normal hematocrit of 43% has a mass concentration of water of 0.84 kg H,O/l. While the molality of glucose (mmol
38
glucose/kg water) is equal in plasma and red cells, the concentration (mmol glucose/l) is higher in plasma than in red cells, and consequently higher in plasma than in whole blood, by a factor of 1.11 ( = 0.93/0.84) depending on the hematocrit. The difficulty with different mass concentrations of water in plasma and whole blood would be overcome by expressing the glucose results as molality instead of concentration. The direct reading glucose electrodes respond to molality of glucose in the sample, but the results are reported as molarity by the manufacturers, relative to the glucose concentration in the calibrator. We have only examined one direct reading glucose electrode, but we believe the results are generally applicable. The present manufacturer had completely ignored the different water concentrations in calibrator, plasma and red cells, which indicates the need of a consensus on how to report the results. It would seem more appropriate to report molality with the correct unit (mmol/kg), but also to provide the optional software enabling adjustment to either plasma or whole blood glucose concentration (mmol/l). The conversion factors should probably be constant, 0.93 kg/l for plasma and 0.84 kg/l for whole blood. Most users would choose this conversion, since the unit for molality is unfamiliar to many physicians, and different results and reference intervals with interchanging methods are highly impractical. Another solution would be to allow a user defined slope-correction factor, as previously suggested [4]. The negative effect of red cells was surprising. Since we had no time for further studies, its cause can only be speculated. It is possible that the increased blood viscosity or the red cell membranes affect the glucose flux into the electrode negatively. However, albumin increases the viscosity without affecting the measured molality of glucose. In conclusion, the direct reading glucose electrodes function well. There are difficulties due to the water concentrations which necessitate a consensus on how to report the results. In this process the clinical chemists and organizations like the IFCC must play a major role (51. References 1 Thevenot DR. Problems in adapting a glucose-oxidase electro-chemical sensor into an implantable glucose-sensing device. Diabetes Care 1982;5:184-189. 2 Kessler E, Levy MR, Allen RL. Red cell electrolytes in patients with edema. J Lab Clin Med 1961;57:32-41. 3 Grimes AJ. Human red cell metabolism. Oxford: Blackwell Scientific Publications, 1980;87-88. 4 Fogh-Andersen N, Wimberley PD, Thode J, Siggaard-Andersen 0. A brief evaluation of NOVA’s Stat Profile 1. In: Maas AHJ, Buckley B, Manzoni A, Moran R, Siggaard-Andkrsen 0, Sprokholt R, eds. Methodology and clinical applications of ion-selective electrodes. IFCC workshop Stresa 1988. Utrecht: Elinkwijk, 1989;269-276. 5 Hunter KW. Biosensors. A new technology for real-time, on-line biochemical monitoring. Am J Clin Path01 1989;91 (Suppl l):S32, S33.
33
CCA 04121
Direct reading glucose electrodes detect the molality of glucose in plasma and whole blood Niels Fogh-Andersen, Peter D. Wimberley, Jlargen Thode and Ole Siggaard-Andersen Department (Received
of Clinical Chemistry,
Herlev Hospital, Herlev (Denmark)
26 July 1989; revision received 1 February
1990; accepted
16 February
Key wordr: Activity; Biosensors; Enzyme electrode; Glucose oxidase; Molality vs. concentration; Water concentration
1990)
Hematocrit;
Summary
It is the activity that determines the direction of chemical processes, transport, etc. and thus provides the clinically more relevant information. Direct reading glucose electrodes consume glucose at a rate proportional to the glucose activity in the sample. The activity equals the molality (mm01 glucose per kg water), so results from direct reading glucose electrodes must differ from the conventionally measured glucose concentration. This was observed in 159 whole blood samples which gave higher results from a direct reading glucose electrode than by our conventional method (y = 1.21x - 0.37 mmol/l). However, adjustment for the different water concentration due to salt, plasma proteins, and hemoglobin occupying space, gave results equal to the concentrations (y = 1.00x - 0.28 mmol/l, r = 0.997). Furthermore, results for samples with constant glucose concentration and varying albumin concentration correlated with the albumin concentration (r = 0.989), but not after adjustment for water concentration (r = 0.037, n.s.).
Introduction
Direct reading glucose electrodes are suited for bedside analysis and may forerunners of implantable devices for insulin delivery in diabetic patients. available incorporate glucose oxidase and an amperometric electrode sensing gen peroxide. Since the enzymatic reaction depends on the glucose activity
Correspondence to: Dr. N. Fogh-Andersen, Herlev, Denmark.
0009-8981/90/$03.50
Department
0 1990 Elsevier Science Publishers
of Clinical
Chemistry,
B.V. (Biomedical
Herlev Hospital,
Division)
be the Those hydrorather
DK-2730
34
than the concentration, and the glucose electrode consumes what it is supposed measure [l], we have studied what the electrodes actually measure.
to
Material and methods We used a Stat Profile 5 glucose analyser (NOVA, Waltham, MA). The apparatus combines sensors for pH, O,, CO,, hematocrit, K+, Na+, Cl-, Ca’+, and glucose. The glucose electrode incorporates immobilized glucose oxidase, catalyzing the reaction glucose + 0, + H,O + gluconate + H++ H,O,. Hydrogen peroxide is oxidized at the platinum anode, generating a current proportional to the glucose activity in the sample (H,O, + 2H++ 0, + 2e-). For routine glucose measurements we take 20 ~1 blood from the earlobe directly into a capillary pipette with sodium heparinate and immediately transfer it to 300 ~1 of diluent (Triton X 100, 0.2%, antifoam emulsion, O.l%, NaCl, 1.00 mol/l, EDTA, 2.8 mmol/l, and phosphate, 50 mmol/l, pH 7.6) which lyses the red cells. An RA 1000 (Technicon, Terrytown, NY) measures the glucose kinetically with glucose dehydrogenase (Merck, Darmstadt, FRG) at 340 nm. We used 159 whole blood samples for comparisons between the glucose electrode and the routine method. Of these, 59 were from patients, and 100 were from a control individual. Some of the samples were equilibrated to a higher or lower poZ, some had a changed hematocrit by adding or removing plasma, and some deliberately had a lower glucose concentration during storage or a higher glucose concentration by adding glucose. All samples were prepared for routine analysis at the same time as the aspiration into the glucose electrode. Linearity was determined with aqueous standards, and between-day precision was determined with aqueous and serum based control solutions. The effect of altered molality was examined by varying the albumin concentration in solutions with constant glucose concentration (constant volume and amount of glucose). 11 solutions with different albumin concentrations were prepared by diluting a stock albumin solution (200 g/l). To each 10 ml of solution we added 100 ~1 of a glucose stock solution, 1 mol/l. The mass concentration of water in each solution was determined by the weight loss of an overnight drying at 105°C. The unmodified glucose-electrode results were adjusted to yield the substance concentrations (mmol/l) taking the different mass concentrations of water into account. First, the molality was calculated by division with the mass concentration of water in the calibrator (0.99 kg H,O/l). Secondly, the molality was changed to concentration by multiplication with the mass concentration of water in the sample. The mass concentration of water in each whole blood sample was calculated from the hematocrit assuming a mass concentration of water in red cells of 0.71 kg H,O/l [2], and a mass concentration of water in plasma of 0.93 kg H,O/l. Results The glucose results from the glucose electrode and the RA 1000 substance concentrations are shown in Fig. 1. The slope indicated that the glucose electrode
35
Fig. 1. Whole blood glucose from the glucose electrode versus whole blood glucose from RA 1000. The unmodified Stat Profile 5 glucose results in 159 whole blood samples were compared to the concentrations measured with RA 1000. Regression: y =1.21x -0.37 mmol/l, c = 0.997, SD (slope) = 0.01, SD (intercept) = 0.08 mmol/l, SD about the regression line = 0.61 mmol/l.
responded to molality of glucose (mm01 glucose/kg water) rather than concentration. We converted the results to concentrations by taking into account the different mass concentrations of water. These adjusted glucose-electrode results are plotted against RA 1000 in Fig. 2. /
I
I
I
I
I
I
I
I
40Water 3$
corrected
B-glucose.
NOVA
mmolll
3c-
25-
20-
15-
10-
5RA 1000 B-glucose.
mmolll
OI 0
I 5
, 10
, 15
I 20
I 25
30
35
40
Fig. 2. Adjusted whole blood glucose from the glucose electrode versus whole blood glucose from RA 1000. Samples were the same as in Fig. 1. Each Stat Profile 5 result was adjusted for the water concentrations of sample and calibrator. The equation was glucose concentration = electrode reading. (Hct.0.71 +(l -Hct).0.93)/0.99. Regression: y =1.00x -0.28 mmol/l. r = 0.997, SD (slope) = 0.01. SD (intercept) = 0.07 mmol/l, SD about the regression line = 0.54 mmol/l.
36
20-
NOVA glucose (mmol/l)
16-
t
16-
t
14-.A
4
12A
lo-
r =
0.969
a642Albumin. g/l
O0
I 20
I 40
I 60
I 60
I 100
1 120
I 140
I 160
I 180
I 200
Fig. 3. Glucose electrode reading versus albumin concentration. All samples had the same glucose concentration (constant volume and amount of glucose) and the indicated concentration of albumin. Albumin decreases the mass concentration of water and increases the molality of glucose. The response of the glucose electrode depends on the albumin concentration (r = 0.989, p < 0.001).
The dependence on albumin is shown in Fig. 3. With constant glucose concentration (mmol/l) higher albumin caused higher glucose molality resulting in increasing glucose-electrode results (r = 0.989), but not after the decreasing mass concentration of water had been taken into account (r = 0.037, n.s., Fig. 4).
NOVA glucose, water-corrected (mmol/l)
,6_ 161412lo-
*
.
.
. r = 0037
6642Albumm, g/l
o-
I 0
I, 20
40
I 60
I 80
I,
100 120
I
140
I
160
I
160
I
200
Fig. 4. Adjusted glucose electrode readings versus albumin concentration. Samples were the same as in Fig. 3. Each Stat Profile 5 result was adjusted for the measured mass concentration of water. The adjusted results do not depend on the albumin concentration (r = 0.037, ns.).
37
CX.mecteciNOVA minus RA loo0
0
B-glucose, mmolll
Fig. 5. Adjusted electrode minus RA 1000 values as function of hematocrit I%.Samples were the same as in Fig. 1. The overall difference between the two methods is - 0.22 mmol/l with an SD of 0.55 mmol/l. The difference depends on hematocrit. Regression: y = -0.030x + 1.16 mmol/l, r = - 0.431, p < 0.001.
Erythrocytes affected the difference between electrode and RA 1000 values. The ratio between unadjusted electrode and RA 1000 values correlated negatively to hematocrit (r = -0.298, p -c O.OOl), and adjusted electrode values minus RA 1000 values likewise depended on hematocrit (r = - 0.431, p < 0.001, Fig. 5). The glucose electrode had a linear response to 35 mmol/l, and samples with very low glucose concentration gave results consistent with those of RA 1000. The partial pressure of oxygen had no effect. The between-day CV was 2.0% from measurements on a control sernm with glucose concentration 11.6 mmol/l (n = 20). Discussion
Analyte activity is increasingly accepted as the quantity to which cells and organisms respond, since it determines the direction of the chemical processes, transport, binding of hormones to receptors, etc. Measuring and reporting the glucose molality (which equals the molal activity times the unit, mmol/kg) should provide the more clinically relevant information, with the further advantage of same values for plasma and whole blood. The activity of glucose in plasma equals that in red cells. This is so because glucose passes through the red cell membrane by facilitated diffusion requiring no energy or insulin. In human red cells the flux is about 250 times the rate of glucose utilization, making the plasma and red cell activities of glucose identical [3]. Since the typical mass concentration of water in plasma is 0.93 kg H,O/l and that in red cells is 0.71 kg H,0/1[2], whole blood with a normal hematocrit of 43% has a mass concentration of water of 0.84 kg H,O/l. While the molality of glucose (mmol
38
glucose/kg water) is equal in plasma and red cells, the concentration (mmol glucose/l) is higher in plasma than in red cells, and consequently higher in plasma than in whole blood, by a factor of 1.11 ( = 0.93/0.84) depending on the hematocrit. The difficulty with different mass concentrations of water in plasma and whole blood would be overcome by expressing the glucose results as molality instead of concentration. The direct reading glucose electrodes respond to molality of glucose in the sample, but the results are reported as molarity by the manufacturers, relative to the glucose concentration in the calibrator. We have only examined one direct reading glucose electrode, but we believe the results are generally applicable. The present manufacturer had completely ignored the different water concentrations in calibrator, plasma and red cells, which indicates the need of a consensus on how to report the results. It would seem more appropriate to report molality with the correct unit (mmol/kg), but also to provide the optional software enabling adjustment to either plasma or whole blood glucose concentration (mmol/l). The conversion factors should probably be constant, 0.93 kg/l for plasma and 0.84 kg/l for whole blood. Most users would choose this conversion, since the unit for molality is unfamiliar to many physicians, and different results and reference intervals with interchanging methods are highly impractical. Another solution would be to allow a user defined slope-correction factor, as previously suggested [4]. The negative effect of red cells was surprising. Since we had no time for further studies, its cause can only be speculated. It is possible that the increased blood viscosity or the red cell membranes affect the glucose flux into the electrode negatively. However, albumin increases the viscosity without affecting the measured molality of glucose. In conclusion, the direct reading glucose electrodes function well. There are difficulties due to the water concentrations which necessitate a consensus on how to report the results. In this process the clinical chemists and organizations like the IFCC must play a major role (51. References 1 Thevenot DR. Problems in adapting a glucose-oxidase electro-chemical sensor into an implantable glucose-sensing device. Diabetes Care 1982;5:184-189. 2 Kessler E, Levy MR, Allen RL. Red cell electrolytes in patients with edema. J Lab Clin Med 1961;57:32-41. 3 Grimes AJ. Human red cell metabolism. Oxford: Blackwell Scientific Publications, 1980;87-88. 4 Fogh-Andersen N, Wimberley PD, Thode J, Siggaard-Andersen 0. A brief evaluation of NOVA’s Stat Profile 1. In: Maas AHJ, Buckley B, Manzoni A, Moran R, Siggaard-Andkrsen 0, Sprokholt R, eds. Methodology and clinical applications of ion-selective electrodes. IFCC workshop Stresa 1988. Utrecht: Elinkwijk, 1989;269-276. 5 Hunter KW. Biosensors. A new technology for real-time, on-line biochemical monitoring. Am J Clin Path01 1989;91 (Suppl l):S32, S33.