Indocyanine green (ICG) is used in cardiology and hepatology for the estimation of cardiac output, liver function, and splanchnic blood flow. ICG is bound to plasma proteins and ultimately excreted by the liver. We studied the whole body kinetics of ICG during constant infusion in pigs weighing 30-40 kg. The conventional kinetic model (backflux model) assumes that deviations from one-compartmental linear kinetics is caused by backflux from a liver storage to plasma, and that no extravascular, extrahepatic distribution takes place. This model was tested against an alternative (redistribution) model postulating that temporary redistribution of ICG into an extrahepatic extravascular storage was responsible for the deviations while the hepatic uptake was a one-way first-order process. A mathematical analysis of the two models showed that they predicted different time courses of the hepatic extraction fraction of ICG. Thus, with blood sampling from both a peripheral artery and a hepatic vein, a discriminative model-testing experiment was possible. This test required a first-order steady-state hepatic removal of ICG which was confirmed in 7 experiments with infusion rates varied in a stepwise fashion (0.133 +/- 0.003, 0.269 +/- 0.010, 0.547 +/- 0.020 and 0.130 +/- 0.003 mumol.min-1). In the model-testing experiments (n = 10) ICG was infused at a constant rate of 0.135 +/- 0.07 mumol.min-1. The mean concentration in peripheral artery (microM) was well fitted by the biexponential function C(t) = 0.476.(1-0.632.e-0.216.1-0.368.e-0.0172.1). The time course of the observed hepatic extraction fraction was significantly different (p = 0.004) from that predicted from the backflux model but in agreement (p = 0.98) with the new model assuming hepatic removal to be a one-way process and implying temporary ICG redistribution into an extrahepatic, extravascular storage with an apparent volume of 0.144 +/- 0.023 L.Kg-1. Accordingly, extravascular ICG was demonstrated in a number of different tissues after 4-hr infusion (n = 3). If ICG is used to estimate hepatic blood flow according to Fick's principle, the use of a backflux model to correct for non-steady-state conditions will lead to an overestimation of hepatic blood flow of 28% after 25-min infusion, 16% after 50 min, and 6% after 100 min. The study indicated that distribution of ICG between plasma and tissues is not instantaneous, and that the time course of the redistribution itself significantly influences whole body kinetics. Comparison with a previously published study by Ott, Keiding, and Bass of ICG kinetics after bolus injection suggested that a two-compartment model was insufficient and that the kinetics for the exchange of ICG between plasma and the redistribution space may be nonlinear. The study demonstrates how blood sampling on both sides of the eliminating organ can expose the influence of redistribution. The discriminative model test for constant infusion experiments is novel and may be useful with other ligands.