BACKGROUND: Transcutaneous carbon dioxide (P tcCO 2 ) monitoring is being used increasingly to assess acute respiratory failure. However, there are conflicting findings concerning its reliability when evaluating patients with high levels of P aCO 2 . Our study evaluates the accuracy of this method in subjects with respiratory failure according to the severity of hypercapnia. METHODS: We included subjects with respiratory failure, admitted to a respiratory intermediate care unit, who required arterial blood gas analysis. Simultaneously, P tcCO 2 was measured using a digital monitor. Relations between P aCO 2 and P tcCO 2 were assessed by the Pearson correlation coefficient. BlandAltman analysis was used to test data dispersion, and an analysis of variance test was used to compare the differences between P aCO 2 and the corresponding P tcCO 2 at different levels (level 1, <50 mm Hg; level 2, 50 -60 mm Hg; level 3, >60 mm Hg). RESULTS: Eighty-one subjects were analyzed. The main diagnosis was COPD exacerbation (45%). P tcCO 2 correlated well with P aCO 2 (r2 ؍ 0.93, P < .001). Bland-Altman analysis showed a mean P aCO 2 ؊ P tcCO 2 difference of 4.9 ؎ 4.4 with 95% limits of agreement ranging from ؊3.6 to 13.4. The difference between variables increased in line with P aCO 2 severity: level 1, 1.7 ؎ 3.2 mm Hg; level 2, 3.7 ؎ 2.8; level 3, 6.8 ؎ 4.7 (analysis of variance, P < .001). CONCLUSIONS: Our study showed an acceptable agreement of P tcCO 2 monitoring with arterial blood gas analysis. However, we should consider that P tcCO 2 underestimates P aCO 2 levels, and its accuracy depends on the level of hypercapnia, so this method would not be suitable for acute patients with severe hypercapnia.