Many currently available point-of-care (POC) testing devices that measure arterial blood gas and electrolytes also assess hematocrit simultaneously. POC devices located in the operating room can be particularly useful during liver transplantation (LT) surgery because profuse bleeding and massive transfusion are frequent. The electrical conductivity method has been adopted in most POC testing devices to measure hematocrit, while other methods like centrifugation, microvolumetry, and electrical sensing zone are used in the laboratory.1–4 However, hematocrit measurement by conductivity method does have limitations because it can be influenced by plasma compositions, including electrolytes and protein concentrations. Previous studies have reported that a bias can be substantial during cardiac surgery involving cardiopulmonary bypass (CPB), where changes in sodium and protein concentrations are common and plasma volume expanders and anticoagulants are used.1,2,5 Inaccurate hematocrit measurements are more common when hematocrit values are less than 30%.6
Several laboratory abnormalities identified during LT may be associated with possible inaccuracies resulting from the use of POC testing devices; for example, reported confounders that can cause hematocrit measurement error with a POC device are low serum protein and high serum osmolality.6,7 Hypoproteinemia or hypoalbuminemia, which is frequent in patients with end-stage liver disease, may also result in falsely low hematocrit measurements during LT. Hypernatremia is the main cause of high osmolality and is known to cause hematocrit bias when using POC devices. However, the effect of hyponatremia, which is more common in LT recipients, on hematocrit measurement has not been evaluated. Hyperglycemia is another common laboratory abnormality during LT that may lead to erroneous hematocrit measurement due to the increase in serum osmolality but has never been reported. These laboratory abnormalities often develop in various combinations, but their mixed effect during LT on the accuracy of hematocrit measured using POC testing devices remains unclear.
Red blood cell transfusions are frequently performed during LT based on the guidance of POC hematocrit measurements. Although transfusion thresholds vary according to surgical situations and patient characteristics, evidence suggests that a hematocrit value of less than 20% is associated with adverse outcomes and an increase in hospital mortality.8 Unnecessary overtransfusion may also occur when POC hematocrit values are <20%, especially while laboratory-measured hematocrit is ≥20%. Intraoperative transfusion of red blood cells has been associated with reduced survival and acute kidney injury following LT.9–11
In this retrospective study, we hypothesized that POC testing devices may report frequent false hematocrit values in recipients of LT due to the mixed effects of hypoproteinemia or hypoalbuminemia, hyponatremia, and hyperglycemia, which are common during the perioperative period. Therefore, the primary aim of this study was to explore the incidence and magnitude of the discrepancies between the hematocrits measured using POC testing devices and laboratory tests during LT procedures and to investigate the associated factors that may contribute to the discrepancy. The secondary aim of this study was to identify the rate and amount of potential overtransfusion related with this discrepancy during LT.
This retrospective observational study was approved by the Institutional Review Board of Seoul National University Hospital (Institutional Review Board number: H-1408-076-604). Requirement of informed consent was waived for this study due to the study’s retrospective design.
We reviewed electronic medical records and laboratory data of consecutive adult and pediatric patients who underwent LT surgery between January 2010 and February 2016. After obtaining all intraoperative POC and laboratory blood test results, hematocrit measurement pairs with concomitant albumin data were analyzed. Single measurements of hematocrit obtained either using a POC device or using a laboratory test, measurements with no concomitant albumin values, and obvious erroneous values were excluded.
Data Collection and Study Protocol
Patient age and weight, estimated blood loss, and amount of erythrocyte transfusion were retrieved from electronic anesthesia records. Blood test results including hematocrit (POC and laboratory), serum total protein (laboratory), serum albumin (laboratory), serum sodium (POC), and serum glucose (POC) values were collected from electronic medical recording system.
Intraoperative blood tests were performed 8 times according to our anesthesia protocol: after anesthesia induction, 1 hour after anesthesia induction, 10 minutes after the beginning of the anhepatic phase, 5 minutes before and after graft reperfusion, 20 minutes after reperfusion, 5 minutes after the completion of biliary reconstruction, and at the end of surgery. Samples for POC tests (measuring arterial blood gas, hematocrit, electrolytes, and glucose), common blood cell counts (measuring hematocrit), and chemistry tests (measuring albumin and protein) were drawn from the patient’s radial arterial line at the same time. The sample for POC test was drawn with a heparin-coated syringe, and the sample for laboratory hematocrit measurement was stored in an EDTA tube.
The POC test device used was the GEM Premier 3000 (Instrumental Laboratory, Bedford, MA) analyzer.12 Laboratory testing was performed using blood samples taken simultaneously with POC sampling and were analyzed within 15 minutes of blood sampling. Total protein, albumin, and sodium levels were measured using the TBA 200FR (Toshiba Medical Systems, Ōtawara, Japan) chemistry analyzer. Laboratory hemoglobin and hematocrit values were measured using the XE-2100 (Sysmex Corp, Kobe, Japan) and XN (Sysmex Corp) automated hematology systems from January 2010 to September 2013 and October 2013 to February 2016, respectively. All tests were performed during the entire study period, except for total protein level, which was measured from May 2012 to February 2016. The hematocrit threshold for red blood cell transfusion was 20% during LT, according to our anesthesia protocol.
Study Groups and Study End Points
The primary outcome variable for this study was hematocrit measurement error between POC hematocrit and laboratory hematocrit, which was the difference between the 2 values calculated using the following equation: hematocrit measurement error (%) = hematocrit measured by POC device (%) − hematocrit measured by laboratory device (%). The measurements within patients were classified into 1 of 2 categories, either hematocrit measurement error <0 (falsely low-measured hematocrit) or hematocrit measurement error ≥0.
In addition, patients who received transfusion with POC hematocrit <20% were classified into 2 groups according to the laboratory hematocrit value. We judged the transfusion as being adequate if both the POC hematocrit and laboratory hematocrit were less than 20%, the transfusion threshold. However, if the POC hematocrit was less than 20% but the laboratory hematocrit was 20% or greater, it was judged that there may have been an “overtransfusion.” Patients with laboratory hematocrit measurements of <20% and ≥20% were assigned to adequate transfusion and overtransfusion groups, respectively. Certain patients were excluded from these groups, however, if their measurements belonged to both groups at the same time.
Descriptive statistics of age, weight, estimated blood loss, and the amount of red blood cell transfusions of all patients were calculated. After the Shapiro-Wilk normality test, 2 categories of hematocrit measurement errors were compared using the Wilcoxon signed rank test in terms of serum total protein, albumin, sodium and glucose, POC and laboratory hematocrits, and hematocrit measurement error. These variables were tested on measurement basis rather than patient basis.
Bland-Altman analysis for repeated measurements was used to test the agreement of each pair of laboratory and POC hematocrit, with bias defined as the mean difference between POC hematocrit and laboratory hematocrit and the upper and lower limits of agreement defined as ±1.96 SD of the bias.13 The mean bias and limits of agreement were estimated by a component of variance technique adjusting for the clustered observations. In addition, clinical limits of agreement were defined as differences within ±3% as compared with laboratory hematocrit. The proportion and confidence intervals (CIs) outside of these limits were calculated using variability adjusting for within-patient clustering.
Continuous variables were categorized according to common laboratory abnormalities found during LT like hypoalbuminemia (serum albumin <3.3 g/dL) or hypoproteinemia (serum protein <6.0 g/dL), hyperglycemia (serum glucose >200 mg/dL), and hyponatremia (serum sodium <135 mEq/L). The frequency and magnitude of falsely low-measured hematocrit in hypoalbuminemia or hypoproteinemia, combined with hyperglycemia and hyponatremia, were calculated to evaluate the effect of common laboratory abnormalities of LT on the hematocrit measurement error. The incidence of hematocrit measurement error was compared between hypoalbuminemia, hypoproteinemia categories, and their counterparts using the Pearson χ2 test. The magnitude of hematocrit measurement error was then compared with the Kruskal-Wallis test, and post hoc pairwise comparison was completed among categories.
The multivariable generalized estimating equation models were used to account for multiple measurements per patient and to assess the effects of factors predictive of the falsely low-measured hematocrit. The independent variables were hypoalbuminemia or hypoproteinemia, hyperglycemia, and hyponatremia, and the dependent outcome was the falsely low measurement of hematocrit. The amount of transfusion between adequate transfusion and overtransfusion groups was compared using Mann-Whitney U test.
The sample size calculation was based on the expected precision of POC hematocrit in liver recipients. When the proportion and the width of 95% CI of POC hematocrit within clinical limits of agreement was assumed to be 70% and 5%, respectively, the sample size was identified to be 323 measurements. If a similar degree of error is observed in all 8 measurements in 1 patient, the total number of samples would be the same number as the measurements, ie, a maximum of 323 patients.
Statistical analyses used SPSS software (version 21.0; IBM Corp, Armonk, NY). MedCalc software (version 16.2.1; MedCalc Software bvba, Mariakerke, Belgium) was used for the Bland-Altman analysis. G*power (version 184.108.40.206; Universität Düsseldorf, Düsseldorf, Germany) was used for the sample size estimation. A P value of <.05 was considered significant.
A total of 7784 blood sampling pairs were collected during LT in 973 patients between January 2010 and February 2016. Following exclusion of cases with loss of hematocrit data (n = 319), single hematocrit only (n = 732), no concomitant albumin data (n = 266), or unreliable test results (n = 6; 2 cases of POC hematocrit = 0% and 4 cases of albumin level more than protein level), 6461 pairs of hematocrit values in 901 patients were analyzed.
The median (interquartile range) of age, weight, estimated blood loss, and red blood cell transfusions in all patients were 54 (47–60) years, 63 (54–70) kg, 1900 (985–4500) mL, and 6 (3, 12) units, respectively. Blood test results between the 2 categories of hematocrit measurement errors are compared in Table 1. The median (interquartile range) of overall hematocrit measurement error was −1.2% (−2.6 to 0.2). The falsely low-measured hematocrit group consisted of 70.3% (4541/6461 pairs), and median (interquartile range) of the hematocrit measurement error was −2.0% (−3.2 to −1.0). All laboratory and POC variables were statistically different between the 2 groups.
Figure 1 shows a Bland-Altman plot with repeated measures, including all pairs of hematocrit measurement. The overall bias between the POC and laboratory hematocrit measurements was −1.1 ± 4.4% (mean ± standard deviation). Upper and lower limits of agreement determined by 1.96 SD (95% CI) were 3.9% (3.74–4.08) and −6.1% (−6.32 to −5.99). In addition, 24.5% (1583/6461) of hematocrit measurement errors were outside of our a priori defined clinically acceptable limits of ±3%.
Figure 2 shows the incidence and magnitude of falsely low-measured hematocrit according to categorized laboratory values. The incidence of falsely low-measured hematocrit was significantly higher in hypoalbuminemia and hypoproteinemia categories as compared with their counterparts (P < .001, both). In patients with hypoalbuminemia, hematocrit measurement error was found to be significantly larger when hyperglycemia was present (P < .001).
The results of a generalized estimating equation analysis to predict falsely low-measured hematocrit are shown in Table 2. Since albumin and protein levels significantly correlated to each other and serum protein concentration values were available in only 49.7% (3209/6461) of all tests, the analyses were performed separately for both the albumin and protein models. Missing values were present in 0.2% for serum sodium and 4.1% for blood glucose. The odds ratios (95% CI) of strong contributors like hypoalbuminemia and hypoproteinemia were 2.55 (2.16–3.00; P < .001) and 3.60 (2.97–4.37; P < .001), respectively.
The adequate transfusion and overtransfusion groups consisted of 85 patients with 158 measurements and 126 patients with 223 measurements, respectively (Table 3). The amount of transfused red blood cells was significantly larger in potential overtransfusion group than in the adequate transfusion group, with a median difference of 2 units (95% CI, 0–4; P = .039), despite similar blood loss (P = .103).
In this large retrospective comparative evaluation of the POC hematocrit and the laboratory hematocrit during LT, we found that the POC hematocrit was generally lower than the laboratory hematocrit. This discrepancy is more pronounced in patients with hypoalbuminemia or hypoproteinemia, as these were identified as strong contributors of falsely low measurement of hematocrit by POC device during LT. Hyperglycemia also contributed to the measurement error; however, the effect of hyponatremia was not identified. The potential for overtransfusion in our study sample was estimated to be as high as 6.4% during LT.
Centrifugation and electrical sensing zone methods are gold standards for measuring hematocrit in many laboratories.6 However, most POC analyzers adopt the electrical conductivity method to measure hematocrit,14 which measures the ability of the blood to transmit electrical current. Electrical conductivity can be influenced by the serum electrolytes, charged protein concentrations, and nonconducting cellular constituent; therefore, blood conductivity can be influenced by electrolyte disturbance and hyperglycemia during surgery.6
Previous literature has reported biases in conductivity-based hematocrit measurements in in vitro testing, cardiac surgery, and emergency situations.2,6,12,14–17 McMahon and Carpenter1 compared the hematocrit of salvaged intraoperative blood before and after cell washing by a cell saver device; specifically, they compared the conductivity method with the centrifuge and electrical sensing zone methods and revealed that the conductivity method gives a falsely low hematocrit measurement when serum protein is altered. Rosenthal and Tobias18 compared the conductivity and the centrifuge method in leukemia patients to evaluate the potential effects of these methods on white blood cell counts. They reported an overestimation of hematocrit due to poor conductivity in the presence of increased white blood cells. Steinfelder-Visscher and colleagues15 compared 127 blood samples obtained during CPB using the GEM Premier 3000 analyzer and laboratory testing. The results from the GEM analyzer deviated −2.2% (95% CI, −4.92 to 0.57) from the laboratory hematocrit and suggested that up to 67% (37/55) of transfusion guided by the device would have been unnecessary if a hematocrit of 20% had been used as the transfusion threshold. The POC blood analyzer was noted as one of the most frequently overlooked devices affecting an institution’s overall transfusion rate. One of the circumstances in which the usefulness of POC devices has not been investigated, in spite of frequent blood transfusion, is LT.
Surgical bleeding is frequent and often massive during LT, requiring rapid assessment of the hematocrit level with POC devices. However, LT recipients frequently show hypoalbuminemia, hypoproteinemia, hyponatremia, and hyperglycemia during pre- and intraoperative periods. These laboratory abnormalities may contribute to hematocrit measurement error by POC devices. Previous studies have evaluated the reliability of the POC hematocrit as a single laboratory parameter. However, the effect of common coexisting abnormalities in LT recipients, as shown in Table 1, on the reliability of the POC hematocrit measurements has not been investigated.
We showed that known major contributors like hypoproteinemia are significant confounders that cause falsely low POC hematocrit measurements. In addition, the effect of hypoalbuminemia was assessed because albumin is more important in the diagnosis and treatment of patients with end-stage liver disease perioperatively as compared with total protein. As expected, the impact of hypoalbuminemia was large enough to be of concern, but less than that of hypoproteinemia. Albumin constitutes only about half of total protein content, and the effect of other protein molecules seems to be obscured when only albumin is measured.
Both hyperglycemia and hypernatremia have been reported to significantly affect falsely low measurements of POC hematocrit. Hyperglycemia is common especially during massive transfusion and after reperfusion of liver grafts.19 Having a history of diabetes mellitus, administration of dextrose in red blood cell transfusion, use of steroids, surgical stress, and an ischemia-reperfusion injury are potential causes of hyperglycemia during LT.20 Interpretation of POC hematocrit in hyperglycemic LT recipients requires caution, because hematocrit could be falsely low in the presence of hyperglycemia. In contrast to previous studies, we tested the effect of hyponatremia instead of hypernatremia because hyponatremia is more common in the recipients of LT.21 However, hyponatremia was not a statistically significant predictor, which may be explained by the fact that GEM Premier 3000 corrects for serum sodium and that severe hyponatremia was rare in our patients.
Our data were analyzed using a hematocrit of 20% as the transfusion threshold. Although the transfusion thresholds differ depending on surgical situations and patient characteristics, a hematocrit <20% has been associated with adverse outcomes and increased in-hospital mortality.8 Overtransfusion may occur when the POC hematocrit is below 20% while the laboratory hematocrit is 20% or higher. The amount of red blood cell transfusion in the potential overtransfusion group was larger than that in the adequately transfused group despite similar amount of blood loss, suggesting that unnecessary transfusion may have occurred. However, there is also a possibility that patients requiring massive transfusion may have shown falsely low hematocrit levels due to diluted hypoproteinemia and hypoalbuminemia.
Our study has some distinct features. First, this is the first study that evaluated the accuracy of hematocrit measurement during LT using a POC device. In addition, data retrieved from clinical practice provided a very large sample size. Second, hematocrit data were analyzed in combination with laboratory abnormalities that are common in LT recipients.
Our study has some limitations. First, it investigated the accuracy of one POC device that uses the conductivity method. More recent devices such as i-STAT analyzer (Abbott Point of Care, Princeton, NJ) with an algorithm to correct for decreased plasma protein as well as serum sodium and potassium levels were not considered. However, a previous study during CPB demonstrated that residual bias remains after correction for serum sodium and potassium levels and after protein correction, while another study evaluated the effect of protein on hematocrit assays with i-STAT-1 and found bias of up to 2 g/dL for hemoglobin and up to 4% for hematocrit by the conductivity method.7,14 Results should be interpreted with caution when hypoproteinemia or substantial hemodilution is suspected and a conductivity-based POC device is used. Second, the POC and laboratory measurements of hematocrit may have been affected by the anticoagulant in the blood sample. The use of EDTA is reported to result in an average 1.8% lower hematocrit than the use of heparin due to osmotic shrinking of erythrocytes when conductivity methods of hematocrit measurement are used.22 The hematocrit measurement error might have been greater than that observed in the current study if the same coagulant was used for POC and laboratory blood samples. However, we believe that the quality control process in the laboratory using EDTA samples eliminated the baseline error due to the use of EDTA. Third, due to the retrospective design, our results can only suggest an association between falsely low POC hematocrit and potential overtransfusion. A prospective study is required to show the actual incidence of overtransfusion. Finally, external validity is limited due to the study’s single-center design. The incidence and magnitude of plasma component change may differ depending on the institution’s anesthesia protocol and surgical techniques.
In conclusion, our large retrospective analysis of POC hematocrit demonstrated that falsely low POC hematocrit occurred in 70.3% of hematocrit measurements from LT patients. The measurement error was greatest when hypoalbuminemia (or hypoproteinemia) and hyperglycemia coexisted. Consideration of these common laboratory abnormalities during LT may help reduce overtransfusion due to inaccurate POC hematocrit.
Name: Won Ho Kim, MD, PhD.
Contribution: This author helped design the study, analyze the data, and write the manuscript.
Name: Hyung-Chul Lee, MD.
Contribution: This author helped collect data, analyze the data, and write the manuscript.
Name: Ho-Geol Ryu, MD, PhD.
Contribution: This author helped analyze the data and write the manuscript.
Name: Eun-Jin Chung, MD.
Contribution: This author helped collect the data and write the manuscript.
Name: Borim Kim, MD.
Contribution: This author helped collect the data and write the manuscript.
Name: Hoiin Jung, DDS, PhD.
Contribution: This author helped analyze the data and write the manuscript.
Name: Chul-Woo Jung, MD, PhD.
Contribution: This author helped design the study, collect the data, analyze the data, and write the manuscript.
This manuscript was handled by: Marisa B. Marques, MD.
1. McMahon DJ, Carpenter RLA comparison of conductivity-based hematocrit determinations with conventional laboratory methods in autologous blood transfusions. Anesth Analg. 1990;71:541–544.
2. McNulty SE, Torjman M, Grodecki W, Marr A, Schieren HA comparison of four bedside methods of hemoglobin assessment during cardiac surgery. Anesth Analg. 1995;81:1197–1202.
3. Vedrinne JM, Motin JColloids and hematocrit measurement by conductivity. Anesth Analg. 1991;72:840.
4. McNulty SE, Sharkey SJ, Asam B, Lee JHEvaluation of STAT-CRIT hematocrit determination in comparison to Coulter and centrifuge: the effects of isotonic hemodilution and albumin administration. Anesth Analg. 1993;76:830–834.
5. Stott RA, Hortin GL, Wilhite TR, Miller SB, Smith CH, Landt MAnalytical artifacts in hematocrit measurements by whole-blood chemistry analyzers. Clin Chem. 1995;41:306–311.
6. Myers GJ, Browne JPoint of care hematocrit and hemoglobin in cardiac surgery: a review. Perfusion. 2007;22:179–183.
7. Hopfer SM, Nadeau FL, Sundra M, Makowski GSEffect of protein on hemoglobin and hematocrit assays with a conductivity-based point-of-care testing device: comparison with optical methods. Ann Clin Lab Sci. 2004;34:75–82.
8. Groom RCHigh or low hematocrits during cardiopulmonary bypass for patients undergoing coronary artery bypass graft surgery? An evidence-based approach to the question. Perfusion. 2002;17:99–102.
9. de Boer MT, Christensen MC, Asmussen M, et al.The impact of intraoperative transfusion of platelets and red blood cells on survival after liver transplantation. Anesth Analg. 2008;106:32–44.
10. Boyd SD, Stenard F, Lee DK, Goodnough LT, Esquivel CO, Fontaine MJAlloimmunization to red blood cell antigens affects clinical outcomes in liver transplant patients. Liver Transpl. 2007;13:1654–1661.
11. Park MH, Shim HS, Kim WHClinical risk scoring models for prediction of acute kidney injury after living donor liver transplantation: a retrospective observational study. PLoS One. 2015;10:e0136230.
12. Bénéteau-Burnat B, Bocque MC, Lorin A, Martin C, Vaubourdolle MEvaluation of the blood gas analyzer Gem PREMIER 3000. Clin Chem Lab Med. 2004;42:96–101.
13. Bland JM, Altman DGMeasuring agreement in method comparison studies. Stat Methods Med Res. 1999;8:135–160.
14. Teerenstra S, Steinfelder-Visscher J, Gunnewiek JK, Weerwind PWEvaluation of point-of-care analyzers’ ability to reduce bias in conductivity-based hematocrit measurement during cardiopulmonary bypass. J Clin Monit Comput. 2014;28:133–138.
15. Steinfelder-Visscher J, Weerwind PW, Teerenstra S, Brouwer MHReliability of point-of-care hematocrit, blood gas, electrolyte, lactate and glucose measurement during cardiopulmonary bypass. Perfusion. 2006;21:33–37.
16. Zatloukal J, Pouska J, Kletecka J, Pradl R, Benes JComparison of the accuracy of hemoglobin point of care testing using HemoCue and GEM Premier 3000 with automated hematology analyzer in emergency room. J Clin Monit Comput. 2016;30:949–956.
17. Maslow A, Bert A, Singh A, Sweeney JPoint-of-care hemoglobin/hematocrit testing: comparison of methodology and technology. J Cardiothorac Vasc Anesth. 2016;30:352–362.
18. Rosenthal RL, Tobias CWMeasurement of the electric resistance of human blood; use in coagulation studies and cell volume determinations. J Lab Clin Med. 1948;33:1110–1122.
19. Yue S, Zhou HM, Zhu JJHyperglycemia and liver ischemia reperfusion injury: a role for the advanced glycation endproduct and its receptor pathway. Am J Transplant. 2015;15:2877–2887.
20. Blasi A, Beltran J, Martin NTransient hyperglycemia during liver transplantation does not affect the early graft function. Ann Hepatol. 2015;14:675–687.
21. Hudcova J, Ruthazer R, Bonney I, Schumann RSodium homeostasis during liver transplantation and correlation with outcomes. Anesth Analg. 2014;119:1420–1428.
Copyright © 2017 International Anesthesia Research Society
22. Gotch F, Torres L, Evans MComparison of conductivity measured hematocrit to microhematocrit. ASAIO Trans. 1991;37:M138–M139.