Short-Term Outcomes of Low Versus High Inspiratory Oxygen Fraction During Induction of General Anesthesia in Noncardiac Surgery: A Pragmatic Open-Label Randomized Non-Inferiority Trial

Background. High inspiratory oxygen fraction (FIO 2 ) is associated with increased perioperative pulmonary morbidity and postoperative mortality. However, the use of pure oxygen is still currently recommended during the anesthesia induction. Methods, This open-label randomized non-inferiority trial was conducted in two metropolitan hospitals at Southern Taiwan. A total of 302 surgical patients (ASA PC ≤ III) who received endotracheal general anesthesia (ETGA) were randomized to receive 100% (FIO 2 1.0) or 60% (FIO 2 0.6) oxygen during induction. The primary endpoint was presence of hypoxemia (SpO 2 ≤ 92%) during the induction of anesthesia. The secondary endpoint was the development of major complications immediately and within 3 days after surgery. Results. A total of 5 patients in the FIO 2 0.6 group developed hypoxemia during induction (3.9% vs 0% for FIO 2 0.6 vs FIO 2 1.0, respectively; P=0.167 for non-inferiority), suggesting that FIO 2 0.6 was inferior than FIO 2 1.0 for anesthesia induction. The mean lowest SpO 2 during induction was also signicantly lower in FIO 2 0.6 group. Four patients reached the primary endpoint had increased body mass indexes (BMI>30 kg/m 2 ). However, the overall incidence of desaturation developed after removal of endotracheal tube was higher in FIO 2 1.0 group (1.4% vs 5.8%, FIO 2 0.6 vs FIO 2 1.0; odd ratio 0.22, 95% condence interval 0.05-1.05; P=0.064). Conclusions. High fractions of oxygen should be used for oxygenation during induction of ETGA in general population, especially in the obese patients. However, the supplement of high FIO2 during induction was hypoxemic have signicant perioperative

Conclusions. High fractions of oxygen should be used for oxygenation during induction of ETGA in general population, especially in the obese patients. However, the supplement of high FIO2 during induction was associated with increased hypoxemic events after removal of endotracheal tube that might have a more signi cant impact on perioperative care.

Background
Surgical patients often experience transient apnea before endotracheal intubation or other airway instrumentation during the induction of general anesthesia. In order to minimize the risks of hypoxemia during establishing arti cial airway, the use of pure oxygen (oxygen partial pressure FIO 2 = 1.0) has been recommended throughout the preoxygenation and induction period (Bouroche and Bourgain 2015; Martin and Grocott 2015). Elevated oxygen reserve in the lungs and oxygen partial pressure in the blood circulation can signi cantly prolong the development of hypoxemia after apnea (Nimmagadda et al. 2017;Edmark et al. 2003). It has been found that it took signi cantly less time for patients preoxygenated with FIO 2 0.6 to drop their peripheral oxygen saturation (SpO 2 ) to below 90% compared with patients oxygenated with FIO 2 1.0 (411 vs 213 min, P < 0.01) (Edmark et al. 2003).
However, a recent study illustrated that high intraoperative inspiratory oxygen fractions are associated with a dose-dependent increase of respiratory complications and increased 30-day mortality (Staehr-Rye et al. 2017). Other sources have also pointed to the potentially damaging effects of high concentration oxygen therapy, as oxygen toxicity may result in direct tracheobronchial and alveolar damage, absorption atelectasis and central nervous system toxicity (van Ooij et al. 2016;Hafner et al. 2015). Therefore, oxygen therapy in clinical settings has been recognized as a two-edged sword and excessive oxygen supplementation should be closely monitored for potential toxicity (Horner and O'Driscoll 2018;Martin and Grocott 2013), and use of FIO 2 > 0.8 is a topic of ongoing debate in the perioperative care (Weenink et al. 2020). In clinical anesthesia, a low optimal oxygen supplement (as low as FIO 2 0.4) is recommended to maintain SpO 2 ≥ 92% during intraoperative mechanical ventilation for nonobese patients (LAS VEGAS Investigators 2017; Guldner et al. 2015).
Currently, there is no consensus on whether lower fractions of inspiratory oxygen during the induction period of anesthesia can decrease the risk of lung injuries and other cellular damage (Ball et al. 2017). In addition, higher perioperative oxygen supplement (FIO 2 ≥ 80%) has been suggested to be associated with a greater incidence of atelectasis and postoperative pulmonary complications (Li et al. 2019). Therefore, this non-inferiority trial hypothesized that in patients receiving anesthesia with endotracheal intubation, an induction FIO 2 of 0.6 was inferior to the standard FIO 2 1.0 in regards to desaturation prevention before endotracheal ventilation establishment and the development of postoperative complications.

Study population
This pragmatic, open-label, randomized (1:1 ratio), non-inferiority clinical trial was conducted in E-Da Hospital and E-Da Cancer Hospital (Kaohsiung, Taiwan) from October to December 2018. Eligible participants included adult aged 18-65 years with American Society for Anesthesiologist physical statuses (ASA PS) ≤ III who were scheduled for surgical procedures and required endotracheal tube intubation general anesthesia (ETGA). The study was approved by the Institutional Review Board of E-Da Hospital (Kaohsiung, Taiwan) on 2 August 2018 (approval number EMRP15107N), and patients were enrolled after obtaining o cial IRB approval. The trial was also registered in September 2018 with clinical trials registration: NCT03665259. Written informed consent was obtained from the patients or their legal surrogates.

Exclusion criteria
Patients with anticipated di cult intubation, active lung disease, history of myocardial infarction or coronary artery disease, any advanced organ dysfunction (i.e. heart failure, renal insu ciency and/or liver cirrhosis), severe anemia (hemoglobin ≤ 8 mg/dl), body mass index (BMI) ≥ 35 kg/m 2 , pregnancy, inadequate preoperative fasting time, or those receiving major operations (i.e. emergency, cardiac surgery, chest surgery and craniotomy) were excluded from the study.

Allocation to intervention
Eligible patients were randomized in permuted blocks of 10 using a computer-generated list to receive either pure oxygen (FIO 2 1.0) or lower oxygen fraction (FIO 2 0.6) in a 1:1 ratio. Treatment allocation for each patient was concealed in an opaque envelope according to the randomization sequence. The envelope was opened by the attending anesthetist immediately before preoxygenation, and the assigned treatment oxygen fraction was disclosed to the anesthesia team. Patient allocation and study ow diagrams are shown in Fig. 1.

Anesthesia and intervention protocols
Before administration of intravenous hypnotic agents, patients were given oxygen (FIO 2 1.0 or FIO 2 0.6) for at least 3 minutes via a face mask at a ow rate of 6 L/min. Anesthesia was induced through intravenous injection of fentanyl (2 µg/kg), propofol (1.5-2.0 mg/kg) and rocuronium (0.8-1.0 mg/kg). Following the loss of consciousness, bag-mask assisted ventilation was provided by the attending anesthetist through the face mask connected to the semi-closed circuit of anesthesia machine. Depth of anesthesia was not routinely monitored with the electroencephalographic devices, and the attending anesthetist decided the optimal duration of assisted mask ventilation during the induction phase. An appropriately size of endotracheal tube was intubated into the trachea under direct laryngoscopy, and the patient was then mechanically ventilated using the volume-control mode. The standard intraoperative ventilator setting followed the recent clinical practice recommendations, included a tidal volume of 8 mL/kg predicted body weight and a positive end expiratory pressure (PEEP) of 2-5cmH 2 O (Li et al.

2019
). Intraoperative anesthesia was maintained by volatile anesthetics (des urane or sevo urane) at optimal levels of minimal alveolar concentration (MAC). Inhaled anesthetics were delivered by 60% oxygen (FIO 2 0.6) at a ow rate of 1L/min during maintenance of anesthesia. High ow rate of oxygen (FIO 2 0.6, 6L/min) was used to wash out the anesthetic gases during the emergence phase of anesthesia. Neuromuscular monitoring was system was not routinely used during the perioperative period. Oxygen fraction was switched to 100% whenever arterial desaturation (de ned as a SpO 2 of ≤ 92%) (Futier et al. 2013) developed during the induction phase of anesthesia. Increases in the fraction of inspired oxygen and changes in ventilatory settings were also made to compensate for any episodes of arterial desaturation happened at any time points during maintenance or emergence phases of anesthesia. After patients had recovered to adequate spontaneous ventilation, the endotracheal tube was removed by the anesthetist and patients were transferred to the postanesthesia care unit (PACU). The standard protocol of anesthesia is shown in Fig. 2.

Measurements
The primary endpoint of this study was the development of arterial desaturation (SpO 2 ≤ 92%) during the induction of anesthesia (Futier et al. 2013). Desaturation during induction was closely monitored by the anesthetic team. The composite secondary endpoint was the development of postoperative complications (from removal of endotracheal tube to three days after surgery), which included events such as desaturation, acute lung injury, pneumonia, surgical site infection (SSI), unplanned admission to intensive care unit (ICU), severe postoperative pain, prolonged hospital stays and mortality (supplementary Table 1). The clinical staff who recorded the secondary endpoints developed after discharged from PACU were blinded to the treatment groups.

Statistics
The reported overall incidence of hypoxemia (SpO 2 < 90%) during induction of anesthesia ranges from 1.4-7.4% (Ehrenfeld et al. 2010;Baillard et al. 2019). With a reference probability of 0.05, the study de ned an arbitrary non-inferiority margin of 5% in the development of desaturation during induction between the FIO 2 0.6 (alternative) and FIO 2 1.0 (standard or reference) treatments. We speculated a Pvalue < 0.05 would indicate non-inferiority of FIO 2 0.6 to induce desaturation during induction of anesthesia (alternative hypothesis), and would correspond to the upper limit of the one-sided 95% con dence interval (CI) of the difference not exceeding the 5% based on the type I error at 0.05 (onesided) and a 90% statistical power. Under 1:1 sampling ratio in the two groups with an expected 10% withdrawal or drop-off rate, the calculated sample size in each group was approximately 500 patients (a total of 1000 patients) (PASS 15, NCSS, Utah, USA). Categorical variables were compared using the χ 2 statistics or Fisher's exact test. Continuous variables were compared using the Student's t-test or Mann-Whitney test. The Kaplan-Meier method was used to analyze the time to development of desaturation during induction, and time-to-event data between the groups were compared with the long-rank test. Intention-to-treat and as-treated analyses are presented for the development of desaturation after removal of endotracheal tube (a secondary endpoint), while the other secondary endpoints were analyzed in the as-treated population. All analyses were carried out using the SAS software, version 9.1 (SPSS software, version 24.0 (IBM, Armonk, NY).

General outcomes and baseline patient characteristics
A total of 541 patients were assessed for eligibility of enrolment and 15 patients decided to withdraw from the study before randomization, and there were no cases dropped off after entering the trial (Fig. 1). This study was prematurely terminated after recruiting of 302 patients due to safety concerns addressed by the institutional Data and Safety Monitoring Board (DSMB), as there was a signi cant increase in the number of patients who reached the primary endpoint. The patient characteristics are listed on Table 1. There were no signi cant differences between the two group in the baseline SpO 2 at room air (97.2 ± 2.0% vs 97.4 ± 2.2% for FIO 2 0.6 vs FIO 2 1.0, respectively; P = 0.370) ( Table 1). The study did not record any patients with unanticipated di cult intubation who required advanced intubating devices, such as beroptic bronchoscope or surgical airway during induction. Types of operation are listed in Supplementary Table 2. Primary endpoint A total of 5 patients developed severe hypoxemia (SpO 2 ≤ 92%) during the induction phase of anesthesia. All the ve patients received FIO 2 0.6 for preoxygenation and they were switched to receive FIO 2 1.0 for rescue therapy. Non-inferiority was not met (P = 0.167 for non-inferiority), as the upper margin 95% CI of 0.074 was greater than the pre-speci ed non-inferiority margin of 0.05 ( Table 2). The incidences of desaturation during induction were 3.9% and 0% in the FIO 2 0.6 and FIO 2 1.0 groups, respectively (P = 0.03, Fisher Exact test) ( Table 2). Time-to-event analysis by the Kaplan-Meier survival curves also con rmed that there was a signi cant difference between the two groups in regards to desaturation during induction (Fig. 3). Furthermore, the mean lowest SpO 2 in the FIO 2 0.6 group was also signi cantly lower than that in the FIO 2 1.0 group (98.7 ± 3.0% vs 99.7 ± 0.8%, respectively; P = 0.01). All the patients who developed the primary endpoint were successfully managed through increasing the fraction of oxygen supplementation to 100% via the bag-mask assisted ventilation. None of the patients required additional treatment or cancellation of surgery (supplementary Table 3).

Secondary endpoints
The ve patients who developed primary endpoint were excluded from the intention-to-treat analysis, but they were included in the as-treated analysis since they were all switched to receive 100% oxygen treatment after developing desaturation (Fig. 1). Two patients in the FIO 2 0.6 group developed desaturation after removal of endotracheal tube in the operating room (1.4%, 2/147). The incidences of desaturation in the FIO 2 1.0 group in the operating room (OR) were 3.3% (5/150) and 3.9% (6/155) for intention-to-treat and as-treated analysis, respectively (Table 3). Lower incidence of desaturation was also recorded in the FIO 2 0.6 group while the patients were cared for in the PACU (0% vs 1.9% for FIO 2 0.6 vs FIO 2 1.0; P = 0.25). According to as-treated analysis, the overall incidences of desaturation after extubation (in OR and at PACU) were 1.4% (2/147) and 5.8% (9/155) in the FIO 2 0.6 and FIO 2 1.0 groups, respectively (P = 0.064) with an odds ratio of 0.22 (95% CI 0.05-1.05) ( Table 3). The characteristics of patients who developed desaturation after extubation are summarized in supplementary Table 4. There were no cases of mortality or unplanned ICU admission in this study, and only one de nite case developed SSI after operation ( Table 3). The lengths of hospital stays were similar between these two groups (5.9 ± 3.0 vs 5.4 ± 2.4 days for FIO 2 0.6 vs FIO 2 1.0, respectively; P = 0.163) ( Table 3).  1.0 group for analysis of secondary endpoints (as-treated analysis); ¶ Data is shown as mean ± SD. Results were analyzed by the Fisher exact test (two-sided) or unpaired t-test, as appropriate.

Discussion
The main factor of interest in this study was the relationship between oxygen supplementation fraction received during anesthesia induction and the development of desaturation or post-operative respiratory distress in non-critically ill patients. Routine pure oxygen supplementation for non-critical or generally healthy patients during induction is still a debated topic (Ball et al. 2017). There is currently no clinical evidence demonstrating the advantages of routine pure oxygen during induction of anesthesia, while intraoperative oxygen overexposure de ned as hyperoxemia (SpO 2 > 98%) and substantial oxygen This study was prematurely terminated after recruitment of 302 patients due to the safety concerns raised by the DSMB members, as all patients who developed primary endpoint received FIO 2 0.6, and time-to-event analysis and the lowest mean SpO 2 con rmed the increased hypoxemic events during the induction period of anesthesia in the FIO 2 0.6 study group. Therefore, the null hypothesis of noninferiority could not be rejected. The incidence of hypoxemia during induction phase by giving 60% inspired oxygen was 3.9% in our study, which is considerably higher than the previous report (Ehrenfeld et al. 2010). It was found that supplementation of 60% FIO 2 during induction phase may provide an estimated number needed-to-harm (NNH) of 31 in developing desaturation that requires urgent medical intervention. The lowest SpO 2 in the FIO 2 0.6 group ranged from 78-92%, and these patients were switched to receive pure oxygen for rescue therapy. None of these patients developed any clinically signi cant consequences. Since there were no differences in patient characteristics and proportion of di cult intubation between the two groups, lower inspiratory oxygen fraction (i.e. FIO 2 0.6) during preoxygenation and assisted ventilation before endotracheal intubation increased risk of desaturation in patients with ASA PS I-III. Our results indicated that four out of the ve patients who developed the primary endpoint had BMI's greater than 30 kg/m 2 . Obese patients have increased oxygen demand and CO 2 production, and as a result they are prone to rapid desaturation during apnea or hyponea due to reduced functional residual capacity and expiratory reserve volume, while the total lung compliance is decreased exponentially (Pelosi et al. 1998;Peppard et al. 2009).
Since the patients who developed primary endpoint were switched to 100% oxygen therapy during induction, these patients were included in the FIO 2 group for the subsequent analysis of the secondary endpoints (as-treated analysis). The main short-term outcomes after ETGA is concerned with the occurrence of acute respiratory distress or desaturation following removal of endotracheal tube. A total of 11 cases of desaturation after extubation in the OR or at PACU were found during the study. In the FIO 2 0.6 group, two case developed a SpO 2 ≤ 92% in the OR, but none developed desaturation at PACU.
However, there were 9 cases of desaturation in the FIO 2 1.0 group, including one patient who developed the primary endpoint was switched to pure oxygen treatment. Although the incidence of desaturation was not statistically different (1.4% vs 5.8% for FIO 2 0.6 vs FIO 2 1.0; P = 0.064) between the two study groups, this unanticipated result highlights that the odds of developing desaturation after removal of endotracheal tube in ETGA patients is 78% lower in patients receiving FIO 2 0.6 than those with pure As high-risk and emergency surgeries were excluded, perioperative uid overload and prolonged operation times are the two main potential confounding factors for postoperative respiratory distress in this trial. Our database showed that there were no differences in average operation time or total uid administered during operation between the two groups (Table 1).
Pulmonary atelectasis occurs in 85-90% of healthy anesthetized adults and is one of the leading causes of postoperative hypoxemic events (Karcz and Papadakos 2013). Besides procedure-and anestheticrelated factors, the composition of inspired gas is another important factor that in uences the formation of pulmonary atelectasis during general anesthesia (Sun et al. 2015; Quintero-Cifuentes et al. 2018).
Edmark et al. found that the mean areas of lung atelectasis immediately after apnea was higher in patients received 100% oxygen compared to those oxygenated with FIO 2 0.6 (10 cm 2 vs 0.3 cm 2 , P < 0.001) (Edmark et al. 2003). Another clinical observational study compared the effects of gas composition on the formation of atelectasis and gas exchange during the induction of general anesthesia (Rothen et al. 1995). Compared with FIO 2 0.3, the degree of atelectasis (1.6 ± 1.6 cm 2 vs 4.7 ± 4.5 cm 2 ) and intrapulmonary shunt (3.2 ± 2.7% to 9.8 ± 5.7%) were signi cantly increased in the FIO 2 1.0 group (Rothen et al. 1995). Therefore, some early studies have suggested that a lower concentration of oxygen mixed with nitrogen may ameliorate the early formation of atelectasis and pulmonary shunt during anesthesia induction (Rothen et al. 1996;Edmark et al. 2011). Our study provides further evidence that high FIO 2 administration during induction may increase the formation of absorption atelectasis and pulmonary shunt, which in turn, may impact the gas exchange and tissue oxygenation at the emergence and recovery phases of anesthesia.
The results of this study indicate that, in generally healthy patients (ASA PS I-III), low fraction oxygen supplementation (FIO 2 0.6) during induction of ETGA appears to affect tissue oxygen tension at two ways: it increases incidence of desaturation during induction, but seems to have protective effects against respiratory distress after endotracheal tube removal. Since increased BMI could be a confounding factor for the development of desaturation during induction, it would be reasonable to repeat follow-up studies in non-obese patients who receive ETGA. Our study examining lower risk population showed that the overall incidence of hypoxemia after removal of endotracheal tube in the FIO 2 1.0 group was 6.0%, which is comparable with the incidences reported in general population ( It should be noted that there are several limitations in this study. First, the fact that the trial was early terminated may have reduced the statistical power needed to determine the difference between the two treatment groups. Secondly, this was an open-label trial in which our anesthetic team was not blinded to the treatment groups. However, it is not practical to conceal the inspired oxygen fractions during anesthesia. Furthermore, blinding the anesthetist-in-charge to oxygen concentrations may be unethical, as tissue desaturation usually progresses very rapidly after the patient has been paralyzed. Nevertheless, the research team members who recorded the secondary endpoints in the ward were blinded to the treatment. Thirdly, obese patients have been recognized as an independent risk factor for early desaturation during apnea (Bouroche and Bourgain 2015; Goudra et al. 2014), but they were included in this study and might confound the primary endpoint. Our initial study design aimed to investigate the effect of inspired oxygen fractions on the perioperative outcomes in relatively healthy patients (ASA PS I-III), as they are more likely to achieve bene cial outcomes from lower oxygen fractions than critically ill patients (Nimmagadda et al. 2017). Although we excluded patients with BMI > 35 kg/m 2 (class II obesity), our study speci ed that class I obese (BMI 30-34.9 kg/m 2 patients may also bene t from high inspired oxygen fractions during induction. Fourthly, this was a pragmatic clinical trial testing the outcomes of two different oxygen fractions used for anesthesia induction in a broad routine clinical practice. Therefore, endotracheal intubation and extubation timing was primarily decided by the anesthetist-incharge based on the patient's clinical responses and anesthetic depth, which resulted in a number of different time points in regards to development of desaturation. Fifthly, this study failed to specify the optimal oxygen concentration for anesthesia induction, and can only provide the estimated range of 60% to100%. Future studies may wish to look into more speci c oxygen concentrations for optimal anesthesia induction. Sixthly, we did not routinely measure the recovery of neuromuscular function before extubation of endotracheal tube. The residual curarisation effect of neuromuscular blocking agents on spontaneous ventilation could affect the occurrence of respiratory distress after extubation. Seventhly, our study did not show any differences in SSI occurrence, as some studies have suggested that higher fraction oxygen supplementation can reduce risk of SSI's (Belda et al. 2005;Chu et al. 2018). Since high risk procedures and patients with major comorbidities were excluded from this study, the incidence of postoperative SSI was relatively low in our study, thereby limiting our ability to detect difference between the two groups.
Lastly, growing evidence suggests that critically ill patients might also bene t from low optimal oxygen therapy (Chu et al. 2018; Damiani et al. 2018). However, our results are not applicable to patients with ASA ≥ IV, advanced systemic diseases, active lung diseases, di cult airway and those who receive emergency and major operations.
As compared with FIO 2 1.0, this clinical study shows that FIO 2 0.6 supplementation in with ASA PS I-III surgical patients receiving ETGA signi cantly increases the risk of hypoxemia during induction. Obese patients (BMI > 30 kg/m 2 ) are associated with higher risk of developing desaturation when FIO 2 0.6 is administered during the preoxygenation and induction phases of anesthesia. However, administration of 100% FIO 2 during anesthesia induction may increase incidence of hypoxemic events after endotracheal tube removal and the return of spontaneous ventilation in the OR and at PACU. Although our study was underpowered to conclude the statistical difference, the brief period of substantially high oxygen exposure at the beginning of anesthesia could be a potential contributing factor to postoperative acute respiratory distress in patients receiving general anesthesia.