Efficacy of high-flow nasal oxygenation against peri- and post-procedural hypoxemia in patients with obesity: a meta-analysis of randomized controlled trials

This meta-analysis aimed at investigating the efficacy of high-flow nasal oxygenation (HFNO) against hypoxemia in patients with obesity compared with conventional oxygenation therapy and non-invasive ventilation. Databases were searched from inception to August 2021. Studies involving peri- or post-procedural use of HFNO were included. The primary outcome was risk of hypoxemia, while the secondary outcomes included status of oxygenation and carbon dioxide elimination. Ten randomized controlled trials (RCTs) were included. We found that HFNO prolonged the safe apnea time at induction compared to control group [mean difference (MD) = 73.88 s, p = 0.0004; 2 RCTs] with no difference in risk of peri-procedural hypoxemia [relative risk (RR) = 0.91, p = 0.64; 4 RCTs], minimum SpO2 (MD = 0.09%, p = 0.95; 4 RCTs), PaO2 (MD = − 8.13 mmHg, p = 0.86; 3 RCTs), PaCO2 (MD = − 6.71%, p = 0.2; 2 RCTs), EtCO2 (MD = − 0.28 mmHg, p = 0.8; 4 RCTs) between the two groups. HFNO also did not improve postprocedural PaO2/FiO2 ratio (MD = 41.76, p = 0.58; 2 RCTs) and PaCO2 (MD = − 2.68 mmHg, p = 0.07; 2 RCTs). This meta-analysis demonstrated that the use of HFNO may be associated with a longer safe apnea time without beneficial impact on the risk of hypoxemia, oxygenation, and CO2 elimination in patients with obesity. The limited number of trials warranted further large-scale studies to support our findings.

www.nature.com/scientificreports/ risks of difficulty in mask ventilation 9 and tracheal intubation 10 compared with those in individuals without. Besides, obesity is associated with a reduced functional residual capacity (FRC), atelectasis, and significant shunting in dependent lung regions with an increase in resting metabolic rate, work of breathing, and minute oxygen demand 11 . Therefore, patients with obesity are at high risk of a rapid drop in arterial oxygen level after the cessation of breathing. Moreover, because coexisting cardiovascular diseases are common in patients with obesity 12 , hypoxemia-induced cardiovascular complications (e.g., myocardial depression) following sedation remain a major concern 13 . Consequently, effective oxygen supplementation is crucial to the prevention of periand post-procedural pulmonary and cardiovascular complications in this patient population. High-flow nasal oxygenation (HFNO) refers to the delivery of oxygen at high flow rates without recourse to invasive or non-invasive ventilation. In the critical care setting, pooled evidence has demonstrated the clinical benefits of applying HFNO in patients with acute respiratory failure or those at high risk of post-extubation respiratory failure [14][15][16] . A recent meta-analysis recruiting mostly patients without obesity also reported the effectiveness of HFNO for prolonging the duration of safe apnea and elevating minimum SpO 2 as well as decreasing the risk of hypoxemia in patients receiving anesthetic induction or sedation 17 . Because upper airway obstruction due to posterior displacements of oropharyngeal structures (i.e., soft palate, base of tongue, and epiglottis) is a definite risk in patients with obesity after anesthesia induction or sedation 18 during which a patent airway remains a key factor for successful oxygenation 19 , the results of the previous meta-analysis 17 may not be applicable to those with obesity. To clarify the benefits of HFNO in this patient population, this meta-analysis aimed at comparing the risk of hypoxemia, oxygenation status, and carbon dioxide elimination between patients with obesity receiving HFNO and those undergoing conventional oxygen therapy (COT) or non-invasive ventilation (NIV) in a variety of clinical settings.

Methods
This meta-analysis was conducted in accordance with the recommendations of the PRISMA statement and registered with the International Prospective Register of Systematic Reviews (CRD42021271777).
Data sources and searches. We searched the Embase, Medline, Google scholar, and the Cochrane Library databases from inception to August 19, 2021, using the following search terms: ("Obesity" or "Obes*" or "Overweight" or "Severe Obesity" or "Morbid Obesity") and ("(high flow or high-flow) ADJ4 (oxygen or cannula* or oxygenation)" or "HFNO" or "HFNC" or "NHF" or "Optiflow" or "THRIVE" or "Transnasal Humidified Rapid Insufflation Ventilatory Exchange") limited to randomized controlled trials (RCTs). No restriction was placed on language, gender, sample size, and study location during literature search. The search strategies for these databases are demonstrated in Supplemental Table 1. Regarding Google scholar, a hand-search strategy was adopted to find the related articles. Once a relevant article was identified, a forward snowballing strategy 20,21 was used to optimize the efficiency of the literature search. Additional records identified by reviewing the reference lists of the retrieved studies were also reviewed for eligibility of being included in the current study.

Inclusion criteria.
To scrutinize the eligibility of the acquired publications for the present meta-analysis, we adopted the following PICO (population, intervention, comparison, outcomes) criteria: (a) Population: adults patients (age ≥ 18 years) with obesity, (b) Intervention: the use of HFNO as the intervention measure, (c) Comparison: the use of COT [e.g., mask/nasal oxygenation] or NIV as a control, (d) Outcomes: inclusion of at least one of these outcomes: incidence of hypoxemia, minimum O 2 saturation, PaO2, safe apnea time, PaCO 2 or EtCO 2 . Only RCTs were included for analysis. The authors of the included articles with missing information were contacted for possible access to the original data.
Exclusion criteria. Exclusion criteria were: (1) studies without a control group; (2) those focusing on patients undergoing cardiothoracic surgeries; (3) those in which information regarding outcomes was unavailable, and (4) RCTs published only as letters or abstracts, or (5) those presented as a review, case report, or other forms of publication other than original research. Study selection. Two authors examined the titles and abstracts of the retrieved RCTs independently for eligibility of being included in the present study. The full texts of the potentially eligible trials were independently reviewed based on the inclusion and exclusion criteria. Differences in opinions about inclusion or exclusion of a particular study were resolved through discussion with a third reviewer. Data extraction. The following items were retrieved from each trial: first author, year of publication, age, gender, body mass index (BMI), sample size, flow rate of HFNO, type of surgery or procedures, incidence of hypoxemia, level of PaO 2 , minimum O 2 saturation, safe apnea time, EtCO 2 , and PaCO 2 . Disagreements were settled by discussion with a third author.
Outcomes and definitions. The primary outcome was the impact of HFNO on the risk of hypoxemia, while the secondary outcomes included minimum SaO 2 , level of PaO 2 , PaO 2 /FiO 2 ratio, EtCO 2 , PaCO 2 , and safe apnea time. The definition of hypoxemia was in accordance with that of each study. As the efficacy for oxygenation may be different between COT and NIV, subgroup analysis of the impact of choosing either approach as control to assess the therapeutic benefit of HFNO was performed. If a study reported an outcome (e.g., level of PaO 2 ) at different time points, we analyzed the data acquired just before invasive mechanical ventilation. www.nature.com/scientificreports/ Assessment of risk of bias. Internal validity of the included RCTs was assessed by two reviewers independently based on the following domains: adequacy of sequence generation, allocation sequence concealment, blinding of participants and caregivers, blinding for outcome assessment, incomplete outcome data, selective outcome reporting, and the other sources of bias 22 . The risk of bias of each RCT was reported as "low," "unclear", or "high". We regarded the risk of "selective outcome reporting" bias of a study as "unclear" if its protocol was not published or registered. Moreover, the sources of funding were scrutinized for the potential of other biases. Disagreements were resolved by discussion.

Data synthesis and analysis. Cochrane Review Manager (RevMan 5.3; Copenhagen: The Nordic
Cochrane Centre, The Cochrane Collaboration, 2014) was used for the present meta-analysis. The pooled risk ratios (RRs) and mean difference (MD) with 95% confidence intervals (CIs) were computed for binary and continuous outcomes, respectively. We assessed heterogeneity with I 2 statistics and defined substantial heterogeneity as an I 2 over 50%. On the assumption of heterogeneity across the included studies, we adopted a priori a randomeffects model for outcome evaluation. The potential publication bias was assessed by visual inspection of a funnel plot on encountering 10 or more trials sharing a particular outcome. Sensitivity analysis was conducted with a leave-one-out approach to weigh the potential influence of the data from an individual trial on the overall outcome. The level of significance was set at < 0.05 for all outcome analyses.

Results
Search results and study characteristics. The process of study selection is shown in Fig. 1. First, of the 578 potentially relevant records retrieved from the databases, 184 duplicates were excluded. Second, screening of the titles and abstracts based on the PICO criteria gave 24 potentially eligible studies. Finally, after all text reviews, 10 studies published from 2014 to 2021 involving 564 patients were included in current meta-analysis.
Characteristics of the included studies are demonstrated in Table 1. All studies recruited participants of both genders with the proportion of females ranging from 29.7 to 89.5%. The mean BMI in the enrolled patients varied from 33 to 52 kg/m 2 . The proportion of patients with obstructive sleep apnea (OSA) was between 12.5 and 72.5% in five trials [23][24][25][26][27] , while five studies did not provide this information [28][29][30][31][32] . Of the ten studies, two assessed the beneficial effects of HFNO on postoperative pulmonary parameters in patients receiving laparoscopic bariatric surgery 23,24 , while eight studies evaluated the efficacy of HFNO against peri-procedural hypoxemia or oxygenation status during anesthesia induction (six trials) [26][27][28][29][30]32 , tracheal intubation in the intensive care unit (one trial) 31 , and colonoscopy under deep sedation (one trial) 25 . In the HFNO groups, the flow rate ranged from 50 to 120 L/min. In the control groups receiving COT/NIV, mask oxygenation was adopted in four studies 23,24,27,29 , while NIV and nasal cannula oxygenation were used in other four 26,30-32 and two 25,28 RCTs, respectively. bias, the risks of bias were low in most studies. However, one study did not give information about randomization of patients 30 , three RCTs did not described the methods used to accomplish allocation concealment [29][30][31] , and two trials reported that the allocation concealment was not masked 26,32 . Considering the impossibility of blinding among patients and caregivers in the included trials, performance bias was high in all studies [23][24][25][26][27][28][29][30][31][32] . Despite the lack of blinding, the risk of detection bias was considered low in all studies that used objective indicators (e.g., PaO 2 ) as assessment parameters. The risk of reporting bias was unclear in one study 30 that did not declare trial registration, while the risk of other biases was unclear or high in five studies 25,27,29,30,32 . Outcomes analyses. Primary outcome: impact of HFNO on risk of hypoxemia. Of the four RCTs included in the present meta-analysis, hypoxemia was defined as an SaO 2 < 90% in three studies 25,28,32 , and < 80% in one trial 31 which was conducted in the intensive care unit. The incidence of hypoxemia was 24.2% and 25.5% in the HFNO and control group, respectively. Pooled results revealed no significant difference in the risk of hypoxemia between patients receiving HFNO and those undergoing COT/NIV (RR = 0.91, 95% CI 0.63 to 1.33, p = 0.64; I 2 = 0%; 4 RCTs; n = 290) (Fig. 3). Consistently, subgroup analysis showed no significant impact of the choice of different approaches (i.e., COT or NIV) on the risk of hypoxemia (p = 0.92) (Fig. 3). Sensitivity analysis demonstrated a consistent finding when the four trials were removed one at a time.
Secondary outcome: impact of HFNO on peri-procedural oxygenation-related parameters. The safe apnea time was defined in two of the included studies either as the time taken for the SpO 2 to drop below 95% 27,28 or the maximum time of observation before invasive mechanical ventilation, which was six minutes in one study 27 and 15 min in the other 28 . Our results demonstrated that the use of HFNO significantly increased the safe apnea time compared to the use of COT (MD = 73.88 s, 95% CI 33.16-114.61, p = 0.0004; I 2 = 0%; 2 RCTs; n = 80) (Fig. 4a) 27,28 . However, forest plot showed no significant difference in PaO 2 (MD = − 8.13 mmHg, 95% CI − 97.93 to 81.68, p = 0.86; I 2 = 94%; 3 RCTs; n = 120) between the two groups ( Fig. 4b) 26,29,30 . Sensitivity analysis demonstrated a consistent finding when one trial was removed one at time. In addition, forest plot also revealed no significant difference in minimum SpO 2 (MD = 0.09%, 95% CI − 2.82 to 3.01, p = 0.95; I 2 = 73%; 4 RCTs; n = 290) between the two groups ( Fig. 4c) 25,27,31,32 . Sensitivity analysis indicated that the minimum SpO 2 was lower in the HFNO group compared to that in the control group when one study 27 was removed. Subgroup analysis showed no significant impact of the choice of different approaches (i.e., COT or NIV) on PaO2 (p = 0.08) (Fig. 4b). For minimum SpO 2 , subgroup analysis revealed a significantly lower minimum SpO 2 associated with the use of HFNO compared to that with NIV (MD = − 1.88%, 95% CI − 3.26 to − 0.51, p = 0.007; I 2 = 0%; 2 RCTs; n = 191), while there was no difference between HFNO and COT (p = 0.15). www.nature.com/scientificreports/ ( Fig. 5a) [26][27][28]32 between the HFNO and control groups. Sensitivity analysis verified a consistent finding when one trial was removed one at time. Forest plot showed no significant difference in the PaCO 2 (MD = − 6.71%, 95% CI − 16.99 to 3.58, p = 0.2; I 2 = 96%; 2 RCTs; n = 98) (Fig. 5b) 26,30 between the HFNO and control groups. The limited availability of trials (i.e., only two) precluded the conduction of a sensitivity analysis of this outcome. Subgroup analysis indicated no significant impact of choosing different approaches of conventional oxygenation (i.e., COT or NIV) on EtCO2 (p = 0.16) (Fig. 5a).

Discussion
Despite oxygen supplementation, patients with obesity may still experience significant hypoxemia after anesthesia-induced apnea because of a reduced FRC and an increased minute oxygen demand 11 that highlight the importance of implementing appropriate postprocedural oxygenation strategy for improving patient safety. Accordingly, the current meta-analysis focused on a comparison between HFNO and COT/NIV in this particular patient population. Our meta-analysis demonstrated that the use of HFNO prolonged the apnea time without a beneficial impact on the risk of hypoxemia, minimum SpO 2 , PaO 2 , EtCO 2 , and PaCO 2 in patients with obesity receiving peri-procedural oxygenation. After tracheal extubation, the application of HFNO was also not associated with an elevated PaO 2 /FiO 2 ratio and a decreased PaCO 2 level at postoperative three hours. Although a previous meta-analysis showed that the use of HFNO could reduce the risk of hypoxemia in patients receiving sedation or anesthetic induction 17 , only one of the included trials recruited patients with obesity; therefore, its findings may not be applicable to patients with obesity. Indeed, our results did not support a superior beneficial effect of HFNO against hypoxemia compared to that in the control group. Consistently, the levels of PaO 2 and minimum SpO 2 were comparable between the two groups, indicating no significant association between the use of HFNO and a reduced risk of hypoxemia. Therefore, one of the striking clinical implications of the present study was that this patient population, who are at risk of hypoxemia, may not benefit from the use of HFNO. In concert with our finding, a closed claims analysis on the management of difficult tracheal intubation showed that a delay in alternative airway intervention and judgment errors may contribute to brain ischemia and mortality 33 . Besides, the use of HFNO may be associated with an elevated risk of delayed airway management (e.g., tracheal intubation) 34 possibly because of a false sense of security that loosens the alert for potential airway problems. In this regard, we suggest that hypoxemia in patients with obesity receiving oxygen supplementation with HFNO should be promptly managed without exposing the patients to unnecessary risks.
A recent international multicenter trial comparing HFNO with standard facemask pre-oxygenation for rapid sequence induction in patients with a normal body build (i.e., mean BMI around 25 kg/m 2 ) 35 demonstrated no www.nature.com/scientificreports/ difference in the incidence of hypoxemia (i.e., SpO 2 < 93%) between pre-oxygenation using HFNO or tight facemask. Our findings were consistent with those in that study 35 . The lack of efficacy of HFNO for the prevention of hypoxemia compared to COT/NIV may be attributable to inadequate positive airway pressure associated with HFNO. First, although a previous study suggested the need for an adequate airway patency to achieve effective oxygenation 32 , the limited positive airway pressure generated by HFNO (e.g., 2.7 cmH 2 O) may be unable to relieve airway obstruction after anesthesia or sedation in patients with obesity 36 . Moreover, although a previous study has demonstrated a positive correlation between the flow rate of HFNO and nasopharyngeal pressure, which could reach over 3 cmH 2 O at a flow rate of 50 L/min 37 , whether a higher flow could improve the risk of hypoxemia in patients with obesity remains unclear. In the current study, there were four trials that provided the outcome of hypoxemia. While three of the trials 25,31,32 used a flow rate of 60 L/min, the other 28 adopted a flow www.nature.com/scientificreports/ rate of 120 L/min. Despite the obvious difference, our sensitivity analysis demonstrated that removal of the study using a higher flow rate 28 had no significant impact on the risk of hypoxemia. Nevertheless, since the number of trials included in the current meta-analysis was relatively small to arrive at a robust conclusion. Second, maintenance of an adequate FRC and avoidance of alveolar collapse is also important for efficient oxygenation 32 .
Although a previous small-scale study with 20 participants reported an increased lung volume and FRC as another potential benefit of HFNO particularly in patients with higher BMIs 38 , that study included only two patients with BMI > 40 kg/m 2 . In contrast, all of our included studies focused on patients with BMI > 30 kg/m 2 . Therefore, our findings implicated that the low-level positive airway pressure generated by HFNO may not be able to increase the lung volume in our patient population. Apart from the lack of a beneficial influence of HFNO on hypoxemia, the present study also showed no positive impact of HFNO on peri-procedural CO 2 clearance. Although a previous study demonstrated that the enhanced CO 2 clearance associated with the use of HFNO may be flow-dependent 39 , our results (flow: 50-120 L/ min) and those of a recent study (flow: 70 L/min) 35 did not support this finding that the use of HFNO was associated with a low CO 2 clearance. Regarding the impact of HFNO on postprocedural CO 2 clearance, our results were derived from two trials that recruited patients undergoing laparoscopic operations in which CO2 was used for abdominal CO 2 insufflation. Although CO 2 clearance may be modified by anesthesiologists immediately after laparoscopic surgery, the present study focused on postoperative three hours so that such an impact would be minimal. Nevertheless, our findings may not be extrapolated to patients receiving non-laparoscopic procedures. Further studies are needed to address this issue.
One recent meta-analysis of three clinical trials enrolling 160 patients with or without obesity reported a safe extension of apnea time by 33.4 s in participants receiving HFNO versus those subjected to COT at anesthesia induction 17 . In spite of the demonstration of a HFNO-associated prolongation of safe apnea time compared to COT in the current study, the finding should be interpreted with caution. First, in spite of our finding of a significant prolongation of safe apnea time, the result was based on two trials enrolling only 80 patients 27,28 . Second, a prolongation of merely 73 s may not be of clinical significance in patients with obesity who usually present with a difficult airway 9,10 . Third, notwithstanding the inclusion of patients with similar BMI and age, there were wide variations between the two studies 27,28 in mean apnea time in both the HFNO (i.e., 581.3 28 vs. 261.4 27 s) and   35 that reported comparable incidences of hypoxemia between patients with or without HFNO despite a prolonged safe apnea in those receiving HFNO 35 .
In patients undergoing bariatric surgery, the prevalence of postoperative atelectasis could be as high as 37% 40 . Not only does obesity predispose to postoperative atelectasis but atelectasis in this patient population also resolves more slowly than in those with normal body build 41,42 . Despite the recommendation of oxygen supplementation in patients at high risk of postoperative atelectasis 43 , we found that the use of HFNO was unable to improve oxygenation parameters in patients with obesity both during the periprocedural period and at three hours after surgery. Consistent with our findings, another meta-analysis investigating patients with obesity undergoing cardiac surgery demonstrated no significant improvements in atelectasis score, dyspnea score, PaO 2 /FiO 2 ratio, and reintubation rate in patients receiving HFNO compared with those undergoing COT 44 . The lack of benefits of HFNO in patients with obesity during the postoperative period underscores the need for timely intervention (e.g., reintubation) in case of respiratory distress after tracheal extubation.
Our study had several limitations. First, the limited number of trials not only blemished the reliability of our findings but also precluded our analysis of some clinical outcomes such as the risk of atelectasis or hypoxemiaassociated complications. Second, the heterogeneity across the included trials in study settings (e.g., anesthetic induction vs. sedation, intensive care unit vs. operating theater) and approaches (i.e., COT or NIV) may introduce bias to our results. Nevertheless, a comparison between HFNO and other widely accepted clinical approaches to oxygen supplementation (e.g., COT and NIV) is still of clinical significance. Third, our findings on the association between HFNO and conventional respiratory parameters may not reflect clinical outcomes. Fourth, all included studies in the current meta-analysis investigated patients with BMI > 30 kg/m 2 , the efficacy of HFNO against hypoxemia in patients with less severe obesity remains to be elucidated.

Conclusion
The current study showed that, compared with conventional oxygen therapy or non-invasive ventilation, the use of high-flow nasal oxygenation was unable to provide additional peri-or post-procedural respiratory benefits for patients with obesity (i.e., BMI > 30 kg/m 2 ) except for a prolongation of safe apnea time. Further large-scale studies are warranted to validate our findings.