During extrarenal therapy, plasma phosphate concentrations have specific kinetics: plasma values reach a steady-state nadir 90-120 min after the beginning of the session (from 0.6 to 1.1 mmol/l) with a subsequent high rebound in the 3-4 h following the session. These kinetics are found during haemofiltration (HF) with high ultrafiltration (UF) rates (greater than 270 ml/min) and UF volumes (greater than 30 1). Other HF studies with different UF rates (100 or 200 ml/min) show that delayed mass transfer cannot explain kinetics which result from a phosphate transfer from cellular to extracellular space. Acetate or bicarbonate reinjection fluid does not modify phosphate kinetics. Immediate decrease of ionised calcium after the session argues against a mobilisation from the exchangeable phosphate pool of bone. Only potassium shows a similar pattern to phosphate, so the hypothesis of a relation between cellular phosphate and potassium fluxes is postulated. 31P-NMR study during and after HF does not allow us to specify phosphate transfer from the cell, but various potassium concentrations in the reinjection fluid (0, 2, 3.5, or 4 mmol/l) confirm the influence of potassium removal on phosphate transfer, and a significant linear relationship can be established between cellular potassium and phosphate fluxes. The influence of phosphate removal on phosphataemia has also been investigated using 0, 2, or 3 mmol/l phosphate in the reinjection fluid. Whatever the phosphate modification achieved by the session, the patient's phosphate concentrations are not significantly different 2 days later.(ABSTRACT TRUNCATED AT 250 WORDS)