Venous blood gases in the assessment of respiratory failure in patients undergoing sleep studies: a randomized study.

STUDY OBJECTIVES: Venous blood gases (VBGs) are not consistently considered suitable surrogates for arterial blood gases (ABGs) in assessing acute respiratory failure due to variable measurement error. The physiological stability of patients with chronic ventilatory failure may lead to improved agreement in this setting.

METHODS: Adults requiring ABGs for sleep or ventilation titration studies had VBGs drawn before or after each ABG, in a randomized order. Veno-arterial correlation and agreement were examined for carbon dioxide tension (PCO2 ), pH, oxygen tension (PO2 ) and oxygen saturation (SO2 ).

RESULTS: We analyzed 115 VBG-ABG pairs from 61 patients. Arterial and venous measures were correlated (with p<0.05) for PCO2 (r=0.84) and pH (r=0.72), but not for PO2 or SO2 . Adjusted mean veno-arterial differences (95% limits of agreement) were +5.0mmHg (-4.4 to +14.4) for PCO2 ; -0.02 (-0.09 to +0.04) for pH; -34.3mmHg (-78.5 to +10.0) for PO2 ; and -23.9% (-61.3 to +13.5) for SO2 . VBGs obtained from the dorsal hand demonstrated a lower mean PCO2 veno-arterial difference (p<0.01). A venous PCO2 threshold of ≥45.8mmHg was >95% sensitive for arterial hypercapnia, so measurements below this can exclude the diagnosis without an ABG. A venous PCO2 threshold of ≥53.7mmHg was >95% specific for arterial hypercapnia, so such readings can be assumed diagnostic. The area under the receiver operating characteristic curve of 0.91 indicated high discriminatory capacity.

CONCLUSIONS: A venous PCO2 <45.8mmHg or ≥53.7mmHg would exclude or diagnose hypercapnia, respectively, in patients referred for sleep studies, but VBGs are poor surrogates for ABGs where precision is important.

CLINICAL TRIAL REGISTRATION: Registry: Australian New Zealand Clinical Trials Register; Name: A comparison of arterial and blood gas analyses in sleep studies; Identifier: ACTRN12617000562370; URL https://www.anzctr.org.au/Trial/Registration/TrialReview.aspx?id=372717.