Hyperoxia reflects poor native cardiac function and demonstrates cardiac index-dependent associations with mortality in VA-ECMO patients.

BACKGROUND

Multiple VA-ECMO studies have recently reported that early hyperoxia is associated with increased mortality, yet the underlying mechanism is unclear. As prior analyses did not account for native cardiac output or fraction of delivered oxygen, it remains unknown whether hyperoxia is directly injurious or primarily a surrogate of poor intrinsic cardiac function and proximal mixing.

OBJECTIVES

This study aims to evaluate whether hyperoxia is an independent predictor of mortality in VA-ECMO patients or whether it primarily reflects impaired native cardiac function.

METHODS

We conducted a single-center retrospective cohort study of 227 VA-ECMO patients to evaluate the relationship between hyperoxia and ICU mortality, accounting for native cardiac index using multivariable regression and principal component analysis. Average PaO2 and cardiac index were analyzed as continuous variables.

RESULTS

Decreased cardiac index was the strongest independent predictor of ICU mortality in a multivariable regression (OR 0.22; 95% CI 0.06-0.70; p = 0.018). After adjusting for native cardiac index and utilization of advanced therapies (i.e., heart transplant or LVAD), the association between hyperoxia and mortality became insignificant. Average PaO2 and cardiac index were strongly inversely collinear (Variance Inflation Factor (VIF) = 7.2 and 11.5, respectively), with their interaction term yielding a VIF of 20.3. Non-survivors and survivors who received advanced therapies share similar PC1 score distributions (a weighted composite of average PaO2 and cardiac index) and impaired hemodynamic profiles, including reduced ejection fraction, pulse pressures, and oxygen delivery indices.

CONCLUSIONS

The association between early hyperoxia and mortality in VA-ECMO is primarily driven by underlying cardiac failure rather than direct oxygen toxicity. The availability of advanced therapies further modifies this relationship, underscoring the need to account for both native cardiac output and heart replacement strategies when evaluating optimal PaO2 targets and the independent effect of hyperoxia on VA-ECMO outcomes.