We prospectively compared two point-of-care haemoglobin concentration measuring devices with laboratory measurements to determine their accuracy in women undergoing caesarean section delivery. The two devices were the Masimo Rainbow SET® Radical -7™ pulse co-oximeter and the HemoCue® HB 201+, which is a cuvette-type system that uses photometry. Co-oximeter readings and HemoCue measurements were taken before and after surgery, and compared with laboratory measurements of haemoglobin concentration taken at the same time. We analysed data from 137 patients using Bland–Altman plots. Limits of agreement for co-oximeter readings were −4.20 to 2.02 g.dl−1 and for HemoCue were −1.49 to 1.48 g.dl−1. The bias (mean difference) for the co-oximeter was −1.09 g.dl−1 (95% CI −1.28 to −0.91) and for the HemoCue was −0.001 g.dl−1 (95% CI −0.089 to 0.088). Overall, 110/274 (40%) co-oximeter readings were within 1 g.dl−1 of laboratory values compared with 247/274 (90%) HemoCue measurements (p < 0.001 for difference). The co-oximeter gave lower readings and was less accurate than the HemoCue system when compared with laboratory measurements.
In the UK, the incidence of major obstetric haemorrhage is 3.7 per 1000 live births, and it remains an important cause of morbidity and mortality . In these circumstances, the decision to transfuse a bleeding patient may be life saving, but estimates of blood loss are often inaccurate and some patients are transfused unnecessarily [2–4]. Although there is controversy concerning a specific ‘transfusion trigger’ [5, 6], measurement of haemoglobin concentration combined with patient symptoms and cardiovascular parameters are used to inform the decision-making process.
Laboratory haemoglobin concentration (Hb) measurements are relatively invasive, in that a blood sample is required; they are also relatively time-consuming and whilst waiting for the result the patient’s Hb may have changed dramatically, rendering the measurement less useful. Therefore, point-of-care devices may be more valuable in this setting, especially when haemorrhage is rapid. The Masimo Rainbow SET® Radical-7™ Pulse CO-Oximeter (Masimo Corp, Irvine, CA, USA) is a new non-invasive point-of-care device that continuously records and displays Hb, thereby giving a more dynamic picture. It functions in a similar manner to the pulse oximeter by using the principle of spectrophotometry to calculate haemoglobin concentration. This co-oximeter is self-calibrating with a spectrophotometric sensor consisting of light-emitting diodes and a photodetector that is placed across the measurement site (e.g. the finger). Light received by the photodetector, after passing through the measurement site, generates electrical signals that, when processed by advanced algorithms, provide an estimate of Hb based on its absorbance characteristics. Any optical interference by other light sources is prevented by the use of an optical shield that covers the sensor attached to the finger (Fig. 1).
The HemoCue® HB 201+ (Ängelholm, Sweden) is another point-of-care Hb-measuring device, and requires only a drop of blood. The drop of blood is placed on a cuvette containing a reagent that prompts an azidmethaemoglobin reaction. The cuvette is inserted into a factory-calibrated analyser with a dual wavelength photometer, that displays the Hb in 15–60 s.
The aim of our prospective study was to compare the accuracy of co-oximeter readings and HemoCue photometry measurements with laboratory values taken to be the gold standard. We prospectively studied women undergoing elective caesarean section delivery under regional anaesthesia.
The London-Surrey Borders Research Ethics Committee approved this study. Written, informed consent was obtained from pregnant women aged over 18 years undergoing elective caesarean section delivery at King’s College Hospital in London. Women undergoing emergency caesarean delivery were excluded from the study.
The anaesthetist collecting the data was trained in the use of the co-oximeter and the photometry device, and was not the same anaesthetist providing direct clinical care. The co-oximeter adhesive sensor was placed on the middle or ring finger of the non-dominant hand, and the optical shield applied. Once the readings had stabilised, the displayed Hb was noted as well as the perfusion index, which is the pulsatile signal (the amount of light absorbed by the pulsating blood vessels) indexed against the non-pulsatile signal (the amount of light absorbed by tissues and non-pulsatile blood vessels), and provides a measure of peripheral perfusion. At the same time, a 16-G venous cannula was placed in the opposite hand and 5 ml blood taken. One millilitre was used for the cuvette photometry measurement using the HemoCue, and 4 ml was sent for laboratory Hb measurement using the Advia 2120 (Siemens Medical Solutions and Diagnostics, Dublin, Republic Of Ireland). Intravenous fluid was attached to the cannula and the blood pressure cuff applied to the same arm (Fig. 1). The choice of regional anaesthesia, fluids and any medication administered were at the discretion of the anaesthetist looking after the patient. At the end of the procedure, the co-oximeter reading was recorded again and another 5 ml blood was taken from the arm opposite to the intravenous fluid infusion, for HemoCue and laboratory Hb measurements.
The sample size was calculated on the basis of a paired t-test (two sets of Hb measurements in the same patient). The null hypothesis was that there was no difference between the two measurements and the alternative hypothesis was that the difference between the two measurements was greater than 0.9 g.dl−1. A standard deviation of 1.7 g.dl−1 was assumed based on reported values in existing studies where the mean Hb was 12 g.dl−1. With 80% power and assuming a significance level of 0.05 (two-sided), a minimum of 114 patients were required, assuming a normal distribution of Hb in our study population. With regard to the data analysis, we correlated all Hb measurements and determined the coefficient of variation for either measurement. As both measurements had positive values sufficiently away from zero, dispersion around the mean could be reliably compared in both sets of values. We used levels of agreement as described by Bland and Altman  to evaluate the accuracy of the measurements when compared with the laboratory measurements and used STATA version 8.0 (Statacorp, TX, USA) for all analyses.
One hundred and fifty women were enrolled in the study, but 12 had incomplete sets of data and one patient received a general anaesthetic; these were excluded from our analysis. We therefore analysed data from 137 patients, which resulted in 274 sets of measurements (Table 1). With regard to the co-oximeter, 258 (94%) readings had a perfusion index > 1, which is recommended by the manufacturer.
Table 1. Characteristics of 137 women undergoing caesarean section under spinal anaesthesia. Values are mean (SD), number (proportion) or median (IQR [range]).
LSCS: lower segment caesarean section.
Body mass index; kg.m−2
Fluid infused; ml
1225 (1000–2000 [300–3500])
Estimated blood loss; ml
500 (400–600 [200–2000])
Mean and coefficient of variation for all three methods at both testing points were similar (Table 2). Seventy-three (53%) co-oximeter readings taken before surgery were within 1 g.dl−1 of laboratory Hb measurements compared with 37 (27%) at the end of surgery. This compares with 118 (86%) HemoCue cuvette photometry readings before surgery and 129 (94%) after surgery (p = 0.36). Table 3 shows the proportion of patients within 1 g.dl−1 from the laboratory gold standard for each method before and after surgery, showing that the HemoCue performed better in staying within the clinically tolerable range of variation from the gold standard. Before surgery, the bias and limits of agreement for the co-oximeter readings (−0.62; −3.55 to 2.30) were greater than those for the HemoCue measurements (−0.09; −1.91 to 1.72) g.dl−1, respectively (Fig. 2). This difference was greater after surgery (co-oximeter readings −1.56; −4.6 to 1.46 and HemoCue readings 0.09; −0.94 to 1.12) (Fig. 3)
Table 2. Haemoglobin concentration using co-oximeter or cuvette photometry (HemoCue) compared with laboratory measurements in 137 obstetric patients before and after surgery. Values are mean (SD).
*Difference from mean laboratory value p = 0.0011.
†Difference from mean laboratory value p < 0.001.
Before surgery; g.dl−1
After surgery; g.dl−1
Table 3. Distribution of co-oximetry and cuvette photometry (HemoCue) by magnitude of difference from laboratory haemoglobin concentration measurements in 137 obstetric patients before and after surgery. Values are number (proportion).
Difference from laboratory value
*p < 0.001.
†p = 0.031.
We have shown in this prospective study that the co-oximetry reading varied significantly compared with both the HemoCue cuvette-based photometry test and the gold-standard laboratory test, which in themselves gave similar results. This implies that the cuvette test, but not the co-oximetry test, can be recommended for use in obstetric patients to help guide blood transfusion.
Our findings are in contrast to a previous study in healthy volunteers that showed a good correlation of co-oximeter Hb with laboratory measurements, with an accuracy of up to 1 g.dl−1 . The differences seen in our study may be due to the patients’ being pregnant, receiving a spinal anaesthetic or requiring oxytocic and vasopressor administration, potentially changing finger perfusion upon which the co-oximeter relies. A recent study comparing co-oximeter with laboratory measurements in 50 women undergoing caesarean delivery found similar results to ours with respect to wide limits of agreement , but demonstrated a positive bias in contrast to our study, which shows negative bias. Butwick et al. demonstrated varying performance for different time points throughout the peri-operative period, which we also observed in terms of greater bias after surgery compared with before. This implies that the performance of the co-oximeter may be affected by shifts in intravascular volume. Indeed, studies in post-cardiac surgery patients  and patients undergoing abdominal and pelvic surgery  also show worsening performance with changes in intravascular volume, especially during haemorrhage , which is when a non-invasive, real-time Hb measurement would be most useful.
There is much debate about the accuracy of the HemoCue cuvette-based measurement system. This centres mainly on the sampling site (capillary versus venous or arterial blood) [12–15]. For consistency and patient comfort, we chose to use venous blood for our measurements. In our study, only 9.9% of HemoCue cuvette photometry values were more than 1 g.dl−1 above or below laboratory values, which is similar to the results in a study of blood donors (9%) . Our limits of agreement, however, were wider than those in patients undergoing aortic surgery  and in pregnant women at term , despite our larger sample size.
When comparing the co-oximeter and the HemoCue with laboratory Hb measurements, the poor correlation we observed with the co-oximeter compared with the HemoCue has also been demonstrated when testing arterial blood during spinal surgery  and capillary blood in major urological surgery . However, another study in critically ill patients showed better correlation with the co-oximeter compared with the HemoCue . As none of these patients were having surgery at the time the observations were taken, this adds to our theory that intravascular volume shifts affect the performance of the co-oximeter, as opposed to the vasoconstrictor noradrenaline, which did not seem to affect it .
Limitations of this study include our not always placing the co-oximeter on the non-dominant hand as recommended by the manufacturers (as the dominant hand may have hardened skin that would interfere with function). In addition, blood was sampled at some distance from the co-oximeter probe, and Hb differences from different sides of the body have been noted in other studies . However, venous blood was always used and this was never taken from the arm used for intravenous fluid administration. Additionally, we used a laboratory haematology analyser as the reference method and although this method has been shown to have acceptable accuracy for Hb measurement , it may not be as accurate as the cyanomethemoglobin assay . The researcher in our study collected the data from both the co-oximeter and the HemoCue, and thus was not blinded from the different results.
In summary, although the co-oximeter was easy to use, we found that it was not accurate enough for clinical decision-making regarding transfusion. In contrast, the HemoCue cuvette system was more accurate than the co-oximeter in comparison with laboratory measurements. We therefore recommend that, in spite of its somewhat more invasive nature, by requirement for a small amount of blood, the HemoCue be used in the setting of obstetric anaesthesia and haemorrhage to measure Hb.
All Masimo Rainbow SET® Radical-7 Pulse CO-Oximeters, optical shields, the monitor and software were provided by the Masimo Corporation, on loan, for the purpose of this study. We thank Valerie Begnoche from the Masimo Corporation for her advice regarding the manuscript. The authors have no conflict of interest to declare.