Simulating individual variability in pharmacokinetics as a ... · Simulating individual variability...

Simulating individual variability in pharmacokinetics as a risk factor for drug toxicity

Iain Gardner Head of Translational DMPK Science

Simcyp Limited

New Perspectives in DMPK: Informing Drug Discovery, Royal Society of Chemistry, London

11 February 2014

© Copyright 2014 Certara, L.P. All rights reserved.

Outline of the presentation

• Integrating population variability into PK simulations

– creation of virtual populations

• Sources of inter-individual variability in Pharmacokinetics – Drug Clearance

• Using virtual populations to predict risk factors for drug toxicity – Cardiac Safety


The Challenge of Population Variability

Environment Disease

Genetics (possibly 2D6 poor

metaboliser!?!)

Cross Section of Patients in the Royal Hallamshire Hospital


CLuint per Liver

CLuint per g Liver

In vitro system

In vitro CLuint

Prediction of human PK (PD) in virtual individuals

Scaling Factor

(MPGGL, HPGL)

Liver weight

Simcyp approach Combine in vitro-in vivo extrapolation (IVIVE) and PBPK approaches in virtual individuals to predict drug concentration and effect

Identifying relevant DISTRIBUTION of values for demographical, biological, physiological and genetic parameters in target population & the COVARIATIONS between the parameters in target POPULATION


Mechanistic IVIVE & PBPK

Systems Data

Drug Data Trial

Design

Population PK (/PD) &

Covariates of ADME &

Study Design

Separating Systems & Drug Information


Separating Systems & Drug Information

Systems Data

Drug Data

Trial Design

Age Weight

Tissue Volumes Tissue Composition

Cardiac Output Tissue Blood Flows

[Plasma Protein]

MW LogP pKa

Protein binding BP ratio In vitro

Metabolism Permeability

Transport Solubility

Dose Route

Frequency Co-administered drugs

Populations studied

Mechanistic IVIVE approach to predict CL Whole body PBPK model

Prediction of drug PK (PD) in population of interest


Genotypes (Population Distribution)

Ethnicity Disease Body Fat Serum

Creatinine

Sex (Population Distribution)

Age (Population Distribution)

Height Brain Volume Body Surface

Area Heart Volme

Weight

Liver Volume

Cardiac Output

Cardiac Index

MPPGL HPGL

Enzyme Abundance

Liver Weight

Intrinsic Clearance

Renal Function

Building virtual populations


Freq

uenc

y

Age

HV Disease

Defined by real data

Demographic Features of Healthy and Disease Populations


0

1500 3000 4500 6000 Addicts

0

200

400

600

800 CVD

Age Distribution in Target Population

MALE FEMALE

Freq

uenc

y

Age Category


CLEARANCE: In Vitro – In Vivo

Extrapolation


? CLuint per

Liver

X

CLuint per g Liver

In vitro system

HLM

In vitro CLuint

µL.min-1

mg protein

HHEP µL.min-1

106 cells

Scaling Factors in Human IVIVE

Scaling Factor 1

Scaling Factor 2

X HPGL

MPPGL X

Liver Weight

?

rhCYP µL.min-1

pmol CYP MPPGL X X CYP

abundance


Sources of Variability: MPPGL and Donor Age

Barter et al. 2007 Curr Drug Met Barter et al. 2008 Drug Met Disposition

0 20 40 60 80 0

50

100

150

Age (years)

MP

PGL

(mg.

g-1 )

Paediatric Adult

40 mg.g-1

29 mg.g-1

0 10 20 30 40 50

1 25 45 65 Age (years)

MP

PG

L (m

g.g-

1 )


CYP Abundance Variability

Different Individuals

- Genetics (Ethnicity)

PLUS

-Age (Ontogeny) - Environment (Ethnicity)

- Sex / Co-medications 0

100

200

300

400

500

0

30

60

90 pm

ol C

YP/m

g m

icro

som

al p

rt

P450 Relative Abundance(average of 167 individuals with CYP3A5 *1 allele within a

1000 randomly selected Caucasian Population)

0.2%3.2%1.6%

15.0%

11.9%

33.1%

12.7% 11.0%4.1%

2.5%4.7%

CYP1A2CYP2A6CYP2B6CYP2C8CYP2C9CYP2C18CYP2C19CYP2D6CYP2E1CYP3A4CYP3A5


0.01

0.1

1

10

100

1000

10000

0.01 0.1 1 10 100 1000 10000 Observed CL (L/h)

Pred

icte

d C

L (L

/h)

s-wrf

caf

zlp

cyc

mdz sld

tlt clz

s-mph

dex omp

chlr

trz

apz

tlb

buf

theo phen

met

dic

sim

PREDICTION OF CLEARANCE (Oral)

Central tendency

&

Measure of variability

Howgate et al (2006)


Drug Distribution

Tissue Volumes

Predict Drug Distibution (Pt:p)

and Vss

Tissue Composition (VEW, VIW, VNL, VNP)

Fu and Fut

LogP, Log Dvo:w to predict K:P

Mechanistic IVIVE & PBPK

Drug Data

Trial Design

System Data


Example: Midazolam pharmacokinetics (simulations in 10 trials of 10 individuals)

Can describe Midazolam Pharmacokinetics using in vitro metabolism data together with systems physiology data


Midazolam Tissue concentrations (mean and 5 and 95 percentiles)

Concentrations in the tissues can also be linked to Pharmacodynamic or Toxicological effects


Cardiotoxicity

• Cardiac side effects major cause of drug withdrawal (regardless of the development level – from pre-clinical up to the post-approval) – E.g. Torsade de points and terfenadine

• Various mechanisms and effects involved – pro-arrhythmia, cardiac cell toxicity

• Drug interactions – important element causing serious

adverse events – E.g. astemizole (CYP related)

• PK variability important for safety assessment – E.g. tolterodine (CYP 2D6 mediated metabolism – genetic variability)


Screening for cardiac safety

• Effects of new compounds on IKR/Herg extensively screened for in drug discovery/development – QSAR models – IKR binding – HERG inhibition – Purkinje fiber studies

• Cardiac safety also often investigated in vivo – Pre-clinical studies – Thorough QT study in humans (~$1000000)

• Can cardiotoxicity also be assessed using mechanistic in

silico models?

19


Links to PD: Assessment of Proarrhythmic Potency

J. of Cardiovasc. Trans. Res. (2012) Toxicology Mechanisms and Methods, 2012

cell metabolism

cell viability

contractility

electro-mechanical

window

mu

ltip

le d

rug

s

demography

physiology

genetics

virtual population

action potential (APD90)

pseudoECG (QTc)

HUMAN HEART LEFT VENTRICULAR

CELL MODEL

hERG in vitro

(measured)

hERG in silico (QSAR)

other ion currents in silico

(QSAR)

other ion currents in vitro

(measured)


(QSAR)


(measured)


(QSAR)


(measured)


• molecular structure –––> QTc prolongation/TdP

21

– mechanistic/physiological

O’Hara and Rudy PLoS computational Biology 7, 2011 Ten Tusscher et al. AmJPhys-HeartPhys 286, 2004

𝑑𝑉

𝑑𝑡=𝐼𝑖𝑜𝑛 + 𝐼𝑠𝑡𝑖𝑚

𝐶𝑚

HUMAN HEART LEFT VENTRICULAR

CELL MODEL

Structure of left ventricular cell model


Variability matters

• Clinical evidence

Polak 2013 DDT accepted

Group Parameter Influence on ECG Reference

Demography

age ↑ age - ↑ QT/QTc Pham 2002

gender Females have longer QT/QTc as compared to

males. James 2007

Anatomy/

physiology

plasma ions concentration

(K+, Ca2+)

↑ K+ - ↓ QT/QTc

↑ Ca2+ - ↓ QT/QTc

Etheridge 2003

Covis 2002

cardiomyocyte size

(volume, area) ↑ size - ↑ QT/QTc Pacifico 2003

heart wall thickness ↑ thickness - ↑ QT/QTc Jouven 2002

cells heterogeneity

across heart wall

M cells presence influence the T wave and ECG

in general. Antzelevitch 2010

heart rate ↑ RR - ↑ QT Harris 2003

Malik 2002

sex hormones ↑ testosterone - ↓ QT/QTc

↑ progesterone - ↓ QT/QTc

Pham 2002

Sedlak 2012

Genetics

common polymorphisms

(ion channels)

Usually slight modification of ECG, may act as

a genetic modifier (with different mutation

both protective effect and QTc prolongation

were observed).

Crotti 2005

mutations (ion channels) ↑ QT/QTc Etheridge 2003

McPate 2005


Building virtual populations – cardiac safety assessment


Example 1: Dolasetron – formulation dependent effect

24

DOLASETRON – CARDIOLOGICALLY ACTIVE Plasma concentration negligible after po dosing

iv formulations – withdrawn from the market due to the TdP reports po formulation – considered as safe


Example 1: Dolasetron – Simcyp PK prediction

po 2.4 mg/kg

Dempsey 1996

Dolasetron

Hydrodolasetron

Metabolite

Parent


Example 1: Dolasetron – CSS PD effect prediction

iv 2.4 mg/kg

po 2.4 mg/kg

Dempsey 1996

-150

-100

-50

0

50

100

150

200

0 500 1000 1500 2000 2500 3000

Max APD90 ~ 370 ms

-150

-100

-50

0

50

100

150

200

0 500 1000 1500 2000 2500 3000

Max APD90 ~ 300 ms


Ranolazine

• IKr inhibitor (in vitro)

• multiple other ionic currents inhibition (in vitro)

• pharmacologically/electrically active metabolites

• clinically – large variability

Example 2 – Ranolazine: IVIVE at the population level

CVT-2738

//upload.wikimedia.org/wikipedia/commons/3/36/Ranolazine.svg


Ranolazine and CVT-2738 plasma concentrations

0

500

1000

1500

2000

2500

0 12 24 36 48 60 72 84 96

Syst

emic

Con

cent

ratio

n (n

g/m

L)

Time (h) 0

100

200

300

400

500

600

700

0 12 24 36 48 60 72 84 96

Syst

emic

Con

cent

ratio

n (n

g/m

L)

Time (h)

Ranolazine CVT-2738

Reasonable description of the plasma concentrations after multiple dosing


Ionic

current

Ranolazine

[IC50]/n Source

CVT-2738

[IC50] Source

IKr 12/1 Measured 1.19

QSAR predicted

IKs 1900/1 Measured -

ICa 311/1 Measured 26.38 QSAR predicted

INa peak 428/1.63 Measured - -

INa late 6.86/0.71 Measured

Electrophysiology inputs to the model


PRO-ARRHYTHMIC POTENCY - IVIVE at the population level - RESULTS

OBSERVED PREDICTED

y = 0.0068x + 4.779

-40

-20

0

20

40

60

80

100

0 500 1000 1500 2000 2500 3000

Ch

ang

e in

QT

c fr

om

bas

eli

ne

(m

sce

)

Ranolazine concentration (ng/ml)


Summary

• Using extrapolated in vitro data coupled with PBPK models it is possible to simulate the pharmacokinetics of many drugs

• By incorporating known physiological co-variates it is possible to make simulations in virtual populations rather than an “average” individual

• Concentrations of drugs in tissue compartments of the PBPK model can be linked to mechanistic models to predict side effects/toxicity

• Physiological variability can also be included within the toxicity models

• Acknowledgements – Sebastian Polak, Amin Rostami

Simulating individual variability in pharmacokinetics as a ... · Simulating individual variability...

Documents

Transcript of Simulating individual variability in pharmacokinetics as a ... · Simulating individual variability...