Bedside instruments that measure or estimate whole blood hemoglobin (Hb) in the operating room may be affected by hemodilution, red blood cell damage associated with cardiopulmonary bypass (CPB), and red blood cell transfusion. The conductivity method has been demonstrated to be accurate under normal conditions [1-4]. However, this method becomes inaccurate when serum electrolytes and/or protein concentrations deviate from normal, as during hemodilution . An adjusted conductivity Hb estimate can be obtained by correcting for changes in non-red-cell constituents. Consequently, multifunction instruments that also measure electrolytes may correct the conductivity measurements, usually with an algorithm based on the sodium concentration. Photometry is another method used to quantify the Hb content of whole blood. This technique was demonstrated to be accurate in nonoperative clinical settings [6-10] and with in vitro hemodilution , but has not been evaluated with in vivo hemodilution and transfusion. Centrifugation estimations for Hb are sensitive to factors that alter the Hb content of red blood cells or result in red cell lysis, such as CPB. The relative accuracy of whole blood centrifugation has been compared to uncorrected conductivity  but not to bedside photometry or adjusted conductivity techniques during the clinical conditions associated with CPB, hemodilution, and blood transfusion.
This study compares bedside estimations or measurements of Hb with the Coulter counter using conductivity, adjusted conductivity, photometric, and centrifugation techniques. The purpose of this study was to evaluate the relative accuracy of four different methods of assessing Hb intraoperatively, under clinical conditions that included euvolemia, hemodilution, and red blood cell transfusion.
After approval by our institutional review board, 25 patients undergoing cardiac surgical procedures, gave written informed consent to participate in this study. There were no exclusion criteria for this study.
Four test methods were evaluated: conductivity by Stat-Crit Registered Trademark (Wampole Laboratories, Cranbury, NJ); adjusted conductivity by Nova Stat Profile 9 (Nova Biomedical, Waltham, MA); photometry by HemoCue Registered Trademark (HemoCue, Inc., Mission Viejo, CA); and centrifugation. Centrifugation hematocrit measurements were performed in triplicate and the average value was divided by three to convert to Hb. The method of conversion of centrifuge hematocrit to Hb corresponded to methods incorporated into conductivity and adjusted conductivity devices. All other Hb tests were performed sequentially using a single measurement by a single operator. The control Hb was measured using the Coulter method, Coulter Registered Trademark STKR 1 (Coulter Electronics, Inc., Hialeah, FL) by our institution's clinical laboratory. Blood sodium and chloride were measured using the Nova Stat Profile 9 and serum total protein was determined by laboratory analysis for each blood sample.
Patient blood samples (15 mL) were obtained before induction of anesthesia (Time 1), after the start of CPB (Time 2), after termination of CPB (Time 3), and after transfusion of all available autologous blood (Time 4). Autologous blood transfusions before Time 4 consisted of cell-saved red blood cells in 10 patients and residual bagged pump blood in 15 patients. Scavenged whole blood was processed using a Dideco BT-795-A Cell Saver and 1000 mL of normal saline for concentrating and washing red blood cells.
All test Hb measurements were performed immediately after sampling from a single 20-mL syringe of blood that was mixed continuously. Blood was obtained from an indwelling radial artery catheter at Time 1 (n = 25) and at Time 4 (n = 25). At Times 2 (n = 25) and 3 (n = 15) blood was obtained from a stopcock connected to a continuous purge line from the arterial filter of the CPB circuit. Patient blood samples from Times 1 to 4 (n = 90) were considered as in vivo. Standardized methods were used for all blood sampling. All radial artery blood samples were obtained after first withdrawing 10 mL of blood, due to the presence of crystalloid in the high-pressure tubing to the transducer. All CPB blood samples were obtained after withdrawing and reinjecting 3 mL of blood.
To compare the different methods under in vitro conditions, 10 blood samples (15 mL) were obtained from bags of cell-saved blood. Tests were performed identically to those for the in vivo blood samples. Blood samples obtained from the bags of processed blood were considered in vitro. Sodium and chloride were measured using the Nova Stat Profile 9 and serum total protein was determined by laboratory analysis for each of these blood samples as well.
The Stat-Crit Registered Trademark, Nova Stat Profile 9, and HemoCue Registered Trademark systems were confirmed to be operating within established limits by using the manufacturers' calibration and maintenance protocols. All study measurements were performed within 3 h of quality control testing. The Coulter counter was calibrated against a reference standard (5C Coulter Registered Trademark Cell Control). Precision analysis for Hb measurement by the Coulter method had a mean difference of 0.02 +/- 0.08 mg/dL using a reference standard.
All patients received isotonic intravenous fluids of Plasmalyte Registered Trademark balanced salt solution (Baxter Healthcare Corp., Deerfield, IL) and normal saline. The CPB prime was 1600 mL Plasmalyte Registered Trademark and 500 mL Hespan Registered Trademark (McGaw Inc., Irvine, CA) with 12.5 g mannitol, 44 mEq sodium bicarbonate, and 10,000 U heparin.
Bias was calculated for each test method by Coulter Hb--test method Hb. The bias values for conductivity, adjusted conductivity, photometric, and centrifugation test methods were compared using a factorial analysis of variance (ANOVA). Between-group differences were subsequently determined using Fisher's protected least significant difference (PLSD).
The changes in bias resulting from hemodilution and blood transfusion for in vivo blood samples at Times 1, 2, 3, and 4 were compared using ANOVA for repeated measures. In vitro bias values were compared to in vivo bias values with the Kruskal-Wallis test.
Percentage differences from the Coulter Hb were calculated by [(Coulter Hb--test method Hb)/Coulter Hb] times 100. Absolute values of the percentage differences between the different test methods were compared using ANOVA.
The changes in total protein, chloride, and sodium resulting from hemodilution and blood transfusion for in vivo blood samples at Times 1, 2, 3, and 4 were compared using ANOVA for repeated measures. In vitro blood chemistries were compared to in vivo blood chemistries with the Kruskal-Wallis test.
Correlations between the Coulter Hb and paired test method values were performed by linear regression analysis.
Coulter Hb served as reference in the Figure plots. All statistical analyses were performed using the Stat View II Trademark statistical package (Abacus Concepts, Inc., Berkeley, CA). A P < 0.05 was considered to be statistically significant. All results are reported as mean +/- SD.
The range of Coulter Hb values for all sample times was 4.2-20.7 g/dL. The frequency distributions of calculated bias values (Coulter--test method Hb) were considered to be normal for the in vivo (n = 90) 1and in vitro (n = 10) blood samples. The frequency distributions of in vivo and in vitro data combined was non-parametric. The in vivo Hb bias values for each of the test methods were: Stat-Crit Registered Trademark = 0.6 +/- 0.8 g/dL Figure 1; Nova Stat Profile 9 = -0.7 +/- 0.4 g/dL Figure 2; HemoCue Registered Trademark = -0.1 +/- 0.2 g/dL Figure 3; and centrifuge = 0.1 +/- 0.5 g/dL Figure 4. These values were significantly different from each other (P < 0.0001, ANOVA, Fisher's PLSD). Hb bias values for in vitro samples (n = 10) obtained from processed blood were Stat-Crit Registered Trademark = 5.1 +/- 0.6 g/dL; Nova Stat Profile 9 = 3.0 +/- 0.6 g/dL; HemoCue Registered Trademark = 0.4 +/- 0.4 g/dL; and centrifuge = 0.6 +/- 0.3 g/dL.
Due to the wide range of Hbs in this study, the percentage difference from the Coulter Hb was determined for each test method. The absolute value (mean +/- SD) percentage differences for each instrument were: Stat-Crit Registered Trademark = 11.7% +/- 8.7%, Nova Stat Profile 9 = 9.0% +/- 5.5%; HemoCue Registered Trademark = 1.6% +/- 1.3%; and centrifuge = 4.8% +/- 4.7%. These differences were significant by ANOVA (P < 0.0001) and, as determined from post hoc tests, each was significantly different from the other methods (P < 0.002, Fisher's PLSD). Of 90 in vivo samples, the number with a > 10% difference from Coulter measured Hb were: Stat-Crit Registered Trademark = 42; Nova Stat Profile 9 = 37; HemoCue Registered Trademark = 0, and centrifuge = 11. The incidence of over- or underestimation of Coulter Hb when the Hb was < 8 g/dL is summarized in Table 3.
Hemodilution and red blood cell transfusion resulted in significant differences in calculated bias values for Stat-Crit Registered Trademark, Nova Stat Profile 9, and HemoCue Registered Trademark from Time 1 (before induction) to Times 2, 3, and 4 Table 1. Comparison of in vivo bias values Table 1 were significantly different from the in vitro values for the different test methods (P < 0.0001, Kruskal-Wallis test).
Hemodilution and red blood cell transfusion also resulted in significant differences in values for total protein, chloride, and sodium for the in vivo (Times 1 to 4, n = 90) and in vitro samples. Sodium and chloride were significantly higher, and the total protein was significantly lower, in the processed cell-saver samples Table 2 compared to the in vivo blood samples (P < 0.0001, Kruskal-Wallis test).
Linear regression analyses of Coulter Hb versus test-method Hb for all measurements yielded R2 values of 0.938, 0.972, 0.996, and 0.960 for Stat-Crit Registered Trademark, Nova Stat Profile 9, HemoCue Registered Trademark, and centrifuge, respectively.
This study evaluated the differences between various bedside methods for Hb estimation or determination compared to clinical laboratory measurements performed with Coulter counter instrumentation. Cardiac surgery patients were studied to achieve a wide range of Hb values under in vivo conditions that included euvolemia, hemodilution, and blood transfusion. Of the four test methods used in this investigation, only the HemoCue Registered Trademark measured Hb directly. Although the two conductivity methods provide a Hb reading, it is calculated from a measured hematocrit then divided by three for conversion to Hb. The Coulter measures Hb and hematocrit independently. Good correlation (R2 = 0.995) and a low bias +/- SD (0.2 +/- 0.18 g/dL) between the Coulter Hb and the Coulter hematocrit converted to Hb provide some justification for using a fixed conversion ratio. Comparisons between the test methods were also made using the Coulter hematocrit as the reference standard; however, the results were not significantly different. Hb, rather than hematocrit, was chosen as a basis for comparing these devices since it is most relevant to the oxygen-carrying capacity of blood.
The uncorrected conductivity percentage difference from Coulter measured Hb was significantly greater than adjusted conductivity, photometric, and centrifugation methods. The observed differences in our study between Coulter and conductivity measurements were consistent with earlier findings [1-5,12,13]. Some of the previous studies reporting on the relative inaccuracy of conductivity measurements have used in vitro conditions in the evaluation. In the present study, blood processing with the cell salvaging device resulted in disproportionate increases in chloride and sodium, as well as a decrease in total protein, that were outside the range of electrolytes and protein encountered with the in vivo blood samples.
Adjusted conductivity estimates of Hb were not as accurate as the centrifugation and bedside photometric methods. The Nova Stat Profile 9 corrects the hematocrit (Hct) electrical resistance measurement based on changes in sodium. The algorithm is Hct % = 1- (1/RCorr) times 100, where RCorr = RSample times ([Na+]/140) (1/2). RCorr and RSample are the corrected and uncorrected blood sample resistances, respectively. This resulted in an overestimation of Hb compared to the Coulter measurement, possibly due to the exclusion of total protein and chloride in the algorithm . It should be noted that the precision of the Nova Stat Profile 9 (SD of bias) was better than that of the centrifuge and Stat-Crit Registered Trademark so that with a simple bias adjustment, precision of the percentage difference would be reduced to 6%. The precision of the Nova Stat Profile 9 measurements enables clinical management comparable to the HemoCue Registered Trademark and centrifuge, as long as relative changes in Hb are considered and an absolute value for Hb is not used as a deciding factor for blood transfusion.
The agreement between bedside photometry and Coulter methods was similar to a previous test of this device in an in vitro hemodilution model . However, the differences in accuracy between bedside photometry and conductivity methods were less in the present study with in vivo blood samples than with those found in the previous in vitro study .
The centrifuge was nearly as accurate as the bedside photometry method at estimating Coulter Hb in this study. There was a tendency toward a slight underestimation of Hb at low Hb levels. The potential for significant inaccuracy based on variations in red cell Hb content and hemolysis related to CPB was evident in only a small percentage of patients.
The expected percentage difference between several instruments measuring Hb in a laboratory is +/- 7% of the target value, as suggested by the Clinical Laboratories Improvement Act of 1988 . Under the conditions of this study, all test methods were not expected to provide identical measurements so we arbitrarily chose to report the number of test samples with a >10% difference from Coulter measured Hb as being clinically significant. There were 49 samples with a Coulter measured Hb <8 g/dL. This represents a clinically sensitive range when red blood cell transfusion is often considered. The incidence of over- or underestimation of the Coulter Hb summarized in Table 3 suggests that if present standards of quality control were strictly applied to the various instruments used under the conditions of this study, the conductivity methods would not be in conformance.
This study did not address any of the economic or practical aspects involved with bedside monitoring of Hb. In comparing the costs of disposables and equipment, the more portable devices such as the centrifuge, Stat Crit Registered Trademark and HemoCue Registered Trademark are much less expensive than the Coulter counter or Nova Stat Profile 9 when fewer tests are performed. However, a device such as the Nova Stat Profile 9 is difficult to compare on a cost basis since it provides more information independent of the Hb and hematocrit that could be useful to patient management. It is also important to consider that the laboratory costs of the Coulter Hb includes quality control testing, equipment maintenance costs, and record keeping costs that would be necessary for the other devices as well.
We conclude that the bedside photometric and centrifuge methods provided the closest approximation to the Coulter Hb. Assessment of conductivity Hb with in vitro conditions exaggerated the inaccuracy of these devices, probably due to an abnormal range of sodium, chloride, and total protein not found with in vivo measurements in this study. The rank order of closest approximation to the Coulter measurement was HemoCue Registered Trademark, centrifuge, Nova Stat Profile 9, and Stat-Crit Registered Trademark during euvolemia, hemodilution, and after scavenged autologous blood transfusion.
1. Rosenthal RL, Tobias CW. Measurement of the electrical resistance of human blood; use in coagulation studies and cell volume determinations. J Lab Clin Med 1948;33:1110-22.
2. Hirsch FG, Texter EC, Wood LA, et al. The electrical conductivity of blood. I. Relation to red cell concentration. Blood 1950;5:1017-35.
3. Kernen JA, Wurzel H, Okada R. New electronic method for measuring hematocrit: clinical evaluation. J Lab Clin Med 1961;57:635-41.
4. Baris RR, Israel AL. Clinical evaluation of the Stat-Crit monitor. Proc Am Acad Cardiovasc Perfusion 1988;9:37-42.
5. McNulty SE, Sharkey SJ, Asam B, Lee JH. Evaluation of STAT-CRIT Registered Trademark hematocrit determination in comparison to Coulter and centrifuge: the effects of isotonic hemodilution and albumin administration. Anesth Analg 1993;76:830-4.
6. von Schenck H, Falkensson M, Lundberg B. Evaluation of "HemoCue," a new device for determining hemoglobin. Clin Chem 1986;32:526-9.
7. Nicholls PD. An evaluation of the HemoCue Trademark for correcting the haemoglobin value of lipaemic samples. Med Lab Sci 1990;47:226-9.
8. Chen PP, Short TG, Leung DHY, Oh TE. A clinical evaluation of the hemocue haemoglobinometer using capillary, venous and arterial samples. Anaesth Intensive Care 1992;20:497-500.
9. Berry SM, Dombrowski MP, Blessed WB, et al. Fetal hemoglobin quantitations using the HemoCue system are rapid and accurate. Obstet Gynecol 1993;81:417-20.
10. Be WKM, Kerkkamp HEM, Booij LHDJ. Hemocue--a new haemoglobinometer in the clinic. Eur J Anesth 1991;8:55-8.
11. McNulty SE, Sharkey SJ, Schieren H. Bedside hemoglobin measurements: sensitivity to changes in serum protein and electrolytes. J Clin Monit 1994;10:377-81.
12. McMahon DJ, Carpenter RL. A comparison of conductivity-based hematocrit determinations with conventional laboratory methods in autologous blood transfusions. Anesth Analg 1990;71:541-4.
13. Vedrinne JM, Motin J. Colloids and hematocrit measurement by conductivity. Anesth Analg 1991;72:840.
© 1995 International Anesthesia Research Society
14. Department of Health and Human Services. Regulations implementing the clinical laboratory improvement amendments of 1988 (CLIA). Federal Register 1992;57:7002-187.