Journal of Clinical Monitoring and Computing

, Volume 27, Issue 1, pp 55–60

Validation of continuous and noninvasive hemoglobin monitoring by pulse CO-oximetry in Japanese surgical patients

  • Tsuyoshi Isosu
  • Shinju Obara
  • Atsuyuki Hosono
  • Satoshi Ohashi
  • Yuko Nakano
  • Tsuyoshi Imaizumi
  • Midori Mogami
  • Masahiro Murakawa
Original Research

DOI: 10.1007/s10877-012-9397-2

Cite this article as:
Isosu, T., Obara, S., Hosono, A. et al. J Clin Monit Comput (2013) 27: 55. doi:10.1007/s10877-012-9397-2

Abstract

We evaluated the accuracy of noninvasive and continuous total hemoglobin (SpHb) monitoring with the Radical-7® Pulse CO-Oximeter in Japanese surgical patients before and after an in vivo adjustment of the first SpHb value to match the first reference value from a satellite laboratory CO-Oximeter. Twenty patients undergoing surgical procedures with general anesthesia were monitored with Pulse CO-Oximetry for SpHb. Laboratory CO-Oximeter values (tHb) were compared to SpHb at the time of the blood draws. Bias, precision, limits of agreement and correlation coefficient of SpHb compared to tHb were calculated before and after SpHb values were adjusted by subtracting the difference between the first SpHb and tHb value from all subsequent SpHb values. Trending of SpHb to tHb and the effect of perfusion index (PI) on the agreement of SpHb to tHb were also analyzed. Ninety-two tHb values were compared to the SpHb. Bias ± 1SD was 0.2 ± 1.5 g/dL before in vivo adjustment and −0.7 ± 1.0 g/dL after in vivo adjustment. Bland–Altman analysis showed limits of agreement of −2.8 to 3.1 g/dL before in vivo adjustment and −2.8 to 1.4 g/dL after in vivo adjustment. The correlation coefficient was 0.76 prior to in vivo adjustment and 0.87 after in vivo adjustment. In patients with adequate perfusion (PI ≥1.4) the correlation coefficient was 0.89. In vivo adjustment of SpHb significantly improved the accuracy in our cohort of Japanese surgical patients. The strongest correlation between SpHb and tHb values was observed in patients with adequate peripheral perfusion suggesting that low perfusion may affect the accuracy of SpHb monitoring.

Keywords

Japanese Pulse CO-oximetry Noninvasive and continuous hemoglobin Perfusion index 

1 Introduction

Previously, total hemoglobin measurement was intermittent and required invasive blood sampling. Noninvasive and continuous hemoglobin measurement with a spectrophotometric sensor may allow care providers to more quickly detect occult bleeding and prevent over-transfusions or anemia due to repeated blood draws. The clinical usefulness of SpHb monitoring has been reported during hemodilution [1], surgery [2, 3] and in the ICU [4] in studies conducted in US and European hospitals but has not been validated in Japanese patients. We studied the agreement of SpHb to a laboratory CO-Oximeter in Japanese surgical patients before and after in vivo adjustment of the SpHb values and also examined the accuracy of SpHb at different peripheral perfusion thresholds, as determined by perfusion index, (PI) in these patients.

2 Methods

This study was approved by the Ethics Committee at Fukushima Medical University and all patients provided written informed consent. Consenting adult patients, who were scheduled for surgical procedures under general anesthesia between February 2009 and June 2010, and required arterial blood sampling during surgery, were enrolled. Patients were excluded if they were under 20 years of age, had finger deformity or hypoperfusion which prevented the proper placement of the SpHb sensor, had hemoglobinopathies such as sickle cell anemia or thalassemia, or were for other reasons determined by the research staff to be unfit for the study.

After each patient was admitted to the operating room, a SpHb sensor (R1 20 or R1 25, revision C), connected to a Radical-7® Pulse CO-Oximeter (Masimo Corp., Irvine, CA, USA) was applied to the third finger of either the right or left hand when feasible (18 of 20 cases). Following induction of general anesthesia, an arterial catheter was placed on the same side of the patient as the SpHb sensor for arterial blood pressure monitoring and blood sampling. When clinically required, blood was drawn from the arterial catheter and analyzed for tHb using a bench top blood gas analyzer (RapidLab® 860, Siemens, Munich, Germany). SpHb at the time of blood draw was recorded for comparison to the tHb value. PI, a measure of peripheral circulation displayed by the Pulse CO-Oximeter, was also recorded at the time of the blood draw. To assess the agreement of SpHb with the laboratory CO-Oximeter, the bias (mean difference of SpHb to tHb), precision (1 SD) and correlation coefficient were calculated. To assess the agreement of SpHb to tHb over the range of observed values, Bland–Altman graphs with limits of agreement for multiple observations per patient during conditions where the true value varies, were plotted [5]. To analyze the ability of SpHb to follow the trend of tHb, a polar plot was created with the distance of data points from the center of the plot representing the mean change in hemoglobin and the angle of the data points to the horizontal axis representing the agreement, as described in Critchley et al. [6]. Good trending was defined as all data points falling with ±0.5 g/dL of the radial axis and acceptable trending was defined as all points falling within ±1 g/dL of the radial axis. To determine if in vivo adjustment of the SpHb values would affect the accuracy, we recalculated the bias, precision, correlation coefficient and limits of agreement after retrospectively subtracting the difference of the first SpHb and tHb pair from all the subsequent SpHb values. To assess the effect of peripheral perfusion on SpHb accuracy in our patients we calculated the bias ± 1SD, limits of agreement and correlation coefficient for all SpHb values associated with PI values ≥1.4 %, compared to all data points as done in a previous study investigating the effect of perfusion on the agreement of SpHb to laboratory CO-Oximeter values in surgical patients [2]. Additionally, a t test was conducted to determine if there was a difference in the bias for data pairs associated with PI < 1.4 and data pairs associated with PI ≥ 1.4. A p value of 0.05 was considered significant.

3 Results

Ten male and 10 female patients with an average age of 59.2 ± 14.1 years and BMI of 23.6 ± 2.8 kg/m2 were enrolled. Patient demographics and procedures are shown in Table 1. Surgical procedures included laparotomy (15), thoracotomy (2), thoracolaparotomy (2) and laparoscopic surgery (1). A total of 92 blood samples were collected and analyzed for comparison to SpHb values. The hemoglobin range from the blood samples was 5.3–13.4 g/dL. Seventy samples were below 10 g/dL, 19 samples were between 10 and 12 g/dL and 3 samples were between 12.1 and 17 g/dL. Bias ± 1SD and average PI was −0.7 ± 1.4 and 0.83 for data pairs associated with PI <1.4 (n = 31) and 0.6 ± 1.4 and 3.44 for data pairs associated with PI ≥1.4 (n = 61). The difference is bias was significant (p = 0.0001).
Table 1

Patient demographics and procedures

Demographics

Age (y)

59.2 ± 14.1

Gender (male/female)

10/10

Weight (kg)

59.2 ± 9.9

Body mass index (kg/m2)

23.6 ± 2.8

Surgery

Laparotomy (n)

15

Thoracotomy

2

Thoracolaparotomy

2

Laparoscopic surgery

1

Data expressed as mean ± 1SD, or number

Bland–Altman analysis showed a bias ±1SD of 0.2 ± 1.5 g/dL, with limits of agreement of −2.8 to 3.1 g/dL from the 92 samples prior to in vivo adjustment (Fig. 1a) and −0.7 ± 1.1 g/dL with limits of agreements of −2.8 to 1.4 g/dL from 71 samples after in vivo adjustment (Fig. 1b). When only samples associated with PI readings of ≥1.4 % were considered, Bland–Altman analysis showed a bias ± 1SD of 0.6 ± 1.4 g/dL, with limits of agreement of −2.2 to 3.3 g/dL prior to in vivo adjustment (61 pairs), (Fig. 2a) and −0.5 ± 0.9 g/dL with limits of agreements of −2.5 to 1.6 g/dL after in vivo adjustment (44 pairs), (Fig. 2b).
Fig. 1

Bland and Altman plots with bias (solid line) and 95 % limits of agreement (dashed line) for noninvasive hemoglobin (SpHb) compared to laboratory hemoglobin (tHb) for 91 data pairs before in vivo adjustment (a) and 71 data pairs after in vivo adjustment (b)

Fig. 2

Bland and Altman plots with bias (solid line) and 95 % limits of agreement (dashed line) for noninvasive hemoglobin (SpHb) compared to laboratory hemoglobin (tHb) for 61 data pairs associated with a perfusion index (PI) ≥1.4 % before in vivo adjustment (a) and 44 data pairs associated with a perfusion index (PI) ≥ 1.4 %, after in vivo adjustment (b)

Linear regression analysis resulted in a correlation coefficient of 0.76 prior to in vivo adjustment and 0.90 after in vivo adjustment. For values associated with PI > 1.4 %, the correlation coefficient was 0.78 before in vivo adjustment and 0.92 after in vivo adjustment (Fig. 3a, b).
Fig. 3

Linear regression plot for noninvasive hemoglobin (SpHb) compared to laboratory hemoglobin (tHb) for 91 data pairs before in vivo adjustment (a) and 71 data pairs after in vivo adjustment (b) with correlation coefficients (R) calculated for all data and data associated with perfusion index values ≤1.4 %. Red triangles represent datapoints associated with perfusion index values ≥1.4 % and blue squares represent datapoints associated with perfusion index values ≤1.4 %

The polar plot showed 100 % of the data points fell within the 1 g/dL boundaries, showing acceptable agreement of SpHb to tHb (Fig. 4).
Fig. 4

Polar plot assessing trending ability of noninvasive hemoglobin (SpHb) compared to laboratory hemoglobin with “acceptable trending limits” of 1 g/dL (dashed line)

When monitoring SpHb with the Radical-7®, if a sufficient signal cannot be obtained, “Low SpHb SIQ” was displayed, and no SpHb value was displayed, alerting the clinician to use other means to determine hemoglobin concentration. There were five times when a blood sample was collected but a SpHb value was not reported by the device.

4 Discussion

Skin pigmentation has been shown to affect the accuracy of noninvasive spectrophotometric measurements such as SpO2 [7]. Other studies conducted in US and European hospitals have shown clinically acceptable accuracy of SpHb but the skin pigmentation or race of the patients was not reported. Using different reference devices for comparison to SpHb, these studies generally showed good agreement of SpHb to tHb when used during surgical procedures and for surveillance monitoring in the ICU. Frasca et al. [4] for example, using a later version of the sensor then used in our study (version E) observed a bias and limits of agreement of 0.0 ± 1.0 g/dL for SpHb when 471 samples from 65 intensive care patients were analyzed with a Sysmex® hematology analyzer. Similarly, Berkow et al. [3] analyzed 186 samples from 29 spine surgery patients with a Radiometer® CO-Oximeter and compared these values to simultaneous SpHb values (sensor version E) and observed a bias of −0.3 ± 1.0 g/dL. Using the same version of sensor as used in our study, Lamhaut et al. [8] compared 85 samples from 44 major urologic surgery patients, analyzed with a Sysmex® hemotology analyzer to SpHb and observed a smaller bias but similar precision as we observed in our patients (−0.02 ± 1.4 g/dL). Some variation in the agreement between SpHb and the reference devices used in these studies can be expected since none (including ours) used the gold standard method for hemoglobin determination, the cyanomethemoglobin assay [9], and analyzers vary in their agreement with the gold standard and their agreement with each other, as has been noted in other studies evaluating SpHb [3, 4]. Our study also demonstrates acceptable trending of SpHb compared to our reference device, a satellite CO-Oximeter, with 100 % of data points falling within 1 g/dL of the a priori boundaries. Since our results in Japanese surgical patients are relatively consistent with those reported in previous studies, we can conclude that it is unlikely that Japanese skin pigmentation affects the accuracy of SpHb.

Data from our cohort of Japanese patients did show the existence of a proportional error however, which was improved by adjusting SpHb by the difference of SpHb to tHb with the first blood draw of each case (in vivo adjustment). In vivo adjustment has received a CE mark and is a commercially available feature on the Radical-7® in some markets. The in vivo adjustment feature when activated allows the clinician to manually adjust the displayed value of SpHb (or other clinical parameters displayed by the Radical-7®) to match that of a corresponding laboratory reference for continuous trending. The positive or negative offset value is displayed along with the adjusted value on the monitor display. The manufacturer recommends confirming the offset value periodically with a reference value and not using the feature when the monitor is displaying a low signal quality message. The in vivo adjustment feature was not available on the Radical-7® devices used in our study so to replicate the effect of this feature we retrospectively adjusted the SpHb value by the difference of SpHb to tHb with the first blood draw of each case. With in vivo adjustment of the SpHb values the bias remained essentially the same but the standard deviation decreased and as such, the limits of agreement were reduced. These results show that in vivo adjustment improved the agreement of SpHb with our reference device, a laboratory CO-Oximeter; although a proportional error was still present.

PI is calculated from the amplitude of the plethysmographic waveform and provides a measure of peripheral circulation. Miller et al. [2] reported that in instances where PI is high, the bias between SpHb and tHb is significantly reduced. Our study confirms that SpHb is affected by peripheral circulation, but the effect was modest. When only data points associated with adequate PI (≥1.4 %) were considered, the correlation coefficient of SpHb to the reference values was just slightly higher than the correlation coefficient from all comparisons (r = 0.78 vs 0.76). SpHb values associated with a PI of ≥1.4 % also had a lower bias ± 1SD compared to all values. Signal quality may be another factor that can affect the agreement of SpHb values to a reference value. The Pulse CO-Oximeter used in this study, the Radical-7®, has a signal quality indicator, called SIQ that flashes on the display when the signal quality for SpHb is compromised. Low signal quality may be due to a misaligned sensor, patient movement, low peripheral perfusion or a combination of factors. A study that evaluated the accuracy of SpHb compared to laboratory CO-Oximetry in spine surgery patients found that data pairs associated with low SIQ had a slightly larger bias but the same standard deviation as data points associated with adequate signal quality (−0.9 ± 1.0 vs −0.1 ± 1.0 g/dL) [3]. We did not separately analyze data points associated with poor signal quality so we cannot speculate on the effect of low signal quality on the accuracy of SpHb compared to the reference in our study. We did observe that SpHb is sometimes not reported by the device during shivering or during peripheral circulation failure due to hemorrhagic shock or prolonged anesthesia.

If SpHb is validated for all types of patients, its use has the potential to confer many advantages to the clinician and the patient. SpHb monitoring should provide more timely data on hemoglobin status compared to hemoglobin measurement by blood sampling and laboratory analysis. Moreover, the ability to monitor hemoglobin continuously may be a major advantage in that it may allow for more timely detection of hemorrhage and guide blood management to prevent over-transfusions. Furthermore, the workload of the anesthesiologist may be lessened as invasive blood draws are reduced. Our study did not include patients with major hemorrhage or those with low SpO2 (such as patients with congenital heart disease). The utility of SpHb monitoring during large hemodynamic shifts and alterations in vascular tone as occurs with major hemorrhage has not been well studied and requires further investigation.

The results of this study showed that SpHb provides a noninvasive means for monitoring hemoglobin concentration with clinically acceptable accuracy in Japanese surgical patients. Low peripheral perfusion appears to have some effect on the accuracy of SpHb however so patients with low perfusion may require more confirmatory laboratory estimations of hemoglobin compared to well perfused patients. In vivo adjustment improved the agreement of SpHb with reference measurements in these patients. In vivo adjustment may represent a significant advance in noninvasive monitoring of hemoglobin as it improved the bias, precision, limit of agreement and correlation coefficient compared to satellite CO-Oximeter measurements in our patients. A prospective analysis of the utility of the commercially available in vivo adjustment feature is needed to confirm our findings.

Copyright information

© Springer Science+Business Media, LLC 2012

Authors and Affiliations

  • Tsuyoshi Isosu
    • 1
  • Shinju Obara
    • 1
  • Atsuyuki Hosono
    • 1
  • Satoshi Ohashi
    • 1
  • Yuko Nakano
    • 1
  • Tsuyoshi Imaizumi
    • 1
  • Midori Mogami
    • 1
  • Masahiro Murakawa
    • 1
  1. 1.Department of AnesthesiologyFukushima Medical University School of MedicineFukushimaJapan

Personalised recommendations