We tested the agreement between non-invasive measurement of intrapulmonary shunt, using oxygen uptake and pulmonary capillary blood flow measurement obtained by nitrous oxide rebreathing, with that measured using mixed venous blood sampling. Nine patients were recruited pre- and post-cardiac surgery resulting in 20 sets of measurements overall. Mean shunt fraction was 12.5%, and bias between methods (+/-95% confidence limits) was -0.7% (+/-0.8%). The standard deviation of the difference was 1.7% with limits of agreement between the two methods of +2.6% and -3.9%. Correlation coefficient r was 0.90. Agreement with the invasive standard was less accurate and precise where cardiac output was measured by bolus thermodilution (mean bias +1.6%, standard deviation of the difference 2.2%, limits of agreement between the two methods of +5.8% and -2.8%, r = 0.86). Good agreement was demonstrated between the non-invasive method and the invasive reference standard.