HCV Model Development: Industry Perspective Larissa Wenning Quantitative Translational Models to Accelerate Hepatitis C Drug Development August 2, 2012 Modeling & Simulation Integrating knowledge, enhancing decisions 1 What Do We Want From HCV Models? INPUTS OUTPUTS Data Portable, integrated form of knowledge Clearly Defined Assumptions MODELING Ideas & Scientific Knowledge Integrated, mathematical representation of all inputs Exploration of Knowledge Gaps Enhanced Understanding Predictions vs. Observations Enhanced Decision Making Modeling & Simulation Integrating knowledge, enhancing decisions 2 What Kind of Questions Might We Answer With Models? • What is the therapeutic window for my compound? – Is there a dose where we can maximize efficacy and minimize adverse events? • What are the optimal combinations of compounds? • What is the optimal dosing regimen and duration of therapy for each of the many patient populations we are interested in? • What is the impact of factors that alter the pharmacokinetics of my compound(s) on efficacy and/or safety? – Drug-drug interactions, formulation changes, special populations Modeling & Simulation Integrating knowledge, enhancing decisions 3 What Will It Take To Get to Enhanced Decision Making? • Flexible, standardized model structures • Understanding of relationship between in vitro and in vivo data • Ability to leverage data from the outside world Modeling & Simulation Integrating knowledge, enhancing decisions 4 HCV Viral Dynamics Model: End Goal System parameters •Rates for infection & production of virus •Rates for clearance of virus & hepatocytes •Regeneration of hepatocytes Patient parameters Patient population (naïve, experienced, null, etc) •PR responsiveness •IL28B genotype •Other factors relevant for IFN-free regimens? New Infection Infection Clearance Wild Type Infected Cells Uninfected Cells New Infection Infection Regeneration Resistant Infected Cells Baseline HCV •Pre-existing RAVs from prior treatment or polymorphisms •High vs low baseline viral load Modeling & Simulation Integrating knowledge, enhancing decisions Clearance Drug Inhibits Production of Virus; Higher potency for Wild-Type than Mutant Efficacy •Dose/exposure response •Against different GTs and RAVs Clearance HCV genotype •GT1, 2, 3, etc Wild Wild-Type Type Virus Virus Drug parameters Resistant Virus Clearance Core viral dynamics model with clear separation between parameters associated with the biological system (virus, hepatocytes), and those associated with the drug; can “plug and play” parameter sets to simulate combinations of drugs in different populations of interest Resistance •Shift in drug efficacy •Baseline amounts and relative fitness of RAVs selected by drug Combinations •Efficacy is additive, synergistic, etc? •Resistance with combination 5 Complex, Multifactorial Problem • As a single company, we do not have enough data to address all of the relevant factors in a timely manner! • Must draw data from the outside world & leverage nonclinical data Modeling & Simulation Integrating knowledge, enhancing decisions 6 Merck HCV Viral Dynamics Model: Current State Drug effect only on production of virus Iwt Vwt T Wild-type infected cells (Iwt) c Vwt pwt (1-wt) (1 – ζ) Iwt Wild-type virions (Vwt) Not shown in diagram, but RBV treatment assumed to result in production of noninfectious virus, which also decays at rate c; then measured total virus = infectious + non-infectious pwt (1-wt) ζ Iwt Uninfected cells (T) pm (1-m) ζ Im T + Iwt + Im = T0 Vm T Mutant infected cells (Im) Im Total # of hepatocytes assumed to remain constant (T0) Modeling & Simulation Integrating knowledge, enhancing decisions pm (1-m) (1 – ζ) Im Mutant virions (Vm) c Vm Model applied to several compounds: MK-5172, MK-7009, Peg-IFN, RBV, boceprevir, etc 7 Example 1: MK-7009 Developing a Model for Multiple Compounds & Patient Populations • M&S objective: improve the understanding of MK-7009 dose and treatment duration needed to cure HCV in combination with SOC treatment of peg-IFN and RBV, accounting for viral dynamics with resistant virus • Approach: pool data across multiple studies, including MK-7009 monotherapy, MK-7009 + PR and PR alone (IDEAL study) Poland et al, American Conference of Pharmacometrics, San Diego, CA, April 2011. Modeling & Simulation Integrating knowledge, enhancing decisions 8 Model structure is flexible enough to represent range of behaviors in viral load 5. Example Individualsubjects Predictions Illustrative example Figure showing fit to 3 individual in an MK-7009 Study) by PR x 44 wks): Phase II study (MK-7009+PR x(Phase 4 wks 2A followed Obs. Obs. Unused in Fit Ind. Pred. Observations, Predictions (log10 IU/mL) 0 Patient 1 ID:72965 5 Pop. Pred. 10 Ind. Pred. Mutant 15 Patient 2 ID:72966 Patient 3 ID:73242 6 4 2 0 -2 -4 0 5 10 15 0 5 10 15 Time (weeks) Modeling & Simulation Integrating knowledge, enhancing decisions 9 Can account for different patient populations using different parameter distributions • Example: response to treatment with PR using data from IDEAL study • IDEAL study is in treatment naïve population, but contains those who will in the future be treatment experienced. “Treatment Naïve” “Treatment Experienced” Proportion R0 δ ED50peg SVR 40% 1.86 0.245 0.389 Subgroup Null responder 20% 3.07 0.184 1.60 Partial responder 10% 2.56 0.193 Relapser 10% 2.52 Other 20% All 100% Subgroup Modeling & Simulation Integrating knowledge, enhancing decisions Proportion R0 δ ED50peg Null responder 33% 3.07 0.184 1.60 0.683 Partial responder 17% 2.56 0.193 0.683 0.227 0.454 Relapser 17% 2.52 0.227 0.454 2.46 0.214 0.771 Other 33% 2.46 0.214 0.771 2.36 0.219 0.743 All 100% 2.69 0.203 0.978 10 Final Model Can Be Used to Simulate Many Scenarios 100% 100% 90% 90% 80% 70% 60% 50% 40% 30% At 48 Wks Treatment At 24 Wks Treatment At 12 Wks Treatment At 4 Wks Treatment 20% 10% Proportion of Patients Reaching SVR Threshold Proportion of Patients with Undetectable HCV RNA Example: Simulated MK-7009 Dose-Response with PR in Treatment-Naïve Patients 80% 70% 60% 50% SVR After 48 Wks Tx SVR After 24 Wks Tx SVR After 12 Wks Tx SVR After 4 Wks Tx 40% 30% Bars: 90% Prediction Interval 20% 10% Bars: 90% Prediction Interval 0% 0% 0 200 400 600 800 1000 MK-7009 Dose (mg BID) Modeling & Simulation Integrating knowledge, enhancing decisions MK-7009+PR through treatment period 0 200 400 600 800 1000 MK-7009 Dose (mg BID) 11 Example 1 Conclusions • A relatively simple viral dynamics model can predict short-term and longer-term response to HCV treatment with peg-IFN+RBV, protease inhibitor monotherapy, and triple combination therapy, in patients with little or no prior treatment. • With a very small estimated ED50, MK-7009 BID administered with Peg-IFN and RBV is predicted to sharply improve SVR over Peg-IFN and RBV alone. • Simulations show that proportion of patients cured increases with treatment time and continues to increase long after proportion with undetectable virus plateaus Modeling & Simulation Integrating knowledge, enhancing decisions 12 Example 2: MK-5172 Leveraging in vitro data • M&S objective: use existing viral dynamics model (developed for MK-7009+/-PR), clinical data from a monotherapy study, and in vitro data to project clinical response for MK-5172 in a number of scenarios • Challenge: Monotherapy data includes patients infected with GT1 and GT3. GT3 data shows dose response, but GT1 does not (all doses appear maximally efficacious) • Approach: Fit monotherapy data for GT3 and GT1 simultaneously and assume relative potency observed in vitro (24-fold more potent for GT1 vs GT3) translates directly in vivo; leverage existing model for PR to simulate combination of MK-5172 +PR Nachbar et al., EASL 2012 & 7th International Workshop on Hepatitis C Resistance & New Compounds Modeling & Simulation Integrating knowledge, enhancing decisions 13 Data & Model Fit GT1 GT3 2 2 2 2 0 2 V0 0 2 V0 0 2 V0 0 2 4 6 4 6 log10 4 6 log10 400 mg QD log10 50 mg QD V0 400 mg QD log10 50 mg QD 4 6 8 8 0 10 20 30 40 50 60 time d time d time d 600 mg QD 100 mg QD 600 mg QD 0 2 4 0 2 4 0 2 4 time d Modeling & Simulation Integrating knowledge, enhancing decisions 6 8 0 10 20 30 40 50 60 time d log10 2 4 6 8 800 mg QD V0 V0 V0 0 10 20 30 40 50 60 2 4 2 0 0 10 20 30 40 50 60 time d 0 10 20 30 40 50 60 time d 200 mg QD log10 6 8 2 0 0 10 20 30 40 50 60 6 8 time d 800 mg QD log10 2 4 log10 time d 200 mg QD 2 0 0 10 20 30 40 50 60 6 8 log10 0 10 20 30 40 50 60 6 8 V0 0 2 4 V0 2 V0 2 6 8 0 10 20 30 40 50 60 time d 2 time d V0 8 0 10 20 30 40 50 60 2 log10 log10 V0 100 mg QD log10 8 0 10 20 30 40 50 60 2 0 2 4 6 8 0 10 20 30 40 50 60 time d 14 Monotherapy Predictions GT 1 GT 3 1 1 0.01 4 50. mg QD 5. mg QD 10. mg QD V0 V0 30. mg QD 10 0.01 5. mg QD 10. mg QD 30. mg QD 10 4 50. mg QD 100. mg QD 100. mg QD 200. mg QD 10 6 400. mg QD 200. mg QD 10 6 400. mg QD 600. mg QD 600. mg QD 800. mg QD 10 8 0 10 20 30 40 time d • Dose differentiation for GT1 predicted to be at or below 10 mg Modeling & Simulation Integrating knowledge, enhancing decisions 800. mg QD 10 8 0 10 20 30 40 time d • 50 mg dose for GT3 predicted to be no different than placebo 15 Setting up Simulation of Combination Therapy: Simulation for Efficacy Against RAVs ED50,m 10 ED50, wt 0.01 0.01 0.01 0.01 10 4 10 6 10 8 2 0 2 4 6 8 10 10 4 10 6 10 8 2 0 2 4 6 8 10 V0 1 V0 1 10 4 10 4 10 6 10 6 10 8 10 8 2 0 2 time d 4 6 8 10 2 1 0.01 0.01 0.01 0.01 10 10 6 10 8 2 0 2 4 6 8 10 10 10 6 10 8 V0 1 4 2 0 2 4 time d 2 6 8 10 10 4 10 4 10 6 10 6 10 8 10 8 2 0 2 4 time d 4 6 8 10 6 8 10 50 mg QD 100 mg QD 200 mg QD 400 mg QD 600 mg QD 800 mg QD time d 1 4 0 100 ED50, wt time d 1 V0 V0 ED50,m 1 time d • 30 ED50, wt 1 time d GT3 ED50,m V0 GT1 3 ED50, wt V0 V0 ED50,m 6 8 10 2 0 2 4 time d Simulate total viral load for a range of ED50,m/ED50,wt ratios to determine reasonable range for ED50,m – Sizable breakthrough on treatment in simulations for 30- and 100-fold shift in potency against resistant virus – Fold shift in potency against resistant virus therefore not greater than 10 Modeling & Simulation Integrating knowledge, enhancing decisions 16 Simulation of Combination Therapy Percent below limit of detection • Very high percentage of patients are expected to become undetectable quickly, and remain so while on therapy Modeling & Simulation Integrating knowledge, enhancing decisions 17 Simulation of Combination Therapy Projected % Breakthrough, Relapse, and SVR 12 weeks of MK 5172 PR Followed by 36 Weeks of PR Alone 48 Weeks Total Treatment dose mg Assumed ED50,m shift from ED50,wt breakthrough 10 30 50 100 200 400 3 4 4 3 3 3 3 fold relapse 6 4 3 2 2 3 SVR 80 88 91 94 95 95 10 fold breakthrough relapse 1 3 3 3 3 3 6 5 4 4 3 2 SVR 74 80 84 89 92 94 12 weeks of MK 5172 PR Followed by 12 Weeks of PR Alone 24 Weeks Total Treatment dose mg Assumed ED50,m shift from ED50,wt breakthrough Modeling & Simulation Integrating knowledge, enhancing decisions 10 30 50 100 200 400 2 2 2 2 2 2 3 fold relapse 14 11 10 8 8 7 SVR 72 83 86 89 91 92 10 fold breakthrough relapse 1 2 2 2 2 2 15 14 12 11 9 7 SVR 63 71 78 83 88 90 18 Example #2 Conclusions • Monotherapy study: – In vitro data has been used successfully to bridge efficacy between genotypes in a viral dynamics model – This tactic may have broader utility to inform relative potency for genotypes and RAVs in these models for early clinical response prediction – For GT1: 10 mg QD dose is predicted to be noticeably less efficacious compared to higher doses – For GT3: 50 mg QD dose is predicted to be similar to placebo in terms of viral load decline • Subsequent studies: – Simulations with the 2-species combination treatment model predict high SVR rates with low viral breakthrough due to RAVs – Comparison of future clinical results to such prospective predictions is planned to further evaluate this early response prediction approach Modeling & Simulation Integrating knowledge, enhancing decisions 19 Conclusions & Wrap-Up • HCV viral dynamics models have the potential to be very useful tools for enhancing decision making by development teams • Flexible, standardized model structures & ability to leverage outside and non-clinical data are critical in the current fast-moving, ever-changing development environment for HCV Modeling & Simulation Integrating knowledge, enhancing decisions 20 Acknowledgments • The M&S Network at Merck • Merck’s HCV M&S team: Bob Nachbar, Luzelena Caro, Julie Stone, many others! • Project teams for MK-7009 and MK-5172 • Bill Poland; Pharsight • John Tolsma, Haobin Luo, Jonna Seppanen; RES Group Modeling & Simulation Integrating knowledge, enhancing decisions 21 Back-Up Slides Modeling & Simulation Integrating knowledge, enhancing decisions 22 Combination Therapy Model wt m wt m wt,ni m,ni healthy hepatocytes hepatocytes infected with wildtype virus hepatocytes infected with mutant virus wildtype virus mutant virus wildtype noninfectious virus mutant noninfectious virus T 0 c pwt pm wt Modeling & Simulation Integrating knowledge, enhancing decisions m total number of hepatocytes hepatocyte infection rate infected hepatocyte death rate viral clearance rate mutation rate wildtype viral production rate mutant viral production rate combined drug and IFN effectiveness on wildtype virus combined drug and IFN effectiveness on mutant23 virus RBV effectiveness