David Damoiseaux (1), Wenlong Li (2), Alejandra MartÃnez-Cha´vez (2), Jos H. Beijnen (1, 3), Alfred H. Schinkel (2), Alwin D.R. Huitema (1, 4, 5), Thomas P.C. Dorlo (1)
(1) Department of Pharmacy & Pharmacology, The Netherlands Cancer Institute, Amsterdam, The Netherlands, (2) Division of Pharmacology, The Netherlands Cancer Institute, Amsterdam, the Netherlands, (3) Utrecht Institute of Pharmaceutical Sciences, Utrecht University, Utrecht, The Netherlands, (4) Department of Clinical Pharmacy, University Medical Center Utrecht, Utrecht University, Utrecht, The Netherlands, (5) Princess Máxima Center for Pediatric Oncology, Utrecht University, Utrecht, The Netherlands.
Introduction: Approximating the human pharmacokinetic (PK) profile as closely as possible in preclinical settings can still be a struggle in the context of predicting first-in-human doses and drug-drug interactions [1-3]. For safety reasons a 4-fold margin is currently recommended for interspecies extrapolation of PK [4, 5].
Human CYP3A4 transgenic (Cyp3aXAV) mice were mainly developed to assess the effect of CYP3A4 in a qualitative manner. This might be a more representative animal model for the pharmacokinetics of CYP3A4-metabolized compounds in human [6]. We hypothesized that this murine model could also yield improved quantitative predictions of human exposure, using a compartmental population PK (PopPK) approach.
Therefore, we aimed to model murine Cyp3aXAV data using a PopPK approach and investigate the predictive value and the accurateness of a quantitative extrapolation of this mouse model to humans for CYP3A4-metabolized compounds. We used an acceptance criterium of a 2-fold difference in terms of exposure and assessed the overall predictiveness of the shape of the extrapolated human PK curve. We investigated this for four different CYP3A4 substrates: lorlatinib, brigatinib, fisogatinib and ribociclib.
Methods: PopPK models for lorlatinib, brigatinib, fisogatinib and ribociclib in mice were developed [7-11]. One- and two-compartment and different absorption-related models were evaluated. To account for differences between wild type (WT) and Cyp3aXAV mice, strain was evaluated as a covariate on clearance (CL) and bioavailability (BB).
All models were scaled to humans using standard allometric scaling of clearances and volumes of distribution. In total, 500 simulations with 50 individuals were run for both extrapolated and earlier published human models, after which the 50% prediction interval curve shapes were visually compared. Also AUCinf, Cmax and Tmax were calculated and compared, with a <2-fold deviation from clinical data as a target [12-15].
Results: The models that best fitted the data were a 2-compartment model with a dual first-order absorption for lorlatinib, a 2-compartment model with an exponential dose effect on BB for brigatinib, a 1-compartment model with transit compartments and an enterohepatic circulation model (EHC) for fisogatinib and a 2-compartment model with an EHC for ribociclib. Although EHC and exponential dose effect on BB led to significant model improvements for ribociclib, brigatinib and fisogatinib in mice, these turned out to be redundant model properties for the extrapolation to human and were omitted for human extrapolation. Significant covariates identified for Cyp3aXAV mice relative to WT mice were 1.34x higher CL and 0.71x lower BB for lorlatinib; 1.91x higher CL for brigatinib; 0.30x lower BB for ribociclib and 0.61x lower CL and 0.57x lower BB for fisogatinib.
The median AUCinf ratios for WT and Cyp3aXAV extrapolation compared to clinical data respectively were: 2.08x and 1.12x for lorlatinib, 1.80x and 0.99x for brigatinib and 1.00x and 0.30x for ribociclib. The extrapolation of the Cyp3aXAV model for both lorlatinib and brigatinib gave adequate predictions of clinical PK profiles indicated by largely overlapping 50% prediction intervals. Extrapolation of WT, despite the similar AUCinf ratios for ribociclib, led to less adequate overlap of the 50% prediction intervals, suggesting that certain PK properties were not well covered in the extrapolation. Due to the lack of clinical data for fisogatinib, only the Cmax could be compared for which both WT and Cyp3aXAV achieved the target.
Conclusions: Extrapolation based on Cyp3aXAV mice led to a more accurate prediction of exposure in humans than WT mice for both lorlatinib and brigatinib, with predicted median AUCinf deviating less than 12%. Moreover, the shape of the extrapolated Cyp3aXAV PK curves of lorlatinib and brigatinib visually matched the human PK curves, suggesting that this model is an appropriate model to predict the human PK of lorlatinib and brigatinib. In contrast, a higher than 2-fold deviation in AUCinf and some deviations in the curve shapes of ribociclib suggested that the Cyp3aXAV model is not fully representative of the human PK.
References:
[1] ICH, Safety Guidelines – S3 Toxicokinetics and Pharmacokinetics – note for guidance on toxicokinetics: the assessment of systemic exposure in toxicity studies s3a 1994.
[2] Ema, guideline on requirements for first-in-man clinical trials for potential high-risk medicinal products. 2007.
[3] FDA, Guidance for Industry Estimating the Maximum Safe Starting Dose in Initial Clinical Trials for Therapeutics in Adult Healthy Volunteers. 2005.
[4] WHO, Guidance Document for the Use of Data in Development of Chemical-Specific Adjustment Factors (CSAFs) for Interspecies Differences and Human Variability in Dose/Concentration–Response Assessment. 2001.
[5] EPA, Guidance for Applying Quantitative Data to Develop Data-Derived Extrapolation Factors for Interspecies and Intraspecies Extrapolation. 2014.
[6] van Herwaarden, A.E., et al., Knockout of cytochrome P450 3A yields new mouse models for understanding xenobiotic metabolism. J Clin Invest, 2007. 117(11): p. 3583-92.
[7] Li, W., et al., P-glycoprotein (ABCB1/MDR1) limits brain accumulation and Cytochrome P450-3A (CYP3A) restricts oral availability of the novel FGFR4 inhibitor fisogatinib (BLU-554). Int J Pharm, 2020. 573: p. 118842.
[8] Li, W., et al., Oral coadministration of elacridar and ritonavir enhances brain accumulation and oral availability of the novel ALK/ROS1 inhibitor lorlatinib. Eur J Pharm Biopharm, 2019. 136: p. 120-130.
[9] Li, W., et al., P-glycoprotein and breast cancer resistance protein restrict brigatinib brain accumulation and toxicity, and, alongside CYP3A, limit its oral availability. Pharmacol Res, 2018. 137: p. 47-55.
[10] Li, W., et al., P-glycoprotein (MDR1/ABCB1) restricts brain accumulation and cytochrome P450-3A (CYP3A) limits oral availability of the novel ALK/ROS1 inhibitor lorlatinib. Int J Cancer, 2018. 143(8): p. 2029-2038.
[11] Martínez-Chávez, A., et al., P-glycoprotein Limits Ribociclib Brain Exposure and CYP3A4 Restricts Its Oral Bioavailability. Mol Pharm, 2019. 16(9): p. 3842-3852.
[12] Gupta, N., et al., Population Pharmacokinetics of Brigatinib in Healthy Volunteers and Patients With Cancer. Clin Pharmacokinet, 2021. 60(2): p. 235-247.
[13] Lu, Y., et al., Ribociclib Population Pharmacokinetics and Pharmacokinetic/Pharmacodynamic Analysis of Neutrophils in Cancer Patients. J Clin Pharmacol, 2021.
[14] Chen, J., et al., Population pharmacokinetic model with time-varying clearance for lorlatinib using pooled data from patients with non-small cell lung cancer and healthy participants. CPT Pharmacometrics Syst Pharmacol, 2021. 10(2): p. 148-160.
[15] Kim, R.D., et al., First-in-Human Phase I Study of Fisogatinib (BLU-554) Validates Aberrant FGF19 Signaling as a Driver Event in Hepatocellular Carcinoma. Cancer Discov, 2019. 9(12): p. 1696-1707.
Reference: PAGE 29 (2021) Abstr 9751 [www.page-meeting.org/?abstract=9751]
Poster: Drug/Disease Modelling - Oncology