Cerebral oximetry uses near-infrared spectroscopy (NIRS) to provide real-time non-invasive interrogation of regional cerebral oxygen saturation (rSO2) and has become an increasingly popular intraoperative monitoring technique.1 Measuring rSO2 in a representative volume of frontal cortex brain tissue (and assuming stable metabolic suppression of the brain under anesthetic conditions), it has been seen as a surrogate of cerebral blood flow and thus as a useful technology to detect cerebral hypoperfusion.1,2 It is thought to be particularly beneficial in the perioperative setting when hemodynamic fluctuations often occur that can lead to postoperative complications such as cognitive impairment or delirium.2,3,4,5,6 Anesthesiologists have utilized cerebral oximetry monitoring in an attempt to optimize both blood pressure and oxygen delivery to maintain adequate cerebral perfusion and decrease the incidence of these neurocognitive complications.2,5

Several observational studies have pointed to the predictive value of cerebral oxygenation monitoring for both short- and long-term functional outcomes.7,8,9 It has been suggested that in cardiac surgery patients, in addition to those following cardiac arrest or with a diagnosis of sepsis, rSO2 (< 60%) may be associated with an increased risk of adverse outcomes.10,11,12 An additional study in aortic arch surgery patients concluded that reduced intraoperative cerebral oxygen saturation was not only associated with extended hospital stay, but also increased overall hospital costs.13 Furthermore, given an inherent increase in the physical and financial burden of patient care associated with cognitive dysfunction following surgery,14 additional efforts directed towards cerebral monitoring and postoperative cognitive dysfunction (POCD) prevention are warranted. Although cerebral oximetry monitoring has been available as a clinical tool for two decades, little consensus exists regarding the role of cerebral oximetry-based management in the perioperative period. Several prospective trials have attempted to assess the impact of cerebral oximetry on postoperative cognitive outcomes; however the results have remained conflicting.

The purpose of this systematic review and meta-analysis was to determine the overall beneficial effect of cerebral oximetry on select outcomes after surgery.

Methods

This meta-analysis followed the guidelines outlined by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement.15 It was also registered on the International Prospective Systematic Reviews Registry database (PROSPERO 2017: CRD42017057293) on 14 February 2017.

Search strategy

The MEDLINE/PubMed, EMBASE, Cochrane Library, Scopus, and Google Scholar databases were searched from inception to 2 December 2017 for randomized-controlled trials (RCTs) assessing the effects of intraoperative cerebral oximetry monitoring on postoperative outcomes following cardiac and non-cardiac surgery. There was no restriction on language. In addition, article citations were reviewed to ensure inclusion of relevant studies not captured in our initial literature search. The clinicaltrials.gov registry was also searched to evaluate for any ongoing RCT where results might be expected to be published in the near future. Two authors (A.Z.V. and R.J.H.) reviewed the literature and screened the abstracts independently. Full-text articles that met the inclusion criteria were reviewed for detailed comprehension and further assessment of the quality and risk of bias. All disagreements between reviewers in the selection and evaluation processes were resolved by discussion with a third reviewer (M.C.G.). All demographic data, including year of publication, sample size, intervention algorithm, type of surgery, anesthetic management, and specified outcomes, were abstracted in a predefined manner.

Eligibility criteria

We limited our meta-analysis to RCTs of adult patients (age > 18 yr) who underwent either cardiac or non-cardiac surgery. The intervention group was monitored with NIRS (interventions specified below) while the control group was not.

Interventions considered

Management guided by the use of intraoperative cerebral oximetry was considered the primary intervention and was triggered by evidence of cerebral oxygen desaturation. Specific interventions included the use of fluids and/or vasopressors for hypotension, an increase in pump flow to maintain the cardiac index above 2 L·m−2·min−1, changes in ventilatory parameters (i.e., optimizing the partial pressure arterial oxygen and carbon dioxide), and blood transfusion if anemic. Thresholds for intervention were generally an rSO2 < 55-60% or rSO2 < 75% of baseline. Prior experimental work has suggested that critical neurologic deficits are more likely to occur with more than a 30% reduction in rSO2,16 and to establish a rational approach, our search also involved the selection of studies wherein a more conservative intervention threshold was utilized.17

Outcomes

The primary outcome in this meta-analysis was the incidence of POCD as defined by the individual studies. Most articles used a combination of standardized assessments of cognitive functions such as the Mini-Mental Status Examination (MMSE), grooved pegboard, anti-saccadic eye movement, color trail, and Montreal Cognitive Assessment. Secondary outcomes included intensive care unit (ICU) length of stay (LOS), overall hospital LOS, as well as the incidence of total transfusion, delirium, surgical site infection, cardiac complications, and mortality. Individual study definitions were also used for secondary outcomes. The time interval to evaluate delirium and POCD outcomes was within one week after surgery.

Assessment of methodologic quality and quality of evidence

Methodologic quality assessment was performed using the Cochrane risk of bias tool for randomized studies.18 Each study was assessed based upon seven domains of potential bias (random sequence generation, allocation concealment, blinding of intervention, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias). The overall risk of bias of individual studies was classified as high if at least one domain was determined at high risk or if there were more than two domains of unclear risk, moderate if at least two domains were determined at unclear risk, and low if all the domains were determined at low risk. The quality of the evidence provided in this meta-analysis was also assessed using five levels of evidence, ranging from level I to III with three subcategories in level II, as previously reported.19

Statistical analysis

Initially, an exploratory qualitative analysis was conducted to describe the characteristics of the studies included in this meta-analysis. The incidence of POCD was extracted as a dichotomous variable (present or absent) and compared using risk ratios (RR) with their respective 95% confidence intervals (CI). We used forest plots to illustrate the estimations and overall effect sizes with pooled RR represented as a solid diamond at the bottom of the forest plot. Outcomes presented as continuous variables were compared using the standardized mean difference (SMD). In cases of publication of median values with their ranges, we converted these measures into mean and standard deviations (SD) using the method of Wan et al.20 In cases where 95% CI of mean values was included, the SD was calculated using a standard formula. Predetermined subgroup analyses were performed based upon type of surgery (cardiac versus non-cardiac surgery) and type of cerebral oximetry-based intervention. Sensitivity analysis was performed based upon overall study quality (high or moderate versus low) as determined by quality of evidence assessment.

Heterogeneity (I2) was assessed using the correspondent Chi-squared test (I2 < 50% and I2 > 50% were considered insignificant and significant heterogeneity, respectively). Publication bias was calculated using the Begg’s and Egger’s tests21 with funnel plots constructed to represent any tendency for publishing in favour of positive effects. Significant publication bias was considered when there was asymmetry in the funnel plot and a statistically significant bias coefficient was noted on the Beggs’s test.21 P < 0.05 was considered statistically significant for all the statistical tests. All analyses were performed using a random-effect model (DerSimonian and Laird method).22 All statistics were performed using Review Manager 5.3 (Cochrane Collaboration, Oxford, UK) or Stata version 13.0 (Stata, College Station, TX, USA).

Results

Literature search and selection

Our initial search yielded 3,177 records. After excluding duplicate studies, we screened a total of 1,454 titles and abstracts. Of these, 26 full-text articles met the full inclusion criteria. Two RCTs were excluded because of a lack of demographic and/or outcomes data.23,24 An additional nine RCTs were excluded as they did not involve the target intervention comparison.25,26,27,28,29,30,31,32,33,34 Two additional RCTs were excluded because they were published as an abstract35,36 and one RCT was excluded because although it correlated anesthetic depth with cerebral oximetry, it did not detail the associated intervention.37 Finally, three additional trials were identified from reference lists of the articles included.7,38,39 Figure 1 outlines the full results of article selection. In total, 15 RCTs were included in this meta-analysis.7,17,38,39,40,41,42,43,44,45,46,47,48,49,50

Fig. 1
figure 1

PRISMA flow chart of the selection of studies

Full size image

Study characteristics

The Table summarizes the characteristics of the included studies. Ten RCTs included patients undergoing cardiac surgery (coronary artery bypass, valve replacement or repair),17,38,39,41,42,43,44,45,46,50 one RCT was in carotid endarterectomy surgery,39 two RCTs included only major abdominal surgery,40,47 and two RCTs included both arthroplasty and abdominal surgeries.7,48 A total of 2,057 patients (1,018 in the intervention group and 1,039 in control group) were included in the overall analysis. The intervention in 13 RCTs was the correction of cerebral oxygen desaturation (i.e., via modifying mechanical ventilation or administering vasopressors), of which seven RCTs followed an algorithm as outlined by Denault et al.17,39,42,44,49,50,51 while the remainder applied other individualized algorithms. The intervention in two RCTs was a combination of fluid administration and/or transfusion if rSO2 decreased by more than 20-25% below baseline.47 The Table also shows the definitions of cerebral desaturation of each study.

Table Characteristics of studies included in the analysis
Full size table

Primary outcome

Among the seven trials examining the primary outcome, management associated with the use of intraoperative cerebral oximetry was associated with a significant reduction in POCD at one week (Fig. 2A; RR, 0.60; 95% CI, 0.40 to 0.89; P < 0.001, I2 = 81%) compared with patients who did not receive therapy guided by cerebral oximetry. Subgroup analysis that included only trials involving cardiac surgery resulted in a similar association (RR, 0.55; 95% CI, 0.36 to 0.86; P = 0.009; I2 = 85%), but we found no significant association in non-cardiac surgery (RR, 0.79; 95% CI, 0.61 to 1.02; P = 0.07; I2 = 0%). Among the studies that did not follow the Denault et al. algorithm, the results again show a significant association between the use of cerebral oximetry to guide intervention and reduction in POCD (five RCTs; RR 0.6; 95% CI, 0.50 to 0.94; P = 0.02; I2 = 72%). The use of cerebral oximetry-driven interventions was not associated with a statistically higher MMSE at one week (Fig. 2B; SMD, 0.39; 95% CI, −0.03 to 0.80; P = 0.07; I2 = 79%) compared with controls. There was no evidence of publication bias in our analyses (Egger’s test bias = −0.05; P = 0.96). Sensitivity analysis revealed no significant differences in the overall analysis for either endpoint.

Fig. 2
figure 2

Forest plots illustrating A) the incidence of postoperative cognitive dysfunction and B) Mini-Mental State Examination score between intervention and control groups

Full size image

Secondary outcomes

The ICU LOS was examined in eight trials, all of which were conducted in cardiac surgery. Our results suggest that patients in the intervention group have significantly shorter lengths of ICU stay compared with the control group (Fig. 3; SMD, −0.21 hr; 95% CI, −0.37 to −0.05; P = 0.009; I2 = 48%). In subgroup analysis, we found that among the studies that followed the Denault et al. algorithm, there was a significant association with a reduction in ICU stay (five RCTs; RR, −0.31 hr; 95% CI, −0.46 to −0.16; P < 0.001; I2 = 0%). In contrast, there was no significant association among the studies that did not follow the Denault et al. algorithm (four RCTs; RR, −0.11 hr; 95% CI, −0.38 to 0.15; P = 0.40; I2 = 63%). Among the eight trials that reported on hospital LOS, pooled analysis found no significant difference between the groups (Fig. 4; SMD, −0.06 days; 95% CI, −0.18 to 0.06; P = 0.29; I2 = 0%).

Fig. 3
figure 3

Pooled effect of cerebral oximetry-guided management on the length of stay in the intensive care

Full size image
Fig. 4
figure 4

Pooled effect of cerebral oximetry-guided management on length of hospital stay

Full size image

Transfusion was examined in six trials, all of which were conducted in cardiac surgery. Patients monitored with intraoperative cerebral oximetry tended to have fewer blood transfusions but this did not reach significance compared with the control group (Fig. 4; RR, 0.88; 95% CI, 0.77 to 1.10; P = 0.05; I2 = 0%) (Fig. 5).

Fig. 5
figure 5

Pooled effect of cerebral oximetry-guided management on incidence of total red blood cell transfusion

Full size image

Four RCTs specifically assessed for postoperative delirium. There was no significant difference between groups in the incidence of postoperative delirium (Fig. 6; RR, 0.90; 95 % CI, 0.63 to 1.29; P = 0.57; I2 = 0%).

Fig. 6
figure 6

Pooled effect of cerebral-oximetry guided management on incidence of postoperative delirium

Full size image

The results of pooled analysis in the eight trials that reported on myocardial infarction suggested no significant difference between groups (Fig. 7; RR, 0.80; 95% CI, 0.42 to 1.52; P = 0.50; I2 = 0%). Subgroup analysis specific to cardiac (RR, 0.98; 95% CI, 0.46-2.06; P = 0.95; I2 = 0%) and non-cardiac (RR, 0.68; 95% CI, 0.09 to 5.40; P = 0.72; I2 = 0%) surgery yielded similar results.

Fig. 7
figure 7

Pooled effect of cerebral oximetry-guided management on incidence of postoperative myocardial infarction

Full size image

Six RCTs compared the surgical site infection rates between the groups. We found no difference in the rate of infection (Fig. 8; RR, 0.91; 95% CI, 0.65 to 1.27; P = 0.58; I2 = 0%) between groups. Mortality within 30 days of surgery was comparable between the intervention and control group (RR, 0.73; 95% CI, 0.34 to 1.58; P = 0.42; I2 = 0%) (Fig. 9).

Fig. 8
figure 8

Pooled effect of cerebral oximetry-guided management on incidence of postoperative surgical site infection

Full size image
Fig. 9
figure 9

Summary of the risk of bias assessment of each study

Full size image

Methodologic quality assessment

The electronic supplemental material (ESM) shows the assessment of study quality (ESM Table) and the funnel plots are shown in the ESM figures. We found no evidence of significant asymmetry or publication bias based upon Begg’s test (for POCD, P = 0.36; for delirium, P = 0.99; for mortality, P = 0.29; for surgical site infection, P = 0.99). Overall, 12 studies were classified at moderate risk and two at high risk of bias.17,39 Given these results, further sensitivity analysis was not performed based upon risk of bias assessment.

Discussion

This meta-analysis assessed the effects of intraoperative cerebral oximetry-guided management on select postoperative outcomes. The results of this study suggest that interventions associated with intraoperative cerebral oximetry monitoring reduce the incidence of POCD resulting in higher MMSE scores at one week compared with a control population. Similarly we found a significant association with shorter ICU LOS in the oximetry-guided intervention group. Nevertheless, the results of our pooled analysis do not suggest a significant difference in hospital LOS or in the incidence of postoperative delirium, transfusion, surgical site infection, or myocardial infarction.

Others have performed meta-analyses to assess the effect of cerebral oximetry monitoring on outcomes after cardiac arrest as well as in extremely low birth weight infants.52,53,54 Cournoyer et al.52 included 20 non-randomized studies in a meta-analysis assessing the effects of cerebral oximetry after cardiac arrest and concluded that higher regional cerebral saturation is associated with improved resuscitation outcomes, especially the return to spontaneous circulation.48 Sorensen et al. concluded that cerebral oximetry monitoring seems important for predicting neurologic complications associated with liver transplantion.54 Although these prior efforts may provide insight into the potential interventions that might stem from the use of cerebral oximetry, it is worth noting that these populations provided only limited information regarding the surrogacy of cerebral oximetry in an operative cohort.55,56

Though not meta-analyses, prior studies designed to illustrate the impact of cerebral oximetry-guided management have qualitatively evaluated the methodology in several populations. At least four systematic reviews have alluded to the potential benefit of cerebral oximetry monitoring in the cardiac surgery population.9,53,57,58 These reviews concluded that despite a limited amount of high-level clinical evidence, the majority of the literature supports the link of cerebral oximetry monitoring to the prevention of POCD. Indeed, Taillefer et al. published a systematic review regarding the use of cerebral oximetry in cardiac surgery, though the authors included only a single RCT,36 and rightfully concluded that this topic had not yet been sufficiently investigated with the rigour necessary to make a more definitive statement regarding the role of cerebral oximetry in adult cardiac surgery.53 It is worth mentioning that the review article of Taillefer et al. was conducted before the publication of the first RCT in cerebral oximetry.40 Our analysis was designed to address this limitation through the inclusion of only RCTs that incorporated interventions guided by the use of cerebral oximetry. Furthermore, we primarily investigated the cognitive impact of these interventions.

There are several possible explanations for the association between cerebral oximetry monitoring during surgery and reduction in POCD. Certainly, it is logical to conclude that the reduction in the incidence of low intraoperative cerebral saturation levels (i.e., indicative of potential cerebral hypoxia) might lead to a subsequent reduction in POCD. This is further supported to be a simple mechanism for benefit by other observational studies. A more nuanced interpretation is that POCD is likely secondary to a relative decrease in effective cerebral perfusion. This may be the downstream result of inadequate arterial blood pressure, cerebral autoregulation impairment, or other unidentified hemodynamic indices.59 While our analysis is unable to specifically evaluate each of these players, cerebral oximetry may represent a useful final common pathway for interpretation of a low perfusion state. Therefore, interventions designed to address one (or all) of these potential variables may provide benefit in reducing the incidence of POCD.

Although our analysis supports the benefit of cerebral oximetry-guided management on the incidence of POCD and a shorter length of ICU stay, it does not show similar impact among a number of other secondary clinical outcomes. Observational studies have previously shown an association between low cerebral oxygen saturation and postoperative delirium.3,60 Others have shown that the severity or duration of postoperative delirium may ultimately be related to subsequent POCD.61 While our analysis did not confirm these results, it is quite possible our pooled analysis included too few patients to adequately assess for postoperative delirium. Furthermore, it is equally plausible that patients may develop POCD without showing signs of early postoperative delirium.

There are several potential implications of these study results. First, they suggest that intraoperative management guided by cerebral oximetry may have applications in the postoperative period. Interventions designed to maintain baseline cerebral perfusion and/or oxygen saturation may prevent POCD and even reduce the length of ICU stay. Second, the interventions described among the included studies are relatively simple and largely include modifications to ventilation strategies, supplementation of additional oxygen, or application of vasopressor support. These do not represent particularly invasive strategies, and therefore it would not be difficult or particularly controversial to begin to develop goal-directed cerebral perfusion protocols based upon the interventions associated with these included trials.

Several important limitations are associated with our meta-analysis. First, the results of our primary analysis were associated with a significant degree of heterogeneity probably due to the different types of cerebral oximetry-based interventions as well as variations in the definition of cerebral oxygen desaturation, the different combinations of cognitive tests that were used to define POCD among the studies, varying surgical case mixes, and other potential differences in individual study-specific patient populations. A number of strategies were utilized to attempt to determine the cause of this level of heterogeneity, including the use of a random effects model, assessment for publication bias, employment of subgroup analysis, and risk of bias assessment. Second, the relatively short time frame that POCD was assessed (i.e., one week postoperatively) could limit the clinical significance of our findings; however it is important to note that only two trials assessed this outcome at three months and both showed significant reduction of POCD in the intervention group.44,48 Another limitation of this meta-analysis is the small sample size of the included RCTs. This highlights the need for further large randomized trials designed to investigate similar intervention strategies surrounding the use of intraoperative cerebral oximetry. Although our analysis failed to show a significant association between the use of cerebral oximetry and other secondary outcomes, this may in part be a function of either a low overall incidence of complications or a lack of adequate patient numbers to detect meaningful differences. After searching the databases and international registries of RCTs, we found two completed but not published RCTs (NCT02155868, ISRCTN23557269) and an ongoing RCT (NCT01707446). Similar future initiatives are likely to add substantial clarity to a rapidly evolving field.

In conclusion, intraoperative cerebral oximetry-guided management is associated with significant reduction in the incidence of POCD. Providers may consider the application of cerebral oximetry to inform specific interventions geared towards minimizing cerebral desaturation and hypoperfusion. Although further large high-quality trials are necessary to elucidate which interventions are most effective and how they directly impact cognitive dysfunction, our findings suggest that simple intraoperative maneuvers based upon cerebral oximetry may provide clear benefit.