Association Between Use of Lung-Protective Ventilation With Lower Tidal Volumes and Clinical Outcomes Among Patients Without Acute Respiratory Distress Syndrome: A Meta-analysis

Context Lung-protective mechanical ventilation with the use of lower tidal volumes has been found to improve outcomes of patients with acute respiratory distress syndrome (ARDS). It has been suggested that use of lower tidal volumes also benefits patients who do not have ARDS.

Objective To determine whether use of lower tidal volumes is associated with improved outcomes of patients receiving ventilation who do not have ARDS.

Data Sources MEDLINE, CINAHL, Web of Science, and Cochrane Central Register of Controlled Trials up to August 2012.

Study Selection Eligible studies evaluated use of lower vs higher tidal volumes in patients without ARDS at onset of mechanical ventilation and reported lung injury development, overall mortality, pulmonary infection, atelectasis, and biochemical alterations.

Data Extraction Three reviewers extracted data on study characteristics, methods, and outcomes. Disagreement was resolved by consensus.

Data Synthesis Twenty articles (2822 participants) were included. Meta-analysis using a fixed-effects model showed a decrease in lung injury development (risk ratio [RR], 0.33; 95% CI, 0.23 to 0.47; I², 0%; number needed to treat [NNT], 11), and mortality (RR, 0.64; 95% CI, 0.46 to 0.89; I², 0%; NNT, 23) in patients receiving ventilation with lower tidal volumes. The results of lung injury development were similar when stratified by the type of study (randomized vs nonrandomized) and were significant only in randomized trials for pulmonary infection and only in nonrandomized trials for mortality. Meta-analysis using a random-effects model showed, in protective ventilation groups, a lower incidence of pulmonary infection (RR, 0.45; 95% CI, 0.22 to 0.92; I², 32%; NNT, 26), lower mean (SD) hospital length of stay (6.91 [2.36] vs 8.87 [2.93] days, respectively; standardized mean difference [SMD], 0.51; 95% CI, 0.20 to 0.82; I², 75%), higher mean (SD) PaCO₂ levels (41.05 [3.79] vs 37.90 [4.19] mm Hg, respectively; SMD, −0.51; 95% CI, −0.70 to −0.32; I², 54%), and lower mean (SD) pH values (7.37 [0.03] vs 7.40 [0.04], respectively; SMD, 1.16; 95% CI, 0.31 to 2.02; I², 96%) but similar mean (SD) ratios of PaO₂ to fraction of inspired oxygen (304.40 [65.7] vs 312.97 [68.13], respectively; SMD, 0.11; 95% CI, −0.06 to 0.27; I², 60%). Tidal volume gradients between the 2 groups did not influence significantly the final results.

Conclusions Among patients without ARDS, protective ventilation with lower tidal volumes was associated with better clinical outcomes. Some of the limitations of the meta-analysis were the mixed setting of mechanical ventilation (intensive care unit or operating room) and the duration of mechanical ventilation.

Mechanical ventilation is a life-saving strategy in patients with acute respiratory failure. However, unequivocal evidence suggests that mechanical ventilation has the potential to aggravate and precipitate lung injury.¹ In acute respiratory distress syndrome (ARDS), and in a milder form of ARDS formerly known as acute lung injury (ALI),² mechanical ventilation can cause ventilator-associated lung injury. Ventilator-associated lung injury is a frequent complication in critically ill patients receiving mechanical ventilation, and its development increases morbidity and mortality.¹

Higher tidal volume (V_T) ventilation causes the alveoli to overstretch in a process called volutrauma, and this overstretching is the main cause of ventilator-associated lung injury.³ The use of a lower V_T was shown to reduce morbidity and mortality in patients with ARDS or ALI, thus justifying the progressive decrease in V_T used by clinicians over the past decades.^4-6 However, in critically ill patients without ALI, there is little evidence regarding the benefits of ventilation with lower V_T, partly because of a lack of randomized controlled trials evaluating the best ventilator strategies in these patients.⁷

Some observational studies have suggested that use of higher V_T in patients without ARDS or ALI, at the initiation of mechanical ventilation, increases morbidity and mortality.^8-10 As suggested by the “biotrauma hypothesis,” ventilation with higher V_T and peak pressures may lead to recruitment of neutrophils and local production and release of inflammatory mediators.¹¹ We conducted a meta-analysis to determine whether conventional (higher) or protective (lower) tidal volumes would be associated with lung injury, mortality, pulmonary infection, and atelectasis in patients without lung injury at the onset of mechanical ventilation.

Studies were identified by 2 authors through a computerized blinded search of MEDLINE (1966-2012), Cumulative Index to Nursing and Allied Health Literature (CINAHL), Web of Science, and Cochrane Central Register of Controlled Trials (CENTRAL) using a sensitive search strategy combining the following Medical Subject Headings and keywords (protective ventilation [text word] OR lower tidal volumes [text word]). All reviewed articles and cross-referenced studies from retrieved articles were screened for pertinent information.

Articles were selected for inclusion in the systematic review if they evaluated 2 types of ventilation in patients without ARDS or ALI at the onset of mechanical ventilation. In 1 group of the study, ventilation was protective (lower V_T). Then, this protective ventilation group was compared with another group using conventional methods (higher V_T). A study was deemed eligible if it evaluated patients who did not meet the consensus criteria for ARDS or ALI at baseline.¹² We included randomized trials as well as observational studies (cohort, before/after, and cross-sectional), with no restrictions on language or scenario (intensive care unit or operating room). We excluded revisions and studies that did not report the outcomes of interest. When we found duplicate reports of the same study in preliminary abstracts and articles, we analyzed data from the most complete data set. When necessary, we contacted the authors for additional unpublished data.

Data were independently extracted from each report by 3 authors using a data recording form developed for this purpose. After extraction, data were reviewed and compared by the first author. Instances of disagreement between the 2 other extractors were solved by a consensus among the investigators. Whenever needed, we obtained additional information about a specific study by directly questioning the principal investigator.

In randomized trials, we assessed allocation concealment, the baseline similarity of groups (with regard to age, severity of illness, and severity of lung injury), and the early stopping of treatment. We used the GRADE approach to summarize the quality of evidence for each outcome.¹³ In this approach, randomized trials begin as high-quality evidence but can be rated down for apparent risk of bias, imprecision, inconsistency, indirectness, or suspicion of a publication bias.

The primary end point was the development of lung injury in each group of the study. Secondary end points included overall survival, incidence of pulmonary infection and atelectasis, intensive care unit (ICU) and hospital length of stay, time to extubation, change in PaCO₂, arterial pH values, and change in the ratio of PaO₂ to fraction of inspired oxygen (FIO₂).

We extracted data regarding the study design, patient characteristics, type of ventilation, mean change in arterial blood gases, lung injury development, ICU and hospital length of stay, time to extubation, overall survival, and incidence of atelectasis. For the analysis of lung injury development, mortality, pulmonary infection, and atelectasis, we used the most protracted follow-up in each trial up to hospital discharge. We calculated a pooled estimate of risk ratio (RR) in the individual studies using a fixed-effects model according to Mantel and Haenszel and graphically represented these results using forest plot graphs.

We explored the following variables as potential modifiers: incorporation of “open lung” techniques (using the authors' definitions) into experimental strategies, between-group gradients in tidal volumes and plateau pressures, and case mix effects. We reasoned that each of these might influence the effect of protective ventilation on outcome. To explore whether these variables modified the outcome, we compared pooled effects among studies with and without them. For continuous variables, we used the standardized mean difference (SMD), which is the difference in means divided by a standard deviation.

The homogeneity assumption was measured by the I², which describes the percentage of total variation across studies that is due to heterogeneity rather than chance. I² was calculated from basic results obtained from a typical meta-analysis as I² = 100% × (Q − df)/ Q, where Q is the Cochran heterogeneity statistic. A value of 0% indicates no observed heterogeneity, and larger values show increasing heterogeneity. When heterogeneity was found (I² > 25%) we presented the random-effects model results as primary analysis.

A sensitivity analysis was carried out by recalculating pooled RR estimates for different subgroups of studies based on relevant clinical features. This analysis demonstrates whether the overall results have been affected by a change in the meta-analysis selection criteria. Also, a sensitivity analysis about the treatment effect according to quality components of the studies (concealed treatment allocation, blinding of patients and caregivers, blinded outcome assessment) was conducted. A potential publication bias was assessed graphically with funnel plots, as well as by a Begg and Mazumdar rank correlation and an Egger regression. Interrater reliability was determined by comparing the number of studies included by one author with those of another author in each stage of the search using κ coefficients.

Parametric variables were presented as the mean and standard deviation, and nonparametric variables were presented as the median and interquartile range (IQR). All analyses were conducted with Review Manager version 5.1.1 (The Cochrane Collaboration) and SPSS version 16.0.1 (IBM SPSS). For all analyses, 2-sided P values less than .05 were considered significant.

Our initial search yielded 2122 studies (458 from MEDLINE, 141 from CENTRAL, 885 from CINAHL, and 638 from Web of Science). After removing 711 duplicate studies, we evaluated the abstracts of 1411 studies. After evaluating the abstract of each study, we excluded 1364 studies because they did not meet inclusion criteria. Subsequently, we carefully read the full text of each of the remaining 47 studies and excluded 27 for the following reasons: no data on outcome of interest in 20 studies and same cohort previously analyzed in 7. Twenty references (2822 participants) were included in the final analysis (Figure 1 and Table 1). For the comparisons of interrater reliability in each stage of the search, the κ coefficient was 0.91 in the citation stage (P = .004), 0.86 during the abstract review (P = .03), and 0.90 in the full-text stage (P = .006).

Table 1 summarizes the studies' characteristics. All but 5 studies^{16,22,23,26,29} were randomized controlled trials, and median follow-up time was 21.0 hours (IQR, 6.28-54.60 hours). The median time of per-protocol mechanical ventilation was 6.90 hours for protective and 6.56 hours for conservative strategy. The development of lung injury was the primary outcome in 4 studies. Eight studies evaluated the levels of inflammatory mediators in bronchoalveolar lavage or blood. Tidal volume was set to 6 mL/kg of ideal body weight (IBW) in the protective group of 13 studies; only in 1 study was the tidal volume in the protective ventilation group above 8 mL/kg IBW. Four studies did not report what weight was used to calculate the tidal volume,^14,15,21,25 1 study used the measured weight,¹⁹ and 15 studies used the predicted weight.^{9,16-18,20,22-24,26-32} Of these, 7 used the ARDSnet formula to calculated the predicted body weight.^{16,18,20,24,28-30}

The tidal volume gradient between protective and conventional ventilation ranged from 2 to 6 mL/kg IBW, with a mean (SD) of 4.15 (1.42) mL/kg IBW. The tidal volume gradient was less than 4 mL/kg IBW in 30.0% of the studies, between 4 and 5 mL/kg IBW in 40% of the studies, and above 5 mL/kg IBW in 30% of the studies. In 15 studies, the reason for intubation was scheduled surgery,^{9,15,17-22,24,25,29-32} and in 5, the reason was mixed (medical or surgery).^{14,16,23,24,28} Lung injury was diagnosed according to the American-European Consensus Conference definition in 6 of the 8 trials that assessed this outcome.^{16,23,26,27,31,32} The diagnosis of infection was made by clinical assessment plus laboratory, radiological, and microbiological evaluation in 2 studies^14,26; was made by decrease in PaO₂/FIO₂ plus radiological assessment in 1 study³¹; and was not specified in the last study.²⁰

eTable 1 summarizes study methods, highlighting features related to the risk of bias. Randomization was concealed in 11 of 15 randomized controlled trials included, and follow-up was excellent with minimal loss. Limitations included a lack of blinding (all trials), a lack of intention-to treat analysis (12 trials), and early stopping for benefit (1 trial). Age, weight, minute-volume (product of respiratory rate and tidal volume), and PaO₂/FIO₂ were all similar between the 2 groups analyzed (Table 2 and eTable 2). As expected, V_T and plateau pressure were lower and positive end-expiratory pressure (PEEP) and respiratory rate were higher in the protective group. PaCO₂ was higher in the protective group but remained within normal limits (35-45 mm Hg). Acidosis (pH <7.35) was found in the protective group in 3 studies, and the pH level in the protective group was similar to that of the conventional group. The mechanical ventilation settings for each study are provided in eTable 3.

Forty-seven of 1113 patients (4.22%) assigned to protective ventilation and 138 of 1090 patients (12.66%) assigned to conventional ventilation developed lung injury during follow-up (RR, 0.33; 95% CI, 0.23-0.47; number needed to treat [NNT], 11). The result of the overall test for heterogeneity was not statistically significant, and the I² was 0% (no sign of heterogeneity) (Figure 2). When stratified by the tidal volume gradient between the 2 groups, the RR for lung injury decreased from 0.35 (95% CI, 0.23-0.51) in the group with less than 4 mL/kg IBW to 0.26 (95% CI, 0.10-0.66) in the group with 4 to 5 mL/kg IBW (eFigure 1). The RR for the development of lung injury with conventional ventilation, analyzing only randomized controlled trials, was 0.26 (95% CI, 0.10-0.66; NNT, 10).

Overall mortality was lower in patients receiving protective ventilation (RR, 0.64; 95% CI, 0.46 to 0.89; NNT, 23). The incidence of pulmonary infection (using the authors' definition) and atelectasis were lower in the group receiving ventilation with a lower V_T (RR [random-effect], 0.45; 95% CI, 0.22 to 0.92; NNT, 26; and RR, 0.62; 95% CI, 0.41 to 0.95, respectively) (Figure 2). The I² test indicated moderate heterogeneity only in the analysis of pulmonary infection (32%). Protective ventilation was associated with a shorter mean (SD) hospital stay (6.91 [2.36] vs 8.87 [2.93] days, respectively; SMD, 0.51; 95% CI, 0.20 to 0.82), and showed no difference in ICU stay (3.63 [2.43] vs 4.64 [3.29] days, respectively; SMD, 0.37; 95% CI, −0.53 to 1.27) and time of mechanical ventilation (51.07 [58.08] vs 47.12 [45.00] hours, respectively; SMD, 0.48; 95% CI, −0.27 to 1.23).

Mean (SD) levels of PaCO₂ were higher in the protective ventilation group (41.05 [3.79] vs 37.90 [4.19] mm Hg, respectively; SMD, −0.51; 95% CI, −0.70 to −0.32), and mean (SD) pH levels were lower (7.37 [0.03] vs 7.40 [0.03], respectively; SMD, 1.16; 95% CI, 0.31 to 2.02). The mean (SD) PaO₂/FIO₂ ratio was similar between the groups (304.40 [65.70] vs 312.97 [68.13], respectively; SMD, 0.11; 95% CI, −0.06 to 0.27). All these analyses yield significant heterogeneity and were analyzed by random-effects model (I² for hospital stay, ICU stay, time of mechanical ventilation, PaCO₂, pH, and PaO₂/FIO₂ of 75%, 95%, 92%, 54%, 96%, and 60%, respectively) (eFigures 2, 3, 4, 5, 6, 7 and eTable 4).

In eTable 5, the GRADE evidence profile is provided. This profile evaluates the effect of protective ventilation in patients without ARDS or ALI, only from a systematic review and a meta-analysis of randomized controlled trials. The findings for lung injury, mortality, and pulmonary infection were considered moderate, high, and low quality, respectively, by the GRADE profile. Sensitivity analyses according to quality components of each study are shown in eTable 6.

In addition, we excluded each trial one at a time and assessed the results. In lung injury and pulmonary infection analyses, the results were always significant despite the exclusion of any trial. After we excluded the trial by Yilmaz et al,²³ the analysis of mortality was no longer significant.

To explore these results, we performed a stratified analysis across a number of key study characteristics and clinical factors, and this analysis is shown in Table 3. Protection from lung injury, in the protective group, was more pronounced in studies that were not randomized controlled trials performed in the ICU. These trials did not incorporate recruitment maneuvers, had a higher plateau pressure gradient, and a smaller tidal volume gradient. In the survival analysis, we found significant changes in studies without recruitment maneuvers, in studies that were not randomized trials, and in studies performed in the ICU with a lower tidal volume gradient.

For pulmonary infections, we found no statistically significant association in studies that were not randomized trials, a tidal volume gradient less than 4 mL/kg IBW, and the use of recruitment maneuvers. A tidal volume gradient from 4 to 5 mL/kg IBW and a randomized controlled trial performed in surgical patients were each associated with a significant reduction in pulmonary infections in the protective group.

Funnel-plot graphical analysis (eFigure 8), Begg and Mazumdar rank correlation, and Egger regression did not suggest a significant publication bias for the analyses conducted in Figure 2 (Kendall τ = 0.17, P = .63; Egger regression intercept = 0.24, P = .68).

We found evidence that a ventilation strategy using lower tidal volumes is associated with a lower risk for developing ARDS. Furthermore, the strategy was associated with lower mortality, fewer pulmonary infections, and less atelectasis when compared with higher tidal volume ventilation in patients without lung injury at the onset of mechanical ventilation. These benefits were associated with a shorter hospital length of stay. Protective ventilation was associated with higher PaCO₂ levels and lower pH values, but no difference in the incidence of acidosis was found. In all studies, although the primary goal of the investigators was to compare 2 different tidal volumes, other ventilator strategy elements were associated with the use of lower tidal volumes. Notably, differences in the levels of PEEP and plateau pressure did not influence the final results of the meta-analysis.

Previously, Esteban et al³³ showed plateau pressures above 35 cm H₂O to be associated with an increased risk of death in ICU patients. Although not definitive, this study at least suggested that higher V_T has the ability to exaggerate lung injury and maybe even cause death in patients who require mechanical ventilation for days. Fernández-Pérez et al³⁴ showed higher V_T to be associated with postoperative respiratory failure in patients receiving ventilation for only a few hours in the operating room. In light of this information, over the past decade, V_T has progressively decreased from greater than 12 to 15 mL/kg IBW to less than 9 mL/kg IBW.^6,35 The results of the present meta-analysis support this change in ventilation practice. Our results may even suggest that V_T should be further reduced.

Protective ventilation in patients with ALI or ARDS is already well established; however, physicians do not always adhere to such guidelines. Mikkelsen et al³⁶ reported that approximately one-third of the patients were receiving protective ventilation at 48 hours, and the main reason for poor adherence was the uncertainty about the diagnosis of ARDS. Another possible reason is that 82% of the patients who never received protective ventilation had a plateau pressure below 30 cm H₂O. However, it is well established that reducing the V_T in patients with plateau pressures below 30 cm H₂O is associated with a survival benefit.¹⁰ In this context, the adoption of protective ventilation in patients without lung injury may be even more difficult.

It is possible that the beneficial effects of protective ventilation, regarding the development of lung injury, are even greater than what is suggested by the current analysis. Mechanical ventilation can damage the lung, cause inflammation, and release cytokines into the systemic circulation.^20,25 This process may cause fever, leukocytosis, and new pulmonary infiltrates, which could be interpreted as ventilator-associated pneumonia instead of ventilator-associated lung injury. The absence of strict criteria for the diagnosis of pneumonia, such as microbiological identification in blood and bronchoalveolar lavage, in the studies evaluated may lead to an incorrect diagnosis. Ventilator-associated lung injury may be incorrectly diagnosed as pneumonia in many cases, underestimating the true incidence of lung injury. It is difficult to diagnose pneumonia in the presence of ARDS or ALI, with a quoted sensitivity using conventional clinical criteria of less than 50%.³⁷

Our findings are in line with a recently published retrospective study of cardiac surgery patients.³⁸ Although it should be noted that the lower tidal volumes in that study were much higher than those used in the protective groups of the studies analyzed in this meta-analysis, a tidal volume of more than 10 mL/kg was found as a risk factor for organ failure and prolonged ICU stay after cardiac surgery.

The results of this meta-analysis should be interpreted within the context of the included studies. Systematic reviews are subject to publication bias, which may exaggerate the study's conclusion if publication is related to the strength of the results. Additionally, it may be important to distinguish between mechanical ventilation performed in the operating room and that performed in the ICU. Patients in the operating room receive mechanical ventilation for a much shorter time than those in the ICU. Both surgical patients and critically ill patients are at risk for several causes of lung injury. However, these may not be the same for both patient groups, and mechanical ventilation may have different effects on both groups. In addition, although our meta-analysis found decreased mortality rate with protective ventilation, the interpretation of this finding should be considered cautiously because it was discovered only after the addition of the study by Yilmaz et al.²³ Also, one important limitation is that the patients received ventilation for a relatively short time in most studies, which complicates the extrapolation of the results for patients receiving ventilation for long periods in the ICU. For the lung injury analysis, 4 of 8 studies (accounting for 85.4% and 87.2% of the events in the conservative and protective groups, respectively) were not randomized controlled trials, and the randomized controlled trials were of moderate quality. Furthermore, funnel plots are limited as a test for publication bias for a small number of studies.

All the dichotomous analyses yielded significant results, and with the exception of pulmonary infection, all the results showed no heterogeneity (I² = 0%). Pulmonary infection yielded moderate heterogeneity (I² = 32%), but the analysis with a random-effects model showed similar results. However, all the continuous analyses showed significant heterogeneity (all I² >60%) and with the use of a random-effects model only differences in pH level, PaCO₂ level, and hospital length of stay showed significant results. Therefore, continuous analyses need to be interpreted with caution because of the heterogeneity.

In conclusion, our meta-analysis suggests that among patients without lung injury, protective ventilation with use of lower tidal volumes at onset of mechanical ventilation may be associated with better clinical outcomes. We believe that clinical trials are needed to compare higher vs lower tidal volumes in a heterogeneous group of patients receiving mechanical ventilation for longer periods.

Corresponding Author: Ary Serpa Neto, MD, MSc, Avenue Lauro Gomes, 2000 São Paulo, Brazil (aryserpa@terra.com.br).

Author Contributions: Dr Serpa Neto had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Serpa Neto, Cardoso, Manetta, Pereira, Espósito, Schultz.

Acquisition of data: Serpa Neto, Pereira, Espósito, Pasqualucci.

Analysis and interpretation of data: Serpa Neto, Cardoso, Manetta, Pereira, Espósito, Damasceno, Schultz.

Drafting of the manuscript: Serpa Neto, Pereira, Damasceno, Schultz.

Critical revision of the manuscript for important intellectual content: Serpa Neto, Cardoso, Manetta, Pereira, Espósito, Pasqualucci, Damasceno, Schultz.

Statistical analysis: Serpa Neto, Pereira.

Administrative, technical, or material support: Cardoso, Manetta, Espósito, Pasqualucci, Damasceno, Schultz.

Study supervision: Damasceno, Schultz.