BACKGROUND: Central venous blood gas (cVBG) values are correlated with arterial blood gas (ABG) values. However, the substitution of cVBG values for ABG values in critically ill patients remains uninvestigated. Thus, we investigated the reliability between cVBG and ABG values and sought to define the conditions that could improve the reliability of cVBG values as a substitute. METHODS: We conducted a prospective comparison of 292 sets of cVBG values and ABG values from 82 subjects admitted to the medical ICU between October 2017-July 2018. Paired cVBG and ABG samples were collected daily during the first 5 d of ICU treatment and on days 8, 15, 22, and 29. Intraclass correlation coefficient (ICC) and Bland-Altman limits of agreement (LOA) were obtained. RESULTS: The ICC between ABG and cVBG was 0.626 for pH, 0.696 for P-CO2, 0.869 for bicarbonate, 0.866 for base excess, and 0.989 for lactic acid. Bland-Altman plots showed clinically unacceptable LOA between all parameters. Subgroup analysis indicated a significant increase in the ICCs of P-CO2 in samples with mechanical ventilation (0.0574-0.735, P = .02) and central venous oxygen saturation (ScvO(2)) >= 70% (0.611-0.763, P = .008). After adjustment, the 95% LOA between ABG and cVBG was 20.06 to 0.07 for pH and 27.09 to 7.05 for PCO2 in mechanically ventilated subjects with ScvO(2) >= 70%. CONCLUSIONS: ABG and cVBG values showed clinically acceptable agreements and improved reliability in mechanically ventilated subjects with ScvO(2) >= 70%. cVBG analysis may be a substitute for ABG analysis in mechanically ventilated patients once tissue perfusion is restored.