PK and PD as predictors of clinical effect

PK and PD as predictors of clinical PK and PD as predictors of clinical effecteffect

2

PKPD workshop at AGAH/Club Phase IPKPD workshop at AGAH/Club Phase I

•drug action– the interaction of the drug molecule at the binding site

e.g. receptor, carrier, channel

•drug effect– a measurable consequence of drug binding or drug action

e.g. EEC change, QT prolongation

•drug response– a desirable or undesirable clinical outcome

e.g. reduced frequency of seizures, reduction in blood pressure

3

How do we find and test new drugs?


clinical

○ dose ranging (empirical)

○ PKPD modeling

○ optimal study design

preclinical

○ screening (empirical)

○ molecular modeling

○ molecular design

screening learning design

4

…modeling may be understood as mechanized intuition

applying the rules of

– biology

– logic

– mathematics

– statistics


5


to learn from experience

to develop models based on data………what data do we need?

Preclinical:

◦ affinities of active drug molecules for the binding site (in vitro, in situ, in vivo)

◦ mechanism between binding and measurable effect including auto-regulation (feedback, synthesis)

◦ in vivo: dose(time) – concentration(time) – measurable effect(time)

6


to learn from experience

to develop models based on data………what data do we need?

Clinical:

◦ ideally everything measured in the preclinical program (in vivo affinities will be difficult to obtain), but at least the

following:

dose(time)

concentration(time)

effect(time)

◦ in addition, drug response data as a function of time

7


How can this strategy be incorporated into R&D planning?

◦ Every preclinical experiment and every clinical study is designed to add data to the PKPD knowledge base.

◦ The modeling and simulation (M&S) scientist participates in the project teams.

◦ For the M&S scientist there exists no boundary between preclinical and clinical development.

The design route will prove to be faster than the “shortcut”.

8


…from the work of EMF-Consulting: Example1

Selection of optimal doses for a new anti-epileptic drug

to be tested in patients.M. Marchand1, O. Petricoul1, E. Fuseau1, D. Bentley2, D. Critchley2

1 EMF consulting, BP 2, 13545 Aix en Provence, France2 EISAI Global Clinical Development, 3 Shortlands, London W6 8EE, UK

○ Rufinamide modulates the activity of sodium channels thus suppressing seizures induced by electroshock (maximal electroshock, MES) or by injection of pentylenetrazole (PTZ) in mice (PD). In clinical studies, rufinamide significantly reduced seizure frequency (PD).

○ Drug X is a new chemical entity with a novel mechanism of action. It shows anticonvulsant effects in rodents. The dose(time)-concentration(time) relationship (PK) was studied in epileptic patients.

9

PKPD workshop: Example 1PKPD workshop: Example 1

○ PKPD modeling in mice:

− Population PKPD modeling used NONMEM. − A one-compartment model with first order

elimination was chosen for both rufinamide and Drug X.

− For PKPD modelling, drug concentrations were predicted in male mice according to weight, the administered dose in mg/kg, population PK parameters (previously estimated), and time of test.

10


PKPD modeling in mice – where are the problems?+ mice are cheap+ mice are genetically well defined+ small interindividual variability

− mice are small− difficult to dose accurately− difficult to obtain more than one blood sample

11


PKPD modeling in mice – Population approach:

Population PK model based on toxicokinetic data

-free choice oral dosing (continuous input during dark hours)

-blood sampling at steady state

-destructive, only one sample per mouse

Cl/F =DR

Css

-oral dosing by gavage (controlled time of drug input)

-blood sampling after single or multiple doses

-destructive, only one sample per mouse

C(t)=D·ka

V/F·(ka – k) · (e-k·t

− e-ka·t

)

12


PKPD modeling in mice – Population approach:

Population PD model using predicted individual concentrations at the time of the effect measurement

Individual concentrations are predicted based on:

-the dose given at the PD experiment

-the gender and weight of the mouse

-the time of the effect measurement after the dose

-the population PK model

13


50

max

CCONC

ECONCDV

Emax = 100% (FIXED)C50 = 1.35 g/mL = 5.98 Rufinamide concentrations (g/mL)

0.0 0.5 1.0 1.5 2.0 2.5 3.0

% o

f m

ice

pro

tect

ed

fro

m t

on

ic h

ind

limb

se

izu

re

0

20

40

60

80

100

ObservationsPredictions

Rufinamide data: Observed and predicted % of protected mice from MES Test

14


50

max

CCONC

ECONCDV

Emax = 76.4%

C50 = 1.64 g/mL

Rufinamide data: observed and predicted % of protected mice from PTZ Test

Rufinamide concentrations (g/mL)

0 2 4 6 8 10 12 14 16 18 20

% o

f m

ice

pro

tect

ed

fro

m c

lon

ic s

eizu

re

0

20

40

60

80

100


15


Drug X data: observed and predicted % of protected mice from MES Test

50

max

CCONC

ECONCDV

Emax = 100% (FIXED)C50 = 141.6 ng/mL = 4.56

Drug X concentrations (ng/mL)

0 50 100 150 200 250

% o

f m

ice

pro

tect

ed

fro

m t

on

ic h

ind

limb

se

izu

re

0

20

40

60

80

100


16


50

max

CCONC

ECONCDV

Emax = 100% (FIXED)C50 = 88.1 ng/mL


0 50 100 150 200 250 300

% o

f m

ice

pro

tect

ed

fro

m c

lon

ic s

eizu

re

0

20

40

60

80

100


Drug X data: observed and predicted % of protected mice from PTZ Test

17


PKPD modeling in patients:

Cavss rufinamide concentrations (g/mL)

0 20 40 60 80 100

Pre

dict

ed to

tal s

eizu

re f

requ

ency

per

28

days

0

2

4

6

8

10

12

avsse CfrequencyseizuretotalLog 0187.0893.0)(

Rufinamide: PD model based on clinical data

18


Link from mice to humans:assumes that the effective concentrations in mice are also effective in humans

A generic mathematical link function (Weibull) was used to relate rufinamide preclinical effects to its clinical response (Loge of total seizure frequency over a period of 28 days).

19

PKPD workshop: Example 1PKPD workshop: Example 1Rufinamide preclinical effect (MES test): the link function shows that effective concentrations in the preclinical MES test are not effective clinically.Is the approach wrong? Not necessarily, but MES is definitely not a suitable preclinical test.

Predicted % of protected mice from tonic hind limb seizure

0 20 40 60 80 100

Pre

dict

ed L

oge

(tot

al s

eizu

re f

requ

ency

per

28

days

)

-2.8

-2.6

-2.4

-2.2

-2.0

-1.8

-1.6

-1.4

-1.2

-1.0

-0.8

an ideal relationship:

50% protected mice are related to half the maximal reduction in seizure frequency in patients.

20


Rufinamide preclinical effect (PTZ test): the link function shows that concentrations which protect more than 50% of mice also reduce total seizure frequency per 28 days in patients. The relationship is not ideal but sensitive enough to be used for the following extrapolation (next slide).

Predicted % of protected mice from clonic seizure

0 20 40 60 80 100

Pre

dic

ted

Lo

g e (t

ota

l se

izu

re f

requ

en

cy p

er

28

days

)

-2.8

-2.6

-2.4

-2.2

-2.0

-1.8

-1.6

-1.4

-1.2

-1.0

-0.8

Weibull Function

21


○ Extrapolation from known to unknown:

− assuming that the link between the preclinical and the clinical test is generally valid and independent of the pharmacologic agent used to cause the response,

− the PTZ preclinical effect measurements of Drug X are used to predict total seizure frequency per 28 days (clinical response).

22


The link function is now applied to drug X: whatever drug X concentration is related to 70% of mice protected from seizures (PTZ test) is expected to be related to a clinical response of 28·e-1.3 =7.6 seizures in 28 days, a minimal response.

Predicted % mice protected from clonic seizure

0 20 40 60 80 100

Pre

dic

ted L

og e

(to

tal s

eiz

ure

fre

qu

en

cy p

er

28

days

)

-6

-5

-4

-3

-2

-1

0

Answer: ~200ng/ml

23


○ Extrapolation from response to concentration:

− the preclinical PD model for PTZ test of Drug X is used to predict the concentration in humans necessary to achieve a certain clinical response.

− knowing… that for rufinamide the link function relates effective concentrations in mice to effective concentrations in humans.

− assuming… that the link function has general applicability and is thus also valid for drug X.

24


Drug X: Finding the necessary concentrations to achieve a certain total seizure frequency per 28 days (response)


0 50 100 150 200 250 300 350

Pre

dic

ted

to

tal s

eiz

ure

fre

que

ncy

pe

r 2

8 d

ays

0

2

4

6

8

10

12

25


○ Extrapolation from concentration to dose:

− a PK model for Drug X established in epileptic patients in a phase IIa pilot study is used to predict the dosing regimen to produce the necessary concentrations.

− this PK model takes the drug interaction with CYP3A4 inducers into account. The recommended dosage is stratified accordingly:

26


In order to achieve similar decrease (2.8 per 28 days) of total seizure frequency as with a typical Cavss (15 μg/ml) of rufinamide, the following daily doses for Drug X are likely to produce a Cavss of 215 ng/mL:

○ Sub-population 1, without co-administration of CYP3A4 inducers: 1.8 units ○ Sub-population 1, with co-administration of CYP3A4 inducers: 7.7 units○ Sub-population 2, without co-administration of CYP3A4

inducers: 4 units○ Sub-population 2, with co-administration of CYP3A4 inducers: 15 units

Note: a Cavss of 215 ng/mL was observed in healthy subjects following repeated daily doses of 4 units which were well tolerated.

27


workshop…english

atelier……..français

Werkstatt…deutsch

This is not a place to shop for work but a place to work.

Before I go on to my second example I would like to solicit contributions, comments, anecdotes… from the attendees.

28


…from the work of EMF-Consulting: Example 2

Selection of an optimal biomarker for neutral endopeptidase (NEP) inhibitors in humans.

A.C. Heatherington, S. Sultana, R. Hidi, M. Boucher, E. Fuseau, M. Marchand, P. Ellis, S.W. Martin Pfizer Ltd, Sandwich, UK; EMF Consulting, Aix-en-Provence, France

Objectives:○ to select a reliable soluble biomarker for NEP inhibitors○ to compare clinical PD models to in vitro PD models○ to build a suitable PKPD model to optimally design future

clinical studies

29


Background:Neutral endopeptidase (NEP) is a metallopeptidase enzyme involved in the degradation of a number of endogenous peptides, including

○ vasoactive intestinal peptide (VIP)○ substance P○ endothelins (hydrolysis of big endothelin, Big ET-1, to endothelin)○ atrial natriuretic peptide (ANP).

It is hypothesized that NEP inhibitors would increase VIP leading to enhanced vasodilatation in genital tissues. Two molecules, UK-447,841 (in vitro IC50 10 nM) and UK-505,749 (in vitro IC50 1.1nM), have undergone pharmacological evaluation to assess their effect on plasma Big ET-1 and ANP levels.

30


Studies: Double-blind, randomized, placebo-controlled phase 1 studies in healthy volunteers

UK-447,841 UK-505,749 Design Cross-over Parallel group Cross-over Dosing Single escalating oral

doses, 3 to 800 mg

Multiple daily oral doses, 100, 400

and 800 mg

Single escalating oral doses,

0.1 to 540 mg PK data 14 samples up to 48 h 13 + 15 up to 48h 14 samples up to 48h PD data

(big ET-1 and ANP) 3 samples up to 8h

7 samples up to 12h

7 samples up to 12h

31


Time (h)

0 10 20 30 40 50

UK

-44

7,8

41 m

ed

ian c

once

ntr

atio

n

(mg/L

)

0.0001

0.001

0.01

0.1

1

10

1003 mg10 mg30 mg100 mg200 mg400 mg800 mg

UK-447,841

median PK data, Phase I, healthy volunteers

32


Time (h)0 10 20 30 40 50

UK

-505,7

49 m

edia

n c

once

ntr

atio

n

(mg/L

)

0.001

0.01

0.1

1

10

100 0.1 mg0.3 mg1 mg3 mg10 mg30 mg90 mg270 mg540 mg

UK-505,749

median PK data, Phase I, healthy volunteers

33


Combined PKPD population model (NONMEM) fortwo drugs (PK) and two biomarkers (PD indirect response model for Big ET-1 and ANP)

•the NEP inhibitors slow down the degradation (kout1,2) of Big ET-1 and ANP •Big ET-1 stimulates production rate (kin2) of ANP

•ANP stimulates production rate (kin1) of Big ET-1

•age enhances production rate (kin2) of ANP •Emax , the maximum decrease in kout , is the same for both drugs but different

for ANP (41%) and for Big ET-1 (66%) •age increases Emax for ANP•IC50, the drug concentration at half-maximal effect, if different for Big ET-1

and ANP and different for UK-505,749 and UK-447,841•also in vivo UK-505,749 is 10 times more potent than UK-447,841

34


kin1kout1

kin2kout2

Emax

IC50IC50

Emax 2

Pop PK model

Pop PD model

IC50 IC50

RCIC

CE

CIC

CEkk

pred

pred

pred

predoutin

)505()505(50

)505(max

)447()447(50

)447(max )()(1

Response compartment

η (IIV)

Cpred (447)

Cpred (505)

Age effect (+)

Age effect (+)

ANP

η (IIV)

η (IOV)

η (IIV)

η (IIV)

η (IIV)

η (IIV)BigET1

+

+

PD indirect response model for Big ET-1 and ANP

35


Time (h)

0 2 4 6 8 10 12

Media

n B

ig E

T-1

(pg/m

L)

0

2

4

6

8

10Doses 3 to 800 mg of UK-447,841Doses 0.1 to 540 mg of UK-505,749

36


Time (h)

0 2 4 6 8 10 12

Me

dia

n A

NP

(pg

/mL

)

20

40

60

80

100

120

140

Doses 3 to 400 mg of UK-447,841Doses 0.1 to 540 mg of UK-505,749

37


UK-505,749 predicted concentration in effect compartment (mg/L)

0 2 4 6 8 10 12 14

Big

En

do

the

lin (

pg/m

L)

0

2

4

6

8

10

12

observed Big ET1Population prediction of Big ET1Individual prediction of Big ET1

38


after UK447,841 treatment

Time (h)

0 5 10 15 20

Big

E c

on

ce

ntr

ati

on

s (

pg

/mL

)

0

2

4

6

8

10

12

14

P5, P50 and P95 SimP5, P50 and P95 Obs

(3 to 800 mg)

39



Time (h)

0 2 4 6 8 10 12

Big

E c

on

ce

ntr

ati

on

s (

pg

/mL

)

0

2

4

6

8

10

12

14

P5, P50 and 95 SimP5, P50 and P95 Obs

(0.1 to 540 mg)

40



Time (h)

0 2 4 6 8 10 12

AN

P c

on

ce

ntr

ati

on

s (

pg

/mL

)

0

50

100

150

200

250


(3 to 400 mg)

41



Time (h)

0 2 4 6 8 10 12

AN

P c

on

ce

ntr

ati

on

s (

pg

/mL

)

0

50

100

150

200

250


(0.1 to 540 mg)

42


Conclusion

○ Big ET-1 plasma concentration and, to a lesser extent, ANP plasma concentration can be used as a pharmacological biomarker for the inhibitory drug effect on enzyme (NEP) activity in healthy volunteers.

○ Big ET-1 has ideal characteristics of a soluble biomarker: it demonstrates dose-concentration-effect, time-linearity, reproducibility of effect with similar Emax for two NEP inhibitors.

○ The ratio of the in vivo IC50 of the 2 compounds is similar to the in vitro ratio. This allows extrapolation between species and between different drugs.

43


I hope to meet many of you again at the PAGE meeting in June 2007 in Copenhagen!

PK and PD as predictors of clinical effect

Documents

Transcript of PK and PD as predictors of clinical effect