Summary of main results
In this meta-analysis, seven studies met the criteria for inclusion.
Of the studies selected, five assessed mortality before hospital discharge (Amato 1998; Brower 2004; Meade 2008; Mercat 2008; Villar 2006) and were of moderate to good quality.
For the main analysis, we decided to exclude studies that applied different tidal volumes between intervention and control arms (Amato 1998; Villar 2006) because lack of clarity as to whether positive results were attributable to a reduction in tidal volume, to higher levels of PEEP, or to both tactics together made conclusions difficult.
The remaining three studies (Brower 2004; Meade 2008; Mercat 2008) assessed mortality before hospital discharge; thus our primary aim was to determine whether use of higher levels of PEEP improved clinical outcome.
High levels of PEEP compared with low levels showed no statistically significant decrease in mortality before hospital discharge in participants with ALI and ARDS (Analysis 1.1; Figure 4). The forest plot, however, showed a trend towards a mortality benefit in the higher-PEEP group. Furthermore, the existence of clinical heterogeneity should be taken into account in considering that outcome because two of the studies (Brower 2004; Mercat 2008) included participants with a PaO2/FIO2 ≤ 300, and Meade 2008 reported on participants with a PaO2/FIO2 ≤ 250.
This outcome is similar to that obtained by Dasenbrook 2011, who found that higher levels of PEEP were not associated with significantly different 28-day mortality. That meta-analysis included four studies; three were the same as in our analysis (i.e. Brower 2004; Meade 2008; Mercat 2008), and in Talmor 2008, a ventilator strategy was used to adjust high levels of PEEP according to the oesophageal pressures of participants with ALI and ARDS. In this review, according to our analysis, clinical heterogeneity is evidenced among the included studies.
The benefit of high levels of PEEP is assessed through a reduction of the collapsed lung and is dependent on the extent of lung recruitment. Gattinoni 2006 demonstrated that use of high levels of PEEP was more beneficial in participants with lower PaO2/FIO2 ratios (ARDS rather than ALI participants) and a high percentage of potentially recruitable lung volume. Furthermore, in a later paper, Caironi 2010 analysed data from that study and provided evidence in favour of the application of high levels of PEEP, especially in participants with great lung recruitability. In addition, Hess 2011 showed that modest levels of PEEP may be more appropriate for participants with ALI (at PaO2/FIO2 ratios ≤ 300), whereas higher levels of PEEP should be used for participants with ARDS (at PaO2/FIO2 ratios ≤ 200). Therefore, the strategy of using high levels of PEEP regardless of the type of patient (i.e. those with ALI or ARDS) could be incorrect.
Of the studies involving participants with ARDS that we analysed, three had assessed mortality in the ICU (Amato 1998; Huh 2009; Villar 2006) and indicated significant differences between groups (RR 0.67, 95% CI, 0.48 to 0.95; Analysis 1.17). In assessing this result, however, we need to consider that in two of those studies (Amato 1998; Villar 2006), high levels of PEEP were part of a ventilatory strategy that also included a low tidal volume. Furthermore, in the assessment of the forest plot, these two studies proved to be of greater benefit. In addition, the meta-analysis of individual-patient data (Briel 2010) indicated that treatment effects varied with the presence or absence of ARDS, as defined by a value of PaO2/FIO2 of 200 or less, and that in participants with ARDS, higher levels of PEEP were associated with improved survival.
We need also to consider clinical heterogeneity because the trials in this analysis used different approaches to determine PEEP levels in the intervention group. Two studies (Brower 2004; Meade 2008) included recruitment manoeuvres (before the setting of PEEP) with high levels of PEEP preset according to PaO2 and FIO2 values. In the remaining study (Mercat 2008), the PEEP level was preset to achieve plateau pressures between 28 and 30 cm H2O. Despite the use of different criteria for PEEP selection, the optimal method still remains unclear. In this review, subgroup analysis that included different ways of applying PEEP (PEEP on the basis of mechanical characteristics of the lung and PEEP according to FIO2 and PaO2; Analysis 1.14; Analysis 1.15) failed to evidence any significant difference between these two approaches.
With respect to these two issues (heterogeneous population and different ways of applying high levels of PEEP), the results of an ongoing study (Kacmarek 2007) using high levels of PEEP only in participants with ARDS, and in an individualized way, should prove to be both relevant and informative.
In the subgroup analysis, we pooled studies that assessed lung-protective ventilation (low tidal volumes and high levels of PEEP) versus conventional ventilation (Amato 1998; Villar 2006; Analysis 1.16) and found a decrease in mortality with the use of lung-protective ventilation (RR 0.62; 95% CI 0.44 to 0.87). For this reason, we conclude—and we must emphasize—that lung-protective ventilation was shown to be beneficial in participants with ALI and ARDS.
In participants with ALI and ARDS, the use of PEEP has been recognized as producing an improvement in oxygenation (Borges 2006; Ranieri 1991; Suter 1975). In this review we accordingly observed that high levels of PEEP did indeed improve oxygenation in participants up to the first, third, and seventh days of mechanical ventilation, but with heterogeneity occurring in the different analyses surveyed (Analysis 1.2; Analysis 1.3; Analysis 1.4). For this reason we undertook subgroup analysis of oxygen efficacy through PaO2/FIO2 on the first and third days in participants with ARDS, and we saw an improvement in oxygenation (Analysis 1.5; Analysis 1.6), consistent with the findings of Gattinoni (Gattinoni 2006) and Grasso (Grasso 2005). Gattinoni showed that in participants with ARDS, the potential for recruitment may be greater, and the earlier physiological study of Grasso applied both lower (9 ± 2 cm H2O) and higher (16 ± 1 cm H2O) PEEP levels and found that participants with a high potential for recruitment had a greater increase in oxygenation (from PaO2/FIO2 of 150 ± 36 to PaO2/FIO2 396 ± 138). By contrast, in the four studies assessing oxygenation on the seventh day (Brower 2004; Huh 2009; Meade 2008; Mercat 2008), PEEP levels did not differ markedly between groups. Furthermore, in participants with late ARDS, improvements in oxygenation and in respiratory mechanics were less well defined, and the potential for alveolar recruitment could have been reduced (Grasso 2002). These observations could justify the marked heterogeneity that can occur in the analysis of oxygen efficiency through PaO2/FIO2 after seven days.
Barotrauma is a complication in mechanically ventilated patients. Two studies have examined the relationship between barotrauma and high levels of PEEP in participants with ARDS (Weg 1998; Eisner 2002). Weg 1998 performed an analysis of data from a prospective trial of aerosolized synthetic surfactant in participants with ARDS and found no relationship between barotrauma and a mean level of PEEP of 12 cm H2O. Eisner 2002 reported that high levels of PEEP may increase the likelihood of early barotrauma, even after control is applied for markers of acute and chronic disease severity. In the present review, six studies assessed barotrauma (Amato 1998; Brower 2004; Huh 2009; Meade 2008; Mercat 2008; Villar 2006); in accordance with Weg 1998, no statistically significant difference was noted between the groups.
In conjunction with statistically significant heterogeneity, high levels of PEEP produced no significant difference between the two groups in terms of the number of VFDs. We included two studies (Brower 2004; Villar 2006) that differed in two ways: First, in terms of population, Brower 2004 used participants with PaO2/FIO2 ≤ 300, and Villar 2006 reported only participants with ARDS ventilated at a standard setting for 24 hours and persisting with a PaO2/FIO2 ≤ 200. Second, with respect to the mode of applying high levels of PEEP, in Brower 2004, PEEP levels were preset according to PaO2 and FIO2 values, whereas in Villar 2006, those levels were preset at 2 cm H2O above the Pflex (upward shift in the slope of the pressure-volume curve). These differences could explain the statistical heterogeneity present in the analysis of VFDs.
Data on LOS in the ICU were insufficient for an examination to be performed.
Summary of findings table
In the main analysis (mortality before hospital discharge), we do not degrade for publication bias owing to the smaller number of included studies. We downgraded the secondary outcomes of oxygen efficiency on the first, third, and seventh days, and ventilator-free days were likewise downgraded because of an inconsistency resulting from minimal overlap among the studies. Here the P value for heterogeneity was less than 0.05, and the I2 was large. We did not include LOS in the ICU in the summary of findings table because we were not able to pool the data for analysis of this outcome (see Summary of findings for the main comparison).
Agreements and disagreements with other studies or reviews
To date, several reviews have examined the use of high levels of PEEP in participants with ALI and ARDS. Evidence from these reviews indicates the trend toward a decrease in mortality with the use of high levels of PEEP, especially in participants with ARDS.
Gordo-Vidal 2007 evaluated four studies published from 1998 to 2006, three of which used a protective ventilatory strategy involving a low tidal volume and a high PEEP. These studies included contained statistical heterogeneity; moreover, the review failed to report a decrease in mortality (RR 0.73, 95% CI 0.49 to 1.10; P = 0.129). Furthermore, the sampling reported in Gordo-Vidal 2007 could not take into account the studies of greatest statistical weight in this present meta-analysis (i.e. Meade 2008, Mercat 2008) because those investigations had not yet been published.
Three reviews (Oba 2009; Phoenix 2009; Yang 2011) included the same studies as ours.
The aim of the study of Oba 2009 was to test the hypothesis that in participants with ALI and ARDS, the use of high levels of PEEP resulted in lower mortality, especially in more severely ill participants. For the analysis, investigators created a hypothetical group in which participants were treated with low VT, instead of conventional tidal volumes, and low PEEP to match tidal volumes between the two groups studied in each trial. In this meta-analysis, a small but significant decrease in mortality with high PEEP may have been shown to exist (RR 0.89; 95% CI 0.80 to 0.99; P = 0.03). As in our review, clinical heterogeneity was present, and investigators found that the effects of high PEEP were greater in participants with higher ICU severity scores.
Phoenix 2009 identified six eligible studies, of which one had evaluated mortality at day 28 (Ranieri 1999). This review furthermore included five studies (Amato 1998; Brower 2004; Meade 2008; Mercat 2008; Villar 2006) that evaluated mortality before hospital discharge. Authors in this review used a random-effects model, included no data about statistical heterogeneity, and showed a reduction in mortality (RR 0.87; 95% CI 0.78 to 0.96; P = 0.007). Those who performed the meta-analysis that was restricted to studies that included PEEP levels as the main variable investigated (e.g. Brower 2004; Meade 2008; Mercat 2008) obtained the same results as we did. Although that study argued that the evidence obtained supported the use of high levels of PEEP in unselected groups of participants with ALI or ARDS in general, or in participants with more severe disease in particular, we are not in agreement with that conclusion because we believe that a determination of which subgroup of participants (ALI or ARDS) would benefit from high levels of PEEP is necessary.
The study of Yang 2011 includes the same six studies as Phoenix 2009 but assesses mortality at day 28 and the rate of barotrauma. In the analysis of mortality at day 28, investigators reported a significant difference between the two groups (odds ratio (OR) 0.81, 95% Cl 0.68 to 0.95; P = 0.01) and furthermore no significant difference in the rate of barotrauma (OR 0.91, 95% Cl 0.55 to 1.51; P = 0.72). We assessed mortality within 28 days of randomization and found no differences between the two groups, although our analysis differs from the review of Yang 2011 in the included studies.
Three reviews (Briel 2010; Dasenbrook 2011; Putensen 2009) as in our main analysis included randomized controlled trials of participants with a diagnosis of ALI or ARDS (as defined by the American-European Consensus Conference) that compared higher versus lower levels of PEEP at the same tidal volume in both groups (control and experimental).
Putensen 2009 identified several studies in participants with ALI and ARDS that attempted to determine whether low tidal volume, high PEEP, or a combination of the two improved outcome. In a meta-analysis that included studies of high versus low PEEP with the same tidal volume, investigators found, as did we in our review, no significant decrease in mortality (OR 0.86, 95% CI 0.72 to 1.02; P = 0.08).
Briel 2010 performed a systematic review and meta-analysis of individual-patient data from three randomized studies that compared higher versus lower PEEP levels in 2299 participants with acute lung injury. Results showed, overall, no statistically significant difference in hospital mortality between groups (RR 0.94, 95% CI 0.86 to 1.04; P = 0.25). However in the subgroup of participants with ARDS, higher levels of PEEP were associated with improved survival (RR 0.90, 95% CI 0.81 to 1.00; P = 0.049). This review included an explicit study protocol and analysis plan, along with access to trial protocols, case report forms and complete, unedited data sets and standardized outcome definitions across trials (except for rescue therapies).
The review of Dasenbrook 2011 included four studies with ventilation strategies that included higher PEEP during volume and pressure limitation in participants with ALI and ARDS. This study reported that higher levels of PEEP were not associated with a significantly different mortality at day 28 (RR 0.90, 95% CI 0.79 to 1.02) or barotrauma.
None of these three reviews (Briel 2010; Dasenbrook 2011; Putensen 2009) demonstrated a decrease in mortality with the use of high PEEP levels. We obtained the same results in the main analysis, which included studies with similar characteristics.