Noninvasive respiratory support for preoxygenation in emergency intubation: a systematic review and network meta-analysis

Abstract

Background

The benefits of preoxygenation with noninvasive respiratory support (NRS), including high-flow oxygen therapy (HFOT) and noninvasive ventilation (NIV), compared to conventional oxygen therapy (COT) during emergency endotracheal intubation (ETI) remain unclear. This network meta-analysis aims to evaluate whether preoxygenation with NRS is more effective than COT in minimising the lowest recorded peripheral capillary oxygen saturation (S_pO2) during emergency ETI.

Methods

A comprehensive literature search was conducted (PROSPERO-CRD42024606842) across Medline, Embase and Scopus. The PICOS criteria were: P: critically ill adult patients requiring emergency ETI; I: randomisation for receiving preoxygenation with NRS; C: randomisation for COT; O: the lowest recorded S_pO2 during emergency intubation (and additional secondary outcomes); S: randomised clinical trials (RCTs).

Results

15 RCTs (2939 patients) met the inclusion criteria. Compared to COT, all NRS methods improved the lowest S_pO2 during emergency ETI (mean difference for HFOT was 1.50, 95% CI 0.43–2.58, p=0.006; for NIV was 3.30, 95% CI 1.81–4.79, p<0.001) (low evidence). Moreover, NIV reduced the occurrence of severe desaturations (S_pO2 <80%) (OR 0.31, 95% CI 0.15–0.61, p<0.001) (very low evidence). Finally, preoxygenation with NRS did not increase the risk of complications (including aspiration, hypotension, barotrauma, arrhythmia or cardiac arrest), and no differences were found in postintubation gas exchange, mechanical ventilation or mortality compared to COT.

Interpretation

During emergency ETI in critical care areas, despite a low certainty of evidence, preoxygenation with NRS overperformed COT in maintaining S_pO2. Only NIV reduced the incidence of severe desaturation, while the risk of complications and adverse events was similar across different preoxygenation devices.

Shareable abstract

In this network meta-analysis, focused on emergency intubation in critical care areas, overall, noninvasive respiratory support showed higher values of the lowest recorded S_pO2, compared to conventional preoxygenation techniques https://bit.ly/3HGQrVk

Introduction

The use of noninvasive respiratory support (NRS), including high-flow oxygen therapy (HFOT) and noninvasive ventilation (NIV) [1–5], has garnered significant attention as a potential strategy to improve oxygenation and decrease desaturation during emergency endotracheal intubation (ETI) [6–9], commonly performed in critically ill patients, prior to induction [10–13]. Over the past decade, numerous studies and scientific societies have explored this topic; however, the actual benefits of NRS for preoxygenation during emergency ETI remain unclear [9–16].

Both HFOT and NIV have been suggested to increase oxygen reserve, to extend the duration of “safe” apnoea and, especially HFOT, to offer greater comfort, as compared to NIV or no intervention, before intubation [1–5, 13–17]. Specifically, NRS can improve preoxygenation during emergency ETI by providing higher and more stable oxygen concentrations and enhancing alveolar recruitment compared to conventional oxygen therapy (COT). In particular, HFOT helps by reducing dead space, creating an oxygen reservoir, and delivering heated and humidified oxygen at high flow rates, even during apnoea [1, 2, 16]. Conversely, NIV increases functional residual capacity, decreases atelectasis and enhances oxygenation, especially in critically ill patients [4, 6]. However, despite these promising results, large randomised controlled trials (RCTs) have not yet demonstrated a clear clinical superiority of HFOT or NIV, as compared to COT, in reducing mortality or adverse events during the pre-intubation phase, particularly in emergency conditions [1, 6, 18]. Interestingly, the “Intube study”, mainly focused on all peri-intubation complications, reported an incidence of severe hypoxaemia of around 9%, of cardiac arrest of 3.1% and aspiration of 4%, values that are absolutely not negligible in clinical practice [9].

Therefore, we designed this systematic review and meta-analysis of RCTs to evaluate whether, among critically ill adult patients (≥18 years) requiring emergency ETI, preoxygenation with NRS, including HFOT and NIV (either continuous positive airway pressure or bilevel positive pressure ventilation), is associated with differences in the lowest recorded peripheral capillary oxygen saturation (S_pO2) during the intubation procedure, compared to COT.

Secondary outcomes included the proportion of patients experiencing severe desaturation during the procedure, intubation-related complications, post-intubation arterial partial pressure of oxygen (P_aO2) and carbon dioxide (P_aCO2), the length of invasive mechanical ventilation (IMV), and short-term mortality.

Methods

The Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA) Statement extension for network meta-analysis was followed for writing this systematic review and meta-analysis (supplementary e-Material 1) [19]. The review protocol was registered in PROSPERO (CRD42024606842) (7 November 2024), an international prospective registry of systematic reviews.

Literature search

An electronic search of Medline, Embase and Scopus from inception until 20 November 2024 was performed with no language restrictions. Grey literature (OpenGrey) and all references of included articles and related reviews and guidelines were searched. A full description of search strategies is reported in the supplementary e-Material 2.

Study selection and data collection

All studies meeting the following Participants, Interventions, Comparisons, Outcomes and Study design (PICOS) question were included: participants were critically ill adult patients requiring emergency ETI; the intervention was preoxygenation through NRS; the comparison was COT; the primary outcome was the lowest recorded S_pO2 during intubation, which means from the induction to ETI [18], while the number of patients experiencing severe desaturation (defined as S_pO2 <80%) [9], the incidence of intubation-related complications (e.g., gastric aspiration, hypotension, barotrauma, arrhythmia and cardiac arrest), post-intubation P_aO2 and P_aCO2 (both measured through the first arterial blood gas taken after ETI), duration of IMV and short-term mortality (defined as 30-day or in-hospital mortality) were secondary outcomes. Eligible study designs were RCTs. Gas exchange data collected during preoxygenation were not analysed due to the high risk of heterogeneity in data collection timing and the relatively small number of studies reporting these parameters [5, 20].

Search results were merged and duplicate records from the same report were removed. Six researchers (T. Pettenuzzo, N. Sella, A. De Cassai, F. Monteleone, P. Zanon and S. Ongaro) were split into three couples, each analysing the same number of overall identified citations. Specifically, each member of the couple independently screened the titles and abstracts of the assigned papers and retrieved the full texts of potentially relevant reports. The reasons for exclusion are reported in figure 1. Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia; available at www.covidence.org) was used for study selection, under the supervision of biostatisticians involved in the study group (H. Ocagli and D. Gregori).

FIGURE 1.

PRISMA flow-chart.

Four researchers (F. Monteleone, P. Zanon, D. Ongaro and S. Talarico) were split into two couples, each analysing the same number of eligible full texts. Specifically, each member of the couple independently assessed the full text of the assigned papers.

Data from included studies were recorded using a Microsoft Excel (Microsoft Corporation, Redmond, WA, USA) report form. Two researchers (E. Pistollato and A. Zambianchi) independently verified all extracted data for accuracy. Any disagreements on study selection and data extraction were resolved by referral to other authors (N. Sella or A. Boscolo), if necessary.

The following information was collected: first author, study year, journal, inclusion and exclusion criteria, outcomes of interest, number of patients and sex.

Certainty of evidence assessment

Four researchers (F. Monteleone, S. Ongaro, S. Talarico and F. Monteleone) were split into two couples and assessed the risk of bias of the same number of included studies. Specifically, each member of the couple independently evaluated the quality of included RCTs using the Risk of Bias (RoB) 2 assessment tool, which examines five domains of bias, i.e., the randomisation process, deviations from intended interventions, missing outcome data, outcome measurement and selection of the reported results. The RoB 2 tool categorises the study-level risk of bias on a three-grade scale, i.e., low risk of bias, high risk of bias or some concerns [21, 22]. Disagreements were resolved by discussion with another author (P. Navalesi), if necessary.

Publication bias was evaluated by visually inspecting a funnel plot for potential asymmetry and Egger's test was applied when the number of studies was >10.

The Grades of Recommendation, Assessment, Development and Evaluation (GRADE) approach was applied to assess the certainty of evidence related to the primary outcome and the following secondary outcomes: severe desaturation, aspiration and mortality [23]. The GRADE addresses the domains of risk of bias, inconsistency, indirectness, publication bias, intransitivity, incoherence and imprecision. Imprecision for each comparison was only incorporated at network level, not at the level of the direct or indirect estimate. A minimally contextualised approach, considering COT as reference, was applied to evaluate the magnitudes of intervention effect [24].

Sensitivity analysis

For the primary outcome only, a pre-planned sensitivity analysis was conducted excluding all RCTs at high RoB.

Subgroup analyses

For the primary outcome only, pre-planned subgroup analyses were conducted to assess the impact of NRS across: 1) studies enrolling only hypoxaemic patients versus those including “mixed” populations; and 2) studies including only patients admitted to the intensive care unit (ICU) or emergency department (ED) or operative rooms (ORs) or enrolled in mixed clinical settings.

Analysis

Conventional pairwise meta-analyses were performed using a random-effects model to account for between-study heterogeneity [25]. The treatment effect for dichotomous outcomes was analysed with the Mantel–Haenszel method and expressed as odds ratio (OR) with 95% confidence interval (CI). The treatment effect for continuous outcomes was analysed with the inverse variance method and expressed as mean difference (MD) or standardised mean difference (SMD) with 95% CI [26]. Whenever necessary, we converted reported median and interquartile range to estimated mean and standard deviation (sd) using Wan's method [27]. When no events were observed in any of the groups in an individual study, a fixed value of 1 was added as a continuity correction to the cells corresponding to the number of events [26]. Statistical heterogeneity for the outcomes among studies was assessed using the chi-squared test and I² statistic. We considered I²≥75% and a p-value<0.1 as high heterogeneity [26].

Availability of evidence, transitivity assumption, intra-network connectivity and network coherence were considered to assess the feasibility of conducting a network meta-analysis [28]. Direct and indirect treatment estimates were compared to check for incoherence in the network meta-analyses [29].

We performed frequentist random-effects network meta-analyses [28]. The direct treatment estimates were based on the common between-study variance Tau² from the network meta-analysis. The indirect estimates were obtained with the Separate Indirect from Direct Evidence and the Separate Indirect from Direct Design Evidence methods [28].

A ranking among treatments was performed based on the frequentist analogue of the surface under the cumulative ranking curve (SUCRA). The ranking was provided as p-scores, ranging from 0 (minimum) to 1 (maximum) and measuring the mean extent of certainty that a NRS modality is better than the competing modalities [28].

For all analyses, two-sided p-values <0.05 were considered significant. All analyses were performed with Review Manager version 5.3 (Nordic Cochrane Centre, Cochrane Collaboration) and R version 4.3.1 (R Foundation for Statistical Computing, Vienna, Austria) with the “netmeta” package for frequentist network meta-analyses. For all analyses, two-sided p-values <0.05 were considered significant.

Results

Study selection and data retrieval

The PRISMA flowchart, shown in figure 1, illustrates the study selection process. Initially, 8346 studies were retrieved from databases. 15 RCTs (2939 patients) [5, 6, 13, 18, 20, 30–39] entered qualitative and quantitative analysis. We requested missing data (twice) from all authors of the included trials, but none was able to provide the missing information.

Study section, characteristics and quality

The characteristics of all studies are summarised in supplementary e-Material 3. Included studies were published between 2006 and 2024. Study populations (>60% male) ranged from 40 to 1301 patients. COT was administered in 1253 (43%) patients, HFOT in 621 (21%), NIV in 1011 (34%), while only two studies, including 54 (2%) patients, investigated the use of a combination of HFOT and NIV, and their findings were discussed marginally [33, 35].

According to the RoB evaluation, one study [35] was defined at low risk, one [30] raised some concerns, while 13 studies were at high RoB. The rationale for each judgement is provided in the supplementary e-Material 4.

The Egger's test confirmed the absence of publication bias for the lowest S_pO2 (p=0.087), the occurrence of intubation-related complications (p=0.229) and mortality (p=0.567); while not for severe desaturation (p=0.032) (supplementary e-Material 5). For the remaining funnel plots, the presence of publication bias was not suspected from the visual inspection (supplementary e-Materials 5 and 6).

Primary outcome

As detailed in figure 2a, preoxygenation with NRS, during emergency ETI, significantly improved the lowest recorded S_pO2 levels, compared to COT (I² 46%). Specifically, the MD for HFTO was 1.50 (95% CI 0.43–2.58, p=0.006; low certainty of evidence), for NIV was 3.30 (95% CI 1.81–4.79, p<0.001; low certainty of evidence), and for the combination of NIV and HFOT was 5.53 (95% CI 2.06–9.00, p=0.002; low certainty of evidence) [5, 6, 13, 18, 20, 30–39]. Supplementary e-Material 7 depicts network estimates for the contribution of direct and indirect evidence to inform on local inconsistency in network meta-analyses.

FIGURE 2.

Outcomes (forest plots, p-scores and netgraphs). COT: conventional oxygen therapy; NRS: noninvasive respiratory support; S_pO2: peripheral oxygen saturation; OR: odds ratio; MD: mean difference; NIV: noninvasive ventilation; HFOT: high-flow oxygen therapy.

Sensitivity and subgroup analyses, detailed in figure 3, include forest plots and p-score estimates for various interventions. In detail, after excluding studies with a high RoB, the noninvasive strategies resulted in a higher lowest recorded S_pO2.

FIGURE 3.

Sensitivity and subgroup analysis (forest plots) (primary outcome). a) Sensitivity analysis: without studies at high risk of bias; b) subgroup analysis: according to pre-intubation P_aO2 values (if <300 mmHg patients were defined as “hypoxaemic”); c) subgroup analysis: according to clinical setting of emergency intubation. COT: conventional oxygen therapy; NRS: noninvasive respiratory support; S_pO2: peripheral oxygen saturation; MD: mean difference; NIV: noninvasive ventilation; HFOT: high-flow oxygen therapy; ICU: intensive care unit; ED: emergency department; OR: operative room.

Then, as shown in figure 3b, NIV and NIV+HFOT, and also HFOT alone, were associated with improved lowest S_pO2, as compared to COT, particularly among hypoxaemic patients [5, 13, 30, 35, 38], with an MD of 6.27 (95% CI 2.86–9.69), 8.93 (95% CI 3.14–14.72) and 3.09 (95% CI 0.08–6.11), respectively.

Moreover, the performance of NRS varied by clinical setting. As compared to COT, HFOT was significantly associated with improved lowest S_pO2 in the ED, with an MD of 2.04 (95% CI 0.66–3.41), but not in the OR or ICU [29, 40]; while NIV, alone or in combination with HFOT, showed promising results especially among ICU patients [5, 6, 13, 20, 30, 33–35, 38, 39] (figure 3c).

Secondary outcomes

The proportion of patients experiencing severe desaturation (2270 patients) was significantly lower in the NIV group, with an OR of 0.31 (95% CI 0.15–0.61, p<0.001; very low certainty of evidence). However, this benefit was not observed with HFOT (0.68, 95% CI 0.33–1.39, p=0.286; very low certainty of evidence) or with the combination of HFOT and NIV (OR 0.57, 95% CI 0.12–2.67, p=0.477; very low certainty of evidence) [5, 6, 13, 18, 20, 30, 32–35, 38, 39] (figure 2b and table 1). Overall, the I² was 59%.

TABLE 1.

Outcome/Comparison	Effect estimate (95% CI)	Direct evidence				Direct rating	Indirect evidence		Indirect rating	Network				Certainty of evidence
		RoB	Inconsistency	Indirectness	Publication		Lowest rating	Intransitivity		Proportion	Rating of most contributing evidence	Incoherence	Imprecision
The lowest SpO2					0
HFOT versus COT	1.57 (0.43–2.72)	2#,¶	0	0		Low	Low	0	Low	0.93	Low	0	0	Low
NIV versus COT	3.34 (1.80–4.89)	2+	0	0		Low	Low	0	Low	0.85	Low	0	0	Low
NIV+HFOT versus COT	5.56 (2.01–9.11)	-	-	-		-	Low	0	Low	0	Low	-	0	Low
NIV versus HFOT	1.77 (-0.05–3.58)	2§	0	0		Low	Low	0	Low	0.22	Low	0	1ƒ	Very low
NIV+HFOT versus HFOT	3.99 (0.31–7.66)	-	-	-		-	Low	0	Low	0	Low	-	0	Low
NIV+HFOT versus NIV	2.22 (-0.98–5.42)	1##	0	0		Moderate	-	-	-	1.00	Moderate	-	1ƒ	Low
SpO2 <80%					1++
HFOT versus COT	0.56 (0.24–1.33)	2¶¶	0	0		Very low	Very low	0	Very low	0.71	Very low	0	1§§	Very low
NIV versus COT	0.29 (0.14–0.60)	2ƒƒ	0	0		Very low	Very low	0	Very low	0.25	Very low	0	0	Very low
NIV+HFOT versus COT	0.52 (0.11–2.58)	-	-	-		-	Very low	0	Very low	0	Very low	-	2###	Very low
NIV versus HFOT	0.52 (0.20–1.32)	2§	0	0		Very low	Very low	0	Very low	0.47	Very low	0	1ƒ	Very low
NIV+HFOT versus HFOT	0.93 (0.17–5.11)	-	-	-		-	Very low	0	Very low	0	Very low	-	1ƒ	Very low
NIV+HFOT versus NIV	1.81 (0.44–7.46)	0	0	0		Moderate	-	-	-	1.00	Moderate	-	1ƒ	Low
Aspiration					0
HFOT versus COT	1.03 (0.22–4.76)	2¶¶¶	0	0		Low	Low	0	Low	0.77	Low	0	2###	Very low
NIV versus COT	0.59 (0.23–1.47)	2ƒƒ	0	0		Low	Low	0	Low	0.93	Low	0	2###	Very low
NIV+HFOT versus COT	0.56 (0.03–11.01)	-	-	-		-	Low	0	Low	0	Low	-	2###	Very low
NIV versus HFOT	0.57 (0.11–2.98)	2§	0	0		Low	Low	0	Low	0.30	Low	0	1ƒ	Very low
NIV+HFOT versus HFOT	0.54 (0.02–14.44)	-	-	-		-	Low	0	Low	0	Low	-	1ƒ	Very low
NIV+HFOT versus NIV	0.96 (0.06–16.24)	0	0	0		High	-	-	-	1.00	High	-	1ƒ	Moderate
Mortality					0
HFOT versus COT	0.83 (0.60–1.15)	2¶¶¶	0	0		Low	Low	0	Low	0.60	Low	0	1§§	Very low
NIV versus COT	0.95 (0.78–1.16)	2+	0	0		Low	Low	0	Low	0.90	Low	0	1§§	Very low
NIV+HFOT versus COT	0.75 (0.23–2.47)	-	-	-		-	Low	0	Low	0	Low	-	2###	Very low
NIV versus HFOT	1.14 (0.82–1.59)	2§	0	0		Low	Low	0	Low	0.51	Low	0	1ƒ	Very low
NIV+HFOT versus HFOT	0.90 (0.26–3.05)	-	-	-		-	Low	0	Low	0	Low	-	1ƒ	Very low
NIV+HFOT versus NIV	0.78 (0.24–2.55)	0	0	0		High	-	-	-	1.00	High	-	1ƒ	Moderate

RoB: risk of bias; S_pO2: peripheral oxygen saturation; HFOT: high-flow oxygen therapy; COT: conventional oxygen therapy; NIV: noninvasive ventilation. ^#: numbers represent the degree of downrating of certainty of evidence; ^¶: lowered two levels for risk of bias (high risk in 7/7 studies); ⁺: lowered two levels for risk of bias (high risk in 3/4 studies and some concerns in 1/4 studies); ^§: lowered two levels for risk of bias (high risk in 1/1 study); ^ƒ: the CI crosses the null effect; ^##: lowered one level for risk of bias (high risk in 1/1 study and low risk in 1/1 study); ^¶¶: lowered two levels for risk of bias (high risk in 4/4 studies); ⁺⁺: a publication bias was strongly suspected from the visual inspection of the funnel plot and the accompanying Egger's regression (p=0.0315); ^§§: the CI crosses the null effect and includes the possibility of harm; ^ƒƒ: lowered two levels for risk of bias (high risk in 2/2 studies); ^###: the CI crosses the null effect and includes the possibility of important harm; ^¶¶¶: lowered two levels for risk of bias (high risk in 3/3 studies).

The rate of intubation-related complications (2536 patients) (figure 2c and supplementary e-Material 7), including aspiration, hypotension, barotrauma, arrhythmia and cardiac arrest, was similar across all NRS techniques and COT [5, 6, 18, 30–35, 37, 39]. Overall, the I² was 39%. A full description of all complications is reported in figure 4.

FIGURE 4.

Additional complications (forest plots, p-scores and netgraphs). MD: mean difference; OR: odds ratio; NRS: noninvasive respiratory support; NIV: noninvasive ventilation; HFOT: high-flow oxygen therapy; COT: conventional oxygen therapy.

Moreover, HFOT and NIV were comparable to COT in terms of post-intubation P_aO2 and P_aCO2 levels (I² 58% and 0%, respectively) (440 patients) [5, 20, 30, 36], duration of IMV (1096 patients; I² 0%) and mortality (2386 patients; I² 0%) [4–6, 18, 19, 30–32, 34, 39] (figure 5a–d, table 1 and supplementary e-Material 7).

FIGURE 5.

Secondary outcomes (forest plots, p-scores and netgraphs). COT: conventional oxygen therapy; NRS: noninvasive respiratory support; S_pO2: peripheral oxygen saturation; OR: odds ratio; MD: mean difference; SMD: standard mean difference; NIV: noninvasive ventilation; HFOT: high-flow oxygen therapy.

Discussion

This systematic review and network meta-analysis, overall including 15 RCTs (2939 patients), evaluated the benefits and safety of different NRS techniques, as preoxygenation methods for emergency ETI among critically ill patients. Our findings indicate that, despite a low certainty of evidence, all NRS methods improved the lowest recorded S_pO2 during ETI, compared to COT. Notably, the most significant benefits of NRS, especially of NIV, were observed in procedures performed in the ICU; while for HFOT, though less pronounced, among ED patients.

NRS may enhance preoxygenation during emergency ETI by delivering higher and more stable oxygen concentrations and improving alveolar recruitment, compared to COT [41, 42]. Specifically, HFOT reduces dead space, creates an oxygen reservoir, and delivers heated and humidified oxygen at high flow rates, even during apnoea [1, 2, 16, 42, 43]. NIV, on the other hand, increases functional residual capacity, reduces atelectasis and improves oxygenation, particularly in critically ill patients [4, 6, 44, 45]. In fact, as described by Frat et al. [6], NIV seems to overperform HFOT only after adjusting for baseline P_aO2 values and, therefore, especially in moderate-to-severe hypoxaemic patients.

Additionally, HFOT can partially assist in carbon dioxide (CO₂) clearance through continuous dead space washout, providing an added safety margin for patients with pre-existing hypercapnia [3, 42, 46]. Similarly, NIV facilitates CO₂ elimination by enhancing alveolar ventilation, reducing the risk of hypercapnia and respiratory acidosis during apnoea [42, 47, 48]. However, our analysis did not show any significant improvement in post-ETI gas exchange, likely due to several confounding factors: the limited number of studies (only four) reporting these outcomes, variability in the timing of arterial blood gas collection (mostly within 5 min post-ETI), unclear documentation of post-procedure ventilation methods, and the inability to distinguish between hypercapnic and non-hypercapnic patients prior to ETI. Moreover, the postintubation blood gas measurements are a useless measure without any data prior to intubation.

This meta-analysis also provides insights into the safety of NRS. The risk of complications was comparable between COT and NRS, despite a very low certainty of evidence, suggesting that preoxygenation through NRS does not introduce additional risks. These findings are in line with recent RCTs and previous meta-analysis demonstrating that NRS effectively reduces critical desaturation during intubation without compromising safety [3, 14, 15, 18, 31, 35, 37, 41, 49, 50]. Additionally, our subgroup analyses revealed that NIV performed better in the ICU, but there is not a clear explanation for justifying these findings or evidence supporting the hypothesis that could be likely due to the availability of resources and expertise.

The observed differences in the location of NRS delivery may reflect variability in resource availability, institutional protocols and patient acuity. For example, ICUs typically offer greater access to advanced respiratory support devices and specialised staff, facilitating more frequent and effective use of NRS. In contrast, EDs or prehospital settings may face logistical limitations or prioritise rapid intubation, particularly in unstable patients. Additionally, variations in clinician training and local practice patterns may further contribute to this heterogeneity.

Focusing on patients’ general anaesthesia, de Carvalho et al. [51] recently showed preoxygenation of adults before induction of general anaesthesia was most effective in terms of safe apnoea time when performed with HFOT with patients in the head-up position in comparison with face mask alone. Furthermore, these results had already been anticipated by Huo et al. [52] a few years earlier.

However, none of these previous meta-analyses were specifically focused on emergency ETI in critical care areas, and this is the strength of our work.

Finally, our study has some weaknesses. First, due to a relatively low number of studies investigating the combination of HFOT+NIV, the results derived from those findings have questionable scientific validity, and they were reported only for transparency. Second, our analysis does not take into account the significant variability in pre-oxygenation across different devices (i.e., nasal cannula, Venturi mask, etc.), which were included in the “COT” group. Furthermore, no information was provided regarding the effectiveness of preoxygenation and denitrogenation. Third, due to missing data, we were not able to provide a subgroup analysis according to the working experience of the airway operator. Fourth, a low (or very low) certainty of evidence for most outcomes and comparisons, mainly related to the risk of bias and imprecision, highlights the need to refine which patients would benefit mostly from different respiratory supports, and not which modality is the best across all patients.

In conclusion, NRS for preoxygenation during emergency ETI overperformed COT in maintaining S_pO2, without increasing the risk of adverse events. Moreover, among NRS techniques, NIV outperformed the other techniques in minimising the risk of severe desaturation.

Supplementary material

Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.

Supplementary material

00329-2025.SUPPLEMENT

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.