Multivariate statistics of disposition pharmacokinetic parameters for structurally unrelated drugs used in therapeutics.

purposeTo explore the quantitative structure pharmacokinetic relationships of the disposition parameters: clearance (CL), apparent volume of drug distribution (V(ap)), fractal clearance (CL(f)), and fractal volume (v(f)) for 272 structurally unrelated drugs used in therapeutics.

methodsLiterature data were used for CL and V(ap) whereas CL(f) and v(f) were estimated as described previously (Pharm. Res. 18, 1056, 2001 and 19, 697, 2002). A variety of molecular descriptors expressing lipophilicity, ionization, molecular size and hydrogen bonding capacity were estimated using computer packages. The data were analyzed using multivariate statistics. For each disposition parameter (CL, V(ap) CL(f), v(f)) PCA (principal component analysis) and PLS (projection to latent structures) were applied to the total set of data as well as to subsets of data.

resultsDrugs were divided into two classes (I and II) according to their v(f)/V(ap) ratio. Class I comprises 131 drugs with v(f)/V(ap) > 1, whereas class 11141 drugs with v(f)/V(ap) < 1. After initial PLS analysis, class I was subdivided in subclusters I(a) (30 drugs) and I(b) (101 drugs). It was found that I(a) included mostly acidic drugs with high protein binding, whereas class II comprises mainly basic, lipophilic compounds. No correlation was found between CL, V(ap), CL(f) and the used descriptors. Adequate PLS models were derived for v(f) considering subclusters I(a), I(b) and class II separately. The low v(f) values of class I(a) drugs were affected negatively from molecular size descriptors and non-polar surface area. For class I(b) drugs with intermediate v(f) values, apparent lipophilicity contributed positively, although molecular size descriptors and polarity were inhibitory factors. The high v(f) values of class II drugs were positively dependent on intrinsic lipophilicity and increased basicity, whereas polarity entered with negative contribution.

conclusionsThe parameters V(ap), CL, and CL(f) fail to reflect the physicochemical properties of drugs. The transformation of V(ap) values to v(f) is the underlying cause for the valid models for v(f). These models allow a global consideration of molecular properties (lipophilicity, ionization, molecular size, polar surface area) which govern the distribution of drugs in the human body. The present study provides additional evidence for the physiologically sound concept of v(f).