Mortality in patients with cardiogenic shock supported with V-A ECMO: A systematic review and meta-analysis evaluating the impact of etiology on 29,289 patients

BACKGROUND Venoarterial extracorporeal membrane oxygenation (VA ECMO) is associated with variable outcomes. In this meta-analysis, we evaluated the mortality after VA ECMO across multiple etiologies of cardiogenic shock (CS). METHODS In June 2019, we performed a systematic search selecting observational studies with ≥10 adult patients reporting on short-term mortality (30-day or mortality at discharge) after initiation of VA ECMO by CS etiology published after 2009. We performed meta-analyses using random effect models and used metaregression to evaluate mortality across CS etiology. RESULTS We included 306 studies (29,289 patients): 25 studies on after heart transplantation (HTx) (771 patients), 13 on myocarditis (906 patients), 33 on decompensated heart failure (HF) (3,567 patients), 64 on after cardiotomy shock (8,231 patients), 10 on pulmonary embolism (PE) (221 patients), 80 on acute myocardial infarction (AMI) (7,774 patients), and 113 on after cardiac arrest [CA] (7,814 patients). With moderate certainty on effect estimates, we observed significantly different mortality estimates for various etiologies (p < 0.001), which is not explained by differences in age and sex across studies: 35% (95% CI: 29-42) for after HTx, 40% (95% CI: 33-46) for myocarditis, 53% (95% CI: 46-59) for HF, 52% (95% CI: 38-66) for PE, 59% (95% CI: 56-63) for cardiotomy, 60% (95% CI: 57-64) for AMI, 64% (95% CI: 59-69) for post‒in-hospital CA, and 76% (95% CI: 69-82) for post-out‒of-hospital CA. Univariable metaregression showed that variation in mortality estimates within etiology group was partially explained by population age, proportion of females, left ventricle venting, and CA. CONCLUSIONS Using an overall estimate of mortality for patients with CS requiring VA ECMO is inadequate given the differential outcomes by etiology. To further refine patient selection and management to improve outcomes, additional studies evaluating patient characteristics impacting outcomes by specific CS etiology are needed.