An International, Multicenter, Observational Study of Cerebral Oxygenation during Infant and Neonatal Anesthesia

NEUROLOGIC injury during pediatric anesthesia and surgery has always been a significant concern, especially during cardiovascular and neonatal surgery.^1,2  During the past few years, there has been concern for potential neurologic injury during anesthesia in infants without congenital heart disease related to potential neurotoxicity of anesthetic drugs³  and cerebral hypoxia–ischemia related to hypotension and hypoxia during surgery.⁴  Determining the incidence of low cerebral saturation during anesthesia in infants and neonates, as well as associated physiologic factors, such as hypotension and hypoxemia, could improve anesthetic safety, because these mechanisms may be preventable causes of neurologic injury.

In pediatric anesthesia, current standard monitoring includes electrocardiogram to monitor heart rate (HR) and rhythm, pulse oximetry saturation (Spo₂), arterial pressure, respiratory rate, and end-tidal carbon dioxide (ETCO₂). Real-time measurement of cerebral tissue hemoglobin oxygenation using near infrared spectroscopy (NIRS) is widely used in cardiac anesthesia and neonatal and pediatric intensive care units but is infrequently deployed outside of these areas. NIRS noninvasively measures cerebral oxygen saturation (rSco₂) in a tissue volume approximately 1 to 2 cm below the sensor reflecting a weighted average saturation in gas-exchanging vessels (arterioles, venules, and capillaries).⁵  Low regional cerebral rSco₂ (less than 50% for greater than 7 h in neonatal intensive care patients⁶  and less than 45% for greater than 3 h in pediatric cardiac intensive care)⁷  has been linked to adverse neurodevelopmental outcomes and to cerebral ischemic lesions on magnetic resonance imaging in neonatal and pediatric cardiac intensive care. In piglet models of hypoxia–ischemia, low regional cerebral rSco₂ (less than 50%) results in decreased brain tissue energetics, electroencephalogram slowing, brain ischemic lesions, and neurobehavioral impairment.⁸  In pediatric cardiac surgery, rSco₂ monitoring is regarded as the standard of care by many institutions,⁹  because perioperative rSco₂ has been associated with neurologic lesions and neurodevelopmental outcomes.^7,10,11 

Several studies have recently examined regional cerebral oxygenation in infants during pediatric, noncardiac, surgical procedures.^12–15  In children under age 2 yr, rSco₂ usually increased with sevoflurane induction of anesthesia, although decreased rSco₂ during induction was associated with very low mean arterial pressure and younger age.¹²  These studies suggest that unrecognized cerebral desaturation during anesthesia of infants occurs not infrequently and is often associated with hypotension. However, these studies were conducted at a single center in low numbers of patients, reflecting local anesthesia practice for the definition and treatment of arterial pressure, as well as ventilation and arterial oxygenation. Thus, these observations may not be generalizable to infant anesthesia practices across the world.

In the present prospective, multicenter, observational study, we sought to determine the incidence of low regional cerebral oxygenation during anesthesia in a large cohort of infants receiving general anesthesia for noncardiac surgery in centers located in Australia, the United States, China, and Italy. Our secondary aims were to describe regional cerebral oxygenation during surgery and to identify factors associated with cerebral desaturation.

This prospective, observational study involved eight pediatric hospitals: Cincinnati Children’s Hospital Medical Center (Cincinnati, Ohio), the Children’s Hospital of Philadelphia (Philadelphia, Pennsylvania), the Children’s Hospital at Westmead (Sydney, New South Wales, Australia), the Royal Children’s Hospital (Melbourne, Victoria, Australia), Princess Margaret Hospital for Children (Perth, Western Australia, Australia), Guangzhou Children’s Hospital (Guangzhou, China), Yuying Children’s Hospital of Wenzhou (Wenzhou, China), and the Giannina Gaslini Children’s Hospital (Genoa, Italy). Each site obtained institutional review board approval for the study. All of the centers agreed to the plan for data collection and statistical analysis a priori. The primary aim of this study was to determine the incidence of low regional cerebral oxygenation using NIRS in infants during anesthesia for noncardiac surgery; the secondary aims were to characterize changes in cerebral oxygenation during surgery and to identify factors associated with low cerebral oxygenation.

Patients were eligible for enrollment if they were less than 6 months of age (corrected to term birth) and scheduled to undergo a general anesthetic for noncardiac surgery expected to last greater than 30 min. Postmenstrual age (PMA) and gestational age were recorded: PMA at the date of surgery in weeks was equal to (date of surgery – estimated date of delivery + 280 days)/7, and gestational age at birth was equal to 40 weeks – (expected date of delivery – date of birth). Patient recruitment began on December 11, 2014, and ended May 31, 2016. Patients were excluded if application of the NIRS monitor was impractical (i.e., surgery on the head or neck of the child), if they were diagnosed with congenital heart disease involving major cardiovascular shunting, if they were diagnosed with a neurologic disease of the brain, or if they had structural malformations of the frontal brain, scalp, or skull by history or physical examination. Eligible children were identified at each of the individual sites from the operating room schedule, or, for emergency cases, the coordinating anesthesiologist contacted the study team. A member of the local study team approached families preoperatively to explain the project and to obtain informed written consent.

Cerebral oxygenation data for each institution were obtained using the SenSmart X-100 Regional Oximetry System (Nonin, USA). This device measures regional tissue hemoglobin oxygen saturation using spatially resolved spectroscopy principles and the modified Beer-Lambert law; absolute rSco₂ as measured by this device has been validated in neonates and infants over the range of rSco₂ 30 to 90% (average root mean square error = 4%).¹⁶  Our study used the pediatric sensor containing light-emitting diodes at four wavelengths with photodiode detectors spaced 2 and 3 cm from the light source. The device captures rSco₂ every 3 s.

Preoperatively, medical history was obtained along with baseline physiologic data. The NIRS sensor was applied to the patient’s forehead 2 cm above the supraorbital ridge lateral to midline before the induction of anesthesia. Apart from placing the NIRS sensor on the patient’s forehead, the infants received standard care per the discretion of the anesthesia team. Baseline NIRS data were obtained before induction on room air. The sensor remained on the forehead throughout the anesthetic period and was removed after emergence from anesthesia in the operating room before transfer to the postoperative anesthesia care unit or the intensive care unit. Source data of the study were collected from the NIRS device, as well as from the perioperative anesthesia records. In the operating room, physiologic data were captured every 3 to 5 min in accordance with site standard of care via electronic medical chart or using the ICM+ physiologic data acquisition system (Cambridge University, United Kingdom). Timestamps were used to ensure data synchronization between NIRS and other physiologic data. At four sites, the anesthesia team was blinded to the NIRS values (Cincinnati Children’s Hospital Medical Center, Children’s Hospital of Philadelphia, Guangzhou Children’s Hospital, and Yuying Children’s Hospital of Wenzhou), whereas at the other four sites (Children’s Hospital at Westmead, Royal Children’s Hospital, Princess Margaret Hospital for Children, and Giannina Gaslini Children’s Hospital) the team was not blinded; however, the unblinded group did not use the rSco₂ values to manage the anesthetic. Unblinded NIRS data were a requirement of the institutional review board at the unblinded institutions. Postoperatively, parents were contacted 7 days after the operation and asked whether their child had any seizures or other significant medical problems related to surgery.

Based on preliminary data of 50 patients from Cincinnati Children’s Hospital Medical Center, the incidence of severe low cerebral oxygenation (rScO₂ less than 50%) was less than 2%. Given this incidence, which is a sample proportion of 0.010, a sample size of 450 patients would produce a two-sided 95% CI with a width equal to 0.020.

The following time points and definitions were defined a priori and used in data collection and analysis. Baseline included the time between placement of the NIRS probe on the patient’s forehead before induction of anesthesia and administration of anesthetic drugs or supplemental oxygen to induce anesthesia. The average value during this time was used to define baseline rSco₂.

Induction included the time from administration of anesthetic induction drugs to the start of surgery (e.g., incision). The average value during this time period was used to define induction rSco₂.

Surgery included the time from the start of surgery (e.g., incision) until the last stitch or dressing was placed. The average value during this time period was used to define surgery rSco₂.

Emergence included the time from 3 to 5 min after the last stitch or dressing was placed and/or until the patient was extubated. The average value during this time period was used to define emergence rSco₂.

Prematurity was defined as infants born before 37 weeks’ gestation. Term was defined as infants born at 37 weeks’ gestation or later. Expected date of delivery was defined as the date mother expected the child to be born, calculated by 40 weeks after the first day of the last menstrual period, from an early ultrasound scan, or from date of conception plus 2 weeks.

Mild hypotension involved an arterial pressure event lasting greater than 3 min (mean arterial pressure of 36 to 45 mmHg or systolic blood pressure of 51 to 60 mmHg). Moderate hypotension included an arterial pressure event lasting greater than 3 min (mean arterial pressure of 26 to 35 mmHg or systolic blood pressure of 41 to 50 mmHg).¹² Severe hypotension involved an arterial pressure event lasting greater than 3 min (mean arterial pressure of 25 mmHg or less or systolic blood pressure of less than 40 mmHg).

Mild low arterial saturation included an arterial saturation event lasting greater than 3 min (Spo₂ 80 to 89%). Moderate low arterial saturation included an arterial saturation event lasting greater than 3 min (Spo₂ 70 to 79%). Severe low arterial saturation involved an arterial saturation event lasting greater than 3 min (Spo₂ less than 70%).

Mild low cerebral saturation involved an rSco₂ event lasting greater than 3 min (relative decrease in rSco₂ of 11 to 20% below baseline value or absolute rSco₂ 60 to 69%). Moderate low cerebral saturation included an rSco₂ event lasting greater than 3 min (relative decrease in rSco₂ of 21 to 30% below baseline value or absolute rSco₂ 50 to 59%).¹⁷ Severe low cerebral saturation involved an rSco₂ event lasting greater than 3 min (relative decrease in rSco₂ greater than 30% below baseline value or absolute rSco₂ less than 50%).

Demographic and other data recorded on the paper Case Report Form were collated in an electronic REDCap database (hosted by the University of Sydney, Sydney, New South Wales, Australia) and then analyzed with R 2015 (R Core Team, Austria) and SAS version 9.3 (SAS Institute Inc., USA). Descriptive summary statistics of baseline and intraoperative data for the patient population, as well as stratified by site, were presented as frequencies and percentages for categorical variables and mean and SD for continuous variables. Repeated measures of cerebral oxygenation and intraoperative data (i.e., rSco₂, Spo₂, systolic blood pressure, mean arterial pressure, HR, and ETCO₂) were analyzed using mixed-effects models with surgery period as the fixed effect and subject nested within the study site as the random effect. Least square means and 95% CIs were estimated using the restricted maximum likelihood method. For the defined variables (low arterial pressure, low cerebral saturation, and low arterial saturation as defined above), incidence rate, frequency of events per patient, percentage of anesthesia time of each event, and percentage of total anesthesia time of all events were reported as median (range) or 95% CI.

For the primary outcomes, univariate regression models were run when incidence was not rare (greater than 5%). Predictors included patient demographics, intraoperative variables, and incidence of low arterial pressure and low arterial saturation based on variables identified a priori in our statistical analysis plan. ETCO₂ values were not included in regression analysis due to the unreliability of this measure in infants. Specifically, American Society of Anesthesiologists (ASA) physical status, history of prematurity, weight at the time of surgery, PMA at the time of surgery, and sex were demographic variables included in analysis; mild, moderate, and severe low arterial pressure and low arterial saturation were also included.

A post hoc area under the curve (AUC) analysis was conducted for time spent below absolute and relative regional cerebral saturation thresholds of any duration, measured in percentage of minutes of the anesthetic (100 times minutes below threshold/minutes of total anesthetic). This analysis includes all time below threshold (not just restricted to events lasting greater than 3 min in duration, as used in the remaining analyses). In addition, a post hoc description of the seven patients with severe cerebral desaturation was added per reviewer recommendation.

We compared patients from blinded and unblinded sites on demographics, intraoperative variables, and outcomes. A two-sample t test or Wilcoxon rank-sum test, as appropriate, was used for continuous variables, and chi-square test or Fisher exact test, as appropriate, was used for categorical variables. A P value of less than 0.05 was considered significant for all of the results.

A total of 453 patients were included in the final data analysis (fig. 1). Patients averaged a PMA of 50 weeks and weight of 5 kg at the time of surgery (table 1). The majority were ASA physical status I or II, term birth, and male sex. Most patients underwent general surgical and urologic procedures not using laparoscopic techniques (table 2). Nearly 85% of patients received inhalational induction with sevoflurane, and 40% also received propofol and opioids at induction; 60% received neuromuscular blocking agents. Nearly 96% were maintained with inhaled anesthesia, and approximately 50% received a local anesthetic for regional analgesia (table 2).

Missing data were present in 6.5% of the 59,466 physiologic data points (2.7% in rSco₂ data) captured every 3 to 5 min from before the induction of anesthesia to after emergence from anesthesia in the operating room. Table 3 displays the incidence of low cerebral oxygenation, arterial pressure, and arterial saturation events greater than 3 min during the study. During the anesthetic time, approximately 27% (n = 106) of patients experienced events of greater than 10% decline in rSco2 from baseline, with more than 40% (n = 196) having an absolute rSco₂ less than 70% lasting greater than 3 min during the anesthetic, representing approximately 4 and 10% of the total anesthetic time, respectively (table 3). Moderate low rSco₂ was less common, with 6% (n = 24) having a greater than 20% decline of rSco₂ from baseline and 11% (n = 51) having an absolute rSco₂ less than 60%, representing 0.8% and 1.0% of the total anesthetic time, respectively; severe low rSco₂ was rare, with only 1.0% (n = 4) experiencing a greater than 30% decline of rSco₂ from baseline and 1.6% (n = 7) having an absolute value less than 50%, representing 0.1% of the total anesthetic time in both cases. Nearly two thirds of patients had mild hypotension, more than one third moderate hypotension, and more than 12% (n = 57) severe hypotension. The average number of mild and moderate hypotension events per patient were 2.5 and 1.0, respectively; 1 in every 5 patients experienced severe hypotension. In addition, these mild, moderate, and severe hypotensive events lasted for approximately 20, 8, and 2% of the total duration of the anesthetic, respectively. Decreased arterial oxygen saturation was far less common than hypotension: only 15, 4, and 2% (n = 66, 19, and 7, respectively) of patients experienced mild, moderate, and severe events of decreased Spo₂. In contrast to hypotensive events, the low Spo₂ events were brief, the duration of which was less than 1% of total anesthesia time. The majority of hypotension occurred during induction and surgery, whereas a minority occurred during emergence (table 4). Similarly, the majority of arterial and cerebral desaturation events occurred during induction and surgery (table 4).

There were significant differences in rSco₂ during the different periods of the study: the rSco₂ increased during induction and remained increased during surgery and emergence (fig. 2). There were no significant changes from baseline in Spo₂ during induction, surgery, or emergence. Statistical differences also occurred in HR between baseline and these time periods. There were statistically significant decreases in arterial pressure during induction, surgery, and emergence (fig. 2). Baseline data were not included for ETCO₂ due to inaccuracy of measurement in the patient population.

Table 5 reports the AUC analysis for regional cerebral saturation. Across the whole study population, the median percentage of minutes below the various thresholds of total anesthetic minutes is minimal. A total of four patients experienced relative declines in rSco₂ of greater than 30% from baseline, for a median (95% CI) of 33% of minutes (1 to 100% of minutes). Eleven patients experienced absolute rSco₂ of less than 50% of any duration, for a median (95% CI) of 6% of minutes (22 to 76% of minutes).

Because severe cerebral desaturation events were uncommon (0.95% for greater than 30% decline of rSco₂ from baseline and 1.55% for absolute value less than 50%), the number of events was insufficient to conduct regression analysis. It was, however, possible to assess the factors associated with mild and moderate cerebral desaturation in univariate analysis (table 6). The factors with odds ratios greater than 3 for mild and moderate cerebral desaturation were severe low Spo₂ and ASA III or IV versus I. All of the other statistically significant factors had odds ratios less than 2 for mild or moderate cerebral desaturation.

A total of nine patients experienced severe cerebral desaturation: four patients had a greater than 30% decline from baseline, and seven patients had an absolute rSco₂ less than 50%; two patients met criteria that fit both definitions. See Supplemental Digital Content 1 (https://links.lww.com/ALN/B547), which is a table listing the nine cases and includes demographic information, operative details, degree of cerebral desaturation, duration of the cerebral desaturation, and details of the events.

A sensitivity analysis was conducted between blinded (USA and China) and unblinded (Australia and Europe) centers to assess the influence of blinding confounder. There were no differences between the centers for sex or incidence of prematurity. There was a disproportion of ASA I and II patients in blinded centers (ASA II = 50%; ASA I = 26%) versus unblinded centers (ASA II = 33%; ASA I = 41%; P = 0.0024). Weight and PMA at surgery were higher in blinded versus unblinded centers (median [interquartile range] = 6 [4 to 7] vs. median [interquartile range] = 5 [4 to 6], P < 0.0001 and 53 [45 to 59] vs. 48 [43 to 56], P = 0.0002, respectively). There were no differences between the centers for mild, moderate, or severe hypotension or incidence of greater than 20% decline of rSco₂ from baseline or absolute rSco₂ less than 60% or less than 50%. There were higher rates of mild, moderate, and severe low arterial saturation (P < 0.0001, P < 0.0001, and P < 0.0201, respectively) in blinded centers versus unblinded centers. There was also a higher rate of greater than 10% decline of rSco₂ from baseline and absolute rSco₂ less than 70% in blinded centers (P = 0.0150 and P = 0.0014, respectively). There were more infants having greater than 30% decline in rSco₂ from baseline in unblinded versus blinded centers (n = 4 vs. n = 0; P = 0.0240).

No seizures were reported in any of the patients in the study population, nor did any experience other cerebral complications at the 7-day follow-up point. One patient died secondary to complications of intestinal obstruction, one patient was still on extracorporeal membrane oxygenation after inadequate oxygenation secondary to omphalocele closure, and two patients were still ventilated after their operations due to septic complications.

In this international, multicenter study of infants during general anesthesia, we found that mild and moderate cerebral and arterial desaturation and mild, moderate, and severe hypotension occurred frequently, whereas severe cerebral desaturation was rare. An AUC analysis was conducted post hoc to quantify the exposure to cerebral desaturation as a percentage of the anesthetic, which confirmed the cerebral desaturation findings. The incidence and exposure to these events were similar among the eight study institutions. Based on sensitivity analysis, whereas mild cerebral desaturation events were more common in blinded centers, there were no differences between the centers for moderate and severe cerebral desaturation events. Thus, whereas there were differences between the centers, there was no clear directional influence of blinding on incidence or exposure to cerebral desaturation, thereby rendering the effect of blinding as incidental. Although our univariate regression indicated that mild and moderate low rSco₂ were associated with many variables, the correlations were weak overall with the exception of severe low Spo₂. Thus, arterial hypotension and mild-to-moderate arterial desaturation were not major contributors to mild or moderate cerebral desaturation or clinically useful to predict low rSco₂.

Our definition of an event includes both duration and magnitude of decline of rSco₂, Spo₂, and arterial pressure. These event definitions were based on clinical practice, published consensus at the time of study design,¹⁸  and previous definitions in the literature.^8,19–21  The 3-min time window was chosen as the shortest period of time over which these events may be clinically relevant and detectable during a standard anesthetic by using a noninvasive blood pressure cuff. In addition, the cerebral oxygenation events were a combination of both relative (percentage below baseline) and absolute to draw comparisons with previous studies that had only reported relative desaturation and newer studies that have been able to report absolute values.¹⁶ 

Michelet et al.¹³  investigated the relationship between relative rSco₂ and arterial pressure in term anesthetized infants and found an association between decreased arterial pressure and relative cerebral desaturation: a decrease of systolic blood pressure and mean arterial pressure of greater than 37.5% and 44.5%, respectively, had the strongest predictive value of cerebral desaturation greater than 20% below baseline. Consistent with our findings, decreases in arterial pressure were poorly correlated with decreases in relative rSco₂. In a study of cerebral and renal NIRS in neonates, Koch and Hansen¹⁴  demonstrated a 2.8% intraoperative incidence of relative cerebral desaturation greater than 20% from baseline that was moderately correlated with decreased Spo₂ (Φ = 0.371) and weakly correlated with decreased arterial pressure (Φ = 0.231). Although similar to our findings, the interpretation and generalizability of the decreased rSco₂ remained in question.

Similar to our observations, several studies have reported a high incidence of low arterial pressure in infants during general anesthesia.^22–24  Because the autoregulatory thresholds for cerebral blood flow in infants are unknown,²⁵  there has been concern that putative learning and behavioral abnormalities in infants after anesthesia and surgery might be a result of cerebral ischemic events and not a direct effect of the anesthetic on the brain.⁴  Our results show that cerebral oxygenation remains unchanged or is mildly decreased during these low arterial pressure events, suggesting that these arterial pressures are within cerebral autoregulation or that the tissue oxygen supply demand relationship is preserved, a result of a reduction in cerebral metabolic rate under general anesthesia resulting in less oxygen requirements in the brain than the awake state.

The last 15 yr of anesthesia research has seen a substantial body of evidence from infant animal studies pointing toward drug neurotoxicity and abnormal learning and behavior later in life after anesthesia, but human evidence of neurodevelopmental compromise remains equivocal.³  This research, and the discussions on its relevance in clinical practice, have refocused the specialty on anesthetic practice²⁶  and clarification of safe physiologic limits relating to blood pressure and oxygenation. Case reports and clinical experience suggest that substantial and prolonged hypotension and/or arterial desaturation may be associated with neurologic compromise and death.²⁷  What is less certain is the severity duration of this relationship with neurologic injury and the cerebral buffer to provide neuroprotection by anesthesia. Reference ranges for arterial blood pressure in children during general anesthesia that preserve blood flow and oxygen delivery to the organs do not currently exist. Although de Graaff et al.²⁸  have reported arterial pressure values in a large population of infants during anesthesia, the corresponding oxygenation of the organs is unknown.

The relationship of decreased rSco₂ to neurologic injury remains uncertain in infants. In neonatal piglets, neurophysiologic function does not become impaired until rSco₂ is less than 50%, and neurologic injury does not develop until rSco₂ is less than 50% for greater than 2 h, followed by a linear increase in the incidence of neurologic injury of approximately 15% per hour thereafter.⁸  In a study of premature infants, Verhagen et al.²⁹  showed an association between both high and low cerebral rSco₂ (less than 50%) in the first 2 weeks of life and developmental outcome as measured by the Bayley Scales of Infant and Toddler Development at 2 to 3 yr of age. This finding adds to that of a similar study by Alderliesten et al.,⁶  where neurodevelopmental outcome was compared with cerebral rSco₂ and blood pressure in a preterm neonatal group. They found an association between low rSco₂ and lower neurodevelopmental scores but no association between hypotension and neurodevelopment. This may indicate the separate use of rSco₂ monitoring in addition to standard parameters of arterial pressure, pulse oximetry, and arterial blood gasses.

In pediatric cardiac anesthesia, several studies have demonstrated links between cerebral rSco₂ and cerebral ischemia lesions on magnetic resonance imaging⁷  and short-^30,31  and long-term^10,30,32  neurodevelopmental outcomes. Dent et al.⁷  found an association between cerebral ischemic lesions and rSco₂ less than 50% for greater than 3 h. In addition, postoperative lower cerebral tissue oxygenation index has been noted to be associated with longer durations of stay in the intensive care unit, longer durations of intubation, and a higher probability of death.³⁰  However, a potential confounder is that cerebral rSco₂ may be affected by cardiovascular functional status and act as a marker for it rather than for cerebral specific injury that causes neurologic damage. The SafeBoosC trial in neonatal intensive care units in Europe has demonstrated that a tailored intervention strategy in NIRS-monitored patients can reduce the exposure to cerebral desaturation³³  in extremely premature infants. Analyses of secondary outcomes are ongoing but do not as yet show any differences between the treatment and control groups.^34,35  No studies have examined the potential relationship between cerebral desaturation and long-term neurodevelopmental outcomes in neonates and infants undergoing general anesthesia for noncardiac surgery.

The incidence of moderate or severe cerebral desaturation noted in this study is lower than that noted in four other recent studies that all used the INVOS 5100 NIRS device (Medtronic, USA). There are differences in precision and bias between NIRS devices that could contribute to the difference in incidences between our study and these studies.^36,37  The fact that the INVOS 5100, the oximeter with the highest variance, also reports a higher rate of desaturation must be kept in mind when interpreting results.

Although our study is the largest to date, it had several limitations. First, the study did not control for anesthesia technique; however, there were no significant differences among techniques across the study institutions. In addition, half of the sites were blinded to the NIRS monitor, whereas half of the sites were unblinded. Although there was potential for bias relating to monitor-driven interventions, there were no significant differences noted in the incidence of moderate or severe low rSco₂ across any of the sites regardless of blinding status. Our event definitions were developed to the best of our ability. The lack of true standardized definitions of these events makes these definitions subjective and open to debate. We also would have hoped to understand the relationship between ETCO₂ and our outcomes. However, given the difficulty in accurate measurements in infants, we did not include this variable in analysis. Lastly, NIRS is limited to measurement of oxygenation beneath the sensor. The possibility exists for low cerebral oxygenation to occur in other areas of the brain that were not being studied or deeper in the brain than the source of the optical signal. However, during general anesthesia, imaging studies show a uniform change in blood flow and metabolism throughout the brain, suggesting that this possibility is unlikely.³⁸  Despite these limitations, this study provides a useful insight into the incidence of low cerebral saturation from major pediatric centers in the world.

In summary, mild cerebral deoxygenation is common during anesthesia of infants, whereas severe cerebral desaturation is uncommon. Low arterial pressure was very common and not well associated with low rSco₂. Although severe cerebral deoxygenation does occur during infant anesthesia, it is both rare and brief, and thus is unlikely to explain the reported learning and behavioral abnormalities associated with general anesthesia and surgery. Only by performing longer-term neurodevelopmental outcome studies would any association with later learning and behavioral difficulties be ascertained.

The authors thank Christopher Brasher, F.A.N.Z.C.A. (Department of Anesthesia and Pain Management, Royal Children’s Hospital, Melbourne, Victoria, Australia); Jack Luxford, B.A. (Sydney Medical School, Sydney, New South Wales, Australia); Michaela Turancova, B.Sc. (University of Sydney, Sydney, New South Wales, Australia); Lara Oversby, R.N. (Department of Anesthesia and Pain Management, Princess Margaret Hospital for Children, Perth, Western Australia, Australia); Thomas Drake-Brockman, B.Phil. (Hons.) (Department of Anesthesia and Pain Management, Princess Margaret Hospital for Children and School of Medicine, University of Western Australia, Perth, Western Australia, Australia); and Lliana Slevin, B.Sc. (Hons.) Pharmacology (Department of Anesthesia and Pain Management, Princess Margaret Hospital for Children and Children’s Lung Group, Telethon Kids Institute, Perth, Western Australia, Australia).

Supported by Australian and New Zealand College of Anaesthetists Project Grant 15/021 (West End, Queensland, Australia) and Society of Pediatric Anaesthetists of New Zealand and Australia Research Grant 2016 (Morrisett, New South Wales, Australia), as well as the Princess Margaret Hospital Foundation (to Dr. von Ungern-Sternberg; Perth, Western Australia, Australia), the Callahan Estate (to Dr. von Ungern-Sternberg; Perth, Western Australia, Australia), and the Stan Perron Fellowship (to Dr. von Ungern-Sternberg; Perth, Western Australia, Australia).

The authors declare no competing interests.