Abstract 2536: End-tidal Carbon Dioxide is Less Sensitive for Predicting Death in the Setting of Optimized Cardiopulmonary Resuscitation Quality
INTRODUCTION: End-tidal carbon dioxide (ETCO2) is known to correlate with cardiac output and has been shown to predict death from cardiac arrest. However, ETCO2 is likely affected by CPR quality, which is frequently suboptimal in clinical practice. As a result, the discrimination power of ETCO2 in the setting of optimized CPR is not known.
METHODS: We conducted a prospective study of consecutive in-hospital cardiac arrests at an academic medical center between 1/06 and 1/08. A CPR-sensing monitor/defibrillator which provided real-time audio visual feedback regarding CPR deficiencies was used. This device collected ETCO2 via a monitor connected in line with the endotracheal tube. Patient demographics and outcomes were abstracted via chart review. Two-sided t-test and Receiver Operator Characteristics (ROC) were used to compare outcomes and assess ETCO2 timepoints.
RESULTS: ETCO2 data were available for 118 patients, with a mean age of 60±16 years. Fifty-two percent were male and ROSC was achieved in 46%. Mean compression depth was 49±10 mm, with a rate of 109±10 and no flow fraction of 0.12±0.20. Ventilation rate was 13±8. The table⇓ below compares various ETCO2 measurements between survivors and non-survivors. Using the previously validated 20 minute ETCO2 value of ≤10 as a cutoff for predicting death yielded a sensitivity of 30%, specificity of 100%, positive predictive value (PPV) of 100%, and negative predictive value (NPV) of 7%. Increasing the cutoff to less than or equal to 25 mmHg raised the sensitivity to 65% and NPV to 13% while maintaining 100% specificity and PPV.
CONCLUSIONS: We demonstrated higher ETCO2 values than have previously been reported in a setting with carefully monitored and preserved CPR quality. As a result, the sensitivity and NPV of the 20 minute ETCO2 cutoff of ≤ 10 mmHg are considerably lower than previously reported, limiting its utility for predicting death. A higher cutoff may be needed to improve accuracy in the setting of optimized CPR quality.