In silico tools to study food-drug interactions, an Industry Perspective

Transcription

1 Paris, April 4th, 2018 Physiologically Based PharmacoKinetic modeling (PBPK): A new Paradigm in Drug Development In silico tools to study food-drug interactions, an Industry Perspective Neil Parrott, Pharmaceutical Sciences, Roche Pharma Research and Early Development, Roche Innovation Center Basel 1

2 Roche Group Roche pred is one of three fully independent research hubs 2

3 What Roche pred Works On Oncology Developing effective cancer therapies Infectious Diseases Effective treatments for life-threatening infectious diseases Immunology & Inflammation Differentiated medicines for patients with immune and inflammatory diseases Ophthalmology Restoring sight + + Neuroscience Developing medicines for serious neurological diseases Rare Diseases Tackling rare genetic disorders 3

4 Overview How do Food Effects Impact Development of a Drug? Predictive Tools for Food Effects and their Application in Pharma Physiologically based Food Effect Modeling A Roche Case Study Future Directions 4

5 Regulatory Guidance Food effect bioavailability studies are needed to support global filings of NDAs 5

6 Food effect throughout Drug Development Classification (e.g. BCS) In vitro Pre-clinical in vivo Simple formulation Early FE in Ph1 Optimized Formulation(s) Repeat FE Market formulation Repeat FE Conducted early in drug development and may be repeated after formulation change & with market formulation for product label Effect of different doses, meal types or times of drug intake in relation to a meal may need to be characterized 6

7 The Need to Understand Mechanisms Food can alter the absorption through various changes : GI physiology, stomach emptying time, ph, bile salt concentration etc Significant optimization efforts may be required & are effective only if mechanisms are understood Tools to predict and understand food effects include in vitro, in vivo and in silico models 7

8 Tools Predictive of Food Effects Drug properties and classification systems Biorelevant solubility / dissolution tests Pre-clinical models (beagle dog) Physiologically based absorption models 8

9 Potential Complexity of Food Effect Abuhelwa, A. Y., et al. (2017). "Food, gastrointestinal ph, and models of oral drug absorption." European Journal of Pharmaceutics and Biopharmaceutics 112:

10 Physiologically Based Pharmacokinetics (PBPK) A mathematical modeling technique for predicting the absorption, distribution, metabolism and excretion (ADME) of synthetic or natural chemical substances in humans and other animal species. 10

11 Small molecule PBPK modeling ABSORPTION release dissolution stomach duodenum jejunum ileum caecum colon permeation DISTRIBUTION venous Lung Liver Brain Kidney Muscle Adipose Other tissues arterial METABOLISM ELIMINATION

12 PBPK Modelling in Industry Discovery Early Development Late Development In vitro In silico In vivo Early risk assessment, Early first in human dose projection, Toxicokinetic dose projection, Early formulation assessment Learn and confirm through data integration First in human, single/multiple ascending dose exposure or efficacy modeling, Drug-Drug interaction, Food effect, Formulation/Absorption modeling Healthy subjects & patients Phase I-IV trials Patient trials, Special populations, Label requirements Continuous Model Refinement & Verification 12

13 Absorption Model parameters include : Undissolved Dissolved Enterocyte Portal vein Physiology Intestinal fluid volume Intestinal transit times Intestinal ph Luminal surface area Metabolizing enzyme expression Drug specific Solubility Particle size Charge Lipophilicity Formulation Agoram, B., W.S. Woltosz, and M.B. Bolger,. Adv. Drug Deliv. Rev., (Supplement 1): p. S41 S67.

14 Experience at Roche 14

15 A Case Study Food effect PBPK prediction Food effect Clinical Study 2 nd Food effect Clinical Study Parrott et al. (2013). "Physiologically Based Pharmacokinetic Modelling to Predict Single and Multiple Dose Human Pharmacokinetics of Bitopertin." Clinical Pharmacokinetics 52(8): Parrott et al. (2014). "Physiologically Based Absorption Modelling to Predict the Impact of Drug Properties on Pharmacokinetics of Bitopertin." The AAPS Journal 16(5):

16 Biopharmaceutical Properties Molecular weight Ionization constant Neutral logd at ph Scaled human permeability (10-4 cm/s) PAMPA 1.2 Caco2 3.5 Solubility (mg/ml) Phosphate buffer ph 7 Fasted state simulating intestinal fluid: (ph=6.5) Fasted state simulating intestinal fluid & : (ph=6.5) Fed state simulating intestinal fluid: (ph=5.0) BCS 2 with enhanced solubility in fed state. However model predicted no food effect on AUC at expected clinical dose of 13 mg & Measured for clinical capsules in FaSSIF at 37C Data became available after EIH prediction 16

17 Physiologically Based Model Prediction PBPK developed based on pre-clinical data and used to predict human pharmacokinetics prior to the first in human studies Predicted : CL: 1 ml/min/kg; Vss = 3 L/kg; F% (<80 mg) = 90% The predicted pharmacokinetics were found to be in good agreement with the clinical data from the single ascending dose study at 3, 6, 12, 24, 50, 80, 120, 180 and 240 mg. 50 mg dose

18 Model Refinement with First Clinical Data Bottom-up prediction Improved simulation of profiles by accounting for slightly increased solubility and permeability Additional modification to intestinal water volume in colon to reduce late absorption Model applied to predict food effect at highest anticipated clinical efficacious dose of 80 mg very slight increase in Cmax and AUC with food Refined model 6, 50 & 180 mg 18

19 Parameter Sensitivity Analysis GastroPlus Baseline parameters Permeability scaled from Caco2 3.5 *10-4 cm/s Solubility in fasted state simulating fluid 25 ug/ml Particle size 6 um radius

20 Clinical Food Effect & Verification of Prediction Studied in 14 healthy volunteers after a high fat/high calorie breakfast Fed / fasted (Geomean) Cmax AUC Simulated Observed (90%CI) 1.39 (1.21 to 1.59) 1.14 (1.09 to 1.19)

21 Further Model Verification Particle Size Relative BA study compared 30 mg tablets containing powder prepared with either jet or hammer milling JET milled HAMMER milled Particle radius (µm) N=22 NHVs Relative BA of HAMMER to JET (90% CI) 78% for AUCinf/dose (72% 80%) 62% for Cmax/dose (57% 67%)

22 Relevance of In Vitro Testing Verify the relevance of in vitro dissolution tests for in vivo drug performance. Dissolution test employed SDS in order to achieve sink conditions in vitro which was otherwise not possible due to the low solubility Reasonable IVIVC confirms relevance of dissolution test

23 Food Effect Study with Market Formulation A film coated tablet was chosen for the market at a dose of 10 mg A relative bioavailability study had shown that 10 mg tablets were equivalent to 10 mg capsules We had confidence that the model was capturing the absorption behavior and had predicted well the food effect at 80 mg However, a 2 nd food effect study was conducted in view of regulatory guidance and the lack of a precedent for waiver of a study based on PBPK modelling

24 Results and Simulation Cmax (ng/ml) AUC0-inf (ng*h/ml) Tmax (hrs) Observed 1 Simulated Observed 1 Simulated Observed 2 Simulated Fasted Fed Geometric mean of individual observed values 2 median of individual observed values 24

25 Conclusion to Roche Case Study This BCS2 molecule had well behaved PK and the PBPK model based on in vitro measurements could be verified with multiple clinical studies A 2 nd food effect study added minimal value and could be waived This represents a good percentage of development compounds (BCS 1 & 2) and overall a significant number of studies might be waived However other BCS2 molecules present more challenges e.g. alectinib Parrott, N. J., et al. (2016). "Physiologically Based Absorption Modeling to Explore the Impact of Food and Gastric ph Changes on the Pharmacokinetics of Alectinib." The AAPS Journal:

26 Future Outlook 26

27 Confidence in the Industry 27

28 Confidence in the Regulators AAPS webinar Sept First-In-Class Regulatory PBPK Modeling Guidelines from both Sides of the Pond Ping Zhao, Anna Nordmark. Very low confidence Not scientifically there yet. 28

29 Confidence in Regulators 48 food effect predictions, ~50% within 1.25-fold, 75% within 2-fold The large knowledge gaps in product, API, and physiology hinder the ability of PBPK to prospectively predict the food effect Our analyses have 3 implications: (1) laying out the strategy of using PBPK to predict food effect (2) identifying key parameters commonly optimized to better describe food effect and (3) providing a knowledgebase that can be expanded 29

30 What is needed to Build Confidence A consistent workflow with standardized inputs Key principles : Mechanism of food effect must be understood Model is validated against clinical food effect data before it can be applied to predict future food effect studies (e.g. for new formulations) Publications and cross-industry verification efforts 30

31 IQ PBPK Working Group Chair Arian Emami Riedmaier (AbbVie) Co-chair Neil Parrott (Roche) EISAI GSK DSI ROCHE PFIZER VERTEX AGIOS NOVARTIS GENENTECH TAKEDA MERCK Group Kick-off: January 2018 Ends: Dec 2019 Aim: To assess the predictive performance of mechanistic model prediction of food effect using a consistent strategy and input data. Highlight cases with high vs. low confidence and provide an industry best practice

32 Acknowledgements Colleagues from Roche pred Pharmaceutical Sciences Colleagues from the GastroPlus User Group Colleagues from the IQ Food Effect Working Group 32