The comparison of preoxygenation methods before endotracheal intubation: a network meta-analysis of randomized trials

Background Preoxygenation before endotracheal intubation (ETI) maintains asphyxiated oxygenation and reduces the risk of hypoxia-induced adverse events. Previous studies have compared various preoxygenation methods. However, network meta-analyses (NMAs) of the combined comparison of preoxygenation methods is still lacking. Methods We searched for studies published in PubMed, Embase, Web of Science, Scopus, and the Cochrane Library. Review Manager version 5.3 was used to evaluate the risk of bias. The primary outcome of this meta-analysis was low oxygen saturation (SpO2) during ETI. The secondary outcomes included SpO2 <80%, SpO2 <90%, and apnea time during ETI. NMA was performed using R 4.1.2 software gemtc packages in RStudio. Results A total of 15 randomized controlled trials were included in this study. Regarding the lowest SpO2, the noninvasive ventilation (NIV) with high-flow nasal cannula (HFNC) group performed better than the other groups. For SpO2 <80%, the NIV group (0.8603467) performed better than the HFNC (0.1373533) and conventional oxygen therapy (COT, 0.0023) groups, according to the surface under the cumulative ranking curve results. For SpO2 <90%, the NIV group (0.60932667) performed better than the HFNC (0.37888667) and COT (0.01178667) groups. With regard to apnea time, the HFNC group was superior to the COT group (mean difference: −50.05; 95% confidence interval: −90.01, −10.09; P = 0.01). Conclusion Network analysis revealed that NIV for preoxygenation achieved higher SpO2 levels than HFNC and COT and offered a more significant advantage in maintaining patient oxygenation during ETI. Patients experienced a longer apnea time after HFNC preoxygenation. The combination of NIV with HFNC proved to be significantly superior to other methods. Given the scarcity of such studies, further research is needed to evaluate its effectiveness. Systematic review registration identifier CRD42022346013


Introduction
Invasive mechanical ventilation is a crucial measure to safeguard patient safety during surgical procedures conducted under general anesthesia, typically necessitating tracheal intubation.Prior to endotracheal intubation (ETI), the induction of anesthesia renders the patient unconscious, and neuromuscular blockade ensues, leading to hypopnea and apnea.This subsequent apnea period poses a heightened risk of hypoxia for the patient, particularly if tracheal intubation poses difficulties, further compounding existing risks (1,2).Preoxygenation refers to the process of enhancing oxygen concentration and reserves by saturating the patient's body with oxygen prior to surgery, ensuring that the patient maintains a safe oxygen saturation (SpO 2 ) level during apnea.Administering preoxygenation before ETI can sustain oxygenation during asphyxia and mitigate the hazards associated with hypoxia-induced adverse events.Consequently, it is highly advisable to routinely recommend preoxygenation as standard practice prior to ETI (3)(4)(5).
The most common form of preoxygenation is mask ventilation with 100% oxygen for 3-5 min, also known as conventional oxygen therapy (COT) (6).The simplicity of mask ventilation lies in its ease of operation; however, prolonged ventilation can compromise patient comfort.In addition to non-invasive ventilation (NIV), a novel approach, the high-flow nasal cannula (HFNC), has been increasingly employed in the preoxygenation process for patients in the operating room (7).HFNC administers heated and humidified gases through a nasal catheter, maintaining a specified fraction of inspired oxygen (FiO 2 ) at a maximum flow rate exceeding 60 L/min.HFNC demonstrates satisfactory oxygenation effects, improves patient comfort, and ensures a more tolerable preoperative experience.Furthermore, mask ventilation obstructs the oral airway, necessitating the removal of the mask during laryngoscopy.This underscores the dual functionality of HFNC as a preoxygenation device capable of maintaining oxygenation during asphyxia.(8).
Previous studies have compared various preoxygenation methods, and some have suggested that HFNC is a more effective preoxygenation device (9,10).In addition, according to the studies on NIV (11) and HFNC (12), preoxygenation is more effective than COT.Based on published studies, several meta-analyses have compared the effects of different preoxygenation modalities (13)(14)(15).A meta-analysis by Kuo  et al. (13) pointed out that transnasal humidified rapid-insufflation ventilatory exchange did not have a significant advantage over the use of facemasks, but it could effectively improve PaO 2 .According to Chiang et al. (14), NIV is more effective than conventional preoxygenation methods.Nonetheless, network meta-analyses (NMAs) comprehensively comparing different preoxygenation methods are scarce.Consequently, a systematic review of published studies along with an NMA assessing various preoxygenation modalities is necessary to provide a holistic understanding of their relative effectiveness and safety.

Study selection
This systematic review and NMA has been registered with PROSPERO (registration number is CRD42022346013), and we performed it according to Preferred Reporting Items for Systematic Reviews and meta-analyses (PRISMA) guidelines.The researchers searched for studies from PubMed, Embase, Web of science, Scopus

Eligibility criteria
We included RCTs involving adult patients who underwent preoxygenation prior to ETI.The preoxygenation devices included COT, NIV, and HFNC.Studies with the following characteristics were excluded: non-intubation; focus on only apneic oxygenation or ventilation; animal studies; protocols, reviews, guidelines, or conference abstracts; lack of control; and including healthy volunteers.

Risk of bias assessment
Review Manager version 5.3 (RevMan 5.3) was used to evaluate the risk of bias in the included studies according to the Cochrane Collaboration tool.A summary of the risk of bias is shown in Figure 2. Three researchers (MZ, RX, and JZ) completed the risk of bias assessment, whereas the other researchers were responsible for deciding on a different opinion.

Data extraction
Five investigators (MZ, RX, JZ, JZ, and XY) independently extracted data.MZ, RX, and JYZ.reviewed all the studies and excluded duplicates, registered studies, and nonclinical studies.Additionally, we reviewed the titles, abstracts, and full texts of the RCTs that were determined to be included in the NMA.XY summarized the characteristics of the 15 included studies in Table 1.AY were responsible for resolving disputes in the data extraction process.

Outcomes
The primary outcome of this meta-analysis was low SpO 2 during ETI.The secondary outcomes included SpO 2 <80%, SpO 2 <90%, and apnea time during ETI.We reported the odds ratios (ORs), mean differences (MDs), and 95% confidence intervals (CI) in a pairwise meta-analysis.The log-OR, MD, and 95% CI were reported for the NMA.

Statistical analysis
Five investigators (MZ, RX, JZ, JZ, and XY) used statistical methodology.First, RevMan 5.3 was used for pairwise metaanalysis.For the heterogeneity test, when P < 0.05 or I2 > 50%, we chose the random-effects model.When the heterogeneity test yielded P > 0.05 or I2 < 50%, the fixed effects model was often selected.
Second, STATA (version 17.0) was used to generate network plots for the different groups, to visualize the relationships between various interventions.The size of the node in the network plot represents the sample size of the group, and the edge width represents the number of studies.Third, NMA was performed using R 4.1.2software gemtc packages in RStudio, based on a Bayesian framework with Markov Chain Monte Carlo (MCMC) simulation.We ran the estimation with a burn-in of 25,000 iterations and sampling of 50,000 iterations from the three chains of initial values.The selection between models was based on the Deviance Information Criteria (DIC).A DIC difference in the consistency test results >5 was considered significant.The fluctuation process of the MCMC chain is represented by a trace plot and the convergence degree of the model is diagnosed together with the density and Brooks-Gelman-Rubin diagnosis plots with the potential scale reduction factor (PSRF).If the degree of model convergence is poor or 1 < PSRF ≤ 1.05, the frequency of pre-iteration and iterations need to be changed.
Fourth, we used R 4.1.2software to calculate the surface under the cumulative ranking (SUCRA) to rank the interventions.A heatmap was used to visualize the SUCRA results.Results of twotailed tests with P <0.05 were considered statistically significant.

Literature search
We searched five databases for 2,622 studies (Databases: 2,204; Registers: 418).After a review by two investigators, 15 RCTs were included in the systematic review and NMA.The search details are represented in a flow diagram in Figure 1.

Study characteristics
The characteristics of the 15 included RCTs, including the study and publication year, participants, and intervention characteristics, are summarized in Table 1.All the studies were published after 2000 and the sample size for each group was at least 20.In addition, the preoxygenation maintenance time was 3-5 min.

Risk of bias assessment and study confidence rating
The risk of bias assessment is shown in Figure 2. All the studies reported random sequence generation and allocation concealment.Blinding was defined as impossible in the included studies; therefore, performance bias for 14 RCTs was defined as high risk, and outcome assessments were defined as unclear owing to a lack of reporting.However, blinding was achievable in a study by Jaber et al. (20); hence, the risk of bias for blinding was defined as low.

Network plot of eligible comparisons of outcomes
Network maps of the outcomes are shown in Figure 3.The outcomes include SpO 2 <80% during ETI (Figure 3A), SpO 2 <90% during ETI (Figure 3B), and the lowest SpO 2 during ETI (Figure 3C).
The NMA results (Figure 4A) showed that the NIV group was superior to the COT group (Log OR 1.23; 95% CI: 0.31, 2.42).For the SUCRA results shown in Table 3, the NIV group (0.8603467) performed better than the HFNC (0.1373533) and COT (0.0023) groups.
The NMA results (Figure 4B) showed that the HFNC group was superior to the COT group (Log OR 0.74; 95% CI: 0.01, 1.91).Regarding the SUCRA results, as shown in Table 3, the NIV group (0.60932667) performed better than the HFNC (0.37888667) and COT (0.01178667) groups.

Discussion
This study marks the first comparison of multiple preoxygenation techniques using NMA.The findings revealed that NIV coupled with HFNC effectively maintained high SpO 2 levels during ETI.Both NIV and HFNC significantly reduced the risk of SpO 2 dipping to <80% and <90%, respectively, with NIV demonstrating superiority over both HFNC and COT.Furthermore, the use of HFNC for preoxygenation extended the duration of apnea.
. /fmed. .Preoxygenation is a crucial measure for safeguarding patient oxygenation throughout the intubation process and ensuring surgical safety, particularly in individuals anticipated to encounter airway challenges.Both the 2015 Difficult Airway Society guidelines and 2022 American Society of Anesthesiologists recommendations emphasize the significance of this practice (6,27).Among them, the 2022 guidelines recommend 3 min of preoxygenation to reach an end-tidal oxygen concentration of 0.90 or higher (EtO 2 ≥ 0.9) and use of various NIV devices, such as nasal catheters and masks.As a modification of conventional nasal catheters, HFNC can deliver high-flow oxygen, generate low levels of positive end-expiratory pressure (PEEP), and allow asphyxiation and oxygenation, making it useful for preoxygenation or oxygen therapy (28).Despite the obvious benefits of HFNC in patients with acute hypoxic respiratory failure and after scheduled extubation (10, 23), its effectiveness is still controversial compared with other preoxygenation methods (29,30).Guitton et al. (19) reported that HFNC preoxygenation reduced tracheal intubation-related adverse events but did not improve lowest SpO 2 .In addition, Vourc'h et al. (23) compared preoxygenation methods in obese patients and found that the HFNC group had a significant advantage in the lowest SpO 2 compared with the NIV group.We included the study by Jaber et al. (20) on NIV combined with HFNC for preintubation oxygenation in critically ill patients with severe hypoxemia and acute respiratory failure.The results of our study indicate that when used in combination with preoxygenation, NIV and HFNC were associated with the lowest SpO 2 levels.Paradoxically, our findings suggest that the combined use of NIV and HFNC could potentially improve the lowest SpO 2 recorded.However, because of the scarcity of relevant studies, our analysis was limited to only one study, thus precluding a direct comparison.Consequently, further studies are required to comprehensively evaluate the effects of this combined approach on preoxygenation.In addition, owing to the small sample size, treatment effects and publication bias need to be considered; therefore, caution must be exercised when interpreting the results of HFNC and NIV studies.Additionally, the best method for combining HFNC and NIV to reduce air leakage has not been well described.Whether continuous nasal positive airway pressure masks play a special role in preoxygenation is worth exploring.
When comparing SpO 2 <90% and SpO 2 <80%, we found no significant advantage of HFNC over NIV, although both were superior to COT.A multicenter, randomized, open-label trial reported the effect of preoxygenation in patients with acute respiratory failure, showing that neither NIV nor HFNC changed severe hypoxemia in these patients, with no significant difference between them (18).A study designed by Kuo et al. pointed out that HFNC can better enhance PaO 2 and has an advantage over mask oxygenation in preventing ETI (15).The reasons for this difference in outcomes may be patient characteristics and oxygen flow settings.Some studies have pointed out that NIV may be a better method of preoxygenation for obese patients (31, 32), and HFNC preoxygenation patients have lower EtO 2 and SpO 2 levels than the NIV group.This result may be related to the limited supraglottic pressure produced by HFNC, the inability to repair airway obstruction after general anesthesia, and the difficulty in maintaining or restoring FRC damage in obese patients (33).However, considering that HFNC is better tolerated and has a median SpO 2 , it is an acceptable alternative for obese patients without NIV or with contraindications (24, 34).In addition, several studies have reported the comparison of preoxygenation methods in critically ill patients with hypoxemia or acute respiratory failure (11,12,16,23).In 2019, Fong et al. (35) conducted an NMA of RCTs on the effects of preoxygenation in patients with acute hypoxic respiratory failure.Seven RCTs encompassing 959 patients were comprehensively analyzed in this study.These findings indicate that NIV is a safe and potentially the most effective preoxygenation technique.One explanation for the inferior performance of HFNC could be the loss of the PEEP effect in patients experiencing respiratory distress due to mouth opening (36).In these patients, the nasal and oral inhalation flows can be as high as 110 and 280 L/min, respectively, which are significantly better than those of HFNC (37).Another possible explanation is that NIV allows the delivery of high levels of FiO 2 and positive intrathoracic pressure, promoting alveolar replenishment, which may improve the efficiency of gas exchange (38,39).
This study has some limitations.Constrained by the paucity of RCTs reporting pertinent comparisons, only 15 studies were included in this analysis, potentially introducing a reporting bias.Furthermore, the combined use of NIV and HFNC for preoxygenation was reported in a single study.Although the initial results appear promising, the absence of a direct comparative evidence necessitates cautious interpretation of the conclusions.

Conclusions
The results of the network analysis showed that NIV for preoxygenation achieved a higher SpO 2 than HFNC and COT and had a more significant benefit in maintaining patient oxygenation during ETI.Patients had a longer apnea time after HFNC preoxygenation.The effect of NIV combined with HFNC was significantly better than that of other methods.Owing to a lack of ./fmed. .studies, further investigation is warranted to evaluate their effects in the future.

FIGURE
FIGUREPRISMA flow diagram of the search strategy and included studies.

FIGURE
FIGURERisk of bias summary review authors' judgements about each risk of bias item for included RCTs.
et al. (15) reported that high-flow nasal oxygenation can enhance PaO 2 and prolong safe apnea time.Li surface under the cumulative ranking curve; RCTs, Randomized Controlled Trials; MCMC, Markov Chain Monte Carlo; DIC, Deviance Information Criteria; PSRF, potential scale reduction factor; PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-analysis; EtO 2 , end-tidal oxygen concentration; PEEP, Positive End-expiratory Pressure; THRIVE, Transnasal Humidified Rapid-Insu ation Ventilatory Exchange.
N = 53 Inclusion criteria: Adults patients (age ≥ 21) requiring RSI due to any condition.Exclusion criteria: active "do-not-resuscitate" orders; crash, awake or delayed sequence intubations; requiring non-invasive positive pressure ventilation; cardiac arrest; suspicion or confirmed diagnosis of base of skull fractures or severe facial trauma that precluded placement of NC; pregnant women; and those incarcerated.≥3min of preoxygenation with usual care by preoxygenating using only non-rebreather mask at flush rate, and then given at least 15L/min of non-humidified and non-heated oxygen from wall supply via NC for apneic oxygenation.≥3min of preoxygenation with HFNC received 60L/min of warm and humidified oxygen at 37 • C and FiO 2 more than 0.90.NR Frat et al. (18) N = 53 Inclusion criteria: Adults patients (age > 18) requiring intubation in the ICU with acute hypoxemic respiratory failure (RR > 25 bpm or signs of respiratory distress, PaO 2 /FiO 2 < 300 mmHg regardless of oxygenation strategy).Exclusion criteria: Cardiac arrest, altered consciousness (GCS <8).NR Nong et al. (22) N = 106 Inclusion criteria: Adults patients (age > 18) requiring intubation in the ICU.Attending physician based on worsening respiratory failure (e.g., blood oxygen saturation (SpO 2 ) < 88% and RR > 36/min) after adequate therapy, fraction of inspired oxygen (FiO 2 ) > 60%, intolerance to NIV, neurological deterioration, or copious respiratory secretions.Exclusion criteria: age < 18 years, pregnancy, severe coagulopathy, cardiac arrest, and contraindications for bag-valve-mask or NIV preoxygenation.≥3 min of preoxygenation with bag-valve-mask driven by 15 L/min oxygen flow, an oxygen reservoir was added to the balloon, and positive end-expiratory pressure was set at 5 cmH 2 O. ≥3 min of preoxygenation with NIV support with the following settings: mode, S/T; f, 20/min; inspiratory positive airway pressure, 12-20 cmH 2 O (adjusted to obtain an expired tidal volume of 7-10 ml/kg); expiratory positive airway pressure, 5 cmH 2 O; and FiO 2 , 100%.
TABLE The direct evidence from pairwise meta-analysis and heterogeneity of outcomes.Odd Ratio; CI, Confidence Interval; NE, Not Estimate; NIV, Noninvasive Ventilation; COT, conventional oxygen therapy; HFNC, High Flow Nasal Cannula; ETI, Endotracheal Intubation.
TABLE The SUCRA results of network analysis outcomes.Intervention SpO 2 < % during ETI procedure SpO 2 < % during ETI procedure Lowest SpO 2 during ETI procedure