PBPK Modelling of Food Effects PBPK SYMPOSIUM 2018 Paris, April 4 th 2018 David Turner (Presented by Dr Sebastian Polak) david.turner@certara.com
Outline Modelling of Food Effects with Population PBPK Models Summary of Food Effects Currently Handled Mechanisms Gaps and Issues Summary 2
Oral Absorption: ADAM Framework & Population Variability Regional Mean and Inter-individual Variability of: ph, [bile salts], fluid volumes, viscosity, [bicarbonate], gut wall morphology, blood flow, etc. Regional Mean and Inter-individual Variability of Derived Parameters: bulk solubility, surface solubility, free fraction, dissolution rate, permeation rate, metabolism etc. 100 simulated subjects CAT : Lu et al., 1996; ACAT: Agoram et al., 2001; ADAM: Jamei et al., 2009 3
Effect of Food on Physiological Parameters Available in Simcyp ph IMMC Cycle (EHC model) Bile Salts Conc. Effect of Food Luminal Fluid Viscosity Fluid Volume Dynamics Buffer Capacity Transit Times Splanchnic Blood Flow
1. ph Gastric ph Change / Return Time to Fasted Gastric ph after Food Gastric ph rises with food 1.5 5.0 (NEC HV Population representative) Gastric ph vs. Time T ff : the period during which the stomach ph returns to the fasted ph (h): (mean ± CV%) ph Young: 1.8h±65 Elderly: 3h±80 If age i < 65: ph ST,t = ph ST,fed,i e -kt Otherwise (age i >= 65): ph ST,t (T ff t) ( ph ST,fed,i T ff - ph ph ST,fasted,i ST,fasted,i ) Data from Russell et al. 1993 Time After Meal (mins.)
1. ph Fasted and Fed ph (dynamic, modifiable V18) 6
2. GIT Transit of Drug / Formulation Segregated Transit Model Formulation tabs Substrate and Inhibitor Independent (except fluid with dissolved drug) 7
2. GIT Transit Complication: Gastric Emptying Fed State Magenstrasse Gastric emptying and intestinal transit measurements should be of the drug and/or formulation not the food itself Fed state gastric emptying is generally assumed slower than fasted BUT The Magenstrasse (stomach road) describes the fast emptying of ingested liquids from the postprandial stomach. The occurrence of the Magenstrasse has great importance for drugs administered together with food as it represents a shortcut through the fed stomach and allows rapid onset of plasma levels. (Provided dissolution in the stomach is sufficiently rapid) the effect of fat content of the meals (on the Magenstrasse) was negligible 8
3. Simulated Dynamic Luminal Fluid Volumes: Fed State Multiple Doses, each after food V fluid,stomach, t=0 = Fluid in Food + Fluid in Drink + Baseline fluid = ~840 ml (PopRep) 9
3. Fed State Intestinal Fluid Volume Dynamics are Poorly Understood Grimm, Weitschies 2017 MolPharm 15 548 10
3. Fed State Intestinal Fluid Volume Dynamics are Poorly Understood 11 subjects 8 m / 3 fe 23 25 years BMI 24-25 MRI imaging at 1 hr intervals Meal: High Fat LBG locust bean gum Fasted Base line 70 100 ml These high-fat meals stimulated substantial increases in SBWC, which increased to a peak at 4 h at 568 ml Highly variable between subjects: peak 150 854 ml. * Hussein, Marciani et al. 2015 11
3. Fed State Intestinal Fluid Volume Dynamics are Poorly Understood 11 subjects 8 m / 3 fe 23 25 years BMI 24-25 MRI imaging at 1 hr intervals Meal: High Fat LBG locust bean gum Fasted Base line 70 100 ml These high-fat meals stimulated substantial increases in SBWC, which increased to a peak at 4 h at 568 ml Highly variable between subjects: peak 150 854 ml. * Hussein, Marciani et al. 2015 12
3. FDA Standard High Fat Meal However, to the best of our knowledge, the small intestinal fluid dynamics after the FDA standard breakfast were not investigated so far. Koziolek, Weitschies 2016 Adv Drug Del Reviews 101 75 Also, confirmed (March 2018) by Prof Luca Marciani (University of Nottingham) Gastric fluid volume dynamics with standard high fat meal Simcyp V18 13
4. Food Effects Bile Salt Concentrations 20 15 FASTED FED Luminal Bile Conc. (mm) 10 5 CMC < 1 mm 0 Stomach Jejunum I Duodenum Jejunum II Ileum I Ileum II Ileum III Ileum IV Colon 1-Kalantzi et al. 2006; 2-Mithani et al. 1996; 3-Zimmermann et al. 1994; 4-Barnwell et al. 1993 and others Mean, CV and regional differences saved in Simcyp Population File Fed is medium to high fat conditions
4. Food Effects Predicting Drug Partition into Bile Salt Micelles A model is required to capture inter-individual variability of ph and [bile salt] Binding to micelles is characterised by a micelle:aqueous partition coefficient (K m:w ) IF [Bile Salts] > CMC (Critical Micelle Concentration) ELSE S (BS)Tot = So + S ionised where S ionised = So 10 pka-ph Glomme, März, Dressman (2004) e.g., for a Monoprotic Base Micelle partition coefficients (K m:w ) predicted from logp o:w or estimated from modelling of in vitro solubility experiments (SIVA Toolkit)
4. Impact of Micelle Partition upon Solubility and Dissolution Rate Effective Diffusion Coefficient (D eff ) is reduced where there is significant bile micelle partition D eff = f neutral D neutral + f ion D ion + f micelle D micelle D micelle = 0.78 D neutral = 4.4 x 10-4 cm 2 /min (typical micelle) (small molecule with MWt around 350 Da) 5 4 Typical Calculated D eff vs. [Bile Salts] Profile D eff (10-4 cm 2 / min) 3 2 1 0 0 5 10 15 20 25 [Bile Salts] (mm)
5. Food Viscosity Food can impact disintegration and dissolution rate Food viscosity may impact diffusivity (D eff ) and h eff hence dissolution rate Requires Fluid Dynamics Model to be Active (Particle Motion Equations) Assumes gastric contents are well-mixed (but Magenstrasse) Dynamic Viscosity based upon dilution to be added to Simcyp V18
5. Accounting for Hydrodynamics: Effective Diffusion Layer Thickness, h eff DR(t) i 1 NBINs Deff ( t) N i S 4 ai ( t) ai ( t) heff, i ( t) (Ssurface Cbulk ( t) ) h ( t) eff, i Fluid Dynamics h eff Equations radius viscosity shear D eff = f aq,n D mono,n + f aq,ion D mono,ion + f micelle D micelle Effective Diffusion Coefficients Re total Sc Sh f 2a( t) U f U p f f D eff D f eff 1/ 2 1/3 2 0.6 Re Sc h eff 2a( t) Sh Relative Velocity Terms? Complex CFD Modelling Representative Velocities (USP2, µdiss) (axial and tangential) CFD models: USP 2: Bai et al. 2007 µdiss: Johansson et al. 2017 18
6. Time-dependent Fed State Intestinal Blood Flow Splanchnic Blood Flow Increase for duration of fed state Clinical Studies Mean & SD Splanchnic Fed/Fasted Blood Flow Ratio Simulations Solid line Mean prediction Dashed lines 5 th and 95 th percentiles 100 healthy volunteers Time (h) after Meal Rose et al. 2017 The AAPS Journal 19 1205.
6. Blood Flow Supply to Small Intestine Evidence: Reports indicating reduced hepatic extraction ratio for high extraction compounds such as propranolol. Accounting for the time-dependent change in splanchnic blood flow is important to recover the increase in exposure to orally administered high extraction compounds in the fed state. Time-independent increase in blood flow Fed/fasted ratio: AUC = 1.01 C max = 1.14 Time-dependent increase in blood flow Fed/fasted ratio: AUC = 1.28 C max = 1.32 Rose et al. ACoP, Las Vegas, USA, Oct 2014; Rose, Turner et al., 2017 AAPS Journal 19 1205.
7. Buffer Capacity Particle Surface ph Particularly in the fasted state buffer capacity is low Particle surface ph reduced (acidic dugs), increased (basic drugs) Particle surface solubility reduced and dissolution rate If surface ph is not accounted for in the fasted state then fasted-fed differences may not be picked up Cristofoletti and Dressman 2016 J Pharm Sci 105(12):3658 21
7. MechPeff Model ONLY Way to Capture Free Fraction driving Permeation Solid Drug Gut Lumen Lumen Wall Unstirred Boundary Layer (UBL) Enterocytes Dissolution 1) Free Fraction = non-micelle bound fraction Free fraction = 1 f micelle Food [bile] f micelle 2) Luminal Wall UBL Permeability (P UBL ) Food f micelle D eff,ubl P UBL P eff (ONLY if P eff is UBL limited High Permeability Drugs) Adapted from Sugano 2009 Interplay!
8. Enterohepatic Recirculation Model (Gallbladder Kinetics) Kinetics of drug in bile differ according fasted or fed state Dose Given Fasted state simulation assumed to start at a random point during the IMMC GB Emptying, No Accumulation in GB GB Accumulation + intermittent GB Emptying Second and subsequent doses are assumed to be taken after additional food and the process returns to Start Fed State 23
Dynamic Bile Salt Model (Simcyp V18 preliminary results) Fully integrated with (and inter-dependent on) dynamic fluid volumes model 24
Wide variability in nature and extent of FE for CR Formulations Clinically Observed FE with CR Formulations of Nifedipine QSAR or classifier models cannot predict FE for CR/MR formulations as they are based on the drug (API) rather than the formulation Data: Toal, 2012, IJCPT 50 202
Zolpidem Negative Food Effect Attributed to Uncharacterized Impact of Meal on Release from Formulation 26
Way forward - Dynamic Viscosity-Disintegration Model: Negative Food Effects`` Relationship between Disintegration Constant and Viscosity Viscosity changes predicted using ADAM fluid volume dynamics model ADAM PBPK Model Predicted in vivo disintegration profile Negative food effect of Trospium chloride Poster: Orbito 2015; ms. in prep
Case Study: Simulating Food Effects Precipitating Compounds Cristofoletti et al. 2016 105 2712 Ketoconazole Posaconazole Fasted Fed Fed Fasted Predict-learn-confirm model development process Estimated precipitation kinetics parameters using fasted PK data and used to predict fed state
FDA View / ASCPT Paper The published literature and perhaps also the results from submissions to the FDA may be biased, because only the good results tend to be published or submitted, whereas the true picture may be lost 29
Some Issues Raised by Li et al The food effect may affect intestinal enzymes and transporters. Comment: The mechanisms are available in PBPK models but not the required kinetic data and (unbound) amounts of various active components in a particular meal (if the active component is digestible or absorbable there is an extra issue). The mobility of food components is largely affected by the presence of fluids in the GI tract. Comment: Food effects on luminal fluid volumes are highly variable and depend on the composition of the meal (see slides). There are no published data on small intestinal fluid volumes even after a standard FDA meal. Especially for highly permeable weak bases, in vitro setups to measure precipitation, such as the transfer model, were demonstrated to over-predict the extent of precipitation in vivo. Comment: Hens et al, 2017 Mol Pharmaceut 14(12):4321 demonstrate with posaconazole BCS 2 (fasted conditions) that over-prediction can be avoided with suitable in vitro data, suitable models and modelling of the IN VITRO data (SIVA Toolkit) prior to PBPK modelling. 30
Summary Food Effects Are complex and either not well understood or not well characterised even for a standard FDA high fat meal Can be assessed in complex experiments such as the TNO TIM Mechanistic models are required to extrapolate to different regions of the GIT, individuals and different populations Mechanistic modelling of complex TNO TIM type experiments is very difficult or impossible; it makes sense to simplify experiments such that mechanisms and parameters can be captured by modelling of those experiments: - requires a range of suitable experiments to be identified and performed - provides mechanistic understanding Population Variability Many (most?) physiological parameters associated with the GIT have high intersubject variability (and sometimes intra-subject variability) so it makes sense to simulate virtual clinical trials rather than a single average subject. 31
Some Publications: Food Effects Cristofoletti Patel N, Dressman JB (2016) Assessment of Bioequivalence of Weak Base Formulations Under Various Dosing Conditions Using Physiologically Based Pharmacokinetic Simulations in Virtual Populations. Case Examples: Ketoconazole and Posaconazole. J Pharm Sci 106(2): 560. Cristofoletti R, Patel N, Dressman JB (2016) Differences in Food Effects for 2 Weak Bases With Similar BCS Drug- Related Properties: What Is Happening in the Intestinal Lumen? J Pharm Sci. 105(9):2712. Andreas CJ, Pepin X, Markopoulos C, Vertzoni M, Reppas C, Dressman JB (2017) Mechanistic investigation of the negative food effect of modified release zolpidem. Eur J Pharm Sci. 102:284-298. Patel N, Polak S, Jamei M, Rostami Hodjegan A, Turner DB (2013) Quantitative prediction of formulation-specific food effects and their population variability from in vitro data with the physiologically-based ADAM model: A case study using BCS/BDDCS Class II drug Nifedipine. Eur J Pharm Sci 57 240. Rose RH, Turner DB, Neuhoff S, Jamei M (2017) Incorporation of the Time-Varying Postprandial Increase in Splanchnic Blood Flow into a PBPK Model to Predict the Effect of Food on the Pharmacokinetics of Orally Administered High-Extraction Drugs. AAPS J 19(4): 1205. Modelling In Vitro Experiments Pathak S, Ruff A, Kostewicz ES, Patel N, Turner DB, Jamei M (2017) Model-based Analysis of Biopharmaceutical Experiments to Improve Mechanistic Oral Absorption Modelling - An Integrated in Vitro in Vivo Extrapolation (IVIV_E) Perspective using Ketoconazole as a Model Drug. Mol Pharmaceut 14(12): 4305. Hens B, Pathak SM, Mitra A, Patel N, Liu B, Jamei M, Brouwers J, Augustijns P, Turner DB (2017) In Silico Modeling Approach for the Evaluation of Gastrointestinal Dissolution, Supersaturation and Precipitation of Posaconazole: A Simulation Study Using the Physiologically-Based ADAM Model. Mol Pharmaceut 14(12): 4321. Simcyp ADAM Model Jamei M, Turner D, Yang J, Neuhoff S, Polak S, Rostami-Hodjegan A, Tucker G. Population-based Mechanistic Prediction of Oral Drug Absorption AAPS J. 2009, 11(2): 225-237. 32
Acknowledgements david.turner@certara.com Simcyp Team (Oral absorption / Gut Wall) Nikunj Patel, Shriram Pathak, Masoud Jamei, Deven Pade, Ali Nimavardi, Sumit Arora, Kostas Stamatopoulos, Sibylle Neuhoff, Matt Harwood, Oliver Hatley, Sebastian Polak Parts of this work has received support from the Innovative Medicines Initiative Joint Undertaking (http://www.imi.europa.eu) under grant agreement n 115369, resources of which are composed of financial contribution from the European Union's Seventh Framework Programme (FP7/2007-2013) and EFPIA companies in kind contribution. 33