Driving Pressure—The Emperor’s New Clothes*

    loading  Checking for direct PDF access through Ovid

Excerpt

Although mechanical ventilation is the cornerstone of supportive care in patients with acute respiratory distress syndrome (ARDS), paradoxically, it can potentiate lung injury (i.e., ventilator-induced lung injury [VILI]), which can contribute to the development of organ failure and death (1). This led clinicians and researchers to develop ventilatory strategies that mitigate the development of VILI, which have traditionally focused on limiting inspiratory plateau airway pressure (Pplat) and tidal volume (VT). Driving pressure can be calculated at the bedside (ΔP = Pplat – positive end-expiratory pressure [PEEP]) and is the ratio of VT to static respiratory system compliance (ΔP = VT/Crs). Targeting ΔP may be desirable, allowing clinicians to normalize VT to Crs, which may better indicate the size of the “baby” lung available for ventilation, and may be a better predictor of mortality than VT or Pplat. An association between ΔP and mortality has been demonstrated in a number of observational studies (2–5). In this issue of Critical Care Medicine, Villar et al (6) evaluated whether ΔP was superior to other variables in predicting outcome in patients with ARDS. In 778 patients with moderate/severe ARDS, both Pplat and ΔP predicted hospital mortality, although performed slightly better than ΔP in their derivation and validation cohorts. In contrast to the study by Amato et al (3), mortality increased linearly with higher levels of Pplat whether ΔP was lower (< 19 cm H2O) or higher (≥ 19 cm H2O).
Although there is a strong physiologic rationale to support the importance of ΔP in mechanically ventilated patients with ARDS, randomized controlled trials designed to manipulate or optimize these variables based on physiologic rationale in the critically ill have been disappointing—high-frequency oscillatory ventilation is a recent example in patients with ARDS (8). Observational studies of interventions are inherently biased by confounding by indication—the promising results in these studies are frequently not replicated in clinical trials. Furthermore, the studies by Villar et al (6) and Amato et al (3) are not observational studies of interventions, but observational studies of physiologic targets. Along with confounding by indication, these types of studies have at least two other major concerns. First, the observation of a physiologic profile in survivors does not mean that an intervention targeted at achieving that profile will reduce mortality. There are numerous examples of observational studies in the critically ill that have identified independent predictors of mortality, such as hypoalbuminemia, anemia, and oxygen delivery, that demonstrated no benefit (or even harm) when manipulated in subsequent randomized controlled trials. Explanations for these results may include: 1) the target is not in the causal pathway or 2) the intervention itself causes harm.
Second, physiologic variables (e.g., VT, Pplat, and PEEP) are interrelated in more complex ways than simple confounding associations—examples of this include mathematical coupling and physiologic coupling. Mathematical coupling occurs because variables are actually derived from each other—ΔP is a linear combination of Pplat and PEEP. Physiologic coupling occurs because the three primary variables (VT, Pplat, and PEEP) modify each other when changed in complicated ways that are unpredictable—and their association itself may be associated with mortality. For instance, a recruitable patient will have experience a reduction in Pplat for a given VT, and that “recruitability” may be what is associated with lower mortality, regardless of how they are managed.
Further refinements to the concept of ΔP, such as transpulmonary ΔP (i.e., the difference between transalveolar pressure at end inspiration and end expiration), may represent a better surrogate for lung stress, as it would exclude any contribution from the chest wall (4, 9).
    loading  Loading Related Articles