To better understand the differences in published estimates of peritoneal mass transport coefficients, a comparative analysis of seven mathematical models of peritoneal transport was performed. Uniform investigation involving measurements of solute concentrations and accurate determination of peritoneal dialysate volume was undertaken in twenty-eight 6 hour dwell studies using 3.86% glucose dialysate in non-diabetic patients undergoing continuous ambulatory peritoneal dialysis (CAPD). The investigated models were all based on the theory of transport across a homogeneous membrane. Diffusive mass transport coefficients (KBD) calculated during a period of dialysate isovolemia served as reference values in the comparative analysis. The evaluation of models involved calculations of transport coefficients and comparison of calculated dialysate-to-plasma concentration ratios (D/P) with experimental D/P. The best fit of theoretically predicted D/P to experimental D/P was obtained with Pyle-Popovich's model, which accounts for both the diffusive (KBD) and convective (sieving coefficient [S]) characteristics of the peritoneal membrane. Garred's model, assuming S = 1, yielded acceptable results for small solutes (except sodium), whereas Henderson's model, in which convective transport is neglected, proved to be surprisingly accurate for KBD for both small solutes (except sodium) and protein. S values (mean +/- SD) calculated using Pyle-Popovich's model were found to be out of the physically interpretable range for glucose (S = -0.38 +/- 0.48) and potassium (S = 1.57 +/- 0.19), and therefore these S values should be treated as phenomenologic rather than physical quantities.