The aim of this study was to develop a pharmacokinetic model that takes into account the negative feedback loop of endogenous erythropoietin production observed after repeated recombinant human erythropoietin administration. A pharmacodynamic data analysis was performed using the changes in i) reticulocyte count, ii) serum levels of soluble transferrin receptors, and iii) soluble transferrin receptors/serum proteins ratio as an index of the therapeutic effect of the hormone. Nine athletes were included in the study; they received repeated subcutaneous administrations (50 IU x kg(-1) per day) of recombinant human erythropoietin. The mean half-life of the terminal part of the curve was 35.5 h, and the total clearance was 17 ml x h(-1) x kg(-1). The total clearance was about two times higher in athletes than in untrained subjects (5.5 - 7.5 ml x h(-1) x kg(-1)) and the half-life period of plasma erythropoietin after subcutaneous administration was five times longer compared to intravenous administration (4 to 7 h). Thus, after subcutaneous administration, the terminal part of the curve should correspond to the absorption phase, instead of to the elimination phase (flip-flop phenomenon). The pharmacodynamic relationship based on a sigmoid Emax model can be reasonably used to relate the changes observed in the markers to recombinant human erythropoietin administration. Recombinant human erythropoietin induces a delayed increase in reticulocytosis and in soluble transferrin receptor levels. In comparison with baseline, the increase of these markers became significant from the third and the tenth day after the initial administration of the hormone, respectively. These results were in accordance with the equilibration delay computed from the pharmacokinetic-pharmacodynamic data modelling (half-life of 25.7 h and 10 days, respectively). The recombinant hormone was well tolerated during this study.