Our systematic review identified 22 RCTs, overall including 2225 patients, that evaluated the association of higher PEEP, compared to lower PEEP, with hospital mortality in unselected adult ICU patients intubated and mechanically ventilated for reasons other than ARDS. We concluded that hospital mortality was similar between the two groups. However, higher PEEP was associated with higher PaO2/FiO2 ratio and lower AaDO2, higher respiratory system compliance, and lower risk of hypoxemia and ARDS occurrence. Other secondary outcomes (e.g., occurrence of atelectasis, barotrauma, ventilation-associated pneumonia, hypotension, duration of ventilation, ICU and hospital stay, and ICU mortality) were similar between the two groups. The overall certainty of evidence for hospital mortality was low, whereas it was judged to be low or moderate for most secondary outcomes.
The role of PEEP during MV in patients with ARDS has been a matter of intense research. Based primarily on the results of an individual patient data meta-analysis reporting lower hospital mortality with higher PEEP among patients with moderate-severe ARDS [8], higher PEEP was suggested with a conditional recommendation in the most recent guidelines on MV in patients with ARDS [46].
The net benefit or harm of PEEP depends on lung recruitability, i.e., the reaeration of non-aerated and poorly aerated lung tissue [5]. Patients without ARDS may exhibit lower lung recruitability than patients with ARDS and develop end-inspiratory alveolar overdistention, increased intrapulmonary shunt and dead space, and higher right ventricular afterload with higher PEEP [12]. These patients may benefit from a MV strategy including lower TVs, which were associated with improved clinical outcomes [43–45]. However, the effect of PEEP in patients without ARDS is still not well understood. To our knowledge, this is the second systematic review and meta-analysis to compare different levels of PEEP in ICU patients without ARDS. Similar to our findings, a previous work concluded that ventilation with higher PEEP is not associated with lower hospital mortality or shorter duration of ventilation, but with higher PaO2/FiO2 and lower occurrence of ARDS and hypoxemia [14]. We believe our work has some strengths compared to the previously published meta-analysis. Three studies were not confirmed for inclusion in our work because of non-invasive application of PEEP [46], crossover design [47], and differential utilization of recruitment maneuvers in the two groups [48]. On the other hand, five additional studies were selected [13, 25, 27, 29, 30]. Among these, the recent RELAx trial found that the use of lower PEEP was non-inferior to higher PEEP in terms of 28-day ventilator-free days [13]. Additionally, no difference between the groups in the occurrence of pulmonary complications, lengths of stay, and mortality was reported [13]. This trial is by far the largest RCT on this topic, accounting for more than 40% of patients included in our systematic review.
Despite the higher number of included patients and stricter selection criteria than the previous meta-analysis [14], we did not identify any association of higher PEEP with clinical outcomes, except for ARDS occurrence, but confirmed its association with physiological outcomes such as oxygenation. Several reasons may account for this.
First, PEEP may not be the most important target for preventing VILI in patients without ARDS. Driving pressure (DP), i.e., the difference between plateau pressure and total PEEP [49], and mechanical power (MP), i.e., the total inflation energy transferred from the mechanical ventilator to the lungs [50], have been recently proposed as reliable mediators of the injurious effects of MV in patients with ARDS [51–54]. Notably, patients without ARDS may benefit from lower DP [55–57] and MP [58], although the evidence is still controversial [59, 60]. Moreover, TVs may be a better surrogate of the risk of VILI in these patients with near-normal respiratory system compliance [61]. Therefore, titrating MV settings on DP and MP rather than PEEP or monitoring DP and MP during PEEP titration may be more effective in minimizing the risk of VILI in non-ARDS patients.
Second, despite the overall low-moderate statistical heterogeneity, clinical heterogeneity among patient populations and outcome definitions (e.g., ARDS, barotrauma, pneumonia) was relevant in the studies included in our meta-analysis. This may have confounded the association between the level of PEEP and clinical outcomes.
Third, the choice of PEEP in most studies was made arbitrarily and not individualized according to the patients’ response to PEEP or the assessment of lung recruitability. Although the improved oxygenation with higher PEEP could depend on alveolar recruitment and reduced intrapulmonary shunt, which may protect against VILI, the use of excessive PEEP could lead to alveolar overdistension, cardiac dysfunction, and reduced oxygen delivery in patients with low lung recruitability regardless of arterial oxygenation [62]. Indeed, oxygenation may not be the best physiologic endpoint as regards the effect of PEEP. In fact, respiratory system compliance or DP may be better indices of the balance between recruitment and overdistension associated with variations of PEEP. Unfortunately, no studies provided data on DP and only 3 studies (189 patients) included compliance among the outcomes, thus limiting our confidence in our results. Furthermore, hypotension and barotrauma occurrence and cardiac index measurements were only reported in a few studies. The trial sequential analysis showed that, although further studies are needed, these are unlikely to show any association of PEEP with mortality; this may suggest that patient selection criteria need to be modified in future studies, for example taking lung recruitability into account.
Finally, physiological benefits may be most likely to translate into clinical advantages when baseline physiological derangements are severe enough. However, baseline oxygenation and compliance, when reported, were near-normal for most patients [13, 24, 28–32, 34, 37, 38, 39].
Our study has several limitations. First, we restricted our research to 6 databases and decided not to include conference proceedings. Therefore, we cannot exclude that other studies could have been identified by exploring other databases and that publication bias may have ensued from our search strategy. In addition, we did not perform an individual patient data meta-analysis, thus potentially missing patient subgroups benefiting from higher or lower PEEP. However, we used a reproducible and comprehensive literature search strategy, including clinical trials and grey literature, and we employed clearly defined inclusion criteria and duplicate independent citation review, data extraction, and quality assessment. Second, as the included studies were published over a 45-year period, the heterogeneous clinical protocols and outcome definitions may limit the reliability of our findings. Importantly, very different levels of PEEP and TV were applied in the studies. However, we excluded studies changing MV settings other than PEEP to avoid potential confounding factors. Additionally, our subgroup analyses on the use of ZEEP vs. PEEP, TV greater vs. lower than 8 mL/kg, and publication year after vs. before 2000 did not identify any clinically significant variation in the intervention effect. Finally, statistical heterogeneity was relevant for some secondary outcomes. Notwithstanding, we performed several subgroup analyses and a post-hoc meta-regression, which partially explained the reported heterogeneity.