3.1 Initial study of the effects of biochemical and Hct% interferences
A total of eight endogenous and exogenous biochemicals were chosen for this study: glucose, ascorbic acid, uric acid, hemoglobin, bilirubin, cholesterol, acetaminophen and acetamidophenol.
Fig.1 shows selected cyclic voltamograms of some of these biochemicals at various concentrations using the GOD(+)-SPCE strips in the potential range from –400 to 400 mV after 20-s sample incubation. The current readings of the oxidative cyclic voltamograms at various concentrations of these biochemicals at 100, 200, 300 and 400 mV are listed in Table 1. It is noted that the endogenous biochemical concentrations in a normal, healthy person are generally as follows:
glucose, below 150 mg dL-1; ascorbic acid, below 5 mg dL-1; uric acid, below 10 mg dL-1; bilirubin, below 5 mg dL-1; cholesterol, below 200 mg dL-1; hemoglobin, below 10 g L-1 (Burtis and Ashwood, 1999). Thus, excluding glucose, the higher (than normal) working concentrations chosen for fresh hemoglobin (20 g L-1) and the rest of the biochemicals (300 and 20 mg dL-1, respectively) for this initial study represent quite stressed physiological conditions. From the current readings, it is observed that at 300 mV, most of the potential interfering biochemicals show relatively the greatest interfering signals. It was found that the degree of biochemical interferences for glucose determination is roughly in the order uric acid ~ ascorbic acid > hemoglobin >
bilirubin > cholesterol. For the exogenous biochemicals, acetaminophen has a higher degree of interference. However, the effect would be greatly reduced after this material has been excreted from the body.
For subsequent chronoamperometric measurements of various concentrations of glucose and uric acid in saline solutions, a potential of 300 mV was used and the tenth second current readings after 20-s sample incubation were collected, as shown in Fig. 2. Under these conditions, the current signals for glucose measurements in the 50–550 mg dL-1 concentration range in phosphate buffer (pH 5.0) and whole blood (Hct% of 40%) were linear with good correlation coefficients, as shown in Fig. 3.
One common complication for diabetic patients is anemia due to hemodialysis (Ma et al., 1999; Anatole et al., 1998). In this case, abnormally low Hct% values were frequently observed for whole blood samples. Thus, it is important to evaluate the determination of blood glucose at
24
low Hct% using SPCE strips. The Hct% interference effects on the current measurements using GOD(+)-SPCE strips at several glucose concentrations are shown in Fig 4. As expected, higher current readings were observed for glucose concentrations with lower Hct% values. An exponential decay relationship, i.e., I=ke-b[Hct%], could be established between the current readings and the Hct% values with correlation coefficients R2≥0.9088. For a given glucose concentration, the current responses at various Hct% values and two uric acid concentrations are shown in Fig 5 with the use of GOD(-)-SPCE strips. A similar result to that of the above-discussed GOD(+)-SPCE strips was observed and corroborated well with those reported earlier (Carter et al., 1997 ). The normal adult hematocrit ranges are 39–51% for men and 36–44% for women (Minetti, 1997). Outside this range, the glucose determinations using this technique would give larger errors, particularly for those with lower Hct%.
3.2 Elimination of interferences of uric acid/biochemicals and Hct%
Three kinds of current signals are defined in this study: (1) I(s+i) is the current reading obtained with the GOD(+)-SPCE strips and includes that of glucose with biochemical and hematocrit interferences, (2) Ii is obtained with GOD(-)-SPCE strips and is due to biochemicals with minor hematocrit interferences and (3) the calculated corrected current, ∆I=I(s+i)-Ii. Note that the hematocrit interference might not be completely eliminated by just calculating ∆I because the degree of hematocrit interferences in glucose determination is usually greater than that of other biochemicals (vide infra). To minimize the interferences of Hct% completely, additional interpolation was exercised (vide infra).
The tenth second I(s+i) and Ii values were obtained after an initial 20-s sample incubation period.
Fig 6 shows the current (i.e., I(s+i) and Ii) plots against spiked uric acid concentrations in two sets of RRBC-saline glucose solutions, i.e., 93 mg dL-1 (RBC/solution volume ratio: 46%) and 316 mg dL-1 (RBC/solution volume ratio: 42%). The solutions were spiked with various amounts of uric acid. It is observed that the slopes of both I(s+i) and Ii plots against uric acid concentrations are similar and the slopes of both ∆I plots against uric acid concentrations are close to zero. This implies that after uric acid concentration corrections, the glucose oxidation currents are independent of uric acid concentration, if Hct% values are not very different.
The ∆I values are also plotted in Fig 7. In an attempt to eliminate Hct% interferences, a series of glucose solutions were prepared with various concentrations and Hct% values. The I(s+i) and Ii
readings were recorded in the same way as those of Figure 6, and corresponding ∆I values were calculated. Fig 7 shows plots of ∆I values against Hct% values at four blood glucose
25
concentrations. From these plots, it is observed that for a glucose concentration of 75.7 mg dL-1, the current readings are similar in the Hct% range tested. However, for glucose concentrations of 170.4, 318.0 and 462.0 mg dL-1, the current readings are increasingly higher at lower Hct% values, as expected (vide supra). Additional efforts to eliminate Hct% interferences are described later.
3.3 Practical applications
Previously a correlation for the determination of whole blood glucose between the more convenient YSI method and that of the more accurate photometric hexokinase method (Schmidt, 1961) was established in our laboratory using 96 plasma samples. The correlation equation was [glucose]YSI = 1.064[glucose]hexokinase-12.63, with R2 = 0.9781. For convenience, the YSI method was used for the practical applications discussed in this section.
The above-discussed methods to eliminate biochemical and Hct% interferences were applied to 37 preclinical human samples (16 from men and 21 from women). The experimentally obtained I(s+i) and Ii values as well as calculated ∆I values are listed in Table 2. The data in column 2 (i.e., [glucose]YSI), column 3 (i.e., [UA]optic) and column 4 (i.e., [Hct]KUBOTA) of Table 2 are the glucose and uric acid concentrations and the Hct% values determined directly using the methods of the YSI 2300 STAT Plus, the Roche UA plus kit/Meterteck SP870 spectrophotometer and the KUBOTA 3110 centrifuge, respectively. The uncorrected glucose concentrations in column 6 (i.e., [glucose]uncorr) were obtained using the I(s+i) values (column 5) and a calibration curve was established between the data in columns 2 and 5 (Collison, et al., 1999; Tieszen and New, 2003;
Kovatchev, et al., 2005), i.e., I(s+i) = 0.0613[glucose]uncorr + 0.0168. For each sample, a percentage bias value is calculated using the corresponding values in columns 2 and 6 and the sum of the absolute percentage bias values for all samples is 204.2. Similarly, the uric acid corrected glucose concentrations in column 9 (i.e., [glucose]uacorr) were obtained using the ∆I values (column 8) and a calibration curve was established between the data in columns 2 and 8, i.e., ∆I = 0.0601[glucose]uacorr - 0.4421. The sum of the absolute percentage bias values for [glucose]uacorr is reduced to 181.7, indicating some improvement in accuracy. If the 37 samples are divided into a high-glucose group (the first 14 samples; more than 120 mg dL-1 glucose) and a normal-glucose group (the last 23 samples; less than 120 mg dL-1 glucose), it is found that the percentage bias is largely contributed by the samples from the high-glucose group, i.e., samples from the higher-risk patients.
Compared with Hct% interferences, uric acid/biochemical interferences and others such as sample per strip background interferences are relatively lower and the use of ∆I values would
26
eliminate most of these minor interferences. However, glucose measurements using SPCE test strips are also interfered with by Hct%, particularly for samples with high glucose concentration and low Hct% (Fig 4 and 7). For example, although both samples 27 and 36 have abnormally low uric acid concentrations (i.e., outside the 3.5– 8.5 mg dL-1 range) and Hct% values (i.e., outside the 36–51% range), the percentage bias values are relatively low because they were in the low glucose concentration group. On the other hand, samples 6 (Hct% 33%) and 11 (Hct% 31%) with normal uric acid concentrations but abnormally low Hct% values tend to have much greater percentage bias values.
To further improve the accuracy of the glucose measurements, we propose the following method to reduce the Hct% interferences, using sample 11 as one example. The initially determined glucose concentration of sample 11 (i.e., 244.8 mg dL-1) fell between 170.4 and 318.0 mg dL-1; therefore, the exponential decay equations represented by curves b and c in Fig. 4 are applicable to calculate the two current values, i.e., 11.5 µA (using the equation represented by curve b) and 22.6 µA (using equation represented by c) at Hct% = 31%. Using the determined I(s+i)
value of 15.0 µA in Table 2 (column 5, sample 11) and by interpolation, we calculated [glucose]Hct-corr of sample 11 to be 217.71 mg dL-1 for the I(s+i) system. The percentage bias is reduced to 8.8 and is lower than the corresponding value, 22.4 (Table 2, column 7, sample 11), without any corrections. If the ∆I value of 13.7 µA for sample 11 and the equations represented by curves b and c in Fig 7 are used, the value of [glucose]Hct-corr. in the ∆I system is 217.69 mg dL-1. The percentage bias is also reduced to 8.8 and is again lower than the corresponding value, 17.9 (Table 2, column 10, sample 11), with only uric acid/biochemical correction, as Table 3 for sample calculations. Similarly, the respective percentage bias values of sample 6 after Hct% corrections are 3.1 and 4.9, and are lower than the respective values, 13.4 (Table 2, column 7, sample 6) and 12.5 (Table 2, column 10, sample 6). The differences of the percentage bias values of Hct%
corrected values and that of Hct% uncorrected values for other samples with normal Hct% are not as large as for samples 6 and 11. Thus, hematocrit corrections are especially important for samples with abnormal hematocrit values.