Modeling and Simulation

hardtofindcurtainUrban and Civil

Nov 16, 2013 (3 years and 10 months ago)

254 views

Modeling and Simulation

beyond PK/PD

CPTR Workshop October 2


4, 2012

Pentagon City


M&S
-
WG Objective
:


For Preclinical Phases: Deliver Quantitative PKPD models to help sponsors
select therapeutic combinations





For Phase I: Deliver PBPK models to help sponsors predict first
-
in
-
human
results for combination regimens (
Pulmosim
/SIMCYP)


For Phase II & III: Deliver clinical trial simulation tools (based on
quantitative drug
-
disease
-
trial models) to be used to help design TB drug
regimen development studies



Here a more in
-
depth look at the clinical setting



Mission and Goals

CPTR M&S
Projects

PBPK

Clinical trial
simulation
tools

Preclinical
PKPD models


SIMCYP Grant
Application
(CPTR+U of F)


Pulmosim

tool
from Pfizer


Developed TB
modeling inventory


Develop drug
-
disease
-
trial model
for TB



White papers


FDA qualification


Data
standards



Data
sources



Database

3


Hollow
Fiber
model

PBPK


Complex ADME processes: PBPK models account for anatomical,
physiological, physical, and chemical mechanisms.



Multi
-
compartment approach to account for organs or tissues,
with interconnections corresponding to blood, lymph flows and
even diffusions.



Develops a system of differential equations for drug
concentration on each compartment as a function of time



Its parameters represent blood flows, pulmonary ventilation
rate, organ volumes etc., for which information is reliable known

[Enter Presentation Title in Insert Tab > Header & Footer

4

PBPK Integrates the Complex
P
rocess of Distribution


Normal
lung
tissue

[Enter Presentation Title in Insert Tab > Header & Footer

5


Inflamed
lung tissue


Granulomatous
tissue


CPTR

PBPK

6

PulmoSim
: Framework for inhaled drugs that can serve as a foundation for
orally administered antibiotics systemically distributed to the lungs

Clinical Trial Simulation Tools

Integrate the disease with pharmacology models

Takes into account design considerations

Gobburu
JV, Lesko LJ.
Annu

Rev
Pharmacol

Toxicol
. 2009;49:291
-
301.

8



Trial Simulations Optimize Design Based on
Quantitative Principles

Test Multiple Replications of

Trial Design Assumptions

Modify Design

0.4

0.5

0.6

0.7

0.8

0

10

20

30

40

50

60

Effect of Dose and Number of Subjects on Power to

Estimate Significant Effect of Drug vs Placebo

1 mg

2 mg

5 mg

10 mg

20 mg

30

4.5

6.5

18

48.5

73.5

40

13

29

76

87

91

50

27.5

52

85


95

99

60

40.5

62

90

97

100

70

55.5


71


94

99

100

N

Drug/Disease Model

Trial Designs


X possible doses


Different N


Sampling time


Inclusion criteria

Range of Outcomes

Analytics/Statistics

CFU

Trial Simulations
Optimize Design
Based on
Quantitative
Principles

For Predictions the Top
-
Down Approach is

Too
L
imiting


Describes existing
data, lacks
mechanistic insights,
limited to explore
new scenarios.

Davies
GR, et al.
Antimicrob

Agents
Chemother
. 2006;50(9):3154
-
6.

But the Bottom
-
up Approach is too expansive


Requires detailed
mechanistic
understanding,
makes models more
“portable”, limited by
unverifiable
assumptions.

Wigginton

JE, et al. A model to predict cell
-
mediated immune regulatory mechanisms during human infection with Mt. J
Immunol
. 2001;166:1951
-
67

Intermediate Approach: Mechanistically
-
Inspired


Retains key
mechanistic verifiable
components, allows
for parameter
estimations and is fit
for simulation
purposes

Marino S et al. A hybrid
multicompartment

model for granuloma formation and T
-
cell priming in TB. J of
Theor

Bio. 2011:280:50
-
62

Leverage can be Obtained
F
rom
O
ther
A
reas


Predator
-
Prey models
in viral infections such
as with HCV may
provide useful
insights for TB
modeling and
simulation

Guedj

J. et al. Understanding HCV dynamics with direct
-
acting antiviral agents due to interplay between intracellular replication and

cellular infection dynamics. J
Theor

Bio 2010;267:330
-
40

The Path
F
orward to a Successful M&S Platform in TB


Obtain the right datasets to model the dynamics of CFU as a function of drug
exposure/dose and disease progression in a mechanistically
-
inspired setting


Longitudinal data


Different combination therapies


Drug susceptible, MDR and XDR strain data



Develop model that is predictive of CFU and linked to outcome taking into
account appropriate other factors as co
-
therapy, demographics
etc



Test and validate the model(s) with regulatory buy
-
in



Develop tool that can interrogate the model to aid in trial design of
compounds under investigation or in development


[Enter Presentation Title in Insert Tab > Header & Footer

13

Regulatory Review Process: What’s success?

Informal discussion with FDA/EMA.

Sponsor submits a letter of intent requesting formal
qualification. FDA/EMA Review Team formed.

Sponsor submits briefing document.

F2F meeting between sponsor and FDA/EMA
Review Team. Review Team may request
additional information.

Sponsor submits full data package. Review
process within FDA/EMA begins.

Consultation
and

Advise Process

14

Regulatory decision qualifying
or endorsing the submitted
tools

Success!!!

Modeling and Simulation

beyond PK/PD

CPTR Workshop October 2


4, 2012

Pentagon City



WHAT PREDICTIVE MODELING SHOULD DO


A DISEASE MODEL AND A MATHEMATICAL MODEL SHOULD GIVE A
QUANTITATIVE PREDICTION:



HOW MUCH RESPONSE?



WITH WHAT DOSE?



ACCURACY SHOULD BE JUDGED BASED ON CLINICAL EVENT RATES and
NOT another model or CONSESUSS



ACCURACY SHOULD BE BASED ON HOW ACCURATE CLINICAL
PREDICTIONS ARE, NOT ON LACK OF COMPLEXITY OF THE MODELING

M. tuberculosis
in the hollow fiber system

Gumbo T
, et al. (2006)
J Infect Dis
2006;195:194
-
201






HFS:
Moxifloxacin

Concentration
-
Time Profile


0.0
0.5
1.0
1.5
0
6
12
18
24
30
36
42
48
Time (hours)
Concentration (mg/L)
HFS,
Simulations
and
Predictions Later
on
“Validated

with
CLINICAL
Data



Efflux pump & cessation of effect of antibiotics


The rapid emergence of quinolone resistance


The potency & ADR of
Cipro
/
O
rflox

versus
M
oxi


The “biphasic” effect of quinolones


The exact dose of Rifampin associated with optimal
effect


The population PK variability hypothesis, and the rates
of ADR arising during DOTS


The role of higher doses of pyrazinamide


The “breakpoints” that define drug resistance

The HFS in
Quantitative
P
rediction

HFS quantitative output on the relationship between
changing concentration and microbial effect




Human pharmacokinetics and their variability






MODELING & SIMULATIONS


Predictive outcome: dose, breakpoints, microbial effect,
resistance emergence, regimen performance



Gumbo T, et al.
Antimicrob
. Agents. Chemotherapy. 2007: 51:2329
-
36


ISONIAZID HFS: Monte Carlo
Simulations


INH inhibitory sigmoid
E
max

based on hollow fiber studies


% patients with nat
-
2 SNPs associated with fast acetylation versus
slow acetylation in different ethnic groups: Cape Town, Hong Kong,
Chennai


M. tuberculosis MICs in clinical isolates


Population PK data from (
Antimicrob.Agents

Chemother
. 41:2670
-
2679) input into the subroutine PRIOR of the ADAPT II


9,999 Monte Carlo simulation for different ethnic groups to sample
distributions for SCL→AUC→AUC/MIC→EBA


Gumbo T, et al. Antimicrob. Agents. Chemotherapy. 2007: 51:2329
-
36


PK
-
PD PREDICTED
vs

OBSERVED EBA IN CLINICAL TRIALS

Gumbo T, et al. Antimicrob. Agents. Chemotherapy. 2007: 51:2329
-
36






PREDICTION

PREDICT:

Etymology

via

Latin
:




præ
-
, "before"






dicere
, "to say".


“PREDICT” to say BEFORE


QUALITATIVE:

Predict
an event in terms of
whether
it occurs


QUANTITATIVE:

Predict
extent and values
prior to
the event

ORACLES AND DEVINING THE FUTURE

http://www.crystalinks.com/delphi.html



If MDR
-
TB
Does
N
ot
A
rise
F
rom
P
oor
C
ompliance
,

W
hy
D
oes
I
t
?


Hypothesis: Perhaps the PK system (i.e., patient’s
xenobiotic metabolism) is to blame


HFS output: kill rates, sterilizing effect rates (i.e.,
log10 CFU/ml/day)


Known clinical kill rates, sterilizing effect rates (i.e.,
log10 CFU/ml/day)


Performed MCS in 10,000 Western Cape Patients on
the FULL REGIMEN


Srivastava S, et al. J. Infect. Dis. 2011; 204:1951
-
9
.




Sputum conversion rate predicted = 56% of patients


Sputum conversion rate from prospective clinical studies in WC= 51
-
63%

External
Validation
of
Model
:

S
putum
C
onversion
R
ates in
10,000
Patients

Srivastava S, et al. J. Infect. Dis. 2011; 204:1951
-
9
.





Many (simulated) patients had 1
-
2 of the 3 drugs at very
low concentration throughout, leading to
monotherapy

of the remaining drug



Drug resistance predicted to arise in 0.68% of all
pts

on
therapy in first 2 months despite 100% adherence

Srivastava

S, et al. J. Infect. Dis. 2011; 204:1951
-
9
.





Prospective
study of 142
patients in the Western Cape
province of South Africa


Jotam

Pasipanodya
, Helen
McIlleron
*,
André Burger, Peter A. Wash, Peter Smith,
Tawanda

Gumbo

Pasipanodya

J, et al. Submitted.



What W
as Done


All patients hospitalized first 2
months


All had 100% adherence first 2
months


Drug concentrations measured at 8 time points
over 24hrs in month
2


Followed for 2 years, 6% non
-
adherence





Pasipanodya J, et al. Submitted.



CART ANALYSIS: Top 3 predictors of Long term outcomes

Pasipanodya J, et al. Submitted.


0.7% patients developed ADR in 2 months versus 0.68% we predicted IN THE PAST from
modeling and simulations
: All ADR had low concentrations of at least one drug


Thank you!

[Enter Presentation Title in Insert Tab > Header & Footer

31

Identifying sources of variability


Individual variability in blood/air
flow with body positions may
affect drug distribution and
elimination in different parts of
the lung

http://www.nigms.nih.gov/NR/rdonlyres/8ECB1F7C
-
BE3B
-
431F
-
89E6
-
A43411811AB1/0/SystemsPharmaWPSorger2011.pdf

32

Identifying sources of variability


Dormant and active bacterial
populations may exhibit
different effect sizes, even at
saturation concentrations

http://www.nigms.nih.gov/NR/rdonlyres/8ECB1F7C
-
BE3B
-
431F
-
89E6
-
A43411811AB1/0/SystemsPharmaWPSorger2011.pdf

33

Identifying sources of variability


Levels of resistance
may explain a drug’s
varying IC
50

magnitudes

http://www.nigms.nih.gov/NR/rdonlyres/8ECB1F7C
-
BE3B
-
431F
-
89E6
-
A43411811AB1/0/SystemsPharmaWPSorger2011.pdf

34

Identifying sources of variability


Additional factors
that induce
variability in a
defined population?

http://www.nigms.nih.gov/NR/rdonlyres/8ECB1F7C
-
BE3B
-
431F
-
89E6
-
A43411811AB1/0/SystemsPharmaWPSorger2011.pdf

35

Identifying sources of variability


Deeper mechanistic
understanding of the
disease processes

http://www.nigms.nih.gov/NR/rdonlyres/8ECB1F7C
-
BE3B
-
431F
-
89E6
-
A43411811AB1/0/SystemsPharmaWPSorger2011.pdf

36

The new CPTR modeling and simulation work group


Integrating quantitative systems pharmacology, spanning different stages of
the combination drug development process for TB



Leveraging previous work to advance existing drug development tools and
develop new ones for specific contexts of use



Data
-
driven modeling and simulation tools: data standards and databases
from available and relevant studies



Spearheading regulatory review pathways with FDA and EMA, to facilitate
the applicability of those drug development tools



Aligning and cross
-
fertilizing with other work groups to increase efficiency

[Enter Presentation Title in Insert Tab > Header & Footer

37