High-flow nasal cannula for hypoxia in the post-anesthetic recovery unit: A systematic review and meta-analysis

Abstract

BACKGROUND

Earlier meta-analyses demonstrated conflicting results on whether high-flow nasal cannula (HFNC) reduces rate of intubation in the postoperative period when compared to conventional oxygen therapy (COT). Since then, newer studies have come out, so an updated comparison of these two modalities is warranted.

AIM

To determine whether HFNC reduces the rate of reintubation compared to COT.

METHODS

We conducted a random-effects meta-analysis, searching databases (MEDLINE, EMBASE, Web of Science) to identify randomized controlled trials comparing HFNC with COT on postoperative patient outcomes of interest including reintubation, escalation to noninvasive ventilation, mortality, length of stay, pneumonia, postoperative pulmonary complications, vital signs and blood gas measurements. We assessed the quality of data using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework.

RESULTS

Twenty-five trials, including 4260 patients met the inclusion criteria. Thirteen investigated reintubation and the pooled result was not significant [odds ratio (OR); 0.93; 95% confidence interval (CI): 0.36-2.38; I² = 47.1%; P = 0.86]. However, HFNC reduced the incidence of atelectasis (OR 0.33; 95%CI: 0.18-0.59; I² = 10.8%; P = 0.0046) and hypoxemia (OR 0.42; 95%CI: 0.21-0.83; I² = 78.3%; P = 0.017). There were no clinical differences detected in other outcomes. GRADE for certainty of evidence was low to moderate.

CONCLUSION

HFNC does not reduce reintubation risk in our analysis with moderate certainty. Use of HFNC should not delay escalation of therapy to noninvasive ventilation or a definitive airway.

Key Words: High-flow nasal cannula; Conventional oxygen therapy; Meta-analysis; General anesthesia; Extubation; Reintubation; Postoperative respiratory adverse events

Core Tip: High-flow nasal cannula has beneficial effect on prevention of atelectasis and hypoxemia compared to conventional oxygen therapy. However, it does not reduce the risk of clinical outcomes including reintubation, therapy escalation, pneumonia, postoperative pulmonary complication, mortality, or prolonged hospital or intensive care unit stay.

INTRODUCTION

Postoperative respiratory adverse events are common complications after major surgery. Common causes of immediate postoperative respiratory adverse events include altered respiratory mechanics from surgery, respiratory depression by drugs and insufficient respiratory effort due to pain. These issues often lead to atelectasis, impaired gas exchange and hypoxemia which may necessitate oxygen therapy ranging from conventional oxygen therapy (COT) such as face mask (FM) and nasal cannula (NC), high-flow NC (HFNC) or non-invasive ventilation (NIV) consisting of bi-level or continuous positive airway pressure (BiPAP or CPAP).

Over the past decade, HFNC has emerged as a promising non-invasive alternative for respiratory support. HFNC delivers heated and humidified oxygen-air mixtures between 20-60 L/minute and precise inspired oxygen (FiO₂) from 21% to 100%. The advantage of HFNC is based on its ability to: (1) Wash out anatomic dead space; (2) Generate a low-level positive end-expiratory pressure (PEEP); (3) Meet patient’s peak inspiratory flow; and (4) Provide comfort with heat and humidity. COT cannot provide precise FiO₂ and at higher flow becomes uncomfortable due to cold gas flow[1]. Alternatively, NIV requires a tight mask fit and causes aerophagia, bloating, and potential worsening of airway obstruction which can lead to skin break down after prolonged use.

Previous systematic review and meta-analyses have examined the use of HFNC in the perioperative period with conflicting results. Chaudhuri et al[2] in their systematic review and meta-analysis (SRMA) of 11 randomized control trials (RCT) concluded that HFNC was associated with a lower reintubation rate when compared to COT, in the postoperative period. A more recent network meta-analysis by Pettenuzzo et al[3] found, when compared to COT, NIV but not HFNC was associated with a lower rate of reintubation. Given the apparent conflicting findings and availability of more recent RCTs, we decided to conduct a SRMA to determine whether HFNC can reduce the risk of re-intubation when compared to COT in the postoperative period.

MATERIALS AND METHODS

This SRMA complies with the PRISMA statement and is registered on PROSPERO (CRD42024552801). The clinical question, study criteria, outcomes and analysis were defined a priori.

Literature search

We searched Ovid MEDLINE, EMBASE and Web of Science from inception to October 3, 2025. Only English studies were included. The full search strategy is available in Supplementary material. The reference lists of included articles were manually searched for additional studies. The included studies met the following criteria.

Population: Studies had to recruit adult patients who had surgery. Patients may have had any form of anesthesia or analgesia. Obstetric populations were excluded.

Intervention and control: Included RCTs compared HFNC to COT NC, simple facemask or Venturi facemask. Studies with an additional NIV comparator were included, but the NIV arm was not used in the analysis. Studies with multiple HFNC arms (i.e. different flow or different devices) were combined into a single group.

Outcomes: The primary outcome was all-cause reintubation at any time point. Secondary outcomes included escalation of therapy to NIV, all-cause mortality, pneumonia, atelectasis, postoperative pulmonary complication (PPC), incidence of hypoxemia, hospital and intensive care unit (ICU) length of stay (LOS), arterial pH, oxygen (PaO₂), carbon dioxide (PaCO₂), oxygen saturation (SpO₂), respiratory rate (RR), heart rate (HR), mean arterial blood pressure and comfort level.

Article selection and data extraction

All titles, abstracts and full texts (if required to assess for inclusion) were reviewed in duplicate and any disagreements resolved by a third reviewer. Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia. Available at www.covidence.org) was used for screening and data extraction. The median and interquartile range (IQR) were approximated into means and their corresponding SD based on recommendations by the Cochrane Handbook for Systematic Reviews of Interventions.

Risk of bias evaluation

The Cochrane Risk of Bias 2 tool was utilized to assess the studies’ risk of bias based on the 5 domains: (1) Randomization process; (2) Deviation from intended intervention; (3) Missing outcome data; (4) Measurement of the outcome; and (5) Selection of the reported result.

Level of evidence evaluation

The Grading of Recommendations Assessment, Development and Evaluation (GRADE) guideline was used to evaluate the certainty of evidence and level of recommendation for each outcome in the meta-analysis.

Statistical analysis

Statistical analysis was done using R (version 4.5.2, the R Foundation for Statistical Computing, Vienna). Standard summary measures were generated with the mean difference (MD) for continuous data and odds ratios (ORs) for categorical data with their corresponding 95% confidence intervals (CI) and an α of 0.05. All analyses were done using random-effects model. I² statistics was used to quantify heterogeneity: 0%-25% was deemed low, 25%-50% moderate and above 50% as high heterogeneity. For categorical outcomes with low incidence (i.e. < 5%) Peto method was used without correction for zero event. Otherwise, the Mantel-Haenszel method with continuity correction was used. Sensitivity analysis were conducted to determine if there were any influential studies.

A funnel plot was constructed for the primary outcome. Given the potential rarity of reintubation rate, Harbord’s test was used to investigate for statistical evidence of publication bias.

RESULTS

Literature search and study selection

Our search yielded 2337 studies, with 25 trials including 4260 patients ultimately meeting the inclusion criteria (Figure 1). The study characteristics were outlined in Table 1. Seven studies[4-10] were rated low risk of bias, five[11-15] with some concerns, and thirteen[16-28] with high risk of bias (Figure 2). Studies were rated higher risk of bias as it was difficult to blind the subjects, care providers and assessors given the interventions. Further, some studies[7,10,12,15,26] allowed clinician discretion for therapy escalation, rather than following protocols. Nine RCTs studied cardiac surgery patients[9,10,12,17,21,22,25-27], five abdominal[4,6,23,24] and five thoracic[11,15,16,18,28] surgeries, three on bariatric[5,8,14] patients, one on orthopedic patients[19] and three on mixed surgical populations[7,20,29]. Eight studies included all patients without contraindication to HFNC[11,13-17,21,25], while the remaining studies focused on specific groups: Two for elderly populations[4,19], three for hypoxic patients[20,26,27], five for obese patients[5,8,10,12,22], two for patients at risk of post operative PPC[7,19], and five for average-risk patients[6,9,18,23,24].

Figure 1  Preferred reporting items for systematic reviews and meta-analyses flowchart of study selection.

Figure 2  Risk of bias assessment figure for all studies.

Table 1 Characteristics of included studies.

Ref.	Country	N (C/I)	RoB	Surgery	Population	Intervention	Comparator	Primary outcome
Ansari et al[11]	Iran	31/28	SC	Thoracic	All comer	HFNC for 24 hours	NC or FM	6-minute walk at POD2
Brainard et al[16]	United States	26/18	High	Thoracic	Admitted to ICU	HFNC for 48 hours	NC or FM	Composite incidence of reintubation, hypoxemia, NIV
Burra et al[17]	India	30/30	High	Cardiac	All comer	HFNC for 4 hours	NC	pH, PaO2, PaCO2 at 1, 2, 4 hours
Chen et al[4]	China	99/98	Low	Laparoscopic	Age > 65	No description	NC	Incidence of hypoxemia until PACU discharge
Corley et al[12]	Australia	74/81	SC	Cardiac	BMI > 30	HFNC for 8 hours	NC or FM	Atelectasis on POD1
El-Nori et al[13]	Egypt	90/90	SC	Thoracic	All comer	HNFC for 48 hours	NC or FM	Incidence of hypoxemia in hospital
Ferrando et al[5]	Spain	74/81	Low	Bariatric	BMI > 35	HFNC for 3 hours	Venturi	Incidence of hypoxemia at 3 hours
Fogagnolo et al[18]	Italy	58/58	High	Thoracic	ASA 1, 2, BMI < 35	HFNC for 24 hours	NC or FM	Diaphragmatic dysfunction at 24 hours
Frassanito et al[6]	Italy	41/42	Low	Gynecologic	BMI < 35	HFNC for 2 hours	Venturi	PF ratio at 2 hours
Fulton et al[14]	United States	25/25	SC	Bariatric	All comer	HFNC for 6 hours	FM	End expiratory lung impedance at 1 hour
Futier et al[7]	France	112/108	Low	Abdominal, thoracic	At risk for PPC	HFNC until 8AM POD1	NC or FM	Hypoxemia at 1 hour
Li et al[19]	China	30/30	High	Orthopedic	Age > 65	HFNC for 1 hour	FM	pH, PaO2, PaCO2 at 1 hour
Mishra et al[20]	India	30/30	High	Non-thoracic	SaO2 < 90	HFNC for 2 hours	Venturi	PF ratio at 2 hours
Parke et al[21]	New Zealand	171/169	High	Cardiac	All comer	HFNC until 9AM POD2	NC or FM	SpO2/FiO2 > 445 at day 3
Pennisi et al[15]	Italy	48/47	SC	Thoracic	All comer	HFNC until 9AM POD2	Venturi	PF ratio < 300 during first 4 days
Rosén et al[8]	Sweden	15/19	Low	Bariatric	BMI > 35	HFNC for 1 hour	NC	PaO2 at 1 hour
Sahin et al[22]	Turkey	50/50	High	Cardiac	BMI > 30	No description	FM	Not clear
Shiho et al[9]	Japan	35/35	Low	Cardiac	No lung disease	HFNC for 1 hour	Venturi	PaO2 at 1 hour
Soliman et al[23]	Egypt	40/40	High	Abdominal	No lung disease, BMI < 35	HFNC for 6 hours	FM	Incidence of PPC during first 5 days
Sun et al[24]	China	39/39	High	Abdominal	ASA 1, 2, BMI < 30	HFNC while in PACU	NC	Atelectasis
Tatuishi et al[25]	Japan	76/72	High	Cardiac	All comer	HFNC for 18 hours	FM	Lung volume reduction
Theologou et al[26]	United Kingdom	33/66	High	Cardiac	PF ratio < 200 after SBT	HFNC until wean or failure	Venturi	Treatment failure
Vourc'h et al[27]	France	43/47	High	Cardiac	SpO2 < 96% on 50% venturi	HFNC for 48 hours	NRB	PF ratio at 1 and 24 hours
Yu et al[28]	China	54/56	High	Thoracic	At risk for PPC	No description	NC or FM	Incidence of hypoxemia during 72 hours
Zochios et al[10]	United Kingdom	45/49	Low	Cardiac	COPD, Asthma, BMI > 35	HFNC for 24 hours	NC or FM	Hospital LOS

C/I: Number of patients in control/number of interventions; RoB: Risk of bias; SC: Some concerns; HFNC: High flow nasal cannula; NC: Nasal cannula; FM: Face mask; POD: Postoperative day; PF: Arterial oxygen pressure to inspired oxygen fraction; SBT: Spontaneous breathing trial; SpO₂: Peripheral oxygen saturation; BMI: Body mass index; COPD: Chronic obstructive pulmonary disease; PPC: Postoperative pulmonary complication; PACU: Post-anesthesia care unit; NRB: Non-rebreather mask; LOS: Length of stay; ASA: American Society of Anesthesiologists; PaO₂: Partial pressure of arterial oxygen; PaCO₂: Partial pressure of arterial carbon dioxide; ICU: Intensive care unit; POD1: Postoperative day 1.

Of the 16 studies where the HFNC devices were documented, one used the Maxventuri[16] and the remaining used the Optiflow/Airvo[5,7,9,11,12,14,15,17,18,21-24,26-28]. The duration of HFNC treatment ranged from 1 hour to 48 hours and the flow rate was between 20-60 L/minute. Six studies compared HFNC with venturi mask[5,6,9,15,20,26], one with non-rebreather mask[27] and the rest with a simple nasal prong and/or FM. All except two studies[26,27] applied HFNC prophylactically rather than therapeutically.

Outcomes

Outcome matrix by study is available as Supplementary Table 1. Any outcomes with more than 3 studies were analyzed and presented in Supplementary Table 2. GRADE evaluation for certainty of evidence is summarized in Table 2.

Table 2 Summary of findings with certainty of evidence.

Outcomes	Anticipated absolute effects		Number of participants(studies)	Certainty of the evidence (GRADE)
	Risk with conventional oxygen therapy	Risk with high flow nasal cannula (95%CI)
Reintubation	2.8%	OR = 0.93 (0.36-2.38)	1693 (13)	Moderate
Therapy escalation	11.3%	OR = 0.64 (0.31-1.32)	1190 (8)	Moderate
Mortality	1.2%	OR = 0.71 (0.18-2.78)	1172 (8)	Moderate
Atelectasis	29.7%	OR = 0.33 (0.18-9.59)	628 (6)	Low
Hypoxemia	34.6%	OR = 0.42 (0.21-0.83)	1276 (11)	Moderate
Postoperative pulmonary complications	43.7%	OR = 0.68 (0.33-1.40)	645 (6)	Low
Hospital length of stay	Mean of 7.7 days with range 2 to 12 days	MD = 0.58 less (-1.17 to 0.01)	1877 (16)	Moderate
ICU length of stay	Mean of 2.6 days with range 1 to 7 days	MD = 0.02 less (-0.19 to 0.15)	1880 (14)	Moderate
pH	Mean of 7.37 with range 7.34 to 7.41	MD = 0.0002 more (-0.01 to 0.01)	423 (4)	Low
PaO2	Mean of 115.1 with range 87.0 to 158.5	MD = 43.0 more (-15.7 to 101.7)	767 (9)	Low
PaCO2	Mean of 39.9 with range 34.7 to 44.1	MD = 1.1 less (-2.6 to 0.4)	831 (10)	Low
PF ratio	Mean of 222.4 with range 46 to 355.5	MD = 33.4 more (-7.99 to 74.8)	447 (6)	Low
RR	Mean of 19 with range 14 to 22	MD = 1.4 less (-3.0 to 0.3)	343 (4)	Low
SpO2	Mean of 98.8 with range 96.8 to 99.7	MD = 0.84 more (-0.61 to 2.23)	344 (3)	Very low
HR	Mean of 85.6 with range 68 to 105.7	MD = 4.6 less (-8.6 to -0.6)	504 (5)	Low
Discomfort	Mean of 4.1 with range 1.5 to 7.3	MD = 0.7 less (-2.3 to 0.9)	1000 (6)	Low

PaO₂: Partial pressure of arterial oxygen; PaCO₂: Partial pressure of arterial carbon dioxide; ICU: Intensive care unit; POD1: Postoperative day 1; PF: PaO₂/FiO₂ ratio; RR: Respiratory rate; SpO₂: Oxygen saturation; HR: Heart rate; CI: Confidence interval; OR: Odds ratio; MD: Mean difference.

Reintubation, therapy escalation and mortality

Thirteen studies including 1693 patients reported this outcome[5,7,10,12,13,15,17,21-23,26-28]. Two studies had 0 reintubations[5,17]. HFNC did not reduce the risk of reintubation when compared to COT (OR 0.93; 95%CI: 0.36-2.38; I² = 47.1%). Rate of re-intubation was 3.1% (27/867) in HFNC and 2.8% (23/826) in COT (Figure 3). Eight studies with 592 patients recorded escalation to NIV. HFNC did not prevent therapy escalation to NIV (OR 0.64; 95%CI: 0.31-1.32; I² = 57.8%). The rate of therapy escalation in HFNC and COT was 8.5% (51/598) and 11.3% (67/592) respectively.

Figure 3 Analysis of reintubation rates with stratification by patient risk. OR: Odds ratio; CI: Confidence interval.

Eight studies with 1172 patients tracked mortality, with 3 studies[13,15,17] having zero mortality. HFNC did not reduce mortality compared to COT (OR 0.71; 95%CI: 0.18-2.78; I² = 0%). Incidence of mortality in HFNC and COT was 0.85% (5/585) and 1.2% (7/587) respectively.

Respiratory complications

Six studies with 628 patients examined the incidence of atelectasis utilizing either ultrasound, chest x-ray or computed tomography[5,13,18,23,24,28]. HFNC reduced the incidence of atelectasis compared to COT (OR 0.33; 95%CI: 0.18-0.59; I² = 10.8%). The rate of atelectasis was 15% (46/315) in HFNC and 30% (93/313) in COT.

Eleven studies recorded the incidence of hypoxemia, with varying thresholds of PF ratio, PaO₂ or SpO₂. HFNC decreased the incidence of hypoxemia (OR 0.42; 95%CI: 0.21-0.83; I² = 78.3%). The rate of hypoxemia was 24.6% (161/655) in HFNC and 34.6% (215/621) in COT.

Pneumonia was explored in six studies[7,13,22-24,28]. The incidence of pneumonia for HFNC and COT was 4.2% (16/383) and 7.0% (27/385) respectively. No difference was found between HFNC and COT (OR 0.58; 95%CI: 0.27-1.28; I² = 0).

Six studies documented the incidence of PPC[6,7,15,16,18,23]. The definition of PPC varied between studies but generally represented a range of postoperative respiratory adverse events including hypoxemia, bronchospasm, atelectasis, pleural effusion, and respiratory failure. HFNC did not protect against PPC (OR 0.68; 95%CI: 0.33-1.34; I² = 49%). The incidence of PPC in HFNC and COT was 35.9% (115/320) and 43.7% (142/325) respectively.

LOS

Sixteen studies[5,7,10,11,13-16,18,21-26,28] with 1877 patients reported hospital LOS. There was a trend (P = 0.053) towards reduced hospital LOS of 0.58 days (95%CI: -1.2 to 0.01; I² = 71%) when HFNC was used. ICU LOS was reported by 14 studies[7,10,12-14,16,17,19,21-23,25-27] with 1770 patients and no difference was identified (-0.02 days; 95%CI: -0.2 to 0.2; I² = 37.3%). Given the likely non-normal distribution of LOS data, we also converted median and IQR to mean and SD using Luo et al’s[30] and Wan et al’s[31] methods and hospital LOS remained shorter and ICU LOS was not statistically different.

Laboratory outcomes

Four studies[6,7,17,19] reported on pH (MD 0.0002, 95%CI: -0.013 to 0.013; I² = 13.3%). Nine studies[6-9,17,19,20,22,23] recorded PaO₂ (MD 43; 95%CI: -16 to 102; I² = 97.1%). Ten studies[5-9,17,19,20,22,23] recorded PaCO₂ (MD -1.1; 95%CI: -2.6 to 0.4; I² = 76.4%). Finally, 6 studies reported on PF ratio (MD 33.4; 95%CI: -8.0 to 74.8; I² = 82.0%). Overall, there was no difference detected in pH, PaO₂, PaCO₂ and PF ratio between HFNC and COT.

Vital signs

Analyses were done for RR, SpO₂ and HR. RR was recorded by 4 studies[6,9,22,27] with 343 patients (MD -1.3; 95%CI: -3.0 to 0.3; I² = 61.6%). SpO₂ was reported by 5 studies[5,17,19,20,22], but 2 studies[17,19] reported with IQR of 0 and could not be converted to SD so analysis was carried out with 3 studies (MD 0.84; 95%CI: -0.61 to 2.23; I² = 58.5%). HR was reduced by 4.6 (95%CI: -8.6 to -0.6; I² = 45.6%; P = 0.034). No clinical differences were detected between HFNC and COT in these outcomes.

Comfort

Six studies[6,7,19-22] sought comfort feedback from patients. We converted all data to a 0-10 scale, with high score indicating more discomfort. HFNC was found to be similarly comfortable as COT, with MD of -0.7 (95%CI: -2.3 to 0.9; I² = 92.6%).

Publication bias

The funnel plot appeared roughly symmetrical on visual inspection. Two studies were shown as outliers[22,28]. Harbord’s test was not significant for publication bias (P = 0.44). Taken together, there was no evidence of publication bias and heterogeneity was likely due to differences in trial designs.

Sensitivity, subgroup and post-hoc analysis

The Baujat plot identified the same two studies as outliers[22,28]. Careful examination revealed zero reintubation events in the HFNC group in both studies, leading to a very low estimate of ORs. Removing the two outliers reduced heterogeneity to 0 but did not change the outcome. HFNC did not reduce reintubation in subgroup analysis by risk of bias, high vs average patient risk factor nor cardiac vs thoracic surgery.

We performed post-hoc meta-regression using the therapy escalation protocol (limited NIV duration[11,13,21,22] vs liberal NIV use[7,10,12,15,26]) as the moderator. The coefficient with liberal NIV use was 1.75 (P = 0.055), leading the OR of reintubation towards neutrality, suggesting a trend towards the rescuing effect of NIV. The MD in PF ratio between COT and HFNC was also used as a moderator for reintubation rate though only 3 studies[5,23,27] provided both results and this moderator did not predict reintubation rate.

DISCUSSION

HFNC is thought of as an attractive oxygen therapy in the postoperative period. Besides its ability to provide humidified high FiO₂ the high flow rate matches patient’s peak inspiratory demand, reduces anatomical dead space and provides PEEP, reducing the work of breathing. HFNC helps offset the respiratory impairments due to anesthesia and surgery, namely atelectasis, and reduced respiratory drive and altered respiratory mechanics because of pain and sedation[1]. From our analysis, while HFNC reduced the radiographic incidence of atelectasis and hypoxemia, it did not change important clinical outcomes such as reintubation, therapy escalation, PPC, pneumonia, mortality and LOS. It also did not alter vital signs or any blood gas findings.

HFNC, when compared to COT, did not reduce reintubation rate in our analysis. We postulated that permitting NIV rescue may have saved patients from reintubation and that HFNC may be inadequate to rescue severe hypoxia or high-risk patients. We tested these hypotheses in post-hoc analysis and subgroup analysis. In studies where liberal NIV was permitted, the meta-regression moderator yields a positive coefficient (1.75; P = 0.055), therefore a more neutral OR (i.e. OR approximately 1) for reintubation, suggesting a trend towards a protective effect against reintubation by NIV. We also explored the effectiveness of HFNC vs COT by baseline risks and severity of hypoxia with subgroup analysis and meta-regression, respectively, but we did not detect any difference in these. As there were only two studies[24,27] that are at low baseline risk and three that reported both PF ratio and reintubation rate, this hypothesis remained to be elucidated with future studies.

The finding that HFNC does not reduce reintubation rate was echoed in several recent meta-analyses. Pettenuzzo et al[3] found, when compared to COT, NIV but not HFNC reduce risk of reintubation, nosocomial pneumonia and hospital LOS and mortality. Lockstone et al[32] studied ICU patients and found HFNC did not reduce reintubation risk in the surgical subgroup. However, earlier meta-analyses by Chaudhuri et al[2] and Zayed et al[33] demonstrated a protective benefit of HFNC on reintubation. Inclusion of studies that were published after their reviews may have changed this outcome.

There are several important limitations to our review. We had chosen Peto method to analyze our primary outcome given its rarity and the method’s handle of zero events. Although this method is prone to group imbalance, the final analysis had similar number of subjects with 867 and 826 in the HFNC and COT groups respectively. There are significant variations in the settings, threshold, timing and the duration of HFNC therapy in the included studies, making it difficult to recommend a universal strategy on HFNC use. Many of our results have moderate to high heterogeneity. We attempted sensitivity analysis and were able to reduce heterogeneity in our primary outcome of reintubation. Nonetheless, this did not change the outcome. Much of the heterogeneity arose because of differences in trial design. The type of surgery, patient risk profiles differed, and subgrouping these factors did not demonstrate a difference. The clinical and laboratory measurements such as vital signs and blood gas were measured at various time points. While we attempted to poole close time points together, this practice may have introduced further heterogeneity in the results. Our analysis agrees with several more recent publications in that HFNC did not prevent reintubation when compared to COT[3,32]. Given the rarity of reintubation events and the minor, if any, protective effect (i.e. OR approximately 1) of HFNC over COT, substantially more trials are needed to demonstrate an effect conclusively. Our trial sequential analysis showed that, assuming a modest protective effect with an OR of 0.8, more than 45000 pooled subjects are required to reach the information threshold. Additionally, the studies focused on high-risk surgical populations are limited to cardiac, thoracic or abdominal surgery. Trials that explore the effect of HFNC in other potentially high-risk surgeries, such as ENT, neurosurgery, vascular, or major urologic surgery may be of interest.

CONCLUSION

Several guidelines have recommended using COT and HFNC in low-risk surgical patients, and HFNC and NIV in high-risk patients[34]. In our analysis, HFNC did not reduce the risk of reintubation, therapy escalation, pneumonia, PPC, mortality and LOS. However, there are reasons beyond these to consider using HFNC. HFNC has been shown to have beneficial effect on prevention of atelectasis and hypoxemia. In settings where financial and resource constraints are not a limiting factor, HFNC can be more liberally applied, especially in high-risk patients, given its minimal risk profile. However, it should not delay prompt escalation to NIV or intubation to prevent further respiratory compromise.