Transcript Diapositive 1 - Physiologie et Thérapeutique Ecole Véto Toulouse

ECOLE
NATIONALE
VETERINAIRE
TOULOUSE
Predictive value of PK/PD drug
modelling: application to
analgesic drugs
PL Toutain
UMR 181 Physiopathologie et Toxicologie Expérimentales
INRA, ENVT
Satellite symposium: Validity and Quality of
Animal Models for Measurement of Pain
Objectives of the presentation
1. Overview on the concept of PK/PD
2. Predictive value of PK/PD modeling
for analgesics
What is PK/PD modeling?
• PK-PD modeling is a scientific tool to
quantify, in vivo, the key PD
parameters (efficacy, potency and
sensitivity) of a drug, which allows to
predict the time course of drug effects
under physiological and pathological
conditions (intensity and duration)
What are the main practical
applications of a PK/PD trial
Preclinical investigations:
It is an alternative to dose-titration studies to
discover a dosage regimen
Clinical setting:
It is a tool to optimize dosage regimen in a
clinical setting (pop PK/PD)
1-An overview on the
concept of PK/PD
Dose titration
Dose
Response
Black box
PK/PD
PK
Response
PD
Dose
Plasma
concentration
Why is plasma concentration profile a
better explicative (independent) variable
than dose for determining a dosage
regimen ?
Dose vs. plasma concentration
profile as independent variable
Dose
Mass
(no biological
information)
Dose
X
F%
Clearance
Time
Concentration profile
(biological information)
Why to prefer a PK/PD
approach to a classical
dose-titration?
The determination of an ED50 or any
ED%
PD
ED50 =
Clearance x target EC50
Bioavailability
PK
ED50 - is a hybrid parameter (PK and PD)
- is not a genuine PD drug parameter
The 3 structural PD parameters:
Dose titration (DT) vs. PK/PD
Emax
1
Emax 1
Emax 2
ED50/EC50
Slope
Sensitivity
1
1
2
Emax/2
steep
shallow
2
Efficacy
2
ED501 ED502
• Range of useful
concentrations
Potency
• Selectivity
Dose
Titration
Emax
ED50
No
PK/PD
Emax
EC50
yes
Why to prefer a PK/PD approach
to a classical dose-titration?
2.The separation of PK
and PD variability
PK/PD variability
• Consequence for dosage adjustment
PK
Dose
PD
BODY
Receptor
Effect
Plasma
concentration
Kidney function
Liver function
...
Clinical covariables
• Pain severity or duration
PK/PD population approach
2-Predictive value of
PK/PD for analgesics
Predictive value of PK/PD modeling
rely on:
1. The question:
–
Mechanistic question vs. Clinical drug development
2. Selection of a pain model & In life validation of
the selected model
3. Appropriate study design & conduct
4. Appropriate PK & PD data
5. Appropriate PK/PD modeling
6. Population PK/PD (clinical setting)
The question: a mechanistic question
Drug discovery
Questions for a veterinary rational drug
development: find an optimal dosage
regimen for a target species
• What is the typical Dosage regimen
• Time information and decision
– Onset of drug action: fentanyl vs. morphine
– Duration of drug action: time of remedication ( Dosage interval)
• Extrapolation
– Between species
• assumption of the same PD parameters
– Within the same species: between route of administration
• Assumption: different PK profile but same qualitative metabolic profile
• Dosage adjustment
– Population investigations
2-Selection of a pain model:
experimental pain models vs.
clinical pain for PK/PD
investigations
Pain models
For PK/PD
investigation
Clinical
Preclinical
Inflammatory
Dose
determination
e.g. NSAIDs
pain≠nociception
Non
Inflammatory
Dose
determination
Opioids
Gabapentine
Surgical models
Possibility to
standardize
Dose
confirmation
Spontaneous
pain
neuropathy
Dose
adjustment
Pop PK/PD
Pain model selection for PK/PD
investigation: value & validity
• Validity:
– to be discussed by the pain’ specialist
– refers to whether a study is able to scientifically
answer the questions it is intended to answer
– Regarding the ultimate objective:
• To investigate neurophysiologic mechanisms of pain or
complicate drug mechanism of action
• Preclinical determination of a dosage regimen
– Simple but reproducible antinociceptive model are often
sufficient
– Validityof a model =capacity to find a useful dose
• Value:
– to be demonstrated by the PK/PD trialist
Pain model selection for PK/PD
investigation: value & validity
• Validity
• Value
– Ethical
– Metrological performances
• Reliable
• Sensitive
• Robust & transferable
– Convenience
– Etc.
Models using pressure noxious
stimulus or thermal noxious
stimulus are considered as
valuable in veterinary medicine
to approximate a starting dose
Inflammatory pressure noxious stimulus.
(here a kaolin inflammation model)
Measure of vertical forces exerted on force
plate
• To measure the
vertical forces, a
corridor of walk is
used with a force
plate placed in its
center.
• The cat walks on the
force plate on leach.
Video
Measure of vertical forces
exerted on force plate
• The measure of
vertical force and
video control are
recorded
 Vertical forces (Kg)
Video
Measure of pain with analgesiometer
• The time for the cat to
withdraw its paw of the ray is
measured.
 withdrawal time of the
paws (second)
 Sensitive and specific model to
activate C-fibers
Video
Validation of the selected model
Validation of the model
1.
A priori validation makes sure the method is
suitable for its intended use
–
When developing a new method
2. In life validation (routine validation for any
new trial)
–
–
–
–
Animal selection
Investigator skill
Reproducibility & repeatability of selected animals
etc
Validation of the model is tedious
Predictive value of PK/PD modeling
rely on:
1. The question:
2. Selection of a pain model & In life
validation of the selected model
3. Appropriate study design & conduct
•
Crossover design and placebo period
4. Appropriate PK & PD data
5. Appropriate PK/PD modeling
6. Population PK/PD (clinical setting)
4-Appropriate data for PK/PD
modeling
Measuring variables in PK/PD trials
Measuring drug exposure
• Full concentration time
curve
– experimental setting
• Cmax , Cmin
– Clinical setting
Measuring drug response
• Biomarkers
• Surrogate
• Clinical outcomes
Measuring exposure
• Generally straightforward.
• May be more complicate if:
– presence of an active metabolite
• Tramadol
– Racemates
• Profens
Tramadol plasma concentration (ng/mL) vs. time
(min) after an IM administration of tramadol (circa
8 mg/kg);
pharmacokinetics of (±)-trans-T and M1
are stereoselective in vivo
•Trans-tramadol [(±)-trans-T] hydrochloride is a
chiral compound
• (+)-, (-)-Trans-T take as the action mainly
through inhibiting the reuptake of serotonin and
norepinephrine, respectively
•The drug is metabolized in the liver to form five
phase I metabolites, with the main pathways (in
man and rats) being O-demethylation to Odemethyltramadol (M1)
•Among the metabolites, M1 is an only active
metabolite, and (+)-M1 has a high affinity to
the opioid receptor
Substances
Action
RR-T
No action
SS-T
Monoamine
re-uptake
µ-opioid
RR-M1
SS-M1
Monoamine
re-uptake
Pharmacodynamic parameters
of tramadol in the rat
Action
Action
IC50 (ng/mL)
RR-T
No action
NA
SS-T
Monoamine
re-uptake
µ-opioid
230
Monoamine
re-uptake
869
RR-M1
SS-M1
20.2
Tramadol and tramadol metabolite M1 concentration (ng/mL)
vs. time (min) in 8 dogs after an IM administration of
tramadol (circa 8 mg/kg) ;
Spaghetti plot; semilogarithmic scale
No CYP2D6 in dogs but an ortholog i.e CYP2D15
Plasma concentrations of R- and Sketoprofen after intramuscular administration
of ketoprofen ( 6 mg/kg)
Concentration ( g/ml)
100
R-ketoprofen
S-ketoprofen
10
1
0.1
0.01
0
10
20
Time (h)
30
Time development of the plasma concentration of ketoprofen and the
mechanical nociceptive thresholds before kaolin injection (negative
control), after kaolin injection (positive control) and after ketoprofen
administration
R-keto
S-Keto
Nociception
Kaolin
EC50
R-keto=2.0±05 µg/mL
S-ket=38.8±10.8
T. K. FOSSE et al JVPT in press
Measuring variables in PK/PD trials
Measuring drug exposure
Measuring drug response
• Full concentration
time curve
• AUC
• Cmax , Cmin
• Biomarkers
• Surrogate
• Clinical outcomes
Which dependent variable for PK/PD
modeling ?
EC50 in vivo effect
EC50 action
whole blood
assay
NSAID
plasma
concentration
Inhibition
of COX
Inhibition of
PGE2
production
Suppression
of lameness
Requires 90% PGE2 inhibition
EC50 response
EC50 response >> EC50 effect
5-PK/PD modelling
Modeling options regarding presence or not of a
delay between PK and PD time development
No PK modeling
E=
Emax x Cobserved
EC50 + Cobservedl
NO
PK modeling
PK and PD
delay
E=
Emax x C(t)model
EC50 + C(t)model
PK origin
Effect compartment model
PD origin
Indirect response model
YES
Thermal threshold
Plasma Fentanyl
•No hysteresis for fentanyl
•Direct incorporation of plasma
fentanyl concentration in an Emax
model
hysteresis loop
ΔT(ºC)
IV
Oral
Buprenorphine concentration
Modeling strategies when
there is a delay of PK origin
The “effect compartment model”
Dose
effect
Time
Effect
Ke0
Concentration
Ce(t)
Ke0
Effect(t)
Effect
Cp(t)
Time
Ce
K10
1:PK model
Parametric (Exponential)
Non parametric (Spline)
2:Link model
Ke0
3:PD model
Parametric (Emax, Hill)
Non parametric (spline)
Estimation of EC50 and Ke0
A mechanistic class of
PK/PD models
An example of dose determination
using a PK/PD modeling approach:
Tramadol in dogs
Thermal stimulus: time course (h) of the paw
withdrawal time expressed as a percentage of the
control value
Tramadol 8mg/kg
150%
100%
Placebo
70%
0
1
2
3
4
5
Time (h)
6
7
8
Data modeling using an indirect
effect model
Kin is the (control) zeroorder rate constant of
the response formation
Kout is the first-order rate
constant of response
disappearance
Rate of change of the
response (withdrawal time,
WT) over time
Model of placebo effect
Observed and fitted response (WT in sec) vs. time (h) to
tramadol after IM administration of tramadol to a dogs .
150%
Withdrawal time (%)
Tramadol 8mg/kg
100
Placebo
70%
0
1
2
3
4
Time (h)
5
6
7
8
Dose effect relationship for tramadol as predicted by
the PK/PD model.
14mg/kg
5mg/kg
1mg/kg
Placebo
time course of effect from 0 to 4h post administration for different IM doses of
tramadol ranging from 1 to 14 µmg/kg
Dose effect relationship for tramadol.
0 to 4h
Emax=362 (%*h)
ED50=4.67mg/kg
0 to 6h
Emax=581 (%*h)
ED50=9.90mg/kg
Doses are from 0 to 14 mg/kg and effects are expressed by the Area Under
the Effect vs. time curves (%*h) from 0 to 4 or 0 to 6h post tramadol
administration
Tramadol: dose-effect Relationship: 7mg/kg
IM vs PO
IM
PO
Placebo
Predictive value of PK/PD modeling
rely on:
1. The question:
–
Mechanistic question vs. Clinical drug development
2. Selection of a pain model & In life validation of
the selected model
3. Appropriate study design & conduct
4. Appropriate PK & PD data
5. Appropriate PK/PD modeling
6. Population PK/PD (clinical setting)
6-Experimental vs. observational
population approach
Two questions regarding experimental
approach
• What is its validity (clinical relevance)
• What about intersubject variability
Dog model “accuracy”
Experimental
Observational Population
• Highly selected (as
homogeneous as
possible) body weight,
sex, age...
• Representative of the target
population different breed,
age, pathological
conditions…
e.g. Beagle dogs
Beagle dogs:
strain (colony) effect
Some strains are very responsive
•
Some strains are very resilient
Some strains are responsive to pain thermal stimulus while some others are
totally unresponsive
– (strain raised for toxicology and selected and trained to be as quiet as possible)
Dog enrolled in a trial based on their individual reproducibility(<25% over 3 days)
Cat model “accuracy”
Not selected for experimental purposes
Are re-homed after trial completion
Experimental pain model “accuracy”
• Experimental nociception
• Clinical pains
– Inflammatory pain
– Visceral pain
– Muscle and joint pain
– Peripheral neuropathy
– Central neuropathy
– Cancer pain
Variability is a biological fact
not a noise …
What is population PK/PD
Goal:
• to determine the sources of PK and PD variability
in the target animal population as well as the
magnitude of that variability, in order to design
dosage regimens that account for individual animal
(or group) characteristics
• to adapt dosage regimen to different subjects of
the population having a given characteristic (e.g.
breed)
Pain subjective assessment
(composite measurement of behavioral & physiological signs)
• Data analysis
– Ordinal (Y/N) or interval scale?
• Scoring rating scale
– Simple descriptive scale (SDS)
– Numerical rating scale (NRS)
– Visual analogue scale (VAS)
• Issues:
– reliability
• Confounding factors (hospitalization, anesthetics, drugs given perioperatively (including some antibiotics as aminoglycosides…);
unresponsiveness of some species; reproducibility between
observers
– Validity:
• No assessment of the subjective part of pain as for self-reporting in
man
Probability of pain alleviation (POA)
• Logistic regression may be used to link measures of drug
exposure to the probability of a clinical success
POA 
Dependent
variable
1
1 e
a bf  through_ conc 
Placebo
effect
sensitivity
Independent
variable (analgesic
exposure)
2 parameters: a (placebo effect) & b (slope of the exposure-effect curve)
Probability of pain relief:
1.2
Probability of pain relief
1
Slope is controlled by the the
intersubject variability,
For morphine in man, the slope factor
is of 3.6 indicating there is
approximately fourfold variability
between subjects.
0.8
0.6
0.4
0.2
0
0
50
100
150
200
Analgesic plasma AUC
1
1  e 2.190.03509 AUC
In analgesic studies in man, the mean effective concentration
(MEC), which is the concentration at time remedication is required,
is usually obtained in this manner.
PK / PD modeling
Conclusions
1. A powerful tool for dose determination and
adjustment or mechanistic purposes
– If a a clear understanding of theoretical background
and computer software.
– If appropriate design (placebo) and metrological
validation of the different endpoints
2. In preclinical setting, the question of the validity
of the selected experimental model holds
3. In clinical setting, there is no longer a “model “
but the main difficulty is the validity (reliability) of
the pain assessment