Effect of oxygenation modalities among patients with postoperative respiratory failure: a pairwise and network meta-analysis of randomized controlled trials

Background Postoperative respiratory failure is associated with increased perioperative complications. Our aim is to compare outcomes between non-invasive ventilation (NIV), high-flow nasal cannula (HFNC), and standard oxygen in patients at high-risk for or with established postoperative respiratory failure. Methods Electronic databases including PubMed, Embase, and the Cochrane Library were reviewed from inception to September 2019. We included only randomized controlled trials (RCTs) that compared NIV, HFNC, and standard oxygen in patients at high risk for or with established postoperative respiratory failure. We performed a Bayesian network meta-analysis to calculate the odds ratio (OR) and Bayesian 95% credible intervals (CrIs). Results Nine RCTs representing 1865 patients were included (the mean age was 61.6 ± 10.2 and 64.4% were males). In comparison with standard oxygen, NIV was associated with a significant reduction in intubation rate (OR 0.23; 95% Cr.I. 0.10–0.46), mortality (OR 0.45; 95% Cr.I. 0.27–0.71), and intensive care unit (ICU)-acquired infections (OR 0.43, 95% Cr.I. 0.25–0.70). Compared to standard oxygen, HFNC was associated with a significant reduction in intubation rate (OR 0.28, 95% Cr.I. 0.08–0.76) and ICU-acquired infections (OR 0.41; 95% Cr.I. 0.20–0.80), but not mortality (OR 0.58; 95% Cr.I. 0.26–1.22). There were no significant differences between HFNC and NIV regarding different outcomes. In a subgroup analysis, we observed a mortality benefit with NIV over standard oxygen in patients undergoing cardiothoracic surgeries but not in abdominal surgeries. Furthermore, in comparison with standard oxygen, NIV and HFNC were associated with lower intubation rates following cardiothoracic surgeries while only NIV reduced the intubation rates following abdominal surgeries. Conclusions Among patients with post-operative respiratory failure, HFNC and NIV were associated with significantly reduced rates of intubation and ICU-acquired infections compared with standard oxygen. Moreover, NIV was associated with reduced mortality in comparison with standard oxygen.


Introduction
Postoperative respiratory failure is associated with increased perioperative complications such as reintubation, invasive mechanical ventilation, and healthcareassociated infections, which can lead to increases in mortality, intensive care unit (ICU) and hospital length of stay, delays in hospital discharges, and higher healthcare resource utilization [1][2][3][4].
Several post-operative pulmonary complications may result in post-operative hypoxemic respiratory failure, including pneumonia, atelectasis, bronchospasm, pneumothorax, and pleural effusion. The incidence of these complications is variable and ranges between 5 and 40% according to the type of surgery, as well as other risk factors including anesthetic technique, duration of surgery, and severity of illness [5][6][7][8][9]. Cardiac surgery has the highest rate of post-operative respiratory complications (up to 40%), followed by thoracic surgery (30%), while abdominal and vascular surgeries have a low incidence of post-operative pulmonary complications (6-7%) [5][6][7].
In nonsurgical patients, oxygenation modalities for hypoxemic respiratory failure are varied. Non-invasive ventilation (NIV) has shown promising results for reducing intubation rates among patients with cardiogenic pulmonary edema and chronic obstructive pulmonary disease exacerbations [10][11][12]. High-flow nasal cannula (HFNC) is a new oxygenation strategy that delivers oxygen at high concentrations and a high flow rate and has been increasingly utilized due to its ease of application, tolerance, and potential clinical benefits [13,14]. Nevertheless, the current European Respiratory Society/ American Thoracic Society (ERS/ATS) guidelines have conditional recommendations regarding the use of NIV in postoperative respiratory failure [15]. In addition, few randomized controlled trials (RCTs) have been conducted to evaluate the efficacy of HFNC vs NIV in postoperative patients. Therefore, we conducted a metaanalysis to compare the efficacy and safety of HFNC, NIV, and standard oxygen therapy in the treatment of patients who developed or were considered high risk for post-operative respiratory failure.

Study design and study selection
Our study is a meta-analysis and systematic review performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocols (PRIS MA-P) 2015 Statement [16]. Two reviewers (M.B., I.G) independently and separately performed a literature search utilizing electronic databases including PubMed, Cochrane Library, and Embase from inception to September 2019 without language restrictions. Articles were first screened by titles and abstracts before exclusion.

Inclusion criteria and study selection
Only RCTs were eligible for inclusion in our analysis. We included studies that compared different oxygenation strategies in patients who developed or were deemed at high risk for developing post-operative hypoxemic respiratory failure. Patients at risk were defined to have intermediate to high risk for development of post-operative pulmonary complications according to either Assess Respiratory Risk in Surgical Patients in Catalonia (ARISCAT) score of ≥ 26 points [4]. Patients who failed spontaneous breathing trial and those who passed spontaneous breathing trials but had risk factors for failed extubation such as cardiac dysfunction, obesity (BMI > 30), or failure of previous extubation were also considered high risk. Post-operative hypoxemic respiratory failure was defined as the development of tachypnea with a respiratory rate of ≥ 25 respirations per minute, intense work of breathing with the use of accessory muscles, hypoxemia (oxygen saturation ≤ 92% or partial arterial oxygen pressure to fraction of inspired oxygen ≤ 300) in the immediate post-operative period or within 7 days post-operatively. We excluded studies that investigated prophylactic use of NIV and HFNC as a routine therapy in the post-operative period.
Data were extracted into a predesigned table independently and separately by two reviewers (L.R and S.S.). Any discrepancies were solved by consensus with a third reviewer (Y.Z.).

Quality assessment
Cochrane Collaboration's tool for assessing risk of bias in randomized controlled trials was used for quality assessment for the included RCTs [17]. Each of the included RCTs was assessed for random sequence generation, allocation concealment, blindness of participants and health-care personnel, blindness of outcome assessment, incomplete outcome data, selective reporting, and other biases if any were present.

Outcomes
Our main outcome was the intubation rate following surgery. Secondary outcomes included mortality at the longest follow-up period provided by each study and ICU-acquired infections.

Statistical analysis
An informative prior Bayesian framework for the network meta-analysis was performed using the Markov Chain Monte Carlo simulation to derive the posterior distribution of the parameter estimates. We used a beta distribution of (0, 2) for binominal likelihood. We used the Brooks-Gelman-Rubin method to assess for convergence. A consistency model which contains treatment as a fixed effect and trial as a random effect was used. Results were reported as odds ratios (ORs) and Bayesian 95% credible intervals (Cr.Is). Inconsistency was assessed using the deviance residuals and deviance information criteria statistics. Sensitivity analysis was performed by including only trials that included patients who had developed respiratory failure. Furthermore, subgroup analysis according to the type of surgery (cardiothoracic or abdominal) was performed. In addition, to show the validity of our results, we performed a direct pairwise meta-analysis for comparisons that have three or more studies comparing directly the two interventions.
Included studies were noted to have inevitable performance bias as blinding of participants and personnel was difficult given the nature of the intervention. Detailed quality assessment was not performed for one study as we only found the abstract with no full article explaining the methods. Supplementary Figure 2 shows the risk of bias in each included RCT based on the authors' judgment.
Sensitivity analysis was performed by including only patients who developed acute hypoxemic respiratory failure (but not patients at increased risk), which showed similar results. In a subgroup analysis for patients undergoing cardiothoracic surgery, both NIV and HFNC were associated with a similar reduction in intubation rates compared with standard oxygen therapy (NIV vs standard oxygen (OR 0.08; 95% Cr.I. 0.03-0.19) and HFN vs standard oxygen (OR 0.08; 95% Cr.I. 0.03-0.21)) ( Fig. 3a). However, in patients undergoing abdominal surgery, NIV (but not HFNC) was associated with significantly reduced intubation rates compared with standard therapy (NIV vs standard oxygen (OR 0.51; 95% Cr.I. 0.26-0.87)) (Fig. 3b).
In an exploratory meta-regression analysis, we found that higher PaCO2 was associated with lower risk for intubation when NIV was compared to standard oxygen therapy (P < 0.05) (Supplementary Figure 3).
In a subgroup analysis based on the type of surgery (cardiothoracic or abdominal), mortality benefit of NIV was limited to those undergoing cardiothoracic surgery compared with standard oxygen therapy (OR 0.31; 95% Cr.I. 0.13-0.70), unlike those undergoing abdominal surgeries (OR 0.56; 95% Cr.I. 0.27-1.08) (Fig. 5).

Direct pairwise meta-analysis results
We have performed direct pairwise meta-analysis comparing NIV versus standard oxygen which showed consistent results of the network meta-analysis (Supplementary Figure 4). However, we did not perform the direct meta-analysis for HFNC vs NIV or HFNC vs standard oxygen because studies that compared directly between these interventions were one or two studies. Supplementary Figures 5 and 6 show the results of these individual studies for different outcomes.

Discussion
In this first network meta-analysis comparing various oxygenation strategies in patients at risk for hypoxemic respiratory failure or established respiratory failure within 7 days of surgery, we have found that NIV and HFNC were associated with a significant reduction in intubation rates and ICU-acquired infections when compared to standard oxygen therapy. However, when compared to standard oxygen therapy, only NIV was found to have a mortality benefit in this patient population. We found HFNC and NIV to have no significant differences in the primary or secondary outcomes. Furthermore, in a subgroup analysis, patients undergoing cardiothoracic surgery had a significantly lower rate of intubation when treated with HFNC or NIV in comparison with standard oxygen therapy, but mortality was significantly lower in patients treated with NIV in comparison with standard oxygen therapy. Additionally, in patients with abdominal surgeries, only NIV was associated with a significant reduction in intubation rates compared to standard oxygen, but there was no significant difference in mortality between competing interventions.
Hypoxemia occurs frequently in the post-operative period and can lead to acute respiratory failure. Several   [27][28][29]. Non-invasive ventilation (NIV) improves oxygenation by recruiting collapsed alveoli and increasing tidal volume participating in gas exchange without hemodynamic adverse events [1,30]. However, previous studies and meta-analyses had not shown a significant reduction in intubation rates with prophylactic use of NIV after surgery, despite the reduction in the incidence of postoperative pulmonary complications [31][32][33]. Currently, NIV is recommended in the treatment of patients with post-operative respiratory failure according to the ERS/ATS guidelines [15]. Our results indicate that intubation rates and mortality are significantly lower in patient populations who are at an increased risk or have developed postoperative respiratory failure treated with NIV in comparison with standard oxygen. In our subgroup analysis, mortality benefit was only noted in patients undergoing cardiothoracic surgeries but not in abdominal surgeries. In an RCT examining NIV vs standard oxygen therapy in patients with respiratory failure after abdominal surgeries, Jaber and Antonelli found that NIV was associated with lower intubation rates, less days on mechanical ventilator, and significantly lower rates of healthcare-associated infections, including pneumonia. Although mortality rates were lower in the NIV group (14 vs 21%), the difference did not reach a statistical significance in their study [22]. Similarly, patients undergoing cardiothoracic surgery and treated with NIV for postoperative respiratory failure had lower rates of intubation and mortality when compared to patients treated with standard oxygen therapy [19,24].
HFNC is a new oxygenation strategy that has been used more frequently in patients with respiratory failure. It is found to be more comfortable than NIV and can deliver concentrated oxygen reaching 100% with a high flow rate up to 60 mL/min [34,35]. Furthermore, it can provide positive end-expiratory pressure up to 2-3mmHG [34,35]. The use of HFNC has shown beneficial effects in patients who developed post-extubation respiratory failure or when used during intubation to prevent hypoxemia when compared to standard oxygen [36][37][38]. Additionally, Frat et al. found lower mortality rates with HFNC in comparison with NIV and conventional oxygen in patients with non-hypercapnic hypoxemic respiratory failure. However, other trials did not find differences between HFNC and standard oxygen therapy [39][40][41][42]. The use of HFNC in the post-operative Data are provided percent (%), mean ± SD, or median (interquartile range) HFNC high-flow nasal cannula, SO standard oxygen, NIV non-invasive ventilation, BMI body mass index, SAPS simplified acute physiology score, PaO2/FiO2 partial pressure of arterial oxygen to fraction of inspired oxygen ratio, PaCO2 partial pressure of arterial carbon dioxide, NA not available period was investigated by several RCTs. In a large RCT involving more than 800 patients after cardiac surgery, the use of HFNC and NIV in the treatment of high-risk patients or those who had developed post-operative respiratory failure was similar between both interventions with similar intubation rates, mortality, and rates of hospital-acquired infections [20]. In our analysis, there was no difference between HFNC and NIV in intubation rates, mortality, and ICU-acquired infections. Similar results were also found in both subgroups (cardiothoracic surgeries and abdominal surgeries). In addition, HFNC was associated with lower intubation rates in patients following cardiothoracic surgeries but not following abdominal surgeries when compared to standard oxygen therapy. This could be explained by the fact that in thoracic surgery, HFNC could minimize lung decruitment post-extubation by providing some level of continuous positive airway pressure through  high-flow ventilation, though this positive pressure can be variable due to the leak around the nasal cannula and nonguaranteed closed mouth of the patients [26]. Furthermore, we found lower rates of infections with HFNC and NIV when compared to standard oxygen are attributed to lower intubation rates in both interventions, which avoids the need for mechanical ventilation and decreases catheter-associated infections.
Although there was no significant difference between HFNC and NIV with regard to rates of intubation, mortality, and ICU-associated infections, when each of these two strategies was compared to standard oxygen, NIV was associated with a survival benefit especially in patients who had cardiothoracic surgery. Additionally, there was a trend toward lower mortality in abdominal surgeries, but HFNC had no mortality benefit in the  total patient population and both subgroups. Whether a lower number of patients included in the comparison between HFNC and standard oxygen or certain other factors could have contributed to the inability to detect a mortality benefit despite a significant reduction of intubation rates is needed to be addressed in further larger and well-controlled trials. Nevertheless, due to the low events, further RCTs are needed to compare between both interventions in different types of surgeries to determine its effect on various long-term clinical outcomes and quality of life and also to examine whether certain patients' risk factors could affect the beneficial effects of these interventions towards reduction of intubation rates and mortality.

Limitations
Our analysis has several limitations. First, we were unable to perform analysis based on various risk factors, duration of surgery, severity scores, and different surgical types as we lack patients' level data. Second, blinding of intervention and personnel was impossible given the nature of intervention. Third, there were few sample size and limited events, and therefore, larger trials and longterm outcomes are needed. Fourth, we used informative prior module for our analysis which could affect the results given the small number of included trials. Fifth, there was a significant time gap between included studies through which there was a significant development in the ICU management, preoperative and postoperative evaluation and care, supportive management, and criteria for admission to the ICU.

Conclusion
Among patients who are at risk for developing postoperative respiratory failure, or have developed postoperative respiratory failure, the use of NIV was associated with reduced rates of intubation, mortality, and ICU-acquired infections in comparison with standard oxygen therapy. In addition, HFNC was associated with reduced rates of intubation and ICU-acquired infections but not mortality in comparison with standard oxygen. There was no significant difference between HFNC and NIV on the various studied clinical outcomes.