The most accurate method for assessing the dialysis dose delivered during high efficiency/flux hemodialysis has not been established. Most current indices of dialysis dose are based on blood-side urea measurements, and thus estimate urea removal. Unfortunately, these methods may lead to inappropriately short dialysis during high flux or high efficiency dialysis, perhaps because of inaccuracies in estimating the amount of urea removal. It is unknown whether these clearance-based approaches can accurately predict either absolute or fractional net urea removal, the latter being equivalent to the solute removal index (SRI). Therefore, we compared the urea removal calculated by five blood-side kinetic methods: (1) urea reduction ration, (2) 1-pool, (3) 2-pool models, and the (4) Smye and (5) Daugirdas formulae. These were compared with the gold standard measurement by direct dialysate quantification. Eight stable patients receiving high-flux hemodialysis were studied over four sessions each. BUN was measured at 0, 45 minutes, 90 minutes, end dialysis, one hour after dialysis (equilibrium value), and 48 hours later. Total body water was determined from the dialysate urea removal; the urea generation rate was calculated using one hour post-dialysis and 48-hour BUN values. Both the total body water and urea generation rate were provided to the 1- and 2-pool models to optimize accuracy. The urea reduction ratio overestimated SRI. The 1-pool model overestimated both absolute urea removal and SRI in 28 of 32 sessions. The 2-pool model slightly underestimated both absolute urea removal and SRI. In contrast, the Smye and Daugirdas formulas accurately estimated SRI.(ABSTRACT TRUNCATED AT 250 WORDS)