Hyperoxia During Cardiopulmonary Bypass Is Associated With Mortality in Infants Undergoing Cardiac Surgery

Asaad G Beshish 1Ozzie Jahadi 2Ashley Mello 3, et al.

Pediatric Critical Care Med.2021 1;22(5):445-453.

DOI: 10.1097/PCC.0000000000002661

PMID: 33443979


Take Home Points

  • Hyperoxia during CPB is an independent risk factor associated with 4-fold greater odds of 30-day mortality.
  • Infants with hyperoxia are more likely to have acute kidney injury, prolonged post-operative stays, and mortality, but the authors failed to identify an association with development of acute kidney injury or prolonged postoperative length of stay when controlled for covariables.
  • A PaO2 of greater than 313 mmHg on CPB has the highest sensitivity with specificity greater than 50% for association with operative mortality.
  • Hyperoxia is associated with greater odds of mortality in neonatal patients.
  • No generally accepted level for pathological hyperoxia exists.

Commentary by Matthew A. Lilien, MD and Laura Downey, MD, Children’s Healthcare of Atlanta Department of Pediatric Cardiac Anesthesiology.

Currently, there are no clearly defined numerical cutoffs for hypoxia, normoxia, or hyperoxia for pediatric patients undergoing cardiopulmonary bypass. Even adequate oxygenation is poorly defined but usually accepted as the balance between oxygen delivery and consumption.1 During metabolism, oxygen (O2) combines with hydrogen (H) in mitochondria to form water (H2O). However, there can be leakage of the enzyme chain creating highly reactive molecule species referred to as reactive oxygen species (ROS).2  The body has an antioxidant system to balance free oxygen, but if the antioxidant system is overwhelmed by high levels of free oxygen, ROS interact with lipids, DNA, and proteins. These reactions trigger cellular responses ranging from subtle modulations of cell signaling to overwhelming oxidative injury resulting in cell necrosis or apoptosis.4 Premature infants are especially susceptible to ROS-induced damage for two major reasons. First, adequate concentrations of antioxidants may be absent at birth since developmental increases in antioxidant capacity (maternal-fetal placental transfer, endogenous production) occur in the latter part of gestation in preparation for the transition to extrauterine life. Second, the ability to increase the synthesis of antioxidants in response to hyperoxia or other oxidant challenges is deficient.5 Given that other risk factors like age, weight, and STAT score are not easily modifiable, this paper looks to identify hyperoxia as a modifiable practice-based risk factor that may lead to improved outcomes.


This is a single-center retrospective cohort study of 469 infants undergoing cardiac surgery with CPB from January 2015 to December 2018. Patients were excluded if ECMO was initiated in the operating room and for patients with corrected gestational age less than 35 weeks. The primary objective was to determine if hyperoxia is an independent risk factor for all-cause inpatient mortality. Secondary outcomes were to determine if hyperoxia is associated with acute kidney injury (stage II or III) or prolonged post-operative length of stay (PPLOS) defined by > 14 days. To control for bias of early mortality appearing as a shortened PPLOS, two methods were used: 1) all patients with post-operative mortality were counted as PPLOS regardless of when the death occurred; 2) a composite rank-based outcome was created for days alive ICU free. PaO2 on CPB was determined at the discretion of the perfusionist. PaO2 for individual patients were determined as an average of all arterial blood gas samples while on CPB.


The study included 471 infants undergoing cardiac surgery with CPB. Two patients were excluded due to initiation of ECMO in the OR. Patients had a median age of 97 days, median weight of 4.9kg, and a STAT score of 4 or 5 in 48% of patients. Median duration of CPB and cross clamp were 128 minutes and 54 minutes, respectively. PaO2 levels on CPB were averaged for each individual patient with a mean of 314 mmHg (IQR 287-340 mmHg). Of the 469 patients, 25 died from various causes (HF, cardiac arrest, multiple organ failure, ECMO).


In order to determine the average intraoperative PaO2 which was most predictive for the primary outcome (30-day mortality), a receiver operating characteristic (ROC) curve was used. The optimal cutoff value of PaO2 of 313 mmHg was selected due to the best combination of sensitivity (80%) and specificity (greater than 50%) for 30-day operative mortality. The area under the curve for PaO2 to predict mortality was 0.67 (95% CI, 0.56-0.77, p=0.003). Of the patients included in the study 237/469 fell under the author’s definition of hyperoxia while on CPB. The hyperoxia group was younger in age, weighed less, underwent more complex surgical procedures (median STAT 4 vs. 3), and had longer CPB and cross clamp durations. In multivariable regression, hyperoxia during CPB was associated with greater odds to predict mortality (OR=3.9, 95%CI, 1.4-10.5, p=0.008) when controlled for age, weight, and STAT mortality score. Survivors had a lower median PaO2 while on CPB than non-survivors (311 mmHg IQR, 286-339 vs 339 mmHg IQR 315-355, p=.003). A subgroup analysis was performed for neonates that demonstrated patients with hyperoxia were more likely to have operative mortality (19% vs 6%, p=0.04) when controlling for weight and CPB time (OR =3.8, 95% CI, 1.02-14.0 p=0.04). Univariable and multivariable analysis failed to find association between hyperoxia and development of AKI or PPLOS.


The results show that there may be an independent association with hyperoxia and mortality. Other studies have also demonstrated hyperoxia as an independent risk factor. Synycer-Taub et al showed worse 30-day mortality and need for dialysis in ECMO for cardiac arrest patients with hyperoxia defined as >198 mmHg.6 Rabi et al showed that in asphyxiated infants, neonates in the room air resuscitation group had a lower mortality than those that received 100% oxygen during resuscitation at both the first week of life (odds ratio 0.70, 95% CI 0.50, 0.98) and at one month and beyond (odds ratio 0.63, 95% CI 0.42, 0.94).7 The Beshish et al study does not claim hyperoxia on CPB directly leads to death, but “hyperoxia may have led to a postoperative course that was more likely to have comorbidities and complications that would lead to mortality.”


This study has several limitations that make it difficult to make any changes or recommendations based on these results.

  • This is a single institution study with a small number of patients. Evidence from a large multicenter study may help elucidate the risk of hyperoxia on bypass.
  • This study does not include all variables that could contribute to morbidity, and hyperoxia on CPB may have been a surrogate for disease severity.
  • Hyperoxia has not been established across patient populations or disease states. These authors used a cut point analysis to define hyperoxia. However, a PaO2 of 313 mmHg is generally much higher than intensivists and anesthesiologists target for infants and neonates perioperatively.
  • Oxygen levels are continuously changing based on patient tissue needs with delivery and consumption while on bypass. These samples are arterial gases. The authors did not measure venous oxygen levels to determine tissue oxygen consumption and need for hyperoxia.
  • There is no data that antioxidant levels and the ability to clear free radicals is consistent between individuals. A specific concern for this patient population is whether patients with cyanotic heart disease have the same ability to cope with free radicals as non-cyanotic patients.
  • There was no clearly defined protocol for perfusionists that dictated parameters for patients to receive higher levels of delivered oxygen, such as decreased Near Infrared Spectroscopy (NIRS), decreased mixed venous oxygen saturation, or increasing lactate. Perfusionists could take samples at set times and periods of CPB cases. It is not clear from this study if sicker patients required higher levels of oxygen.
  • Of the subgroup of neonates, the survivors have a wide distribution of PaO2. However, the distribution of PaO2 in the neonatal deaths had overall higher PaO2 levels. It is not clear from this study that hyperoxia is a risk factor for mortality, but instead hyperoxia may be a surrogate for severity of illness. The authors considered STAT scoring as a variable but that doesn’t account for patient’s overall disease state.

Based on this paper and others that are referenced throughout this study, it is suggested that delivering higher levels of oxygen may have a detrimental effect on patient outcomes. This is particularly seen when ischemic tissue, as in post cardiac arrest and brain ischemia, is resuscitated with hyperoxia causing reperfusion injury. Similar situations can be extrapolated to CPB and patients with cyanotic heart disease. Congenital heart patients that are chronically hypoxic will partially switch to anaerobic metabolism, resulting in acidosis, depending on the level of hypoxia. Whether this hypoxia is acute or chronic, there is a lower antioxidant reserve due to increased ROS production.8 The result of CPB with hyperoxia will be increased ROS with decreased capacity of the antioxidant system. Thus, it is a logical jump to think hyperoxia on CPB may lead to reperfusion injury and cell death. More research is required on the subject, and current studies are limited to suggest a change in practice. However, the consequences of hyperoxia are not clearly balanced by the benefits of higher PaO2 levels with the potential for cell injury. There is research that suggests normoxia during CPB procedures in adults results in reduced myocardial oxidative stress compared to hyperoxia9. However, the concept of controlled reoxygenation and the practice of aiming for lower levels of oxygen on CPB are not common in the pediatric population. Nevertheless, both are worth investigating given the existing evidence of reperfusion injury.



  1. Pediatric Oxygen Therapy: A Review and Update. Brian K Walsh and Craig D Smallwood
    Respiratory Care June 2017, 62 (6) 645-661; DOI: https://doi.org/10.4187/respcare.05245
  2. Hyperoxia in anaesthesia and intensive care E. Horncastle1 and A.B. Lumb1,2, * 1 St James’s University Hospital, Leeds, UK and 2 University of Leeds, Leeds, UK
  3. Oxygen Toxicity Louise Thomson MBChB, MRCPH, James Paton MD* School of Medicine, University of Glasgow
  4. Pacher P., Beckman J.S., Liaudet L.: Nitric oxide and peroxynitrite in health and disease. Physiological reviews 2007; 87: pp. 315-424.
  5. Maturation of the antioxidant system and the effects on preterm birth Jonathan M. Davis a,*, Richard L. Auten b
  6. Sznycer-Taub NR, Lowery R, Yu S, et al: Hyperoxia is associated with poor outcomes in pediatric cardiac patients supported on venoarterial extracorporeal membrane oxygentation. Pediatr Crit Care Med 2016; 17:350-358
  7. Rabi Y, Rabi D, Yee W: Room air resuscitation of the depressed newborn: A systematic review and meta-analysis. Resuscitation 2007; 72:353-363
  8. Mokhtari, Amir, and Martin Lewis. “Normoxic and Hyperoxic Cardiopulmonary Bypass in Congenital Heart Disease.” BioMed Research International, Hindawi Limited, 2014, pp. 1-11. Crossfer, doi: 10.1155/2014/678268.
  9. Topcu AC, Bolukcu A, Ozeren K, Kavasoglu T, Kayacioglu I. Normoxic management of cardiopulmonary bypass reduces myocardial oxidative stress in adult patients undergoing coronary artery bypass graft surgery. Perfusion. 2021;36(3):261-268. doi:10.1177/0267659120946733