Pulmonary arteriovenous malformations (PAVMs) are often associated with hereditary haemorrhagic telangiectasia (HHT). The quantification of right-to-left shunts in patients with PAVMs is important in diagnosis and follow up. Traditionally, this shunt is measured by the 100% oxygen method, in which the value for the arteriovenous difference in oxygen content, Cao(2)-CVO(2) (where Cao(2) is the oxygen content of arterial blood and CVO(2) is the oxygen content of mixed venous blood) is estimated. Alternative methods consist of measurement of the systemic or renal uptake of (99m)Tc-labelled macroaggregates of albumin (MAA), which are trapped in pulmonary capillaries, but pass through PAVMs. We first measured Cao(2)-CVO(2) in 12 HHT patients before and after embolization of PAVMs. We obtained a mean value of 4.4 ml/100 ml, instead of the usual 5 ml/100 ml. Subsequently, we measured right-to-left shunt in 21 HHT patients using the 100% oxygen method and with two different methods involving (99m)Tc. We used the kidney-lung method (K/L method), in which it is assumed that the right kidney receives 10% of the cardiac output, and we also used a method with two tracers (HSA/MAA method): (1) (99m)Tc-labelled human serum albumin (HSA) (which passes through pulmonary capillaries) to measure the fraction of the cardiac output perfusing the kidneys, and (2) MAA to measure the shunt fraction. In 35 shunt measurements we evaluated this new technique and the K/L method, by comparing the results with those from the 100% oxygen method. There was poor agreement between the 100% oxygen method and the K/L method, with 95% limits of agreement for the shunt fraction of -15.2% to +15.2%. There was moderate agreement between the 100% oxygen method and the HSA/MAA method, with limits of agreement of -8.3% to +7.7%. We conclude that the different methods cannot replace each other, because the limits of agreement are too wide for clinical use.