A Multicenter, Randomized, Controlled Phase IIb Trial of Avoidance of Hyperoxemia during Cardiopulmonary Bypass

CARDIAC surgery utilizing cardiopulmonary bypass (CPB) is one of the most common forms of major surgery, with more than 1.25 million patients undergoing this surgery worldwide each year.¹  Despite improvements to both surgical techniques and the equipment used for CPB, mortality and significant morbidity remain high. During CPB, it is common practice to oxygenate the arterial return blood to supranormal levels, in part because monitoring of the blood oxygen levels, and thus the function of the oxygenator, is commonly done intermittently using blood gas sampling. Oxygen is one of the most widely available and prescribed therapeutic drugs in medicine.²  While the risks associated with hypoxemia are well recognized, there is growing concern that hyperoxemia could also induce detrimental systemic effects.^3–5 

Cardiac surgery–associated multiorgan dysfunction (CSA-MOD) is well recognized and includes cardiac surgery–associated acute kidney injury (CSA-AKI), acute hepatic impairment, myocardial damage, and postoperative neurologic deficit.⁶  The etiology of CSA-MOD is complex but is thought to involve numerous injurious pathways including the development of oxidative stress and cellular damage caused by excess reactive oxygen species.⁷  Animal data suggest that these effects may be amplified by exposure to hyperoxemia, which is common during CPB.⁷  Although there is a considerable overlap between them, the principal processes include exposure of circulating blood to nonbiocompatible surfaces and erythrocyte damage and hemolysis caused by mechanical stress from the blood circulating pumps and the blood–gas interfaces in cardiotomy suction. The subsequent activation of inflammatory pathways, ischemia–reperfusion injury, decreased end-organ perfusion, and hemolysis are contributors to CSA-MOD.⁶  These complications result in significant morbidity and mortality, as well as in increased intensive care and hospital length of stay and consequential increases in healthcare costs.⁸ 

Although the deleterious effects of CPB affect all body systems, they are most marked on the kidney, with renal damage occurring in up to 30 to 70% of patients, depending on the definition used.⁹  Oxidative stress, the creation of oxygen free radicals and heme- or iron-containing reactive oxygen species, is recognized as an important factor in the development of CSA-AKI.^10–12  The production of these toxic molecules is in part dependent on arterial oxygen levels. There is also evidence suggesting that high levels of arterial oxygenation may exacerbate ischemic-reperfusion injuries, and this effect is attenuated by avoiding hyperoxemia.^13,14  Hyperoxemia has also been shown to have significant adverse hemodynamic effects after coronary artery bypass surgery.¹⁵ 

Current standard practice is to use supranormal arterial blood oxygen tensions during CPB; however, increasing availability of real-time continuous in-line blood gas analysis now provides the opportunity to safely use lower, more physiologic arterial oxygen tensions.

We hypothesized that the avoidance of perioperative arterial hyperoxia decreased the degree of oxidative stress and therefore the severity of the CSA-MOD after cardiac surgery utilizing CPB.

A prospectively randomized, open-label, parallel-group, phase IIb interventional study was undertaken in two metropolitan hospitals: one in New Zealand and one in Australia. The study was approved by the local ethics committees at each site (12/NTA/15 Northern A Health and Disability Ethics Committee, New Zealand, and Calvary Health Care Adelaide Human Research and Ethics Committee 13-CHREC-F001, Australia).

Patients were approached in hospital before surgery, the study was discussed, and written informed consent to participate was obtained by trained research staff from all study participants before enrollment. The trial was prospectively registered with the Australia and New Zealand Clinical Trials Registry: 12612000756820 (http://anzctr.org.au/).

Participants were eligible for inclusion if they were 16 yr of age or older and undergoing scheduled cardiac surgery utilizing CPB. Exclusion criteria were any of the following: acute renal failure (acute increase in serum creatinine more than 50% from baseline in the 6 weeks before surgery), preoperative end-stage renal disease (serum creatinine more than 3.4 mg/dl) or receiving any form of renal replacement therapy, preoperative hepatic dysfunction (aspartate aminotransferase [AST] more than 2× upper limit of normal), recent (less than 6 weeks) cerebrovascular event (including cerebrovascular accident, transient ischemic attack, or intracerebral bleed), pregnancy, planned hypothermic circulatory arrest, or preoperative intraaortic balloon pump.

After enrollment, patients were randomized 1:1 in blocks of 8 to the intervention or usual care, with the sequence generated by an independent statistician. To facilitate balance in severity between treatment arms, patients were further stratified according to risk of AKI using well-recognized criteria.¹⁶ 

Allocation concealment was maintained until the time of randomization by using opaque, sealed, sequentially numbered envelopes, prepared by a person independent of the study.

The intervention group received protocolized control of arterial oxygenation from induction of anesthesia until the end of surgery to maintain arterial oxygen tensions of 75 to 90 mmHg. It was recommended that hyperoxemia be avoided during anesthetic induction and the prebypass period to target peripheral oxygen saturation (Spo₂%) 92 to 95%. It was acknowledged that not all anesthetists would be comfortable with this approach and might wish to administer 100% Fio₂ (as opposed to the recommended target Spo₂ of 92 to 95%) during part or all of the pre- and post-CPB period. During CPB, the oxygen tension of the arterial return blood flow target was 75 to 90 mmHg using continuous, in-line, real-time blood gas monitoring (Terumo CDI 500, Terumo Corporation and Spectrum M4, Spectrum Medical, UK), providing arterial saturation more than or equal to 97% (if less than 97% then the Pao₂ was to be increased beyond 90 mmHg).

The control group received usual care, including selection of appropriate supplemental oxygen by the treating anesthetist and clinical perfusionist. Both the centers participating in this study do not routinely use a Fio₂ of 1.0 during CPB, rather Fio₂ is titrated to maintain a circuit arterial saturation of more than or equal to 99%.

The conduct of CPB followed normal clinical practice (crystalloid prime, mild hypothermia, −34° to 32°C; target hematocrit, more than 0.23, flow index, 2.0 to 3.0 L min⁻¹ m⁻²; mean arterial pressure, more than 50 mmHg). Alpha-Stat blood gas management was used, and the oxygenator gas Fio₂ was titrated to meet the study protocol. All other anesthetic, CPB, and surgical care were provided at the discretion of the clinician.

We used dedicated research nurses to collect data using paper case report forms. Data included demographic data (i.e., age, ethnicity, body mass index, comorbidities), intraoperative data (i.e., surgery type, CPB, and aortic cross-clamp time), postoperative data (i.e., requirement for renal replacement therapy, length of mechanical ventilation, and length of stay in intensive care unit [ICU] and hospital), and biochemical data (arterial blood gas measurement, serum creatinine, urea, hemoglobin, selenium, C-reactive protein, high-sensitivity troponin T, AST, and amylase were all measured immediately before surgery and at 6 and 24 h after commencement of CPB). Serum creatinine was also recorded, if measured, on postoperative days 1 to 5. Hourly urine output measures were performed up to day 5 if the patient remained catheterized and in the ICU. All patients were contacted 28 and 90 days after randomization to ascertain mortality and requirement for renal replacement therapy.

The primary outcome was the difference in CSA-AKI during hospital admission as determined by the Kidney Disease: Improving Global Outcomes (KDIGO) AKI Work Group guidelines.¹⁷  Secondary outcomes included changes in multiple markers of end-organ damage (troponin, AST, amylase, C-reactive protein) and length of mechanical ventilation, intensive care, and hospital stay. Process of care data was collected to determine the actual difference in arterial oxygen tension between the two groups and amount of inspired oxygen (Fio₂%) used by the anesthetist before commencement of CPB.

As a phase IIb equivalent study of a novel therapeutic intervention, we were seeking to provide proof of concept, evidence of potential efficacy, and sufficient information to enable rigorous design of a definitive phase III study.

Sample size calculations were based on the findings of a previous study of CSA-AKI in which a 20% reduction in the prevalence of AKI was observed.¹⁸  Using our hospital database, the baseline incidence of CSA-AKI using KDIGO criteria was 67%. Sample size calculations were based on the desire to have an 80% power to reduce this incidence from 2 in 3 to 1 in 2 with a two-sided P value of 0.05. This equates to a raw difference of 17% (67 vs. 50%) and a relative risk reduction of 25%. A relative risk reduction of 25% is considered clinically significant and is less than that used in previous studies of CSA-AKI.^19,20  To facilitate this, 286 patients were required. To account for possible loss to follow-up, 298 patients were enrolled. When considering secondary variables, these 298 patients further enabled an 80% power to detect a difference in continuous outcomes equivalent to one third of a standard deviation with a two-sided P value of 0.05. A difference of this magnitude was perceived to be of clinical importance.

All analysis was performed according to the intention-to-treat principle using SAS version 9.4 (SAS Institute Inc., USA). Group comparisons in the primary outcome were made using chi-square tests for equal proportion (or Fisher exact tests where numbers were small), Student’s t tests for normally distributed data, or Wilcoxon rank sum tests otherwise. Results are presented as n (%), mean (SDs), or median (interquartile range). Repeated measures data were analyzed using mixed linear models fitting main effects for treatment, time, and an interaction between treatment and time to determine if the groups behaved differently over time, with results present as least square means (SEs). Risk stratification was not utilized for analysis nor were center effects included in the model. A two-sided P value of 0.05 was considered to be statistically significant.

During the period from December 2012 to May 2014, 305 patients were enrolled, of which 298 patients were randomized at two hospitals (fig. 1).

Seven patients were not randomized before surgery due to changes caused by alterations in the operative schedule. Of those randomized, 148 were allocated to usual care and 150 to intervention. No subjects were lost to follow-up, and all had the primary and secondary outcomes available for analysis. Data were available for analysis of primary outcome for all 298 participants. Baseline characteristics of study participants are described in table 1.

No adverse events or side effects of treatment were recorded in study participants. A total of nine participants (3%) died: eight died postoperatively in the ICU and one died after discharge from hospital but before day 28 postoperative.

Excellent treatment separation was achieved between the two groups during the bypass phase, but Pao₂ values pre- and post-CPB were similar in both groups (fig. 2).

The pre-CPB value was measured immediately before commencement of CPB. The post-CPB value is an average of the values recorded every 10 min from the end of CPB to the completion of surgery. The ICU value is the average of values recorded for 6 h from admission to ICU. Values are presented as means, and error bars show SEM. Both groups had Fio₂ titrated to meet the protocol (intervention arm) or as per usual practice (control arm; fig. 3).

There was no difference between the intervention and control arms in respect to the management of blood flow (2.6 vs. 2.5 l m⁻² min⁻¹: P = 0.06) and oxygen delivery index (320 vs. 324 ml min⁻¹ m⁻², P = 0.53) during CPB. Both arterial (97.2 vs. 99.0%, P < 0.001) and venous (78.6 vs. 81.7%, P < 0.001) saturations of blood in the CPB circuit were lower in the intervention arm.

There was no significant difference in the occurrence of renal dysfunction between groups as measured by KDIGO and based on change in either urine output or creatinine (table 2).

There was no significant difference found between the two groups in the multiple biomarkers measured at baseline, 6 or 24 h after commencement of CPB (table 3). There was also no difference seen in ventilation hours, ICU or hospital length of stay, and requirement for renal replacement therapy or mortality to day 90.

This study assessed the avoidance of hyperoxemia in adult patients undergoing cardiac surgery using CPB. We achieved excellent treatment separation once the participant was on CPB but failed to demonstrate any significant difference in the incidence of AKI, other measures of CSA-MOD, or ICU and hospital length of stay.

To our knowledge, this is the first study to utilize real-time in-line arterial blood gas analysis to avoid perioperative arterial hyperoxemia during CPB in an attempt to decrease oxidative stress and severity of CSA-MOD. Previously published work attempting to establish the most appropriate Fio₂ during CPB, which would result in normoxemia, found that to avoid hyperoxemia, Fio₂ should be kept at 0.35 during CPB.²¹  Currently, there is no published evidence or consensus to inform clinical perfusionists as to where to target oxygen levels or what the ideal Fio₂ or Pao₂ is on bypass,²¹  resulting in wide practice variation and results.

A pragmatic study design was used to ensure generalizability of study results to large numbers of patients presenting for cardiac surgery. This study enrolled a large sample size, there was no loss to follow-up, and the study was undertaken over 17 months. Some delay to recruitment was experienced due to availability during the trial of the continuous in-line blood gas monitoring system, which was not present on all bypass machines. This meant that not all patients available could be recruited due to the inability to provide intervention treatment.

Treatment separation was easily achieved by clinical perfusionists using the continuous in-line real-time blood gas monitoring. The moderate degree of hyperoxemia (mean Pao₂, 178 mmHg) observed during CPB in the control arm reflects the current clinical practice in both our study sites and many other centers in New Zealand and Australia. A mean arterial oxygen saturation in the intervention arm of 97.2% demonstrated that we achieved the maximal possible reduction in hyperoxemia as it was considered clinically unacceptable to conduct CPB with arterial saturations lower than 97%.

Although arterial hyperoxemia has long been used to protect against potential side effects of CPB such as hypoxic injury and gaseous embolization, emerging evidence from diverse populations, such as those postcardiac arrest, suggests that exposure to high levels of oxygen tensions may in fact be harmful.^22,23 

Hyperoxemia has been associated with many side effects such as alterations in capillary blood flow, reduced cardiac index, increased systemic vascular resistance, hemolysis, and formation of micro air emboli.²⁴  We hypothesized that by controlling oxygen levels in a normal range while on bypass, these side effects would be negated in the intervention group. We utilized a wide range of biomarker-based endpoints to detect a treatment effect for organ damage between the groups but failed to show any difference. Although there are some newer biomarkers that may directly measure oxidative stress (e.g., isoprostanes), we chose to measure commonly used and available measures of individual organ damage. We chose to use the incidence of AKI using the KDIGO definition as the primary outcome as it is well recognized that the development of AKI after cardiac surgery is associated with poorer outcomes. KDIGO has superseded earlier classifications of AKI and has been used in a number of clinical studies in this patient population; however, most previously published work has only used the creatinine component of KDIGO rather than both the creatinine and urine output components. We chose to use the full definition in accordance with guidelines, which yielded a much higher incidence of AKI than is found using creatinine criteria alone. However, even if only the creatinine component of the KDIGO definition is used, we found no difference between groups.

The lack of efficacy of the intervention may be explained by an absence of true ischemia-reperfusion injury after routine CPB as compared to that seen for instance in survivors of cardiac arrest whom have a “whole of body ischemia-reperfusion insult” as opposed to a more limited insult focused on the heart in CPB patients. It is plausible that routine hypothermia during CPB diminishes the effects of hyperoxemia, moderating the development of postoperative neurocognitive dysfunction and MOD.²⁵  Furthermore, there is recent evidence suggesting that exceeding a critical threshold of oxygen delivery might limit the incidence of postoperative AKI.²⁶  There was no difference in oxygen delivery index during CPB between the intervention and the control groups, and this was well in excess of the DO₂i threshold identified, possibly negating the impact of hyperoxemia on markers of renal outcome.

We recognize some inherent limitations to our study. First, although attending anesthesiologists were encouraged to target a low normal level of arterial oxygenation pre- and post-CPB, there was only a small difference between groups at these time points. The explanation given by the anesthesiologists was that they were uncomfortable using low Fio₂ during these periods. The intervention, therefore, only occurred during CPB and thus there was significant hyperoxemia both pre- and post-CPB, and it is possible that these times represent “at-risk” periods for hyperoxemia-mediated cellular damage. In particular, it is theoretically plausible that the most critical time for hyperoxemia is when the blood is first exposed to the artificial surfaces of the bypass circuit. In our study, we did not achieve significant separation of oxygen levels immediately before initiating CPB. We also did not insist on the provision of normoxemia to patients in either the prebypass or postbypass period but left that to the discretion of the treating clinicians. Nevertheless, in our sample of cardiac surgery patients, we did not demonstrate any difference in outcomes with avoidance of hyperoxemia during CPB.

Although we successfully achieved normoxemia in the intervention group, the actual difference in Pao₂ between the two groups was only moderate (approximately 80 mmHg), and it is plausible that greater degrees of hyperoxemia are required to cause organ damage. Greater hyperoxemia in the control arm could have been achieved by mandating a Fio₂ of 1.0 in these patients; however, this would have been contrary to usual care at the two participating sites.

This is an unblinded study in a group of all-comers rather than high-risk cardiac surgical patients. Perhaps if we had targeted only high-risk patients, results may have been different. CSA-MOD occurs through multiple injurious pathways, which may have differential effects in different subpopulations of patients. It is plausible that certain preoperative patient characteristics, surgical procedures, or time on CPB may identify a group of patients who are more likely to benefit from the avoidance of hyperoxemia. Although we have provided data on the primary outcome for isolated coronary artery bypass surgery and other patients, the sample size does not allow further meaningful analysis of other subgroups.

We chose to undertake indirect measures of oxidative stress and measured markers of organ dysfunction rather than attempting to quantify the degree of oxidative stress directly.

In summary, the avoidance of modest hyperoxemia (mean, 178 mmHg) during CPB appears safe and achievable but failed to demonstrate any difference in AKI, markers of organ damage, or length of stay. Future clinical studies in this area should consider extending the intervention to the pre- and post-CPB period and should also consider if maximizing the treatment separation by using a Fio₂ of 1.0 is appropriate.

Supported by the Health Research Council of New Zealand, Auckland, New Zealand, and The Green Lane Research & Education Fund, Auckland, New Zealand.

Research in the Cardiothoracic and Vascular Intensive Care Unit, Auckland City Hospital, Auckland, New Zealand (Drs. McGuinness and Parke), is supported in part by an unrestricted grant from Fisher & Paykel Healthcare Ltd., Auckland, New Zealand. The other authors declare no competing interests.

Full protocol available from Dr. McGuinness: shaymc@adhb.govt.nz. Raw data available from Dr. McGuinness: shaymc@adhb.govt.nz.

Members of the SO-COOL investigators are as follows: Auckland City Hospital, Auckland, New Zealand: Shay P. McGuinness, M.B.Ch.B., Rachael L. Parke, Ph.D., Cornelius Kruger, M.B.Ch.B., Tim Willcox, F.A.N.Z.C.P., Megan Baker, B.A., Keri- Anne Cowdrey, B.A., Eileen Gilder, M.A., Lianne McCarthy, B. Sc.; Calvary Wakefield Hospital, Adelaide, South Australia: Kate Drummond, M.B.B.S., Thomas Painter, M.B.Ch.B.