Acute respiratory distress syndrome (ARDS) and acute lung injury (ALI) are common, devastating clinical syndromes of lung injury that affect both medical and surgical patients. The prevalence of ARDS/ALI is approximately 9% in intensive care patients and 39.6% in ventilated patients (1). The mortality rate is approximately 40%–50% (2). Those who survive the illness have a reduced quality of life as well as cognitive impairment and frequent disability (3).
ARDS is associated with reduction in static compliance of the respiratory system (4). As a result of this low compliance, mechanical ventilation (MV) with high pressures is needed to obtain a sufficient tidal volume. The larger the tidal volume, the higher the pressure required, which may lead to alveolar rupture (barotrauma). MV could lead to injury resulting from overdistension. This results from the distribution of the increased tidal volume to the high-compliance regions, causing stretching and sheer forces on the alveolar wall (volutrauma) (5,6). Cyclic inflation of the injured lung can exacerbate lung injury (6). Damage caused by MV to the lungs is known as ventilator-induced lung injury.
Pressure- and volume-limited ventilation has been proposed as ventilatory support for ARDS patients (7–9). This type of approach has been termed lung-protective ventilation strategy (LPVS).
However, decreasing the tidal volume is not without potential hazards (10). Severe hypercapnia and acidosis can have adverse effects, including increased intracranial pressure, depressed myocardial contractility, pulmonary hypertension, and depressed renal blood flow (11). The view that these risks are preferable to the higher plateau pressure required to achieve normocapnia is a substantial shift in the approach to ventilatory management.
A previous meta-analysis (12) investigated this issue. The authors did not combine the trial results or perform a thorough exploration of overall clinical heterogeneity, and they stratified the trials into 2 groups, beneficial and nonbeneficial, trying to explain the opposite results in a post hoc fashion. This might have surreptitiously introduced a selection towards the 3 nonbeneficial trials. Their results could be attributable in part to the effect of this selection bias. In addition, the authors expressed concern that patients in the control groups of the beneficial trials may have been exposed to unnecessary risks. However, the Office for Human Research Protection (OHRP) substantiated the ethical appropriateness and research design of the concerned trials (13).
The aim of this systematic review was to compare ventilation with smaller tidal volume and ventilation with conventional tidal volume to determine whether such a LPVS reduced morbidity and mortality in critically ill patients affected by ALI/ARDS. A secondary objective was to determine whether the comparison between small tidal volume and conventional tidal volume is different if a plateau pressure of more than 30–35 cm H2O was used.
We identified trials by searching the Cochrane Central Register of Controlled Trials (CENTRAL), the Cochrane Library issue 3, 2002, the electronic databases MEDLINE (January to June 2003), EMBASE and CINAHL (1982 to June 2003), intensive care journals and conference proceedings, databases of current research, reference lists, and “gray literature.” No language restrictions were applied.
We selected studies with protective ventilation that used smaller tidal volume or low airway driving pressure (plateau 30 cm H2O or less) or a combination of the two and that resulted in tidal volume of 7 mL/kg or less versus conventional mechanical ventilation that used tidal volume in the range of 10 to 15 mL/kg. Hypercapnia, as an unavoidable part of the protective ventilation intervention, was accepted as long as the resulting acidosis was controlled and kept within acceptable ranges.
Only randomized, controlled trials (RCT) were accepted to guarantee control of selection bias. Quasi-randomized or crossover studies were excluded. All patients entered into the trials had to be 16 yr or older, with tracheal intubation and pulmonary ventilation, affected by ARDS or ALI from any cause, as defined by the North American-European Consensus Conference on ARDS (14) or by the Lung Injury Severity Score (15).
Primary outcome was overall mortality, evaluated at hospital discharge, if this information was available; otherwise, we used mortality at the end of the follow-up period scheduled for each trial. Secondary outcomes were:
- Development of multiorgan failure.
- Duration of MV and total duration of mechanical support.
- Total duration of stay in intensive care unit and hospital.
- Long-term mortality.
- Long-term health-related quality of life.
- Long-term cognitive outcome.
Data Abstraction and Analysis
Two independent reviewers performed trial searches, assessment of methodology, and extraction of data, using standard forms, with comparison and resolution of any differences found at each stage. High-quality trials were defined as those that were controlled, were appropriately randomized, had adequate concealment of allocation, and demonstrated completeness of follow-up according to intention-to-treat analysis. Concealment of allocation was judged adequate if trials took adequate measures to conceal allocation through central randomization, serially numbered opaque envelopes, or tables of random numbers. Generation of allocation sequences was judged adequate whether trials were deemed to have satisfactory sequence generation (random numbers generated by computers or drawing of lots of envelopes). A thorough attempt was made to identify clinical heterogeneity based on differences in population study, delivery of interventions, and measures of outcomes.
Review Manager 4.2 (Cochrane Collaboration software) was used to combine data from the trials. Meta-analysis was performed using a fixed effect model. We performed formal exploration of heterogeneity using a statistical test of non-combinability, with the level of statistical significance set at P = 0.1 (16). We used the statistics I2 as a further contribution to interpretation of the extent of heterogeneity (17). Values of I2 <56%, and more than 31%, account for “moderate” heterogeneity. Meta-analysis with a random effects model was applied when this happened (18). Relative risk (RR), risk difference (RD), and 95% confidence intervals (CI) were reported for categorical outcomes, and weighted mean differences (WMD) were reported for continuous variables. A subgroup analysis was accomplished based on plateau pressure in control groups. We performed sensitivity analyses based on study quality, clinically important features (which may explain differences across study results), and imputing values for dropouts.
We found 8 studies of potential relevance. Three studies were excluded after a closer assessment (19–21). Five studies (22–26) met the study inclusion criteria (Table 1). The total number of patients randomized in each study varied from 52 (22) to 861 (26). Time elapsed from eligibility to randomization ranged from 1 h to 36 h.
All the trials assessed the baseline risk by combining several prognostic variables into a severity score (APACHE II or APACHE III score) (27,28). The APACHE II score at baseline was in the range of 17 ± 8 sd (23) to 28 ± 7 sd (22). The APACHE III score ranged from 81 ± 28 sd (26) to 90 ± 26 sd (25).
Four trials (23–26) set the tidal volume based on body weight. In the ARDS Network and the Brower et al. (25) and Stewart et al. (24) trials, the tidal volume was set according to predicted (or ideal) body weight (IBW), which is, on average, 20% less than measured body weight. When transformed to mL/kg measured body weight, the mean tidal volume in the ARDS Network trial ranged from 9.4 to 9.9 mL/kg in the control group, which was quite similar to values used in the other trials, and 5.2 mL/kg in the small tidal volume group, which was less than other trials. Protocols for the management of acidosis using bicarbonate infusions were developed in all trials. ARDS Network investigators (26) were most aggressive in attempting to keep pH more than 7.30 for all patients and allowed violations of the tidal volume and airway pressure limits when pH decreased to less than 7.15. In contrast, another study (24) did not dictate volume and pressure violations until pH decreased to 7.00.
Mortality was measured at a cut-off point of day 60 in one study (23) and at hospital discharge in two other studies (24,25). In one study (22), mortality both at day 28 and at hospital discharge were reported. Importantly, in the ARDS Network study patients were followed until discharged home or for 180 days, whichever occurred first. This may be considered equivalent to hospital mortality. Only 3 trials reported duration of MV (23–25) (Table 2). In the ARDS Network study, this outcome was reported as median number of days without further description. Similarly, organ failures were reported using different measures. Brochard et al. (23) reported that 24 patients in each group suffered from organ failures without further details, whereas Stewart et al. (24) reported a mean value of 2 organ failures per patient in each group. The ARDS Network study reported the number of days without nonpulmonary organ failure at day 28. Other secondary outcomes considered relevant for this review were not reported in the studies.
Four studies (22–24,26) reached a grade A of quality and one study (25) was graded B according to the Cochrane system for grading concealment of allocation. In the ARDS Network trial 31 patients (3.6%) were lost to follow-up. Twenty-two of them were patients still hospitalized when the trial was stopped, and for 9 patients the outcome was unknown (Fig. 1 in the original trial). In only one trial (22) was it clearly stated that the study was not blind. Protocols for management of MV were found in all the studies, thus minimizing performance bias. A protocol for weaning was found only in the ARDS Network study. Publication bias was minimized by the fact that the review included trials reporting positive as well as negative results.
The result of the overall test for heterogeneity was not statistically significant (P = 0.12; 4 df), but it was within the range of “moderate heterogeneity” (I2 = 45.9%). Overall mortality at the end of the follow-up period for each trial was lower in patients with the LPVS when compared with control treatment: fixed effect model RR, 0.85 (CI, 0.74–0.98), RD −6% (CI, −12% to −1%) (Fig. 1). Using a random effects model, the RR was 0.91 (CI, 0.72 to 1.14), RD −4% (CI, −14% to 6%). The different duration of follow-up used in each trial for outcome assessment should suggest caution when interpreting this overall estimate of effect. Therefore, we stratified the trials based on comparable outcomes; summary estimates for hospital mortality and mortality at day 28 are shown in Table 3.
Mortality at day 28 (22,23,26) (1030 patients) was significantly less in patients with the LPVS: fixed effect model RR, 0.74 (CI, 0.61 to 0.88); RD was −10% (CI, −15% to −4%). When applying a random effects model the RR was largely unchanged: 0.73 (CI, 0.61 to 0.87). The benefit of small tidal volume ventilation on hospital mortality (22,24–26) (1086 patients) was uncertain: RR, 0.84 (CI, 0.68 to 1.05). The test for heterogeneity gave a P = 0.22 (df = 3), I2 = 31.8%; and P = 0.41 (df = 2), I2 = 0% for hospital mortality and mortality at day 28, respectively, thus confirming more homogeneous data.
There were only sufficient data to assess the association between LPVS and the duration of MV. There was a trend toward a smaller duration of MV in patients with the LPVS, but this reduction was not statistically significant (fixed effect model: WMD −0.83 [−1.92 to 0.27], random effects model: WMD 0.38 [−3.06 to 3.82], test for heterogeneity:P = 0.22, df = 3, I2 = 33.1%).
We performed subgroup analysis based on the delivery of interventions and comparing the overall estimate of effect of treatment on all cause mortality at the end of the follow-up period. Effect of treatment in trials with a “low pressure” control group (mean plateau pressure, 31 cm H2O or less) (23–25) (288 patients) was compared with the effect of treatment in trials with a “high pressure” control group (plateau pressure more than 31 cm H2O, mean value) (22,26) (914 patients). Overall mortality was significantly less in the LPVS when a “higher” plateau pressure in the control arm was applied (fixed effect model: RR 0.76 [CI, 0.64–0.91], RD −10% [CI, −16% to −4%]) and not significantly different with “lower” plateau pressure in control arm (fixed effect model: RR 1.13 [CI, 0.88–1.45], RD 6% [CI, −6% to 17%]). The test for direct comparison of the two groups gave a P value of 0.004. The interaction P value was 0.0003. The test for heterogeneity was P = 0.89 (df = 2), I2 = 0%, and P = 0.43 (df = 1), I2 = 0%, respectively, for “low” and “high” plateau pressure groups (Fig. 2).
We excluded one trial that was clearly unblinded and that used a different method to deliver the intervention in the treatment group (22) (53 patients). The test for heterogeneity changed slightly, P = 0.13 (df = 3). Overall mortality at the end of follow-up periods was also changed slightly; fixed effect model: RR 0.87 (CI, 0.75–1.01); random effects model: RR 0.97 (CI, 0.76–1.24). We evaluated the impact of the largest study (26) (861 patients) by excluding it from the analysis. The effect of LPVS and conventional treatment strategies on all cause mortality disappeared; fixed effect model: RR 1.01 (CI, 0.81–1.27); random effects model: RR 0.99 (CI, 0.74–1.30). The test for heterogeneity was P = 0.20 (df = 4), I2 = 35.2.
Sensitivity analysis performed by shifting the 31 censored patients of the ARDS Network trial between groups, according to best/worse case analysis, did not affect the overall estimate.
To test whether underlying risk (as expressed by control groups’ mortality rates) was associated with the effect of intervention, we performed a regression analysis. Figure 3 shows a graph of RR of death against proportion of deaths in the conventional ventilation group from the 5 trials included in this review. The graph includes the line of predicted values from a linear regression. Point T in Figure 3 defines a cut-off value of risk in the control group showing that the treatment is effective (RR <1) only in patients with an underlying risk more than this value. However, the association was not statistically significant (estimated slope is −0.013 [CI, −0.03 to 0.01], P = 0.17). Weighted regression equation: (1.07 − 0.005 × Control), P = 0.67.
We found evidence that ventilation strategy using a tidal volume of <7 mL/kg of measured body weight and plateau pressure <31 mm Hg reduced mortality at day 28 and a suggestion that it might reduce hospital and long-term mortality. In all the selected studies, although the primary goal of the investigators was to compare two different tidal volumes, other elements of ventilatory strategy were associated with smaller tidal volumes. The experimental intervention included permissive hypercapnia, variable levels of positive end-expiratory pressure (PEEP), and low plateau airway pressure. The traditional intervention consisted of larger tidal volume, normocapnia, lower levels of PEEP, and potentially higher plateau pressures. Therefore, the studies performed a comparison of two approaches rather than two single interventions, and caution is required in interpreting these results, especially when analyses have been inspired by looking at the available aggregate data. The meta-analysis is dominated by the large ARDS Network trial (26). The overall estimate at the end of the follow-up period for each trial is of borderline statistical significance when using the fixed effect model. The statistic I2 showed that an important percentage of variability in point estimates is a result of heterogeneity rather than sampling error. That value is sufficiently high to justify using a random effects model. The point estimate is similar to that given by the fixed effect model with wider CIs. Overall, the discordant figures reflect an unstable conclusion and uncertainty regarding the generalizability of the results (18). Several hypotheses can explain and interpret these data:
- There is a “hidden” source of heterogeneity. Although statistical tests of heterogeneity were not significant, we showed that studies were clinically different in some points. Furthermore, most of the trials did not report protocols of concomitant treatments and associated diseases (i.e., ventilator-associated pneumonia).
- There is a subgroup of patients who can be ventilated with smaller tidal volumes or volumes in the conventional range, without differences in mortality. Because alveolar recruitment occurs during tidal inflation (29), reduction of tidal volume may prove beneficial when it prevents hyperinflation and overdistention, but it could be harmful if it is unable to recruit previously collapsed or compressed alveoli. In this situation the level of PEEP may play a critical role (30) and individual titration of ventilation is crucial (31).
Because of several differences in the choice of time of follow-up for outcome assessment, meta-analysis was repeated using a homogeneous outcome (mortality at day 28), which showed a relevant, stable benefit from the protective ventilatory approach (RR Reduction = 26%). The benefit from small tidal volume ventilation was uncertain when considering hospital mortality as the end-point. A possible explanation could involve the lack of power of the analysis or the absence of beneficial effect of smaller tidal volumes on the development of organ failure, leading to an unfavorable long-term outcome. Conversely, smaller tidal volumes have a beneficial effect on short-term endpoint, protecting aerated lung parenchyma and leading to lung recovery.
Duration of MV was not significantly associated with ventilatory strategy. Surprisingly little is known regarding long-term outcome. Two studies reporting follow-up at 1–2 years were unable to demonstrate that a limited ventilation strategy improves either long-term function or quality of life in survivors of ALI/ARDS (32,33).
Two studies (22,26) used a plateau pressure in the control groups that was higher than the other trials. In the study design, the ARDS Network investigators would have allowed plateau pressure to be increased to 45 mm Hg if necessary to deliver the target tidal volume of 12 mL/kg IBW. Meta-analysis performed stratifying trials according to “high plateau pressure” and “low plateau pressure” in the control group confirmed that when delivery of conventional tidal volume was associated with plateau pressure of 31 cm H2O or less there was no evidence of decreased mortality from protective ventilation.
A previous meta-analysis (12) investigated this issue, concluding that treatment in controls differed from current practice in terms of too-large tidal volumes and plateau pressure, and this difference may have influenced outcomes in two trials (22,26). This study suggested that as long as tidal volumes produce airway pressures considered safe, there is no benefit from using smaller tidal volumes. Therefore, the trials may have identified risks associated with larger tidal volume rather than benefits associated with smaller tidal volume. However, despite the low power resulting from the small number of trials, regression analysis indicates that the effect of treatment was not related to the mortality rate in the control groups, and underlying risk is not a significant source of heterogeneity (34) (Fig. 3). In addition, it should be taken into account that plateau pressure reflects both pulmonary and chest compliance and therefore the same limit of pressure may reflect smaller tidal volumes if the chest is stiff and vice versa. Patients with potentially altered chest compliance were excluded in all the studies. Nevertheless, in some patients with ARDS, flattening of the pressure-volume curve may be attributable to an increase in chest wall elastance related to abdominal distension (35). Intraabdominal pressure was not accounted or controlled for in the selected studies. The difference between the two groups of trials (high pressure and low pressure) could be attributed to different plateau pressure in control groups, as well as differences in chest wall elastance, or simply random variability.
In the Eichacker et al. study (12), overall clinical heterogeneity is not discussed but simply recognized using a statistical test (Breslow-Day test, P = 0.06). We know that these formal tests for heterogeneity have a low power when the number of trials is small. The new I2 statistics accounts for the number of studies included in the analysis, showing that a variability actually exists across the trials, but it is not enough to prevent the results from being combined using a random effects model. In addition, we stratified the trials according to comparable outcomes, eliminating an important source of heterogeneity.
Important concerns were raised in the design of landmark clinical trials of patients with ARDS (12). First, are the control groups representative of the standard of care? Second, is there a standard ventilation strategy in ARDS or only current clinical practice? Third, might the controls have been subjected to a range of tidal volumes conferring a disadvantage versus not being in the trial? These safety-related concerns triggered a controversy and an investigation by the OHRP (13). However, after the investigation, the OHRP declared that risks to subjects participating in the ARDS Network trial were minimal and reasonable in relation to anticipated benefits. The concerned studies were needed because many aspects of the care of ARDS patients were highly variable, with no standard of care (13,36).
A possible interpretation of the discordant results, as proposed by Gattinoni et al. (37) could involve variations of transpulmonary pressure in the individual patient. The large volume might induce lung damage when the resulting transpulmonary pressure is high. Conversely, when trans-pulmonary and airway pressure are within the safe limits, large or intermediate tidal ventilation (8–10 mL/Kg) could be used, thus avoiding potentially deleterious effects of small tidal volume. This hypothesis should be tested in a new trial.
There were some concerns about clinical heterogeneity of trials:
- The ceiling value of 30 to 31 cm H2O of plateau pressure was violated by two studies (22,26) that used larger differences in tidal volume between groups;
- Mortality was the main end-point, but the cut-off point varied across trials;
- Two studies (23,26) reported the smallest mortality rate in control groups, presumably reflecting a lower baseline risk;
- Inclusion criteria were very similar or even identical across studies, whereas exclusion criteria were slightly different.
- Presumably, treatment was not blinded in any of the studies included in this review because of the practical difficulties of “sham” ventilation.
Some issues still remain to be addressed:
- The treatment may be effective only in a subgroup of patients. The smaller tidal volume ventilation may be clinically worthwhile only in the more severely ill patients.
- The adverse effects of smaller tidal volume were not addressed in the studies. Specifically, the impact of acidosis and hypercapnia on the development of organ failure, a major determinant of outcome (38), was not clear.
- Ventilation with smaller tidal volume can be very effective for short-term lung recovery, but its impact on long-term recovery is still uncertain or unknown.
- The independent contributions of smaller tidal volume (avoiding overdistension) and PEEP (avoiding cyclic opening and closure of alveoli) could not be detected.
- The strict control of acidosis may have had a key role in decreasing mortality in one trial (26).
Ventilation with larger tidal volume and higher plateau pressure is associated with increased risk of death, but the independent contribution of larger tidal volume (overdistension) and/or higher plateau pressure (barotrauma) cannot be clearly identified. Smaller tidal volume ventilation may be preferable when lung recovery is a priority. Long-term health-related quality of life, long-term cognitive outcomes, and cost need to be assessed in future trials. Large trials with subgroup analysis would be able to determine which specific patients would benefit from one of the two ventilatory approaches. Alternatively, a systematic review that uses individual patient data (39) can achieve the ultimate aim to:
- Undertake survival and other time-to-event analyses. If individual survival times were available for each trial, heterogeneity of the considered outcomes (a major issue of this review) could be solved and the issue of whether protective ventilation leads to a prolongation of long-term survival or only to lung recovery could be investigated;
- Undertake subgroup analysis to assess differences in hypercapnia, actually administered tidal volumes, and other confounders;
- Ensure the appropriateness of analysis.
We would like to thank Professor Marcus Müllner, Professor Nathan Pace, Dr. Asima Bokhari, Dr. Mark Davies, Janet Wale, Nete Villebro, and Kathie Godfrey for their help and editorial advice during the preparation of this review. Thanks to John Senior for his assistance in reviewing the manuscript and for his contribution of ideas when we planned the study. Thanks to Mario Mergoni for constructive critique.
1. Roupie E, Lepage E, Wysocki M, et al. Prevalence, etiologies and outcome of the acute respiratory distress syndrome among hypoxemic ventilated patients. Intensive Care Med 1999;25:920–9.
2. Lewandowski K. Epidemiological data challenge ARDS/ALI. Intensive Care Med 1999;25:884–6.
3. Davidson TA, Caldwell ES, Curtis JR. Reduced quality of life in survivors of acute respiratory distress syndrome compared with critically ill control patients. JAMA 1999;281:354–60.
4. Marini JJ. Lung mechanics in the adult respiratory distress syndrome. Clinical Chest Med 1990;11:673–90.
5. Parker JC, Hernandez LA, Peevy KJ. Mechanism of ventilator induced lung injury. Crit Care Med 1993;21:131–43.
6. Dreyfuss D, Soler P, Basset G, Saumon G. High inflation pressure pulmonary edema: respective effects of high airway pressure, high tidal volume and positive end-expiratory pressure. Am Rev Respir Dis 1998;137:1159–64.
7. Hickling KG, Henderson SJ, Jackson R. Low mortality associated with volume pressure limited ventilation with permissive hypercapnia in severe adult respiratory distress syndrome. Intensive Care Med 1990;16:372–7.
8. Hickling KG, Walsh J, Henderson S, Jackson R. Low mortality rate in adult respiratory distress syndrome using low-volume, pressure limited ventilation with permissive hypercapnia: a prospective study. Crit Care Med 1994;22:1568–78.
9. Artigas A, Bernard GR, Carlet J. The American-European consensus conference on ARDS, part 2. Intensive Care Med 1998; 24:378–98.
10. Roupie E, Dambrosio M, Servillo G, et al. Titration of tidal volume and induced hypercapnia in acute respiratory distress syndrome. Am J Respir Crit Care Med 1995;152:121–8.
11. Feihl F, Perret C. Permissive hypercapnia: how permissive should we be? Am J Respir Crit Care Med 1994;150:1722–37.
12. Eichacker PQ, Gerstenberger EP, Banks SM, et al. Meta-analysis of acute lung injury and acute respiratory distress syndrome trials testing low tidal volumes. Am J Respir Crit Care Med 2002;166:1510–4.
13. Steinbrook R. Trial design and patient safety. The debate continues. N Engl J Med 2003;349:620–30.
14. Bernard GR, Artigas A, Brigham KL, et al. The American-European consensus conference on ARDS: definitions, mechanisms, relevant outcomes, and clinical trial coordination. Am J Respir Crit Care Med 1994;149:818–24.
15. Murray JF, Matthay MA, Luce JM, Flick MR. An expanded definition of the adult respiratory distress syndrome. Am Rev Respir Dis 1988;138:720–3.
16. Fleiss JL. Analysis of data from multiclinic trials. Control Clin Trials 1986;7:267–75.
17. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med 2002;21:1539–58.
18. Deeks JJ, Altman DG, Bradburn MJ. Statistical methods for examining heterogeneity and combining results from several studies in meta-analysis. In: Egger M, Smith GD, Altman DG, eds. Systematc reviews. London: BMJ books, 2001:285–312.
19. Rappaport SH, Shpiner R, Yoshihara G, et al. Randomized, prospective trial of pressure-limited versus volume-controlled ventilation in severe respiratory failure. Crit Care Med 1994;22:22–32.
20. Ranieri VM, Suter PM, Tortorella C, et al. Effect of mechanical ventilation on inflammatory mediators in patients with acute respiratory distress syndrome. JAMA 1999;282:54–77.
21. Esteban A, Alia I, Gordo F, et al. Prospective randomized trial comparing pressure-controlled ventilation and volume-controlled ventilation in ARDS. Chest 2000;117:1690–6.
22. Amato MB, Barbas CS, Medeiros DM, et al. Effect of a protective-ventilation strategy on mortality in the acute respiratory distress syndrome. N Engl J Med 1998;338:347–54.
23. Brochard L, Roudot-Thoraval F, Roupie E, et al. Tidal volume reduction for prevention of ventilator-induced lung injury in acute respiratory distress syndrome. Am J Respir Crit Care Med 1998;158:1831–8.
24. Stewart TE, Meade MO, Cook DJ, et al. Evaluation of a ventilation strategy to prevent barotrauma in patients at high risk for acute respiratory distress syndrome: pressure and volume limited ventilation strategy group. N Engl J Med 1998;338:355–61.
25. Brower RG, Shanholtz CB, Fessler HE, et al. Prospective, randomized controlled clinical trial comparing traditional versus reduced tidal volume ventilation in acute respiratory distress syndrome patients. Crit Care Med 1999;27:1492–8.
26. The acute respiratory distress syndrome network. Ventilation with lower tidal volumes as compared with traditional tidal volumes for acute lung injury and the acute respiratory distress syndrome. N Engl J Med 2000;342:1301–8.
27. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: a severity of disease classification system. Crit Care Med 1985;13:818–29.
28. Knaus WA, Wagner DP, Draper EA, et al. The APACHE III prognostic system: risk prediction of hospital mortality for critically ill hospitalized adults. Chest 1991;100:1619–36.
29. Gattinoni L, Pelosi P, Crotti S, Valenza F. Effects of positive end-expiratory pressure on regional distribution of tidal volume and recruitment in adult respiratory distress syndrome. Am J Respir Crit Care Med 1995;151:1807–14.
30. Ranieri VM, Mascia L, Fiore T, et al. Cardiorespiratory effects of positive end-expiratory pressure during progressive tidal volume reduction (permissive hypercapnia) in patients with acute respiratory distress syndrome. Anesthesiology 1995;83:710–20.
31. Glasziou PP, Irwing LM. An evidence-based approach to individualising treatment. BMJ 1995;311:1356–9.
32. Cooper AB, Ferguson ND, Hanly PJ, et al. Long-term follow up of survivors of acute lung injury: lack of effect of a ventilation strategy to prevent barotrauma. Crit Care Med 1999;27:2616–21.
33. Orme J, Romney JS, Hopkins RO, et al. Pulmonary function and health related quality of life in survivors of ARDS. Am J Respir Crit Care Med 2003;167:690–4.
34. Sharp SJ, Thompson SG, Altman DG. The relation between treatment benefit and underlying risk in meta-analysis. BMJ 313:735–8.
35. Ranieri VM, Brienza N, Santostasi S, et al. Impairment of lung and chest wall mechanics in patients with acute respiratory distress syndrome: role of abdominal distension. Am J Respir Crit Care Med 1997;156:1082–91.
36. Steinbrook R. How best to ventilate? N Engl J Med 2003;348:1393–401.
37. Gattinoni L, Chiumello D, Russo R. Reduced tidal volumes and lung protective ventilatory strategies: where do we go from here? Curr Opin Crit Care 2002;8:45–50.
38. Monchi M, Bellenfant F, Cariou A, et al. Early predictive factors of survival in the acute respiratory distress syndrome: a multivariate analysis. Am J Respir Crit Care Med 1998;158:1076–81.
39. Stewart LA, Parmar MKB. Meta-analysis of the literature or of individual patient data: is there a difference? Lancet 1993;341:418–22.