Optimal Positive End Expiratory Pressure Levels in Ventilated Patients Without Acute Respiratory Distress Syndrome: A Bayesian Network Meta-Analysis and Systematic Review of Randomized Controlled Trials

Background: To find the optimal positive end expiratory pressure (PEEP) in mechanical ventilated patients without Acute Respiratory Distress Syndrome (ARDS), we conducted a Bayesian network meta-analysis and systematic review of randomized controlled trials (RCTs) comparing different level of PEEP based on a novel classification of PEEP level: ZEEP group (PEEP = 0 cm H2O); lower PEEP group (PEEP = 1–6 cm H2O); intermediate PEEP group (PEEP = 7–10 cm H2O); higher PEEP group (PEEP > 10 cm H2O). Result: Twenty eight eligible studies with 2,712 patients were included. There were no significant differences in the duration of mechanical ventilation between higher and intermediate PEEP (MD: 0.020, 95% CI: −0.14, 0.28), higher and lower PEEP (MD: −0.010, 95% CI: −0.23, 0.22), higher PEEP and ZEEP (MD: 0.010, 95% CI: −0.40, 0.22), intermediate and lower PEEP (MD: −0.040, 95% CI: −0.18, 0.040), intermediate PEEP and ZEEP (MD: −0.010, 95% CI: −0.42, 0.10), lower PEEP and ZEEP (MD: 0.020, 95% CI: −0.32, 0.13), respectively. Higher PEEP was associated with significantly higher PaO2/FiO2 ratio(PFR) when compared to ZEEP (MD: 73.24, 95% CI: 11.03, 130.7), and higher incidence of pneumothorax when compared to intermediate PEEP, lower PEEP and ZEEP (OR: 2.91e + 12, 95% CI: 40.3, 1.76e + 39; OR: 1.85e + 12, 95% CI: 29.2, 1.18e + 39; and OR: 1.44e + 12, 95% CI: 16.9, 8.70e + 38, respectively). There was no association between PEEP levels and other secondary outcomes. Conclusion: We identified higher PEEP was associated with significantly higher PFR and higher incidence of pneumothorax. Nonetheless, in terms of other outcomes, no significant differences were detected among four levels of PEEP. Systematic Review Registration: The study had registered on an international prospective register of systematic reviews, PROSPERO, on 09 April 2021, identifier: [CRD42021241745].


INTRODUCTION
Although invasive mechanical ventilation is a lifesaving strategy for critically ill patients, previous studies have considered it a potentially harmful intervention (1,2). Positive end expiratory pressure (PEEP) has shown efficacy in maintaining alveoli opening, improvement of gas exchange and reduction of injurious shear forces in acute respiratory distress syndrome (ARDS) patients since 1960 s (3). To date, however, the optimal PEEP levels remain unclear, owing to occurrence of potential negative effects that cause overdistention of the lungs, exacerbate lung stress as well as strain and impair hemodynamics by reducing venous return and increasing pulmonary vascular resistance. Therefore, PEEP's net benefits or harm are depended on the balance between alveolar recruitment and overdistension, and should be particularly beneficial in disease states with substantial alveolar collapse (4). Nevertheless, this trade-off is often difficult to achieve clinically. Similarly, the optimal PEEP level for mechanical ventilated patients without ARDS remains unclear. Several studies have demonstrated that higher PEEP levels could improve oxygenation, reduce occurrence of ventilator-associated pneumonia (VAP), prevent ARDS in this population (5). In fact, application of PEEP has increased in clinical practice (6). However, PEEP level in a relatively healthy lung is expected to be lower because of less lung collapse which requires less pressure to open the collapsed lung. In addition, previous research evidences from animal studies have shown that ventilation with higher PEEP levels might worsen existing lung injuries or cause development of new ones (7)(8)(9). A recent RELAx trial demonstrated that a higher PEEP strategy generated clinically superior outcomes than lower levels with regards to the number of ventilator-free days (VFD) at day 28 in ventilated patients without ARDS, although there was a possibility of elevated hypoxemia in the lower PEEP group (10).
A previous systematic review and meta-analysis compared efficacy of different PEEP levels in patients without ARDS (11). However, the findings therein should be interpreted with caution, owing to a moderate to high heterogeneity, a low to very low quality of evidences (QoE), and the fact that the included studies could not allow the authors to comprehensively address the effects of moderate PEEP levels. In the present study, we conducted a Bayesian network meta-analysis and systematic review of RCTs to compare efficacy of different PEEP levels in ventilated patients without ARDS, and identify the optimal level for this population. Specifically, we divided the patients into four groups according to their PEEP levels. We chose a novel classification, based on patients' PEEP levels, which is closer to clinical practice, and can allow for reduction of heterogeneity in the analysis as well as precise evaluation of the effects of different PEEP levels.

MATERIALS AND METHODS
This meta-analysis was performed in accordance with the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-analyses extension statement for reporting network meta-analyses (PRISMA-NMA) (12). The study was also prospectively registered on PROSPERO database (Registration number: CRD42021241745).

Data Sources and Study Search
We searched PubMed, Web of Science, Embase, Cochrane Library, Embase up to January 2021. Reference lists of relevant articles were also reviewed. The inclusion criteria were as follows: (i) studies were RCTs; (ii) the study population comprised ventilated patients without ARDS; (iii) intervention included higher vs. lower PEEP; and (iv) studies were published in English. The exclusion criteria were as follows: (i) studies that analyzed pediatric patients; (ii) patients were not in ICU; (iii) data were unavailable; and (iv) duplicate publications.

Study Selection and Data Extraction
Meta-analysis was performed by two researchers (JZ and ZML), who independently screened the citations and abstracts in duplicate and extracted the data. All references that were judged potentially relevant were evaluated for full-text eligibility. Discrepancies were resolved by consensus with a third author (YBH). In cases where relevant data or information was missing, we attempted to contact the authors of the studies.

Outcome Measures
Primary outcome was the duration of mechanical ventilation, whereas secondary outcomes included PaO 2 /FiO 2 ratio (PFR), length of stay (LOS) in ICU, LOS in hospital, hospital mortality, 28-day mortality, ICU mortality, occurrence of ARDS, pneumothorax, atelectasis and hypoxemia.

Assessment of Risk of Bias
Two authors (JZ and ZML) independently assessed the risk of bias (RoB) in individual studies, using the revised Cochrane risk-of-bias tool for randomized trials (13), and classified them as either low or high. Any disagreements between them were resolved by discussion and consensus with a third author (YBH). Low-biased studies were defined as those with no <4 low-risk items, based on the Cochrane risk-of-bias tool.

Statistical Analysis
A random effects network meta-analysis was performed using a Bayesian framework. We also calculated mean differences for continuous outcomes and odds ratios (ORs) for dichotomous outcomes, then converted medians and interquartile ranges to means and standard deviations as previously described (14). Network meta-analysis was performed using the "gemtc" package (version 0.8-2) implemented in R version 3.4.4 (https://www.rproject.org/). This package is based on an approach that follows the graph-theoretical methodology. We ranked the treatments using the P-score to reveal the degree of certainty that a specific treatment was better than the others. Based on this, Pscores close to 1 and 0 denoted the best and worst treatments, respectively. Moreover, studies followed by a value of I 2 ≥ 50% were considered to have substantial heterogeneity. To limit the possibility of type I error, we performed a Trial sequential analysis (TSA) using TSA version 0.9.5.10.

Primary Outcomes
A total of 11 eligible articles (5,10,16,19,21,23,25,29,33,35,38), with 1,848 participants, reported duration of mechanical ventilation. Among them, 6 studies (16,19,21,23,25,29), with 572 patients (16,19,21,23,25,29). A summary of the RoBs is shown in Figure 3 while the resulting funnel plot is illustrated in  from both direct and indirect comparisons across the three groups were highly consistent ( Figure 5C). However, results from ranking analysis showed that intermediate PEEP levels could shorten the duration of mechanical ventilation, followed by ZEEP, higher PEEP and lower PEEP (Figure 6). Furthermore, TSA showed that conventional and O'Brien-Fleming significance boundaries were not crossed by the cumulative Z-curve, indicating that the evidence was not sufficient and conclusive. Therefore, further trials are needed to validate these findings. A graphical representation of this analysis is shown in Figure 7.

Secondary Outcomes
Eleven eligible studies, with 1,648 patients, reported on PFR (5, 10, 15, 20-22, 24, 25, 32-34), with 6 of them (comprising 347 patients) focusing on post-cardiac surgery patients (5,10,21,22,24,33). Results of RoB are shown in Figure 3 and Supplementary Figure 1A. Direct comparison revealed no significant differences in PFR among PEEP levels, in both general or post cardiac surgical patients (Supplementary Figure 1B)  sigificant differences based on the other comparisons (Figure 8). Moreover, node-splitting analysis, based on both direct and indirect comparisons in these groups, revealed consistent results (all p > 0.05) (Supplementary Figure 1C). Ranking analysis showed that higher PEEP was associated with the best PFR, followed by intermediate and lower PEEP, and lastly ZEEP (Supplementary Figure 1D).

DISCUSSION
We employed a Bayesian network meta-analysis to compare ZEEP levels in 28 RCTs (with 2,709 patients) that focused on ventilated patients without ARDS. Our results revealed that intermediate PEEP levels were associated with the highest decrease in the duration of mechanical ventilation, although there were no significant differences among PEEP levels based on direct and indirect comparisons. Meanwhile, higher PEEP levels were associated with significantly higher PFR and increased incidence of pneumothorax. Notably, we found no significant differences among the PEEP groups with regards to LOS of hospital and ICU, hospital, 28-day and ICU mortalities, occurrence of ARDS, atelectasis and hypoxemia. However, our results should be interpreted cautiously, owing to the TSA outcomes and presence of heterogeneity.
One meta-analysis published in 2016 (11) demonstrated that ventilation with higher PEEP levels in ICU patients without ARDS was not associated with neither reduced in-hospital mortality nor shorter ventilation duration, but with lower incidence of ARDS and hypoxemia, as well as higher PaO 2 /FiO 2 . Notably, the study had a moderate to high heterogeneity, while its quality of evidence was low to very low. Consequently, the authors could not address the effects of moderate levels of PEEP (11). When compared to the aforementioned metaanalysis, our study had several strengths. Firstly, we included seven recent studies, which included one large RCT describing use of high PEEP in patients without ARDS. The lower and higher PEEP groups in the former study corresponded to low and intermediate PEEP groups, respectively, in our study (10). This could also explain why our results were not completely consistent with previous meta-analyses. Secondly, we employed a novel  classification, and divided the patients into four groups according to the specific PEEP levels. The ZEEP and very high PEEP (>10 cm H 2 O) groups are not routine choices across clinical practice for non-ARDS patients, and these 2 extreme PEEP levels have always been applied in post cardiac surgery patients in our included studies. Moreover, since most of these studies were published 20 years ago, our novel classification allowed us to address the effects of moderate PEEP levels closer to clinical practice. Thirdly, a previous meta-analysis reported PEEP levels that ranged from 0 to10 cm H 2 O for the low group, and 5 to 30 cm H 2 O for the high group, while the heterogeneity was so large that the authors could not make a definite conclusion. Our novel classification solved this problem to a certain extent, and made the conclusion more credible.
Although previous studies have demonstrated the potential benefits and adverse effects of PEEP in ARDS, selecting appropriate PEEP levels seems to be a complex process in patients without ARDS owing to a huge heterogeneity in this population. Although an increase in PEEP levels has been reported in such population in the real-world, evidence of how to choose an optimal concentration was lacking (6,41). In our study, Bayesian analysis revealed that intermediate PEEP (PEEP = 7-10 cm H 2 O) was associated with shorter duration of mechanical ventilation, whereas network meta-analysis found no significant differences among the studied PEEP levels, which was partially in line with the RELAx trial (10). Interestingly, one study demonstrated that a higher PEEP could reduce the duration of mechanical ventilation (16), was although this corresponded to the intermediate PEEP group in our study. To our knowledge, there were many confounding factors that affected the duration of mechanical ventilation, affirming PEEP's lack of significant impact observed herein.
Our results further showed that PFR was positively correlated with PEEP levels, which was consistent with a previous metaanalysis (11). In ARDS patients, PEEP has been shown to recruit the collapse alveoli, maintain the end expiratory lung volume and improve gas exchange (3). Interestingly, the same principle seems to work in patients without ARDS. On the other hand, inadequate elevated PEEP has been found to cause alveoli overdistension in ARDS patients, thereby causing barotrauma (42,43). In our opinion, this challenge might be even more pronounced in non-ARDS patients as the collapse alveoli in these patients might be less than those in ARDS patients. This explains why higher PEEP levels were associated with significantly increased incidence of pneumothorax relative to the other PEEP levels in our study. Although the meta-analysis published in 2016 demonstrated that high PEEP was associated with a lower incidence of ARDS and hypoxemia (11), we found no evidence to support this finding.
Although our findings provide evidence of the potential benefits or harmful effects of different PEEP levels, PEEP should not just be applied according to its height, as many physiologic effects of PEEP could be U-shaped (44,45), Individualized PEEP regimes should be optimized based on a specific patient's physiology rather than focusing simply on the dosage. To date, however, no trial has attempted to evaluate the efficacy of PEEP in patients without ARDS prior to randomization (4), which necessitates future trials.

CONCLUSION
In summary, results of our Bayesian network meta-analysis and systematic review revealed that intermediate PEEP levels are associated with the highest decrease in duration of mechanical ventilation in patients without ARDS. However, there were no significant differences among studied PEEP level groups based on both direct and indirect comparisons. Meanwhile, it is evident that higher PEEP levels are associated with significantly higher PFR and increased incidence of pneumothorax. Furthermore, the four studied PEEP levels have no significant impact on LOS of hospital, LOS of ICU, hospital mortality, 28-day mortality, ICU mortality, occurrence of ARDS, as well as atelectasis and hypoxemia.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.