Background

Venous blood gas sampling has replaced arterial sampling in many critically ill patients, though interpretation of venous pCO₂ still remains a challenge. Lemoël et al., Farkas and Zeserson et al. have proposed models to estimate arterial pCO₂ based on venous pCO₂. Our objective was to externally validate these models with a new dataset. This was a prospective cross-sectional study of consecutive adult patients with a clinical indication for blood gas analysis in an academic emergency department in Sweden. Agreement of pairs was reported as mean difference with limits of agreement (LoA). Vital signs and lead times were recorded.

Results

Two hundred and fifty blood gas pairs were collected consecutively between October 2021 and April 2022, 243 valid pairs were used in the final analysis [mean age 72.8 years (SD 17.8), 47% females]. Respiratory distress was the most common clinical indication (84% of all cases). The model of Farkas showed the best metrics with a mean difference between estimated and arterial pCO₂ of − 0.11 mmHg (95% LoA − 6.86, + 6.63). For Lemoël the difference was 2.57 mmHg (95% LoA − 5.65, + 10.8), Zeserson 2.55 mmHg (95% LoA − 7.43, + 12.53). All three models showed a decrease in precision in patients with ongoing supplemental oxygen therapy.

Conclusion

Arterial pCO₂ may be accurately estimated in most patients based on venous blood gas samples. Additional consideration is required in patients with hypo- or hypercapnia or oxygen therapy. Thus, conversion of venous pCO₂ may be considered as an alternative to arterial blood gas sampling with the model of Farkas being the most accurate.