Evoked potentials improve multimodal prognostication after cardiac arrest

Abstract Aim Predicting recovery in comatose post-cardiac arrest patients requires multiple modalities of prognostic assessment. In isolation, absent N20 cortical responses in somatosensory evoked potentials (SSEPs) are a specific predictor of poor outcome. It is unknown whether SSEP results, when assessed in the context of prior knowledge (demographic and clinical information), change the pretest predicted probability of recovery. Methods In a single center retrospective study, a cohort of 323 patients admitted to post-cardiac arrest service at a tertiary care center were classified into a group based on SSEP testing. We built adjusted logistic regression models including clinical examination findings on the day SSEPs were recorded to generate a pre-test outcome probability for awakening, withdrawal of life-sustaining therapy (WLST) and survival to discharge. We then added the upper extremity N20 cortical response results to the model to obtain updated outcome probabilities. ROC curve was used to determine the additive effect of using SSEPs to the model. Survival to discharge, awakening, and WLST due to neurological reasons were designated as primary, secondary and tertiary outcomes, respectively. Results Analyses showed that evoked potentials are ordered in sicker patients. Adding SSEP to the model increased the proportion of patients with less than 1% and 5% chance of survival, as well as the proportion of patients with over 95% chance of WLST. AUC for survival increased from 0.85 to 0.93 when SSEP was included (p = 0.006). Conclusion Adding the N20 SSEP response results to prior knowledge changed the predicted probability of WLST and survival to discharge in comatose post-arrest patients.