Monitoring hemoglobin levels in the operating room currently requires repeated blood draws and involves several steps. Because of the variable lag time to receive the laboratory results, blood transfusion management decisions may be delayed or made before hemoglobin results become available. The ability to measure hemoglobin continuously and noninvasively may allow for a more rapid assessment of a patient's condition, enabling prompt and appropriate blood management.
The development of spectrophotometric methods for the noninvasive measurement of blood constituents such as total hemoglobin (tHb) has been a highly desired yet largely unachieved goal of medical bioengineering. Although prototype technologies for the noninvasive measurement of tHb have been described, none had achieved Food and Drug Administration clearance and become commercially available until the introduction of the Pulse CO-Oximeter, which can be used for continuous hemoglobin (SpHb) monitoring.1,2 The methodology of hemoglobin measurement by Pulse CO-Oximetry is similar to that used for oxyhemoglobin estimation by conventional pulse oximetry, except that instead of 2 wavelengths of light, multiple wavelengths are transmitted through the finger to measure the light absorbance characteristics of the analyte, tHb. The exact number of wavelengths and the specific wavelengths used are proprietary and are not available from the manufacturer. Recent studies compared the accuracy of SpHb with that of laboratory CO-Oximetry in healthy subjects undergoing hemodilution3 and in surgical and intensive care unit patients.4,5 Additionally, the use of SpHb has been shown to reduce blood transfusion frequency in orthopedic surgery patients.6 The intent of this study was to expand on those findings by testing the accuracy of SpHb in surgical patients undergoing complex procedures. Specifically, we performed a prospective, observational study to compare the accuracy of SpHb with that of laboratory CO-Oximetry tHb measurements in patients who were undergoing complex spine surgery and were at high risk for blood loss.
After receiving IRB approval and written informed consent, we enrolled a convenience sample of 29 patients who were scheduled to undergo complex spine surgery at a large academic medical center (Johns Hopkins Medical Institution, Baltimore, MD) over a 5-month period (May–September, 2010). A spine procedure was defined as complex if multiple spinal levels were involved, the estimated surgical time was >4 hours, and blood products were requested for the procedure. Only patients 18 years and older were included. Exclusion criteria included (a) no anticipated blood loss, and (b) no planned invasive arterial or central venous monitoring.
All patients were monitored with ASA standard monitors plus invasive arterial monitoring and/or central venous monitoring as part of their standard care. Patient demographics (age, gender, ASA physical status) and surgical variables (surgery type, procedure duration, estimated blood loss, total volume IV fluids given, and total volume of packed red blood cells transfused) were recorded. For noninvasive hemoglobin monitoring, Rainbow adult ReSposable™ sensors (rev E) were attached to the ring or middle fingertip of the subject and then connected to a Radical-7 Pulse CO-Oximeter (SET version 7601; Masimo Corporation, Irvine, CA). The sensors were covered with black plastic shields to prevent optical interference. The Pulse CO-Oximeter was connected to a laptop computer with an automated data collection software program (PhysioLog; Masimo Corporation) for the noninvasive recording of hemoglobin (SpHb) and SIQ (a signal-quality indicator), and to log the exact time of the start of each blood draw to be used for comparison to SpHb. The Radical-7 sampling frequency was 62.5 Hz. Data were then exported into Microsoft Excel (Microsoft, Redmond, WA) for statistical analysis. Throughout each surgery, blood samples were obtained hourly (or more often if clinically indicated) by withdrawing 3 mL of blood from an indwelling arterial or central venous catheter into heparinized blood gas syringes. Samples were analyzed for tHb by the central laboratory with a Radiometer ABL800 CO-Oximeter (Radiometer America, Westlake, OH). This method of hemoglobin determination has been shown to have “good” correlation with the international standard for hemoglobin measurement, the cyanomethemoglobin (HiCN) assay.7 The bias of the ABL800 is reported to be between 0.04 and 0.37 g/dL compared with the HiCN assay when hemoglobin values are in the range of 11.3 to 24.2 g/dL.8 The bias may be different for the range of hemoglobin values reported in this study (6.9–13.9 g/dL). The laboratory analyzer was calibrated and subjected to daily quality-control testing per the manufacturer's specifications by the central laboratory. The Pulse CO-Oximeter is self-calibrating and requires no routine user calibration. Laboratory tHb measurements were compared with SpHb values obtained at the time of the blood draw by calculating the bias and precision for the data pairs. Bias was defined as the average difference between the SpHb and tHb pairs, and precision was defined as 1 SD of the bias.
The Pulse CO-Oximeter has a signal-quality indicator for the SpHb value (SIQ) that provides the user with an indication of the reliability of the SpHb reading. According to the manufacturer, low SIQ readings are subject to more variability and may not be as reliable; therefore, these values were analyzed separately. For the purposes of this study, low SIQ was defined as an SIQ value of <50% as recorded by the data logger. This threshold of <50% is used by the manufacturer to trigger the low SIQ indicator light on the device.
To compare SpHb with tHb over the range of values, we constructed Bland-Altman graphs,9 which are used to assess agreement between 2 methods of measurement that entail multiple observations and for which the true value varies. We used this statistical method to estimate the bias and limits of agreement for all data and for data pairs associated with adequate SIQ. The limits of agreement define the range within which 95% of the difference between the 2 methods of measurement lies and is equal to the average difference ±1.96 times the standard deviation of the difference. The standard deviation is the square root of the total variance, which is estimated to be the sum of the variance that results from repeated differences between the 2 methods on the same subject and the variance that results from differences between the averages of the 2 methods across subjects. Additionally, the mean percent errors of SpHb measurements compared with tHb measurements in 4 ranges of tHb values (for all data pairs and separately for data pairs associated with SIQ values of >50%) were calculated by dividing 2 SDs of the bias of SpHb to tHb by the mean of the range of tHb values. A mean percent error of 20% and a bias of 1.5 g/dL were considered clinically acceptable for continuous measurement of hemoglobin. A bias of 1.5 g/dL compared with a reference device was posited to be clinically acceptable in another recent study that assessed SpHb in surgical patients.4
To assess the ability of SpHb monitoring to follow the trend of tHb measurements, we compared the magnitude of changes in tHb between consecutive measurements that were larger than 1.0 g/dL with the magnitude change in the time-matched, consecutive SpHb values. Consecutive changes in tHb that were <1.0 g/dL were considered clinically insignificant and were not analyzed.
Lastly, to assess overall reliability of SpHb monitoring during surgery, we calculated the percent monitoring time that the SpHb signal was associated with low SIQ and the percent monitoring time that the SpHb was not displayed (signal dropout rate). The median perfusion index, a relative measure of the pulse strength at the monitoring site that corresponds to peripheral perfusion, was also calculated during the low SIQ and signal dropout periods.
Thirty-three patients were enrolled in the study. Two patients were excluded by request from the IRB because they had been enrolled before an adjustment to the consent process that clarified that any venous laboratory results were research tests as opposed to routine clinical care laboratory tests. One patient was withdrawn from the study because of canceled surgery. One patient received no blood draws during surgery and was therefore excluded from the analysis. The remaining 29 patients (15 men, 14 women) that composed the study group underwent lumbosacral fusion (14), thoracolumbar fusion (6), thoracopelvic fusion (6), and sacrectomy (3). Estimated blood loss ranged from 100 to 4500 mL, with a median of 1800 mL. Twenty-four of the 29 patients received transfusion of packed red blood cells. Intraoperative transfusions ranged from 0 to 25 U of packed red blood cells, with a median of 4 U. Surgery duration ranged from 144 to 537 minutes, with a median duration of 342 minutes, and the monitoring time ranged from 146 to 721 minutes, with a median duration of 392 minutes.
In total, 186 blood samples were drawn and evaluated for tHb. These values were paired with an equal number of SpHb measurements taken from the continuous record of SpHb measurements recorded by the data collection software. The SpHb measurement used for comparison to the laboratory value was the single value recorded at the exact time of the start of the blood draw. A median of 6 samples was collected per patient. tHb values ranged from 6.9 to 13.9 g/dL, and SpHb measurements ranged from 6.9 to 13.4 g/dL. Mean bias and precision of 186 SpHb measurements compared with the reference device measurements were −0.3 g/dL and 1.0 g/dL, respectively. Fifty-six data pairs included SpHb values that were associated with a low SIQ message displayed on the device and were analyzed separately. The bias and precision of the data pairs that were associated with a low SIQ display were −0.9 and 1.0 g/dL, respectively. After exclusion of the low SIQ readings of <50% (as recorded by the data collection software), the bias and precision of the remaining 130 data pairs were −0.1 and 1.0 g/dL, respectively. Table 1 shows the mean percent error and the number of SpHb measurements with a given range of bias for 4 tHb ranges. These data are shown for all data pairs (Table 1A) and for the SpHb values associated with SIQ >50% (Table 1B). The mean percent error for all hemoglobin ranges remained essentially the same when low SIQ values were removed except for that of SpHb values <10 g/dL, which decreased from 19.0% to 11.7%.
Accuracy over the continuous range of values was analyzed by using Bland-Altman plots, which showed limits of agreement of −2.4 to 1.7 g/dL for all data pairs (Fig. 1) and limits of agreement of −2.0 to 1.8 g/dL for the 130 data pairs associated with adequate (≥50%) SIQ values (Fig. 2).
Table 2 shows the ability of SpHb to follow the trend of tHb values. Each time that tHb decreased by >1.5 g/dL between consecutive measurements (6 times), SpHb values also decreased. However, SpHb decreased by >1.5 g/dL only 1 of the 6 times and decreased by less than that amount the other 5 times. On all 5 occasions that tHb values increased by >1.5 g/dL between consecutive measurements, SpHb also increased. SpHb increased by >1.5 g/dL in 2 of these cases and by <1.5 g/dL in the remaining 3 cases.
During the monitoring time, there were periods when the low SIQ indicator was present and there were occasional and brief losses of the SpHb signal (device reported no value instead of a dash or a flashing value as occurs with low SIQ). The low SIQ indicator was present in 20 of the 29 cases for a median of 7% of the time (upper and lower quartiles of 43% and 0%). The median perfusion index during low SIQ periods was approximately 0.5%, indicating that low peripheral perfusion may contribute to low SIQ values.
Episodes of SpHb signal loss occurred in 19 of the 29 cases. The median percent time of signal loss was 1%, with upper and lower quartiles of 3% and 0%. The median perfusion index during the dropout period was approximately 2%, with upper and lower quartiles of 2.7% and 0.5%, indicating that factors other than perfusion may have contributed to signal loss.
Figure 3 shows a representative graph of SpHb and tHb values plotted against time during a posterior fusion surgery in a 69-year-old female patient. Figure 4 shows a representative graph of SpHb and tHb values plotted against time during a thoracopelvic fixation and fusion in a 61-year-old female patient who received transfusions at 2 time points.
In this study, we compared the accuracy of continuous noninvasive hemoglobin measurement with that of intermittent hemoglobin values obtained by blood sample analysis with laboratory CO-Oximetry in patients undergoing complex surgery of the spine. We found that SpHb measurements had an average bias of −0.3 g/dL for all data pairs, and −0.1 g/dL when low SIQ data pairs were removed. Bland-Altman analysis showed limits of agreement of approximately ±2 g/dL. Other types of statistical analysis may also be appropriate for these data,10 but we chose to limit the analysis to those methods most frequently used to enable this study to be easily compared with others that have evaluated this technology.
Our results are similar to the bias of −0.15 g/dL and precision of 0.92 g/dL reported by Macknet et al.,3 who tested the accuracy of SpHb in healthy subjects undergoing hemodilution. They are also similar to the bias and precision of 0.0 ± 1.0 g/dL reported by Frasca et al.,5 who evaluated SpHb in intensive care unit patients. The findings of other studies that have investigated the accuracy of SpHb were not consistent with our own. Miller et al.4 investigated the accuracy of SpHb in 20 patients undergoing spine surgery. The investigators concluded that although SpHb often correlated well with tHb values, differences were >1.5 g/dL in 39% of their observations. Another study by Gayat et al.11 showed a substantially larger bias and precision (1.8 ± 2.6 g/dL) when SpHb values were compared with measurements made with a Siemens ADVIA hematology analyzer in 300 emergency room patients. None of the above-mentioned studies compared SpHb with the “gold standard,” HiCN. There are several methodological differences between our study and that by Gayat et al. that may account for the differences in reported accuracy. First, Gayat et al. used an earlier version of the SpHb sensor. Second, the Radical-7 was used as a “spot check” device, not for continuous monitoring as was done in our study and is the manufacturer's indication for use. Additionally, Gayat et al. did not separately analyze SpHb values associated with low SIQ, which, depending on how many of those values were included in their data set, could significantly affect the reported accuracy.
SIQ, the manufacturer's signal-quality indicator, is a continuous indication of the SpHb signal quality and thus provides some degree of measurement reliability. Low SIQ values may be caused by improper placement of the sensor, motion or other types of environmental interference, decreased peripheral perfusion, or a combination of factors. The precise algorithm used to determine SIQ is proprietary to the manufacturer and has not been disclosed. When the data points associated with low SpHb SIQ values were removed, the precision of SpHb compared with the reference device remained the same at 1.0 g/dL, but the bias improved slightly from −0.3 to −0.1 g/dL, indicating that SIQ may provide valuable information regarding the reliability of the SpHb values reported by the device. The mean percent error decreased from approximately 19% to 12% for SpHb values <10 g/dL but stayed essentially the same for the hemoglobin ranges >10 g/dL. Fifty-two percent of SpHb values <10 g/dL were associated with low SIQ, whereas only 27% of values between 10.0 and 11.9 g/dL and 15% of values between 12.0 and 13.9 g/dL were associated with low SIQ.
We found that the SpHb data stream would sometimes drop out (device reported dashes instead of a hemoglobin value) when a patient was being repositioned or during other interventions that caused movement at the sensor site, such as motor-evoked potential–stimulated hand movements. This observation is consistent with the manufacturer's instructions that patient motion may interfere with device performance. The median percent of time that the device did not display a value was approximately 1%, and approximately half the cases had no dropouts. The brief dropout periods are unlikely to affect perioperative transfusion decision-making because even with brief lapses in the continuous SpHb measurement, the device provides dramatically more data with constant updates on the patient's hemoglobin status than is possible with intermittent blood sampling.
The finding that the accuracy of SpHb measurements was clinically acceptable according to study criteria may have implications for the management of intraoperative blood loss and transfusion requirements in patients undergoing complex spine surgery. One of the principal complications associated with complex spine surgery is the considerable loss of blood during the procedure. During spinal fusion, it is not uncommon for patients to sustain perioperative blood loss that approaches their total estimated blood volume.12,13 For this reason, much concern has been focused on reducing the number of blood units administered during this type of surgery through advancements in surgical and anesthetic techniques, as well as via the administration of antifibrinolytic medications.14 The delay in receiving laboratory hemoglobin values may lead to the unnecessary transfusion of blood products if concerns arise that the values are low; this unnecessary use of blood products could potentially be avoided by the use of continuous hemoglobin monitoring.
Although this was an observational study that did not evaluate the effect of SpHb monitoring on blood management, we believe that continuous access to hemoglobin status during surgery may provide clinicians with the information required to more closely manage blood loss and titrate transfusions. The knowledge gained has the potential to reduce the risk of over- or undertransfusion. Our institution uses a spine surgery protocol that recommends transfusion to maintain a tHb ≥10 g/dL, and the availability of SpHb monitoring for these cases could improve compliance with this protocol. A recent randomized trial of 327 subjects supports this hypothesis. Blood transfusion frequency was reduced by 86% in the group monitored by SpHb, with no differences between groups in the amount of blood transfused in the 12-hour postoperative period or in the 30-day complication rates.6
Aside from clinically acceptable accuracy, our study also demonstrates that in our patients, SpHb trended with changes in tHb. Because there were few large excursions in hemoglobin concentration in our patients (average change in hemoglobin was <3 g/dL), additional studies are required to fully assess the trending ability of SpHb.
One limitation of the study is that it did not include cases with massive blood loss or with large numbers of data points at which intraoperative hemoglobin was very low (<8 g/dL) or in the range of potential transfusion (<10 g/dL). Our institution uses a spine surgery protocol that recommends blood transfusion to maintain hemoglobin levels of approximately 10 g/dL intraoperatively. This practice probably explains the lack of patients with low intraoperative hemoglobin values in our study. The accuracy of SpHb readings in anemic patients may differ from what was demonstrated in our sample. Another limitation of our study is that we used laboratory CO-Oximetry as the reference device instead of the international standard method for determining hemoglobin, the HiCN assay.7,15 The HiCN assay is time- and labor-intensive and requires the use of hazardous reagents; therefore, it was not deemed practical for use in our study.
Besides laboratory CO-Oximetry, several other methods of measuring patient hemoglobin are frequently used in hospitals and other health care settings, including spun hematocrit, Coulter counter, and spectrophotometric point-of-care devices. The accuracy of each of these methodologies is affected by multiple variables, including device calibration, sample handling, and other sources of variation specific to the technology. Gehring et al.16 evaluated both inter- and intradevice variation in hemoglobin measurements of 5 different models of CO-Oximeters. When the same blood sample was analyzed on 2 identical devices to test intradevice variability, the standard deviation between measurements ranged from 0.2 to 1.2 g/dL. In a separate evaluation of interdevice variability, Torp et al.17 analyzed 471 arterial blood samples from liver transplant patients by both laboratory CO-Oximetry and a Coulter analyzer and found that mean Coulter measurements were approximately 1 g/dL lower than the CO-Oximeter values (P < 0.001). Hemoglobin values from point-of-care devices have also been shown to have significant variability when compared with those from a Coulter analyzer. Patel et al.18 measured hemoglobin in patients undergoing open heart surgery using 8 different point-of-care analyzers and a Coulter analyzer as the reference device. The mean differences (±SD) between the point-of-care test devices and the reference test method ranged from 0.0 ± 0.2 to 0.8 ± 0.4 g/dL. Point-of-care hemoglobin analysis from capillary samples has been shown to have even greater variability than that from arterial or venous samples.19 When the accuracy of a new method of hemoglobin measurement is evaluated, the variation in values obtained from existing, accepted methods of measurement and the inherent intradevice variation should be considered.
An additional limitation to the study is that it did not address the economic aspects of using SpHb monitoring in place of or in addition to analysis of blood samples by the central laboratory. A cost comparison of Pulse CO-Oximetry and laboratory analysis of hemoglobin may be a valuable avenue of investigation for a future study.
The SpHb measurements from Pulse CO-Oximetry demonstrated clinically acceptable accuracy of hemoglobin measurement with a bias and precision of −0.1 ± 1.0 g/dL during complex spine surgery when compared with measurements obtained from a standard laboratory reference device. This technology provides more timely and complete information on hemoglobin concentration than does the current standard of intermittent blood sample analysis and thus has the potential to improve blood management during surgery.
Name: Lauren Berkow, MD.
Conflicts of Interest: Consultant for Masimo Corporation, study funded by Masimo.
Contribution: Study design, conduct of study, data analysis, manuscript preparation.
Name: Stephanie Rotolo.
Conflicts of Interest: Research assistant salary supported by Masimo.
Contribution: Data collection.
Name: Erin Mirski.
Conflicts of Interest: Research assistant salary supported by Masimo.
Contribution: Data collection.
This manuscript was handled by: Dwayne R. Westenskow, PhD.
1. McMurdy JW, Jay GD, Suner S, Crawford G. Noninvasive optical, electrical, and acoustic methods of total hemoglobin determination. Clin Chem 2008;54:264–72
2. Rabe H, Alvarez RF, Whitfield T, Lawson F, Jungmann H. Spectroscopic noninvasive measurement of hemoglobin compared with capillary and venous values in neonates. Neonatology 2010;98:1–5
3. Macknet MR, Allard M, Applegate RL II, 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
4. 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
5. Frasca D, Dahyot-Fizelier C, Catherine K, Levrat Q, Debaene B, Mimoz O. Accuracy of a continuous noninvasive hemoglobin monitor in intensive care unit patients. Crit Care Med 2011 . [Epub ahead of print]
7. Recommendations for reference method for haemoglobinometry in human blood (ICSH standard 1986) and specifications for international haemoglobincyanide reference preparation (3rd edition). International Committee for Standardization in Haematology; Expert Panel on Haemoglobinometry. Clin Lab Haematol 1987;9:73–9
8. ABL 800 Flex Reference Manual. Bronshoj, Denmark: Radiometer Medical ApS, 2004
9. Bland JM, Altman DG. Agreement between methods of measurement with multiple observations per individual. J Biopharm Stat 2007;17:571–82
10. Morey TE, Gravenstein N, Rice MJ. Let's think clinically instead of mathematically about device accuracy. Anesth Analg 2011;113:89–91
11. Gayat E, Bodin A, Sportiello C, Boisson M, Dreyfus JF, Mathieu E, Fischler M. Performance evaluation of a noninvasive hemoglobin monitoring device. Ann Emerg Med 2011;57:330–3
12. Guay J, Haig M, Lortie L, Guertin MC, Poitras B. Predicting blood loss in surgery for idiopathic scoliosis. Can J Anaesth 1994;41:775–81
13. Guay J, Reinberg C, Poitras B, David M, Mathews S, Lortie L, Rivard GE. A trial of desmopressin to reduce blood loss in patients undergoing spinal fusion for idiopathic scoliosis. Anesth Analg 1992;75:405–10
14. Elgafy H, Bransford RJ, McGuire RA, Dettori JR, Fischer D. Blood loss in major spine surgery: are there effective measures to decrease massive hemorrhage in major spine fusion surgery? Spine 1976;35:S47–56
15. International Committee for Standardization in Haematology. Recommendations for haemoglobinometry in human blood. Br J Haematol 1967;13:71–5
16. 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
17. Torp KD, Shine TSJ, Aniskevich S, Peiris P, Shapiro DP. Comparison of Coulter counter and CO-Oximeter (pHOx) measurements for hemoglobin. Proceedings of the 2010 Annual Meeting of the American Society of Anesthesiologists. Abstract A937. Available at: http://www.asaabstracts.com/strands/asaabstracts/search.htm
18. Patel KP, Hay GW, Cheteri MK, Holt DW. Hemoglobin test result variability and cost analysis of eight different analyzers during open heart surgery. J Extra Corpor Technol 2007;39:10–7
© 2011 International Anesthesia Research Society
19. Yang ZW, Yang SH, Chen L, Qu J, Zhu J, Tang Z. Comparison of blood counts in venous, fingertip and arterial blood and their measurement variation. Clin Lab Haematol 2001;23:155–9