Glucose Determination by Automatic

Glucose Determination by Automatic Chemical Analysis Harold J. Grady and Martha A. Lamar 1HE GENERAL CONSTRUCTION and operation of an automatic chemical analyzer essentially identical with the one used in this study* has been described (1). Glucose is conveniently measured with this analyzer using the principle of the Hoffman method (2, 3). When using this method for the determination of blood glucose by automatic analysis, a change with time of the transmittance reading of a given standard is noted. This change is of sufficient magnitude that 3 or more standard curves are required over a period of 2 hours to avoid significant error in determining unknown concentrations. The variation in standard curves is caused, for the most part, by variations in dialyzer temperature. The use of a water bath for temperature control of the dialyzer greatly reduces this change. A water bath similar to the one constructed at this institution but with better temperature regulation is now available as standard equipment on the Technicon AutoAnalyzer and has been used for some of the experiments reported in this paper. Additional data are presented concerning the accuracy and precision of the automatic analyzer compared to a standard manual method for the determination of true blood glucose. EXPERIMENTAL The Hoffman method for glucose measurement uses the quantitative reduction of alkaline ferricyanide (yellow) to ferrocyanide (colorless) by glucose. The photometer measures the disappearance Prom the Department of Medicine, University of Kansas Medical Center, Kansas City, Ka Received for publication Nov. 17, 1958. Autoanalyzer, Model 1, Technicon Instruments Corp., Chauneey, N. Y. 542

Vol. 5, No. 6, 1959 GLUCOSE DETERMINATION 543 of the yellow color and an inverse concentration-transmittance curve is obtained which is linear on a semilogarithmic plot to about 500 mg. per 100 ml. if the reaction is carried out in the absence of cyanide. However, cyanide is added routinely, which increases the sensitivity and results in good precision for determination of concentrations of glucose in the normal and low normal range, although this causes some nonlinearity and lowered range. When dialyzer temperatures of 38#{176} are employed the increase in the quantity of glucose dialyzed allows the use of a smaller blood sample. For most of these studies the composition of the reagents and their rates of flow through the instrument as indicated by the manufacturer are listed in Table 1. Except where indicated, 40 samples were analyzed per hour which required a total of 0.80 ml. of blood or standard solution of glucose per sample. The usual sample consisted of oxalated whole blood to which was added about 1 mg. per ml. of sodium fluoride. Standards were prepared from reagent grade glucose made up in a saturated solution of benzoic acid to concentrations of 50, 100, 150, 200, and 300 mg. per 100 ml. For the experiments on the temperature effect on the rate of glucose dialysis a modification of the above conditions was used. For a standard curve the above standards were diluted fivefold, sampled at a rate of 0.8 ml. per minute, segmented by air pumped at a rate of 0.8 ml. per minute, and diluted with water pumped at 1.6 ml. per minute. This mixture was injected directly into the ferricyanide line which was pumped at 2.5 ml. per minute, and to this ferricyanide and sample mixture was added the potassium cyanide reagent pumped at 2.0 ml. per minute. This final solution passed through the heating bath and then through the flow cell in the usual fashion. Mixers were inserted in the line after addition of the sample to ferricyanide and after addition of the cyanide to the sample-f erricyanide mixture. To determine quantity dialyzed, the sample, air, and water diluent Table 1. RATES or FLOW FOB HOFFMAN METHOD FOR GLUCOSE BY AUTOMATIC ANALYSIS Rai8a of flow Sample or reagent,nz./min. Blood 0.8 Air in sample stream 0.8 0.9% NaCI saturated with capyrlic alcohol 2.5 KON 0.5% solution in 0.9% NaCI 2.0 K3Fe(CN)6 0.5% solution in 2% Na,CO3 2.5 Air in reagent stream 1.2

544 GRADY & LAMAR Clinical Chemistry were pumped at the same rates as described in the preceding paragraph and were made to flow through the sample side of the dialyzer which was thermostated by immersion in the constant temperature bath that had been modified for operation at various temperatures. Through the opposite side of the dialyzer, water at a rate of 2.4 ml. per minute (exactly equivalent to the combined volume of the sample and diluent on the sample side) and air at a rate of 0.8 ml. per minute were introduced to receive the dialyzed glucose. The latter stream (dialysate) flowed into the ferricyanide, which was then treated as for the standard curve obtained without the dlialyzer, as described above. This allowed a measurement of the actual concentration of glucose in the dialysate, and by comparison with the concentration on the sample side of the dialyzer the percentage dialyzed was determined. All units of the automatic analyzer were standard except the bath for the regulation of the dialyzer temperature. These standard units include a flow cell with a 6-mm. light path and matched interference filters, one each for the reference and sample side, which have halfband widths of 10 m with a 35% transmittance at the peak of 420 m. No aperture on the sample side and a number 4 aperture on the reference side allowed balance of the instrument at 100% transmittance on the recorder with a reading between 500-800 on the potentiometer with water in the flow cell. With the instrument so adjusted, the blank reading (all reagents flowing with water as the sample) was about 20% transmittance. The manual method for glucose was that of Nelson and Somogyi. The reagents were prepared as described in Reiner (4), and the method was modified as follows. A 1 to 10 filtrate was prepared by mixing 1 volume of blood, 5 volumes of water, and 2 volumes each of barium hydroxide and zinc sulfate. To 0.5 ml. of this ifitrate were added 2.0 ml. of alkaline copper reagent and the mixture heated for 20 minutes in a boiling water bath. Then 2.0 ml. of arsenomolybdate were added and the mixture diluted to 25 ml. and read at 520 m. Before the commercial bath was available a water bath for temperature control of the dialyzer was made at this institution. A stirrer and temperature controlling unit with overall regulation of about ±0.3#{176}was mounted on the bath. When changing the cellophane dialyzing membrane, the bath was drained and the dialyzing unit disconnected and removed from the bath. Such changes are not usually required more often than once a week.

Vol. 5, No. 6. 1959 GLUCOSE DETERMINATION 545 90 80 6o E 50 Fig. 1. Concentration-transmittance curves for glucose determination by auto- 40 rye A matic procedure. Curve B obtained approximately 2 hours after curve A. All I I I I points are averages for 5 consecutive 0 40 80 120 160 200 240 Glucose Concentration in Mg. Percent An example of the change in transmittance readings with time which occur without the water bath is indicated in Fig. 1. These curves were prepared from averages of data for S consecutive mornings in the routine clinical chemistry laboratory. Curve A was obtained at about 9 A.M. and curve B at about 11:00 A.M. Between these two times about 60 blood samples were processed each morning. A maximum error of about 8% in the determination of an unknown blood sugar could have occurred as a result of this change. This error can be reduced considerably by running several standard curves during the period of analysis and using the standard curve nearest the unknown in time for the determination of that unknown concentration. To keep the error caused by standard curve shift within a few per cent, therefore, at least 3 standard curves are required during a 2-hour period. Changes of the type described above for glucose do not occur in the standard curves obtained in the determination of urea by the carbamido diacetyl monoxime reaction (5, 6). It is known that the dialysis of urea is sufficiently rapid that in passing through the dialyzer, urea is essentially in equilibrium across the cellophane membrane (1). Room temperatures ordinarily encountered are probably not low enough to reduce the diffusion rate below that which results in equilibrium, and hence changes in room temperature do not affect the total amount of urea dialyzed. With glucose, due to its lower diffusion rate, such equilibrium is not achieved at room tem-

546 GRADY & lamar Clinical Chemistry Table 2. VARIATION IN TRANSMITFANCE OF GLUCOSE 5mDARDs WITH TEMPERATUU Glucose concentration mg.iloomi. 50 100 200 250 800 Temperature e#{216} Mean % transml.ttance ± standard deviation 23.8-24.4 28.7±0.07 31.5±0.07 44.0±0.40 54.3±0.60 72.3±0.61 25.0-25.4 28.6±0.07 32.0±0.15 44.9±0.39 57.1±1.27 77.4±0.90 25.8-26.2 28.6±0.11 32.2±0.20 45.8±0.50 56.9±0.32 78.1±1.40 27.0-27.4 28.2-28.6 28.7±0.08 28.7±0.11 32.3±0.12 32.6±0.17 46.7±0.38 59.5±0.95 48.0±0.43 62.5±0.69 80.5±0.93 84.9±1.53 29.4-29.8 28.9±0.17 32.9±0.16 49.8±0.43 64.8±0.77 89.4±0.72 31.2-31.8 9.0±0.19 33.4±0.12 50.7±0.73 65.8±0.97 90.0±1.27 5A11 values are averages of 6 consecutive determinations of the series of 5 standards analyzed at the rate of 20 per hour. Standard deviation of each group of 6 readings was calculated as follows: /S X2 X\2 standard deviation = - I N \N where X = the individual reading in percent transmittance and N = number of readings. perature, and it seemed reasonable that the change in the amount of color produced by a given standard was caused by a change in the rate of dialysis of the glucose. One of the most obvious factors affecting diffusion rate is the temperature of the dialyzer. A study was made of the variation in reading of the glucose standards with temperature of the dialyzer using the constant temperature bath described above. The results are presented in Table 2. It is clear that change in temperature causes significant variations in the percentage of transmittance of a given standard. That these changes can be explained principally by the temperature effect on the rate of dialysis of glucose is indicated by Table 3 and Fig. 2. Table 3 indicates the actual percentage of the glucose on the sample Table 3. Err8cv OF TEMPERATURE ON AMOUNT 01 GLucosE DIALYZZD Temperatur. #{176}0. % glucose dialyzed 30#{176} 9.7 32#{176} 10.1 34.2#{176} 10.7 36#{176} 11.2 38.6#{176} 11.8 All values are averages of 3 consecutive determinations of 100 mg. per 100 ml. and 200 mg. per 100 ml. standards.

Vol. 5, No. 6. 1959 GLUCOSE DETERMINATION 547 34 Temperature #{176}C. Fig. 2. A comparison of the experimental and theoretical increase in glucose dialyzed due to increase in dialyzer temperature. Each point of the experimental curve is the average per cent of increase of glucose concentration in the dialysate of the 100 and 200 mg. per 100 nil. glucose standards compared to their concentration at 30#{176} C. as the reference. The experimental curve is the best straight line through these points calculated according to the method of least squares. I side which is dialyzed under the conditions described earlier. From 10 to 12% diffuses across compared to a theoretical maximum of 50%, and therefore about one-fifth of the equilibrium value is obtained. That the increase in glucose dialyzed is due to the change in the diffusion constant of glucose with temperature is shown in Fig. 2. The theoretical curve is the expected per cent increase in glucose dialyzed as the temperature rises. This was calculated on the assumption that the size of the pores of the cellophane is large compared to the glucose molecule and that the temperature effects are due only to changes in the diffusion constant of glucose in water. The points were calculated using the following equation (7): DT1 = D.1 2 T1nT2 T2nT2 Table 4. Mwc CHANGE IN TRANSMITTANCE READINGS FROM INITIAL READING FOR 2-Houa PERIOD EXPRESSED IN PER CENT T CHANGE ± STANDARD DEVIATION#{176} Concentration (,ngfloo of standard mi.) Tsmperatur. not controlled (without H00 bath) 1 einperatter. controlled (with H00 bath) 100 0.66±0.37 0.28±0.20 200 3.60±1.85 1.30±0.77 250 6.00±3.54 2.10±1.19 REach mean is calculated from data of 20 consecutive days. Mean change is reported without regard to sign (direction of change).

548 GRADY & LAMAR Clinical Chemistry where Dr1 and DT2 represent the diffusion constants of glucose at the two temperatures, T1 and T2, and nt1 and flt2 represent the viscosity of water at the two temperatures. After use of the bath for controlling the dialyzer temperature, the variations in the standard readings decreased significantly. As can be seen in Table 4, the changes in per cent transmittance readings with time have decreased to about one-third of their original value and tend to be more consistent (lower standard deviation). Thus with temperature control a change of about 2% transmittance oc- Table 5. RECOVERY Or GLUCOSE BY AUTOMATIC ANALYSIS AND MAN1LTAL METHODS Au to,,,mtic analysis Manual Glucose Glucose Glucose Glucose deter-mined recovered deter-in med recovered Test (average (average % (average (average % sw.terial Sample mg. %) isp. %) recovery ing. %) ing. %) recovery Bloood A 78 75 (4 replicates) (duplicates) Blood A plus 177.7 99.7 99.7 173.5 98.5 98.5 100 mg./100 ml. (duplicates) (4 replicates) Blood A plus 278 200 100 273 198 99 200 mg./100 ml. (duplicates) (4 replicates) Blood B 1 75 Blood B plus 272 197 98.5 200 mg./100 ml. 1 BloodO 1 50 Blood C plus 152 102 102 100 mg./100 ml. 1 Blood C plus 247 197 98.5 200 mg./100 ml. 1 Labtrol 117 102 113 98.5 (Dade value: 115) 1 Blood D (1:3 dilu- 1 20 20 18 16.5 tion with buffer)b 2 20 15 16.5 Blood D (1:3 dilu- 1 0 0 tion with 30% 2 0 0 yeast in buffer)b 3 0 4 0 Average recovery 100.1 98.7 Crystalline reagent grade glucose was weighed and diluted to proper volume with the appropriate blood. bo.1 M. phosphate, ph 7.5.

Table 6. PRECISION OF DETERMINATION OF GLUCOSE BY AUTOMATIC ANALYSIS Usmo HOFFMAN METHOD AND COMPARISON WITH MANUAL NELSON-SOMOGYI METHOD Automatic (Hoffman) Manual (Nsison-Romogyi) glucose concentration glucose concentration (mg./100 nil.) (mg./100 nil) Blood Sample Sample Blood Sample Baeepi. No. 1 2 No. 1 2 1 72 74 46 63 64 2 70 70 47 103 108 3 196 197 48 168 171 4 78 75 49 127 127 5 44 43 50 147 152 6 112 113 51 168 174 7 96 96 52 59 79 8 151 153 53 61 64 9 133 133 54 107 105 10 103 103 55 79 83 11 92 92 56 122 131 12 87 87 57 57 55 13 129 129 58 101 107 14 82 82 59 49 51 15 116 117 60 63 64 16 107 107 61 57 59 17 111 109 62 49 53 18 117 117 63 77 74 19 152 153 64 69 74 20 254 255 65 154 153 21 79 76 66 174 173 22 72 76 67 173 167 23 56 56 68 121 117 24 108 105 69 131 131 25 103 105 70 114 111 26 176 175 71 58 57 27 72 72 72 68 67 28 55 55 73 117 118 29 68 69 74 62 62 30 85 86 75 89 85 31 103 103 76 92 95 32 78 78 77 63 58 33 73 72 78 85 85 34 139 140 79 124 123 35 74 72 80 74 79 36 72 72 81 106 101 37 80 79 82 150 160 38 76 78 83 194 194 39 75 75 84 140 138 40 81 82 85 65 66 41 94 94 86 185 188 42 80 80 87 198 188 43 80 81 88 143 143 44 89 90 89 150 150 45 85 84 90 70 66 Standard deviation = ±0.803 Standard deviation = ±2.76 Mean 99.0 99.1 Mean 107.2 108.2 Standard deviation of each method was calculated as follows where d = difference between duplicates Standard deviation (8)

550 GRADY & LAMAR Clinical Chemistry curred with the 250 mg. per 100 ml. standard and this would cause a maximum error of about 1.6% in the estimation of the glucose concentration of an unknown. Table 5 shows the recovery of glucose by the automatic procedure employing the Hoffman method compared to the manual Nelson-Somogyi method. It is seen that added glucose is recovered well by both procedures but that the automatic procedure is superior in this respect. Incubation of the blood with yeast removes the glucose, and the results from these samples is good evidence that the blood blank is zero and that everything that is measured is actually glucose. The combination of a zero blank and essentially 100% recovery demonstrates the high accuracy of the automatic procedure. The precision of the automatic method has been calculated from the differences between duplicates as shown in Table 6, and it is seen that the automatic method is more precise. The duplicates were run one immediately following the other, and in the manual procedure duplicates were all analyzed at the same time. SUMMARY A method of automatic chemical analysis has been improved by stabilizing the temperature of the dialyzing unit. With this addition the automatic analyzer has been shown to have greater precision and accuracy than a manual method. REFERENCES 1. Skeggs, L. T. Am. J. CUn. Paihol. 28, 311 (1957). 2. Hoffman, W. S. Photeioinetrio Clinical Chemi.stry. New York, William Morrow, 1947. 3. Johnson, J. Am. J. Med. Tech. 24, 271 (1958). 4. Beiner, M. (Ed.) Standard Methods of Clinical Chemistry. Vol. I. New York, Academic Press, 1953. 5. Friedman, H. S. Anal. Chem. 25, 662 (1953). 6. Marsh, W. H., Fingerhut, B., and Kirach, E. Am. J. Chn. Paihol. 28, 681 (1957). 7. Bull, H. B. Physical Biochemistry. New York, John Wiley, 1951. 8. Youden, W. J., Statistical Methods for Chemists. New York, John Wiley, 1951.