Model-based dialysis adequacy prediction by continuous dialysate urea monitoring

A modeling approach for on-line estimation of urea kinetics from continuous measurement of urea concentration in the effluent dialysate stream (DUN) is presented. On-line identification of urea kinetics response parameters is used to predict and update dialysis adequacy during the treatment. Dialysis adequacy can be quantified in several ways, but its strict dependence on final urea concentration is a major fact. For this reason, a good predictive skill on the time course of DUN may enable better performances in the control of dialysis outcome by treatment parameters adjustment. A post-filter enzymatic sensor performs continuous measurement of DUN on patients undergoing standard haemodialysis. To get an early prediction of the end dialysis urea level, the solution of a variable volume double- pool (VVDP) model is used, whose parameters are identified at each time on the basis of the past DUN history. Unlike the variable volume single-pool (VVSP) model, this enables a prompt and accurate estimation of the final DUN. In fact, after 75 min the estimates always differ by less than 10% from the values measured by the sensor at the end of the treatment. Moreover, values predicted by the model in the last hour always lie within 1% of measured final values. Real-time knowledge of an analytic expression for whole DUN time course also enables the accurate prediction of total removed urea, with no need of cumbersome dialysate collection techniques.