Comparison of two single-solute models of potassium kinetics during hemodialysis

Abstract Optimal potassium removal in hemodialysis (HD) is an important but difficult to achieve goal, influenced by numerous factors. Two types of single-solute mathematical models have been previously proposed to assess potassium kinetics in HD: pseudo-one compartment (p1) and two-compartment models (2c). We compared these two models in simulating potassium kinetics during HD sessions with different treatment settings. After estimation of unknown parameters via fitting to clinical data during 4 h sessions with a dialysate potassium of 2.6 ± 0.6 mmol/L, the models were used to simulate 4 HD sessions for each patient, resulting from the combination of session length (4 h vs. 8 h) and potassium dialysate concentration (2.6 vs. 0 mmol/L). The simulated potassium concentration profiles were similar under different treatment conditions, and predicted potassium removal during the treatments was 77 ± 24 mmol with the standard settings; both increases in session length and potassium dialysate to plasma concentration gradient resulted in a significant increase in potassium removed. Both models indicated similar minimum values of dialysate potassium concentration required to avoid post-HD hypokalemia: 1.18 ± 0.66 mmol/L for 4 h HD and 1.71 ± 0.52 mmol/L for 8 h HD. The models described similar kinetics for potassium during different combinations of treatment settings. Total removal of potassium and minimum dialysate concentration to avoid post-HD hypokalemia, were predicted without significant differences by both models. Although no model has a clear advantage in terms of describing clinical data, our analyses suggest that 2c might offer a better trade-off between physiological accuracy and overparametrization.