Relative plasma, brain and cerebrospinal fluid (CSF) exposures and unbound fractions in plasma and brain were examined for 18 proprietary compounds in rats. The relationship between in vivo brain-to-plasma ratio and in vitro plasma-to-brain unbound fraction (fu) was examined. In addition, plasma fu and brain fu were examined for their relationship to in vivo CSF-to-plasma and CSF-to-brain ratios, respectively. Findings were delineated based on the presence or absence of active efflux. Finally, the same comparisons were examined in FVB vs. MDR 1a/1b knockout mice for a selected P-glycoprotein (Pgp) substrate. For the nine compounds without indications of active efflux, predictive correlations were observed between ratios of brain-to-plasma exposure and plasma-to-brain fu (r(2) = 0.98), CSF-to-brain exposure vs. brain fu (r(2) = 0.72), and CSF-to-plasma exposure vs. plasma fu (r(2) = 0.82). For the nine compounds with indications of active efflux, nonspecific binding data tended to over predict the brain-to-plasma and CSF-to-plasma exposure ratios. Interestingly, CSF-to-brain exposure ratio was consistently under predicted by brain fu for this set. Using a select Pgp substrate, it was demonstrated that the brain-to-plasma exposure ratio was identical to that predicted by plasma-to-brain fu ratio in MDR 1a/1b knockout mice. In FVB mice, plasma-to-brain fu over predicted brain-to-plasma exposure ratio to the same degree as the difference in brain-to-plasma exposure ratio between MDR 1a/1b and FVB mice. Consistent results were obtained in rats, suggesting a similar kinetic behavior between species. These data illustrate how an understanding of relative tissue binding (plasma, brain) can allow for a quantitative examination of active processes that determine CNS exposure. The general applicability of this approach offers advantages over species- and mechanism-specific approaches.

Quantitative prediction of human pharmacokinetics is critical in assessing the viability of drug candidates and in determining first-in-human dosing. Numerous prediction methodologies, incorporating both in vitro and preclinical in vivo data, have been developed in recent years, each with advantages and disadvantages. However, the lack of a comprehensive data set, both preclinical and clinical, has limited efforts to evaluate the optimal strategy (or strategies) that results in quantitative predictions of human pharmacokinetics. To address this issue, the authors conducted a retrospective analysis using 50 proprietary compounds for which in vitro, preclinical pharmacokinetic data and oral single-dose human pharmacokinetic data were available. Five predictive strategies, involving either allometry or use of unbound intrinsic clearance from microsomes or hepatocytes, were then compared for their ability to predict human oral clearance, half-life through predictions of systemic clearance, volume of distribution, and bioavailability. Use of a single-species scaling approach with rat, dog, or monkey was as accurate as or more accurate than using multiple-species allometry. For those compounds cleared almost exclusively by P450-mediated pathways, scaling from human liver microsomes was as predictive as single-species scaling of clearance based on data from rat, dog, or monkey. These data suggest that use of predictive methods involving either single-species in vivo data or in vitro human liver microsomes can quantitatively predict human in vivo pharmacokinetics and suggest the possibility of streamlining the predictive methodology through use of a single species or use only of human in vitro microsomal preparations.

ABSTRACT:Although approaches to the prediction of drug-drug interactions (DDIs) arising via time-dependent inactivation have recently been developed, such approaches do not account for simple competitive inhibition or induction. Accordingly, these approaches do not provide accurate predictions of DDIs arising from simple competitive inhibition (e.g., ketoconazole) or induction of cytochromes P450 (e.g., phenytoin). In addition, methods that focus upon a single interaction mechanism are likely to yield misleading predictions in the face of mixed mechanisms (e.g., ritonavir). As such, we have developed a more comprehensive mathematical model that accounts for the simultaneous influences of competitive inhibition, time-dependent inactivation, and induction of CYP3A in both the liver and intestine to provide a net drug-drug interaction prediction in terms of area under the concentration-time curve ratio. This model provides a framework by which readily obtained in vitro values for competitive inhibition, time-dependent inactivation and induction for the precipitant compound as well as literature values for f m and F G for the object drug can be used to provide quantitative predictions of DDIs. Using this model, DDIs arising via inactivation (e.g., erythromycin) continue to be well predicted, whereas those arising via competitive inhibition (e.g., ketoconazole), induction (e.g., phenytoin), and mixed mechanisms (e.g., ritonavir) are also predicted within the ranges reported in the clinic. This comprehensive model quantitatively predicts clinical observations with reasonable accuracy and can be a valuable tool to evaluate candidate drugs and rationalize clinical DDIs.

ABSTRACT:The P-glycoprotein (P-gp)-deficient mouse model is used to assess the influence of P-gp-mediated efflux on the central nervous system ( The efflux transporter P-glycoprotein (P-gp) attenuates the central nervous system (CNS) distribution of many drugs, including opioids, triptans, protease inhibitors, and antihistamines. One method used to assess the influence of P-gp on the CNS distribution of compounds is the P-gp-deficient mouse model. The P-gp efflux ratio, calculated from the ratio of brain/plasma partition coefficient (K p,brain ) in P-gpdeficient (mdr1aϪ/Ϫ) mice to K p,brain in P-gp-competent (mdr1aϩ/ϩ) mice, reflects the degree to which P-gp-mediated efflux attenuates CNS distribution. However, when other processes influence CNS distribution, the P-gp efflux ratio may be a poor indicator of the degree to which CNS distribution of a compound is impaired.

ABSTRACT:Unbound fractions in mouse brain and plasma were determined for 31 structurally diverse central nervous system (CNS) drugs and two active metabolites. Three comparisons were made between in vitro binding and in vivo exposure data, namely: 1) mouse brainto-plasma exposure versus unbound plasma-to-unbound brain fraction ratio (fu plasma /fu brain ), 2) cerebrospinal fluid-to-brain exposure versus unbound brain fraction (fu brain ), and 3) cerebrospinal fluid-to-plasma exposure versus unbound plasma fraction (fu plasma ). Unbound fraction data were within 3-fold of in vivo exposure ratios for the majority of the drugs examined (i.e., 22 of 33), indicating a predominately free equilibrium across the blood-brain and blood-CSF barriers. Some degree of distributional impairment at either the blood-CSF or the blood-brain barrier was indicated for 8 of the 11 remaining drugs (i.e., carbamazepine, midazolam, phenytoin, sulpiride, thiopental, risperidone, 9-hydroxyrisperidone, and zolpidem). In several cases, the indicated distributional impairment is consistent with other independent literature reports for these drugs. Through the use of this approach, it appears that most CNS-active agents freely equilibrate across the blood-brain and blood-CSF barriers such that unbound drug concentrations in brain approximate those in the plasma. However, these results also support the intuitive concept that distributional impairment does not necessarily preclude CNS activity.