ABSTRACT
AIMS: To assess the comparability of venous and arterial samples for pH, bicarbonate and potassium measurements in critically ill patients. METHODS: Simultaneous arterial and venous samples from 206 critically ill patients were analysed in duplicate. Coefficients of variation and 95% limits of agreement were calculated for arterial and venous samples. Bland-Altman plots were constructed to assess agreement between sampling sites. RESULTS: The median (range) of arterial pH, bicarbonate concentrations, potassium concentrations and glucose concentrations were 7.40 (7.01-7.56), 25 (9-41) mmol/l, 4.2 (3.1-6.8) mmol/l and 7.4 (3.0-13.5) mmol/l, respectively. Coefficients of variation for arterial and venous pH were both 0.1%, with bias (95% limits of agreement) of -0.01 (-0.03 to 0.01) for arterial and -0.01 (-0.02 to 0.01) for venous samples. The bias (95% limits of agreement) between arterial and venous samples was 0.03 (-0.02 to 0.08). Coefficients of variation for arterial and venous bicarbonate results were 0.8 and 0.7%, respectively, with bias (95% limits of agreement) of 0 (-0.5 to 0.5) mmol/l for both sample types. The bias (95% limits of agreement) between venous and arterial samples was 0 (-1.3 to 1.3) mmol/l. Coefficients of variation for arterial and venous potassium samples were 0.8 and 1.1%, respectively, with bias (95% limits of agreement) of 0 (-0.1 to 0.1) for both sample types. The bias (95% limits of agreement) between venous and arterial samples was 0.1 (-0.4 to 0.6) mmol/l. CONCLUSIONS: A venous blood sample, analysed on a blood gas machine, is sufficiently reliable to assess pH, bicarbonate and potassium concentrations in critically ill patients, suggesting that venous sampling alone is appropriate in the management of diabetic ketoacidosis.
