Brain injury after cardiac arrest - the predictive information of computed tomography

Background and aimHypoxic-ischaemic encephalopathy (HIE) is a leading cause of morbidity and mortality in unconscious patients resuscitated after out-of-hospital cardiac arrest. Many patients die after withdrawal of life-sustaining treatments (WLST), often due to a presumed poor prognosis. To minimise the risk of premature WLST, guidelines recommend using a multimodal prognostication. Head computed tomography (CT) is widely used after cardiac arrest, but the evidence supporting its reliability as a predictor of poor outcome remains limited. This thesis aims to evaluate qualitative and quantitative CT indicators of HIE to strengthen the evidence for CT as a prognostic tool in unconscious cardiac arrest patients.MethodsI) Post-hoc analysis of a prospective multicentre study, the Target Temperature Management at 33 °C versus 36 °C after Cardiac Arrest (TTM) trial. Adult patients from Swedish sites with CTs were included. Two blinded radiologists assessed early (<24 hours) and late (≥24 hours) CTs with various qualitative- and quantitative (grey- white matter ratio (GWR)) methods to predict poor outcome, defined as a modified Rankin scale (mRS) score of 4-6 (moderate severe disability, severe disability, or death) at six months. (II) Protocol to establish pre-specified radiological criteria for identifying signs indicative of HIE on CT, aimed at predicting poor functional outcomes after cardiac arrest. (III) Prospective international multicentre sub-study of the Hypothermia versus Normothermia after Out-of-hospital Cardiac Arrest (TTM2) trial. CTs performed >48 hours ≤7 days after cardiac arrest were assessed with manual (standardised qualitative and basal ganglia GWR) and automated atlas-based GWR. Prognostic performance for poor outcome prediction (mRS 4-6 at six months) for the qualitative assessment and for the pre-defined GWR cut-off <1.10 was calculated. Inter- and intrarater agreement were analysed. IV) In-depth analysis of prognostic performance and interrater agreement for individual CT items included in the standardised qualitative assessment in Paper III. V) Retrospective single-centre study. GWR at the basal ganglia level was calculated using two different-sized Regions of Interests (ROIs) by three raters on CTs without current significant pathology, in an age- and sex- matched cohort to a cardiac arrest population.ResultsI) N=106. All tested CT assessment methods had better predictive performance for poor outcome on late compared to early CTs. A GWR cut-off <1.10 showed 100% specificity across all methods and raters. The highest sensitivity at this cut-off, 50- 63%, was achieved on late CTs with the GWR basal ganglia 8 ROIs. III) N=140. Standardised qualitative CT assessment and all GWR methods at cut-off <1.10 predicted poor outcome with 100% specificity. Median sensitivity for the seven raters was: 37% (qualitative), 39% (GWR <1.10 8 ROIs), 30% (GWR <1.10 4 ROIs), and 41% (automated GWR <1.10). The highest interrater agreement was achieved with the GWR <1.15 8 ROIs method, kappa 0.83. IV) Loss of grey-white matter distinction predicted poor outcome with 100% specificity and 45-50% sensitivity. The specificity for sulcal effacement was 93-99% and the sensitivity 29- 49%. The highest interrater agreement was achieved with loss of grey-white matter distinction at the high convexity level, kappa 0.74. V) N=155. No participant had GWR <1.10, regardless of ROI size. Median GWR ranged from 1.30 to 1.32 (0.1 cm2 ROIs) and from 1.27 to 1.32 (0.2 cm2 ROIs). Variability between raters and ROI sizes was ± 0.1.ConclusionsCT is a highly specific tool for predicting poor functional outcomes after cardiac arrest and should be considered in patients who remain unconscious >48 hours post- arrest, as part of a multimodal neuroprognostication strategy.Combining a structured qualitative assessment of definite severe HIE with a GWR <1.10, assessed manually or via an automated method at the basal ganglia level, enables prediction of poor functional outcome with high specificity and moderate sensitivity.Improving interrater agreement will require further refinement of standardised qualitative assessment, with focus on the robust predictive marker of loss of grey- white matter distinction.