High-Flow Nasal Oxygen versus Noninvasive Ventilation in Acute Exacerbation of Chronic Obstructive Pulmonary Disease Patients: A Meta-Analysis of Randomized Controlled Trials

Background High-flow nasal cannula (HFNC) can be used in stable chronic obstructive pulmonary disease (COPD) patients, but the effect of HFNC on clinical outcomes in patients with acute exacerbation of chronic obstructive pulmonary disease (AECOPD) is still uncertain. Methods We searched electronic literature databases for randomized controlled trials (RCTs) comparing HFNC with noninvasive ventilation (NIV) in hypercapnic patients with AECOPD. The primary endpoint of this meta-analysis was PaCO2, PaO2, and SpO2. The secondary outcomes were the respiratory rate, mortality, complications, and intubation rate. Results We included 7 RCTs with a total of 481 patients. There were no significant differences on measures of PaCO2 (MD = −0.42, 95%CI −3.60 to 2.75, Z = 0.26, and P = 0.79), PaO2 (MD = −1.36, 95%CI −4.69 to 1.97, Z = 0.80, and P = 0.42), and SpO2 (MD = −0.78, 95%CI −1.67 to 0.11, Z = 1.72, P = 0.08) between the HFNC group and the NIV group. There was no significant difference in measures of the mortality and intubation rate between the HFNC group (OR = 0.72, 95%CI 0.30 to 1.69, Z = 0.76, and P = 0.44) and the NIV group (OR = 2.38, 95%CI 0.49 to 11.50, Z = 1.08, and P = 0.28), respectively. But the respiratory rate in the HFNC group was lower than that in the NIV group (MD = −1.13, 95%CI −2.13 to −0.14, Z = 2.23, and P = 0.03), and fewer complications were found in the HFNC group (OR = 0.26, 95%CI 0.14 to 0.47, Z = 4.46, and P < 0.00001). Conclusion NIV was noninferior to HFNC in decreasing PaCO2 and increasing PaO2 and SpO2. Similarly, the mortality and intubation rate was similar among the two groups. The respiratory rate and complications were inferior in the AECOPD group treated with HFNC.


Introduction
Te high-fow nasal cannula (HFNC) is a high-concentration oxygen technique that reliably achieves a FiO 2 as high as 100%. HFNC is more likely to beneft patients with severe symptoms who need high-concentration oxygen rather than patients requiring low oxygen fow rates [1]. So, HFNC is more frequently used in severe acute respiratory failure patients who are at high risk of intubation. But for hypercapnic respiratory failure secondary to acute exacerbation of chronic obstructive pulmonary disease (AECOPD), the standard of care is noninvasive ventilation (NIV) [2]. In some patients, NIV is sometimes difcult to tolerate. Especially younger patients with breathing frequency and a high heart rate may more frequently experience NIV intolerance [3]. Tere is a need for providing another treatment option that is both easy to administer and well tolerated for hypercapnic chronic obstructive pulmonary disease (COPD) patients. Te benefts of HFNC include providing oxygen at high fows with an optimal degree of heat, carbon dioxide washout of upper airways by providing airfows 30−60 L/min, and enhanced secretion clearance through the provision of reliable humidifcation [4]. In one investigation with COPD patients treated for home HFNC in a crossover clinical trial, HFNC with the titration of 20 L/min was not inferior to NIV in COPD patients with stable hypercapnia for a lower CO 2 clearance [5]. A meta-analysis in stable hypercapnic COPD patients comparing HFNC to conventional oxygen demonstrated that the addition of HFNC did not increase the arterial partial pressure of carbon dioxide (PaCO 2 ) in these patients [6]. However, the efect of HFNC on clinical outcomes in patients with AECOPD with hypercapnia is still uncertain. We performed a meta-analysis to compare the clinical outcomes for HFNC with NIV in the AECOPD patients.

Search Strategies.
We search electronic literature databases for randomized controlled trials (RCTs) comparing HFNC with NIV in hypercapnic patients with AECOPD. A literature search was performed through the following databases: Cochrane, Google Scholar, MEDLINE database, PubMed, and Embase from inception to July 2022. Te following search terms were used: high-fow nasal cannula or nasal high-fow therapy or high-fow nasal cannula or highfow oxygen through nasal cannula or high-fow nasal therapy, non-invasive ventilation or noninvasive ventilation or noninvasive positive pressure ventilation, acute exacerbation of obstructive pulmonary disease or COPD patients with chronic respiratory failure or hypercapnic respiratory failure or chronic obstructive pulmonary disease patients with hypercapnia or COPD exacerbation, and randomized controlled trial or randomized clinical trial.

Data Extraction.
Two reviewers independently evaluated the included studies and extracted data into RevMan 5.3 (review manager: Cochran handbook for systematic reviews). Any disagreement about whether the RCT met the inclusion or exclusion criteria between the two reviewers was resolved by discussing it with a third reviewer. If still more information was required, communication with the authors through Email would be carried out.

Study Selection.
We included RCTs comparing HFNC with NIV in AECOPD patients. Te inclusion criteria included (1) randomized control trials, (2) human studies, (3) the comparison between HFNC with NIV in hypercapnic patients with AECOPD was performed in the study, (4) all patients were adults, and (5) if more than one eligible study using the same protocol from the same centre, the study with the longest follow-up was used. Te exclusion criteria were as follows: (1) studies reported none of these outcomes: PaCO 2 , PaO 2 , SpO 2 , respiratory rate, mortality, complications, and intubation rate. (2) We excluded studies with the baseline pH < 7.30, for most of these patients need invasive ventilation.

Outcome Measure.
Te primary endpoint of this metaanalysis was PaCO 2 , PaO 2 , and SpO 2 . Te secondary outcomes were the respiratory rate, mortality, complications, and intubation rate.

Quality Assessment.
We used the Cochran handbook for systematic reviews of intervention guidelines to assess the risk of bias. Each study was evaluated for random sequence generation, concealment of allocation sequence, blinding of the participants and personnel, blinding of outcome assessment, incomplete outcome, and selective reporting. Also, they were classifed by two authors as having a high risk of bias and unclear risk of bias or a low risk of bias based on the Cochrane tool.
2.6. Statistical Analysis. Statistical analysis of the metaanalysis was done using Cochrane systemic review software RevMan 5.3. We used the Mann-Whitney U test to help us verify the hypothesis and rendered statistical signifcance as a P value and a Z value < 0.05. Te odds ratio (OR) and 95% confdence intervals (CIs) were calculated for dichotomous outcomes, and weighted mean diferences (WMDs) and 95% confdence intervals (CI) were calculated for continuous outcomes in each included study. Te I 2 value was used to assess statistical heterogeneity. If the I 2 value ≤ 50% was considered as having no statistical heterogeneity, a fxed efects model was used to estimate the overall summary efect sizes. Otherwise, we used a random efects model. And subgroup analysis or sensitivity analysis would be carried out.

Study Selection.
Te search algorithm identifed 104 records. 92 records were identifed from electronic databases and 12 records from reference lists. After deduplication, 26 records were excluded. 78 records were screened. 46 records were excluded by reading the abstracts for not about acute exacerbation of COPD (n � 18), not RCT (n � 11), nonhuman studies (n � 5), and retrospective studies (n � 12). 32 full-text articles were assessed for eligibility. 25 full-text articles were excluded for reasons. 9 articles were excluded for outcomes that have not met this review, 10 articles were excluded for lack of essential data, and 6 articles were excluded for not adults. 7 articles were included in the fnal meta-analysis [7][8][9][10][11][12][13] Figure 1.

Included
Studies. We included 7 RCTs with a total of 481 patients. All included studies had been published and is shown in Table 1.

Quality Assessment.
Te risk of bias about the methodological quality of the included studies are elaborated and summarized, respectively, in Figures 2 and 3. Due to the nature of respiratory support management, blinding the participants is not possible.

Te Primary Endpoint
(1) PaCO 2 . Te primary endpoint is PaCO 2 , PaO 2, and SpO 2 . "PaCO 2 " was reported in fve studies. 170 patients in the HFNC group and 177 patients in the NIV group were available to compare the PaCO 2 . Tere was no signifcant diference in measures of PaCO 2 between the HFNC group and the NIV group (MD � −0.42, 95%CI −3.60 to 2.75, Z � 0.26, and P � 0.79), Figure 4.
(2) PaO 2 . Te "PaO 2 " was reported in three studies. 124 patients in the HFNC group and 128 patients in the NIV group were available to compare the PaO 2 . Tere was no signifcant diference in measures of PaO 2 between the HFNC group and the NIV group (MD � −1.36, 95%CI −4.69 to 1.97, Z � 0.80, and P � 0.42), Figure 5.
(3) SpO 2 . Te " SpO 2 " was reported in three studies. 128 patients in the HFNC group and 128 patients in the NIV group were available to compare the PaO 2 . Tere was no signifcant diference in measures of SpO 2 between the HFNC group and the NIV group (MD � −0.78, 95%CI −1.67 to 0.11, Z � 1.72, and P � 0.08), Figure 6.

Te Secondary Endpoint
(1) Respiratory rate. Te second endpoint contains three outcomes: respiratory rate, mortality, and complications. First, the respiratory rate was reported in four studies. 110     We have no registration number for our study is not a clinical study but a meta-analysis.
patients in the HFNC group and 117 patients in the NIV group were available to compare the respiratory rate. Te respiratory rate in the HFNC group was lower than that in the NIV group (MD � −1.13, 95%CI −2.13 to −0.14, Z � 2.23, and P � 0.03), Figure 7.
(2) Mortality. Second, Mortality was reported in three studies. 86 patients in the HFNC group and 93 patients in the NIV group were available to compare the respiratory rate.
(3) Complications. Tird, complications were reported in three studies. 148 patients in the HFNC group and 150 patients in the NIV group were available to compare the respiratory rate. Fewer complications were found in the NIV group (OR � 0.26, 95%CI 0.14 to 0.47, Z � 4.46, and P < 0.00001), Figure 9.
(4) Intubation rates. Fourth, intubation rates were reported in three studies. 104 patients in the HFNC group and 105 patients in the NIV group were available to compare the respiratory rate. Fewer complications were found in the NIV group (OR � 2.38, 95%CI 0.49 to 11.50, Z � 1.08, and P � 0.28), Figure 10.

Discussion
Te major fnding in our meta-analysis was that HFNC is not inferior to NIV (which is the standard of treatment for acute decompensated hypercapnic respiratory failure) in decreasing PaCO 2 and increasing PaO 2 and SpO 2 . Similarly, the mortality and intubation rate was similar among the two groups. Te respiratory rate and complications were inferior in the AECOPD group treated with HFNC. NIV is the standard treatment for hypercapnic respiratory acidosis patients according to current guidelines and can signifcantly reduce mortality and the need for intubation among AECOPD patients with respiratory acidosis. Tis beneft appears similar for patients with a mild (pH 7.30 to 7.35) and a more severe nature (pH < 7.30) [14]. NIV corrects the mechanism leading to hypercapnia by increasing the tidal volume and by decreasing the work of breathing while reducing CO 2 [15]. Many studies indicated that the early use of NIV in mild AECOPD patients with a partial pressure of carbon dioxide (PaCO 2 ) >45 mmHg and a pH > 7.25 can efectively alleviate respiratory muscle fatigue. However, intolerability and discomfort have restricted the widespread application of NIV in AECOPD patients [16].
HFNC oxygen therapy is used in hypoxemic respiratory failure patients. But some studies have explored that different fow rates of HFNC resulted in a fow-dependent reduction in PaCO 2 in stable hypercapnic COPD patients [17]. For safety and efcacy, HFNC has not been demonstrated in AECOPD patients. When compared to NIV, HFNC also demonstrated a reduction of the inspiratory muscle efort similar to spontaneous breathing [18]. Recent studies found that the short-term (within 2 h) application of HFNC could efectively decrease PaCO 2 . Several mechanisms are involved in the explanation of these results, such as the reduction of the anatomical dead space in the upper airways and inspiratory resistance, which improves alveolar ventilation [19]. Te adequate fow and humidifed warm gas can attenuate inspiratory resistance and increase expiratory resistance can also reduce the physiological dead space, improve airway clearance, and attenuate the work of breathing, allowing for a higher fraction of minute ventilation to facilitate gas exchange [1]. At the same time, HFNC also increases the tidal volume to a lesser extent than NIV. HFNC improves the washout of the upper airway dead space and generates a low level of positive end-expiratory pressure  (PEEP) [20]. Other studies found HFNC led to a fowdependent reduction in PaCO 2 , accompanied by an increase in the tidal volume [21]. A recent study found HFNC combined with NIV had a higher degree of comfort in patients with AECOPD, can improve blood gas parameters, and increase patient compliance [22].
HFNC may be better tolerated than NIV. A recent study found blood pressure signifcantly decreased after using NIV for NIV had more impact on venous return than HFNC [9]. In our meta-analysis, we found that the respiratory rate in the HFNC group was lower than that in the NIV group. Other studies have explored the physiological efects of HFNC, and HFNC could decrease the neuroventilatory drive and work of breathing in COPD patients [23]. Tis can explain why the respiratory rate in the HFNC group was lower, and the patients felt more comfortable than the NIV group. We also found that fewer complications were found in the HFNC group. A retrospective study found that there were fewer nursing interventions and skin breakdown episodes reported in the HFNC group compare to the NIV group [24]. Some studies found that more patients needed bronchoscopy for secretion management in the NIV group since the patients on NIV feel uncomfortable and claustrophobic within the secured mask. Tis may be due to the     inconvenience of coughing by taking the mask of. By contrast, the humidifying and heating function of HFNC enables the gas delivered to reach a temperature of 37°C and an absolute humidity of 44 mg H 2 O/L, and HFNC can provide optimal humidity, so patients can drink, cough, and talk [25]. For this easy compliance, a subgroup analysis found fewer patients required intubation in the HFNC group than in the NIV group, although this was not a statistically signifcant fnding [26]. But in our meta-analysis, we found complications that were statistically fewer in the HFNC group than in the NIV group.

Conclusion
NIV was noninferior to HFNC in decreasing PaCO 2 and increasing PaO 2 and SpO 2 . Similarly, the mortality and intubation rate was similar among the two groups. Te respiratory rate and complications were inferior in the AECOPD group treated with HFNC.

Data Availability
Te data used to support this study are available within the article.