Trending Ability of End-Tidal Capnography Monitoring During Mechanical Ventilation to Track Changes in Arterial Partial Pressure of Carbon Dioxide in Critically Ill Patients With Acute Brain Injury: A Monocenter Retrospective Study.

BACKGROUND: Changes in arterial partial pressure of carbon dioxide (Pa co2 ) may alter cerebral perfusion in critically ill patients with acute brain injury. Consequently, international guidelines recommend normocapnia in mechanically ventilated patients with acute brain injury. The measurement of end-tidal capnography (Et co2 ) allows its approximation. Our objective was to report the agreement between trends in Et co2 and Pa co2 during mechanical ventilation in patients with acute brain injury. METHODS: Retrospective monocenter study was conducted for 2 years. Critically ill patients with acute brain injury who required mechanical ventilation with continuous Et co2 monitoring and with 2 or more arterial gas were included. The agreement was evaluated according to the Bland and Altman analysis for repeated measurements with calculation of bias, and upper and lower limits of agreement. The directional concordance rate of changes between Et co2 and Pa co2 was evaluated with a 4-quadrant plot. A polar plot analysis was performed using the Critchley methods. RESULTS: We analyzed the data of 255 patients with a total of 3923 paired ΔEt co2 and ΔPa co2 (9 values per patient in median). Mean bias by Bland and Altman analysis was -8.1 (95 CI, -7.9 to -8.3) mm Hg. The directional concordance rate between Et co2 and Pa co2 was 55.8%. The mean radial bias by polar plot analysis was -4.4° (95% CI, -5.5 to -3.3) with radial limit of agreement (LOA) of ±62.8° with radial LOA 95% CI of ±1.9°. CONCLUSIONS: Our results question the performance of trending ability of Et co2 to track changes in Pa co2 in a population of critically ill patients with acute brain injury. Changes in Et co2 largely failed to follow changes in Pa co2 in both direction (ie, low concordance rate) and magnitude (ie, large radial LOA). These results need to be confirmed in prospective studies to minimize the risk of bias.