To investigate an appropriate solute kinetics model for clinical application, we analyzed urea nitrogen (UN) and creatinine (Cr) kinetics by a variable-volume two-compartmental model (2CM) and a regional blood flow model (RBF) in 44 hemodialysis patients with varying proportions of first compartmental volume and regional volume (p(1)). Solute kinetics could not be solved in some of the patients with higher p(1) values, and there were more solution failures by the RBF than by the 2CM. The solute generation rate (g) and solute distribution volume in the dry state (V(D)) increased with increases in p(1) in both models, but there were some differences between the two models. When g was normalized by V(D), it became relatively constant, irrespective of the p(1) value or model used (0.133 +/- 0.029 mg/min/l by the 2CM and 0.132 +/- 0.029 mg/min/l by the RBF for UN; 0.0200 +/- 0.0049 mg/min/l by the 2CM and 0.0198 +/- 0.0048 mg/min/l by the RBF for Cr). The intercompartmental mass transfer coefficient (K(c); liters/min) calculated by the 2CM decreased as p(1) increased (K(c) = -1.77.p(1) + 1.16, p < 0.0001, R = 0.999 for UN; K(c) = -0.847.p(1) + 0.556, p < 0.0001, R = 1.000 for Cr). The systemic blood flow (Q(sys); liters/min) calculated by the RBF also decreased as p(1) increased (Q(sys) = -11.1.p(1) + 6.21, p < 0.0005, R = 1.000 for UN; Q(sys) = -5.22.p(1) + 2.90, p < 0.001, R = 0.999 for Cr). Since the RBF more frequently failed to solve the solute kinetics and since there was a difference in its Q(sys) values for UN and Cr, the 2CM was considered to be a superior model. When p(1) was extremely low, the 2CM could be transformed into a modified variable-volume one-compartment model (1CM) which presented a similar g/V(D) (0.133 +/- 0.029 for UN; 0.0200 +/- 0.0048 for Cr). This modified 1CM was considered to satisfy appropriate conditions for clinical application, since it is simpler than the 2CM and provides useful information on the dialysis dose.