Background

Veno-arterial ExtraCorporeal Membrane Oxygenation (V-A ECMO) as a bridge to recovery or long-term assist devices in severe cardiocirculatory failure typically lasts 5–7 days. There is scarce guidance on which patients benefit from prolonged support, as prediction models are based on parameters at the start of ECMO. We aimed to characterize the subgroup of patients still alive and on ECMO at day 7, and to develop a multivariable prediction model predicting one-year mortality in this subgroup.

Methods

This multicenter cohort study examines the characteristics and outcomes, including quality of life (QoL), of patients receiving short (1–7 days) and prolonged (> 7 days) V-A ECMO support. In patients still on ECMO at day 7 (the prolonged ECMO group), a multivariable logistic regression model was developed to predict one-year mortality. Model performance was evaluated using Area Under the Receiver Operating Characteristic curve (AUROC), Brier score and calibration plot, following internal validation performed via bootstrapping.

Results

Out of 546 patients, 160 (29%) underwent prolonged V-A ECMO support. These patients had lower Acute Physiology And Chronic Health Evaluation (APACHE) IV scores (mean 79 [SD 29] vs. 92 [SD 36]) and differed regarding ECMO indication. One-year mortality, (1-7days: 61%, >7days: 58%), and health related QoL at one year (1–7 days: Median EQ-5D index score [IQR] 0.82 [0.71, 0.91], > 7 days: 0.82 [0.73, 0.91]) were comparable between both groups. A multivariable logistic regression model predicting one-year mortality was constructed with age, APACHE IV, pre-ECMO cardiac arrest, acute kidney injury during first week; and cumulative Sequential Organ Failure Assessment (SOFA), and pulse pressure at day 7. The internally validated model showed moderate performance, with an AUROC of 0.68 and a Brier score of 0.22.

Conclusion

Patients with prolonged V-A ECMO differ in terms of indication and baseline risk scores. Mortality was not higher in patients who received ECMO support for longer than 7 days, and QoL at one year was good in both groups. Our developed multivariable prediction model showed moderate performance in predicting one-year mortality for those still on ECMO at day 7, and may help in multidisciplinary decision-making regarding continuation of ECMO support pending external validation.