Abstract Or109: Lactate-Pyruvate Ratio is Associated with Noninvasive Optical Measures of Cerebral Oxygenation in an Experimental Pediatric Model of Cardiac Arrest

Abstract

Introduction: Early and accurate prediction of neurological injury during cardiopulmonary resuscitation (CPR) is vital for guiding clinical decisions and improving outcomes following cardiac arrest. Previous studies have shown the utility of non-invasive optical neuromonitoring as an indicator of CPR quality. This study further examines the utility of intra-arrest optical metrics of cerebral oxygenation, blood volume, and microvascular flow to predict post cardiac arrest cerebral metabolic injury. Hypothesis: Non-invasive optical neuromonitoring during CPR predicts cerebral microdialysis lactate-pyruvate ratio (LPR). Methods: A total of 46 pediatric swine ( Sus scrofa , 8–12 kg) underwent asphyxia-associated cardiac arrest followed by CPR. Continuous optical waveforms were recorded on the left forehead during CPR, including tissue oxygen saturation (StO 2 ), total hemoglobin concentration (THC), oxyhemoglobin concentration (HbO 2 ), and blood flow index (BFI) signals. Waveforms were block-averaged in non-overlapping 15-second windows. A cerebral microdialysis catheter was inserted in the contralateral hemisphere to a depth of 1-1.5cm. Microdialysis perfusate were obtained between ROSC time and 1h after ROSC, and peak LPR during this period was measured. To evaluate time dependency, we modeled multiple temporal CPR segments: 2–4, 2–6, 2–8, 2–10, 2–12, and 2–14 minutes. For each segment, the mean and slope (linear trend) of each optical feature were used to train extreme gradient boosting (XGB) regression models to predict post-ROSC peak LPR. Five-fold cross-validation was employed, and model performance was assessed using the coefficient of determination (R 2 ) from stacked predictions. Results: Mean StO 2 and HbO 2 were the most predictive univariate features (Table 1). Mean StO 2 explained 27–28% of the variance in the 2–8 and 2–10 minute segments, while mean HbO 2 explained 27–30% of LPR variance across all segments except 2–10 minutes. The best-performing multivariate model combined all optical features and achieved an R 2 of 0.40 using the 2–8 minute segment. Conclusion: Optical measurements during CPR, particularly indicators of cerebral oxygenation (StO 2 , HbO 2 ), show promise as non-invasive predictors of acute cerebral metabolic injury following cardiac arrest. These measures outperformed markers of blood volume (THC) and microvascular flow (BFI), supporting their potential role in real-time monitoring and decision-making during and following resuscitation.