The utility of near infrared spectroscopy for estimating the mixed venous saturation in children: insights using data from extracorporeal membrane oxygenation

doi:10.21203/rs.3.rs-1965393/v1

Download PDF

Article

The utility of near infrared spectroscopy for estimating the mixed venous saturation in children: insights using data from extracorporeal membrane oxygenation

https://doi.org/10.21203/rs.3.rs-1965393/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background: As near infrared spectroscopy (NIRS) offers a noninvasive means by which to monitor regional venous saturation, this study set forth to determine the correlation between near infrared spectroscopy values and mixed venous saturations (SvO2) obtained while on veno-arterial extracorporeal membrane oxygenation (VA-ECMO).

Methods: This was a single-center retrospective study which included pediatric patients who required VA-ECMO and had cerebral and renal NIRS values between November 2020 through November 2021. Demographic data, SvO2, and cerebral and renal NIRS were collected in all patients. A Bayesian linear regression was done to model the SvO2 using the NIRS values. A venous saturation estimate was calculated by averaging the NIRS values. A Bland-Altman analysis and a Bayesian Pearson correlation coefficient calculation were performed between the actual SvO2 and both the modeled and the average-estimated SvO2.

Results: Data from a total of nine patients were used in the final analyses. Bayesian linear regression analyses demonstrated that both cerebral and renal NIRS values were statistically significantly associated with the measured SvO2. The mean difference between the estimated and measured SvO2 was -19 (-34 to -5). The Bayesian Pearson correlation coefficient was 0.574 (0.492 to 0.642). When the SvO2 was estimated with a simple average of the cerebral and renal NIRS values, the mean difference between the average estimated and measured SvO2 was 2 (-14 to 18). The Pearson correlation coefficient was 0.530 (0.443 to 0.604).

Conclusion: Regional NIRS can be used to estimate SvO2 in critically ill children with moderate correlation in patients in VA-ECMO.

near infrared spectroscopy

extracorporeal membrane oxygenation

mixed venous saturation

pediatric critical care

The maintenance of adequate systemic oxygen delivery is at the basis of maintaining life and reducing morbidity and mortality. Systemic oxygen delivery is the product of oxygen content and cardiac output and while oxygen content can be quantified relatively easily utilizing arterial blood gas values, cardiac output is not as easy to quantify in children. Cardiac output can be measured by use of indwelling catheters, although there are several practical limitations of this approach in children ^{1, 2}.

The Fick principle offers an alternative approach to quantifying cardiac output. The Fick equation describes cardiac output as oxygen consumption divided by the arteriovenous oxygen content difference ³. While oxygen consumption can be estimated, both the arterial and venous oxygen contents can be calculated using arterial and blood gas values, respectively.

While regional venous saturations can be helpful as a trending tool to estimate cardiac output, the mixed venous saturation is best used to help estimate global systemic cardiac output. Mixed venous saturations can be obtained through a right atrial or pulmonary artery catheter but are limited by practical considerations ⁴. Rather, regional venous saturations obtained by central venous lines or by near infrared spectroscopy, are more frequently and easily obtained in critically ill children ⁵. These have been demonstrated to be useful in managing critically ill patients although there are some who still prefer to guide management with a true mixed venous saturation and disregard the use of, or minimize the utility of, regional venous saturations.

A unique situation in which a true mixed venous saturation can be obtained in critically ill children is when on veno-arterial extracorporeal membrane oxygenation (VA ECMO). If the venous catheter is in the right atrium, then the venous saturation obtained from the venous limb of the VA-ECMO circuit represents a mixed venous saturation in most circulations. Children on VA-ECMO will also often be monitored utilizing near infrared spectroscopy. NIRS utilization in patients on ECMO is routine at some practicing institutions. This practice is predicated by its utility for the early identification of possible perturbations to the NIRS signal such as thromboembolic phenomena, air entrapment or cannula malposition. Additionally, having NIRS placed while on ECMO allows for continuing to trend them after decannulation from ECMO and loss of the ability to continuousuly monitor mixed venous saturation.

The advent of newer, noninvasive monitor technology allows providers make management decision in a prompt manner. Correlation studies between NIRS and invasive global blood flow are necessary as they help providers assess the linear relationship between the two and the strength of that relationship. Understanding the correlation between venous saturation and near infrared spectroscopy values allows for a better understanding of the utility infrared spectroscopy in patient management.

As near infrared spectroscopy offers a noninvasive means by which to monitor regional venous saturation, this study set forth to determine the correlation between near infrared spectroscopy values and mixed venous saturations obtained while on VA-ECMO. It was hypothesized that cerebral and renal near infrared spectroscopy values would be able to accurately estimate the mixed venous saturation.

Study design

This was a single-center retrospective study. Institutional review board approval was received, and the study is in compliance with the Helsinki Declaration. Because this study was a retrospective medical record review, and thus presented no more than minimal risk, the institutional review boards waived the need for patient informed consent.

Patient identification

Pediatric patients (under 18 years of age) who were cared for at Advocate Children’s Hospital and required VA-ECMO between November 2020 through November 2021 were identified. This specific time period was utilized as the cardiac intensive care unit at our institution transitioned between near infrared spectroscopy devices and this time period ensured that all near infrared spectroscopy values were obtained using the same device. All patients in our institution on VA-ECMO have venous blood gases run from the venous side of the ECMO pump every four to six hours and no patients were excluded for lack of these data. These patients were then reviewed to make sure that they had cerebral and renal near infrared spectroscopy values while on VA-ECMO. Those without both cerebral and renal infrared spectroscopy values available while on VA-ECMO were excluded from this study. Patients with parallel circulation were included only if they had a normally draining systemic veins and an atrial level communication that was all right to left such that the right atrial blood would not have any pulmonary venous contribution in it. Parallel circulation was defined as a circulation in which the pulmonary and systemic circulations were being supplied with blood with the same oxygen saturation.

Near infrared spectroscopy

All near infrared spectroscopy values were obtained using the Casmed Fore-sight Elite platform (Edwards Lifesciences, Irvine, CA, USA). Patients must have had two-site near infrared spectroscopy monitoring to be included in this study. The cerebral near infrared spectroscopy values were obtained by use of a probe placed on the forehead while the renal infrared spectroscopy values were obtained by use of a probe placed on either flank. Probes are replaced at least every three days, sooner if otherwise indicated.

Data Collection

As the aim of this study was specifically to explore the relationship between near infrared spectroscopy values and venous saturations data collection was limited. A retrospective chart review was conducted to gather data on the patients deemed eligible for inclusion. Demographic data about all patients were collected including age at time of cannulation for VA-ECMO, the presence or absence of congenital heart disease, the presence of absence of parallel circulation (defined as the situation in which the saturation of blood to the pulmonary and systemic circulations is equal), and duration of VA-ECMO. The date and time at which the venous saturation was obtained was then noted and cerebral and renal near infrared spectroscopy values recoded within 15 minutes of the venous saturation were also noted. If a cerebral and renal near infrared spectroscopy value was not recorded within 15 minutes of the venous saturation, then none of the data at that timepoint were utilized.

Hemoglobin, methemoglobin, and patient temperature were also all recorded from the venous blood gas results. Total bilirubin as noted on laboratory testing was also recorded for each day. All of these variables were collected as they have been documented previously to impact near infrared spectroscopy values.

Statistical analyses

A Bayesian linear regression was done to model the mixed venous saturation using the cerebral and renal near infrared spectroscopy values. The mixed venous saturations was entered as the dependent variable and the cerebral near infrared spectroscopy values, renal near infrared spectroscopy values, temperature, hemoglobin, methemoglobin, and total bilirubin were entered as the independent variables.

The resulting marginal posterior distributions were then reviewed for each independent variable to develop a model to estimate the most probable mixed venous saturation. The resulting model from the linear regression was then used to calculate a mixed venous saturation at each time point. A Bland-Altman analysis was done to estimate the mean difference and the 95% confidence interval of the mean difference between the modeled mixed venous saturation and the actual mixed venous saturation. A Bayesian Pearson correlation coefficient was then calculated to assess the correlation between this modeled venous saturation and the obtained venous saturation.

Next a venous saturation estimate was calculated by simply averaging the cerebral and renal near infrared spectroscopy values. A Bland-Altman analysis was done to estimate the mean difference and the 95% confidence interval of the mean difference. A Bayesian Pearson correlation coefficient was then calculated to assess the correlation between this average-based estimated venous saturations and the obtained venous saturation. The purpose of this was to see what the utility of a more easy-to-calculate estimate of the mixed venous saturation would be.

Pearson correlation coefficients were qualitatively interpreted as follows: a very strong correlation was considered to be present for a correlation greater than or equal to 0.90, a strong correlation for a correlation coefficient of between 0.70 and 0.89, a moderate correlation for a correlation coefficient between 0.50 and 0.69, a weak correlation for a correlation coefficient between 0.30 and 0.49, and a negligible correlation for a correlation coefficient below 0.30 ⁶.

A frequentist linear regression analysis was also conducted for comparison. The dependent variable was the actual mixed venous saturation and the dependent variables were cerebral near infrared spectroscopy values, the renal near infrared spectroscopy values, temperature, hemoglobin, methemoglobin, and total bilirubin.

All statistical analyses were done using JASP Version 0.16 (University of Amsterdam, Amsterdam, The Netherlands). As Bayesian analyses and not frequentist analyses were used, P-values are not reported. Bayes Factors are reported with results

Cohort characteristics

Data from a total of nine patients were used in the final analyses. Eight of the nine (89%) included patients had congenital heart disease. Of all nine patients, six (66%) had parallel circulation. Average age at time of cannulation was 3 months and the average weight was 3.8 kg (Table 1). A total of 313 data points were available from these nine patients.

The mean mixed venous saturation was 72.7 ± 8.8, the mean cerebral infrared spectroscopy value was 71.9 ± 7.8, and the mean renal near infrared spectroscopy value was 77.3 ± 9.8.

Measured mixed venous saturation compared to model-based estimate of mixed venous saturation

Bayesian linear regression analyses demonstrated that both cerebral and renal near infrared spectroscopy values were statistically significantly associated with the measured mixed venous saturation. The most probable model was as follows: (0.996 x hemoglobin) + (0.182 x cerebral near infrared spectroscopy value) + (.403 x renal near infrared spectroscopy value). The r-squared value for the model was 0.53.

Utilizing the Bayes Factor for the model, the inclusion of hemoglobin, cerebral near infrared spectroscopy, and renal near infrared spectroscopy increased the odds of in favor of the model by a factor of 52.2. Observed data were 2.1 times more likely with the model containing hemoglobin, cerebral near infrared spectroscopy value, and renal near infrared spectroscopy value than the next most probable model.

Hemoglobin was found to have a 99% probability of being included the model for mixed venous saturation, cerebral near infrared spectroscopy 95%, and renal near infrared spectroscopy 100%. Temperature was found to have a 36% probability of being included in the model, methemoglobin 50% and total bilirubin 34%. The Bayes Factor for inclusion in a model was 1,311 for hemoglobin, 19 for cerebral near infrared spectroscopy, and 2,900 for renal near infrared spectroscopy (Table 2).

The mean difference between the estimated and measured venous saturation was -19 (95% confidence interval -34 to -5). The Bayesian Pearson correlation coefficient was 0.574 (95% confidence interval 0.492 to 0.642), demonstrating a significant and moderate correlation. The Bayes Factor for this correlation was greater than 100.

Measured mixed venous saturation compared to average-based estimate of mixed venous saturation

When the mixed venous saturation was estimated with a simple average of the cerebral and renal near infrared spectroscopy values, the mean difference between the average-estimated and measured mixed venous saturation was 2 (95% confidence interval -14 to 18). The Pearson correlation coefficient was 0.530 (0.443 to 0.604), demonstrating a significant and moderate correlation. The Bayes Factor for this correlation was greater than 100.

Frequentist linear regression

The results of the frequentist linear regression to model mixed venous saturation are outlined in the supplemental file. The resulting regression equation is similar to the equation from the Bayesian analysis.

While regional venous saturations can be helpful in guiding clinical management, some clinicians still prefer to utilize mixed venous measurements. Regional near infrared spectroscopy values have moderate correlation with regional venous saturations and, thus, the current study aimed to determine how regional near infrared spectroscopy values correlate with mixed venous saturation ⁷. To do this, the current study utilized patients on VA-ECMO in whom the ECMO pump venous blood reflects a mixed venous blood sample. The current study demonstrates that regional near infrared spectroscopy has a significant, moderate correlation with the actual mixed venous saturation. While the correlation is moderate, the absolute correlation has a wide 95% confidence interval. Thus, the regional near infrared spectroscopy values are more useful as indicators of the directional change in the mixed venous saturation rather than predicting the exact value of the mixed venous saturation.

Near infrared spectroscopy values obtained from the Fore Sight elite device were not significantly impacted by temperature, methemoglobin, or total bilirubin. Hemoglobin, did, however, impact near infrared spectroscopy values. Interestingly, the estimated mixed venous saturation was only minimally more accurate when hemoglobin was included in the model along with weighted near infrared spectroscopy values. A simple average of the cerebral and near infrared spectroscopy values actually had moderate correlation with the measured mixed venous saturation, demonstrating that the average of the two regional near infrared spectroscopy values can be trended to help determine the direction of change in the mixed venous saturation. Thus, it may be reasonable that at the bedside one can simply average the two rather than try to use the specific coefficients from the model to help estimate what the mixed venous saturation is.

What is interesting about the findings from the current study are the relative contributions of the cerebral and renal near infrared spectroscopy values on the mixed venous saturation calculation. Previous studies have described models to estimate the mixed venous saturation with variability in the coefficients for superior and inferior caval vein saturations. Thus, it is not surprising that a similar phenomenon is noted with near infrared spectroscopy values. Additionally, there may be physiologic reasons for variability in regional oxygen economy. Nonetheless, as the normal difference between the superior and inferior caval vein saturations is approximately 10, assigning different relative contributions of each to the mixed venous saturation will yield estimates with low enough difference to truly be clinically significant ⁸.

Near infrared spectroscopy utilizes the Beer-Lambert law, employing infrared light with a wavelength between 660 and 940 nm. The sensor detects how much light is absorbed at wavelengths generally associated with oxyhemoglobin and deoxyhemoglobin, providing an estimate of the underlying regional oxygen saturation.

Being able to monitor the trend of the regional venous saturation, or the mixed venous saturation, allows for assessment of the adequacy of systemic oxygen delivery. The adequacy of systemic oxygen delivery is determined by oxygen consumption and oxygen delivery. While near infrared spectroscopy doesn’t allow for determining whether changes are due to oxygen consumption or oxygen delivery, it allows for assessment of the overall adequacy of systemic oxygen delivery. Previous studies in children have demonstrated that near infrared spectroscopy can help predict morbidity and mortality and that titration of care utilizing near infrared spectroscopy can lead to a reduction in morbidity and mortality ^9-13. A continuous, noninvasive means to do this is not otherwise available. Invasive methods such as central venous catheters are associated with risks including vascular damage, thrombosis, and infection. Compared to serum laboratory evaluations such as lactate, venous saturation is a more robust way to estimate cardiac output and, subsequently, systemic oxygen delivery. It is important to recall that venous saturation is part of the Fick equation to estimate cardiac output while serum lactate is not. Serum lactate can be elevated for reasons other than inadequate cardiac output or inadequate systemic oxygen delivery.

While this study utilizes children on ECMO, this is simply because these are the patients that actually have true mixed venous saturations measured. There is little reason, or previous data, to demonstrate that this correlation between near infrared spectroscopy and mixed venous saturation should change once off ECMO. Previous studies have demonstrated good correlation of regional near infrared spectroscopy values and regional venous saturations in patients not on ECMO, with coefficients similar to that noted in the current study.

Cost may be another consideration. Near infrared spectroscopy devices cost approximately $10,000 to $20,000. These are reusable while the sensors themselves cost approximately $100 each. The average blood gas analyzer costs approximately $22,591. The billed charge for a venous gas with lactate can be approximately $500. Thus, the near infrared spectroscopy and blood gas analyzer cost may equal out. Two-site near infrared spectroscopy then costs $200 per every three days since the sensors can last approximately three days. This is far less than the cost of a single venous blood gas with a venous saturation and lactate included. Thus, it is possible to hypothesize that near infrared spectroscopy may offer some cost-savings. This, however, has not been demonstrated objectively in reported studies in pediatric cardiology. Other disciplines of medicine and surgery, however, have reported some cost savings associated with the use of near infrared spectroscopy ¹⁴.

Near infrared spectroscopy is not without its limitations. Different device models will have different predictive values and device-specific limitations must be understood ¹⁵. Different devices will have values impacted by skin pigmentation, total bilirubin, temperature, and hemoglobin values to different extents. The sensors may also result in cutaneous burns. Nonetheless, these limitations don’t pose a significant increase in the risk of harm to patients.

This study is also not without its limitations. First, while the mixed venous saturation and near infrared spectroscopy values were collected with temporal proximity, there was still a 30-minute window that was used. It is possible that in some patients that the near infrared spectroscopy values collected didn’t accurately reflect the ones precisely when the mixed venous gas was drawn. This is unfortunately a limitation of the retrospective nature of this study. It was felt that a 30-minute window, with 15-minutes on either side of the time that the mixed venous saturation was drawn, would be acceptable based on anecdotal experience. It is likely that this limitation would lead to a slightly lower correlation coefficient and thus the correlation coefficients reported in this study may be lower than what would have been noted with precisely simultaneous measurements.

Another limitation of this study is that we did not have a robust means by way to quantify and adjust for skin pigmentation. So, while were able to account for other things that could have impacted near infrared spectroscopy values, we were unable to do so for skin pigmentation. Regardless, if this played a role with the device utilized for this study, the correlation coefficients reported in this study may be lower than what they would have been if skin pigmentation had been adjusted for. Additionally, most of the patient cohort were patients with parallel circulation, although those patients had normally draining systemic veins and an atrial level communication that was all right to left.

Local factors on the venous side of the extracorporeal membrane oxygenation circuit, including continuous infusions, blood transfusions, continuous renal replacement therapy, type of pump, and extracorporeal membrane oxygenation flow may all impact the mixed venous saturation without altering the near infrared spectroscopy. If anything, this would mean that the true correlation between the mixed venous saturation and the near infrared spectroscopy would be higher than noted in this study.

The findings of this study may not be exactly portable to other near infrared spectroscopy devices. While the direction of associations should be similar, the magnitude may not be.

Regional near infrared spectroscopy can be used to estimate the mixed venous saturation in critically ill children with moderate correlation in patients in VA-ECMO.

Acknowledgements: None

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Conflicts of interest

The authors have no relevant financial or non-financial interests to disclose.

Availability of data and material

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Ethics approval

Institutional review board approval was received, and the study is in compliance with the Helsinki Declaration. Because this study was a retrospective medical record review, and thus presented no more than minimal risk, the institutional review boards waived the need for patient informed consent.

Bronicki RA. Hemodynamic Monitoring. Pediatr Crit Care Med. Aug 2016;17(8 Suppl 1):S207-14. doi:10.1097/PCC.0000000000000779
Bronicki RA, Chang AC. Management of the postoperative pediatric cardiac surgical patient. Crit Care Med. Aug 2011;39(8):1974-84. doi:10.1097/CCM.0b013e31821b82a6
Tsang R, Checchia P, Bronicki RA. Hemodynamic Monitoring in the Acute Management of Pediatric Heart Failure. Curr Cardiol Rev. 2016;12(2):112-6. doi:10.2174/1573403x12666151119165007
Bronicki RA. Venous oximetry and the assessment of oxygen transport balance. Pediatr Crit Care Med. Jul 2011;12(4 Suppl):S21-6. doi:10.1097/PCC.0b013e3182211667
Rivers EP, Ander DS, Powell D. Central venous oxygen saturation monitoring in the critically ill patient. Curr Opin Crit Care. Jun 2001;7(3):204-11. doi:10.1097/00075198-200106000-00011
Schober P, Boer C, Schwarte LA. Correlation Coefficients: Appropriate Use and Interpretation. Anesthesia & Analgesia. 2018;126(5):1763-1768. doi:10.1213/ane.0000000000002864
Loomba RS, Rausa J, Sheikholeslami D, et al. Correlation of Near-Infrared Spectroscopy Oximetry and Corresponding Venous Oxygen Saturations in Children with Congenital Heart Disease. Pediatr Cardiol. Aug 30 2021;doi:10.1007/s00246-021-02718-7
Iskandrian A, Kimbiris D, Bemis CE, Mintz G. A comparison of formulas used to estimate mixed venous saturations. Cathet Cardiovasc Diagn. 1976;2(4):347-51. doi:10.1002/ccd.1810020407
Ranucci M, Isgro G, Carlucci C, et al. Central venous oxygen saturation and blood lactate levels during cardiopulmonary bypass are associated with outcome after pediatric cardiac surgery. Critical care. 2010;14(4):R149. doi:10.1186/cc9217
Hoffman GM, Ghanayem NS, Kampine JM, et al. Venous saturation and the anaerobic threshold in neonates after the Norwood procedure for hypoplastic left heart syndrome. The Annals of thoracic surgery. Nov 2000;70(5):1515-20; discussion 1521. doi:10.1016/s0003-4975(00)01772-0
Tweddell JS, Hoffman GM, Mussatto KA, et al. Improved survival of patients undergoing palliation of hypoplastic left heart syndrome: lessons learned from 115 consecutive patients. Circulation. Sep 24 2002;106(12 Suppl 1):I82-9.
Tweddell JS, Ghanayem NS, Mussatto KA, et al. Mixed venous oxygen saturation monitoring after stage 1 palliation for hypoplastic left heart syndrome. The Annals of thoracic surgery. Oct 2007;84(4):1301-10; discussion 1310-1. doi:10.1016/j.athoracsur.2007.05.047
Dabal RJ, Rhodes LA, Borasino S, Law MA, Robert SM, Alten JA. Inferior vena cava oxygen saturation monitoring after the Norwood procedure. Ann Thorac Surg. Jun 2013;95(6):2114-20; discussion 2120-1. doi:10.1016/j.athoracsur.2013.01.076
Pelletier A, Tseng C, Agarwal S, Park J, Song D. Cost analysis of near-infrared spectroscopy tissue oximetry for monitoring autologous free tissue breast reconstruction. J Reconstr Microsurg. Oct 2011;27(8):487-94. doi:10.1055/s-0031-1284234
Nagdyman N, Ewert P, Peters B, Miera O, Fleck T, Berger F. Comparison of different near-infrared spectroscopic cerebral oxygenation indices with central venous and jugular venous oxygenation saturation in children. Paediatr Anaesth. Feb 2008;18(2):160-6. doi:10.1111/j.1460-9592.2007.02365.x

Table 1. Characteristics of included patients

Characteristics	Value
*Age (months)*	3 (2)
*Weight*	3.8 (1.6)
*Gender*
Male	5 (56%)
Female	4 (44%)
*Congenital Heart Disease*
Yes	8 (89%)
No	1 (11%)
*Parallel Circulation*
Yes	6 (66%)
No	3 (33%)

Values are presented as mean (SD) or n (%)

Table 2. Data regarding the variables including in the most probably model

	Probability of inclusion	Bayes Factor inclusion	Mean coefficient
*Hemoglobin*	99%	1,311	.966
*Cerebral near infrared spectroscopy*	95%	19	0.182
*Renal near infrared spectroscopy*	100%	2,900	0.403

No competing interests reported.

SupplementalFile1.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

The utility of near infrared spectroscopy for estimating the mixed venous saturation in children: insights using data from extracorporeal membrane oxygenation

Status:

Version 1

Abstract

Introduction

Methods

Results

Discussion

Conclusion

Declarations

References

Tables

Additional Declarations

Supplementary Files

Status:

Version 1