Prediction of neurological outcome after cardiac arrest and targeted temperature management

Sammanfattning: Background: Prediction of neurological outcome in unconscious patients after cardiac arrest (CA) forms the basis for decisions on further level-of-care based on results from clinical neurological examinations, neuroimaging (CT or MRI), neurophysiology (EEG or SSEP) and blood biomarkers of brain injury. Methods must be highly specific to avoid misclassifying patients with possibilities of a good outcome. In 2015, the European Resuscitation Council (ERC) and the European Society of Intensive Care Medicine (ESICM) published an algorithm to identify poor outcome patients after CA. Novel methods to analyse biomarkers of brain injury have become available and may prove superior to the recommended marker Neuron-specific enolase (NSE). Purpose: I) To describe findings on head computed tomography (CT) after CA, and evaluate the prognostic accuracy of generalised oedema. II & III) To explore the prognostic accuracies of 3 novel blood biomarkers after CA: the axonal marker serum Neurofilament light (NFL), the astrocytic Glial fibrillary acidic protein (GFAP) and the neuronal marker Ubiquitin C-terminal hydrolase (UCH-L1). IV) To assess the predictive performance of the ERC/ESICM algorithm and modifications thereof. To describe prognostic accuracies of single and combined prognostic methods as recommended by the ERC/ESICM.Methods: All patients participated in the Target Temperature after Out-of-hospital cardiac arrest (TTM) trial, a prospective international multicentre trial randomising adult patients with CA of presumed cardiac origin to targeted temperature managements of 33°C or 36°C for 24 hours. Papers I and IV are retrospective studies utilizing information from the TTM database. Serum samples studied in Papers II and III were collected prospectively at 24, 48 and 72 hours post-arrest, and stored in a biobank for batch analysis after trial completion. Primary outcome was poor neurological outcome, defined as Cerebral Performance Category Scale 3-5 (severe cerebral disability, vegetative state or death) at 6 months follow-up.Results: I) Early CT examinations ≤24h were usually normal. Subacutely, generalised oedema was the most common finding, and strongly associated with poor outcome. II) Already at 24 hours post-arrest, serum NFL analysed with an ultrasensitive assay (Simoa) predicted poor neurological outcome with higher prognostic accuracy than any prognostic method currently recommended in the ERC/ESICM algorithm. NFL also differentiatied between various levels of brain injury. II) GFAP and UCH-L1 may be useful as early markers 24h after CA, yet at 48 and 72 hours their prognostic accuracies were not superior to neuron specific enolase. GFAP and UCH-L1 were not elevated in hemolysis, which may prove an advantage compared to NSE. IV) The ERC/ESICM algorithm predicted poor outcome without false positive predictions (100% specificity) and identified approximately fourty percent of patients with poor outcome. Any two pathological findings according to the ERC/ESICM criteria predicted poor outcome without false positive predictions, regardless of level of unconsciousness. Withdrawal of life-sustaining-therapy (WLST) was common in the TTM trial and may have influenced our results.Conclusions: Prognostication after CA should always be multimodal. The current ERC/ESICM algorithm safely predicted poor outcome, but could benefit from minor modifications. Serum NFL has the potential to guide treatment decisions, both to predict poor outcome and to identify patients with a presumed good neurological prognosis where further treatment is life-saving.