Use of nonlinear mixed effect modeling for the meta-analysis of preclinical pharmacokinetic data: application to S 20342 in the rat.

The standard two-stage analysis of separate preclinical pharmacokinetic (PK) and toxicokinetic (TK) studies may lead to good information on the bioavailability in the rat at a low (pharmacologic) dose but only an idea on the dose/exposure relationship, on gender, and on time effect. In view of these drawbacks, we decided therefore to explore the usefulness of the implementation of a meta-analysis in preclinical studies in a given species (the rat in this case) taking as an example S 20342, an investigational new drug with potential antipsychotic properties. A nonlinear mixed-effect PK model was built from all intravenous (IV) and oral (PO) data collected until the completion of the 4-week toxicity study. The database included data from 201 Wistar rats (161 males and 40 females). Forty animals received the drug IV and 161 PO. The treatment duration ranged from 1 day to 4 weeks. IV doses were 3, 5, and 20 mg/kg, and 11 different oral doses were tested in the range of 5 to 200 mg/kg. Three different salts were administered PO: hydrochloride, sulfate, and mesylate. The modeling was performed with NONMEM IV. The best pharmacokinetic model was a two-compartment model with simultaneous first-order and zero-order absorption. The combination of these two input functions allowed the model to fit the peak plasma concentrations observed in the first hour (first order), especially after oral administration of low doses, and to take into account the prolonged absorption phase when the dose increased (zero order). A significant gender effect was found on CL. In addition, significant positive correlations were found between weight and CL, weight and Vc, and dose and the dose fraction after a zero-order absorption. No covariate significantly influenced the other parameters. In conclusion, the meta-analysis of preclinical data allowed for an objective assessment of statistically significant effects throughout the model-building process, leading to a better knowledge (and thus a better understanding) of preclinical PK in the rat. Moreover, the model obtained could be used to interpret further preclinical specific studies involving a sparse sampling design (e.g., further TK studies and PK/PD studies). Although this meta-analysis is more complicated than the noncompartmental approach and requires a case-by-case effort, it could be very useful to integrate this approach in the preclinical development process.