Peritoneal dialysis kinetic modeling: validation in a multicenter clinical study

To clinically validate the use of a computer-based kinetic model for peritoneal dialysis (PD) by assessing the level of agreement between measured and modeled values of urea and creatinine clearances and ultrafiltration (UF).

An open multicenter observational study.

There were 111 adult continuous ambulatory peritoneal dialysis (CAPD) patients (47 female, 64 male) in four centers. All patients underwent a four-hour peritoneal equilibration test (PET) using 2.5% dextrose but with variable fill volumes (range: 1-3 L). Patients with a residual renal function greater than 10 mL/min were excluded.

Correlations and limits of agreement between measured and modeled values of total weekly urea KT/V, total weekly normalized creatinine clearance (L/week/1.73 m2), daily drain volume (L), net ultrafiltration (L), daily peritoneal urea clearance (L/day), and daily peritoneal creatinine clearance (L/day). Measured values were obtained from 24-hour urine and dialysate collections while modeled values were based on results from the PET in combination with the PD ADEQUEST kinetic program.

The results show there is excellent agreement between measured and modeled urea KT/V and creatinine clearances, with concordance correlations of 0.94 and 0.92, respectively. Given the excessive variation and limited range in ultrafiltration values, the concordance correlation between measured and modeled UF was only 0.50. In terms of daily peritoneal clearances and ultrafiltration, the level of precision (i.e., standard deviation) in the differences between modeled and measured values is +/- 1.05 L/day for urea clearance +/- 1.03 L/day for creatinine clearance, and +/- 0.919 L/day for ultrafiltration. By contrast, the level of precision (i.e., standard deviation) in the differences between two measured values is estimated to be +/- 0.979 L/day for urea clearance, +/- 0.802 L/day for creatinine clearance, and +/- 0.707 L/day for ultrafiltration. Defining the limits of clinical agreement to be +/- 2 standard deviations of the differences between two clinically measured 24-hour clearances (or ultrafiltration), we find that 94% of the modeled urea clearances, 87% of the modeled creatinine clearances, and 86% of the modeled ultrafiltration values fall within the limits of clinical agreement.

Data for a carefully performed PET and overnight exchange can, in combination with a scientifically validated kinetic model, provide clinicians with a powerful mathematical tool for use in CAPD dialysis prescription management. Although not intended to replace actual measurements, kinetic modeling can prove useful as a means for predicting clearances for various alternative prescriptions and perhaps also as a means for checking certain types of noncompliance.