Optimization of selectivity in high-performance liquid chromatography using desirability functions and mixture designs according to PRISMA

A computer program for the mobile phase optimization of high-performance liquid chromatography (HPLC) is described. The desirability function technique combined to the PRISMA mixture design was employed to enhance the quality of HPLC separations. The use of statistical models to predict the behaviour of retention times (t(R)) and band broadening at the different eluent compositions obtained by PRISMA was examined for dansyl amides and coumarins. The study showed that the dependence between the eluent composition and t(R) values of dansyl amides and coumarins can be expressed using quadratic regression models with a high degree of accuracy. Band broadening given by means of the band width at half-height (w(h)) war described by a linear regression model. Both models were used in calculating and predicting the resolution (R-s) in various solvent combinations. The desirability function converted the calculated (R-s) value into the desirability value (D), and the overall optimum was then defined by means of the overall desirability. The optimal eluent mixtures for the separation of compounds were easily read from the contour plot inside the horizontal plane of the PRISMA model. A good separation was achieved using the optimized solvent combination. Depending on the aims of the optimal separation, the program allows either optimization of critical pairs or achieving the overall optimum giving a reasonable separation for as many compounds as possible. (C) 1998 Elsevier Science B.V.