High-flow Nasal Cannula versus Conventional Ventilation in Laryngeal Surgery: A Systematic Review and Meta-analysis

High-flow nasal cannula (HFNC) is an emerging option for maintaining oxygenation in patients undergoing laryngeal surgery, as an alternative to traditional tracheal ventilation and jet ventilation (JV). However, the data on its safety and efficacy is sparse. This study aims to aggregate the current data and compares the use of HFNC with tracheal intubation and jet ventilation in adult patients undergoing laryngeal surgery. We searched PubMed, MEDLINE (Medical Literature Analysis and Retrieval System Online, or MEDLARS Online), Embase (Excerpta Medica Database), Google Scholar, Cochrane Library, and Web of Science. Both observational studies and prospective comparative studies were included. Risk of bias was appraised with the Cochrane Collaboration Risk of Bias in Non-Randomized Studies - of Interventions (ROBINS-I) or RoB2 tools and the Joanna Briggs Institute (JBI) Critical Appraisal Checklist for case series. Data were extracted and tabulated as a systematic review. Summary statistics were performed. Meta-analyses and trial sequential analyses of the comparative studies were performed. Forty-three studies (14 HFNC, 22 JV, and seven comparative studies) with 8064 patients were included. In the meta-analysis of comparative studies, the duration of surgery was significantly reduced in the THRIVE (Transnasal Humidified Rapid-Insufflation Ventilatory Exchange) group, but the number of desaturations, need for rescue intervention, and peak end-tidal CO2 were significantly increased compared to the conventional ventilation group. The evidence was of moderate certainty and there was no evidence of publication bias. In conclusion, HFNC may be as effective as tracheal intubation in oxygenation during laryngeal surgery in selected adult patients and reduces the duration of surgery but conventional ventilation with tracheal intubation may be safer. The safety of JV was comparable to HFNC.


Introduction And Background
Trans-nasal humidified rapid insufflation ventilatory exchange (THRIVE) delivers warm and humidified oxygen through high-flow nasal cannula (HFNC) at a high flow rate of up to 120L/min to patients who are both spontaneously ventilating or apneic. This has been used in both adult and neonatal intensive care units for respiratory distress syndrome [1][2][3], as an alternative non-invasive ventilatory option for patients with congestive heart failure and pulmonary oedema [2,3], in the peri-operative setting for pre-oxygenation [4], intra-operatively in the setting of difficult intubation [5], in post-extubation [6,7], and in sedation for endoscopies [8][9][10].
There are many modes of conventional ventilatory options during laryngopharyngeal surgery, including mechanical ventilation (MV) with tracheal intubation (TI), superimposed high-frequency jet ventilation (SHFJV) through the supraglottic, infraglottic, and transtracheal routes. Tracheal intubation has traditionally been performed for airway surgery, and intermittent intubation and cessation of oxygenation and ventilation can increase the risk of hypoxemia and carbon dioxide (CO 2 ) accumulation with progressive respiratory acidosis. Furthermore, intubation can obstruct the surgical view, hinder surgical access, and prolong surgical duration. There have been many studies on the use of jet ventilation for laryngeal surgery, most of them case series , which show complications including hypoxia, hypercapnia, need for intubation, and barotrauma. HFNC depends on the patient having a patent upper airway, which is provided by suspension laryngoscopy during laryngopharyngeal surgery. It avoids disturbing the surgical field, facilitates surgical access, and theoretically shortens procedure time.
Since Huang's review of the use of THRIVE in apneic patients undergoing shared airway surgery [5], there have been more recent studies comparing HFNC to conventional ventilation strategies such as tracheal intubation and SHFJV in the setting of laryngologic surgery [32][33][34][35], but there has not yet been a comprehensive comparative systematic review and meta-analysis of each method's efficacy. This study aims

Data items
Study characteristics included type of study, number of patients, intervention used, patient age, BMI, American Society of Anesthesiologists (ASA) grade, type of surgery, and neuromuscular blockade. Efficacy was defined by the following primary outcomes: duration of surgery, number of desaturations (defined as oxygen saturation (SpO2) <91% by the majority of the studies), and secondary outcomes including the need for rescue intervention (defined by conversion to intubation or jet ventilation), need for intubation, lowest SpO2, peak partial pressure of arterial carbon dioxide (PaCO2), peak venous partial pressure of CO2 (PvCO2), peak end-tidal carbon dioxide (EtCO2), peak transcutaneous CO 2 (or values at end of case), rate of EtCO2 increase, pH, bicarbonate, and other complications.

Data Collection Process
Two reviewers (KC and TY) screened the included studies independently, extracted the data and summarised it as a systematic review table. A third reviewer (KK) adjudicated if there were conflicts. Numerical data was extracted from diagrams, the appendices of the studies were checked, and authors were contacted for missing data. If a study did not explicitly discuss complications, we reported it as "not reported".

Synthesis Methods
We summarised effect estimates using descriptive statistics (median, interquartile range (IQR), and range) and applied a median effect estimate. For the comparative studies, the risk of bias table and meta-analysis were performed using the latest version of ReviewManager (RevMan) (5.3). Median and IQRs were converted to estimate mean and standard deviation using the method described by Wan et al. [57]. A random effects model (DerSimonian and Laird method) was utilised due to an anticipated high degree of heterogeneity. Pooled estimates for dichotomous outcomes were presented as risk ratios (RRs) with 95% confidence intervals (CIs) and for continuous outcomes, mean differences with standard deviations were used. Studies were assessed for heterogeneity using Cochran's Q test and I2 tests. Substantial heterogeneity was defined as I2 > 75%, The risk of publication bias was assessed by funnel plots. Robustness tests were planned to investigate heterogeneity. Low heterogeneity and few studies precluded sensitivity analysis. TSA was performed to determine the reliability of the meta-analysis, utilising TSA software version 0.9 (The Copenhagen Trial Unit, Centre for Clinical Intervention Research, The Capital Region, Copenhagen University Hospital -Rigshospitalet, Copenhagen, Denmark) [58].

Certainty Assessment
Confidence in the results obtained was assessed by two authors independently (KC and TY) using the Grading of Recommendations, Assessment, Development and Evaluations (GRADE) framework [59]. The quality of evidence was rated very low, low, moderate, and high. Disagreements were adjudicated by the third author (KK).

Characteristics of HFNC Studies
The characteristics of the 14 HFNC studies of 464 patients are summarised in

Outcomes of HFNC Studies
The outcomes of HFNC studies are summarised in Table 5. The duration of surgery was reported by two studies; 12.4 (4.4) minutes by Yang et al. [48] and 20.8 (7.8) minutes by Gustafsson et al. [56]. The duration of apnea was reported by 12 studies, the median was 21.7 minutes (IQR = 17-24.55, range = 10-30). The number of desaturations was reported by 12 studies and the median rate of desaturation was 8.1% (IQR= 4.55-16.6, range = 0-34.2). The total number of desaturations was 53 out of 448, and the need for rescue intervention was reported by 12 studies, with a median percentage for rescue intervention of 14.2% (IQR = 3.3-22.6, range= 2.9-100). The total number of rescue interventions was 60 out of 405. The need for intubation was reported by 11 studies, median percentage for intubation was 3.9% (IQR 1.9-19.6, range = 0-100). The total number of intubations was 38 out of 383 (mean = 9.9%). The lowest SpO2 or SpO2 after the end of apnea was reported by nine studies, median 91% (IQR = 76-95.25, range = 60-100). The peak PaCO2 or PaCO2 at the end of apnea was reported by three studies, median 10.2 kpa (range 9.44-11.86). The peak EtCO2 or EtCO2 at the end of apnea was reported by eight studies, median 7.4 (IQR = 7.025-7.9, range = 6.66-8.47). The peak transcutaenous CO2 was reported by two studies, ranging from (9.33-11.92). The rate of EtCO2 increase was reported by eight studies, the median was 0.15 kPa/min (IQR = 0.12-0.16, range = 0.03-0.17). pH at the end of apnea was reported by three studies, the median was 7.2 (range = 7.11-7.23). Bicarbonate at the end of apnea or the end of the case was reported by three studies, the median was 23.5 mmol/L(range 23.5-25.4).

Characteristics of Jet Ventilation Studies
The characteristics of the 21 jet ventilation studies of 6336 patients are summarised in

Outcomes of Jet Ventilation Studies
The outcomes of jet ventilation studies are summarised in Table 7

Characteristics of Comparative Studies
The characteristics of the comparative studies are summarised in Table 8

Outcomes of Comparative Studies
The outcomes of the comparative studies were summarised in Table 9 [32][33][34][35][54][55][56]. The duration of surgery was reported by five studies. Duration of apnea was reported by all studies, number of desaturations was reported by six studies, need for rescue intervention was reported by five studies, and need for intubation was reported by five studies. The lowest SpO2 or SpO2 at the end of apnea was reported by all studies. Peak ETCO2 was reported by three studies. Peak PaCO2, peak transcutaneous CO2, rate of EtCO2 or PaCO2 increase, pH at the end of apnea, and complications were reported by two studies. Bicarbonate at the end of the case was reported by one study. Only five of the comparative studies [33][34][35]57,58] could be included in the meta-analysis. We performed meta-analysis on the duration of surgery, number of desaturations, need for rescue intervention, lowest SpO2, and peak EtCO2. In the five studies included in the meta-analysis, standard ventilation referred to tracheal intubation and SHFJV, with only four patients in Nekhendzy et al.'s study undergoing SHFJV [34].    Three studies [33][34][35] including 236 patients reported the duration of surgery as an outcome. The duration of surgery was significantly decreased in the THRIVE group as show in Figure 2 (OR -4.92, 95%CI -7.73 to -2.11) (GRADE: Moderate). There was insignificant heterogeneity with an I^2 of 0%, Z-value (p = 0.0006). Four studies [32][33][34][35] including 266 patients reported the number of desaturations intraoperatively as an outcome. Although a rare event, the number of desaturations was significantly lower in the standard ventilation group as shown in Figure 3 (OR 6.58 95%CI 1.11 to 39.07). There was insignificant heterogeneity with an I^2 of 0%, Z-value (p = 0.04). Three studies [32][33][34] including 168 patients reported the need for rescue interventions intraoperatively as an outcome. The need for rescue intervention was significantly lower in the standard ventilation group as reported in Figure 4   Two studies [32,33] including 148 patients reported the peak EtCO2 as an outcome. The peak EtCO2 was significantly higher in the THRIVE group as shown in Figure 5 (OR 2.54, 95%CI 1.84 to 3.25) (GRADE: Moderate). There was moderate heterogeneity with an I^2 of 59%, Z-value (p < 0.00001). TSA is a methodology which weighs type I and II errors and regards the addition of each trial in the metaanalysis as an interim meta-analysis [41]. This helps to quantify the statistical reliability of data and assess the need for further trials [60]. In Figure 6, for the duration of surgery, the Z-curve (blue line) crossed the boundary for conventional benefit, the monitoring boundary, and exceeded the estimated information size, indicating there was sufficient statistical power.  In Figure 9, the O'Brien-fleming alpha-spending boundaries were not renderable as the first information fraction already exceeded 100% of the required information size, indicating that conventional ventilation significantly reduced the peak EtCO2 with sufficient statistical power.

Discussion
The use of HFNC in the intraoperative setting as an alternative to mechanical and jet ventilation has been gaining in popularity, especially for shared-airway procedures such as endoscopy, bronchoscopy, and laryngeal procedures [61][62][63][64]. While there are prior meta-analyses and systematic reviews of the use of HFNC in the induction period or for gastroscopy and bronchoscopy [61,62,64], this pre-registered study is the first that focuses on laryngeal surgery, comprehensively compares HFNC and conventional ventilation, systematically reviews jet ventilation in the setting of laryngeal surgery, and includes a trial sequential analysis. In the meta-analysis of comparative studies comparing THRIVE to standard ventilation, the duration of surgery was reduced in the THRIVE group (Figure 2), and the information size was reached in the TSA indicating sufficient statistical power. There were significantly fewer desaturation events (Figure 3), need for rescue intervention (Figure 4), and a lower peak EtCO2 ( Figure 5) in the standard ventilation group. The studies included in the meta-analyses were found to have low risk of selection bias. Complications and mortality were rare in both groups.
The patient populations were similar between both the HFNC and jet ventilation sets of case series, with a median age of 54.9 (HFNC) and 50.6 (jet ventilation), with a majority of ASA 1 and 2, and median BMI of 26.8 and 26.9, respectively. The profile of interventions was similar, with the majority of studies looking at microlaryngeal and laryngeal surgery. In most studies, the median apnea time was between 15-20 minutes. After this period, transcutaenous CO 2 and PaCO2 increased but desaturation was rare. Analyses by Liu et al. [62] and Spence et al. [63] showed that the minimum O 2 saturation was higher and safe apnea time was extended in the HFNC group compared to oxygenation by regular nasal cannula. Hung et al.'s study of patients undergoing gastrointestinal endoscopy showed that HFNC was associated with reduced risks of oxygen desaturation, severe hypoxemia, other airway interventions, procedure interruption, and CO 2 level but did not affect procedure time [61]. Our study shows for short laryngeal procedures, HFNC is a viable strategy for apneic oxygenation and reduces the duration of surgery, hypothetically due to an improved surgical view and turnover time by reducing the need for intubation and extubation.
The studies included in this review and meta-analysis were mostly of elective surgeries performed on ASA 1 and 2 patients with a median age of 54.9 and a median BMI of 26.8. Although Lee and Quek reported a case of the use of THRIVE in a morbidly obese patient to facilitate airway surgery [65], obese patients have reduced functional residual capacity, higher risk of lung atelectasis, obstructive sleep apnea, respiratory depression, reflux and regurgitation. While the use of THRIVE is safer than pre-oxygenation or low-flow oxygen [61][62][63], our study shows there is an increased risk of desaturation, number of rescue interventions, and peak EtCO2 compared to mechanical ventilation and SHFJV. In patients with risks of aspiration and laryngospasm, less physiological reserve, pulmonary hypertension, and severe obstructive sleep apnea, and the morbidly obese, MV with a definitive airway remains the potentially safer option.
Laser is occasionally used in laryngeal surgery and airway fire is a potentially fatal complication. Several of the studies in both the HFNC and jet ventilation groups included patients undergoing laser surgery [15,46,35], with no cases of airway fire reported. Some studies have reported the use of THRIVE during laser cases without adverse events or airway fire by turning off the oxygen flow for 40 seconds before the use of laser or by decreasing the fraction of inspired oxygen and suctioning airway gases [5,46]. In a physical model simulation, laser use in high-flow nasal oxygen can lead to "violent" self-sustained fires with continuous laser [66]. Further precautions such as using low-wattage lasers, minimizing lasering time, and reducing oxygen concentration should be taken to minimize the risks of combustion when utilizing lasers with THRIVE [67,68], but the short use of diathermy was safe, although there has been a case report of intra-oral ignition of monopolar diathermy during the use of THRIVE [68].

Limitations
Most of the data in this study was largely dependent on the results of non-comparative case series and retrospective studies. There were only a few comparative studies and fewer RCTs. In the retrospective comparative studies, the grouping of the interventions and patients made meta-analysis impossible decreasing the amount of analysable data [54][55][56]. Publication bias was unavoidable as negative results are less likely to be submitted or accepted for publication. The jet ventilation studies were older compared to HFNC studies, introducing bias. We reduced this risk by only including recent studies, utilizing the Joanna Briggs Institute critical appraisal checklist for case series and only including well-performed case series with reliable data. Many of the older jet ventilation studies did not report the outcomes, reducing the amount of aggregatable data. There were several studies that did not report or discuss complications; however, this was true for both HFNC and jet ventilation studies, which likely reduced any effect on the safety analysis. Summarising the effect estimates by using the median and IQR, while an accepted method according to the Cochrane Handbook, is limited as it does not account for differences in the sizes of the studies and its performance has not been evaluated.
The statistical heterogeneity in the meta-analysis was low, but the number of studies included was low and had a moderate to high risk of bias; however, the risk of selection bias was low for the comparative studies included in the meta-analyses as seen by the risk of bias assessments. We increased the reliability of our results through trial sequential analyses, implementing strict inclusion/exclusion criteria reducing the clinical heterogeneity and utilizing the GRADE framework. While the standard ventilation arm included SHFJV patients, the number was low (n=4) so there was minimal risk of introducing a confounder. For the retrospective comparative studies, there was a risk of patient selection bias in choosing conventional ventilation or HFNC. This impact was minimized as the meta-analysis consisted mostly of prospective RCTs and only one retrospective study, which had a low risk of selection bias [35]. Despite the differences, the impact on length of hospital stay, mortality, or complication rate compared to conventional ventilation is uncertain.

Conclusions
This pre-registered, comprehensive study shows that while HFNC is a viable and safe alternative in adult laryngeal surgery including microlaryngoscopy, suspension laryngoscopy, and laryngotracheal procedures, conventional ventilation may be safer. Compared to conventional ventilation, HFNC was associated with reduced operative time, but with an increased risk of desaturation, hypercarbia, and requirement for rescue intervention. Leaving aside the benefits of tracheal intubation, apnoeic HFNC oxygenation should be used with cautious monitoring during laryngeal microsurgery. The use of HFNC for laryngeal surgery settings mandates close communication between anesthesiologists and surgeons for patient selection and rescue interventions and patient-specific rescue plans should be instituted. In higher-risk patients, conventional ventilation with endotracheal intubation remains the gold standard from a safety perspective.
Larger prospective RCTs of HFNC versus jet ventilation and HFNC versus MV in laryngeal surgery are needed, with a focus on the postoperative outcomes and consequences of desaturation and high EtCO2. Further research on the selection of certain groups of patients who are safe for HFNC such as those with high BMI, elderly, pediatrics, and patients with obstructive sleep apnea is required.