Comparison of crosspoint and least-squares regression methods in computation of membrane protein flux parameters from...

To compute two parameters of capillary protein transport of reflection coefficient (sigma) and permeability--surface area product (PS), two major methods have been employed. The first, or crosspoint method, uses two sets of lymph flow (Jv) and lymph/plasma protein concentration ratios (R) and solves the irreversible thermodynamic relationship between sigma and PS by simultaneous solution. The second, or least squares method, analyzes the total R versus Jv curve by finding a best fit of all points to the irreversible thermodynamic relationship of R versus Jv. Three theoretical membranes were analyzed with both methods by imposing random errors of 10 or 20% in R and Jv. It was found that such errors, especially in R, were associated with substantial decreases in the percent of successful computations by the crosspoint method. In tight membranes (sigma = 0.9) both methods gave comparable values. For intermediate (sigma = 0.5) and loose membranes (sigma = 0.2) the least-squares method was less specific than the crosspoint method, since up to 9% of random values of R and Jv would be included within the 95% limits for the method. Moreover, attributing homogeneous variance to input data and sampling Jv at geometric intervals exaggerated the variability of the least-squares method output parameters over those obtained by the crosspoint method. Because of nonspecificity, and the fact that the least squares method does not allow for the possibility that sigma and PS are direct functions of hydrostatic capillary pressure or Jv, the crosspoint method appears superior to the least-squares method for computation of sigma and PS from R and Jv.