Abstract 311: Selection of Optimal Predictor and Critical Thresholds for Return of Spontaneous Circulation Using Non-Invasive Frequency-Domain Diffuse Optical Spectroscopy During Cardiopulmonary Resuscitation
Introduction: We have shown that during CPR, novel non-invasive monitoring of cerebral tissue oxygenation (StO 2 , %) and total hemoglobin concentration (THC, μmol/L) by frequency-domain diffuse optical spectroscopy (FD-DOS) is associated with ROSC in a swine model of pediatric cardiac arrest. Our objective is to find the optimal non-invasive predictor for ROSC and assess feasibility of a stable critical threshold over time in early CPR. Hypothesis: Stable critical thresholds with high sensitivity or specificity for ROSC may be established in early CPR (<10 min) from non-invasive cerebral StO 2 and THC measurements initiated at CPR start. Methods: One-month-old swine (n=31) underwent 7 minutes of asphyxia, induction of ventricular fibrillation, and up to 20 minutes of CPR till ROSC or death (no ROSC). Absolute StO 2 and THC and absolute and relative change from 1 minute into CPR (time for chest molding and FD-DOS placement) were evaluated as ROSC predictors over time. For each variable, an ROC curve and two critical thresholds, maximizing specificity (=1) or sensitivity (=1), were determined at 1-min intervals from 2-10 minutes of CPR using univariate logistic regression. Optimal predictor was selected by highest mean AUC. A stable specificity or sensitivity threshold was feasible if the mean threshold had a specificity or sensitivity >0.9 over all intervals, respectively. Results: Absolute change in StO 2 (ΔStO 2 ) had the highest mean (SD) AUC of 0.90 (0.07). Consistently >0.8, the AUC exceeded 0.9 after 7 minutes of CPR ( see Fig. ). The mean specificity threshold (ΔStO 2 = +5.1%) and sensitivity threshold (ΔStO 2 = +1.4%) achieved an overall specificity of 0.93 and sensitivity of 0.98, respectively. Conclusions: Non-invasive monitoring of absolute change in StO 2 was most predictive of ROSC and stable critical thresholds with high specificity or sensitivity were established in early CPR. Future work will independently validate this promising tool for CPR optimization.