Post-extubation Oxygenation Strategies and Mortality and Reintubation Rates in Acute Respiratory Failure: A Systematic Review and Network Meta-Analysis

Background: High-ow nasal cannula oxygenation (HFNC) and noninvasive positive-pressure ventilation (NPPV) possibly decrease tracheal reintubation rates better than conventional oxygen therapy (COT); however, few large-scale studies have compared HFNC and NPPV. We conducted a network meta-analysis (NMA) to compare the effectiveness of three post-extubation respiratory support devices (HFNC, NPPV, COT) in reducing the mortality and reintubation risk. Methods: The Cochrane Central Register of Controlled Trials, MEDLINE, EMBASE, and Ichushi databases were searched. COT, NPPV, and HFNC use were assessed in patients aged ≥ 16 years who underwent invasive mechanical ventilation for >12 hours for acute respiratory failure and were scheduled for extubation after spontaneous breathing trials. The GRADE Working Group Approach was performed using a frequentist-based approach with multivariate random-effects meta-analysis. Short-term mortality and reintubation and post-extubation respiratory failure rates were compared. Results: After evaluating 4,631 records, 15 studies and 2,600 patients were included. The main cause of acute hypoxic respiratory failure was pneumonia. Although NPPV/HFNC use did not signicantly lower the mortality risk (relative risk [95% condence interval], 0.75 [0.53–1.06] and 0.92 [0.67–1.27]; low and moderate certainty, respectively), HFNC use signicantly lowered the reintubation risk (0.54 [0.32–0.89]; high certainty) compared with COT use. The associations of mortality with NPPV and HFNC in either outcome did not differ signicantly (short-term mortality and reintubation, relative risk [95% condence interval], 0.81 [0.61–1.08] and 1.02 [0.53–1.97]; moderate and very low certainty, respectively). Conclusion: NPPV or HFNC use may reduce endotracheal reintubation risk, but not short-term mortality risk.


Background
Invasive mechanical ventilation (IMV) is a life-saving procedure for patients with acute respiratory failure (ARF) [1]. Approximately 10-20% of patients who are extubated after a successful spontaneous breathing trial (SBT) require reintubation within 48-72 hours [2], which may be associated with prolonged mechanical ventilation, extended intensive care unit (ICU) and hospital stay, and increased mortality [2,3].
NPPV has been reported to be effective in preventing reintubation after planned extubation in high-risk patients [6,7,10,11]. However, NPPV may increase the risk of complications, including aspiration pneumonia, interface intolerance, and patient discomfort [12,13]. HFNC can minimize the complications of NPPV by delivering high concentrations of humidi ed oxygen via a nasal cannula. However, contradictory results have been reported despite the large number of clinical trials [14,15].
Some systematic reviews and meta-analyses which compared two of the three respiratory support devices (COT, NPPV, and HFNC) [16][17][18][19][20] have shown that in terms of reducing the rate of tracheal reintubation, HFNC was better than COT but equivalent to NPPV. Moreover, there were no signi cant differences between the therapies in terms of mortality rates. Although several studies have compared HFNC with COT and NPPV with COT, few large-scale studies have compared HFNC with NPPV. Therefore, small sample sizes may have affected the results of systematic reviews.
To overcome these limitations, we performed a systematic review and network meta-analysis (NMA) to compare the effectiveness of three respiratory support devices in reducing mortality and reintubation rates by including studies that compared two of the three respiratory support devices (COT, NPPV, and HFNC) in patients with ARF.

Protocol and registration
This systematic review was designed in accordance with the Preferred Reporting Items for Systematic review and Meta-Analyses (PRISMA) extension statements for reporting systematic reviews that incorporate NMA (e- Table 1 in additional le 1) [21]. The review protocol is registered with PROSPERO (CRD42020139112).

Studies, participants, interventions/comparators, and outcomes
We included all reports of randomized controlled trials (RCTs) in English and Japanese regardless of publication status (e.g., published, unpublished, and academic abstracts). Randomized crossover, cluster randomized, and quasi-experiment trials were excluded. This metaanalysis included reviews of adult patients (age ≥16 years) who underwent IMV for more than 12 hours due to ARF and were scheduled for extubation after an SBT. The de nitions of acute hypoxic respiratory failure and SBT were individualized for each study. This meta-analysis excluded studies that included patients who underwent tracheostomies, experienced accidental extubation or self-extubation, those who experienced hypercapnia during SBT, and those who had do-not-resuscitate (DNR) orders. Studies in which more than half of the study population had acute chronic obstructive pulmonary disease (COPD) exacerbation, those that included patients with a postoperative status or who were being treated for trauma, and those that included patients with congestive heart failure were also excluded. We included RCTs that compared two of the three available respiratory support devices: (1) COT: low-ow nasal cannula, face mask, and venturi mask (no ow rate restriction); (2) NPPV: the type of mask, mode, duration of ventilation, and weaning methods were not limited; and (3) HFNC: no limitations on the ow rate or F I O 2 . The outcome measures evaluated were as follows: the primary outcome was the short-term mortality rate ( [1] at the end of the follow-up period for each trial within 30 days, [2] at ICU discharge, and [3] at hospital discharge). Secondary outcomes included the reintubation rate within 72 hours (reintubation included the need for intubation and NPPV) and post-extubation respiratory failure rate (the de nition was individualised for each study).

Data sources and search details
We searched the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE via PubMed, EMBASE, and Ichushi, a database of Japanese papers for eligible trials. We searched for ongoing trials in the World Health Organization International Clinical Trials Platform Search Portal. In cases of missing data, we attempted to contact the authors of each study. Searches were performed in December 2020. Details regarding search strategy and when the searches were performed are shown in e- Table 2 in additional le 1.
Study selection, data collection process, and data items Two of the three physicians (YO, CN, and HY) screened the title, abstract, and full text during the rst and second screenings for relevant studies and independently extracted data from eligible studies into standardized data forms. For abstract-only studies that could not be evaluated according to the eligibility criteria, we contacted the authors. Disagreements, if any, between two reviewers were resolved via discussion among themselves or with a third reviewer as necessary. Data extraction from identi ed studies during the second screening was done by three reviewers using two tools: (1) the Cochrane Data Collection Form (RCTs only) [22] and (2) Review Manager (RevMan) software V.5.3.5 [23].

Risk of bias within individual studies
The risk of bias for primary outcomes were independently assessed using the Cochrane Risk of Bias tool 1.0 [24,25]. Each bias was graded as 'low risk', 'unclear risk', or 'high-risk'. Discrepancies between reviewers were resolved by mutual discussion.

Statistical analyses
Direct comparison meta-analysis A pairwise meta-analysis was performed by using RevMan 5.3 (RevMan 2014). Forest plots were used for the meta-analysis, and effect sizes are expressed as relative risk (RR) and weighted mean differences, both with 95% con dence intervals (CI), for categorical and continuous data, respectively. Outcome measures were pooled using a random effect model to include study-speci c effects in measures. A two-sided p-value <0.05 was considered signi cant.
We planned to use a funnel plot, Begg's adjusted rank correlation test, and Egger's regression asymmetry test for the possibility of publication bias if ≥10 studies were available (RevMan) [27]. However, as <10 studies were included for each outcome, we did not test for funnel plot asymmetry.

Network comparison meta-analysis
Data synthesis A network plot was constructed to determine the number of studies and patients included in this meta-analysis. An NMA, using the netmeta 0.9-5 R-package (version 3.5.1), was performed using a frequentist-based approach with multivariate random effects meta-analysis, and effect size was expressed as the RR (95% CI). Covariance between two estimates from the same study shows variance of data in the shared arm, as calculated in a multivariable meta-analysis performed using the GRADE Working Group Approach for an NMA.

Transitivity
The transitivity assumption underlying the NMA was evaluated by comparing the distribution of clinical and methodological variables that could act as effect modi ers across treatment comparisons.

Ranking
Ranking plots (rankograms) were constructed using the probability that a given treatment had the highest event rate for each outcome. The surface under the cumulative ranking curve (SUCRA), which is a simple transformation of the mean rank, was used to set the hierarchy of the treatments [28] and was created using standard software (Stata 15.0, Stata, TX, USA).

Risk of bias across studies
Assessment of the risk of bias across studies followed considerations on pairwise meta-analysis. Conditions associated with 'suspected' and 'undetected' bias across studies were determined by the presence of publication bias as shown by direct comparison.

Indirectness
The indirectness of each study included in the network was evaluated according to its relevance to the research question, which consisted of the study population, interventions, outcomes, and study setting, and was classi ed as low, moderate, or high. Study-level judgments could be combined with the percentage contribution matrix.

Imprecision
The approach to imprecision comprised a comparison of the range of treatment effects included in the 95% CI with the range of equivalence. We assessed the heterogeneity of treatment effects for a clinically important risk ratio (<0.8 or >1.25) in CI.

Heterogeneity
To assess the amount of heterogeneity, we compared the posterior distribution of the estimated heterogeneity variance with its predictive distribution [29]. The concordance between assessments based on CI and prediction intervals, which do and do not capture heterogeneity, respectively, was used to assess the importance of heterogeneity. We assessed the heterogeneity of treatment effects for a clinically important risk ratio of <0.8 or >1.25 in prediction intervals.

Assessment of inconsistency
The inconsistency of the network model was estimated from inconsistency factors and their uncertainty, and consistency was statistically evaluated using the design-by-treatment interaction test [30]. For comparisons informed only by direct evidence, there was no disagreement between evidence sources and, thus, there was 'no concern' for incoherence. If only indirect evidence was included, there was always 'some concern'. 'Major concern' applied when the p-value of the design-by-treatment interaction test was <0.01.

Study characteristics
The protocols and characteristics of each study included in this meta-analysis are summarized in Table 1. A total of 2,600 patients were included in the quantitative analysis. The main cause of acute hypoxic respiratory failure was pneumonia, followed by postoperative respiratory failure. Of the 15 studies, two mainly included patients with exacerbation of chronic respiratory disorders.
The risk of bias within included studies is shown in e- Fig 2 in additional le 1. Although not all studies blinded participants and clinicians to the intervention, almost all other domains of the risk of bias were low (e- Fig 2 in additional le 1). All studies were judged as having a low risk of bias for outcomes (risk of bias across studies).

Network meta-analysis
The results of pairwise comparisons are shown in e-Figs 3, 4, and 5 in additional le 1 (short-term mortality, reintubation, and post-extubation respiratory failure, respectively). The funnel plot of each outcome was not described because the number of studies included for each comparison was <10.
Con dence in the RR of each comparison and short-term mortality, assessed by the GRADE system, is shown in Table 3. Incoherence between direct and indirect RRs was not observed for any of the three comparisons, according to the p-values of inconsistency. The heterogeneity for all three comparisons resulted in a 'no concern' rating due to the 95% prediction interval of the risk ratio.  Table 2 summarizes the ndings of the NMA for short-term mortality. Moreover, e- Table 3 in additional le 1 summarizes the estimate and certainty of the evidence of direct, indirect, and network comparisons.

Endotracheal reintubation
Fourteen studies were included in the analysis of endotracheal reintubation. Compared with COT, HFNC (RR, 0.54 [95% CI, 0.32-0.89]: high certainty) was signi cantly associated with a lower risk of reintubation (Fig 2b), although there were no signi cant difference in the association with reintubation between NPPV and HFNC (RR, 1.02 [95% CI, 0.53-1.97]: low certainty), and between NPPV and COT (RR, 0.55 [95% CI, 0.30-1.00]: moderate certainty). Anticipated absolute effects and 95% CIs between each of the two comparisons decreased by 62 per 1,000 (95% CI, −96 to 0) for NPPV vs. COT and 60 per 1,000 (95% CI, −88 to −14) for HFNC vs. COT, but increased by 5 per 1,000 (95% CI, −107 to +221) for NPPV vs. HFNC (Table 4). Table 3 shows the con dence in the RR of each comparison and reintubation assessed by the GRADE system. Incoherence between direct and indirect RRs was not observed for all three comparisons and was decided by the p-value of inconsistency. The heterogeneity of two comparisons (NPPV vs. COT and HFNC vs. COT) resulted in 'some concern' and 'major concern', but that of one comparison (HFNC vs. NPPV) resulted in a 'no concern' rating due to the 95% prediction interval of the risk ratio.  Table 4 summarizes ndings of the NMA for reintubation; e- Table 3 in additional le 1 presents the estimate and certainty of the evidence of direct, indirect, and network comparisons.

Post-extubation respiratory failure
Seven studies were included in the analysis of post-extubation respiratory failure. Compared with COT, NPPV (RR, 0.86 [95% CI, 0.54-1.38]: low certainty) and HFNC (RR, 0.66 [95% CI, 0.43-1.02]: moderate certainty) were not associated with a lower risk of mortality (Fig 2c). There was no signi cant difference in the association with mortality between NPPV and HFNC (RR, 1.30 [95% CI, 0.79-2.14]: low certainty). Anticipated absolute effects and 95% CIs between each of two comparison decreased by 26 per 1,000 (95% CI, −87 to +71) for NPPV vs. COT and 57 per 1,000 (95% CI, −95 to +3) for HFNC vs. COT, but increased by 81 per 1,000 (95% CI, −56 to +307) for NPPV vs. HFNC (Table 5). Table 3 shows the con dence in the RR of each comparison and post-extubation respiratory failure assessed by the GRADE system. Incoherence between direct and indirect RRs was not observed for all three comparisons, as indicated by the p-value of inconsistency. The heterogeneity of one comparison (HFNC vs. COT) and that of two comparisons (NPPV vs. COT and HFNC vs. NPPV) resulted in 'some concern' and 'no concern' ratings due to the 95% prediction interval of the risk ratio.  Table 5 summarizes ndings of the NMA for post-extubation respiratory failure, and e- Table 3 in additional le 1 summarizes the estimate and certainty of the evidence of direct, indirect, and network comparisons.

Discussion
In our NMA, there was no between-group differences in short-term mortality (groups: NPPV, HFNC, and COT). NPPV/HFNC use did not signi cantly lower the mortality risk. However, the use of HFNC signi cantly lowered the reintubation risk relative to the COT use. There were no signi cant differences in the rate of endotracheal reintubation or post-extubation respiratory failure between HFNC and NPPV. The mortality rate did not reduce in any two of the three comparisons.
Herein, NPPV contributed to a reduction in the reintubation rate compared to that observed with COT, without reducing mortality, which is consistent with several previous pairwise systematic reviews comparing NPPV and COT. Previous RCTs show that NPPV is more effective in reducing reintubation and mortality rates than COT in a high-risk group of patients with post-extubation respiratory failure, including COPD [7] [ [43][44]. However, Kondo et al. showed that NPPV decreased reintubation and mortality rates more effectively than COT despite the complete exclusion of COPD patients from the study [45]. In our study, we excluded studies with COPD patients constituting >50% of the study population, as COPD is a risk factor for post-extubation respiratory failure [46]. Thus, the abovementioned exclusion potentially caused a difference between the effectiveness of NPPV and COT in the systematic reviews included in the NMA.
When HFNC was compared to COT, differences in outcomes between previous pairwise systematic reviews and this NMA-based study were observed. A systematic review by Ni and colleagues showed that HFNC is associated with a lower reintubation rate than COT, despite no reduction in mortality rate [16], which is identical to our study. Although a systematic review by Zhu et al. revealed that HFNC contributed to a reduction in post-extubation respiratory failure compared to that observed with COT, reductions in reintubation and mortality rates were not apparent [17]. In the study by Zhu et al. and our NMA, the effect of HFNC differed in terms of the reintubation rate; however, this difference is likely attributable to the eligibility of included patients. We excluded RCTs that included >50% of postoperative patients, whereas the study by Zhu et al. included all RCTs with postoperative patients (three RCTs; n=715) [47]. In postoperative abdominal surgery patients, diaphragmatic dysfunction and decreased lung vital capacity can cause atelectasis, resulting in hypoxemic respiratory failure [48]. Including patients with such different mechanisms of respiratory failure may increase patient heterogeneity and result in different outcomes compared to those observed with HFNC and COT.
Zhou et al. recently reported a systematic review using NMA that compared NPPV, HFNC, and COT in post-extubation patients [49], but their inclusion criteria differ from ours. Zhou et al. included all studies with COPD patients, whereas we excluded studies with >50% COPD patients. Moreover, Zhou et al. showed that NPPV was associated with reductions in mortality and post-extubation respiratory failure rates compared to COT. COPD is a risk factor for reintubation after extubation and predisposes patients to hypercapnia during SBT [46]. Thus, NPPV is more effective than COT for patients with hypercapnia after extubation [50], which possibly led to differences in results between our study and that of Zhou et al. Furthermore, including trials with many COPD patients potentially increases the patient heterogeneity. Therefore, we excluded trials where COPD patients accounted for >50% of the study population. This study utilized a four-step approach for assessing the certainty of the NMA estimate developed by the GRADE Working Group [51], whereas the study by Zhou et al. did not conduct a similar assessment. A systematic approach using the GRADE system is necessary for evaluating the quality of the evidence to assess whether the evidence is convincing or of low quality, thereby guiding subsequent decision-making.

Implications
The results of our systematic review are useful for selecting an appropriate noninvasive oxygenation strategy for post-extubation patients because the use of NPPV or HFNC will prevent reintubation in a greater proportion of patients (66-69 patients per 1,000) compared to that of COT. Early weaning from IMV improves patient mortality, whereas reintubation signi cantly increases mortality risk [3]. Therefore, it is important to choose an appropriate strategy to prevent reintubation after extubation. Both NPPV and HFNC are associated with a lower reintubation rate than COT; therefore, physicians can choose a strategy according to the patient's respiratory physiology status and wishes.

Limitations
Our systematic review using NMA has several limitations. First, we combined studies that included patients with different etiological conditions necessitating intubation, which may have increased the heterogeneity of the studies. Despite excluding RCTs with >50% of patients with postoperative intubation and COPD, the inclusion of a xed number of postoperative and COPD patients may have in uenced the results. Second, we combined studies with different degrees of respiratory failure during the extubation of patients. The effect of NPPV and HFNC may differ depending on illness severity, and differences in severity may be an effect modi er. This NMA included other RCTs, with different characteristics, such as duration of intubation, risk factors for reintubation after extubation, and methods of SBT, which may also be effect modi ers. Third, because only one RCT directly compared NPPV and HFNC, there may not have been a signi cant difference due to insu cient sample size; there were no signi cant differences in mortality or post-extubation respiratory failure rates, but this may have been different if the sample size was larger. There was incoherence between direct and indirect estimation in the pairwise comparison of NPPV and HFNC, which led to a grading down of network estimation due to the lack of RCTs that directly compared NPPV and HFNC.

Conclusion
In conclusion, noninvasive respiratory support strategies may be associated with a lower reintubation rate compared with COT, but not shortterm mortality risk. Availability of data and materials: The datasets generated during and/or analyzed during the current study are not publicly available due to post-hoc analyses by co-authors but are available from the corresponding author on reasonable request.

Competing interests:
The authors declare that they have no competing interests.

Funding:
There was no funding source for this study.
Authors' contributions HY participated in the study design, data acquisition, statistical analyses and interpretation of results, and manuscript writing. HO, TM, and YO contributed to the study conception and participated in the data acquisition, interpretation of results, and manuscript writing. CN contributed to the study conception and participated in the study design, data acquisition, interpretation of results, and manuscript writing. MN and NS contributed to the study conception and participated in the study design, interpretation of results, and manuscript writing.
There is a discrepancy between the ndings of this review and those indicated by PROSPERO data because the update of PROSPERO was not completed in time for the submission of this review. All authors in this review meet the International Committee of Medical Journal Editors  Tables   Table 1 Study populations, protocols, and study characteristics