Workshop 6 Modeling and Simulation to help …

1 Modelling and Simulation to help define MABEL and starting dose in FIH studiesB Laurijssens, BEL Pharm W Martin, Pharmacometrics Group, Dept Clinical Pharmacology, Pfizer, Sandwich Labs, Kent, 2 How should we select a starting dose ?Base the starting dose on toxicology in animalsBase the starting dose on Pharmacology in animalsBase the starting dose on toxicology in animals and expected pharmacology in humansSlide 3 What is the right thing to do ? accurately predict exposure in humans Scaling to man should use state of the art approaches PK scaling to humans of Pharmacology in Humans (MABEL) Inter-species differences in binding, relative time ( normalized to lifespan), signalling or pathway differences should be taken into account relationship in animals Need to take account of NOAEL To estimate a safety margin All predictions are relative to observed toxicology Pharmacology usually evident prior to a Safety Margin/Factor to predicted dose To ensure the safety and wellbeing of subjects in the trialSlide 44.

2 What drives the safety factor ?Allow a safety adjustment based on level of Margin depends upon patient population In oncology patients willing to accept smaller safety margin than in healthy risk moleculeswould be those that: Are novel Are very potent Are agonists Have low species cross-reactivity Have steep dose -response curves Have a high degree of uncertainty in the calculation of the consideration should be given to molecules with pleiotropic effect Biological cascade or cytokine release leading to amplification of an effect that might not be sufficiently controlled by a physiologic feedback mechanism (eg in the immune system or blood coagulation system)Slide 5 Pharmacokinetics1. Prediction in humans Allometric Scaling Mechanistic Models PBPK ModelsPharmacology1.

3 Prediction in humans Receptor occupancy (Kd) PK model + RO Mechanistic PKPD models1. Calculate MABEL2. Perform risk assessment of drug3. Potentially apply safety factor4. Quote FIH dose relative to NOAELHow do we determine safe starting dose ?Slide 6 Modelling is vital when the concentration or dose -response relationships are not straight delays between pharmacokinetic and Pharmacodynamics Active metabolites Equilibrium delays ( Oxycodone) indirect response Pharmacodynamics Response Biological systems that develop tolerance or rebound (Benzodiazepine & exaggerated anxiety) Pharmacological System Transduction delays ( AMG531, EPO, GNRH agonists) Drug response is the results of the interaction on many physiological pathways; inflammation, bone remodelling, Endocrine in translation of PD effect across speciesSlide 7 When would we want to use Modelling to estimate MABEL ?

4 When the drug response takes time to develop and is not directly related to plasma concentrations. May be due to an active metabolite or equilibrium delay between plasma concentration and biophase (eg Oxycodone) causing ng/mLElapsed time after dosingAnticlockwise HysteresisLalovic Clin Pharm Ther. 2006 DoseCpCePharmacokineticsDispositionBioph aseDistributionDirect PDResponseEquilibrium DelayPupil Diam(mm from Baseline)PharmacodynamicsSlide 801 2 3 4 Time (days)PK PDAMG531 When would we want to use Modelling to estimate MABEL ? When there is a lag time in drug response even after the drug reaches the biophase ( AMG531, EPO, INFa)DoseCpPkDispositionPDResponseBio-si gnalPDTransductiondelaysSlide 9 Outline: starting dose Calculation for FIH Predicting PK in humans Predicting human pharmacology (MABEL) Pulling it all together Working example ConclusionsSlide 10 Methods for predicting human PK scaling Simplest approach Very reliable for some small molecules and biotherapeutics(not TMD) Relates PK parameter to body weight2.

5 (Semi-)mechanistic-based PK models Takes into account species and disease state differences Better for extrapolation beyond dose -range Simcyp Simulator for small PK models May be most accurate method to predict across species [not being covered in this course]Slide 111. Allometric Scaling for PK PredictionAllometric scaling relates a variable to size Variable: Vd, CL, Size: Body weight, maximum life span, brain weight, body surface areaX = a * Wba: coefficient, b:allometric scaling exponent Usually performed on log-log scale Exponent tends to be similar across molecules CL: b= [ BSA] Vd: b = 1[ BW] t : b = [ ]Log[Bwt (Kg)]-2-10123 Log [Vss (mL)]-2024 Vss = 188*Wt( )MiceRhesus MonkeyHuman PredictedSlide 12 Methods for predicting human PK scaling Simplest approach Very reliable for some small molecules and biotherapeutics (not TMD) Relates PK parameter to body weight2.

6 (Semi-)mechanistic-based PK models More reliable for target-mediated disposition (membrane bound targets) Takes into account species and disease state differences Better for extrapolation beyond dose -range Simcyp Simulator for small PK models May be most accurate method to predict across species [not being covered in this course]Slide 132. Mechanistic PK models: why are they needed?Ti me (Days)071421283542 Serum Concentration (ng/mL)0. 111010010001000010000010000000. 003 mg/ kg0. 01 mg/ kg0. 1 mg/ kg1. 0 mg/ kg3. 0 mg/ kgTi me (Days)0714212835420. 11101001000100001000001000000 Note: Non-linearity occurs where Cp >>> Kd (3x10-12M)Membrane bound, internalizing antigen which binds MAb and undergoes targetmediated disposition (TMD)01002003004005001251020 50100 Time (h)Concentration (ug/mL)Predicted concentrationsObserved concentrationsParameters Scaled to Humans Inter-compartmental Cl and VoL Production and elimination of antigeNHuman parameters Linear component cl and voMonkeysHumans PredictedHuman ObservedSlide 14 Outline.

7 starting dose Calculation for FIH Predicting PK in humans Predicting human pharmacology (MABEL) Pulling it all together Working example ConclusionsSlide 15 Methods used for predicting human pharmacology Predicted receptor occupancy (Kd) PK model coupled with receptor occupancy PK model coupled with ex vivo assay PK model as empirical PK/PD model in absence of biomarker or other in vivo PD data Mechanistic PKPD modelsSlide 16 Why is receptor occupancy not straight forward, especially for Biotherapeutics?1.[MAb] is similar to target levels More complex equation distribution leads to Cplasma>> Cbiophase Target Cplasmaneeds to be much and off-rates at receptor quite slow for Mab Equilibrium calculations ( Kd based) may over predict receptor (receptors or ligands) have inherent turnover of target by antibody changes kinetics of target ( internalization) resulting in RO that is not predictable by pharmacological equilibrium approachesSlide 171.

8 Estimating Receptor Occupancy: MAbStandard Receptor Occupancy Calculation for [MAb]>>target (Also termed the Kd of Duff calculation) What do we need to What do we need to calculate RO calculate RO Only KDSlide 182. Diffusion barrier to biophasePanitumumab (Anti-EGF MAb) Plasma concentration 100- fold >> Tissue ConcentrationSlide 193. Impact of changing Koffrateon RO calculationskoffkonAgAbRinkdegDosekel+Ab AgkintAbPKPD Model (TMD)DosekoffkonAg+AbAgAbDuff FormulaSlide 204. Impact of Target Turnover on RO Estimation At high receptor turnover rates, higher molar excess of MAb is required (Not accounted for in simple Kd model) turnover half- life (hours)10-310-210-1 PKPD predictionDuff formulaDose For 10% Receptor OccupancykoffkonAgAbRinkdegDosekel+AbAgk intAbPKPD Model (TMD)DosekoffkonAg+AbAgAbDuff FormulaSlide 214.

9 Impact of target/receptor turnover: variability across species and targetSlide 22 Methods used for predicting human pharmacology Predicted receptor occupancy (Kd) PK model coupled with receptor occupancy PK model coupled with ex vivo assay PK model as empirical PK/PD model in absence of biomarker or other in vivo PD data Mechanistic PKPD modelsSlide 234. Impact of target turnover: Impact on FIH dose selectionSubcutaneous DosingSlide 24 PKPD model-based approach to MABEL Establish a mechanism-based model in a relevant animal species to demonstrate the relationship between dose and RO. Determine RO and pharmacological effect relationship. Extrapolate model to humans using all relevant data (literature,in vitro human etc) Perform simulations considering both uncertainty in model parameters and in scale-up This approach should help select a more rational starting dose for FIH within the minimum anticipated biological effect level (MABEL) principles using all relevant literature and project level dataSlide 25 dose (mg/kg) Occupancy (%)020406080100 What is an appropriate level of receptor occupancy for first dose ?

10 90% RO may be appropriate for certain ANTAGONISTS<10% RO is more appropriate for an AGONIST~10% RO may be more appropriate for most ANTAGONISTWill also depend on: target (novel? ) therapeutic area (oncology?) preclinical toxicology profile pharmacology (immunomodulatory?) Effector Mechanisms ADCC or ADC are NOT directly related to ROSlide 26 FIH starting dose Working Example Molecule:IgG1 Fusion Protein(Peptide) NPLATE (AMG531) Action:Agonist Mechanism: Promotes the viability and growth of megakaryocyte colony-forming cells, increasing platelet production Target: Binds the TPO receptor (peptide is distinctly different from TPO) Competitor info: Similar mechanism Genentech rhu-TPO, Amgen PEG-MGDF Background information: Molecular weight = 59,000 Da Platelets x 1011L TPO Receptors/ Platelet 56 (Baseline) TPO Receptors nM Kd = 14 nMSlide 27 AMG-531 Preclinical Data Available In vitro studies Platelet binding relative to TPO Rabbit, mice, rat, monkey, human MK-CFU proliferation assays Monkey and humanIn vivo studies PK FcRn knock-out and wild type (single dose ) 1 mth tox in rat and monkeySlide 28In vitro results: Platelet binding across speciesCompetition assay with TPO number of receptors/platelet human : ~60 monkey : ~30 rat : ~6 rabbit : ~2MK-CFU proliferation assays Similar results in both monkeys and humans for AMG531 and MGDFWhat can you conclude from this information?