While haemodynamic variability interferes with the assumption of constant flow underlying thermodilution cardiac output calculation, variability in (peripheral) arterial vascular physiology may affect pulse contour cardiac output methods. We compared non-invasive finger arterial pressure-derived continuous cardiac output measurements (Nexfin® ) with cardiac output measured using thermodilution during cardiothoracic surgery and determined the impact of cardiovascular variability on either method. We compared cardiac output derived from non-invasive finger arterial pressure with cardiac output measured by thermodilution at four grades (A-D) of cardiovascular variability. We defined Grade A data as heart rate and mean arterial pressure variability < 5% and the absence of arrhythmias (implying stable flow) and Physiocal® interval (as measure of variability in finger arterial physiology) > 30 beats. Grade B included all levels of heart rate/mean arterial pressure variability and arrhythmias (Physiocal < 30 excluded). Grade C included all Physiocal intervals (heart rate/mean arterial pressure variability > 5% and arrhythmias excluded). Grade D included all data. Comparison results were quantified as percentage errors. We analysed measurements in 27 patients undergoing coronary artery bypass surgery. Before extracorporeal circulation, the percentage error was 23% (n = 14 patients) in grade A, 28% (n = 20) in grade B, 32% (n = 22) in grade C and 37% (n = 26) in grade D, with a significant increase in variance (p = 0.035). Bias did not differ between grades. After extracorporeal circulation (n = 27), percentage errors became larger, but were not different between grades. Variability during cardiothoracic surgery affected the comparison between thermodilution and non-invasive finger arterial pressure-derived cardiac output. When the main sources of variability were included, percentage errors were large. Future cardiac output methodology comparison studies should report haemodynamic variability.