Methods
Population: We performed a retrospective analysis of ECMO patients admitted to the Johns Hopkins Hospital between 2016 and 2019. We included all adult patients (age >18 years), excluding patients who underwent multiple runs to minimize the potential bias resulting from severe illness. This study was carried out with the approval of the Johns Hopkins institutional review board (IRB00216321). Informed consent and clinical trial registration are not applicable for this study. Demographics, past medical history, laboratory values, ECMO indication, hemodynamic data, and Sequential Organ Failure Assessment (SOFA) score (day 1) were collected.11 ABI included ischemic stroke, intracranial hemorrhage (ICH), cerebral edema, hypoxic ischemic brain injury, seizures, and coma (Glasgow Coma Scale<8) despite being off sedation for >24 hours.12 Systemic hemorrhage included surgical site bleeding, gastrointestinal bleeding, disseminated intravascular coagulation, pulmonary hemorrhage, and genitourinary bleeding.
Exposure: Temperature measurements were collected during the first 24 hours after ECMO cannulation. Bladder temperatures were collected hourly, while additional recordings were obtained when clinically indicated. Metrics for hypothermia included: mean temperature and time (hours) spent between 32-36°C.5
Outcomes: The primary outcome was neurological function at discharge defined by modified Rankin Scale (mRS), with a score of 0-3 defined as good neurological outcome (4-6: poor).13Secondary outcomes included in-hospital mortality and systemic hemorrhage.
Statistical analysis: Demographic and clinical variables for hypothermia vs. no hypothermia were compared by unpaired t-test, chi-squared test, or Mann-Whitney U test as appropriate. Results were expressed as the mean with standard deviation if normally distributed or as the median with quartiles if non-normally distributed for quantitative variables and as proportions for categorical variables. Carefully selected pre-specified variables a priori that showed significant association with primary outcome in univariate analyses were used in the multivariable models reported. A p value <0.05 was considered statistically significant. Odds ratios (ORs) with 95% confidence intervals (CIs) were reported. All analyses were carried out in STATA 15 (College Station, TX, USA).