We undertook the first clinical evaluation of a novel, non-invasive device for the continuous measurement of plasma haemoglobin concentration in 25 patients undergoing elective cardiac surgery. At four pre-determined intervals, samples of blood were taken for plasma haemoglobin estimation on a blood gas analyser and a laboratory device and were compared with the plasma haemoglobin estimation on the novel device using the Bland-Altman method. The 95% limits of agreement for estimation of plasma haemoglobin concentration for the device vs. laboratory, the device vs. the blood gas analyser and the blood gas analyser vs. the laboratory were 101.3 g.l(-1), 103.1 g.l(-1) and 14.5 g.l(-1), respectively. The bias (mean difference) in each case was 27.4 g.l(-1), 25.1 g.l(-1) and 2.4 g.l(-1), respectively. We conclude that the novel device in its current form is not a suitable replacement for more invasive methods of determining plasma haemoglobin concentration in patients in the setting of cardiac surgery; however, lessons learnt from the study will help to improve the device's future performance.