Accuracy of end-tidal and transcutaneous PCO2 monitoring during sleep

Although it is intuitively desirable, the measurement of arterial carbon dioxide tension (PaCO2) during diagnostic polysomnography and nocturnal trials of positive pressure therapy is invasive and potentially expensive. The accuracy of end-tidal carbon dioxide tension (PETCO2) and transcutaneous carbon dioxide (tcPCO2) monitoring in these contexts has not been systematically evaluated. This investigation was undertaken to evaluate the accuracy of PETCO2 and tcPCO2 in patients undergoing polysomnography.

Values of PETCO2 were compared with PaCO2 in 19 patients spontaneously breathing room air (condition 1), in 13 patients receiving supplemental oxygen via nasal cannula (condition 2), and in 22 patients receiving nocturnal positive pressure ventilatory assistance (all but one with continuous positive airway pressure or bilevel positive airway pressure) (condition 3). The accuracy of tcPCO2 monitoring during sleep was also examined by comparing tcPCO2 values with simultaneously recorded PaCO2 values obtained during sleep in patients undergoing nocturnal polysomnography. Data were collected using three commercially available brands of tcPCO2 monitors (capnograph R, n = 17 patients; capnograph S, n = 17; and capnograph N, n = 15).

Accuracy of PETCO2--There was significant scatter in the PaCO2 vs PETCO2 relationship such that only 23 percent of the variability in PaCO2 was explained by variation of PETCO2 during condition 1 and only 15 percent and 20 percent of the variability in PaCO2 was explained by variation of PETCO2 during conditions 2 and 3, respectively. 21.3 percent of patients had average PETCO2 values in error by > 10 mm Hg during condition 1, while during conditions 2 and 3, 46.2 and 63.7 percent of patients had average values in error by > 10 mm Hg, respectively. Accuracy of tcPCO2--While capnographs S and N generally overestimated PaCO2 with a wide scatter, capnograph R tended to have offsetting overestimations and underestimations of PaCO2 with a wide scatter. With each capnograph, a relatively small portion of the variability of the PaCO2 was explained by variability of the tcPCO2 (r2 = 0.2, 0.45 and 0.64 for capnographs S, N, and R, respectively). Across the three capnographs, 43.1 to 66.7 percent of measurements were in error by > 10 mm Hg, and 5 to 20 percent of measurements reflected errors > 20 mm Hg. There was no consistent relationship between the tcPCO2 error and the level of PaCO2, nor was the tcPCO2 error consistent in individual patients. There was no relationship between tcPCO2 accuracy and body mass index.

Neither PETCO2, measured within a face mask, nor tcPCO2 is a consistently accurate reflection of PaCO2. This limits the utility of these variables in monitoring patients during diagnostic and therapeutic sleep studies, and in particular, during trials of nocturnal ventilatory assistance where adequate levels of support are to be established and unacceptable hyperventilation and respiratory alkalosis must be recognized.