OBJECTIVE To describe the use of a transcutaneous blood gas monitoring system in critically ill dogs, determine if transcutaneous and arterial blood gas values have good agreement, and verify if clinical or laboratory variables are correlated with differences between transcutaneous and arterial blood gas measurements. METHODS Prospective observational study. METHODS University teaching hospital ICU. METHODS Twenty-three client-owned dogs. METHODS In critically ill dogs undergoing arterial blood gas monitoring, a transcutaneous blood gas monitor was used to measure transcutaneous partial pressure of carbon dioxide (PtcCO2 ) and transcutaneous partial pressure of oxygen (PtcO2 ) values 30 minutes after sensor placement, which were compared to PaCO2 and PaO2 values measured simultaneously. Clinical and laboratory variables were concurrently recorded to determine if they were correlated with the difference between transcutaneous and arterial blood gas measurements. RESULTS Bland-Altman analysis revealed a mean bias of 4.6 ± 26.3 mm Hg (limits of agreement [LOA]: -46.9/+56.1 mm Hg) between PtcO2 and PaO2 and a mean bias of 9.3 ± 8.5 mm Hg (LOA: -7.5/+26.0 mm Hg) between PtcCO2 and PaCO2 . The difference between PtcCO2 -PaCO2 was strongly negatively correlated with HCO3 (-) (r(2) = 0.52, P < 0.001) and PaCO2 (r(2) = 0.58, P < 0.001) and weakly positively correlated with diastolic blood pressure (r(2) = 0.21, P = 0.044), whereas the difference between PtcCO2 -PaCO2 was moderately negatively correlated with diastolic blood pressure (r(2) = 0.33, P = 0.008). CONCLUSIONS Agreement between transcutaneous and arterial PO2 and PCO2 measurements in these critically ill dogs was inferior to that reported in similar adult and pediatric human studies. The transcutaneous monitor consistently over-estimated PaO2 and PaCO2 and should not be used to replace arterial blood gas measurements in critically ill dogs requiring blood gas interpretation.