Prediction and risk stratification of survival in accidental hypothermia requiring extracorporeal life support: An individual patient data metaanalysis

Background: Extra-corporeal life support (ECLS) is a life-saving intervention for patients with hypothermia induced cardiac arrest or severe cardiovascular instability. However, its application is highly variable due to a paucity of data in the literature to guide practice. Current guidelines and recommendations are based on expert opinion, single case reports, and small case series. Combining all of the published data in a patient-level analysis can provide a robust assessment of the influence of patient characteristics on survival with ECLS. Objective: To develop a prediction model of survival with good neurologic outcome for accidental hypothermia treated with ECLS. Methods: Electronic searches of PubMed, EMBASE, CINAHL were conducted with a hand search of reference lists and major surgical and critical care conference abstracts. Studies had to report the use of ECLS configured with a circuit, blood pump and oxygenator with an integrated heat exchanger. Randomized and observational studies were eligible for inclusion. Non-human, non-English and review manuscripts were deemed ineligible. Study authors were requested to submit patient level data when aggregate or incomplete individual patient data was provided in a study. Survival with good neurologic outcome was categorized for patients to last follow-up based on the reported scores on the Cerebral Performance Category (1 or 2), Glasgow Outcome Scale (4 or 5) and Pediatric Overall Performance Category (1 or 2). A one-stage, individual patient data meta-analysis was performed with a mixed-effects multi-level logistic regression model reporting odds ratio (OR) with a 95% confidence interval (CI). Results: Data from 44 observational studies and 40 case reports (n = 658) were combined and analyzed to identify independent predictors of survival with good neurologic outcome. The survival rate with good neurologic outcome of the entire cohort was 40.3% (265 of 658). ECLS rewarming rate (OR: 0.93; 95% CI: 0.88, 0.98; p = .007), female gender (OR: 2.78; 95% CI: 1.69, 4.58; p < 0.001), asphyxiation (OR: 0.19; 95% CI: 0.11, 0.35; p < 0.001) and serum potassium (OR: 0.62; 95% CI: 0.53, 0.73; p < 0.001) were associated with survival with a good neurologic outcome. The logistic regression model demonstrated excellent discrimination (c-statistic: 0.849; 95% CI: 0.823, 0.875). Conclusions: The use of extracorporeal life support in the treatment of hypothermic cardiac arrest provides a favourable chance of survival with good neurologic outcome. When used in a weighted scoring system, asphyxiation, serum potassium and gender can help clinicians prognosticate the benefit of resuscitating hypothermic patients with ECLS.