The prognosis of the acute respiratory distress syndrome (ARDS) has improved dramatically within the past decades, with in-hospital mortality rates ranging from 90% in the seventies 1 to approximately 30% in a recent study 2. Reduction of the tidal volume delivered to mechanically ventilated patients, and thus of the stress applied to their lungs, unambiguously contributed to improving outcomes, as demonstrated by the ARDSnet study, which showed a 22% higher survival in patients who received lower (6 mL/kg) than in those who received larger (12 mL/kg) tidal volumes 3. Interestingly, almost one decade before the ARDSnet study was published, the concept of "permissive hypercapnia" 4 had already led to the use of lower tidal volumes by clinicians and well-conducted observational studies had evidenced significant decrease in the mortality of patients suffering from ARDS 5. Indeed, compelling physiological evidence had been drawn from experimental studies that had described the deleterious effects of mechanical ventilation using high peak inspiratory pressures on lungs, regrouped under the term ventilator-induced lung injury (VILI) 678. In addition to this "volutrauma," so-called "low-volume" injury associated with the repeated recruitment and derecruitment of distal lung units has been incriminated in the development of VILI and forms the rationale for the use of positive end-expiratory pressure (PEEP) 91011. We reviewed seminal experimental studies that led to our current understanding of VILI and contributed to the current recommendations in the respiratory support of ARDS patients.

Only 3 years after the first description of ARDS was made 12, Mead et al. developed the conceptual basis for VILI from the analysis of the mechanical properties of the lungs using a theoretical model of lung elasticity 13. They suggested that the forces acting on lung parenchyma can be actually much greater than those applied to the airway, and theorized that the pressure tending to expand an atelectatic region at a transpulmonary pressure of 30 cm H₂O surrounded by fully expanded lung would be approximately 140 cm H₂O 13. In a visionary statement, the authors concluded that "mechanical ventilation, by applying high transpulmonary pressures to heterogeneously expanded lungs, could contribute to the development of lung hemorrhage and hyaline membranes." In 1974, Webb and Tierney demonstrated for the first time that mechanical ventilation could generate lung lesions in intact animals 6. Rats ventilated with peak inspiratory pressures of 30 or 45 cm H₂O developed pulmonary edema within 60 and 20 min, respectively. Interestingly enough, when a 10-cm H₂O PEEP was applied and the level of end-inspiratory pressure kept constant, the amount of lung edema was lessened 6. Although the authors suggested that low tidal ventilation should be used, they recently mentioned that "this article seemed to interest few clinicians or investigators for a decade or more, perhaps because a similar degree of injury in patients was not apparent" and acknowledged that "in retrospect, it seems almost irresponsible that we didn't publicize our concerns that such ventilator patterns might be harmful to humans" 14. Since this seminal publication, our knowledge of VILI has considerably increased 15.

In early publications, the development of pulmonary edema during mechanical ventilation was ascribed to increased lung microvascular pressure, resulting from high lung volume ventilation and surfactant depletion 6716. As a matter of fact, Parker et al. showed that mean lung microvascular pressure increased by 12.5 cm H₂O during ventilation of open-chest dogs at 64 cm H₂O positive inspiratory pressure 16. Indeed, it has been demonstrated that lung inflation decreases interstitial pressure (thus, increases transmural pressure) around extra-alveolar vessels because of the interdependence phenomenon and around alveolar vessels because of surfactant inactivation. Inflating lungs dilates extra-alveolar vessels 17 and during inflation from low transpulmonary pressure, the increase in vessel diameter is such that an effective outward-acting pressure in excess of pleural pressure (1 to 2 cm H₂O for each centimeter of water increase in transpulmonary pressure) expands these vessels 18. Surfactant is inactivated commensurately with the magnitude of tidal volume and duration of ventilation during ventilation of excised lungs 192021. It was demonstrated that the increase in alveolar surface tension resulting from surfactant inactivation leads to increased filtration through alveolar microvessels 222324. However, the magnitude of microvascular pressure changes during lung inflation is modest and insufficient to explain the occurrence of severe pulmonary edema during mechanical ventilation with a high tidal volume. In addition to pressure changes, alterations of alveolo-capillary barrier permeability are involved in the development of pulmonary edema during high-volume ventilation of intact animals and are the most important responsible for VILI. The increase in alveolar epithelial permeability to small hydrophilic solutes has been studied by Egan during static inflation of fluid-filled in situ sheep lobes 2526. The equivalent-pore radius (an index of epithelial permeability) increased from approximately 1 nm at 20 cm H₂O inflating pressure to 5 nm at 40 cm H₂O alveolar pressure. Albumin diffused freely across the epithelium at the highest pressures, indicating the presence of large leaks. Such permeability increases persisted or even increased after cessation of inflation, implying that epithelial injury was irreversible. Unexpectedly, Parker et al. demonstrated in isolated blood-perfused dog lobes that microvascular permeability alterations, as assessed by increased capillary filtration coefficient, also occurred during high peak pressure ventilation (> 45 cm H₂O) 7. It was subsequently demonstrated that, within minutes after the onset of ventilation, rats ventilated with 45 cm H₂O peak pressures exhibited not only macroscopic pulmonary edema (Figure 1) but also a dramatic increase in microvascular permeability assessed by the distribution space of intravenously injected ¹²⁵I-labeled albumin (Figure 2) 8. Electron microscopy studies consistently revealed widespread disruption of epithelial cells leading to denudation of basement membranes and the presence of many gaps in the capillary endothelium (Figure 3A, B) 8. Such findings demonstrated that high-volume mechanical ventilation leads to pulmonary edema of the permeability type.

The increase in transmural microvascular pressure, even if modest, contributes to the severity of pulmonary edema, which may be fulminating during VILI, because any increase in the driving force will have a dramatic effect on edema formation in the face of an altered microvascular permeability 27.

Several studies have shown that protracted mechanical ventilation using high peak pressures led to lung infiltration with neutrophils 4955. Moreover, neutrophil depletion was associated with better gas exchange and less lung injury in a rabbit model of surfactant depletion 56, suggesting that the inflammatory reaction per se could be deleterious. On the other hand, early studies had evidenced that mechanical ventilation could have deleterious systemic effects. For instance, Kolobow et al. demonstrated that sheep ventilated with 50 cm H₂O peak pressure, corresponding to tidal volumes of 50 to 70 mL/kg, died from multiple organ dysfunction within 48 h 36. The biotrauma hypothesis, i.e., lung tissue stretching might result in lung damage solely through the release of inflammatory mediators and leukocyte recruitment, has been put forward to provide an explanation why most patients with ARDS die from multiple organ failure rather than hypoxemia. Tremblay et al. 57 showed in unperfused rat lungs that high tidal volume ventilation (40 mL/kg) with zero end-expiratory pressure resulted in dramatic increases in the lung lavage levels of tumor necrosis factor-α (TNF-α), interleukin-1β (IL-1β), interleukin-6, and macrophage inflammatory protein-2, compared with controls (Figure 8), suggesting that mechanical ventilation can influence the inflammatory/anti-inflammatory balance in the lungs. However, using the same model of unperfused rat lungs, others found only slightly higher IL-1β and MIP-2 bronchoalveolar lavage fluid concentration in rats ventilated with 42 mL/kg tidal volume than in those ventilated with 7 mL/kg tidal volume 5859 (Figure 9). Moreover, there was no difference in TNF-α lung level between both ventilator strategies. The authors concluded that ventilator strategies that injure lungs do not necessarily result in primary production of proinflammatory cytokines in the lungs 59. The two-hit hypothesis has been put forward to reconcile these discrepant findings. Injurious mechanical ventilation may not be sufficient per se to promote intense lung proinflammatory cytokine secretion but will do so in combination with another aggression 6061. For instance, hemorrhagic shock and resuscitation and high PIP ventilation in rats interact to increase lung and systemic release of proinflammatory mediators (Figure 10). The biotrauma hypothesis seemed supported by the finding that ARDS patients ventilated with a protective strategy (i.e., tidal volume of 7 mL/kg and PEEP of 15 cm H₂O determined from analysis of the PV curve) exhibited lower bronchoalveolar lavage fluid and plasma concentrations of inflammatory mediators than patients ventilated with a "control" strategy (i.e., tidal volume of 11 mL/kg and PEEP of 6 cmH₂O) only 36 h after randomization 62. However, this study was not designed to address whether these decreases were associated with improved clinical endpoints and the clinical meaning of these changes in cytokine levels remains elusive. Similarly, modest, although significant, differences were observed in the plasma levels of IL-6 and IL-8 in ARDS patients ventilated with 6 vs. 12 mL/kg tidal volume 63. In contrast, transiently changing ventilatory settings from a low tidal volume (5 mL/kg)-high PEEP (15 cm H₂O) to a high tidal volume (12 mL/kg)-low PEEP (5 cm H₂O) strategy in ALI patients led to an increase in plasma and bronchoalveolar lavage fluid levels of cytokines 64. Intriguingly, most of these were anti-inflammatory cytokines (IL-1 receptor antagonist and IL-10), emphasizing that it remains unclear whether mechanical ventilation affects the balance of cytokines toward pro- or anti-inflammation 65.

Finally, there is growing evidence that ventilator settings may localize or disperse proteinaceous lung edema or bacteria 666768. Indeed, ventilation with high tidal volume and no PEEP promoted contralateral bacterial seeding in a unilateral model of Pseudomonas aeruginosa pneumonia in rats 68. In contrast, ventilation at the same end-inspiratory pressure but with a high PEEP and thus a lower tidal volume prevented such contralateral dissemination. The potential for adverse ventilator patterns to disperse localized alveolar edema to the opposite lung was studied using scintigraphic methods 35. A ^{99 m}Tc-labeled albumin solution was instilled in a distal airway and produced a zone of alveolar flooding that stayed localized during conventional ventilation. Ventilation with high end-inspiratory pressure dispersed alveolar liquid in the lungs. This dispersion began almost immediately after high-volume ventilation was started and was likely the consequence of a convective movement induced by ventilation. Interestingly, PEEP application prevented this spread even when tidal volume was equivalent and thus end-inspiratory pressure higher (Figure 11). In heterogeneously ventilated lungs, fluid transfer may be propelled toward regions of normal compliance. PEEP may prevent fluid dispersion by avoiding lung collapse and stabilizing edema in the distal airways.

Reducing the stress applied to the lungs by lowering tidal volume improved the survival of ARDS patients 3. However, the mortality rate remains high, between 30 and 60% depending on the study 269. Thus, a considerable amount of studies aiming at developing new ventilator strategies or pharmacological treatments of VILI have been published.

As discussed earlier in this article, the prevention of ventilator-associated lung injury during treatment of ALI may theoretically stem from two approaches: 1) easing the strain 96 applied to diseased lungs through the reduction of tidal volume and therefore of end-inspiratory lung volume (which may be evaluated indirectly by inspiratory plateau pressure); and 2) reducing the so-called atelectrauma via an adequate use of PEEP. Whereas the first concept, which is undisputed on physiological grounds, received a resounding illustration with the demonstration of an improved prognosis by a simple reduction of tidal volume in ARDS patients 3, things are far less clear regarding the second. Indeed, as explained above, the concept of repetitive opening and closure or distal airspaces and the putative low lung volume lesions related to insufficient PEEP levels remains debated. This uncertainty formed the basis of three high-quality, independent, randomized, controlled studies 25354, which all failed to demonstrate improved survival with a high PEEP strategy. Although a meta-analysis of these trials showed a reduction of mortality using higher PEEP levels in the most severe population 97, this does not constitute a definite proof but rather a hypothesis that would require another well-conducted, randomized trial to be confirmed. Moreover, the marked heterogeneity of ventilator protocols among these three studies (with resulting considerable differences in plateau pressures between one study 54 and the others 253) casts some doubt on the physiological rationale for this meta-analysis. Interestingly, a recent study using positron emission tomography imaging in ARDS patients to quantify lung metabolism, a surrogate of lung inflammation, showed that lung regions undergoing cyclic recruitment-derecruitment did not exhibit higher metabolism than those continuously collapsed throughout the respiratory cycle, thus questioning the concept of low lung volume injury 98.

Few experimental concepts have led to dramatic changes in clinical practices as the concept of VILI did. The understanding that end-inspiratory distension is the main determinant of VILI led to the reduction of tidal volume and improved the survival of ARDS patients. Yet, whether tidal volume should be set at 6 mL/kg in all patients remains unsettled 99, and there is currently no bedside tool that allows an accurate assessment of the aerated lung volume and thus no way to tailor tidal volume routinely. Rather than sticking to arbitrarily fixed approaches, clinicians should tailor mechanical ventilation according to patient's individual characteristics, including close monitoring of plateau pressure and maybe PEEP titration guided by esophageal pressure monitoring 100. In addition, opinion leaders should strive to convince clinicians not to use excessive tidal volume, which is still unfortunately the case 101. Indeed, two epidemiological studies suggested an association between ventilator settings (i.e., use of a tidal volume > 6 mL/kg or 700 mL 102 and plateau pressure > 30 cm H₂O 103) and the development of ARDS in mechanically ventilated patients who did not meet ARDS criteria upon hospital admission. Similarly, a relationship between the use of large tidal volumes and postoperative respiratory failure has been established after mechanical ventilation in patients undergoing a pneumonectomy, suggesting that ventilation does not need to be protracted to be deleterious 104. Even though the tidal volume to be used in patients with so-called normal lungs has not been defined yet, these studies suggest that protective ventilator settings could prevent the development or ARDS, particularly when a systemic insult is associated.

The story of VILI and its clinical correlates may be viewed as a success of applied physiology 105106. Indeed, based on sound physiology, physicians had started to reduce tidal volume long before this strategy was proved by randomized, controlled trials 5107. The considerable improvement in survival of ARDS patients with a simple and low-cost physiologic approach is at striking contrast with the failure of many randomized trials of expensive drugs to improve the prognosis of critically ill patients 105106. Nevertheless, many questions persist and VILI will likely not soon be out of the spotlight.

ANTU: α-naphthylthiourea; ARDS: acute respiratory distress syndrome; LIP: lower inflection point; PEEP: positive end-expiratory pressure; PV: pressure-volume; UIP: upper inflection point; VILI: ventilator-induced lung injury.

The authors declare that they have no competing interests.

NDP and DD drafted the manuscript, and JDR and GS revised the manuscript.