pk-pd modeling and dosage determination for proof-of ...€¦ · pk-pd modeling and dosage...

©2008 metrum research group LLC

PK-PD Modeling and DosageDetermination for Proof-of-Concept Trials

Marc R. Gastonguay, Ph.D.([email protected])

IMMPACT-VIII

Early Clinical Study Designs, Emphasizing Proof-of-Concept Trials

June 12-14, 2008

Arlington, VA

©2008 metrum research group LLC 2

PKPD in Proof of Concept Trials: IMMPACT 2008

Overview

- PK/PD (Exposure-Response) and Model-Based DrugDevelopment

- Role of Exposure-Response Modeling in Proof-of-ConceptTrials Planning and Design

Analysis and Quantitative Support for PoC Determination

Building Knowledge for Later Stage Development

Other examples of E-R utility

- Summary Points



Innovation: Planes are modeled long before takeoffNASA Aerospace Engineering Grid

•Lift Capabilities•Drag Capabilities•Responsiveness

•Deflection capabilities•Responsiveness

•Thrust performance•Reverse Thrust performance•Responsiveness•Fuel Consumption

•Braking performance•Steering capabilities•Traction•Dampening capabilities

Crew Capabilities- accuracy- perception- stamina- re-action times- SOP’s

Engine Models

Airframe Models

Wing Models

Landing Gear Models

Stabilizer Models

Human Models

Whole system simulations are produced by couplingall of the sub -system simulations

It takes a distributed virtual organization to design,simulate and build a complex system like an aircraft

http://grids.ucs.indiana.edu/ptliupages/presentations/cendiapril25-05.ppt



Pharmacometrics…the science ofinterpreting and describingpharmacology in a quantitative fashion(e.g. through modeling and simulation)

DOSE CONCENTRATION RESPONSE

PK PD

DIS

EA

SE

PR

OG

RE

SS

ION

TRIAL DESIGNS &DEVELOPMENT STRATEGY KNOWLEDGE



Modeling and Simulation: A Tool to Facilitatethe Learn-Confirm Continuum

Collect data

Build models to describe data andconfirm prior knowledge

Use M&S to learn from new data andexplore future outcomes

Make informed decisions

Perform new experiment (study)

Sheiner LB. Learning versus confirming in clinical drug development.Clin Pharmacol Ther 1997; 61(3):275-91.



M&S Throughout Drug Development

Simulation & Optimizationof Phase III Trial Designs

Simulation-Guided(Adaptive) Phase IIDesigns: EarlyProbability of Success &Dose-Response

Pop PK, E-R forLabeling andConfirmatoryEfficacy Support

Preclinical &Early Development:PK-PD, SystemsBiology M&S

Therapeutic AreaKnowledge: DiseaseProgression &Target ResponseProfile

ToxicologyPKBiomarkerE-R

HumanMTD, PKBiomarker,Tolerability, E-R

Biomarker, Efficacy,Tolerability E-RDose-ResponseCovariates, Pop PK-PD

Efficacy,Safety & DoseSpecialPopulations

NewFormulationsBridge to NewIndications

Proof of Concept:Probability-BasedDecision Rule



What to Learn?

05

1020

30

LPSres$AUC

LP

S (

min

)

390

410

430

450

TSTres$AUC

TS

T (

min

)

05

1020

30

NAW1res$AUC

NA

W

NAW 1

NAW 2

13

1020

NG2-73 AUC (ng*hr/mL)

Dose

(m

g)

0 50 100 150


02

46

810

WRres$AUC

WR

Im

media

te

02

46

810

WRres$AUC

WR

0.5

hr

02

46

810

WRres$AUC

WR

7 h

rs

13

1020


Dose

(m

g)

0 50 100 150


05

1020

30LPSres$AUC

LP

S (

min

)

390

410

430

450

TSTres$AUC

TS

T (

min

)

05

1020

30

NAW1res$AUC

NA

W

NAW 1

NAW 2

13

1020


Dose

(m

g)

0 50 100 150


02

46

810

WRres$AUC

WR

Im

media

te

02

46

810

WRres$AUC

WR

0.5

hr

02

46

810

WRres$AUC

WR

7 h

rs

13

1020


Dose

(m

g)

0 50 100 150

NG2-73 AUC (ng*hr/mL)EXPOSURE (e.g. AUC, Cmax, Css avg)

RES

PON

SE

Dos

e

Dos

e

Bio

mar

kers

Pro

b (A

E)

Tole

rabi

lity

Clin

ical

Out

com

es

PD

Effi

cacy

link



- Stanski.Model-BasedDrugDevelopment:ACriticalPathOpportunity,March18,2004

Regulatory Support for M&S

http://www.fda.gov/oc/initiatives/criticalpath/presentations.html



Regulatory Support for M&S:Guidance Documents

- Population Pharmacokinetics (FDA and EMEA)- Exposure-Response Relationships (FDA)- Dose-Response Information to Support Drug

Registration (ICH-E4)- General Considerations for the Clinical Evaluation of Drugs (FDA 77-

3040)- General Considerations for Pediatric Pharmacokinetic Studies (FDA)- Pharmacokinetics in Patients with Impaired Renal Function (FDA)- Pharmacokinetics in Patients With Impaired Hepatic Function (FDA)- Studies in Support of Special Populations:

Geriatrics (ICH-E7)- Ethnic Factors in the Acceptability of Foreign

Clinical Data (ICH-E5)- Clinical Investigation of Medicinal Products in the Pediatric Population

(ICH-E11)



Determination of PoC

- Primary Challenge: Define decision criteria for PoCdetermination Proof of mechanism

Statistically significant efficacy response with approval endpoint

Acceptable probability of achieving multivariate target responseprofile

Comparability to active control

- Once defined, probability of meeting PoC decision criteriafor different trial designs can be explored throughmodeling and simulation



Target Response Profile

05

1020

30

LPSres$AUC

LP

S (

min

)

390

410

430

450

TSTres$AUC

TS

T (

min

)

05

1020

30

NAW1res$AUC

NA

W

NAW 1

NAW 2

13

1020


Dose

(m

g)

0 50 100 150


02

46

810

WRres$AUC

WR

Im

media

te

02

46

810

WRres$AUC

WR

0.5

hr

02

46

810

WRres$AUC

WR

7 h

rs

13

1020


Dose

(m

g)

0 50 100 150


05

1020

30LPSres$AUC

LP

S (

min

)

390

410

430

450

TSTres$AUC

TS

T (

min

)

05

1020

30

NAW1res$AUC

NA

W

NAW 1

NAW 2

13

1020


Dose

(m

g)

0 50 100 150


02

46

810

WRres$AUC

WR

Im

media

te

02

46

810

WRres$AUC

WR

0.5

hr

02

46

810

WRres$AUCW

R 7

hrs

13

1020


Dose

(m

g)

0 50 100 150

NG2-73 AUC (ng*hr/mL)EXPOSURE (e.g. AUC, Cmax, Css avg)

RES

PON

SE

Dos

e

Dos

e

Bio

mar

kers

Pro

b (A

E)

Tole

rabi

lity

Clin

ical

Out

com

es

PD

Effi

cacy



PK and Exposure-Response M&SOpportunities in PoC

- PK Modeling Understand PK in target population and possibly reduce inter-

individual variability in exposure to increase signal/noise: dosingindividualization

Select PoC doses with minimal exposure overlap Explain unexpected outcomes (e.g. unknown phenotypic

differences in PK) Adjust for formulation differences

- E-R Modeling Assessment of E-R relationships for multiple endpoints (e.g. after

dose-ranging based on efficacy endpoint) Basis for trial simulations: explore performance/options in silico

before initiating clinical trial



Impact of E-R Varies with PoC Trial DesignsMTD-Type PoC Design

Typically 1 active treatment dose vs. referencetreatment

Dose selected based on Phase I MTD Standard pair-wise statistical comparison

Dose-Ranging PoC Design Multiple doses investigated Dose-range informed by preclinical data, Phase

I, biomarker, competitor data Model-based data analysis Often multi-variable PoC assessment



Exposure-Response in MTD-Type PoC:Proceed with Caution

- Single active treatment arm atMTD (300 mg) vs. Placebo

- Obtain PK in all individuals

- Explore resulting relationshipbetween exposure (Cavg) andResponse (1 observation perindividual)

- Can we make an accurateassessment of the PK-PDrelationship from this design?

Problem described in: Nedelman JR, Rubin DB, Sheiner LB. Diagnostics for confounding inPK/PD models for oxcarbazepine. Stat Med. 2007 Jan 30;26(2):290-308.



Exposure-Response in MTD-Type PoC- Consider possible inter-individual correlation between PKand PD



Exposure-Response in MTD-Type PoC

- Resulting exposure-response relationships are misleading



Exposure-Response in MTD-Type PoC- One solution: Obtain within-individual E-R (e.g. crossover)analyzed with mixed-effects modeling



Exposure-Response in MTD-Type PoC- Another solution: Population E-R with broad dose-range



PK-PD in Planning and Design of PoC Trials

- Use prior information, when available

Phase I PK, tolerability, biomarkers

Pre-clinical estimates of effective concentrations,relative potency

Competitor data

Therapeutic area knowledge



Toxicity E-R to Inform PoC Dose Selection

Data from SAD, MADin healthy volunteers



Probability of QTc Prolongation

• Explore probability of QTc – related toxicity atvarious doses from Phase I data

• Project QTc prolongation at expected Cmax, giventop dose and DDI

• Define dose-limit and early probability ofcompound viability



Modeling Biomarker Data: Phase I MD Study- PK-PD relationship evident & quantifiable (‘Emax’ model)

- Establish target PoM

- Set doses for investigation in PoC = concentrations within apparentefficacious range

Concentration

Mar

ker

ConcEC

ConcEE

++=

50

max

0

*Marker

0 1 2 3 4

7

6

5

4

3

2

1

0

O PlaceboO Dose 1O Dose 2O Dose 3--- Model Prediction



E-R Analysis of PoC Trials

- Example Parallel groups: 4 active doses + placebo + active control

(competing therapy)

Multiple Endpoints: biomarker 1 (efficacy), biomarker 2(undesired), clinical outcome 1

PoC determination based on model-based posterior probability ofreaching target response profile



E-R Based PoC: Test & Active ComparatorResponse: Biomarker 1 (efficacy)

Drug X (red),Comparator Y (blue)

Median Concentration

Media

n R

esponse

EC50 = dashed lines

Median observation at each collection time for each treatment (circles)

PK-PD Model Prediction (solid line)

01

23

45

- Drug X (red) was more potent than Comparator Y (blue)- Relative potencies (EC50 of X vs. Y) very consistent across multiple response variables

Plasma Drug Concentration

Bio

mar

ker 1

Res

pons

e

RXij = E0i + EMAXi*Ci/(EC50Xi + Ci) + eij

RYij = E0i + EMAXi*Ci/(EC50Yi + Ci) + eij

Target



- Identified Drug X concentrations associated with BM II effect- Consider doses that provide for target concentrations

PoC: E-R for Biomarker 2 (undesired)

Concentration rangeassociated with “no effect”

Concentration rangeassociated with “effect”

Concentration

Mar

ker I

I

Target



Res

pons

e I

Dos

e

Cmax (concentration)Drug X

E-R for Clinical Outcome I

Target



- Drug X posterior probability distribution for target response meetsPoC criterion, but which doses should go into Phase 2b, whereprimary endpoint will be an approval outcome measure?

- Comparator Y Dose-Response Literature data Model = Nonlinear ‘Emax’ model for mean relationship Uncertainty range: Based on standard errors of parameter estimates

- Scaled for Approximate Dose-Response of Drug X Based on biomarker relative EC50 of Drug X vs. Comparator Y Accounted for PK differences Additional variability for uncertainty in scaling ratios

Building Knowledge for Phase 2b



Dose Y

Response

Dose-Response Model for Comparator Y: 2b Response

0 1 2 3 4

0

1

2

3

4

Literature data (o)

Mean Prediction (___)

Uncertaintyrange: basedon 95% CI’s ofparameterestimates

Dose

Res

pons

e



- Select doses to further characterize (reduce uncertainty in) response surface

- Target doses ~ 50% (ED50), 80% (ED80) & max effects (Emax)

0

1

2

3

40 1 2 3 4

Scaled Dose-Response for Drug X:Predicted 2b Response

Dose

Res

pons

e



Other Examples of E-R in Analgesic PoC Trials

- Dissociation of rescue drug effects from test treatment

- Model-based inferences with dropout (missing data)



Dissociating Treatment Effects fromRescue Dose Effects

- Chronic pain PoC design (PBO plus 4 dose levels)

- Acetaminophen rescue (500 mg) allowed as needed

- Reduction in pain intensity is primary endpoint

- Problem: How to interpret pain response in presence of rescue?

- Proposal: Analyze entire data set with model-based analysis using 2simultaneous exposure-response relationships: Study Drug E-R

Rescue E-R



Consideration

- Potential delay between plasma exposure and exposure atsite of action (e.g.,CNS) May be more pronounced with

▶ with acute or ‘prn’ dosing▶ shorter t1/2 and/or rapid Tmax

Figure from: Shinoda S, Aoyama T, Aoyama Y, Tomioka S, MatsumotoY, Ohe Y 2007. Pharmacokinetics/pharmacodynamics of acetaminophenanalgesia in Japanese patients with chronic pain. Biol Pharm Bull30(1):157-161

Also see: Staahl C, Upton R, Foster DJ, Christrup LL, Kristensen K,Hansen SH, Arendt-Nielsen L, Drewes AM. Pharmacokinetic-pharmacodynamic modeling of morphine and oxycodone concentrationsand analgesic effect in a multimodal experimental pain model. J ClinPharmacol. 2008 May;48(5):619-31.



Dual E-R Model Schematic

Gut

StudyDrug dose

Central Peripheral

Effect

PD

Effe

ct

Gut

RescueDrug dose

Central

Effect

PD

Effe

ct

Total Observed PD Response



Individual Contributions to Total Response

- Integrated model ofStudy Drug and RescueE-R

- Allows interpretation ofindividual and jointeffects

- Success of thisapproach highlydependent on adequateDose-Ranging design

- Results preliminary:Evaluation ofperformance throughsimulation ongoing



Model-Based Inferences in the Presence ofDropout

- Acute pain PoC study

- Dropout after first rescue

- Population nonlinear-mixed effects exposure-response modeldeveloped from observed repeated-measures data (missing Atrandom assumption)

Approach first described in:

Sheiner LB. A new approach to the analysis of analgesic drug trials,illustrated with bromfenac data. Clin Pharmacol Ther. 1994Sep;56(3):309-22.



Observed Data

0.25 1 2 4 6 8 10 12

Observed Pain Intensity Placebo (n=34)

Time (hr)

NP

RS

Score

Fre

quency (

%)

020

4060

80100

Rescued

NPRS = 10NPRS = 9

NPRS = 8NPRS = 7

NPRS = 6

NPRS = 5

NPRS = 4NPRS = 3

NPRS = 2NPRS = 1

NPRS = 0

0.25 1 2 4 6 8 10 12

Predicted Pain Intensity Placebo (n=34)

Time (hr)

NP

RS

Score

Fre

quency (

%)

020

4060

80100

0.25 1 2 4 6 8 10 12

Extrapolated Pain Intensity Placebo (n=34)

Time (hr)

NP

RS

Score

Fre

quency (

%)

020

4060

80100

0.25 1 2 4 6 8 10 12

Observed Pain Intensity 60 mg Patients (n=65)

Time (hr)

NP

RS

Score

Fre

quency (

%)

020

4060

80100

Rescued

NPRS = 10NPRS = 9

NPRS = 8NPRS = 7

NPRS = 6

NPRS = 5

NPRS = 4NPRS = 3

NPRS = 2NPRS = 1

NPRS = 0

0.25 1 2 4 6 8 10 12

Predicted Pain Intensity 60 mg Patients (n=65)

Time (hr)

NP

RS

Score

Fre

quency (

%)

020

4060

80100

0.25 1 2 4 6 8 10 12

Extrapolated Pain Intensity 60 mg Patients (n=65)

Time (hr)

NP

RS

Score

Fre

quency (

%)

020

4060

80100

0.25 1 2 4 6 8 10 12


Time (hr)

NP

RS

Score

Fre

quency (

%)

020

4060

80100

Rescued

NPRS = 10NPRS = 9

NPRS = 8NPRS = 7

NPRS = 6

NPRS = 5

NPRS = 4NPRS = 3

NPRS = 2NPRS = 1

NPRS = 0

0.25 1 2 4 6 8 10 12


Time (hr)

NP

RS

Score

Fre

quency (

%)

020

4060

80100

0.25 1 2 4 6 8 10 12


Time (hr)

NP

RS

Score

Fre

quency (

%)

020

4060

80100

0.25 1 2 4 6 8 10 12


Time (hr)

NP

RS

Score

Fre

quency (

%)

020

4060

80100

Rescued

NPRS = 10NPRS = 9

NPRS = 8NPRS = 7

NPRS = 6

NPRS = 5

NPRS = 4NPRS = 3

NPRS = 2NPRS = 1

NPRS = 0

0.25 1 2 4 6 8 10 12


Time (hr)

NP

RS

Score

Fre

quency (

%)

020

4060

80100

0.25 1 2 4 6 8 10 12


Time (hr)

NP

RS

Score

Fre

quency (

%)

020

4060

80100

Placebo

Dose 1 Dose 2

Freq

uenc

y of

NP

RS

Sco

re



PD Response Time-Course

0 2 4 6 8 10 12

02

46

810

PI121 ID=2036

Time (hr)

Pain

Inte

nsity

0 2 4 6 8 10 12

02

46

810

PI121 ID=2037

Time (hr)

Pain

Inte

nsity

0 2 4 6 8 10 12

02

46

810

PI121 ID=2038

Time (hr)

Pain

Inte

nsity

0 2 4 6 8 10 12

02

46

810

PI121 ID=2039

Time (hr)

Pain

Inte

nsity

0 2 4 6 8 10 12

02

46

810

PI121 ID=2040

Time (hr)

Pain

Inte

nsity

0 2 4 6 8 10 12

02

46

810

PI121 ID=2041

Time (hr)

Pain

Inte

nsity

0 2 4 6 8 10 12

02

46

810

PI121 ID=2042

Time (hr)

Pain

Inte

nsity

0 2 4 6 8 10 12

02

46

810

PI121 ID=2043

Time (hr)

Pain

Inte

nsity

0 2 4 6 8 10 12

02

46

810

PI121 ID=2044

Time (hr)

Pain

Inte

nsity

Pai

n In

tens

ity (N

PR

S)



Model-Based Extrapolation

0.25 1 2 4 6 8 10 12


Time (hr)

NP

RS

Score

Fre

quency (

%)

020

60100

0.25 1 2 4 6 8 10 12


Time (hr)

NP

RS

Score

Fre

quency (

%)

020

60100

0.25 1 2 4 6 8 10 12


Time (hr)

NP

RS

Score

Fre

quency (

%)

020

60100

0.25 1 2 4 6 8 10 12


Time (hr)

NP

RS

Score

Fre

quency (

%)

020

60100

Rescued

NPRS = 10

NPRS = 9

NPRS = 8

NPRS = 7NPRS = 6

NPRS = 5

NPRS = 4

NPRS = 3

NPRS = 2

NPRS = 1NPRS = 0

0.25 1 2 4 6 8 10 12


Time (hr)

NP

RS

Score

Fre

quency (

%)

020

60100

0.25 1 2 4 6 8 10 12


Time (hr)

NP

RS

Score

Fre

quency (

%)

020

60100

0.25 1 2 4 6 8 10 12


Time (hr)

NP

RS

Score

Fre

quency (

%)

020

60100

0.25 1 2 4 6 8 10 12


Time (hr)

NP

RS

Score

Fre

quency (

%)

020

60100

Rescued

NPRS = 10

NPRS = 9

NPRS = 8

NPRS = 7NPRS = 6

NPRS = 5

NPRS = 4

NPRS = 3

NPRS = 2

NPRS = 1NPRS = 0

Dose 1 Dose 2

- Repeated-measures nonlinear mixed effects model usedto extrapolate individual responses over time

- View simulated response time-course in the absence ofdropout

Freq

uenc

y of

NP

RS

Sco

re

Freq

uenc

y of

NP

RS

Sco

re



Summary (1)The utility of exposure-response in PoC trials depends onstudy design and PoC goals.

- MTD-Type PoC E-R modeling of PoC data has minimal value; may be misleading

PK modeling still useful for understanding target population PK,reducing variability, or for explaining extreme outcomes

- Dose-Ranging PoC E-R has high value for design, analysis and PoC determination

Comparative E-R relationships across multiple endpoints/activecontrols provides insight into probability of achieving target productprofile

Advances knowledge building for future drug development phases

Basis for trial simulations to explore future designs



Summary (2)

- PK and E-R modeling and simulation: are tools for knowledge-building and decision support in drug

development

provide basis for trial simulations to explore and optimize trialdesign performance

are best supported by trial designs that explore individual E-Rrelationships

of multiple endpoints allows quantitative assessment of drug’smultivariate response profile, supporting dose-selection decisions

may be useful in assessing test treatment response in presence ofrescue dosing (preliminary)

may be useful for making inferences in the presence of dropout(for non-regulatory purposes)



Additional References- Krall RL, KH Engleman, HC Ko, and CC Peck. “Clinical Trial Modeling and Simulation – Warner KE, Peck

CC, Work in Progress.” /Drug Info J/, 32: 971-976, 1998.- Peck CC. “Drug development: Improving the process.” /Food and Drug Law J/. 52 (2):163-167, 1997.- Reigner BG, PEO Williams, IH Patel, J-L Steimer, CC Peck, and P van Brummelen. “An evaluation of the

integration of pharmacokinetic and pharmacodynamic principles in clinical drug development.” /ClinPharmacokinet/, 33(2): 142-52, 1997.

- Biomarkers Definitions Working Group. “Biomarkers and surrogate endpoints: Preferred definitions andconceptual framework.” /Clin Pharm Ther/ 69 (3): 89-95, 2001.

- Galluppi GR, MC Rogge, LK Roskos, LJ Lesko, MD Green, DW Feigal, and CC Peck. “Integration ofpharmacokinetics and pharmacodynamics studies in the discovery, development and review of proteintherapeutic agents: a conference report.” /Clin Pharm Ther/, 2001.

- Holford NHG, JPR Monteleone, HC Kimko, and CC Peck. “Simulation of Clinical Trials” in /Annual RevPharmacol Toxicol./ Vol. 40: 209-234, 2000.

- Lockwood, Ewy, Herman, Holford. Application of Clinical Trial Simulation to Compare Proof of ConceptDesigns for Drugs with a Slow Onset of Effect; An Example in Alzheimer’s Disease. Pharm Res. 23:9, 2006.

- Klingenberg B. A Unified Framework for Proof of Concept and Dose Estimation with Categorical Responses.www.williams.edu/~bklingen .

- Lalonde RL, Kowalski KG, Hutmacher MM, Ewy W, Nichols DJ, Milligan PA, Corrigan BW, Lockwood PA,Marshall SA, Benincosa LJ, Tensfeldt TG, Parivar K, Amantea M, Glue P, Koide H, Miller R. Model-baseddrug development. Clin Pharmacol Ther. 2007 Jul;82(1):21-32.

- Atkinson AJ Jr, Lalonde RL.Introduction of quantitative methods in pharmacology and clinical pharmacology:a historical overview. Clin Pharmacol Ther. 2007 Jul;82(1):3-6.

FDA Presentations on Model-Based Drug Development- http://www.fda.gov/oc/initiatives/criticalpath/presentations.html- http://www.fda.gov/ohrms/dockets/ac/03/slides/3998s1.htm- http://www.aapspharmaceutica.com/meetings/files/38/Booth.ppt



Acknowledgements

- Heidi Costa- Leonid Gibiansky- Bill Knebel- Matthew Riggs- Bill Gillespie

- Industry collaborators