We prospectively compared the accuracy of two noninvasive monitors of arterial CO2 (end-tidal and transcutaneous) in mechanically ventilated infants and toddlers with respiratory failure. The study included infants and toddlers less than 48 mo of age who required tracheal intubation and mechanical ventilation for respiratory failure. In each patient, both ETCO2 and transcutaneous CO2 (TC-CO2) were simultaneously monitored and compared with PaCO2 when an arterial blood gas analysis was performed. The cohort for the study included 25 toddlers and infants ranging in age from 1 to 40 mo and in weight from 3.3 to 19.1 kg. A total of 100 sample sets (PaCO2, ETCO2, TC-CO2) was compared. The ETCO2 to PaCO2 difference was 6.8 +/- 5.1 mm Hg, while the TC-CO2 to PaCO2 difference was 2.3 +/- 1.3 mm Hg (P < 0.0001). The absolute difference of the TC-CO2 and PaCO2 was 4 mm Hg or less in 96 of the 100 values, while the ETCO2 to PaCO2 difference was 4 mm Hg or less in 38 of the 100 values (P < 0.0001). Bland-Altman analysis revealed a bias of -0.68 with a precision of +/-2.35 when comparing the TC-CO2 and the PaCO2 and a bias of -6.68 with a precision of +/-5.01 when comparing ETCO2 with PaCO2. In neonates and infants with respiratory failure, TC-CO2 monitoring provided a more accurate estimation of PaCO2 than ETCO2 monitoring.