1.

Introduction

Patients presenting hemodynamic instability should be aggressively managed without delay.^1,2 This often involves the use of invasive monitoring tools, such as arterial lines or central venous catheters, in order to monitor tissue perfusion. Although they provide useful information, these tools are often unavailable and require time to install. During resuscitation, the utility of laboratory values is blunted by the fact that they provide static information on a continuously evolving clinical process. The time required to access these values can result in outdated information that provides little real-time feedback on the efficacy of the resuscitation. Near–infrared spectroscopy (NIRS) is a noninvasive, continuous and painless method of monitoring the oxygen saturation of hemoglobin in any given superficial tissue (tissue saturation) and, as such, can be used as a way of detecting tissue hypoperfusion. Unlike pulse oximetry, which only measures the arterial saturation of hemoglobin, NIRS measures the saturation in all vessels smaller than 1 mm (arterioles, capillaries, and venules).³ NIRS devices can also provide saturation measures in situations of circulatory arrest, as they do not depend on pulsation to provide its readings. Tissue perfusion (and saturation) is affected in situations of hemodynamic instability.^4–6 Thus, the quantification of tissue saturation by NIRS oximetry could prove to be an interesting surrogate measure for tissue perfusion and a useful tool in guiding resuscitation in settings where more invasive tools are unavailable such as when out of the hospital or in the emergency room. They can also prove to be an interesting addition to other specialities of critical care, such as anaesthesiology and the intensive care.

NIRS technology was developed after an experiment by Jöbsis, which showed that light spectrum could be used as a way to monitor patients. This is based on two unique properties. The first is that near–infrared photons (i.e., photons with a 700 to 1300 nm wavelength) have the capacity to go through organic tissues without being significantly absorbed. The second is that oxyhemoglobin and deoxyhemoglobin have different absorptions properties along this wavelength spectrum, allowing them to be differentiated.⁷ Using the Beer-Lambert law and an algorithm, NIRS devices calculate the total concentration of both these molecules, thus providing a value representing tissue saturation in oxygen (${\mathrm{StO}}_2$).

Although promising, there are still areas of uncertainty with this technology. One of these areas is the reproducibility of tissue oximeters, which has not been thoroughly studied. Current data has shown that the difference between two successive measurements can be quite large, although some oximeters have shown better reproducibility than others.^8–12 Also, newer tissue oximeters have only been studied with forehead and forearm readings in adults, other sites have been included in past clinical research.^11,12

The main objective of this study was to compare the reproducibility of two commonly used tissue oximeters, the INVOS 5100c (Covidien, Mansfield, Massachusetts) and the Equanox 7600 (Nonin Medical, Plymouth, Minnesota). For the primary outcome, we hypothesized that the Equanox would display a better reproducibility than the INVOS owing to its more recent development. The effects of sensor switching, measurement site and side of body (left versus right) on reproducibility were evaluated as secondary outcomes. The second objective was to describe the interchangeability of the oximeters and sides of the body to provide similar raw tissue saturation measurements. The third objective was to assess the time necessary for each of the oximeters to provide a stable value for tissue oximetry. Finally, the last objective was to describe and compare reference values for NIRS saturation for all sites evaluated.

2.

Methods

2.4.

Measures of Cerebral and Somatic Oximetry

Prior to measures being taken, participants were asked to provide their basic sociodemographic characteristics and their phototype, using the Fitzpatrick phototyping scale.¹⁵ At the beginning of the experiment, participants were asked to lie on their backs while their vital signs and skinfold thickness (using a Harpenden Skinfold Caliper) were measured. Six sites (forehead, deltoid, forearm, knee, calf, and foot) on both sides of the body were then marked with either a pen or tape to ensure that the same exact location was used for each sensor throughout the measure collection (cf., Fig. 1).

Fig. 1

Sites of tissue saturation measurements. 1. Forehead: above the eyebrow, long side perpendicular to the body axis and following the median line of the face. 2. Deltoid: at the height of the head of the humerus, long side perpendicular to the body axis and centered on the coronal plane of the body. 3. Forearm: distally to the elbow crease, centered on the palmar face of the forearm and long side perpendicular to the body axis. 4. Knee: on the medial side of the thigh at the base of the junction between the patella and the vastus medialis of the quadriceps, long side parallel to the leg axis and just medial to the patella. 5. Calf: on the medial line of the calf, 3 cm below the articular line of the knee and long side parallel to the leg axis. 6. Foot: on the medial line of the plantar face, long side parallel to the leg axis and proximal side just distally to the heel.

The order in which each tissue oximeters was used was randomized. The sensors were placed on each of the measurement sites until the reading was stable. A stable reading was defined as an identical value in two consecutive measurements, measures alternating between two contiguous values, or a maximum delay of 20 s from the first reading. In the latter case, the last recorded value was retained. The time delay to stabilization was noted. If a measure was not provided within 30 s of sensor apposition, the sensor, cables, and connections were verified. Should the oximeter persist in its failure to provide a measure, its functionality was tested on the experimenter. In instances where the oximeter provided a measure on the experimenter but not on the subject, the data were considered as missing, and the experimenter proceeded to the next site with the same sensor. Values from various measurement sites were taken in a predetermined order (forehead, deltoid, forearm, knee, calf then foot) with a first sensor (s1). Once all sites had been evaluated on both sides of the body (t1), the process was repeated at 5 min (t2) a second time with the same sensor (s1) to evaluate the intrasensor reproducibility. After the second run, the procedure was repeated a third time at 10 min (t3) with a different sensor of the same oximeter model (s2) to assess the intersensor reproducibility. The same procedure was then used for the second oximeter. Because up to 24 measures would ultimately need to be taken with the same sensor, it was decided that the sensors would be held in place manually by the experimenter instead of using the adhesive, in order to avoid skin irritation. That being said, great care was taken neither to compress the sensors nor to expose them to ambient light during measurements, so as to avoid impacting the values measured. Every measure was either collected or supervised by the primary author (AC). This process was repeated for each of the two oximeters. A total of 72 measures were recorded for each subject in the study (six sites per side of the body and three values per site for each of the two oximeters).

3.

Results

3.2.

Measures of Reproducibility

3.2.1.

INVOS

Intersensor ICCs for the INVOS oximeter ranged from 0.78 to 0.91 while the intrasensor ICCs ranged from 0.76 to 0.96 (cf., Table 2). Global ICCs for the INVOS oximeter, as well as $S_w$, mean bias and limits of agreement are presented in Table 3, Figs. 2 and 3. The INVOS forearm intrasensor $S_w$ was 2.97% (95% CI 2.77 to 3.18) (data not shown).

Table 2

Calculated measures of reproducibility of the INVOS oximeter (ICC and 95% CI).

	Intersensor		Intrasensor
	Left	Right	Left	Right
Forehead	0.91 (0.84 to 0.95)	0.83 (0.72 to 0.90)	0.91 (0.86 to 0.95)	0.88 (0.79 to 0.93)
Deltoid	0.84 (0.73 to 0.90)	0.84 (0.69 to 0.91)	0.86 (0.76 to 0.92)	0.90 (0.83 to 0.94)
Forearm	0.78 (0.65 to 0.87)	0.86 (0.76 to 0.92)	0.81 (0.68 to 0.88)	0.78 (0.65 to 0.87)
Knee	0.92 (0.87 to 0.96)	0.94 (0.91 to 0.97)	0.92 (0.87 to 0.95)	0.92 (0.86 to 0.95)
Calf	0.89 (0.82 to 0.93)	0.91 (0.86 to 0.95)	0.93 (0.88 to 0.96)	0.82 (0.71 to 0.89)
Foot	0.88 (0.78 to 0.94)	0.88 (0.80 to 0.93)	0.96 (0.93 to 0.98)	0.92 (0.86 to 0.95)

Note: ICC: intraclass correlation coefficient; 95% CI: 95% confidence interval.

Table 3

Global measures of reproducibility of the INVOS oximeter.

		ICC and 95% CI	Sw and 95% CI (%)	Mean bias and limits of agreement (%)
INVOS	Intersensor	0.92 (0.90 to 0.93)	2.11 (2.05 to 2.17)	$-0.85$ ($-9.11$ to 7.41)
	Intrasensor	0.92 (0.91 to 0.93)	2.22 (2.16 to 2.29)	$-0.35$ ($-8.42$ to 7.72)

Note: ICC: intraclass correlation coefficient; 95% CI: 95% confidence interval; Sw: within-subject standard deviation.

Fig. 2

Plot of identity describing the (a) intersensor and (b) intrasensor reproducibility of the INVOS oximeter.

Fig. 3

Bland-Altman plot describing the (a) intersensor and (b) intrasensor reproducibility of the INVOS oximeter.

3.2.2.

Equanox

Intersensor ICCs for the Equanox oximeter ranged from 0.73 to 0.94 while intrasensor ICCs ranged from 0.77 to 0.96 (cf., Table 4). Global ICCs for the Equanox oximeter, as well as $S_w$, mean bias and limits of agreement are presented in Table 5, Figs. 4 and 5. The Equanox forearm intrasensor $S_w$ was 2.68% (95% CI 2.50 to 2.87) (data not shown).

Table 4

Calculated measures of reproducibility of the Equanox oximeter (ICC and 95% CI).

	Intersensor		Intrasensor
	Left	Right	Left	Right
Forehead	0.76 (0.62 to 0.85)	0.73 (0.57 to 0.84)	0.82 (0.71 to 0.89)	0.77 (0.63 to 0.86)
Deltoid	0.89 (0.91 to 0.93)	0.92 (0.86 to 0.95)	0.95 (0.92 to 0.97)	0.87 (0.78 to 0.92)
Forearm	0.78 (0.65 to 0.87)	0.77 (0.63 to 0.86)	0.82 (0.72 to 0.89)	0.79 (0.66 to 0.97)
Knee	0.94 (0.90 to 0.97)	0.93 (0.89 to 0.96)	0.96 (0.93 to 0.97)	0.95 (0.92 to 0.97)
Calf	0.92 (0.87 to 0.95)	0.90 (0.83 to 0.94)	0.95 (0.92 to 0.97)	0.93 (0.88 to 0.96)
Foot	0.86 (0.76 to 0.91)	0.82 (0.71 to 0.89)	0.91 (0.86 to 0.95)	0.88 (0.81 to 0.93)

Note: ICC: intraclass correlation coefficient; 95% CI: 95% confidence interval.

Table 5

Global measures of reproducibility of the Equanox oximeter.

		ICC and 95% CI	Sw and 95% CI (%)	Mean bias and limits of agreement (%)
EQUANOX	Intersensor	0.90 (0.88 to 0.91)	1.96 (1.91 to 2.02)	$-0.39$ ($-7.78$ to 7.00)
	Intrasensor	0.92 (0.91 to 0.93)	1.67 (1.63 to 1.72)	0.32 ($-6.07$ to 6.72)

Note: ICC: intraclass correlation coefficient; 95% CI: 95% confidence interval; Sw: within-subject standard deviation.

Fig. 4

Plot of identity describing the (a) intersensor and (b) intrasensor reproducibility of the Equanox oximeter.

Fig. 5

Bland-Altman plot describing the (a) intersensor and (b) intrasensor reproducibility of the Equanox oximeter.

3.2.3.

Comparisons of reproducibility between the two oximeters

For the primary outcome of our main objective, we found no significant difference when comparing both oximeters in terms of global intersensor ICC using the fisher r-to-z transformation ($p=0.06$) (cf., Tables 3 and 5). For the secondary outcomes, using the same method, we also found no significant difference between the two oximeters in terms of global intrasensor ICCs ($p=0.78$). However, the intrasensor ICC of the Equanox was better than its intersensor ICC ($p=0.003$), which was not the case for the INVOS ($p=0.42$) (cf., Tables 3 and 5). Using the Bland-Altman method, the Equanox displayed a significantly better inter and intrasensor $S_w$ than did the INVOS ($p=0.019$ and $p<0.001$, respectively). Again, the intrasensor $S_w$ of the Equanox was better than its intersensor $S_w$ ($p<0.001$) while no such difference was found for the INVOS ($p=0.28$) (cf., Tables 3 and 5 and Figs. 2–5). Both oximeters displayed adequate consistency across the range of observed results, as shown by the similarities displayed in dot dispersion around the reference line of the Bland-Altman plots (Figs. 3 and 5).

3.2.4.

Comparisons of reproducibility between sides of the body and sites of measurement

Last, for the main objective, ICCs and $S_w$ were similar when comparing both sides of the body ($p=0.58$ and $p=0.74$, respectively). When comparing ICCs, among all possible comparisons between sites, the knee site displayed better ICCs than all the other sites ($p$ ranging from 0.005 to $<0.001$). Also, the deltoid and calf sites were superior to the forearm site ($p=0.009$ and $p<0.001$, respectively). Using the Bland-Altman method, measures from the knee and calf sites had better $S_w$ than did all other sites ($p$-values ranging from 0.005 to $<0.001$). Furthermore, the forehead, deltoid, and foot had better $S_w$ than the forearm site ($p$-values ranging from 0.002 to $<0.001$) (cf., Table 6).

Table 6

Comparison of measures of reproducibility of two tissue oximeters between sides and sites.

		ICC and 95% CI	Sw and 95% CI (%)	Mean bias and limits of agreement (%)
Left	Intersensor	0.91 (0.90 to 0.92)	2.10 (2.05 to 2.16)	$-0.49$ ($-8.43$ to 7.45)
Right	Intersensor	0.91 (0.89 to 0.92)	2.08 (2.02 to 2.13)	$-0.75$ ($-8.49$ to 6.99)
Forehead	Intersensor	0.86 (0.83 to 0.89)	2.16 (2.05 to 2.26)	$-0.81$ ($-9.06$ to 7.43)
Deltoid	Intersensor	0.88 (0.84 to 0.91)	2.07 (1.97 to 2.17)	$-0.84$ ($-8.41$ to 6.73)
Forearm	Intersensor	0.81 (0.76 to 0.86)	2.86 (2.72 to 3.00)	$-1.23$ ($-11.35$ to 8.89)
Knee	Intersensor	0.94 (0.92 to 0.95)	1.54 (1.47 to 1.62)	$-0.17$ ($-5.86$ to 5.52)
Calf	Intersensor	0.91 (0.88 to 0.93)	1.72 (1.63 to 1.80)	0.27 ($-6.36$ to 6.89)
Foot	Intersensor	0.87 (0.83 to 0.90)	2.20 (2.09 to 2.30)	$-0.94$ ($-8.72$ to 6.83)

Note: ICC: intraclass correlation coefficient; 95% CI: 95% confidence interval; Sw: within-subject standard deviation.

3.3.

Interchangeability of the Oximeters and Sides of the Body to Provide Similar Raw Tissue Saturation Measurements

For the second objective, tissue saturation measures provided by the two oximeters varied considerably [ICC 0.43 (95% CI 0.36 to 0.49), Bias $-0.77$ (Limits of agreement $-19.88$ to 18.34)]. When evaluated together, all tissue saturation measurements from the Equanox showed less variation (SD 8.0) than the ones from the INVOS (SD 10.2). Thus, in comparison to the INVOS’ tissue saturation measurements, the Equanox generally displayed higher tissue saturations when in the lower saturations range (e.g., around 40% to 60%) and lower tissue saturations in the higher saturation range (e.g., around 75% to 90%). This can be appreciated in Fig. 6 in the Bland-Altman plot by noticing that the dot cloud seems to follow a negative linear regression. On the other hand, the tissue saturations varied much less when we compared the readings from the left and right sides of the body [ICC 0.91 (95% CI 0.89 to 0.92), Bias $-0.34$ (limits of agreement $-7.94$ to 7.25)] (cf., Fig. 7). Those results are similar to the ones previously described for inter- and intrasensor reproducibility (cf., Tables 3 and 5 and Figs. 2–5).

Fig. 6

(a) Plot of identity and (b) Bland-Altman plot describing the interchangeability among two tissue oximeters.

Fig. 7

(a) Plot of identity and (b) Bland-Altman plot describing the interchangeability between the oximetry readings of the left and right side.

4.

Discussion

The present study sought to compare the reproducibility of two commonly used tissue oximeters. Both oximeters exhibited good reproducibility. Interestingly, no differences were observed when ICCs were used to assess this metrological characteristic, but the Equanox displayed better reproducibility when the Bland-Altman method was used, displaying smaller $S_w$ and limits of agreement. The reproducibility also varied among measurement sites, but not between each side of the body. Despite the good intraoximeter reproducibility, the interchangeability of raw saturation value between both oximeter was deficient, reflecting poor interoximeter agreement. To our knowledge, this is the first time that that both of these characteristics (reproducibility and interchangeability) were evaluated for tissue saturation at sites other than the forehead and the forearm, or for multiple sites of the body. The wide range of observed values in our experiment allowed a good appraisal of these characteristics. In addition, we showed the Equanox was faster at providing a stable saturation measurement than the INVOS was, a characteristic that had never been evaluated in these two devices. Finally, reference values for the two oximeters used were also described for the first time at sites most often currently used in clinical practice.

The Equanox showed better absolute reproducibility, as reflected by better $S_w$ values than the INVOS, but similar relative reproducibility, as reflected by the similar ICC values. This is likely explained by the fact that, as explained earlier, the ICC is a correlation coefficient and therefore takes into account the variability and range of the observed measures, thus providing a relative measure while the $S_w$ and limits of agreement are absolute measures of reproducibility. The Equanox also provided a smaller range of observed values across the measurement sites and sides of the body than did the INVOS. Based on these results, the Equanox could be favoured over the INVOS, especially when NIRS monitoring is used intermittently or in unstable conditions since good reproducibility of a device is more important in these conditions than in continuous or stable monitoring. That being said, the magnitude of the differences observed in terms of reproducibility between the two oximeters is quite small and should not be the sole determinant of the type of oximeter used. Other issues that should be considered include price, versatility, convenience, or institutional preferences. The Equanox’s intrasensor reproducibility was better than its intersensor reproducibility. This should encourage clinicians using this device to use a single sensor per patient whenever possible.

These results contrast in some aspects with prior published studies. First, studies evaluating NIRS reproducibility only presented $S_w$ values, and not ICCs, making ICC comparisons impossible. These studies reported $S_w$ ranging from 3% to 7%.^{8,9,11,12,17–20} Notably, in three different studies, Hyttel-Sorensen observed $S_w$ of 5.4%,¹¹ 2.9%,¹² and 4.0%¹⁹ using value measured on the forearms of healthy adults using the INVOS. We, on the other hand, observed slightly lower $S_w$ of 3.0% for the same measurement site. As for the Equanox, a $S_w$ of 4.6% was observed in the sole study it was used in, as compared to the 2.7% $S_w$ we observed.¹² Therefore, the reproducibility observed in our study is among the better values previously mentioned in the available published data. This is possibly because we standardized the data collection procedures and technique. In our experiment, all measurement sites were marked, participant body position was stable and all measurements were taken or directly supervised by the same investigator (AC). This may have contributed to our better measures of reproducibility, which are similar to the best values observed by Hyttel-Sorensen.¹² This emphasizes the necessity to be both familiar and rigorous with the procedure used when NIRS monitoring is utilized, especially in case where intermittent measures are taken. The Fore-Sight (Casmed, Branford, Connecticut), another oximeter less commonly used that we did not evaluate, compared favourably to both the INVOS and Equanox in terms of reproducibility in other experiments. In light of our results, this oximeter likely merits more rigorous evaluation in further studies of reproducibility.^12,18,21

We identified variations in the reproducibility between measurement sites, but not between sides of the body. Variations in the reproducibility between measurement sites had already been observed in a paediatric study, which only included measures from the forehead and arm.¹⁹ This should influence the way intermittent monitoring with this technology is done. Indeed, when intermittent peripheral NIRS monitoring is used to assess body perfusion, such as in the intensive care unit, the knee and calf sites should be favoured over other sites because of their better reproducibility. Of course, this does not apply when upper extremity specific saturations are warranted, such as in situations where upper extremity ischemia is suspected.

Despite similar mean values in both samples, interoximeter agreement was poor on an individual basis, with limits of agreement of $\pm 19\%$. This is even larger than what had already been observed in multiple settings and between other brands of oximeters.^22–26 This is likely explained by the measurements over sites—such as the deltoid, knee or foot—for which this had never been evaluated. Moreover, it is probable that ischemia worsen disagreements between oximeters’ readings.²⁴ Thus, our results seem to reinforce the fact that readings from different brands of oximeters are not interchangeable and that absolute thresholds of clinically significant ischemia are oximeter-specific and need to be studied individually. This high variability is likely caused by the fact that each type of NIRS oximeter has unique characteristics (number of light-emitting diodes and receptors, distances between diodes, wavelengths used). This will influence the size and depth at which the underlying tissue is evaluated (wide distances between emitters and receptors increase the volume of tissue evaluated).^27,28 Moreover, tissue oximeters use a specific algorithm to calculate the overall concentration of oxyhemoglobin and deoxyhemoglobin.²⁹ This is explained by the absence of a recognized gold-standard for tissue oxygenation, which led to lack of standardization for the NIRS technology.

On the other hand, interside agreement was much better with limits of agreement of $\pm 8\%$, a value that is somewhat smaller than what has been previously observed in the cardiac and vascular surgery population.³⁰ This difference is probably explained by the assumed absence of significant vascular disease in our sample of healthy volunteers, a characteristic that could potentially create asymmetry limbs perfusion. This description could help guide clinicians when caring for a healthier population when searching for perfusion asymmetry between limbs.

The Equanox was quicker to give a stable reading than the INVOS. While the difference observed ($\sim 5\mathrm{s}$) might not be clinically relevant in a hospital setting, the Equanox could be favored if this technology is used as a triage assessment tool in mass casualty events, helping to identify patients who require emergent, life-saving interventions.^6,31

Finally, we described the reference values for the INVOS and Equanox at most sites used in clinical practice. The variations observed between sites can be explained by the perfusion status of these sites, but also of the nature of the tissue evaluated (brain, muscle, connective tissue, fat, and so on). These results will likely help clinicians when using this technology at less conventional sites. The results observed in the present study were similar to most results described previously in healthy volunteers.^11,12,23 Davie and Fellahi^25,32 observed higher tissue saturations in their experiment on the forehead and calf, respectively. This could be explained by different sensor positioning and a smaller sample size ($n=12$ and 20). Interestingly, patients with stable severe systemic disease about to undergo cardiac surgery seem to have similar cerebral saturation as healthy volunteers.^30,33 This suggests that only critical systemic insults will affect baseline cerebral saturations.³³ The possible determinants of raw tissue saturations across all measurements sites (e.g., age, comorbidity, skin color, body mass index, or skinfold thickness) will need to be better studied and defined in future studies.²¹

5.

Conclusion

In summary, the Equanox could be favored over the INVOS, especially when NIRS monitoring is used intermittently or in unstable conditions, mainly because of its slightly better absolute reproducibility. The knee and calf sites displayed better reproducibility, and could potentially be favoured in situations where NIRS monitoring at these sites is clinically indicated. The description of reference values and normal range of asymmetry for each measurement site will help clinicians identify situations of abnormal perfusions. The differences in observed tissue saturation between devices suggests that clinicians should use the same device for each patient when possible. Also, tissue saturation thresholds are not interchangeable from one NIRS oximeter to another, given their poor interoximeter agreement. NIRS offer new avenues for assessing perfusion in an array of clinical situations in which noninvasive monitoring is required. However, an eventual standardization of NIRS oximeters will be essential to future development of this technology and the expansion of its use in critical care.