Physiologically-Based Pharmacokinetic Modelling in Regulatory Decision-making at the European Medicines Agency

Sue Cole, Expert Pharmacokinetics Assessor

Disclaimer

The views expressed in this presentation are those of

the speaker, and are not necessarily those of the

MHRA or EMA.

Overview

• Review of PBPK models in Regulatory

submissions.

• Biopharmaceutical applications

• Drug Interactions

• Special populations

• Comments on presentations of models in

submissions- the Guideline

• Evaluation of uncertainty in models

• Qualification of models

PBPK seen as a valuable tool:Important potential value for benefit:risk decisions

From EMA-EFPIA Modelling and Simulation Workshop, December 2011

PBPK provides a mechanistic basis to reduce/quantify uncertainty in

extrapolation and to identify “at risk” populations.

Extrapolation - always a component of

benefit:risk decisions and can be an

important contributor to uncertainty.

Examples: elderly, polypharmacy (DDI),

critically ill, obese patients, paediatric, ethnic

groups (pharmacogenetics), … .

4

ACAT or ADAM model

2 main sections of model:

intestinal lumen (unreleased, undissolved

or dissolved drug)

enterocyte (absorbed drug)

Each section is divided up into

compartments:

stomach

small intestine (duodenum, jejenum, ileum)

large intestine (caecum, colon)

ACAT model: Advanced Compartmental Absorption and Transit model

PBPK to understand absorption

Generally lack of mechanistic understanding of

absorption in new drug applications

- link to drug properties

Example- Conditional approval.

• ADME study in progress, uncertainties around food effect

and linearity- possible precipitation in gut.

• Delay in drug absorption and high variability.

• Applicant used a zero-order followed by a first-order

absorption rate constant for absorption.

• Mechanistic approach has been suggested- e.g. PBPK

Cole APS 2017

Drug Disposition diagram

7

PBPK Model

System model: physiology

Drug specific parametersADME, PK, PD

PBPK for extrapolation of PK across

populations

Zhao et al, Clin Pharmacol Ther 89:259-267 (2011)

PBPK models in Regulatory

submissions

2

0

5

10

15

20

25

30

Number of submissions containing a PBPK model

Year of starting of the procedure

Purpose of PBPK models

submitted to EMA

Main categories Specific purpose Number

Intrinsic factors General description of PK parameters 8

Organ impairment 8

Differences across groups (ethnicity, disease states,

age groups)

5

Effect of polymorphisms 7

Extrinsic factors

(interactions)

DDI involving enzymes drug as victim 37

drug as perpetrator 23

DDI involving transporters drug as victim 3

drug as perpetrator 8

DDI based on pH changes 2

Food-drug interactions 2

Interaction with cigarette smoke 1

Drug parameters Comparison between strengths/formulations 8

Luzon et al CPT 2016

Purpose of PBPK in Regulatory

submissions

– In about 75% of procedures where a PBPK model is

suggested/ submitted, at least one of the purposes relates

to DDI (as victim or as perpetrator), specially for CYP3A4

mediated interactions

– Other purposes include

• Better understanding PK, role of enzymes/transporters…

• Dose recommendations

• Food effect

• Effect of polymorphisms / ethnic differences

• PK in special population (renal/hepatic impairment)

– Many cases in scientific advice- increasing in paediatric and

biopharmaceutical applications

One model- multiple applications

Drug

interactions as

a victim

Drug

interactions as

a perpetrator

Racial

differences

Hepatic

impairment

Renal

impairment

Reduced

cardiac outputFood effect

Biopharmaceutical applications

Application Conclusion

Variation Describe pharmacokinetics

following alternative dose route.

e.g intranasal or sub-lingual

Under consideration.

Could replace DDI

studies

Generics Replace an in vivo fed study Not accepted- in vivo

results did not show BE

Generics IVIVC Under review

Generics Justify lack of need for a study at a

higher dose

In house

Generics Importance of micronisation for high

solubility.

In house

Product Spec Support differences in product

specification.

SAWP Advice

DDI applications

14

Application Conclusion

Simulation of the effect of weak and moderate CYP 3A4 inhibitors

based on data for a strong inhibitor. Or worse case for inhibitor

Accepted

Waive an in vivo study for CYP inhibitor not meeting in vitro

criteria

Accepted

Simulations of DDI in poor metabolisers based on in vivo data in

extensive metabolisers with inhibition of 2 CYP pathways

Accepted

Simulation of time dependent inhibition Accepted

Simulation of induction Unlikely to be

accepted

Simulation of simultaneous induction and inhibition Unlikely to be

accepted

Waive an in vivo study for UGT inhibitor Under review

Waive an in vivo study for a transporter inhibitor Unlikely to be

accepted

Waive an in vivo study for inhibition of transport Unlikely to be

accepted

UGTs- Literature

• Complexities in assessing the

contribution of extra-hepatic

clearance mechanisms

• Empirical scaling factors required

• Blood data identifying drug

metabolism in kidney is limited

• Well stirred assumptions are not

appropriate for the kidney.

• The kidney is not homogenous

• Abundance data for the UGTs

and transporters in the various

regions of the kidney, are lacking.

Gill 2013 DMD 41 (4) 744- 53

UGTs Regulatory

• A number of examples with UGT as a major

clearance pathway.

• Also to support racial differences and effect of

polymorphisms

• At least 2 ongoing applications where effect of

UGT inhibition e.g by atazanavir, is predicted by

PBPK.

UGT Example

• Example Compound X

UGT1A1 contribution 61%

Atazanavir 400 mg QD prediction:

Cmax ratio of 1.04, AUC ratio of 1.11.

Qualification and uncertainty exploration requested

Response: The atazanavir PBPK model has been verified

with respect to UGT1A1 inhibition potency based on the

clinically observed DDI between atazanavir and UGT1A1

substrate raltegravir. The DDI between raltegravir and

atazanavir mainly occurs through inhibition of gut UGT1A1

The prediction of lack of interaction is mainly due to the

model assumption that X does not undergo gut extraction.

Transporters- FDA review

• Examples of interactions at intestinal Pgp

• Understanding the role of OATPs in DDIs

Pan 2016 JCP. 56, S123- 131

Models for transporters

• Tissue concentrations

important

• Difficult to validate

• Complex interplay between

transporters and enzymes

• Identifiability concerns

• Scaling factors

• Detailed models:

Simepravir

Simvastatin

Tsamandouras 2015 Pharm Res

32, 1864- 83

Snoeys 2016 CPT 99, 224- 34

Understanding interactions

with statins

• Compound Y inhibits OATP

• PBPK model to predict the impact of Y on statins

• Clinical data from drug-drug interaction (DDI) studies

with atorvastatin and simvastatin

• Does the rate of absorption of the statin affect the DDI

due to OATP inhibition by Y

• An interaction is expected when Y is taken

concomitantly with medicinal products that are OATP

substrates and have a similar tmax

• A maximal extent of 2-fold increase of Cmax is expected,

AUC <1.5-fold.

• No meaningful impact of Y on exposure of OATP

substrates is predicted if they are absorbed slowly.

Special Populations

• Knowledge gaps for renal impairment:

• Physiology data

• Mechanistic in vitro renal transporter parameters

• Understanding of mechanism causing changes in

transporter mediated secretion

• Comparable effects on systemic exposure but dynamics

in proximal tubular cells are different, depend on model

assumptions

• Similar for hepatic impairment

• Obeticholic acid- dose adjustment- cautious approach

Galetin 2017 J.Pharm Sci

Special Populations- uses

Application Conclusion

Paediatrics To support design of clinical studies With POPPK- in some

cases extrapolate efficacy

Paediatrics New Formulation Support adult

bioequivalence

Race Racial differences in enzyme or

transporter polymorphisms e.g.

UGTs

Reduced clinical study

Obesity Is a dose adjustment required? Under review

Renal

Impairment

Severe based on mild and

moderate or CYP 3A4 inhibition

Under review

Hepatic

Impairment

OATP substrate Under review

Disease state To support design of clinical studies EMA Qualification advice

PBPK modelling process-

Paediatrics

Paediatric extrapolation-

example

• Monoclonal Antibody

• High unmet medical need

• Very limited data <2 years old

• Limited literature data for similar molecules in this age

group, uncertainties in model, one example of a maturation

function, changes in endogenous IgG levels

• Extrapolation accepted to birth based on PopPK and PBPK

models and literature data for other similar molecules

• Used models to investigate worse case scenarios

Model evaluation in the EMA

25

European Commission

EMA

COMP HMPCCHMP CVMP PDCO CAT

PKWP

MSWG

SAWP

BSWP

PRAC

EWG

Reflections on PBPK models in

submissions.

• Encourage- benefits of mechanistic understanding.

• Lack of consistency in reporting. Generally:

➢ Lack of quantitative assessment of precision

➢ Variability not adequately captured

➢ Lack of uncertainty

exploration

➢ Lack of qualification for

intended use

What do we generally ask for?

• To confirm that the drug model is valid by

comparing with more clinical pharmacokinetic

studies (different doses, repeated dosing etc)

• Sensitivity analysis for key parameters

• What are the assumptions? Impact?

• Software qualification for the intended use

Evaluation of of models

Precision

Bias

Uncertainty

Variability

Assumptions

Covariance

Qualification

Verification of

inputs

Sensitivity

analysis

Identifiability

Predictability of the model

Prediction

PBPK Guideline

Guideline on the qualification and reporting of

Physiologically Based Pharmacokinetic (PBPK)

Modelling and Simulation

• Draft was released in July 2016

• Public consultation was due 31 Jan 2017

• NB! In Europe draft Guidelines are not

in force.

Some aspects from the guideline

Report should include:

• Purpose of the simulation including regulatory use

• Justification of system parameters, incl. library files,

physiological parameters of population

• Justification of drug specific parameters- predictability

• Mechanistic description of the system

• Justification of assumptions made and impact on results

• Qualification of the system i.e. the predictive performance

of the system for the particular purpose /intended use

Definitions

Qualification: The process of establishing confidence in a PBPK

platform to simulate a certain scenario, in a specific context, on

the basis of scientific principles, and ability to predict a large

dataset of independent data thereby showing the platforms ability

to predict a certain purpose. In the context of PBPK models,

qualification is purpose and platform version specific.

Verification: The model verification is a focused on the

correctness of the mathematical model structure.

Predictive performance of drug model: The process of

establishing confidence in the drug model. The reliability is

assessed on the basis of how well important characteristics of the

drug model has been tested against in vivo pharmacokinetic data

and whether adequate sensitivity and uncertainty analyses have

been conducted to support the models ability to provide reliable

predictions.

Evaluation of PBPK modelling

PBPK Model

System modelAnatomy

Biology

Physiology

Pathophysiology

Patient/disease extrinsic factors

Drug specific parametersADME, PK, PD and MOAMetabolism

Active transport/Passive diffusion

Protein binding

Drug-drug interactions

Receptor binding

32

Assessing the plasma

concentration time profile

– Criteria for accuracy of prediction?

– When is there under/ over prediction?

• How much is acceptable?

• Refer to effect/safety–exposure response

– Consider all parameters Cmax, AUC and t1/2

– Important parameters can be application dependent

Capturing variability in the

prediction

• Populations in PBPK platforms allow simulation of a

large number of individuals

• Examples where additional variance applied to account

for uncertainties in Bioavailability and clearance

Uncertainties in models

Evaluation of uncertainty

Fit for purpose models

• Different levels of precision can be accepted depending of the

use of the model

• Different levels of characterisation of impact of biologic variability,

inclusion of relevant pathways and other alternate assumptions

can be accepted

• Should we have fit for purpose UQ methodologies for models of

varying complexity?

Address what-if scenarios - best and worst case - to inform risk

assessment and decision making

• Assessment of the consistency, robustness and distribution of the

source data

• Expert informed scenario analysis

• Quantify impact of input, parameter and assumption uncertainty

in the resulting predictions

Local Sensitivity analysis

• Determine impact of any uncertain parameters or impact of

changes in key parameters

• Ideally want to determine the important parameters and

correlations

Global sensitivity analysis

McNally 2011 Frontiers in Pharmacol

Galetin 2017 Pharm Res

• Important to factor covariance

• Important where physiology is

not well characterised or

complex inter-play between

enzymes and transporters?

• Computationally complex but

quicker methods are

available e.g. eFast

• Compare integrated

population PBPK modelling.

Qualification for the intended

use- What do we mean?

Qualification is related to the PBPK platform

• Is there enough scientific support for a certain use

of the model?

DDI

• Enzyme inhibition

• Induction

• Transporter

IVIVC

Formulation changes

Biowaivers

Extrapolation of PK data in

young childrenFood effects

Prediction of PK in Special

populations

Slide curtesy of Anna Nordmark

Qualification is important for

high regulatory impact decisions

• High regulatory impact decisions

Examples:

» All changes to SmPC (ie label)

» Use of a PBPK model in place of clinical data (DDI, BE study)

» Non studied scenarios

» Extrapolation outside the studied area

• Medium regulatory impact decisions

» Such as paediatric dose setting that will be confirmed by a clinical study

The Qualification data set

• Qualification dataset should be pre-specified

• Selection criteria for the drugs and the in vitro and in vivo

parameters for these drugs should be described.

• The dataset should, if possible, cover a range of

pharmacokinetic characteristics, such as permeability,

extraction ratio, protein binding etc.

• Acceptable data sets have been presented for drug

interactions- literature publications

• CYP3A4 dataset- Fahmi DMD 2009: 48 in vivo clinical

studies used in prediction, 31 different inhibitors and

inducers. Midazolam was used as probe drug

Example Qualification for Time

Dependent Inhibition

Example -Qualification for TDI

• Simulation of 27 clinical trials was performed, 8

cases showed under prediction. However only 4

different inhibitors.

• The qualification was performed with an earlier

version of the software

• The risk of predicting a TDI as a false negative is

reduced by the sensitivity analyses performed on

the critical parameters for the new drug X

Biopharmaceutics and Special

Populations- under development

• Prediction of a drug’s oral absorption characteristics from

its formulation requires knowledge on the interplay among

physiology, the drug product, and the drug substance.

• Some GAPs

– GI physiology factors

– The confidence in IVIVC

– Interplay

• Lack of qualification

• Qualification also needed for Special populations

• Typical data set suggested- 10 compounds.

One model- multiple applications

Drug

interactions as

a victim

Drug

interactions as

a perpetrator

Racial

differences

Hepatic

impairment

Renal

impairment

Reduced

cardiac outputFood effect

Multiple applications- Multiple evaluations!

Conclusions

• Drug Interaction prediction have formed the majority of PBPK

models in submissions

• Examples of non-P450 and transporters in PBPK models

• Also applications in Biopharmaceutics and for exposure in

Special Populations

• Often one model- multiple uses

• Level of evaluation and reporting of models is variable

• Need to assess accuracy of prediction

• Need for uncertainty quantitation

• Need for qualification for the intended use

QUESTIONS?