This study compares two noninvasive techniques for monitoring the partial pressure of carbon dioxide (PCO2) in 24 anesthetized adult patients. End-tidal PCO2 (PetCO2) and transcutaneous PCO2 (PtcCO2) were simultaneously monitored and compared with arterial PCO2 (PaCO2) determined by intermittent analysis of arterial blood samples. PETCO2 and PtcCO2 values were compared with PaCO2 values corrected to patient body temperature (PaCO2T) and PaCO2 values determined at a temperature of 37 degrees C (PaCO2). Linear regression was performed along with calculations of the correlation coefficient (r), bias, and precision of the four paired variables: PETCO2 versus PaCO2 and PaCO2T (n = 211), and PtcCO2 versus PaCO2 and PaCO2T (n = 233). Bias is defined as the mean difference between paired values, whereas precision is the standard deviation of the difference. The following values were found for r, bias, and +/- precision, respectively. PETCO2 versus PaCO2: 0.67, -7.8 mm Hg, +/- 6.1 mm Hg; PETCO2 versus PaCO2T: 0.73, -5.8 mm Hg, +/- 5.9 mm Hg; PtcCO2 versus PaCO2: 0.87, -1.6 mm Hg, +/- 4.3 mm Hg; PtcCO2 versus PaCO2T: 0.84, +0.7 mm Hg, +/- 4.8 mm Hg. Although each of these PCO2 variables is physiologically different, there is a significant correlation (P less than 0.001) between the noninvasively monitored values and the blood gas values. Temperature correction of the arterial values (PaCO2T) slightly improved the correlation, with respect to PETCO2, but it had the opposite effect for PtcCO2. In this study, the chief distinction between these two noninvasive monitors was that PETCO2 had a large negative bias, whereas PtcCO2 had a small bias. We conclude from these data that PtcCO2 may be used to estimate PaCO2 with an accuracy similar to that of PETCO2 in anesthetized patients.