Abstract 290: Early Monitoring Of Neurological Recovery Using Poincare Analysis Index Of Heart Rate Variability - A Potential Biomarker For Asphyxial Cardiac Arrest Related Brain Injury

Abstract

Heart rate variability (HRV) is influenced by autonomic nervous system including subcortical structures such as brainstem. We hypothesized that HRV would reflect graded response to hypoxic ischemic brain injury after cardiac arrest (CA). We investigated HRV based on the Poincare analysis method during the hyperacute recovery period after resuscitation to define its role in assessing neurologic recovery. Adult male Wistar rats (360±20g) were subjected to either 5 /7-min of asphyxial CA and grouped based on neurologic recovery - good (n=17) and bad (n=16), assessed at 4 hr post resuscitation using a standardized Neurological Deficit Scale (NDS). A good outcome (GO) was defined as a NDS sub score (arousal+brainstem) of 36-40 and bad outcome (BO) as 23-35, with 40 as good and 0 as brain dead. A Poincare plot showing successive heart rate intervals was developed and analysed on 5-min ECG segments (first 1 hr) owing to the nonstationary HRV patterns during early recovery period. Several machine learning models were trained to assess the results from the standard deviations (SD1, SD2), to decipher the parasympathetic and sympathetic activities, and evaluated by the area under receiver operating characteristic curve (AUC). GO group demonstrated an NDS of 39.65±0.99, while the BO group had an NDS of (30±2.65) showing a statistical difference between groups (p<0.01). GO group exhibited an immediate increase post resuscitation in SD1 (0.55±0.23) and SD2 (0.67±0.29) with a gradual decrease ~1 hr, SD1 (0.20±0.12) and SD2 (0.13±0.09). No significant differences in initial SD1 (0.30±0.14) and SD2 (0.28±0.16) vs after 1 hr of recovery (SD1 (0.29±0.08), SD2 (0.30±0.16)) were observed across BO group. For GO, all parameters indicated a faster return to baseline values- SD1 (0.21±0.08) SD2 (0.15±0.16), showing a better recovery of the cardiac autonomic control, in agreement with the evolution of NDS. Logistic regression and naïve bayes performed best with SD2, giving an accuracy of 72% (AUC: 0.75) and 75% (AUC: 0.88) for recovery prognostication in 5-fold cross validation. Unlike previous studies to find biomarkers for CA related brain injury focused on cortical assessment, our results indicate that the novel HRV method can fill the gap in assessing subcortical injury after CA.