Hyperphosphatemia in the hemodialysis population is ubiquitous, but phosphate kinetics during hemodialysis is poorly understood. Twenty-nine hemodialysis patients each received one long and one short dialysis, equivalent in terms of urea clearance. Phosphate concentrations were measured during each treatment and for one hour thereafter. A new model of phosphate kinetics was developed and implemented in VisSim. This model characterized additional processes involved in phosphate kinetics explaining the departure of the measured data from a standard two-pool model. Pre-dialysis phosphate concentrations were similar in long and short dialysis groups. Post-dialysis phosphate concentrations in long dialysis were higher than in short dialysis (P < 0.02) despite removal of a greater mass of phosphate (P < 0.001). In both long and short dialysis serum phosphate concentrations initially fell in accordance with two-pool kinetics, but thereafter plateaued or increased despite continuing phosphate removal. Implementation of an additional regulatory mechanism such that a third pool liberates phosphate to maintain an intrinsic target concentration (1.18 +/- 0.06 mmol/L; 95% confidence intervals, CI) explained the data in 24% of treatments. The further addition of a fourth pool hysteresis element triggered by critically low phosphate levels (0.80 +/- 0.07 mmol/L, CI) yielded an excellent correlation with the observed data in the remaining 76% of treatments (cumulative standard deviation 0.027 +/- 0.004 mmol/L, CI). The critically low concentration correlated with pre-dialysis phosphate levels (r=0.67, P < 0.0001). Modeling of phosphate kinetics during hemodialysis implies regulation involving up to four phosphate pools. The accuracy of this model suggests that the proposed mechanisms have physiological validity.