The effect of vascular access recirculation (AR) on the modeled urea distribution volume (V) is not straightforward. When blood is sampled properly so that it is not admixed with recirculated blood, AR will cause an unexplained increase in V in cases in which AR is present throughout the dialysis session (when AR is limited to the terminal portion of a dialysis session it will cause little or no change in the modeled V). On the other hand, when blood is sampled from the arterial line after simply stopping the pump, postdialysis blood urea nitrogen (BUN) represents arterial line BUN and does not always reflect the BUN in the patient's blood. Under these conditions, when AR is present throughout the dialysis session the modeled V usually shows an unexplained decrease, but V may be unchanged, depending on the urea reduction ratio (URR). We performed a mathematical analysis to predict when V would be decreased and when it would be unchanged when the postdialysis BUN is contaminated with admixed blood. The analysis revealed that when AR is present uniformly throughout the dialysis session, the modeled V should be underestimated when the URR is < or = O.70. When the URR is greater than 0.70, even severe degrees of AR may not be reflected by a change in V. When AR is limited to the terminal part of the dialysis session or when AR increases during the dialysis session, and when V is based on admixed postdialysis blood, underestimation of V will occur in almost all circumstances. In a cross-sectional comparison of modeled to anthropometric volume ratios in eight patients with severe AR and in 11 controls, and in sequential modeling studies in a single patient in whom severe AR developed gradually over time, the volume ratio was low in most, but not all instances when modeled V was based on an admixed postdialysis BUN sample.