systems for determining glucose in whole blood by flow analysis in a clinical ... for steady-state and flow-injection analysis. (17). A manual injection valve (Valve.
LIN.
CHEM. 36/9, 1657-1661
(1990)
linical Evaluation of Amperometric 3lucose in Undiluted Whole Blood Ian Gunaslngham,1
Enzyme
Electrodes
Keyphraees:
wall-jet electrode
.
glucose
carbon-paste oxidase
enzyme
electrode
Over the last decade, considerable research effort has bcused on the development of biosensors (1-4). The main notivating factors are the need for sensing devices that can nirpass the limits of speed, sensitivity, and stability in xisting devices (5-7). Perhaps the most successful type of biosensor is the nuperometric enzyme electrode, especially for monitoring lucose (8-11). Glucose sensors, which are widely used for linically monitoring diabetic patients (10, 12), must meet he need for accurate, rapid, and economical detection of glucose in undiluted whole blood. Such a device would have pplications in both online extracorporeal systems and linical analyzers. Advances in the field of amperometric enzyme probes nclude the development of a new generation of enzyme lectrodes, in which a synthetic mediator (Mod) is used as m electron shuttle between the enzyme-active center and Jie electrode surface, instead of the naturally occurring xygen (13). The enzyme system used is the glucose oxidase GOD; EC 1.1.3.4)-catalyzed reaction: :lluoose + GOD/FAD-+gluconolactone
+ GODIFADH2
(1)
+2
+2 Med#{176} +2W
(2)
3ODIFADH2
Analysis
for
ChIn-Huat Tan,1 and Tar-Choon Aw2
Ne have evaluated the carbon-paste enzyme/wall-jet elecrode in continuous-flow analysis for glucose in undiluted whole blood. Responses of the electrode to protein-based Ind aqueous samples under wall-jet configuration are decribed. Imprecision in the flow-injection mode was 99%) was from Fluka Chemie AG, Buchs, Switzerland. Polycarbonate membrane (0.03-Mm pore size) was from Nucleapore, Pleasanton, CA. All other reagents used were of analytical grade. The buffer solution, 0.1 mol/L sodium phosphate buffer (pH 7.4) containing 0.1 g of sodium azide per liter, was made up in Millipore “Milli Q”-grade water. All standards were made up in this buffer. Glucose solutions were allowed to equilibrate overnight before assay. Glucose
serum
oxidase
albumin
Apparatus Amperometric measurements were performed with a Model 174A polarographic analyzer (Princeton Applied Research, Princeton, NJ). Current-time curves were recorded with the x-y recorder (Graphtec Model WX2400; Graphtec Corp., Tokyo, Japan). The reference electrode was Ag/AgC1 in saturated KC1; the counter electrode was graphite. CLINICAL
CHEMISTRY,
Vol. 36, No. 9, 1990
1657
Flow System A large-volume (15 mL) wall-jet cell (Figure 1) was used for steady-state and flow-injection analysis (17). A manual injection valve (Valve V-7; Pharmacia Fine Chemicals, Uppsala, Sweden) was used to load the sample, with use of a peristaltic pump (Eyela Model MP-3; Tokyo Rikakikai, Tokyo, Japan) to deliver the sample plug and carrier streams to the detector. The pump was calibrated before use on the basis of the time taken for a fixed volume to be delivered. The sample volume for flow-injection measurements, 25 L, afforded the characteristic of peak proffles. For the steady-state measurements, a 200-L sampleilljection loop was used. A steady-state response was generally achieved within 40 s.
analyses
Procedures Assembly of enzyme electrode. Graphite rods (3 mm in diameter) were obtained from Johnson Matthey, Essex, U.K. A 2-mm-thick disc was cut from the rod and polished with abrasive paper. Electrical contact was made by connecting a wire to one side of the disc with silver-loaded epoxy resin. The disc was then inserted into a Teflon holder and sealed with epoxy resin, leaving a 1-mm-deep well. Before use, the electrode was cleaned successively with ethanol and nitric acid, rinsed ultrasonically in distilled water, and then oven dried. Stock carbon paste was prepared by mixing tetrathiafulvalene with graphite powder (20/80, by wt) that had been ground from a graphite rod and sieved with a 300-tracks/in. stainless-steel sieve screen (Haver, Darmstadt, F.R.G.). We then mixed the combination with small volumes of acetone to dissolve the tetrathiafulvalene. The acetone was evaporated by passing a stream of purified nitrogen into the mixture. We then added 200 mg of paraffin liquid (Nujol, 850 g/L at 20#{176}C; “SpectrosoL”; BDH Chemicals Pty., Poole, U.K.) to 1 g of the graphite-tetrathiafulvalene mixture and mixed thoroughly. Enzyme electrodes were prepared by packing the paste into the well of the electrode and then compressing and smoothing the paste with a flat spatula. We deposited 10 pL of 40 g/L GOD solution on top of the carbon-paste
Counter
Pefspek Housing
Waste 1 cm Fig. 1. Diagram of wall-jet/enzyme 1658
CLINICAL
surface and dried it at 4#{176}C. The GOD was immobilized by depositing 2.5 pL of a freshly prepared mixture of equal volumes of 50 g/L BSA solution and 25 mL’L glutaraldehyde solution on top of the dried GOD electrode, which was then covered with a polycarbonate membrane. The membrane was held in place with a Teflon cap. The enzyme electrode was allowed to sit at 4 #{176}C for 24 h. Blood glucose analysis. Patients’ blood samples, obtained in fluoridated tubes (Sarstedt, Rommelsdorf, F.R.G.), were divided into two portions and analyzed in parallel with both the Ektachem XR700 multichannel analyzer (Eastman Kodak, Rochester, NY) and the amperometric enzyme/wall-jet detector. The Ektachem XR700 measures glucose by a GOD/colorimetry system (18). It has a precision of 2.1% or 0.139 mmol/L for serum glucose (19). All the
CHEMISTRY,
Outlet
detector Vol. 36, No. 9, 1990
Results Optimizing
were
performed
at room
temperature.
and DIscussion the Enzyme/Wall-Jet
Detector
One of the difficulties of analyzing undiluted blood in continuous-flow systems is the difference in the matrix composition of the samples and the standard solutions, particularly the high viscosity of blood and the presence of large particles, such as blood cells and proteins. In use of the enzyme/wall-jet detector for the analysis of whole blood, the way the electrode inlet is oriented has a crucial bearing on the performance of the detector. When the jet inlet and working electrode are aligned in a horizontal plane, the sample jet tends to be diverted downwards, away from the electrode, owing to the effectof gravity. This is particularly significant at low flow rates, leading to the breaking up of the jet stream. To overcome this problem, we placed the enzyme electrode facing upwards, vertically aligned to the jet inlet so that the jet does not degenerate into turbulence before impinging on the electrode. To investigate the effect of sample viscosity, we prepared two sets of glucose standards, one containing BSA (250 g/L) and one not. The BSA had no effect on the electrode response in the steady-state mode; in the flow-injection mode, however, there was a marked difference in the peak current response at different flow rates (Figure 2). At low flow rates, the peak currents for the glucose standards with and without BSA are comparable. With increases in flow rate, the peak current for both standard solutions declines, with the current for standards containing BSA being substantially higher. When the flow rate exceeds 2.5 mL/min, the current responses for both standards approach a similar limiting value. The higher peak current for standards containing BSA at intermediate flow rates (0.6-2.4 mL/min) is an effect of viscosity: the higher the viscosity, the greater the contact time between glucose and the electrode. At lower flow rates, however, diffusion effects predominate over viscosity effects. The effect of electrode-inlet separation on the flow-injection peak currents at low flow rate (0.4 mL/min) is shown in Figure 3. The curves for both standard solutions with and without BSA cross at the electrode-inlet separation of 0.5 mm. The peak currents for standards containing BSA reach a limiting value independent of electrode-inlet separations between 2 and 8 mm. However, for standards not containing BSA, the peak current remains steady between 2 and 4 mm; thereafter, it declines. When the enzyme/wall-jet detector was operated at higher flow rates (>2.5 mL/min), response to the two sets of
Flow
Rul.
(wi/ow,)
Fig. 2. Effect of flow rate on flow-injection peak current glucose, 10 mmol/L, (#{149}) in phosphate buffer, (0) in phosphate buffer with BSA, 250 gIL
Tetrathiafulvalene carbon-paste (20/80 by wt) enzyme electrode, operating potential 200 mV, sample volume 20 ML. Inlet electrode separation = =
=
GlucoSe RSeov,rSd
Fig. 4. Precision
Glucose control 1
mm
=
profile mode;
potential
Analytical
of the mediated enzyme/wall-jet detector in duplicate 15 times over a period of six days (n flow rate = 0.83 mUmln, sample volume = 50 ML
assayed
was
30). Flow-injection
operating
(mmol)
=
0.2 V
Recovery
Studies
Table 1 summarizes the results mined with, the present detector. served with a mean of 100.7%. 2.0
Accuracy
1)6
1.6
#{163}I.cI,od.-Ir,I
Sepoollon
(o.n)
of Analysis
of Patients’
of recovery studies obGood recovery was ob-
Specimens
Results of glucose analysis of patients’ specimens with the enzyme/wall-jet detector were compared with those by the colorimetric method on a standard clinical laboratory analyzer, Ektachem XR700 multichannel analyzer (Table 2). The enzyme/wall-jet detector measures glucose in whole blood, whereas the colorimetric method measures serum glucose with use of GOD and the chromogen system (19). Linear-regression equations of y = 0.974x + 0.500 andy 1.152x + 0.295 were obtained for the steady-state and flow-injection modes, respectively. For the steady-state mode, the correlation coefficient obtained after the analysis of 143 patients’ samples is 0.986, with a positive bias of 0.315 (21). The t-value of 5.584 indicates that’ the positive bias is significant. For the flow-injection mode, analysis of 130 patients’ samples yields an r of 0.974, with a significant positive bias of 1.355. Generally, the mediated enzyme/ wall-jet detector gave a positive bias. In comparing the operating modes of mediated enzyme/ wall-jet detector, glucose values and the standard error were higher for the flow-injection mode. This could be because of matrix variations from sample to sample. How=
Fig. 3. Effect of inlet electrode separation on flow-injection peak current for glucose, 10 mmol/L, without (#{149}) or with (0) BSA, 250 g/L Flow rate 0.45 mL/min, tetrathiafulvalene carbon-paste enzyme electrode; other conditions same as in Fig. 2
glucose standards was significantly different. Initially, the peak current for both remained constant with increasing electrode-inlet separation up to 8 mm (data not shown), although the standards containing BSA produced higher peak currents. Thus, for aqueous standards to be used successfully in calibrating the enzyme-electrode detector in continuousflow analysis, the electrode-inlet separation should be held at 0.5 mm, facing upwards in a plane vertical to the jet inlet, with the flow rate adjusted to 0.5 mL/min.
Table
Recovery
Mediated Glucose,
Precision We determined the precision profile of the mediated enzyme/wall-jet detector by analyzing eight pools of glucose standards containing BSA (40 g/L) in phosphate buffer, by flow-injection mode (Figure 4). A total of 30 analyses for each glucose pool were performed over a period of six days (n = 30). The total CV was
