Quantitative structure-pharmacokinetic relationships for disposition parameters of cephalosporins.

This study explores the utility of quantitative structure-pharmacokinetic relationship models of the disposition parameters: clearance (CL), apparent volume of drug distribution (V(ap)), fractal clearance (CL(f)), and fractal volume (v(f)), for a series of 23 cephalosporins used in therapeutics. Data for CL, V(ap) and elimination half-life were obtained from literature, whereas CL(f) and v(f) were calculated from the literature data for CL and V(ap), respectively. A variety of descriptors expressing acidity/basicity, lipophilicity, molecular size and hydrogen bonding properties were estimated using computer packages. For each pharmacokinetic parameter, projection to latent structures (PLS) was applied to the total dataset. Adequate PLS models, with one principal component, were derived for CL, CL(f), V(ap) and v(f). Identical descriptors were found to be significant for the two clearance as well as for the two volume of distribution terms. CL and CL(f) expressed similar performance while the predictive performance of v(f) was much higher than that of V(ap). Multiple linear and non-linear regression models were developed. The regression results were in agreement with the PLS models. The non-linear models were superior to the relevant linear relationships. The worst models found were for V(ap) (R(2)=0.523 and R(2)=0.571 for the linear and non-linear model, respectively) and the best models found were for v(f) (R(2)=0.729 and R(2)=0.824 for the linear and non-linear model, respectively).