Transfusion of red blood cells remains the first-line therapy for treating hypoxia caused by anemia. Unfortunately, few real-time methods are available to accurately monitor the adequacy of end-organ oxygenation, especially in the anesthetized patient. Therefore, transfusion thresholds, determined by patient comorbidities and evidence of active bleeding, are used to determine when inadequate oxygen delivery may be occurring that requires treatment with red blood cells.1,2 According to the American Society of Anesthesiologists (ASA) 2007 practice guidelines, transfusion is reasonable at a hemoglobin concentration <7 g/dL and likely not indicated at a hemoglobin concentration >10 g/dL.1 Based on a review of the most current evidence, the 2012 AABB (formerly the American Association of Blood Banks) guidelines recommend considering perioperative transfusion at a hemoglobin concentration of <8 g/dL.2 Because of the risks associated with the transfusion of red blood cells and the associated increase in morbidity and mortality in surgical patients who receive intraoperative red blood cell transfusions, the accurate and rapid measurement of hemoglobin concentration would limit exposure to inadequate oxygen delivery while minimizing the exposure to excessive red blood cell transfusion.3–7 Given the narrow range in transfusion thresholds, coupled with the dynamic nature of the operating room where ongoing or anticipated surgical blood loss rapidly changes the need for supplemental red blood cells, it is imperative that transfusion be guided by accurate, objective, and rapid assessment of hemoglobin concentration before the presentation of signs or symptoms of tissue hypoxia.
A multitude of factors affect the accuracy of hemoglobin concentration measurements, including the type of measurement device and the laboratory methodology used.4,7 With the advent of point-of-care and continuous, noninvasive monitors measuring hemoglobin levels, the accuracy and precision of values obtained through these means compared with standard laboratory methods have been called into question, prompting new research efforts.6–10 Many of the current studies have compared noninvasive and continuous pulse co-oximetry and point-of-care analyzers with arterial blood gas (ABG) co-oximetry or central laboratory-determined total hemoglobin concentration via the complete blood count (CBC).4,8–10 As Morey et al.6 indicated in a 2011 editorial, however, the preponderance of hemoglobin concentrations >10 g/dL when comparing the 2 methods may not provide sufficient information to determine whether the device is accurate in the range where transfusion decisions are made (e.g., 7–10 g/dL). Although the difference between the 2 measurement devices may be small for measurements >10 g/dL, if the difference in the performance of the device is large enough when the hemoglobin concentration is between 7 and 10 g/dL, it can inaccurately bias the clinical decision to transfuse or not transfuse red blood cells. Additionally, despite the use of ABG co-oximetry or standard laboratory measurement of hemoglobin as a reference point for the true plasma hemoglobin concentration (CBC), scarce literature evaluates the potential significant difference in the values obtained using these 2 laboratory methodologies.
The aim of this study was to determine the level of agreement of arterial blood hemoglobin measurements obtained from complex spine fusion surgery patients and simultaneously measured by the ABG laboratory co-oximeter and the central laboratory analyzer (CBC). We hypothesized that the 2 techniques would produce measurements that varied more than the stated resolution of 0.1 g/dL, especially in the range of hemoglobin values (e.g., 7–10 g/dL) where most transfusion decisions are made.
After approval from the IRB of Northwestern University Feinberg School of Medicine for waiver of written informed consent, we identified patients who underwent spinal fusion of ≥3 bony levels between September 2006 and September 2010. The data were extracted from a database consisting of clinically available data of patients who underwent complex spine surgery during this time period. Because of the complexity of importing and filtering these data into the respective fields from the clinical records, we limited the extraction to patients in these 4 years.11
Perioperative management of hemodynamic variables and blood product administration was based on the Northwestern High Risk Spine Protocol.11,12 Administration of cell salvage red blood cells (when available) and allogeneic red blood cells was at the discretion of the attending anesthesiologist, with each individual patient’s specific transfusion trigger determined based on his or her cardiovascular status.1,2 Part of the philosophy of this protocol was that frequent surveillance of “transfusion triggers” would allow the anesthesiologist to avoid moderate-to-severe anemia, thrombocytopenia, and coagulopathy in the patient. After measuring the preincision hemoglobin, the protocol calls for the measurement of an ABG with co-oximeter−derived hemoglobin every hour, as well as a central laboratory CBC every 2 hours during the initial portion of the surgery, when there is less likely to be rapid bleeding. During the portions of the operation when there is likely to be either rapid blood loss or continued perturbations in oxygen-carrying capacity as the result of a combination of blood transfusion and continuous bleeding, the protocol calls for measuring simultaneous ABG and CBC hemoglobin concentrations on at least an hourly basis.
The Northwestern Memorial Hospital hematology laboratory uses Beckman Coulter technology via the Coulter LH750 Hematology Analyzer (Beckman Coulter, Inc., Brea, CA) and measures hemoglobin concentration via photocurrent methodology at a set wavelength of 525 nm.a The blood gas laboratory uses the Instrumentation Laboratories GEM Premier4000 analyzer (Instrumentation Laboratories, Bedford, MA) with co-oximetry and optical absorbance measures of total hemoglobin with wavelengths between 480 and 650 nm and the stated measurable range of hemoglobin from 3.0 to 23.0 g/dL with resolution of 0.1 g/dL.b Both the blood gas laboratory and the main laboratory ensure standards of measure consistent with accreditation by the College of American Pathologists.
Statistical Methods—Sample Size Determination
The primary hypothesis of the global IRB was that the intraoperative portion of the Northwestern High Risk Spine Protocol would decrease the number of units of red blood cells transfused compared with the previous clinical care at Northwestern Memorial Hospital.11 Because all patients who underwent complex spine surgery at Northwestern Memorial Hospital after August 31, 2006, received care as part of this hospital-approved clinical pathway, this was a paired availability study. Using the data from Yang et al.,13 if the estimated blood loss of the historical controls were 5800 ± 2472 mL and 1 unit of red blood cells is transfused for every 500 mL of blood loss, then the average blood transfusion for patients undergoing large spinal fusion procedures is 11.6 ± 4.9 units. With these assumptions, 300 patients in each group would result in a 90% power to detect a 500-mL difference in estimated blood loss.
Univariate analysis was performed using NCSS 2009 (Number Cruncher Statistical Systems, Inc., Kaysville, UT), and the limits of agreement analysis was performed with MedCalc Statistical Software (version 12.7.5, MedCalc Software bvba, Ostend, Belgium).
All of the data were tested for normality using the Shapiro-Wilk W test. Normally distributed data are presented as mean ± SD, and these data were compared between groups using a paired t test. Non-normally distributed data are presented as median (95% confidence interval [CI]), and these data were compared using the Wilcoxon signed rank test. Categorical data are presented as number (percent, 95% CI calculated using the Clooper-Pearson method) and compared using Fisher exact test or χ2 test. A 2-sided P < 0.05 was considered statistically significant.
The agreement between the CBC and the ABG was determined using limits of agreement analysis that accounted for repeated measures.14–16 Error grid analysis was performed by plotting the pairs of hemoglobin values against 3 zones of clinical relevance, based on a combination of the relative difference between the simultaneously measured pairs of values and the clinical significance of the observed hemoglobin measurements on the decision to transfuse the patient.6,7 The zones were determined by the most current recommendations of the ASA regarding transfusion thresholds.1
After identifying 484 patients who underwent spinal fusion of ≥3 bony levels, the database was queried to identify those patients who had concurrently measured hemoglobin values by the central laboratory (CBC) and by co-oximetry from the ABG laboratory (Fig. 1). From 348 eligible patients, 1832 pairs of measured hemoglobin values were obtained from contemporaneous blood samples. A median of 5 pairs of measured hemoglobin values was obtained from each patient (99% CI, 1–12.5 pairs of measurements). The median hemoglobin measured in the central laboratory (CBC) was 10.3 g/dL (95% CI, 8.1–12.8 g/dL), whereas the median hemoglobin measured in the ABG laboratory was 10.6 g/dL (95% CI, 8.4–13.2 g/dL). The median difference (ABG-CBC) between measured hemoglobin values was 0.4 g/dL (95% CI, 0.35–0.40 g/dL; P < 0.0001). The magnitude of the difference between the measured hemoglobin values was <0.5 g/dL in 55.5% (95% CI, 53.2%–57.8%) of patients (Table 1); however, 6.8% (95% CI, 5.8%–8.1%) of paired measurements had a difference of >1.0 g/dL. There was good agreement between the ABG and CBC hemoglobin values across the entire dataset (Cohen κ = 0.69; 99% CI, 0.66–0.72; Fig. 2).
Limits of agreement analysis correcting for repeated observations in multiple patients demonstrated that the mean difference between the pairs of measured hemoglobin values (i.e., bias, ABG-CBC) was 0.39 g/dL (95% CI, 0.36–0.41 g/dL; Fig. 3). The 95% limits of agreement of the difference between paired measurements were –0.70 to 1.47 g/dL. There did not appear to be any relationship between the actual hemoglobin values obtained and the bias between measurements (Fig. 3).
Error grid analysis demonstrated that 41 pairs (2.2%) were outside of zone A (Fig. 4, A). All but one of these paired measurements was in zone B. At least 1 measured hemoglobin value from each pair that was outside of zone A was <10 g/dL. There were 765 pairs (41.8%) of measured hemoglobin values where at least 1 of the pair of measured hemoglobin values was between 7 and 10 g/dL (Fig. 4, B). In this region, where the majority of transfusion decisions are made, the median CBC hemoglobin was 9.2 g/dL (95% CI, 7.7–9.9 g/dL), the median ABG hemoglobin was 9.6 g/dL (95% CI, 8.0–10.6 g/dL), and the median difference (ABG-CBC) was 0.3 g/dL (95% CI, 0.25–0.35 g/dL; P < 0.0001). There was only moderate agreement between the CBC and ABG hemoglobin values for all values in this region (Cohen κ = 0.46; 99% CI, 0.41–0.50; Fig. 4, B, gray solid regression line). The bias of this subset of data was 0.38 g/dL (95% CI, 0.34–0.42 g/dL; Fig. 5). The 95% limits of agreement between paired measurements in this subset of data were –0.73 to 1.50 g/dL.
Analyzing the data from simultaneous measurement of arterial hemoglobin with co-oximetry (ABG) and central laboratory analyzer (CBC) during complex spine fusion surgery resulted in several important findings. First, the mean difference between the pairs of measured hemoglobin values (i.e., bias, ABG-CBC) was 0.39 g/dL (95% CI, 0.36–0.41 g/dL), which is consistent with the findings reported in the small study by Giraud et al.8 Second, 44.5% of the pairs of measured hemoglobin values differed by >0.5 g/dL and 6.8% differed by >1.0 g/dL. Third, error grid analysis of the entire dataset demonstrated that 2.2% of the pairs of measured hemoglobin values were outside of zone A, thereby potentially resulting in a different transfusion decision depending on which measurement technique was used. Fourth, when focusing on the 41.8% of pairs of measured hemoglobin values that were between 7 and 10 g/dL, the bias of this subset was 0.38 g/dL with the 95% limits of agreement of –0.7 to 1.50 g/dL. Taken altogether, these data prove our hypothesis that only moderate agreement exists between pairs of measured hemoglobin values when comparing the 2 methods, and with >40% of the values differing by >0.5 g/dL, it is possible for transfusion decisions to be influenced by which of the 2 most commonly accepted methods of measuring hemoglobin is actually used.
Previous studies have compared both venous and arterial samples in healthy volunteers.13,17 The present study used hemoglobin values gathered exclusively from an arterial source from patients undergoing very similar surgical procedures. Consequently, samples were not affected by comorbidities such as peripheral vascular disease or pulmonary dysfunction, as they would be if venous samples were considered.8 Additionally, we used a large sample size with ongoing acute perioperative blood loss as opposed to healthy volunteers, thereby allowing us to examine hemoglobin values <10 g/dL, thus making this dataset more clinically relevant than previously published cohorts, which are often limited by the use of healthy volunteers and a preponderance of hemoglobin values outside of the range of transfusion thresholds (7–10 g/dL).4,6,8,10,18 For example, Giraud et al. used hemoglobin measurements from a variety of surgical procedures considered high risk for bleeding; despite this effort, only 32% of hemoglobin values were <10 g/dL and 2.3% were <8 g/dL. In contrast, the current study included simultaneously measured hemoglobin values with 41.8% (762 samples) in the range of transfusion thresholds.
Importantly, this study does not establish a “gold standard” for perioperative hemoglobin measurement; however, these results confirm that when comparing novel measurement modalities, ABG and CBC values are not interchangeable as a reference point for validating point-of-care technologies.
Given the increasingly recognized morbidity associated with packed red blood cell transfusion, the use of a restrictive approach to transfusion is supported by both the ASA and AABB guidelines.1,2 In this setting, the accuracy and reliability of hemoglobin measurement is paramount because even small bias may prompt unnecessary transfusion. Error grid analysis was used to help depict the clinical significance of the observed difference between ABG and CBC hemoglobin values.6 When considering values <10 g/dL, 40 pairs (2.2% of all data points) were observed in zone B, where the difference between ABG and CBC may incorrectly guide transfusion decisions. Although a difference of 0.4 g/dL when hemoglobin values are greater than 10 g/dL is unlikely to alter clinical course, this difference may have important implications at hemoglobin values closer to 8.0 g/dL. For example, in the setting of active hemorrhage using a transfusion trigger of 8.0 g/dL, one might consider adjusting for the known mean bias of ABG co-oximetry and begin transfusion at values closer to 8.4 g/dL. Additionally, in the operating room, ABG hemoglobin values often are used over the central laboratory because of faster results and simultaneous blood gas analysis, whereas postoperatively, most physicians rely on CBC values. Perioperative physicians caring for patients should be aware of this difference so as not to misinterpret this known bias as ongoing acute blood loss, thereby prompting unnecessary transfusion. Furthermore, although the mean difference across the entire dataset was 0.4 g/dL, in the range of 7 to 10 g/dL, only fair-to-moderate agreement was found between ABG and CBC (Cohen k = 0.39), indicating that hemoglobin values are more weakly correlated in this clinically significant range than across the entire dataset. This finding translates to a relatively greater proportion of data points outside of zone A in the range of hemoglobin values 7 to 10 g/dL, further emphasizing the potential clinical implications of the significant difference between ABG and CBC hemoglobin values.
The current study does not identify a gold standard of hemoglobin measurement. It does, however, highlight the fact that 2 of the most common perioperative hemoglobin measurement modalities, co-oximetry and central laboratory analysis, are not interchangeable. Based on our findings, further studies investigating novel point-of-care and continuous hemoglobin monitoring technology should be consistently compared with either ABG or CBC alone and attempt to focus on hemoglobin values between 7 and 10 g/dL. Additionally, during management of acute blood loss, particularly when implementing a restrictive approach, consideration should be given to the mean difference in hemoglobin values between these 2 commonly used measurement techniques.
Name: Louanne M. Carabini, MD.
Contribution: This author helped with study design, data acquisition, data analysis, and manuscript preparation.
Attestation: Louanne M. Carabini has reviewed the original study data and data analysis, has approved the final manuscript, and is the archival author responsible for maintaining study records.
Name: William J. Navarre, MD.
Contribution: This author helped in data analysis and manuscript preparation.
Attestations William J. Navarre has reviewed the original study data and data analysis and has approved the final manuscript.
Name: Michael L. Ault, MD, FCCP, FCCM.
Contribution: This author helped with study design and manuscript preparation.
Attestation: Michael L. Ault has reviewed the original study data and data analysis and has approved the final manuscript.
Name: John F. Bebawy, MD.
Contribution: This author helped with study design and manuscript preparation.
Attestation: John F. Bebawy has reviewed the original study data and data analysis and has approved the final manuscript.
Name: Dhanesh K. Gupta, MD.
Contribution: This author helped with study design, data analysis, and manuscript preparation.
Attestation: Dhanesh K. Gupta has reviewed the original study data and data analysis and has approved the final manuscript.
This manuscript was handled by: Maxime Cannesson, MD, PhD.
a Operating Principles (pages 3.1–3.22). Coulter LH 750 System Reference and Software Manual. Issue DD, January 2013.
b Measurement Methodology (pages 13–24). Instrumentation Laboratory Reference Guide GEM Premier 4000. May 2012.
1. American Society of Anesthesiologists Task Force on Perioperative Blood Transfusion and Adjuvant Therapies. . Practice guidelines for perioperative blood transfusion and adjuvant therapies: an updated report by the American Society of Anesthesiologists Task Force on Perioperative Blood Transfusion and Adjuvant Therapies. Anesthesiology. 2006;105:198–208
2. Carson JL, Grossman BJ, Kleinman S, Tinmouth AT, Marques MB, Fung MK, Holcomb JB, Illoh O, Kaplan LJ, Katz LM, Rao SV, Roback JD, Shander A, Tobian AA, Weinstein R, Swinton McLaughlin LG, Djulbegovic BClinical Transfusion Medicine Committee of the AABB. . Red blood cell transfusion: a clinical practice guideline from the AABB*. Ann Intern Med. 2012;157:49–58
3. Carabini LM, Ramsey GBarash PG, Cullen BF, Stoelting RK, Cahalan MK, Stock MC, Ortega R. Hemostasis and transfusion medicine. In: Clinical Anesthesia. 20137th ed. New York, NY Lippincott, Williams & Wilkins:408–46
4. Berkow L. Factors affecting hemoglobin measurement. J Clin Monit Comput. 2013;27:499–508
5. Ferraris VA, Davenport DL, Saha SP, Austin PC, Zwischenberger JB. Surgical outcomes and transfusion of minimal amounts of blood in the operating room. Arch Surg. 2012;147:49–55
6. Morey TE, Gravenstein N, Rice MJ. Let’s think clinically instead of mathematically about device accuracy. Anesth Analg. 2011;113:89–91
7. Morey TE, Gravenstein N, Rice MJ. Assessing point-of-care hemoglobin measurement: be careful we don’t bias with bias. Anesth Analg. 2011;113:1289–91
8. Giraud B, Frasca D, Debaene B, Mimoz O. Comparison of haemoglobin measurement methods in the operating theatre. Br J Anaesth. 2013;111:946–54
9. Dewhirst E, Naguib A, Winch P, Rice J, Galantowicz M, McConnell P, Tobias JD. Accuracy of noninvasive and continuous hemoglobin measurement by pulse co-oximetry during preoperative phlebotomy. J Intensive Care Med. 2013;29:238–42
10. Miller RD, Ward TA, Shiboski SC, Cohen NH. A comparison of three methods of hemoglobin monitoring in patients undergoing spine surgery. Anesth Analg. 2011;112:858–63
11. Zeeni C, Carabini LM, Gould RW, Bebawy JF, Hemmer LB, Moreland NC, Koski TR, Koht A, Schafer MF, Ondra SL, Gupta DK. The implementation and efficacy of the Northwestern high risk spine protocol. World Neurosurg. 2014
12. Halpin RJ, Sugrue PA, Gould RW, Kallas PG, Schafer MF, Ondra SL, Koski TR. Standardizing care for high-risk patients in spine surgery: the Northwestern high-risk spine protocol. Spine (Phila Pa 1976). 2010;35:2232–8
13. Yang BP, Ondra SL, Chen LA, Jung HS, Koski TR, Salehi SA. Clinical and radiographic outcomes of thoracic and lumbar pedicle subtraction osteotomy for fixed sagittal imbalance. J Neurosurg Spine. 2006;5:9–17
14. Myles PS, Cui J. Using the Bland-Altman method to measure agreement with repeated measures. Br J Anaesth. 2007;99:309–11
15. Bland JM, Altman DG. Agreement between methods of measurement with multiple observations per individual. J Biopharm Stat. 2007;17:571–82
16. Bland JM, Altman DG. Comparing methods of measurement: why plotting difference against standard method is misleading. Lancet. 1995;346:1085–7
17. Gehring H, Duembgen L, Peterlein M, Hagelberg S, Dibbelt L. Hemoximetry as the “gold standard”? Error assessment based on differences among identical blood gas analyzer devices of five manufacturers. Anesth Analg. 2007;105:S24–30
© 2015 International Anesthesia Research Society
18. Macknet MR, Allard M, Applegate RL 2nd, Rook J. The accuracy of noninvasive and continuous total hemoglobin measurement by pulse co-oximetry in human subjects undergoing hemodilution. Anesth Analg. 2010;111:1424–6