The Accuracy of Noninvasive Hemoglobin Measurement by Multiwavelength Pulse Oximetry After Cardiac Surgery : Anesthesia & Analgesia

Secondary Logo

Journal Logo

Technology, Computing, and Simulation: Research Reports

The Accuracy of Noninvasive Hemoglobin Measurement by Multiwavelength Pulse Oximetry After Cardiac Surgery

Nguyen, Ba-Vinh MD*,§,¶; Vincent, Jean-Louis MD, PhD; Nowak, Emmanuel PhD; Coat, Michelle MD*; Paleiron, Nicolas MD§; Gouny, Pierre MD; Ould-Ahmed, Mehdi MD§; Guillouet, Maité MD*; Arvieux, Charles Christian MD, PhD*; Gueret, Gildas MD, PhD*,¶

Author Information
doi: 10.1213/ANE.0b013e31822c9679

Perioperative blood loss is common after major surgery. Rapid identification of postoperative bleeding is necessary to limit complications because delay before transfusion can lead to increased morbidity and mortality.1,2 However, unnecessary transfusion exposes patients to certain risks and increases costs.3,4 Evaluation of the degree of blood loss can be challenging and is currently based on vital signs (arterial blood pressure, tachycardia), estimation of apparent bleeding, and hemoglobin (Hb) concentration.

Hb concentration may provide important information to quantify active bleeding and to facilitate decisions regarding the need for blood transfusion. However, measurement of Hb concentration from a blood sample is invasive, labor- and time-consuming, and not continuous. Point-of-care testing is preferable, and the HemoCue photometer was developed for this purpose. This very accurate system is widely used in intensive care units (ICUs) and in the operating room,57 but it still requires a blood sample (albeit small) and can provide only intermittent measures. Continuous, noninvasive Hb monitoring would not only facilitate more rapid decision making, for example regarding blood transfusion, but would also reduce the number of blood samples needed. In March 2008, the Masimo Corp. (Irvine, CA) released an innovative device that uses multiwavelength pulse oximetry for the noninvasive measurement of Hb: the Radical 7 (Rad7). For the Rad7 to be of clinical use, however, it needs to be accurate. Accuracy data have been established in healthy adults and in surgical patients,8,9 but accuracy has not been assessed in critically ill cardiac surgery patients. Critically ill patients after cardiac surgery are at particular risk of acute and major bleeding. Cardiac patients also often have coronary disease or cardiac failure, and are less tolerant of anemia than other patients.10,11 The ability to monitor Hb concentrations noninvasively and continuously could therefore be of particular use in this group of patients, enabling early detection of bleeding.

The aim of this prospective, observational study, therefore, was to evaluate the accuracy of the Rad7 analyzer in the cardiovascular surgical ICU by comparing Hb concentrations obtained with the Rad7 with those obtained from laboratory analysis. Our hypotheses were that the Rad7 analyzer would be sufficiently accurate to be of use in the cardiac surgical ICU and that the updated software version V would be more accurate than the previous V version.


This observational study was conducted in a university hospital (CHU Brest, France). Our local ethical committee gave approval (Clermont Tonnerre Hospital). After informed written consent, all consecutive cardiovascular ICU patients were included during two 1-week periods in January and May 2009.

All patients were monitored with a radial artery catheter. The clip-on pulse oximeter probe was placed at the extremity of the second finger on the same side as the arterial catheter. The results from the Rad7 were compared with the arterial Hb concentration obtained from the automated hematology analyzer, XE-2100 (Roche, Neuilly sur Seine, France), used in our hospital. According to our biological department quality controls, coefficients of variation of repeatability and reproducibility for the XE-2100 are <1% and 1.33%, respectively, giving accuracy for Hb measurement of approximately 0.05 g/dL.

Blood samples were collected via arterial lines in serum tubes with gel (EDTA tube, volume 10 mL; Becton Dickinson Vacutainer®, Le Pont de Claix, France) after a purge of 5 mL and taken to the laboratory immediately after collection. All sample tubes were completely filled and their volumes were constant. They were analyzed once for the purposes of this observational study. When an arterial blood sample needed to be collected for determination of Hb concentration, the Hb concentration and perfusion index shown by the Rad7 were recorded simultaneously. Two software versions of Rad7 were studied. During the first week, version V was used, and for the second week, the software had been updated to version V With the second software version, Rad7 values were recorded only when the device indicated that the values were reliable.

Statistical analyses were performed using the “base,” “psy,” and “boot” packages of R software (R foundation for Statistical Computing, Vienna, Austria; available at, and SAS software, version 9.1 (SAS Institute, Inc., Cary, NC) for mixed models.

Bias was defined as the difference between the Rad7 Hb (Masimo SpHb) and the XE-2100 Hb (laboratory Hb), that is, Masimo SpHb − laboratory Hb. A mixed model was used to calculate bias to consider repeated measurements in the same patient. Intrapatient correlation was estimated with a TOEP2 structure for the covariance matrix.12 Data are presented as mean ± SD. Agreement between the 2 analyzers was assessed using the Bland-Altman method for repeated measures13,14 and an intraclass correlation coefficient for continuous data.15 We used 95% prediction intervals instead of conventional 95% limits of agreement in the Bland-Altman graph because of the small sample size. A 95% prediction interval has a 95% chance of containing the value of a future observation. Correlation analysis was performed using a Fisher transformation.

We also studied Hb bias according to the perfusion index values to determine whether the Rad7 correctly calculated Hb concentration at a low perfusion index. A P value <0.05 was considered as statistically significant. All confidence and prediction intervals were 2-sided.


The study included 41 patients admitted to the 7-bed postcardiac surgery unit: 14 patients for the first software version (age 75 ± 6 years, 7 male, 42 points of comparison) and 27 for the second (age 66 ± 10 years, 21 male, 61 points of comparison). All patients were Caucasian. Patient temperatures were between 36°C and 38°C throughout the study period. All patients received midazolam, sufentanil, paracetamol, nefopam, and morphine, and 40% of the patients received norepinephrine. The first measurement was made while patients were receiving mechanical ventilation; for subsequent measures, patients were breathing spontaneously. Mean Hb concentrations, R2 coefficient, intraclass coefficient, and bias are shown in Table 1.

Table 1:
Hemoglobin Concentration Measured with a New Multiwavelength Pulse Oximeter (SpHb) Using Old and New Versions of the Device's Software, Hemoglobin Concentration Measured Using a Laboratory Hemoglobinometer (Hb), the Squared Coefficient Correlation Between SpHb and Hb (R 2), the Intraclass Coefficient (ICC), the Difference Between SpHb and Hb (Bias), the 95% Prediction Interval (PI), and the Range for the Differences Between SpHb and Hb

The difference between Rad7 and XE-2100 measurements of Hb could vary considerably for different analyses in a single patient with no correlation with Hb values. For example, the bias varied from −6.3 to +1.2 g/dL in 1 patient and from +0.2 to −3.8 g/dL in another. Hence, the difference between the methods for different sampling times in the same patient was not always constant. The intrapatient correlation was 0.27 for V and 0.43 for V The mean bias was not significantly different for the 2 Rad7 software versions (95% confidence interval [CI] for the difference in the bias: −0.4 to 1.3, P = 0.28).

There was a weak correlation between the Hb concentrations measured using the 2 versions of Rad7 and those from the XE-2100 (R2 = 0.11, 95% CI: 0.00–0.33 for the V software; R2 = 0.27, 95% CI: 0.09–0.46 for the V software) (Fig. 1). Figure 2 shows the Bland-Altman plot of bias for the 2 software versions against the reference XE-2100. Figure 3 shows a weak logarithmic relationship in the difference between Hb concentration (SpHb) and Hb and the perfusion index (R2 = 0.33, 95% CI: 0.10–0.56 for the V software; R2 = 0.13, 95% CI: 0.01–0.32 for the V software).

Figure 1:
Correlation between the hemoglobin concentration measured using a Rad7 analyzer (Masimo) (SpHb) and the hemoglobin concentration measured with a laboratory hemoglobinometer (XE-2100; Roche) (Hb). R 2 is the squared correlation coefficient between SpHb and Hb. Black triangles correspond to the measurements made using Rad7 version software. Gray triangles correspond to version software. The dashed line shows the line of identity.
Figure 2:
Bland-Altman plot for repeated measures comparing the hemoglobin concentration measured using a Masimo Rad7 analyzer (SpHb) and the hemoglobin concentration measured with a laboratory hemoglobinometer (XE-2100; Roche) (Hb). Mean hemoglobin is equal to (SpHb + Hb)/2. Black triangles are used for Rad7 V software and gray triangles for Rad7 V software. The lines (black for Rad7 V software and gray for V software) represent 95% prediction intervals for the bias.
Figure 3:
Logarithmic correlation between the perfusion index and the difference in hemoglobin measurement using a Masimo Rad7 analyzer (SpHb) and a laboratory hemoglobinometer (XE-2100; Roche) (Hb) (SpHb − Hb) for the 2 software versions showing that the bias increased when the perfusion index decreased (R 2 = 0.33, P < 0.0001 for the Rad7 V software; R 2 = 0.13, P = 0.004 for the Rad7 V software).

Notably, the difference increased when the perfusion index decreased. For the V software, the average absolute Hb difference was 1.9 g/dL for perfusion index <2 and 0.8 g/dL for perfusion index >2 (P = 0.03). For the V, the average absolute difference was 2.1 g/dL when the perfusion index was <2, and 1.55 g/dL when the perfusion index was >2 (P = 0.26).


Our study demonstrates the poor accuracy of a device that measures Hb noninvasively using multiwavelength pulse oximetry in postcardiac surgery patients. Differences between Hb measured noninvasively by Masimo's Rad7 pulse oximeter and a laboratory hematology analyzer were similar for the 2 software versions, but the intraclass correlation and the correlation coefficient R2 were higher for the newer version ( than for the earlier version (, although these differences were not statistically significant and both software versions were too inaccurate to be of clinical use. The Rad7 was even less reliable when the perfusion index was weak.

Reliable noninvasive tools are desirable to monitor acutely ill patients. The Hb concentration monitoring based on multiwavelength pulse oximetry developed by Masimo has considerable potential applications in anesthesiology and intensive care. However, there are few published clinical data related to this device and no study has evaluated the accuracy of this technique after cardiac surgery. In cardiac surgery, low Hb concentrations are associated with worse outcomes,11,16 so it is important to detect bleeding early, particularly in the postoperative period when hemorrhage may be occult. The Rad7 could potentially enable appropriate management of bleeding to be instituted earlier than with traditional laboratory Hb measures. Early recognition of the need for massive transfusion, combined with aggressive management of hemorrhage by surgical and nonsurgical means, has a significant potential to reduce early mortality.17 Moreover, because blood transfusion may be associated with increased morbidity and mortality,4 the Rad7 could potentially help limit blood transfusions to those patients for whom they are really necessary. The Rad7 could also reduce the need for repeated blood samples, which can contribute to progressive anemia in ICU patients.18,19

Pulse oximetry was designed to measure arterial oxygen saturation. Multiwavelength oximetry was designed to measure SpHb. Skin pigmentation, gender, and the type of oximeter probe all affect pulse oximeter accuracy at low arterial oxygen saturation.20 Because the Rad7 Hb determination is based on the oximetry waveform, if there is error in the pulse oximeter saturation measurement, it is likely that the SpHb determination would also have error. Our results showed a mean absolute difference between SpHb and Hb of 1.2 and 1.8 g/dL for the 2 Rad7 software versions, respectively. Moreover, in patients with hypovolemia secondary to bleeding, a decrease in tissue perfusion, particularly of the skin, can occur with a resultant decrease in tissue perfusion. According to the manufacturer's information, the perfusion index is a relative assessment of the pulse strength at the monitoring site, and as such is an indirect and noninvasive measure of peripheral perfusion. It is calculated by giving the pulsatile signal (during arterial inflow) as a percentage of the nonpulsatile signal, both of which are derived from the amount of infrared (940 nm) light absorbed. The perfusion index is a numerical value that indicates the strength of the “pulsatile component” of the infrared signal returning from the monitoring site. The Masimo infrared signal is influenced primarily by the amount of blood at the monitoring site, not by the level of oxygenation of the blood. An important observation of our study was that the lower the perfusion index, the larger the mean absolute bias between the Rad7 and the central laboratory analyzer. A decrease in perfusion index suggests increased vascular resistance. A hypothermic, bleeding, or shocked patient is at risk of skin vasoconstriction and, therefore, has less tissue perfusion and a low perfusion index. The Rad7 analyzer may not be sufficiently accurate in such cases. In our study, the bias was larger when the perfusion index measured by the Rad7 decreased, limiting the ability of the Rad7 to detect the bleeding. Although the ability of the Rad7 to accurately measure the Hb concentration was better with the second software version, our results still demonstrate insufficient accuracy for this system to be used in clinical practice.

Our findings contrast with those of 2 recent studies by the same authors,8,9 who reported much greater accuracy for the Rad7. The first study,8 published only in abstract form, compared SpHb measurements with invasive laboratory measurements in surgical patients and in healthy adults undergoing normovolemic hemodilution and reported a mean bias of 0.03 g/dL and a precision of 1.12 g/dL. The second study,9 in healthy adults undergoing normovolemic hemodilution, reported a bias of 0.15 g/dL and a precision of 0.92 g/dL. Of note, in this study, some 2.4% of measurements were excluded from the analysis because of device low signal quality. The authors concluded that the accuracy of SpHb was unaffected by perfusion index and that SpHb is an acceptable alternative to invasive Hb tests for many clinical scenarios. The differences between these studies and ours are likely related to the different patient populations, mostly healthy volunteers versus critically ill cardiac surgery patients, and the use of hemodilution to induce anemia. Hence, from these studies, if plasma volume remains constant as a result of fluid infusion, the Rad7 may be accurate. If, however, anemia is associated with hypovolemia, as in our study, the Rad7 gives discrepant results compared with those obtained with the central hematology analyzer. Moreover, by assessing the Rad7 in cardiac surgery patients, our results are more relevant to clinical practice than results from healthy volunteers. In another abstract, Lamhaut et al.21 reported a good correlation between the Masimo reading and the laboratory result in patients undergoing a urologic surgical procedure with risk of hemorrhage. However, the standard deviation was 1.11 g/dL so that the Hb bias in 95% of patients ranged between ±2 g/dL, which is quite high. These authors used only linear regression and not a Bland-Altman analysis. Moreover, the probe and software were different from those used in our study. A recent study by Miller et al.22 reported that absolute differences between SpHb and laboratory Hb were >2.0 g/dL for 22% of patients undergoing spinal surgery, in agreement with our findings. In this study,22 the differences were also reduced when the perfusion index was higher.

We chose to analyze arterial blood samples because we thought that results from arterial blood samples would be more reliable within the same patient than venous samples. Moreover, blood samples could be easily drawn from the arterial lines that are sited routinely in all our cardiac surgery patients. We also wanted to prevent any possibility of “false” anemia induced by hemodilution around venous lines. Nevertheless, it is interesting to note that Evron et al.23 reported no significant differences between the Hb concentrations obtained from various sampling sites (peripheral vein, central vein, and radial artery) or at different times.

One of the limitations of our study is that the blood samples were not run in duplicate. This point will be considered in future studies. However, weekly quality control in our laboratory gives a variability coefficient <1% for Hb measurement.


Although monitoring Hb concentrations in real time has wide applications in anesthesiology and intensive care, the values obtained need to be very accurate before such a system can be used in daily clinical practice. Our study shows poor correlation between Hb measured noninvasively by multiwavelength pulse oximetry and a laboratory hematology analyzer. The difference was greater when the pulse oximetry perfusion index was low, as may occur in shock, hypothermia, or vasoconstriction patients. The accuracy of the multiwavelength pulse oximetry, therefore, needs to be improved before we can recommend its use in the anesthesiology or intensive care setting.


From the *Department of Anesthesiology and Critical Care Medicine, University Hospital, CHU Brest, France; †Department of Intensive Care Medicine, Erasme Hospital, Université libre de Bruxelles, Brussels, Belgium; ‡Inserm CIC 0502, CHU Brest; §Department of Anesthesiology and Critical Care Medicine, Clermont Tonnerre Military Hospital, Brest; ‖Department of Cardiothoracic and Vascular Surgery, University Hospital, CHU Brest; and ¶Université de Brest, Faculté de Médecine et des Sciences de la Santé, Laboratoire de Physiologie, Brest, France.


Ba-Vinh Nguyen, MD, is currently affiliated with the Department of Anesthesiology and Critical Care Medicine, Clermont Tonnerre Military Hospital, Brest, and Université de Brest, Faculté de Médecine et des Sciences de la Santé, Laboratoire de Physiologie, Brest, France.


Name: Ba-Vinh Nguyen, MD.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Attestation: Ba-Vinh Nguyen has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: Jean-Louis Vincent, MD, PhD.

Contribution: This author helped analyze the data and write the manuscript.

Attestation: Jean-Louis Vincent has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: Emmanuel Nowak, PhD.

Contribution: This author helped analyze the data, write the manuscript, and perform statistical analysis.

Attestation: Emmanuel Nowak has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: Michelle Coat, MD.

Contribution: This author helped conduct the study and data collection.

Attestation: Michelle Coat has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: Nicolas Paleiron, MD.

Contribution: This author helped conduct the study and data collection.

Attestation: Nicolas Paleiron has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: Pierre Gouny, MD.

Contribution: This author helped analyze the data.

Attestation: Gouny Pierre has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: Mehdi Ould-Ahmed, MD.

Contribution: This author helped design the study and analyze the data.

Attestation: Mehdi Ould-Ahmed has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: Maité Guillouet, MD.

Contribution: This author helped conduct the study and data collection.

Attestation: Maité Guillouet has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: Charles Christian Arvieux, MD, PhD.

Contribution: This author helped design the study and conduct the study.

Attestation: Charles Christian Arvieux has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: Gildas Gueret, MD, PhD.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Attestation: Gildas Gueret has seen the original study data, reviewed the analysis of the data, approved the final manuscript, and is the author responsible for archiving the study files.


1. Auroy Y, Lienhart A, Péquignot F, Benhamou D. Complications related to blood transfusion in surgical patients: data from the French national survey on anesthesia-related deaths. Transfusion 2007;47:184S–9S
2. Mathew JP, Mackensen GB, Phillips-Bute B, Stafford-Smith M, Podgoreanu MV, Grocott HP, Hill SE, Smith PK, Blumenthal JA, Reves JG, Newman MF. Effects of extreme hemodilution during cardiac surgery on cognitive function in the elderly. Anesthesiology 2007;107:577–84
3. Dunne JR, Malone D, Tracy JK, Gannon C, Napolitano LM. Perioperative anemia: an independent risk factor for infection, mortality, and resource utilization in surgery. J Surg Res 2002;102:237–44
4. Murphy GJ, Reeves BC, Rogers CA, Rizvi SI, Culliford L, Angelini GD. Increased mortality, postoperative morbidity, and cost after red blood cell transfusion in patients having cardiac surgery. Circulation 2007;116:2544–52
5. Mendrone A Jr, Sabino EC, Sampaio L, Neto CA, Schreiber GB, Chamone Dde A, Dorlhiac-Llacer PE. Anemia screening in potential female blood donors: comparison of two different quantitative methods. Transfusion 2009;49:662–8
6. Munoz Gomez M, Naveira E, Romero A, Cordero J, Gonzalez Molina M, Ramirez Ramirez G. Accuracy and reliability of the immediate determination of haemoglobin using the HemoCue B hemoglobin in patients undergoing haemodialysis [in Spanish]. Nefrologia 2004;24:579–82
7. Van de Louw A, Lasserre N, Drouhin F, Thierry S, Lecuyer L, Caen D, Tenaillon A. Reliability of HemoCue in patients with gastrointestinal bleeding. Intensive Care Med 2007;33:355–8
8. Macknet MR, Kimball-Jones PL, Applegate RL, Martin RD, Allard MW. Non-invasive measurement of continuous hemoglobin concentration via pulse CO-oximetry. Anesthesiology 2007;107:A1545
9. Macknet MR, Allard M, Applegate RL, 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
10. Beattie WS, Wijeysundera DN, Karkouti K, McCluskey S, Tait G, Mitsakakis N, Hare GM. Acute surgical anemia influences the cardioprotective effects of beta-blockade: a single-center, propensity-matched cohort study. Anesthesiology 2010;112:25–33
11. Oliver E, Carrio ML, Rodriguez-Castro D, Javierre C, Farrero E, Torrado H, Castells E, Ventura JL. Relationships among haemoglobin level, packed red cell transfusion and clinical outcomes in patients after cardiac surgery. Intensive Care Med 2009;35:1548–55
12. Verbeke G, Molenberghs G. Linear Mixed Models for Longitudinal Data. New-York: Springer, 2000
13. Myles PS, Cui J. Using the Bland-Altman method to measure agreement with repeated measures. Br J Anaesth 2007;99:309–11
14. Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res 1999;8:135–60
15. Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull 1979;86:420–8
16. Murphy GJ, Angelini GD. Indications for blood transfusion in cardiac surgery. Ann Thorac Surg 2006;82:2323–34
17. Perkins JG, Cap AP, Weiss BM, Reid TJ, Bolan CD. Massive transfusion and nonsurgical hemostatic agents. Crit Care Med 2008;36:S325–39
18. Nguyen BV, Bota DP, Melot C, Vincent JL. Time course of hemoglobin concentrations in nonbleeding intensive care unit patients. Crit Care Med 2003;31:406–10
19. Vincent JL, Baron JF, Reinhart K, Gattinoni L, Thijs L, Webb A, Meier-Hellmann A, Nollet G, Peres-Bota D. Anemia and blood transfusion in critically ill patients. JAMA 2002;288:1499–507
20. Feiner JR, Severinghaus JW, Bickler PE. Dark skin decreases the accuracy of pulse oximeters at low oxygen saturation: the effects of oximeter probe type and gender. Anesth Analg 2007;105:S18–23
21. Lamhaut L, Apriotesei M, Lejay M, Vivien B, Carli P. Comparison between a new noninvasive continuous technology of spectrophotometry-based and RBC count for hemoglobin monitoring during surgery with hemorrhagic risk [abstract]. Eur J Anaesthesiol 2010;27:3AP7–1
22. 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
23. Evron S, Tress V, Ezri T, Szmuk P, Landau O, Hendel D, Schechter P, Medalion B. The importance of blood sampling site for determination of hemoglobin and biochemistry values in major abdominal and orthopedic surgery. J Clin Anesth 2007;19:92–6
© 2011 International Anesthesia Research Society