Physiologically based pharmacokinetic modeling of midostaurin and metabolites at steady-state to bridge drug interaction scenarios in lieu of clinical trials.

Abstract

Midostaurin and its active metabolites are substrates, mixed inhibitors/inducers of cytochrome P450 (CYP)3A4. The main objective of this study was to develop/refine a physiologically based pharmacokinetic (PBPK) model that incorporated recent clinical drug-drug interaction (DDI) data with midazolam after multiple dosing, to qualify the pharmacokinetic (PK) model simulations of midostaurin and its metabolites, and to apply it to predict untested clinical DDI scenarios with potential comedications. In this study, Simcyp PBPK model of midostaurin and its 2 metabolites was refined from a previously published model associated with endogenous biomarker 4β-hydroxycholesterol data through further optimization of CYP3A4 inhibition/induction potency and was qualified to simulate midostaurin steady-state PK. The incorporation of these parameters enabled DDI predictions of high midostaurin doses on the PK of midazolam and oral contraceptives containing ethinyl estradiol. Additionally, scaling factors for in vitro breast cancer resistance protein and the organic anion transporting polypeptide (OATP1B) inhibition were applied to account for the observed single-dose DDI with rosuvastatin and further extrapolated to predict steady-state DDI with other OATP1B drug substrates. The overall prediction results showed minimal impact of midostaurin at high doses on CYP3A substrates or an effect on the exposure of OATP1B substrates. In summary, the midostaurin PBPK model was retrospectively refined, requalified, and used to simulate the steady-state perpetrator DDI of midostaurin and its metabolites. This PBPK modeling approach and the resulting model predictions were implemented into the midostaurin product label (up to 100 mg twice a day) without the need for confirmatory clinical studies. SIGNIFICANCE STATEMENT: The manuscript describes how a midostaurin PBPK model was updated, verified, and applied to untested scenarios by a predict-learn-confirm cycle as new clinical data become available. It also provides a learning experience of prospective prediction by utilizing endogenous biomarker 4β-hydroxycholesterol to evaluate a complex CYP3A4-mediated drug interaction.