To test the accuracy of urea kinetic modeling (UKM), the classic fixed-volume model UKM(f), two variable-volume models (UKM(vb) and UKM(vd)), direct dialysis quantification (DDQ) and a partial dialysate collection method (PDC) were evaluated in 15 stable, high-hematocrit patients. Urea generation rate (G) was also determined from a 1-week collection of total dialysate and urine (OWC). The results, except distribution volumes, were highly correlated. However, Kt/V, the normalized whole-body urea clearance, was about 8% higher with UKM(vb) and UKMvd(vd). Two of three simple equations for Kt/V rendered grossly deviating, but highly correlating, results. The normalized protein catabolic rate was 8% higher with UKM(vd). With OWC as reference, UKM(vb) and UKM(vd) overestimated G by 19 and 15%, respectively. All results of PDC closely followed those of DDQ. This method may be an alternative for exact quantification. Before using a new UKM method it should be compared to an established reference method.