The impact of dynamic driving pressure on mortality during VV ECMO—an ELSO Registry Analysis

Dear Editor,

Exposure to higher driving pressure during invasive mechanical ventilation for acute respiratory failure is associated with increased mortality. The Extracorporeal Life Support Organization (ELSO) ECMO Guideline recommends ultra-low tidal volumes [1]—generally defined as <4 mL/kg of predicted body weight (PBW)—though this has not been tested in a randomized trial. By enabling gas exchange independent of the native lung, ECMO may enhance lung-protective ventilation strategies [2]. Whether reductions in driving pressure during venovenous ECMO support are associated with lower mortality is unclear.

We conducted a registry-based, emulated trial using the ELSO Registry (eTable 1). Patients were included if, at ECMO initiation, they were <70 years and were endotracheally intubated for <7 days. We compared the effect of exposure to dynamic driving pressure (DDP) (peak inspiratory pressure – positive end-expiratory pressure) ≤15 cmH₂O vs >15 cmH₂O during the first 24 hours of ECMO support on hospital mortality of adult patients with acute respiratory failure. We estimated average treatment effect (ATE) weights to balance patients on severity of illness using propensity score based inverse probability weighting (Supplemental Online Content). We performed sensitivity analyses using multiple imputation for missing data and examining the effect of DDP >15 cmH₂O on clinically relevant subsets: patients with a primary documented diagnosis of acute respiratory distress syndrome, and/or receipt of neuromuscular blocking agents, and patients with a PaO₂/FiO₂ <100 mmHg.

We identified 9307 patients and 3113 (33%) deaths in our data set from Jan 1^st, 2008-December 31^st, 2021. Inclusion flowchart (eFigure 1) and baseline characteristics (eTable 4) are listed in the Supplement. There was good separation in DDP between groups: ≤15 cmH₂O group: median 11 cmH₂O [Interquartile range (IQR): 10 to 13] and >15 cmH2O group: 20 cmH₂O [17 to 23]; p<0.001. Respiratory rate was lower among patients managed with ≤15 cmH₂O driving pressure (10 [10 to 15] vs 13 [10 to 16] bpm; p<0.001). Mortality was 31% in patients managed with ≤15 cmH₂O driving pressure, and 39% in patients managed with >15 cmH₂O driving pressure (p<0.001).

Following the construction of the propensity scores (eFigure 2), the groups were well balanced (eFigure 3). After adjusting for patient characteristics, hospital mortality was significantly higher in patients exposed to DDP >15 cmH₂O (odds ratio [OR] 1.42 [95% CI 1.29, 1.56]; p<0.001) (Table 1, eFigure 4). The association of DDP >15 cmH₂O with hospital mortality was robust in sensitivity analyses (eTable 6, eTable 7). E-value for the primary outcome was 1.64.

Table 1:

Hospital Mortality Odds Ratios for Driving Pressure >15 cmH₂O

Cohort	Odds ratio1 (95% CI)	Odds ratio2 (95% CI)	E value
All patients	1.41 (1.26,1.59)	1.42 (1.29,1.56)	1.67
ARDS	1.28 (1.08,1.52)	1.33 (1.15,1.57)
NMB	1.36 (1.18,1.58)	1.34 (1.19,1.53)
ARDS & NMB	1.25 (1.01,1.58)	1.26 (1.01,1.54)

¹Unweighted

²ATE after inverse probability of treatment (IPT) weights

Abbreviations: ARDS-acute respiratory distress syndrome; ATE-average treatment effect in the treated; CI-confidence interval; cmH₂O-centimeters of water; NMB-neuromuscular blockade

To our knowledge, this is the first study to utilize a trial emulation approach of DDP to test this hypothesis among patients with venovenous ECMO. Our findings are strengthened by consistent, significant results in subsets, with a 42% higher odds of mortality in the main group, and 26-34% higher odds among the subgroups with DDP >15 cmH₂O. We acknowledge limitations, including potential for residual confounding, limited applicability, the assumption that data were measured without error, the acknowledgement that DDP is a proxy of static DP, the lack of data on prone positioning, and that our analysis measured the association between measurements at baseline, but was not adherence-adjusted and does not consider different treatment durations.

Abbreviations:

ARDS

acute respiratory distress syndrome

ATE

average treatment effect in the treated

CCI

Charlson comorbidity index

CI

confidence interval

cmH₂O

centimeters of water

DDP

dynamic driving pressure

ELSO

Extracorporeal Life Support Organization

GEE

generalized estimating equation

IPT

inverse probability of treatment

NMB

neuromuscular blockade

VV ECMO

veno-venous extracorporeal membrane oxygenation