Simultaneous measurements of arterial, end‐tidal, and transcutaneous carbon dioxide (Pa CO 2, Pet CO 2, Ptc CO 2, respectively) were obtained in 134 children receiving mechanical ventilation (ages, 2 days to 16 years; mean, 2.5 years). The mean ± SD Pet CO 2± bias (Pa CO 2 − Pet CO 2) was 3.4 ± 6.6 mmHg. When the Pet CO 2 bias was plotted against the Pa O 2/P AO 2 ratio, a change in the scatter was obvious at about 0.3. The Pet CO 2 bias for patients with Pa O 2/P AO 2 under 0.3 was 7.8 ± 7.3 mmHg compared to 0 ± 3.4 in patients with Pa O 2/P AO 2 above 0.3 (P < 0.001). Pet CO 2 differed significantly from Pa CO 2 (P < 0.001) only for patients with Pa O 2/P AO 2 under 0.3. The slope (Pa CO 2 versus Pet CO 2) for these patients was 1.59, while the slope for patients with Pa O 2/P AO 2 above 0.3 coincided with the line of identity (1.00). The mean ± SD Ptc CO 2 bias (Pa CO 2 − Ptc CO 2) was − 1.3 ± 7.2 mmHg. Skin perfusion was recorded at the area close to the transcutaneous CO2 monitor electrode and was defined as normal when capillary refill was below 3 seconds. The Ptc CO 2 bias for patients with normal skin perfusion was −0.2 ± 5.4 mmHg (P = 0.73) compared to −4.1 ± 9.9 for patients with decreased skin perfusion (P = 0.01). The slope of Ptc CO 2 against Pa CO 2 was closer to identity in patients with normal skin perfusion (1.17) than in patients where it was decreased (slope, 1.40). We suggest that Pa CO 2 estimation by both Pet CO 2 and Ptc CO 2 is sufficiently precise and reliable for clinical use in critically ill children. Certain limitations stem from the nature of the techniques. Measurement of alveolar to arterial O2 ratio may improve the precision of Pa CO 2 estimation by capnography; assessment of skin perfusion is important in order to increase the accuracy of the transcutaneous method, especially in critically ill children.