Intra-operative high inspired oxygen fraction does not increase the risk of postoperative respiratory complications

Alternating intervention clinical trial

Abstract

BACKGROUND 

The WHO recommends routine intra-operative and early postoperative use of high inspired oxygen concentrations (hyperoxia). However, a high intra-operative inspired oxygen fraction (FiO₂) might result in an increased risk of postoperative respiratory complications.

AIM 

To test the hypothesis that intra-operative FiO₂ of 80% compared with 30% inspired oxygen decreases the postoperative ratio of arterial saturation to fraction of inspired oxygen (SpO₂/FiO₂). Secondarily, to evaluate whether an intra-operative inspired FiO₂ of 80% increases the incidence of pulmonary complications.

DESIGN 

Posthoc subanalysis of a large alternating cohort trial.

SETTING 

Cleveland Clinic, Cleveland, United States, from 2013 to 2016.

PATIENTS 

Adults having colorectal surgery. Cases lasting less than 2 h, re-operations on the same hospitalisation, and cases with missing intra-operative or postoperative data were excluded.

INTERVENTION 

Maintaining intra-operative FiO₂ at 30 or 80% and alternating this management every 2 weeks for a study period of 39 months.

MAIN OUTCOME 

Minimal SpO₂/FiO₂ ratio value in the postanaesthesia care unit. Secondary outcome was a composite of postoperative pulmonary complications throughout hospitalisation.

RESULTS 

A total of 5056 patients were included. Groups were well balanced on all demographic, baseline and procedural variables. Median time-weighted averages of intra-operative FiO₂ in the 30 and 80% groups were 43% (IQR 38 to 54%, N=2486) and 81% (IQR 78 to 82%, N=2570), respectively. No difference was found in the lowest SpO₂/FiO₂ ratio (estimated median difference 0 [95% confidence interval: 0, 0], P = 0.91). The incidence of postoperative pulmonary complications was 16.3 and 17.6% in the 30 and 80% FiO₂ groups, respectively (relative risk 1.07 [95% confidence interval: 0.95, 1.21], P = 0.25).

CONCLUSION 

Intra-operative hyperoxia did not change the postoperative SpO₂/FiO₂ ratio or the risk for pulmonary complications. Clinicians should not refrain from using hyperoxia for fear of provoking respiratory complications.

TRIAL REGISTRATION 

ClinicalTrials.gov identifier: NCT01777568.

This article is accompanied by the following Invited Commentary:

Aveline C. Intra-operative high inspired oxygen during open abdominal surgery and postoperative pulmonary complications: From physiology to individualised strategies. Eur J Anaesthesiol 2019; 36:317–319.

Introduction

The WHO and the United States Centers for Disease Control and Prevention (CDC) recently recommended administration of 80% inspired oxygen fraction (FiO₂) during and immediately after surgeries performed with general anaesthesia and endotracheal intubation.^1,2 The recommendation was based on some data suggesting that intra-operative high FiO₂ reduces the incidence of surgical site infections. Many anaesthesia clinicians are reluctant to provide higher concentrations of oxygen because of potential adverse effects on the lungs. Most randomised trials utilising intra-operative high FiO₂ evaluated its effect on various wound-healing related complications, and many do not even report pulmonary outcomes.^3–7 Neither the WHO nor the CDC recommendations fully address these potentially harmful effects. Furthermore, a recent large registry analysis of adults having noncardiothoracic surgery reported a dose-dependent association between increased intra-operative FiO₂ and the incidence of postoperative pulmonary complications (PPC), as well as 30-day mortality.⁸

Respiratory complications after anaesthesia are common and have reported incidences ranging from 4% to about 40%, depending on the definition and the baseline risk.^9,10 Although oxygen administration is a key component in the treatment of hypoxaemic patients, its effects on the lungs of normoxaemic patients remains controversial. Even short exposure to high oxygen concentration promotes absorption atelectasis.¹¹ Reduced mucociliary function and cellular damage from reactive oxygen species increase the risk of inflammatory and infectious complications.¹² In critical care settings excessive exposure to oxygen is associated with worsening acute respiratory distress syndrome and ventilator-associated pneumonia.^13–15

We therefore evaluated the effect of intra-operative high FiO₂ on postoperative respiratory outcomes in patients who participated in a recent large trial. Specifically, we tested the primary hypothesis that 80% inspired intra-operative oxygen decreases the postoperative ratio of arterial saturation to fraction of inspired oxygen (SpO₂/FiO₂) compared with 30% inspired oxygen. And secondarily, we tested the hypothesis that 80% FiO₂ increases a composite of PPC.

Methods

Study setup and patients

Adult patients scheduled for surgery in a surgical suite dedicated to colorectal surgeries in Cleveland Clinic's Main Campus were enrolled into an alternating cohort quality improvement project, in which the target FiO₂ was altered every 2 weeks between 30 and 80%, for a total of 39 months from 2013 to 2016. This resulted in a large cohort of more than 8000 patients assigned to intra-operative administration of either 30 or 80% oxygen that were well balanced on more than 50 potential confounders, as described in the main study analysis of wound-related complications.¹⁶ The rest of the intra-operative ventilation management, as well as the postoperative FiO₂, were not controlled and were left to the discretion of caregivers. In addition, clinicians were instructed to provide sufficient intra-operative oxygen to maintain patients’ saturation at or above 95%, and were allowed to use FiO₂ as high as 100% during induction of and emergence from anaesthesia. Cases lasting less than 2 h, cases with missing intra-operative FiO₂ data, re-operations during same hospitalisation, and cases with missing FiO₂ or SpO₂ values in the postanaesthesia care unit were excluded from the current analysis.

Outcomes measurements

SpO₂ data were collected from the patients’ electronic medical records, and FiO₂ in the postanaesthesia care unit was calculated from the reported oxygen enrichment device used and the oxygen flow, based on a conversion table (Supplemental Digital Content 1, https://links.lww.com/EJA/A193). We assumed FiO₂ remained unchanged until the next recording. Demographic, surgical and anaesthetic data as well as postoperative complications were also extracted from the electronic medical records.

For the primary outcome, we used the lowest SpO₂/FiO₂ ratio recorded in the postanaesthesia care unit for each patient. The secondary outcome was a composite of respiratory complications; pulmonary infection; pneumonia; respiratory failure or distress; bronchospasm; tracheitis or bronchitis; pulmonary collapse or atelectasis; emphysema; pulmonary oedema; pneumothorax or air leak; pleural effusion; acute respiratory distress syndrome; exacerbation of chronic obstructive pulmonary disease or asthma; continuous invasive mechanical ventilation; transfusion-related acute lung injury and pulmonary embolism or infarction. These are listed and defined in Supplemental Digital Content 2, https://links.lww.com/EJA/A193, by diagnosis codes or events documented in the medical record from admission to the postanaesthesia care unit until hospital discharge.

Ethics

This study was approved by the Cleveland Clinic Institutional Review Board (IRB #12-891, Cleveland, Ohio, United States, executive director Daniel Beyer) on 29 August 2012 and was registered in clinicaltrials.gov (NCT01777568). This posthoc analysis was also approved by the Cleveland Clinic institutional review board (IRB #17-824, Cleveland, Ohio, United States, executive director Bridget Howard) on 26 June 2017.

Statistical analysis

We descriptively compared the 30 and 80% oxygen groups on demographic, baseline and procedural variables using standard descriptive statistics and absolute standardised difference (ASD). The ASD was calculated as absolute difference in means or proportions divided by the pooled SD.

To assess the effect of supplemental oxygen, we compared the lowest SpO₂/FiO₂ ratio between two treatment groups using Wilcoxon rank-sum test. Median difference with its confidence interval (CI) was reported from Hodges–Lehmann estimator. Secondarily, the relative risk (RR) of PPC comparing 30 to 80% oxygen groups was assessed using a generalised linear model with log link. The significance criterion was 0.05 for both primary and secondary analyses.

Power considerations

We expected that each group would have 2500 eligible patients who would meet all inclusion/exclusion criteria of this study. Based on literature and a preliminary query of our peri-operative database, we assumed that the mean of lowest SpO₂/FiO₂ was 330 with a SD of 200. The estimated power was thus 0.94 for a difference of 20 U in the SpO₂/FiO₂ ratio and 0.76 for a difference of 15 U, with an alpha level of 0.05.

Results

A total of 8097 colorectal surgeries performed at Cleveland Clinic Main Campus from 28 January 2013 to 11 March 2016 were retrieved. After excluding 3041 nonqualifying cases, 5056 surgeries were analysed, with 2486 surgeries allocated to 30% oxygen and 2570 allocated to 80% oxygen. Patients’ flow and reasons for exclusion are described in Fig. 1.

Fig. 1: Flow diagram. FiO₂, fraction of inspired oxygen; PACU, postanaesthesia care unit; SpO₂, arterial haemoglobin saturation.

Patient characteristics by group are summarised in Table 1, and their surgical and anaesthetic information are presented in Table 2. The 30 and 80% oxygen groups were well balanced on all demographic, baseline and procedural variables (ASD < 0.10). Therefore, adjustment for potential confounders was unnecessary. The median intra-operative time-weighted average inspired oxygen concentration was 43% (IQR 38 to 54%) in patients assigned to 30% or the lowest tolerated concentration, and 81% (IQR 78 to 82%) in those assigned to 80% oxygen. FiO₂ in the postanaesthesia care unit was kept at a level of 27% by low oxygen flow through a nasal cannula in 77% of patients (median time-weighted-average FiO₂ [IQR] 27% [27%, 27%]).

Table 1: Baseline characteristics of patients in the 30 and 80% oxygen groups

Table 2: Summary of anaesthetic and surgical information

There was no significant difference in the lowest SpO₂/FiO₂ ratio in the postanaesthesia care unit, comparing the high and low FiO₂ groups (Fig. 2). The estimated median difference was 0 (95% CI: 0, 0; P = 0.91, Table 3). The primary outcome did not change over the course of the study. Hyperoxia also had no effect on the composite outcome of PPC. The overall observed incidence of pulmonary complications was 16.3% (406/2486) in patients assigned to 30% inspired oxygen and 17.6% (451/2570) in those assigned to 80% oxygen. The estimated RR was 1.07 (95% CI: 0.95, 1.21), P = 0.25. The incidence of the composite and its components is shown in Table 3, including the RR (95% CI) for components with incidence of 1% or higher. In-hospital mortality rate was similar in both groups. Twelve patients (0.48%) in 30% oxygen group and 12 (0.47%) in 80% oxygen group died in hospital after surgery.

Fig. 2: Box plot of the arterial saturation to fraction of inspired oxygen ratio by treatment groups. Lowest arterial saturation to fraction of inspired oxygen ratio measured in the postanaesthesia care unit. FiO₂, fraction of inspired oxygen; SpO₂, arterial haemoglobin saturation.

Table 3: Effects of supplemental oxygen on arterial saturation to fraction of inspired oxygen ratio in postanaesthesia care unit and postoperative pulmonary complications

Discussion

High intra-operative FiO₂ did not reduce the average minimal SpO₂/FiO₂ ratio in the postoperative care unit. Furthermore, supplemental oxygen did not worsen the risk of pulmonary complications. Since the underlying trial included more than 5000 patients, we had 94% power to detect changes as small as 20 U in the SpO₂/FiO₂ ratio. It is thus unlikely that 80% intra-operative oxygen substantively worsens postoperative oxygenation in our study population of adults having colorectal surgery.

Our finding is consistent with the smaller PROXI trial which randomised 1400 patients to 30 or 80% oxygen intra-operatively and for 2 postoperative hours.¹⁷ There was no difference in the incidence of atelectasis, pneumonia, or respiratory failure in the two groups. A meta-analysis by Hovaguimian et al.¹⁸ reported no difference in the incidence of postoperative atelectasis and partial arterial oxygen pressure to FiO₂ (PaO₂/FiO₂) ratio ratio in patients receiving high or normal intra-operative FiO₂.

In contrast, a recent large registry-based analysis reported a dose–response association between intra-operative FiO₂ and incidence of pulmonary complications.⁸ Of note, the overall incidence of pulmonary complications in that registry analysis was about 4%, compared with 16% in our patients. Complications were presumably more common in our patients because they were sicker. For example, 70% of our patients had American Society of Anesthesiology physical status scores at least 3, compared with less than 30% in the report by Staehr-rye et al.⁸ But a more serious question is why Staehr-rye et al. report an association between inspired oxygen concentration and pulmonary complications whereas we saw no such effect. Unadjusted confounding in the registry analysis seems the most likely explanation. For example, many clinicians tend to give supplemental oxygen to sicker patients, a practice that potentially introduces substantial bias into retrospective analyses.

Our study differed in that we prospectively allocated patients to the high and low oxygen administration groups. This method proved effective in creating remarkably comparable groups across more than 50 baseline and intra-operative potential confounding factors. In addition, the outcomes in our study were collected from the nursing vital signs reports as well as from diagnosis codes, which were probably both indifferent, if not completely blinded, to the study allocation. We therefore were able to effectively minimise selection and measurement biases, along with confounding, that are always dangers in retrospective analyses. Since we were not powered for subgroup analyses, our results might not generalise well to specific patient populations, who may need specific ventilatory strategies. But considering the large sample size and the remarkable balance on all tested potential confounders, the risk for subgroup imbalance is low.

The underlying trial used a novel alternating intervention design rather than conventional individual-patient randomisation. The advantage of this approach is that investigators can rapidly and inexpensively enrol large numbers of patients (our alternating intervention trial included nearly as many patients as all 20 previous randomised trials). We note though that treatment allocation was not randomised on an individual basis; in fact, even the 2-week exposure blocks were not randomised. Instead, exposure periods simply alternated. In this respect, the approach resembles cluster randomisation, with clusters distributed in time rather than in space.

A limitation of the underlying trial is that we did not routinely obtain arterial blood samples; instead we used the SpO₂/FiO₂ ratio as a surrogate measure of oxygenation for the more common PaO₂/FiO₂ ratio. However, the SpO₂/FiO₂ ratio is considered a reliable diagnostic tool for acute respiratory distress syndrome and can replace the PaO₂/FiO₂ ratio in the respiratory part of the Sequential Organ Failure Assessment score.^19,20 SpO₂ has also been shown to be an effective assessment of early respiratory insufficiency, either as a component of a SpO₂/FiO₂ ratio or as a variable in the Ellis or Rice equations to calculate an actual PaO₂/FiO₂ ratio.^21,22 An advantage of the SpO₂/FiO₂ ratio is that it is a continuous measure, rather than being restricted to times at which arterial blood might be sampled. Another consequence of the design is that aside from FiO₂, we did not control for any other ventilatory parameter, such as the use or level of positive end-expiratory pressure, or recruitment manoeuvres. It is possible that anaesthesia providers used these tools differentially between the two groups, to improve oxygenation in the low FiO₂ group.

We assessed oxygenation only during the initial hours in the postanaesthesia care unit; delayed effects of hyperoxia on oxygenation would thus be overlooked in our analysis. However, complications were considered throughout hospitalisation. The two most common components of the composite secondary outcome were chronic obstructive pulmonary disease or asthma exacerbation and bronchospasm, each defined either by diagnosis codes or by treatment. These indirect diagnoses were surely less accurate than specific pulmonary function measures such as expired volume in one second. However, it seems highly unlikely that our postoperative outcome measures were biased by exposure.

The WHO and Center for Disease Control recommendations to routinely expose patients to intra-operative and postoperative high FiO₂ have been questioned.^23–26 Much of the concern relates to unconvincing evidence that supplemental oxygen reduces surgical site infection. But others have argued that the recommendations ignored potential deleterious pulmonary consequences. Although we only tested the intra-operative part of the proposed intervention, our results suggest that intra-operative administration of 80% inspired oxygen does not worsen postoperative oxygenation or promote pulmonary complications.

Acknowledgements relating to this article

Assistance with the study: none.

Financial support and sponsorship: supported by internal funding only. BC received a fellowship grant from the American Physicians Fellowship for Medicine in Israel. ER received a grant from Instituto Salud Carlos III (BA17/00032).

Conflicts of interest: none.

Presentation: none.