Non-invasive estimation of venous admixture: validation of a new formula

We have developed a computer program that estimates venous admixture (intra-pulmonary shunt) from four measurements: haemoglobin concentration, end-tidal carbon dioxide tension (PE'CO2), fractional inspired oxygen concentration (FIO2) and pulse oximetry (SpO2). The formula was tested on patients in an intensive therapy unit by using it to estimate shunt while it was measured simultaneously by a standard, invasive method. A total of 101 measurements were made in 29 patients. After correcting the systematic errors in the assumed differences between PE'CO2 and arterial PCO2, and between SpO2 and co-oximetrically measured SaO2, and correcting for a trend in the arteriovenous oxygen concentration difference (C(a-v))2) with shunt, the bias of the non-invasive minus invasive shunt differences was negligible, with no significant dependence on shunt. The limits of agreement were then +/- 16% shunt overall (+/- 13% within patients). When SaO2 was used instead of SpO2, the limits were +/- 11% (+/- 8% within patients).