Abstract
BACKGROUND: The unifying goal of lung-protective ventilation strategies in ARDS is to minimize the strain and stress applied by mechanical ventilation to the lung to reduce ventilator-induced lung injury (VILI). The relative contributions of the magnitude and frequency of mechanical stress and the end-expiratory pressure to the development of VILI is unknown. Consequently, it is uncertain whether the risk of VILI is best quantified in terms of tidal volume (VT), driving pressure (ΔP), or mechanical power.
METHODS: The correlation between differences in VT, ΔP, and mechanical power and the magnitude of mortality benefit in trials of lung-protective ventilation strategies in adult subjects with ARDS was assessed by meta-regression. Modified mechanical power was computed including PEEP (Powerelastic), excluding PEEP (Powerdynamic), and using ΔP (Powerdriving). The primary analysis incorporated all included trials. A secondary subgroup analysis was restricted to trials of lower versus higher PEEP strategies.
RESULTS: We included 9 trials involving 4,731 subjects in the analysis. Odds ratios for moderation derived from meta-regression showed that variations in VT, ΔP, and Powerdynamic were associated with increased mortality with odds ratios of 1.24 (95% CI 1.03–1.49), 1.31 (95% CI 1.03–1.66), and 1.37 (95% CI 1.05–1.78), respectively. In trials comparing higher versus lower PEEP strategies, Powerelastic was increased in the higher PEEP arm (24 ± 1.7 vs 20 ± 1.5 J/min, respectively), whereas the other parameters were not affected on average by a higher PEEP ventilation strategy.
CONCLUSIONS: In trials of lung-protective ventilation strategies, VT, ΔP, Powerelastic, Powerdynamic, and Powerdriving exhibited similar moderation of treatment effect on mortality. In this study, modified mechanical power did not add important information on the risk of death from VILI in comparison to VT or ΔP.
Introduction
The unifying goal of lung-protective ventilation strategies in ARDS is to minimize the strain and stress applied by mechanical ventilation to the injured lung, thereby reducing ventilator-induced lung injury (VILI).1 Compelling evidence gathered in the last 20 years has shown that mechanical ventilation with low tidal volume (VT) improves survival of patients with ARDS.2,3 Recently, driving pressure (ΔP, normalizes VT for compliance, a surrogate of the size of the “baby lung”) has been shown to be a better predictor of mortality than VT.4 The exact mechanism through which these parameters exert injury on the lung, however, is not clear.5 Although it is widely accepted that stress and strain resulting from tidal volume and applied end-expiratory pressure play a key role in the development of VILI, additional components of the mechanical breath (ie, flow and breathing frequency) also appear to contribute to tissue damage.6-8
Both VT and ΔP neglect the impact of flow and breathing frequency on the development of VILI. Mechanical power has been suggested as a summary variable that incorporates all of these parameters into a single, overarching entity. Mechanical power represents the energy applied to the respiratory system to achieve a change from resting lung volume over 1 min.9 In other words, it characterizes the relationship between the sum of pressures applied to the respiratory system and the respective changes in lung volume they generate, accounting for the potential influences of flow and breathing frequency as well as VT and ΔP on VILI. The association between mechanical power and VILI is not new, and it has been demonstrated in animal experiments.10 Recently, retrospective observational studies in humans have also linked mechanical power to mortality in subjects with and without ARDS.11-13 However, mechanical power also carries a number of limitations, such as its lack of adjustment to the actual functional lung size and the controversial relationship between PEEP, power, and the risk of VILI.14 It is therefore uncertain whether the risk of VILI is best quantified in terms of VT, ΔP or, mechanical power.15
We sought to explore this question by quantifying the extent to which the effect of lung-protective ventilation on mortality can be explained by its effect on each parameter in a meta-regression of clinical trials of lung-protective ventilation strategies in ARDS.
QUICK LOOK
Current knowledge
Tidal volume (VT) and driving pressure (ΔP) play a key role in the development of ventilator-induced lung injury (VILI), yet additional components of the mechanical breath, such as breathing frequency, may also contribute to tissue damage. Both VT and ΔP neglect the effect of breathing frequency on the development of VILI. Mechanical power represents the energy applied to the respiratory system to achieve a change from resting lung volume, accounting for breathing frequency and flow as well as ΔP and VT.
What this paper contributes to our knowledge
In a meta-regression of trials evaluating lung-protective ventilation strategies in subjects with ARDS, a modified version of mechanical power did not provide additional information about the risk of death from VILI in comparison to VT or ΔP.
Methods
Search Strategy and Inclusion Criteria
We consulted a recent systematic review on the use of mechanical ventilation in subjects with ARDS16 and performed a comprehensive search of MEDLINE and PubMed. The search combined Medical Subject Headings and specific key words to identify trials that evaluated different lung-protective ventilation strategies (eg, lower vs higher VT ventilation, lower vs higher PEEP ventilation, and lung recruitment maneuvers) in adult subjects with ARDS, that reported mortality as primary end point, and provided information on the variables required to compute mechanical power on day 1 after randomization. Trials of prone positioning, neuromuscular blockade, extracorporeal life support, and nonconventional modes of ventilation such as high-frequency oscillatory ventilation were excluded because we deemed it unlikely that the mechanism of benefit in these trials could be quantified in terms of VT, ΔP, or mechanical power. This study was performed at the Interdepartmental Division of Critical Care Medicine, University of Toronto.
Data Abstraction
Two reviewers (JD, MT) independently performed the data abstraction process using a specific data-recording form. Disagreements were resolved by consulting with a third reviewer (ECG). VT, ΔP, and Power on day 1 in both treatment and control arms were recorded.
Mechanical Power Computation
Considering that the effect of flow in the generation of VILI is unclear because most of the energy generated by the resistive component of the respiratory system dissipates throughout the proximal airway, we excluded airway resistance from the formula proposed by Gattinoni et al9 to compute a modified mechanical power (Powerelastic) based purely on elastic characteristics of the respiratory system:
where f is breathing frequency and ELrs is the elastance of the respiratory system. To test the hypothesis that static strain (ie, PEEP-derived strain) has less of an influence in the development of VILI, we also computed modified dynamic mechanical power (Powerdynamic) by considering PEEP = 0 cm H2O in this equation:
Finally, driving power (Powerdriving) was calculated with the formula proposed by Marini and Jaber5:
where CRS is the compliance of the respiratory system and 10 represents a predicted population averaged normal respiratory-system compliance (0.1 L/cm H2O) suggested by Marini and Jaber5 to adjust for the observed “baby lung.” Of note, individual patient respiratory-system compliance would be modified by age, gender, height and other conditions.
Statistical Analysis
The effect size of these 5 potential moderators (VT, ΔP, Powerelastic, Powerdynamic, and Powerdriving) on the magnitude of the treatment effect in each trial (odds ratio for mortality at 28 d or in the ICU, whichever is the primary outcome in the individual trial) was assessed by meta-regression of the trials identified in the systematic review. The primary analysis incorporated all included trials. Because PEEP could have a different effect on the development of VILI depending on how it modifies these parameters, a secondary subgroup analysis was restricted to trials of lower versus higher PEEP strategies. Finally, we also examined the association between each variable and the risk of death across trials using generalized linear mixed-effects regression (GLMER). All statistical analyses were conducted in R 3.4.4.
Results
The literature search generated 2,090 titles, of which 2,072 were excluded; 18 studies were selected for full-text review. Of these, 9 met exclusion criteria or had missing data,17-25 for a total of 9 trials2,3,26-32 involving 4,731 subjects for analysis (Fig. 1). Characteristics of the selected trials are described in Table 1. We observed moderate heterogeneity between the included studies (I2 = 55.8%).
Flow chart.
Characteristics of Included Trials
In these 9 trials, the effect of the intervention on each moderator varied widely between trials (Fig. 2). Odds ratios for moderation derived from meta-regression showed that variations in VT, ΔP, and Powerdynamic from the trial interventions were associated with increased mortality (Table 2). Of these, Powerdynamic exhibited the highest odds ratio, but 95% CIs overlapped substantially for all moderators. The association between treatment effect size and Powerelastic or Powerdriving was the weakest. Similar results were observed in the GLMER analysis, although the effect of Powerdriving showed narrower confidence intervals (Table 2).
Moderating effect of tidal volume, driving pressure, mechanical power, dynamic power, and driving power on mortality effect size in trials of lung-protective ventilation strategies.
Predictors of Treatment Effect and Mortality in Trials of Lung-Protective Mechanical Ventilation
In the subset of trials comparing higher versus lower PEEP strategies, Powerelastic was increased in the higher PEEP arm. On average, the reduction in ΔP and modified mechanical power parameters obtained from lowering VT was much larger than that obtained by increasing PEEP (Table 3).
Differences in Parameters Between Intervention and Control Arms
Discussion
In trials of lung-protective ventilation strategies, several parameters of risk for VILI (VT, ΔP, Powerelastic, Powerdynamic, Powerdriving) exhibited similar moderation of treatment effect on mortality. Our findings suggest that a modified mechanical power does not add important information about the risk of death from VILI in comparison to other commonly used static parameters such as VT or ΔP. Unlike the other parameters, Powerelastic did not reach statistical significance, raising the possibility that including PEEP in the computation of mechanical power degrades, rather than enhances, information about risk of VILI.
The relevance of the inspiratory flow and breathing frequency in the development of VILI has been under scrutiny for many years. Animal studies have reported that higher peak inspiratory flow could worsen gas exchange and lung compliance when injurious VT is delivered.7,8,33 This association, however, has not been replicated in studies including humans. Most of the pressure generated by flow is spent to overcome the endotracheal tube’s and the proximal airway’s dependent resistance. The relevance of inspiratory flow on the development of VILI, a phenomenon that occurs at the alveolar level, is therefore doubtful.15 Admittedly, higher flows may entail a more rapid rise in alveolar stress that may be injurious, but in the absence of detailed measurements of airway and alveolar mechanics it is impossible to determine the rate of change in alveolar pressure from the flow. On the other hand, breathing frequency is one of the main components associated with mortality in a retrospective study including > 8,000 critically ill subjects.11 Because mechanical power is meant to reflect the cumulative energy delivered to the respiratory system over a period of time, breathing frequency (and not flow) appears to have a greater physiologic rationale in the development of VILI. Nevertheless, our results do not support the hypothesis that breathing frequency provides additional important information when assessing VILI risk at the bedside.
The lack of meaningful differences in effect moderation by different markers of VILI risk may be attributable to the fact that all of the relevant variation in the parameters resulted from variation in VT (because these trials targeted lower VT). Theoretically, the optimal lung-protective strategy should target the parameter that best quantifies the risk of VILI.34 However, other mechanisms such as mechanical heterogeneity within the injured lung might also contribute to VILI35 and this may not be captured by any of the aforementioned parameters.
Importantly, we also found that trials of higher versus lower PEEP did not produce meaningful between-group reductions in VT, ΔP, Powerelastic, or Powerdynamic (Table 3). If these parameters mediate VILI, failure to modify these parameters might explain in part why these interventions consistently failed to significantly modify mortality.26,28-32 Meaningful reductions in ΔP and power following increased PEEP would only be expected in subjects who exhibit substantial lung recruitment. The effect of higher versus lower PEEP on ΔP, mechanical power, and mortality require further study in this specific subgroup.
We observed moderate heterogeneity among the included studies. This enhances the relevance of selecting a meta-regression approach to analyze the data, given that a single summary measure might fail to capture the diversity among different subsets of subjects.36 This is particularly true when evaluating the effect of mechanical power because its value will be more influenced by changes in VT than by changes in PEEP or breathing frequency,9 as shown in our secondary analysis. The goal of meta-regression is to determine whether a given parameter (eg, modified mechanical power) can explain the observed statistical heterogeneity.
This study has several limitations. First, the number of trials is small, which limits statistical power to detect moderation of treatment effect. Individual patient data meta-analysis might provide more information. Second, we used physiological and ventilatory data collected at a single time point on the first day of the randomized trial to compute the parameters of interest, and it is unclear how accurately such data reflect overall exposure to the parameters. Moreover, we did not compute mechanical power using the original formula described by Gattinoni et al9 because the information needed to perform this calculation was not available in all of the reviewed studies. Although neglecting the impact of flow on the risk of VILI has a physiological rationale, the derivations presented in this study should only be considered as alternatives.
Conclusions
In summary, this analysis does not support the hypothesis that mechanical power is a more useful surrogate measure for the risk of death from VILI or for the benefit of lung-protective ventilation strategies in comparison to targeting driving pressure or tidal volume. Individual patient data meta-analysis might provide further information to confirm or refute this conclusion.
Footnotes
- Correspondence: Ewan C Goligher MD PhD, Toronto General Hospital, 585 University Ave, 11-PMB Room 192, Toronto ON M5G 2N2, Canada. E-mail: ewan.goligher{at}uhn.ca
Dr Goligher has disclosed relationships with Getinge and Timpel. Dr Fan has disclosed relationships with ALung Technologies and MC3 Cardiopulmonary. Dr Brochard has disclosed relationships with Medtronic|Covidien, Air Liquide, Philips, Sentec, General Electric, and Fisher & Paykel. The other authors have disclosed no conflicts of interest.
- Copyright © 2021 by Daedalus Enterprises
