Accuracy of end-tidal carbon dioxide tension analyzers

Substantial mean differences between arterial carbon dioxide tension (PaCO2) and end-tidal carbon dioxide tension (PETCO2) in anesthesia and intensive care settings have been demonstrated by a number of investigators. We have explored the technical causes of error in the measurement of PETCO2 that could contribute to the observed differences. In a clinical setting, the measurement of PETCO2 is accomplished with one of three types of instruments, infrared analyzers, mass spectrometers, and Raman spectrometers, whose specified accuracies are typically +/- 2, +/- 1.5, and +/- 0.5 mm Hg, respectively. We examined potential errors in PETCO2 measurement with respect to the analyzer, sampling system, environment, and instrument. Various analyzer error sources were measured, including stability, warm-up time, interference from nitrous oxide and oxygen, pressure, noise, and response time. Other error sources, including calibration, resistance in the sample catheter, pressure changes, water vapor, liquid water, and end-tidal detection algorithms, were considered and are discussed. On the basis of our measurements and analysis, we estimate the magnitude of the major potential errors for an uncompensated infrared analyzer as: inaccuracy, 2 mm Hg; resolution, 0.5 mm Hg; noise, 2 mm Hg; instability (12 hours), 3 mm Hg; miscalibration, 1 mm Hg; selectivity (70% nitrous oxide), 6.5 mm Hg; selectivity (100% oxygen), -2.5 mm Hg; atmospheric pressure change, less than 1 mm Hg; airway pressure at 30 cm H2O, 2 mm Hg; positive end-expiratory pressure or continuous positive airway pressure at 20 cm H2O, 1.5 mm Hg; sampling system resistance, less than 1 mm Hg; and water vapor, 2.5 mm Hg. In addition to these errors, other systematic mistakes such as an inaccurate end-tidal detection algorithm, poor calibration technique, or liquid water contamination can lead to gross inaccuracies. In a clinical setting, unless the user is confident that all of the technical error sources have been eliminated and the physiologic factors are known, depending on PETCO2 to determine PaCO2 is not advised.