Cardinality Matching Study of Early v.s. Delayed VV ECMO in Severe Respiratory Failure

VV ECMO is increasingly used in refractory respiratory failure. Despite advances in lung protective ventilation strategies, patients who have severe respiratory failure often develop complications from mechanical ventilation, including volutrauma and barotrauma. ECMO allows gas exchange to occur extracorporeally and may reduce the potential burden of iatrogenic lung injury by allowing a reduction of volume and pressure support - a 'lung rest' strategy. It is theorised that earlier intitiation of ECMO may allow for better outcomes, as there will have been less time for iatrogenic lung injury to occur.

The UK ECMO registry has been collected of patients treated under the NHS England commissioned respiratory ECMO service since 2011. This study has been previously registered and publication is intended shortly.

Patients will be extracted from this registry if they received VV ECMO. Propensity matching scores will be created and patients will be stratified into groups of 'early' vs 'delayed' ECMO, based on their probability of being in either group.

Patients will be divided into cohorts based on the median PaO2/FiO2 ratio at decision to cannulate ('less severe hypoxaemia') and ('very severe hypoxaemia'). Matched cohorts will be created correcting for key confounding factors (age, primary diagnosis, duration of pre-ECMO ventilation and PaCO2), using cardinality matching (a novel technique described by Zubizaretta et al. in 2014) and traditional propensity-score-based methods.

The technique with greater balance and statistical power (as defined by sample size) will be selected for the primary analysis.

Further analyses will assess the relationship between hypoxaemia at decision-to-cannulate and confounding factors as above.