Comparing near-infrared spectroscopy-measured cerebral oxygen saturation and corresponding venous oxygen saturations in children with congenital heart disease: a systematic review and meta-analysis.

Background: Near-infrared spectroscopy (NIRS) is a non-invasive approach that measures cerebral regional oxygen saturation (rScO2). In this study, we evaluated the evidence on the validity of NIRS and the interchangeability between NIRS and common invasive approaches by exploring the correlation and consistency and comparing the mean and standard deviation between the NIRS rScO2 and jugular bulb venous oxygen saturation (SjvO2) as well as central venous oxygen saturation (ScvO2) in the perioperative period of children with congenital heart disease (CHD).
Methods: We searched electronic bibliographic databases (PubMed, The Cochrane Library and Embase) and screened the studies that met the inclusion criteria. We included cross-sectional studies of CHD pediatric patients in the perioperative period receiving both tests for NIRS rScO2 and SjvO2 or NIRS rScO2 and ScvO2. Methodological quality assessment and heterogeneity analyses were performed. We qualitatively summarized the results of Bland-Altman's analysis. Meta-regression, subgroup analyses, and sensitivity analyses were carried out to explore the causes of heterogeneity.
Results: There was no significant difference in Cohen's d between rScO2 and ScvO2 or between rScO2 and SjvO2 (Cohen's d =0.06, 95% CI: -0.16 to 0.28; Cohen's d =0.03, 95% CI: -0.25 to 0.31, respectively) and notable heterogeneity existed (I2=76.0%, P<0.001; I2=73.6%, P<0.001, respectively). A positive linear correlation was present between rScO2 and ScvO2 or between rScO2 and SjvO2 (r=0.58, 95% CI: 0.54 to 0.63; r=0.60, 95% CI: 0.54 to 0.66, respectively) and the heterogeneity was not significant (I2=36.7%, P=0.065; I2=12.7%, P=0.328, respectively). In most studies, the 95% limits of agreements of Bland-Altman's analysis were large. No evidence of publication bias was observed. Conclusions: The rScO2 measured by NIRS reflected the SjvO2 and ScvO2 monitored by invasive measurements in the perioperative period of children with CHD to some extent. However, wide limits of agreements between rScO2 and SjvO2 as well as ScvO2 indicated that NIRS and SjvO2 as well as ScvO2 are not interchangeable. Whether NIRS plays a prominent role in monitoring cerebral oxygen saturation in children with CHD needs further research. 2022 Translational Pediatrics. All rights reserved.

Introduction

In the past several years, the postoperative survival rate of children with congenital heart disease (CHD) has improved, but whether their postoperative quality of life has also improved is not known (1-3). For CHD patients undergoing cardiac surgery or cardiac catheterization, brain injury could occur before and after interventions, and neurodevelopmental impairment could affect patients’ cognition, motor skills, social interaction and behavior, language, concentration, and executive function (4,5). Preoperative and postoperative hypotension and hypoxemia are significant risk factors for brain injury in pediatric CHD patients (6,7). Thus, real-time neurological monitoring is necessary for children with CHD undergoing cardiac surgery or cardiac catheterization (5). Cerebral perfusion and oxygen saturation are important factors affecting neurological functions that must be monitored perioperatively to assure satisfactory patient outcomes.

Direct or indirect invasive approaches are used to monitor cerebral oxygen saturation. The best indicator reflecting global tissue oxygen saturation is mixed venous saturation, but it is inconvenient to acquire. Central venous oxygen saturation (ScvO2) is considered the gold standard substitute for mixed venous oxygen saturation to monitor tissue oxygen saturation in pediatric cardiac surgery (8-10). Pulmonary artery catheters can provide ScvO2 directly. In the absence of pulmonary artery catheters, superior vena cava saturation can be used to reflect cardiac index and mixed venous oxygen saturation after cardiac surgery as a measure of tissue oxygen saturation. Sampling through retrograde cannulation of the jugular vein and measuring jugular bulb venous oxygen saturation (SjvO2) by reflectance oximetry is another accepted invasive method for measuring global cerebral oxygen saturation. An SjvO2 below 50% indicates that insufficient oxygen is supplied to the brain, and treatment to increase cerebral oxygen supply and/or decrease metabolic demand is needed (11).

Near-infrared spectroscopy (NIRS) is a non-invasive method for the measurement of cerebral regional oxygen saturation (rScO2) (12). Invented by Jöbsis et al. in 1997 (12), NIRS is a widely used and reliable tool for the measurement of cerebral hypoperfusion in infants with CHD (13). The NIRS monitoring system is based on tissue transmission and absorption of near-infrared light (wavelength 700–950 nm) via biomolecules, for example, oxygenated and deoxygenated hemoglobin. Owing to different optical densities in the near-infrared spectrum, the concentrations of hemoglobin molecules can be determined by their relative absorption wavelength (14). In addition, rScO2 measures the oxygen saturation of brain tissue after weighting that of arteries, veins, and capillaries, and it is simultaneously related to physiological variables such as arterial oxygen saturation, partial pressure of arterial carbon dioxide (PCO2), blood pressure, hematocrit, cerebral blood flow, cerebral blood volume, and cerebral metabolic rate (15,16). In pediatric cardiac surgery, it is recommended that NIRS oxygenation monitoring start before delivery of oxygen and continue to the postoperative period (7). As a sustainable, rapid, and non-invasive monitoring tool for rScO2 measurement, NIRS may yield improved measurement of cerebral oxygen saturation. Nevertheless, due to the lack of generally accepted reference values and assumption of fixed arterio-venous ratio as well as diverse algorithms, the accuracy of NIRS remains controversial (17). The paradox is that biologic variation exists in arterio-venous ratio related to hypoxia while manufacturers hypothesize a fixed arterio-venous ratio, which will inevitably affect the results of NIRS (18). Besides, readings of NIRS vary considerably between repeated measurement of the same subjects and between subjects with different cerebral oximeters (19). Currently, the Food and Drug Administration (FDA) have not regulated the standards for accuracy of cerebral oximeters. A systematic review and meta-analysis is needed to validate the accuracy of NIRS and assess whether rScO2 can replace SjvO2 and ScvO2 to monitor cerebral oxygen saturation.

This systematic review and meta-analysis intended to assess the validity of NIRS in measuring cerebral oxygen saturation in children with CHD undergoing surgery and evaluate the interchangeability between noninvasive NIRS and common invasive approaches in observational studies. Our specific objectives were to compare Cohen’s d between NIRS rScO2 and ScvO2 as well as SjvO2 and explore the correlation and consistency between NIRS rScO2 and ScvO2 and SjvO2. We hypothesized that the validity of NIRS is comparable to that of ScvO2 and SjvO2 measurements. We presented the following article in accordance with the MOOSE reporting checklist (available at https://tp.amegroups.com/article/view/10.21037/tp-22-345/rc) (20).

Methods

Search strategy

A literature search was conducted by 2 investigators independently from the inception of the relevant databases until February 2022. The electronic bibliographic databases of PubMed, The Cochrane Library, and Embase were searched. We searched these databases using the search strategies described in Supplementary Appendix 1, Appendix 2, and Appendix 3. All the retrieved results were crosschecked by Y Ma, L Zhao.

Inclusion criteria

Types of studies

Cross-sectional analyses.

Types of participants

CHD pediatric patients younger than 18 years of age undergoing cardiac surgery or cardiac catheterization.

Types of examination method

(I) Cerebral oxygen saturation of patients was monitored perioperatively by NIRS and SjvO2. (II) Patients were monitored perioperatively by NIRS and ScvO2 as an assessment of cerebral oxygen saturation. (III) All included studies provided details of the NIRS devices used.

Statistics

The included studies reported mean difference (MD), standard deviation (SD), and correlation coefficient (r) values between rScO2 and SjvO2 and between rScO2 and ScvO2.

Study selection

After removal of duplicate papers, the remaining articles retrieved in the databases were screened through the titles and abstracts according to the inclusion criteria. Editorials, reviews, animal experiments, commentaries, conference papers, non-English language articles, abstracts, unpublished articles, and irrelevant articles were excluded. Articles with unobtainable original texts were also excluded. Full text articles were assessed for eligibility, and articles that lacked the data mentioned stipulated by the inclusion criteria were excluded. The screening was performed manually and independently by 2 reviewers, and differences were resolved through discussion and consensus.

Data extraction

The following data were extracted from the included studies: author, publication year, study design, number of patients, data points, median age, types of treatment, diagnosis, sites of NIRS, types of NIRS device, types of venous oxygen saturation measurement, MD, SD, r values between 2 methods, and 95% limits of agreement (LOA). Data extraction was conducted by 4 experienced investigators.

As the MD and SD were not provided in some studies, we extracted rScO2, ScvO2, and SjvO2 values from scatter diagrams created by MATLAB® (MathWorks, Inc., Natick, MA, USA) instead of contacting authors and then calculated the MD and SD between rScO2 and ScvO2 and between rScO2 and SjvO2. To standardize effect size, we used Cohen’s d to evaluate the difference between the 2 methods. Cohen’s d was calculated as the MD divided by the SD (21). Most studies used Pearson’s correlation coefficients to describe the correlation between the 2 methods. For several studies that used Spearman’s correlation coefficients, we converted the Spearman’s correlation coefficients into Pearson’s correlation coefficients for the sake of uniformity (22). Before calculating the pooled Pearson’s correlation coefficient, we transformed the correlation coefficients into Fisher’s Z, as the variance or standard error was closely related to the r values.

Methodological quality assessment

Considering that all the included articles were cross-sectional studies, we used the 11-item checklist for methodological quality assessment of cross-sectional studies recommended by the Agency for Healthcare Research and Quality (AHRQ) (23). When the answer was ‘NO’ or ‘UNCLEAR’, the item was scored ‘0’; if it was answered ‘YES’, then the item was scored ‘1’. Studies rated as 8–11 scores were regarded as high-quality studies; scores of 4–7 indicated intermediate-quality; and scores of 0–3 suggested low-quality (24).

Statistical analysis

We divided all the studies into rScO2 versus ScvO2 and rScO2 versus SjvO2 groups to estimate the difference and correlation among methods. For continuous variables, we evaluated the pooled Cohen’s d and r values with their 95% confidence intervals (95% CIs) in the rScO2 vs. ScvO2 and rScO2 vs. SjvO2 groups, respectively, using the Stata 15.0 software (Stata Corporation, College Station, TX, USA). We then performed heterogeneity analyses by I2 test with Stata under a fixed effects model. A score of 25%

Results

The search and methodological quality assessment

Based on our search strategy, 1,590 articles were retrieved from the PubMed, Cochrane Library, and Embase databases (). The number of papers remaining after removing duplicates was 1,115. A total of 47 papers were left after we excluded editorials, reviews, animal experiments, comments, conference papers, non-English language articles, and irrelevant articles through reading of the titles and abstracts. After full text screening of the remaining articles, and additional 24 articles were excluded. Two articles lacked a description of NIRS devices; 8 articles included patients without CHD; 8 articles lacked a comparison between rScO2 and SjvO2 as well as ScvO2; and data in 6 articles could not be extracted. A total of 23 studies with 997 children from 1995 to 2022 were eligible for inclusion (8,10,27-47). The search flowchart of this meta-analysis is shown in . Since we excluded non-English language articles, there was language bias in this study.

Figure 1

Search flowchart.

Regarding the methodological quality assessment, the scores of the included studies as assessed by the 11-item checklist recommended by AHRQ are given in . Among them, 4 studies were of high quality, and the remainder were of intermediate quality.

Table 1

Characteristics of the included studies

Author	Year	Study design	Cases	Data points	Median age (range)	Types of treatment	Diagnosis [n]	Sites of NIRS	NIRS devices	Types of venous oxygen saturation	95% LOA (%)	Quality scores
Yoxall (27)	1995	Cross-sectional study	15	150	2 (0.3–14.0) y	Cardiac catheterization	PDA [3], VSD [4], after corrective surgery [3], PH [1], complex cyanotic banalities [4]	Right fronto-temporal	NIRO 500 (Hamamatsu Photonics, Japan)	SjvO2	−18.8 to 20.5	6
Daubeney (28)	1996	Cross-sectional study	40	147	4.5 (0.04–14.5) y	Cardiac catheterization or cardiac surgery under CPB	Acyanotic and cyanotic forms of congenital heart disease	Bilateral forehead	INVOS 3100 (Somanetics Corp., Troy, MI, USA)	SjvO2	NA	6
Nagdyman (10)	2004	Cross-sectional study	43	70	2.8 (0.02–16.8) y	Cardiac corrective surgery	ASD [16], VSD [13], complete endocardial cushion defect [1], AS [6], MS [3], CoA [1], complex congenital heart defect [9], HOC [1], PS [2]	Supra-orbital region	NIRO 300 (Hamamatsu Photonics, Japan)	ScvO2	NA	6
Nagdyman (29)	2005	Cross-sectional study	60	60	4.4 (0.1–16.0) y	Cardiac Catheterization	ASD [14], VSD [6], complete endocardial cushion defect [7], AS [3], MS [2], CA [2], complex congenital heart defects [8], HCM [2], TGA [6], TOF [10]	Right forehead	NIRO 300 (Hamamatsu Photonics, Japan)	SjvO2	−15.3 to 11.7	8
Shimizu (30)	2005	Cross-sectional study	5	14	0.6 (0.2–1.3) y	Cardiac surgery under CPB	TGA [1], VSD [1], AVSD [1], TOF [2]	Forehead	NIRO 300 (Hamamatsu Photonics, Japan)	SjvO2	−17.8 to 11.0	7
Tortoriello (31)	2005	Cross-sectional study	20	100	0.8 (0.4–8.0) y	Reparative or palliative cardiac surgery under CPB	HLHS [5], PA [6], BTS [7], BDG [4], PHTN [3], CAVC [3]	Right, left, or bilateral forehead	INVOS 5100 (Somanetics Corp., Troy, MI, USA)	ScvO2	−10.1 to 13.4	6
Bhutta (32)	2007	Cross-sectional study	29	52	8.6 (1.3–17.0) y	Cardiac catheterization, myocardial biopsy	post-orthotopic heart transplant [29]	Forehead	INVOS 5100B (Somanetics Corp., Troy, MI, USA)	ScvO2	−12.5 to 16.3	7
Kirshbom (33)	2007	Cross-sectional study	20	20	0.6 (NA) y	Cardiac catheterization	HLHS [8], PA or TA [10], DORV variants [2]	Bilateral forehead	INVOS (Somanetics Corp., Troy, MI)	ScvO2	NA	6
McQuillen (34)	2007	Cross-sectional study	70	NA	0.3 (0.01–1.2) y	Cardiac surgery under CPB	AVSD [10], TGA [8], TOF [8], VSD [5], ASD [3], PA [3], truncus arteriosus [2], TAPVD [2], PS [1], PDA [1], AI [1], CoA [4], IAA [1], TA/PA [5], RAI [2], TA [1], Ebstein’s anomaly [1], HLHS [7], uAVSD [2], HRV [2]	Left forehead	INVOS 5100 (Somanetics Corp., Troy, MI, USA)	ScvO2	−25.6 to 23.5	7
Knirsch (35)	2008	Cross-sectional study	60	120	4.3 (0.2–16.0) y	Cardiac catheterization	CHD [60]	Right forehead	INVOS 5100 (Somanetics Corp., Troy, MI, USA)	ScvO2	−15.5 to 15.9	6
										SjvO2	−18.6 to 17.4
Nagdyman (36)	2008	Cross-sectional study	30	3660	3.1 (0.1–16.0) y	Cardiac catheterization	ASD [6], VSD [2], PDA [1], heart transplantation [5], complex congenital heart defects [8], cardiomyopathy [2], TGA [1], TOF [6]	Left forehead	NIRO 200 (Hamamatsu Photonics, Tokyo, Japan)	ScvO2	−20.1 to 10.3	8
										SjvO2	−20.1 to 15.7
Ranucci (37)	2008	Cross-sectional study	15	117	1.5 (0.02–7.0) y	Cardiac surgery under CPB	ASD [2], VSD [3], TOF [5], TAPVR [1], CPC [2], AS [2]	Forehead	INVOS (Somanetics Corp., Troy, MI, USA)	ScvO2	−15.2 to 26.4	8
Ricci (38)	2010	Cross-sectional study	100	890	13.0 (NA) days	Cardiac surgery under CPB	TGA [39], HLHS and UVH [26], TOF [24], other diagnosis [11]	Right forehead	INVOS 5100 (Somanetics Corp., Troy, MI, USA)	ScvO2	−25.0 to 25.0	5
Ginther (39)	2011	Cross-sectional study	8	690	8.1 (2.0–15.0) y	Bicaval cardiac catheterization	PS [2], AS [2], VSD [1], ASD [1], MR [1], RV-PA conduit insufficiency and stenosis [1]	Right forehead	INVOS 5100 (Somanetics Corp., Troy, MI, USA)	ScvO2	NA	6
Marimón (40)	2012	Cross-sectional study	20	605	4.5 (0.02–16.3) y	Cardiac surgery under CPB	TOF [4], atrioventricular canal defect [4], VSD [2], TA [1], aortic vascular ring [1], AS [1], CoA [1], ASD [1], HLHS [3], TOF with MAPCA and VSD [2]	Forehead	INVOS 5100 (Somanetics Corp., Troy, MI, USA)	ScvO2	NA	6
Hansen (41)	2013	Cross-sectional study	32	NA	0.2 (0.1–0.80) y	Superior cavopulmonary anastomosis with CPB	HLHS [26], MA [2], TGA [1], DILV [1], DORV [1], AS [1]	Midline forehead	INVOS 5100 (Somanetics Corp., Troy, MI, USA)	ScvO2	−17.9 to 19.8	5
Moreno (8)	2013	Cross-sectional study	23	980	12.0 (2–46) d	Open heart surgery	HLHS [8], TGA [6], TAPVC [5], IAA with VSD [2], multiple VSD [1], TA [1]	Forehead	INVOS 5100 (Somanetics Corp., Troy, MI, USA)	ScvO2	−17.2 to 38.1	7
Iodice (42)	2014	Cross-sectional study	10	36	2.2 (0.1–8.5) y	Cardiac surgery under CPB	MR [1], PA with VSD [1], TGA [2], IAA [1], TA [2], TOF [1], MS [1], univentricular heart with HAA [1]	Forehead	INVOS 5100 (Somanetics Corp., Troy, MI, USA)	ScvO2	−13.0 to 10.0	5
Kussman (43)	2017	Cross-sectional study	57	NA	4.8 (NA) y	Cardiac catheterization	Acyanotic or cyanotic congenital heart disease [57]	Forehead	FORE-SIGHT (CASMED, Inc., Branford, CT, USA)	SjvO2	−7.7 to 9.9	7
Naguib (44)	2017	Cross-sectional study	34	361	3.5 (NA) y	Cardiac surgery utilizing CPB and require bicaval cannulation	VSD [24], ASD [6], ASD + VSD [12], AVSD [9], BDG [3], Fontan [3], valve replacement [18], TOF [12], subaortic membrane resection [3], cortriatriatum [6], Ebstein’s anomaly [3], aortic arch augmentation [3]	Right and left forehead	FORE-SIGHT (P/N 01-06-2030C; CASMED, Inc., Branford, CT, USA)	SjvO2	NA	8
Rescoe (45)	2017	Cross-sectional study	73	520	7.9 (NA) d	Stage 1 palliation under CPB	HLHS [73]	Forehead	FORE-SIGHT (CASMED, Inc., Branford, CT, USA)	ScvO2	−17.6 to 34.1	7
Gagnon (46)	2020	Cross-sectional study	47	506	11.3 (NA) d	Stage 1 palliation under CPB	HLHS [38], DILV [4], TA [3], others [2]	Forehead	FORE-SIGHT ELITE (CASMED, Inc., Branford, CT, USA)	ScvO2	−10.6 to 38.6	6
Terada (47)	2022	Cross-sectional study	186	NA	6.0 (NA) y	Cardiac catheterization	ASD [16], VSD [54], AVSD [9], AS [8], PA [2], PDA [4], TAPVR [3], TGA [15], TOF [16), DORV [1], Single right ventricle [14], Single left ventricle [18], others [10]	Forehead	INVOS 5100C (Covidien, Tokyo, Japan)	ScvO2, SjvO2	NA	4

NIRS, near-infrared spectroscopy; LOA, limits of agreement; PDA, patent ductus arteriosus; VSD, ventricular septal defect; PH, pulmonary hypertension; SjvO2, jugular bulb venous oxygen saturation; CPB, cardiopulmonary bypass; ASD, atrial septal defect; AS, aortic stenosis; MS, mitral stenosis; ScvO2, central venous oxygen saturation; HOC, hypertrophic obstructive cardiomyopathy; PS, pulmonary stenosis; HCM, hypertrophic cardiomyopathy; TGA, transposition of great arteries; TOF, tetralogy of Fallot; AVSD, atrioventricular septal defect; NA, not available; HLHS, hypoplastic left heart syndrome; PA, pulmonary atresia; BTS, Blalock-Taussig shunt; BDG, bidirectional Glen; PHTN, pulmonary hypertension; CAVC, complete atrioventricular canal; TA, tricuspid atresia; DORV, double outlet right ventricle; TAPVD, total anomalous pulmonary venous drainage; AI, aortic insufficiency; CoA, coarctation of aorta; IAA, interrupted aortic arch; RAI, right atrial isomerism; uAVSD, unbalanced atrioventricular septal defect; HRV, hypoplastic right ventricle; CHD, congenital heart disease; TAPVR, total anomalous pulmonary venous return; CPC, cavo-pulmonary connection; MA, mitral atresia; DILV, double inlet left ventricle; TAPVC, total anomalous pulmonary venous connection; INVOS, a type of NIRS device; NIRO, a type of NIRS device; FORE-SIGHT, a type of NIRS device.

Descriptors

The studies included in this meta-analysis are listed in . All 23 included articles were cross-sectional studies. There were a total of 997 participants in the 23 included studies. The sample sizes ranged from 5 to 186. The ages of patients were spread over a broad range in each study. The mean age ranged from 7.9 days to 8.6 years. Studies were carried out in diverse countries, including the UK, USA, Germany, Italy, Argentina, Canada, and Japan. The treatment for CHD involved 2 approaches: 9 studies used cardiac catheterization; 13 studies used cardiac surgery under cardiopulmonary bypass; and 1 study included patients undergoing cardiac surgery or cardiac catheterization. The devices of NIRS mentioned in the involved studies were the NIRO 500/300/200 (Hamamatsu Photonics, Hamamatsu, Japan), the INVOS 3100/5100 (Somanetics Corp., Troy, MI, USA), the INVOS (Covidien, Boulder, CO, USA), the INVOS 5100C (Covidien, Tokyo, Japan), the FORE-SIGHT (CAS Medical Systems, Inc, Branford, CT, USA), and the FORE-SIGHT ELITE (CAS Medical Systems, Inc., Branford, CT, USA). The INVOS 5100C and FORE-SIGHT were approved by the FDA for use in pediatric patients. The NIRS probes were placed over the forehead of patients.

Main outcomes

The pooled Cohen’s d with 95% CIs as well as the I2 for the rScO2 vs. ScvO2 group and the rScO2 vs SjvO2 group were performed under a random effects model ().

Figure 2

Cohen’s d between rScO2 and ScvO2. CI, confidence interval; rScO2, cerebral regional oxygen saturation; ScvO2, central venous oxygen saturation.

Figure 3

Cohen’s d between rScO2 and SjvO2. CI, confidence interval; rScO2, cerebral regional oxygen saturation; SjvO2, jugular venous oxygen saturation.

The pooled Cohen’s d was 0.06 (95% CI: −0.16 to 0.28; ) in the rScO2 versus the ScvO2 group and 0.03 (95% CI: −0.25 to 0.31; ) in the rScO2 versus the SjvO2 group, which indicated no statistically significant difference between the Cohen’s d of rScO2 measured by NIRS and ScvO2 as well as SjvO2 measured by invasive approaches. The heterogeneity for the rScO2 vs. ScvO2 group was high (I2=76.0%, P<0.001), and the heterogeneity for the rScO2 vs. the SjvO2 group was intermediate (I2=73.6%, P<0.001). The fixed effects model showed that the rScO2 was positively correlated with the ScvO2 (Fisher’s Z =0.67, 95% CI: 0.60 to 0.74; ) with an r value of 0.58 (95% CI: 0.54 to 0.63) after Fisher’s Z transformation. Accordingly, rScO2 and SjvO2 were correlated (Fisher’s Z =0.70, 95% CI: 0.61 to 0.79; r=0.60, 95% CI: 0.54 to 0.66; ). The heterogeneity was within the acceptable range in the rScO2 vs. ScvO2 group (I2=36.7%, P=0.065) and the rScO2 vs. SjvO2 group (I2=12.7%, P=0.328).

Figure 4

Fisher’s Z between rScO2 and ScvO2. CI, confidence interval; rScO2, cerebral regional oxygen saturation; ScvO2, central venous oxygen saturation.

Figure 5

Fisher’s Z between rScO2 and SjvO2. CI, confidence interval; rScO2, cerebral regional oxygen saturation; SjvO2, jugular venous oxygen saturation.

Sixteen of the included studies performed Bland-Altman’s analysis to discuss the consistencies between rScO2 and SjvO2 or rScO2 and ScvO2. The 95% LOA of each study is summarized in , and it was fairly wide in most studies. Among them, 3 studies postulated that a difference of ±5% was considered clinically acceptable (8,29,36). Apparently, consistencies between rScO2 and SjvO2 or rScO2 and ScvO2 were poor in these studies with this standard. With one exception, Shimizu et al. declared that rScO2 and SjvO2 showed a reasonable consistency because the difference in the two parameters within the limits of ±10% accounted for 86% (30). Notably, this study only included 5 patients with 14 data points which was unrepresentative.

Publication bias

To assess publication bias, we drew funnel plots for the meta-analyses of Cohen’s d and Fisher’s Z only in the rScO2 vs. ScvO2 group (). Since there were fewer than 10 studies in the rScO2 vs. SjvO2 group, we did not analyze publication bias in this group. Egger’s tests revealed no evidence of publication bias (P=0.76 for Cohen’s d analysis between rScO2 and ScvO2; P=0.067 for Fisher’s Z analysis between rScO2 and ScvO2).

Figure 6

Funnel plot with pseudo 95% confidence limits for Cohen’s d and Fisher’s Z. (A) Cohen’s d between rScO2 and ScvO2; (B) Fisher’s Z between rScO2 and ScvO2. rScO2, cerebral regional oxygen saturation; ScvO2, central venous oxygen saturation.

Sensitivity analyses, meta-regression, and subgroup analyses

Removal of any one of the studies alone did not significantly affect the overall results, which suggested that the results of this meta-analysis were stable (Figures S1-S4). In view of the low heterogeneity of Fisher’s Z analyses, further analysis was not carried out. As relatively significant heterogeneity existed in Cohen’s d analyses in both the rScO2 versus ScvO2 and the rScO2 versus SjvO2 groups, we conducted meta-regression to explore the cause. As a result, the types of NIRS devices were related to high heterogeneity of Cohen’s d in the rScO2 versus ScvO2 group (P<0.001), whereas the number of patients (P=0.67) and the types of treatment (P=0.46) were not related. The meta-regression result of Cohen’s d in the rScO2 versus SjvO2 group also indicated that the types of NIRS devices were responsible for the heterogeneity (P<0.001), and that the number of patients (P=0.98) and types of treatment (P=0.88) were not responsible.

In the analyses of the rScO2 versus ScvO2 group, significant heterogeneity was noted between the 3 subgroups of types of NIRS devices (P<0.001), and, among them, the FORE-SIGHT subgroup had the highest heterogeneity (I2=67.3%, P<0.001). After removal of studies using the FORE-SIGHT device (45,46), the I2 was reduced to 43.9% (P=0.04). Similarly, in the analyses of the rScO2 versus SjvO2 group, heterogeneity between different types of NIRS devices was significant (P=0.03, and the heterogeneity of the FORE-SIGHT subgroup was high (I2=79.9%, P=0.03). The I2 decreased to 0.0% (P=0.76) when studies of the FORE-SIGHT subgroup were removed (43,44). No significant heterogeneity existed between the number of patients and the types of treatment subgroups. The results of subgroup analyses can be found in .

Table 2

Subgroup analysis

Subgroup	Number of studies	Cohen’s d, (95% CI)	P value	I2	Heterogeneity between subgroups, P
Number of patients
rScO2 vs. ScvO2					0.264
Patients ≤20	6	−0.10 (−0.40, 0.20)	0.496	6.30%
Patients >20	11	0.13 (−0.15, 0.40)	0.362	83.40%
rScO2 vs. SjvO2					0.508
Patients ≤20	2	−0.17 (−0.80, 0.45)	0.583	0.00%
Patients >20	7	0.06 (−0.25, 0.37)	0.7	79.80%
Types of treatment
rScO2 vs. ScvO2					0.108
Cardiac surgery under CPB	11	0.17 (−0.15, 0.49)	0.294	80.60%
Cardiac catheterization	6	−0.13 (−0.32, 0.05)	0.166	17.10%
rScO2 vs. SjvO2					0.639
Cardiac surgery under CPB	2	0.17 (−0.29, 0.64)	0.467	2.40%
Cardiac catheterization	6	0.03 (−0.34, 0.40)	0.869	82.50%
Types of NIRS devices
rScO2 vs. ScvO2					<0.001
NIRO	2	−0.56 (−0.89, −0.23)	0.001	0.00%
INVOS	13	−0.00 (−0.15, 0.14)	0.952	25.30%
FORE-SIGHT	2	0.86 (0.38, 1.33)	<0.001	67.30%
rScO2 vs. SjvO2					0.032
NIRO	4	−0.29 (−0.55, −0.03)	0.031	0.00%
INVOS	3	−0.03 (−0.20, 0.13)	0.699	0.00%
FORE-SIGHT	2	0.63 (−0.05, 1.32)	0.071	79.90%

NIRS, near-infrared spectroscopy; CI, confidence interval; rScO2, cerebral regional oxygen saturation; ScvO2, central venous oxygen saturation; SjvO2, jugular venous saturation; CPB, cardiopulmonary bypass; NIRO, a type of NIRS device; INVOS, a type of NIRS device; FORE-SIGHT, a type of NIRS device.

Discussion

This systematic review and meta-analysis assessed the validity of NIRS in measuring rScO2 in children with CHD undergoing surgery and the interchangeability between NIRS and common invasive approaches that measure SjvO2 and ScvO2. After removing studies that met the exclusion criteria, such as reviews, animal experiments, non-English language articles, unpublished articles and irrelevant articles, we included 23 studies of intermediate to high quality. No evidence of publication bias was observed. No statistically significant difference was found between rScO2 and SjvO2 and between rScO2 and ScvO2 in the pediatric patients with CHD and NIRS rScO2 was positively correlated with the saturation of jugular bulb blood and central venous blood. The results indicated that NIRS exhibited comparative accuracy to a certain extent. Nevertheless, significant heterogeneity was found in the results, which could be attributed to variation in the diverse NIRS devices, especially the FORE-SIGHT device, used in the measurement of rScO2. This shortcoming detracted from the validity of the research. Besides, in terms of Bland-Altman’s analysis, the 95% LOAs of most studies were wide, which suggested that the interchangeability between NIRS and SjvO2 as well as ScvO2 was up for debate.

Through the application of sensitivity analyses, meta-regression, and subgroup analyses to investigate the cause of heterogeneity, we learned that different NIRS devices calculate rScO2 through different algorithms. The NIRO device calculates the tissue oxygenation index (TOI) through spatially resolved spectroscopy by the equation TOI = HbO2/(HbO2 + HHb) × 100 (48,49). The INVOS device calculates rScO2 according to the ratio of oxyhemoglobin to total hemoglobin (47,50). In the FORE-SIGHT device, cerebral mixed arterial-venous oxygen saturation is measured by the different absorption of oxygenated and deoxygenated hemoglobin to near-infrared light, thus detecting rScO2 by the formula rScO2 = 0.3 SaO2 + 0.7 ScvO2 or rScO2 = 0.3SaO2 + 0.7 SjvO2 according to a 30% arterial to 70% venous ratio (45,51,52). The effectiveness of monitoring cerebral oxygen saturation varies from device to device. Theoretically, NIRS monitors oxygenation saturation at the tissue and cellular level; at the cellular level, cytochrome aa3 is a key variable to measure changes in mitochondrial oxygenation (53). However, the INVOS 3100 device cannot provide information about this parameter on account of technical difficulties, whereas the NIRO 500 device not only measures the level of cytochrome aa3 but also monitors the oxygenation state of hemoglobin (54). In addition, Naguib et al. (44) reported that the FORESIGHT device had higher sensitivity, whereas the INVOS device had better specificity. The use of different algorithms among commercial NIRS devices makes comparing the rScO2 of the devices difficult, and industry standards among devices are lacking.

This study has reference significance for the application of NIRS in clinical practice to monitor cerebral oxygen saturation properly in children with CHD undergoing surgery. The results of our study revealed that rScO2 measured by NIRS reflected the SjvO2 and ScvO2 measured by co-oximetry in the perioperative period of children with CHD to a certain extent. However, the agreement between NIRS and invasive oxygen saturation measurements was below expectation, and the LOA was wide, which indicated that NIRS cannot replace SjvO2 and ScvO2. The NIRS method has inherent limitations that restrict its application. As a monitoring tool, NIRS is affected by factors such as hemoglobin, blood volume, arterial blood pressure (55), cardiac output, arterial oxyhemoglobin saturation, the position of the probe head, body position, and vasoactive drugs (19). Due to the discrepancy in sensor technology, near infrared wavelength, fixed arterio-venous (A/V) ratio, and reference values of different NIRS devices, each manufacturer lacks large sample data to determine reference values of cerebral oxygen, which is the major limitation in NIRS validation. Studies have found that the A/V ratio differed among subjects, which implies that a fixed A/V ratio is not appropriate for validating the technology (56).

Despite the limitations of NIRS, the ability of providing noninvasive continuous monitoring for cerebral oxygen saturation is of importance and the potential to identify cerebral ischemic events is prominent. Venous thrombosis, infection, and extracerebral contamination are common and intractable problems of invasive measurements, which can be avoid by using NIRS (57,58). NIRS has been extensively used in pediatric patients undergoing cardiovascular operations with a high risk of compromised cerebral perfusion (59). Cruz et al. (60) reported that the variance of peripheral capillary oxygen saturation (SpO2) was larger than that of NIRS, meaning that NIRS monitoring was more stable and better at predicting events. Similar results were replicated in a 24-hour observational study that found NIRS superior to SpO2 in monitoring hypoxia and ischemic events (61). Moerman et al. (62) demonstrated that NIRS monitoring identified compromised cerebral perfusion despite hemodynamic measurements being normal. However, Robust studies, including randomized clinical trials, are needed to prove the clinical benefit of NIRS. There is still no evidence that early monitoring of cerebral oxygenation during pediatric surgery under general anesthesia improves prognosis after surgery (63). A systematic review of surgical treatment of pediatric CHD concluded that the benefit and cost-effectiveness of NIRS monitoring and management capable of improving short-term clinical neurological outcomes has not yet been demonstrated (64). Zheng et al. (65) reported that the correlation between decreased rScO2 and postoperative neurological complications was low, and improving rScO2 desaturation in attempts to prevent stroke, delirium, or postoperative cognitive dysfunction could not be supported by available evidence. Future studies should look at solving the heterogeneity problems, that is, setting uniform reference standards among NIRS devices. In addition, more research should attempt to determine whether perioperative monitoring with NIRS can improve postoperative outcomes.

This systematic review and meta-analysis had some limitations. First, the proposed data extraction and analysis method may be inadequate, as the discrepancy in data acquisition led to downstream data issues. We used MATLAB to extract data from images in articles when the MD and SD were not provided. In that case, the ordinate and abscissa values of the scatter points were measured manually, which may have introduced errors. When the scattered points were stacked together, accurately identifying them separately was difficult, and accurately measuring the diameter of the scattered points themselves was challenging. The second limitation was the relatively high heterogeneity in the analyses of Cohen’s d. Under normal circumstances, combining effect size in the case of high heterogeneity due to the broad distribution and variation of sample characteristics is not recommended. However, since we discussed the cause of high heterogeneity through sensitivity analyses, meta-regression, and subgroup analyses, the results were still of reference value. Thirdly, the lack of randomized controlled trials reduced the validity of the study. More high-quality studies are required in this field.

Conclusions

This systematic review and meta-analysis revealed that rScO2 measured by NIRS reflects SjvO2 and ScvO2 recorded by invasive measurements in the perioperative period of children with CHD to some extent. Nevertheless, wide LOA indicated that the interchangeability between NIRS and invasive oxygen saturation measurements was below expectation. Despite the technical limitations of NIRS, it provides a non-invasive, convenient approach for the real-time monitoring of hypoxia, ischemia, and changes in cerebral perfusion. More evidence is needed to prove the possible clinical benefits of NIRS in monitoring cerebral oxygen saturation in children with CHD.

The article’s supplementary files as