Drug-Disease Modeling Applied to Drug Development and Regulatory Decision Making in the Type 2 Diabetes Mellitus Arena

Transcription

1 Drug-Disease Modeling Applied to Drug Development and Regulatory Decision Making in the Type 2 Diabetes Mellitus Arena Tokyo, Japan, December 8, 2015 Stephan Schmidt, Ph.D. Assistant Professor Center for Pharmacometrics and Systems Pharmacology University of Florida at Lake Nona

2 Overall Trend in R&D Efficiency Scannel et al. (2012) Nat Rev Drug Disc 11:

3 Current Trends in Drug Development: Costs Average cost to develop a new drug is ~ $1.2 Billion * Adapted from Biopharmaceutical Research Industry 2013 Profile. Pharmaceutical Research and Manufacturers of America

4 The Learn/Confirm/Apply Paradigm From today s phased approach: Discovery Preclinic Phase I Phase II Phase III Phase IV To Learn & Confirm: Discovery Learn Confirm Optimize understanding to maximize medical value Optimize execution 4

5 Quantitative Pharmacology (M&S) Visser et al. (2014) CPT Pharmaco Sys Pharmacol 5

6 Keys to Successful Use of M&S For Strategy and Decision-Making in Drug Development Who is the patient? Define the shape of D/R or PK/PD curves Disease biomarkers Identify the right target Understand pathophysiology 6

7 3 Key Questions that Define the Context for M&S What do we want to know? How certain do we need to be? What are we willing to assume? ALL MODELS ARE BUILT FIT FOR PURPOSE 7

8 What Do We Want to Know? EFFECTS REGIMEN PATIENT DRUG RESPONSE SURFACE OR USERS MANUAL 8

9 What Do Clinicians Need to Know? What is a reasonable initial dosage regimen? How to adjust for intrinsic and extrinsic factors? When will effect be seen? When will effect plateau? How to know when to change dose? What happens when dose is skipped? 9

10 How Certain Do we Need to Be? 10

11 What Are We Willing to Assume? Lung Fat Drug in Drug out Barrett, ASCPT,

12 Application: Clinical Trial Simulations Adapted from: Garhyan et al. In: Applied Pharmacometrics (2014)

13 What Impact Does This Have on Regulatory Decision Making? Critical Part of Regulatory Decision-Making In theory, any and all clinical situations where physicians need information about dosing can be tested during drug development. However, ethical and practical limit the number of studies that a sponsor can conduct. CONDUCT LEARN CONFIRM PREDICT WAIVE 13

14 Challenge: Relating Knowledge & Data Aronow, AAPS,

15 Challenge: Different Time Scales for PK&PD 15

16 Implications for Drug-Disease Models OUTCOME (YEARS) PHARMACODYNAMICS EFFECT Slow biomarker(s) (MONTHS) PHARMACOKINETICS Fast biomarker(s) (MINUTES to HOURS) DOSE (PK) (MINUTES to HOURS) 16

17 Drug-Disease Model Setup Pharmacokinetics (PK) Pharmacodynamics (PD) Drug Drug Conc. Biomarker Effect PK/PD Models 17

18 Case Study: Diabetes Mellitus Is a chronic progressive disease One of the top 10 leading causes of death Type 2 is the most common form To date, >8% of the global adult population (>380 million people) are affected (T2DM) Expected rise to ~600million people worldwide by 2035 Cost: 11% of the global healthcare budget in 2013 ($US 548 billion) and rising ($US 612 billion) 18

19 19

20 Key Processes in Glucose Homeostasis 20

21 Biomarkers Biomarkers are generally defined as: Characteristics that are objectively measured and evaluated as indicators of Fasting Plasma Glucose normal biological processes, Fasting Serum Insulin pathogenic processes, or Biomarker for diabetes Biomarker for diabetes pharmacologic responses to a Normal level mmol/l and insulin tolerance t 1/2 dependent on insulin concentrations therapeutic intervention HbA1 c Biomarker for sustained glycemic control Non-enzymatic glycation of hemoglobin Normal level % t 1/ days (cf. RBCs) Normal level < 11 mu/l t 1/2 4 6 min Lesko and Atkinson; Ann Rev Pharmacol Toxicol. (2001) 41:

22 Pharmacometric (Drug-Centric) Models Study in 624 Type 2 Diabetes Mellitus patients evaluating the long-term effect of modified release gliclazide on fasting plasma glucose (FPG) levels FPG t FPG 0 t f T α is a hybrid constant that does not provide any information about the underlying physiological parameters and may also change over time Modified from: Frey et al., Br J Clin Pharmacol (2003) 55:

23 Application of Pharmacometric Models Pharmacometric models are currently used to: 1) Quantify treatment response (Δ from placebo) 2) Support dose selection 3) Inform clinical trial design However, they face limitations with characterizing: 1) Complex, multilevel (disease) processes 2) The impact of the patient s disease status on treatment response Need for more mechanistic modeling approaches to explain the dynamic interaction between drug, biological system and underlying disease processes 23

24 Mechanism-Centric Models Homeostatic feedback Drug Target exposure Target binding & activation Transduction EFFECT Physiologicallybased pharmacokinetic modeling Receptor theory Dynamical systems analysis PK PD Modified from: Danhof et al.; Annu Rev Pharmacol Toxicol (2007) 47:

25 Link to Biomarkers Modified from: Danhof et al.; Pharm Res. (2005) 22:

26 Identification of Clinically Relevant Covariate Relationships Drug Dose EFFECT PK PD Covariates Modified from: Danhof et al.; Pharm Res. (2005) 22: Modified from: Swiss Med Wkly. (2012)142:w

27 Physiological Response to Food Intake GLP-1 Glucagon-like peptide 1 Stimulates insulin Inhibits glucagon, somatostatin, gastric emptying and food-intake Serum conc. 6 ng/ml t 1/2 2 min GIP Gastric inhibitory polypeptide Stimulates insulin and fat storage Inhibits somatostatin Serum conc. 16 ng/ml t 1/2 7 min GI-Tract 27

28 Integrated Incretin-Glucose-Insulin Model Jauslin-Stetina et al., J Clin Pharmacol (2011) 51:

29 Application of Mechanism-Centric Models Mechanism-centric models are currently used to: 1) Characterize the dynamic interaction between drug, biological system and disease at multiple (biomarker) levels 2) Evaluate the effect of combination therapy 3) Distinguish between treatment effects (given an appropriate study design) However, they face limitations with characterizing: 1) Multiple (disease) pathways contributing to the clinical condition 2) T2DM patients are treated as a homogeneous patient population 3) Primarily focused on evaluation of efficacy Network-centric models may be needed to sufficiently characterize on- as well as off-target effects 29

30 Systems Pharmacology (Network-Centric) Models Modified from: Kohl et al., Clin Pharmacol Ther. (2010) 88:

31 31

32 Application of Network-Centric Models Network-centric models are currently used to: 1) Characterize the pertinent physiology that comprise the key pathways or targets of interest 2) Quantitatively integrate relevant biology across systems 3) Explore the impact of (novel) therapeutic interventions on the system However, they face limitations with characterizing: 1) Clinical data due to the inherent complexity of the model 2) Parameter values obtained from the literature for informing these models can be highly variable between settings 3) Link to long-term clinical outcome is frequently missing Simplified version of these network-centric models may have to be developed that conserve key dynamic properties 32

33 Wait a Second! Aren t physiologically-based pharmacokinetic (PBPK) models using information on metabolic and transporter networks? Yes. So could one call them network-centric (Systems Pharmacology) models? Yes. They provide the pharmacokinetic front-end to systems pharmacology models. 33

34 Physiologically-Based Pharmacokinetic Models Intrinsic/extrinsic Factors PBPK Model components System component (drug-independent) Drug-dependent component Lung Rapidly perfused organs ADME, PK, PD and MOA Blood Slowly perfused organs Kidney Liver Blood Metabolism Active transport Passive diffusion Protein binding Drug-drug interactions Receptor binding Huang and Temple, 2008 Individual or combined effects on human physiology Dosing Intestines Elimination PBPK Model Zhao et al. (2011) Clin Pharmacol Ther 89: Predict, Learn, Confirm, Apply 34

35 How is PBPK Being Utilized by Sponsors? Huang et al, J Pharm Sci, 2013 Increased use of PBPK by drug developers Pan, ASCPT Annual Meeting, 2014, Atlanta, GA Majority of the cases were related to drug-drug interactions (~ 60%); pediatrics ranks the second 35

36 Extension to a PBPK/PD Model for Diabetes Mellitus Schaller et al. (2013) CPT Pharmacometrics Syst Pharmacol. Aug 14;2:e65. doi: /psp

37 Glucagon [pmol/l] Predicted Concentration [mmol/l] Insulin [mu/l] Glucose [mmol/l] Fitted Concentration [mmol/l] Resimulation of Clinical Trial Selected Individual Fit Subject Model Data Meal Meal Meal Prediction Subject Meal Meal Meal E D C B B D A C E E C B Time [h] D B D 2 A C E Reference Concentration [mmol/l] Time [h] Reference Concentration [mmol/l] 37 Schaller et al. (2013) CPT Pharmacometrics Syst Pharmacol. Aug 14;2:e65. doi: /psp

38 Model Applications: T1DM - Automatic Blood Glucose Control Initialization with patient data (physiological parameters, e.g. weight, height, gender) Blood glucose measurements taken frequently, stored and delivered to the controller The process works on two time scales: Short: online calculation of optimal insulin dose based on recent glucose measurements Long: offline model adaptation based on full measurement data history Schaller et al. (2015) IEEE Trans Biomed Eng Nov 2. [Epub ahead of print] 38

39 39