PK-PD analysis and modelling - UCLouvain€¦ · Nov 2017 PK/PD and modelling Why modelling ? (*)...
Transcript of PK-PD analysis and modelling - UCLouvain€¦ · Nov 2017 PK/PD and modelling Why modelling ? (*)...
Nov 2017 PK/PD and modelling 1
PK-PD analysis and modelling
Nov 2017 PK/PD and modelling
Why modelling ? (*)
• to move from mere description to underlying phenomena…– nature can often be better explained in terms of equations than
mere description
– this has been essential in physics (think about gravity law, radioactive decay, study of electromagnetic field and optics, … up to the equivalence of mass and energy…)
• to allow predictions over and beyond what is immediately accessible by the experience…
• to generate rules that can be applied widely…
* CAUTION: modelling in UK English but modeling in US English …
2
Nov 2017 PK/PD and modelling
In vitro studies
3
Nov 2017 PK/PD and modelling
Response to an antimicrobialan example with ceftobiprole and S. aureus (one strain)
Effect-over-time
0 6 12 18 24-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.50.010.2122050200
concentration(multiples of MIC)
time (h)
log1
0 C
FU
4
Nov 2017 PK/PD and modelling
Response to an antimicrobialan example with ceftobiprole and S. aureus (2 strains)
Effect-over-time(2 strains)
0 6 12 18 24-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.50.010.2122050200
concentration(multiples of MIC)
time (h)
log1
0 C
FU
This where the phenomenon happens
5
Nov 2017 PK/PD and modelling
Response to an antimicrobial: the modelan example with ceftobiprole and S. aureus (2 strains)
-2 -1 0 1 2 3
-1
0
1
2
3 MSSA #1
MSSA #2
log10 multiples of MIC measured at pH 7.4
∆ lo
g cf
u (2
4 h
- 0 h
)
6
Nov 2017 PK/PD and modelling
Response to an antimicrobial: the modelan example with ceftobiprole and S. aureus (multiple strains)
MIC
CS
Emax
-2 -1 0 1 2 3
-1
0
1
2
3 MSSA HA-MRSA CA-MRSA
log10 multiples of MIC measured at pH 7.4
∆ lo
g cf
u (2
4 h
- 0 h
)Emin
7
Nov 2017 PK/PD and modelling
Analyses
Equation for Prism
Equation:Sigmoidal dose-responseY=Bottom + (Top-Bottom)/(1+10^((LogEC50-X)))
;X is the logarithm of concentration. Y is the response;Y starts at Bottom and goes to Top with a sigmoid shape
Sigmoidal dose-response:
also called "4-parameters logistic equation", i.e.• bottom (Emin)• Top (Emax)• EC50• Hill slope
Sigmoid dose-response
-1 0 1 2 3 40.0
2.5
5.0
7.5
10.0
Hill slope = 0.5 1.0 2.0
concentration (log10)
resp
onse
Emin
Emax
EC50
8
Nov 2017 PK/PD and modelling
AnalysesEquation
Equation:Sigmoidal dose-responseY=Bottom + (Top-Bottom)/(1+10^((LogEC50-X)))
;X is the logarithm of concentration. Y is the response;Y starts at Bottom and goes to Top with a sigmoid shape
9
Nov 2017 PK/PD and modelling
Type of functions
how would you fit those
data
Do not forget to use the appropriate axes !
10
Nov 2017 PK/PD and modelling
Type of functions
This would be a good
model
11
Nov 2017 PK/PD and modelling
Run statistics
12
Nov 2017 PK/PD and modelling
Run tests
13
Nov 2017 PK/PD and modelling
Two examples
14
Nov 2017 PK/PD and modelling
Impact of MIC on the response of intracellular bacteria to moxifloxacin
-3 -2 -1 0 1 2-3
-2
-1
0
1
2
3
SA069
SA481
NRS386
NRS192
NRS384
≤ 0.06 0.125 1.0
SA1
MIC (mg/L)
HMC 551KKH II-7924
≥ 2.0
log10 extracellular concentration (mg/L)
-3 -2 -1 0 1 2 3 4-3
-2
-1
0
1
2
3
log10 extracellular concentration (x MIC)
after normalization for MIC
Lemaire et al. Journal of Antimicrobial Chemotherapy (2011) 66:596-607
15
Nov 2017 PK/PD and modelling
Colistin and inoculum effect
colistin concentration (mg/L)
low inoculum (∼ in vitro testing)
high inoculum (∼ in vivo
The extent and rate of killing of P. aeruginosa by colistin were markedly decreased at high CFUo compared to those at low CFUo.Bulita et al. Antimicrob. Agents Chemother. (2010) 54:2051-2062
16
Nov 2017 PK/PD and modelling
In search of models with Prism
17
Nov 2017 PK/PD and modelling
In search of models (including your own)
18
Nov 2017 PK/PD and modelling
In search of models (including your own)
19
Nov 2017 PK/PD and modelling
And here you are …azithromycin
-5 -4 -3 -2 -1 0 1 2-4
-3
-2
-1
0
1
2
Cmax0.5 mg/L
MIC, 0.01-0.03 mg/LCs, 3 mg/L
Log concentration (mg/L)
∆ lo
g cf
u fr
om ti
me
0 (4
8 h)
clarithromycin
-5 -4 -3 -2 -1 0 1 2-4
-3
-2
-1
0
1
2
Cmax1 mg/L
MIC, 0.008 mg/LCs, 0.06 mg/L
Log concentration (mg/L)∆
log
cfu
from
tim
e 0
(48
h)
telithromycin
-5 -4 -3 -2 -1 0 1 2-4
-3
-2
-1
0
1
2
Cmax1 mg/L
MIC, 0.008 mg/LCs, 0.04 mg/L
Log concentration (mg/L)
∆ lo
g cf
u fr
om ti
me
0 (4
8 h)
ciprofloxacin
-5 -4 -3 -2 -1 0 1 2-4
-3
-2
-1
0
1
2
Cmax4 mg/L
MIC, 0.01 mg/LCs, 0.002 mg/L
Log concentration (mg/L)
∆ lo
g cf
u fr
om ti
me
0 (4
8 h)
moxifloxacin
-5 -4 -3 -2 -1 0 1 2-4
-3
-2
-1
0
1
2
Cmax4 mg/L
MIC, 0.01 mg/LCs, 0.001 mg/L
Log concentration (mg/L)
∆ lo
g cf
u fr
om ti
me
0 (4
8 h)
finafloxacin
-5 -4 -3 -2 -1 0 1 2-4
-3
-2
-1
0
1
2
Cmax12 mg/L
MIC, 0.01 mg/LCs, 0.05 mg/L
Log concentration (mg/L)
∆ lo
g cf
u fr
om ti
me
0 (4
8 h)
20
Nov 2017 PK/PD and modelling
In vivo pharmacokinetics
21
Nov 2017 PK/PD and modelling 22
What is PK analysis and modeling ?
• Noncompartmental analysisNoncompartmental PK analysis examines total drug exposure and looks for function(s) fitting the change of concentration over time without reference to where the drug may distribute.
Analysis is simple and does not imply anything concerning the actual fate of the drug.
The results are purely descriptive and non-predictive unless the function selected is linked to physical phenomena (e.g. 1st order kinetics).
Nov 2017 PK/PD and modelling 23
What is PK analysis and modeling ?• Compartmental analysis
Describes and predicts the concentration-time curve based on the movements of the drug between compartments (kinetic or physiological model)
Once the model is indentified, it can be used to predict the concentration at any time.
The model may be (very) difficult to develop
The simplest PK compartmental model is the one-compartmental PK model with IV bolus administration and first-order elimination.
The most complex PK models rely on the use of physiological information to ease development and validation.
23
Nov 2017 PK/PD and modelling 24
What is PK analysis and modeling ?
• Compartmental analysis
The simplest PK compartmental model is the one-compartmental PK model with IV bolus administration and first-order kinetic elimination
This can be developed with simple software accessible to lay users such as Prism (with some sophistication sometimes)
More complex PK models rely on the use of physiological information to ease development and validation.
This requires "high capacity" software that is often impossible to use without serious introduction
24
Nov 2017 PK/PD and modelling
Simple compartmental models
25
Nov 2017 PK/PD and modelling
Integrating … (calculus)
26
Nov 2017 PK/PD and modelling
From model to data and finding "best parameters" with a computer (curve fitting)
• choose (or enter) your equation• enter your data• enter initial parameter values
(best estimate; optional but useful)• the computer will then
– compare equation-based curve to actual data– modify parameters by successive iterations
until a "best" fit is obtained …– the limit is the number of iterations
numerical integration
27
Nov 2017 PK/PD and modelling
From data to model with a computer (no calculus)
28
Nov 2017 PK/PD and modelling 29
Example of monocopartmental analysis … (*)
equation: C = C0.e-kt
theoretical curve
Exponential-decay (1 compartment)
0.0 2.5 5.0 7.5 10.00
25
50
75
100
125
150
Co = 142 mg/L - 2 g - 70 kg - Vd = 0.2 L/kgke = 0.185 (t1/2 = 2h)plateau = 0
time (h)
conc
entr
atoi
n (m
g/L)
* this analysis and the following ones concern ceftazidime IV
29
Nov 2017 PK/PD and modelling 30
Fitting to ideal population data (*)Ceftazidime: ideal patients
0.0 2.5 5.0 7.5 10.00
25
50
75
100
125
150
* data from a few volunteers
30
Nov 2017 PK/PD and modelling 31
Ideal population: tests for 95 % CICeftazidime: ideal patients
0.0 2.5 5.0 7.5 10.00
25
50
75
100
125
150
31
Nov 2017 PK/PD and modelling 32
Ideal population: residuals
ideal-valuesNonlin fit of ideal-valuesData Table-1:Residuals
0 1 2 3 4 5 6 7-7
-5
-3
-1
1
3
5
time (h)
why are they much larger
here ?
32
Nov 2017 PK/PD and modelling 33
Real population (*)ceftazidime: real population
0.0 2.5 5.0 7.5 10.00
25
50
75
100
125
150
n = 27 patientsCo = 98 mg/L (2 g - 71.5 ± 10.9 kg)ke = 0.1865 (t1/2 = 3.716 h)plateau = - 8.2
time (h)
conc
entr
atio
n (m
g/L)
* data from several patients
33
Nov 2017 PK/PD and modelling 34
Real population: 95 % CI ceftazidime: real population
0.0 2.5 5.0 7.5 10.00
25
50
75
100
125
150
34
Nov 2017 PK/PD and modelling
Real population: residuals
0 1 2 3 4 5 6 7-35
-25
-15
-5
5
15
25
35
time (h)
resi
dual
s
35
Real population: residuals
Ideal population: Residuals
0 1 2 3 4 5 6 7-35
-25
-15
-5
5
15
25
35
time (h)
resi
dual
s
35
Nov 2017 PK/PD and modelling 36
More complex models: accumulation / decay
equation: C = D/Vd x ka/(ka-ke) [e-ket – e-kat]
theoretical curve
Bateman function(applied to ceftazidime)
0.0 2.5 5.0 7.5 10.00
25
50
75
100
125
150
D = 2 g / 70 kg (28.57 mg/kg) - Vd = 0.2 L/kgka = 8ke = 0.3465 (t1/2 = 2h)plateau = 0
time (h)
conc
entr
atio
n (m
g/L)
36
Nov 2017 PK/PD and modelling
In search of more complex models with Prism
37
Nov 2017 PK/PD and modelling
Ceftazidime with Bateman function
0 1 2 3 4 5 6 70
25
50
75
100
time (h)
conc
entr
atio
n (m
g/L)
38
Accumulation / decay with Prism … (*)
equation: C = D/Vd x ka/(ka-ke) [e-ket – e-kat]
real data
R2 = 0.57
38
Nov 2017 PK/PD and modelling 39
Examples d'analyse monocompartimentale … (*)
equation: C = D/Vd x ka/(ka-ke) [e-ket – e-kat]
Ceftazidime with Bateman
0 1 2 3 4 5 6 70
25
50
75
100
Prism has a
problem here !
39
Nov 2017 PK/PD and modelling 40
When the data become really too complex…
40
Nov 2017 PK/PD and modelling
The Mixed non-lin approaches
• A mixed model is a statistical model containing both fixed effects and random effects.
• These models are useful in a wide variety of disciplines in the physical, biological and social sciences.
• They are particularly useful in settings where repeated measurements are made on the same statistical units (longitudinal study), or where measurements are made on clusters of related statistical units.
• Because of their advantage in dealing with missing values, mixed effects models are often preferred over more traditional approaches such as repeated measures ANOVA.
41
Nov 2017 PK/PD and modelling
The Mixed non-lin approachesDifferent softwares, but all working by numerical integration based on pre-defined models
42
Exemples avec la témocilline
Nov 2017 PK/PD and modelling 43
Nov 2017 PK/PD and modelling 44
Temocillin in a nutshell
• Temocillin or 6-α-methoxy-ticarcillin• Registered in 1984 for the first time (Beecham)• Maintained on the market since 1998 (Eumedica)
– BE, LU, UK and now FR
• Narrow-spectrum antibiotic (Gram-negative oriented)– Enterobacteriaceae
– B. cepacia
– Neisseria, Haemophilus, Pasteurella, Legionella
– Inactive against most strains of P. aeruginosa, Acinetobacter,Stenotrophomonas,
– no useful activity against Gram-positive and anaerobes
• Stable to most β-lactamases– Class A (including ESBL, KPC), class C (AmpC), class D (OXA-1)
– Hydrolysed by OXA-48-like (class D) and class B enzymes (metallo-enzymes)
H3C
OHN
S
NO
O-
O
H
O
OO
S
But what if you place the bulky group on the β-lactam ring ?
Nov 2017 PK/PD and modelling 45
HHN
S
NO
O-
O
H
O
OO
S
H3C
OHN
S
NO
O-
O
H
O
OO
S
ticarcillin temocillin
water cannot
access…
Matagne et al. Biochem. J. (1993) 293, 607-411
Nov 2017 PK/PD and modelling 46
Why me and temocillin ?
As a result …
Nov 2017 PK/PD and modelling 47
Belgian SmPC, last revision 2012; Van Landuyt et al, AAC 1982; 22:535-40
Susceptible organisms
MIC < 1 mg/L 1 mg/L < MIC < 10 mg/L 10 mg/L < MIC < 100 mg/LMoraxella catarrhalisHaemophilus influenzaeLegionella pneumophilaNeisseria gonorrhoeaeNeisseria meningitidis
Brucella abortusCitrobacter spp.Escherichia coliKlebsiella pneumoniaePasteurella multocidaProteus mirabilisProteus spp (indole +)Providencia stuartiiSalmonella TyphimuriumShigella sonneiYersinia enterocolitica
Serratia marcescensEnterobacter spp
Intrinsically resistant organisms
anaerobesGram(+) bacteriaAcinetobacter sppPseudomonas aeruginosa
ESKAPE pathogens
Chemical stability of temocillin in concentrated solutions
Nov 2017 PK/PD and modelling 48
De Jongh et al. Journal of Antimicrobial Chemotherapy (2008) 61:382-388 – Supplementary Material
Comparative chemical stabilities of β-lactams upon storage of concentrated solutions at 25 and/or 37°C
Nov 2017 PK/PD and modelling 49
Conclusion Molecule Stability limit 1 reference
good temocillin > 24 h at 37°C 2 De Jongh et al. JAC 2008
aztreonam > 30 h at 37°C Chanteux et al. (abstract)
piperacillin 24 h at 37°C Viaene et al. AAC 2002
weak ceftazidime 24 h at 25°C / 8 h at 37°C Servais et al. AAC 2001
problematic cefepime color appearance within 6 h Baririan et al. JAC 2003
insufficient imipenem < 5 h Viaene et al. AAC 2002
meropenem < 5 h Viaene et al. AAC 2002
doripenem ∼ 6-10 h Berthoin et al. JAC 2010
1 > 90 % of original compound (European Pharmacopoiea) 2 stable for 3 weeks at 4°C (for home medication) (Carryn et al., J Antimicrob Chemother 2010;65:2045-2046)
JAC: J Antimicrob ChemotherAAC: Antimcrob Agents Chemother
• For β-lactams, – only the free fraction is (probably) active…
Nov 2017 PK/PD and modelling 50
Temocillin pharmacodynamics: the lessons of β-lactams
Exemple #1 (très court):bolus et infusion continue
Nov 2017 PK/PD and modelling 51
Application to clinical trials (ICU patients)
Nov 2017 PK/PD and modelling 52
De Jongh et al, JAC 2008; 61:382-8
2 g q12 h
LD: 2 g; CI 4g/24h
MIC90
MIC90PK/PD Bkpt 8-16 mg/L
Monte Carlo simulation
MIC90
PK/PD target
Exemple #2 (plus long): patients de soins intensifsavec données manquantes
Nov 2017 PK/PD and modelling 53
Nov 2017 PK/PD and modelling
Temocillin project (full)
20 30
20 30
101
102
101
102
101
102
101
102
ID: 1 ID: 2 ID: 3
ID: 4 ID: 5 ID: 6
ID: 7 ID: 8 ID: 9
ID: 10 ID: 11Con
cent
ratio
n (m
g/L)
Time (h)
54
Nov 2017 PK/PD and modelling
Outputs: individual curves
24 26 28 30 320
100
2002
24 26 28 30 320
50
100
1503
24 26 28 30 320
100
2004
24 26 28 30 320
100
200
5
55
Nov 2017 PK/PD and modelling
Outputs: spaghetti plot (*)
24 25 26 27 28 29 30 31 320
50
100
150
200
250
time
obse
rvat
ions
Total number of subjects: 4Average number of doses per subject: 1Total/Average/Min/Max numbers of observations: 15 3.75 3 4
* not noodles !
56
Nov 2017 PK/PD and modelling
Outputs: population curves
24 26 28 30 32-100
0
100
200
3002
24 26 28 30 32-100
0
100
200
3003
24 26 28 30 32-100
0
100
200
3004
24 26 28 30 32-100
0
100
200
3005
57
Nov 2017 PK/PD and modelling
Outputs: population
24 25 26 27 28 29 30 31 32 33-100
-50
0
50
100
150
200
250
300
TIME
DV
Percentile 90 : Obs.Percentile 50 : Obs.Percentile 10 : Obs.Percentile 90 : 90% CIPercentile 50 : 90% CIPercentile 10 : 90% CIOutliersOutlier (area)
58
Nov 2017 PK/PD and modelling
Outputs: observations vs. predictions
50 100 150 200
50
100
150
200
obse
rvat
ions
predictions
Using the population parameters
50 100 150 200
50
100
150
200
obse
rvat
ions
predictions
Using the individual parameters
59
Nov 2017 PK/PD and modelling
Outputs: residuals
20 30 40-2
0
2
4
PW
RE
S
time
50 100 150 200-2
0
2
predictions
PW
RE
S
20 30 40-2
0
2
4
IWR
ES
time
50 100 150 200-2
0
2
predictions
IWR
ES
20 30 40-2
0
2
4
NP
DE
time
50 100 150 200-2
0
2
predictions
NP
DE
60
Exemple #3 (long): volontaires vs soins intensifset impact de la fraction libre
Nov 2017 PK/PD and modelling 61
Cette partie est reprise du travail de Thèse en coursde Mr Perrin Ngougni-Pokkem
62Nov 2017 PK/PD and modelling
There is growing evidence that standard antibiotic regimens may not provide adequate drug concentrations …
J.W. Mouton et al: Int J Antimicrob Agents. 2002 Apr;19(4):323-31.Roberts et al, Br J Clin Pharmacol. 2012;73:27-36.
63Nov 2017 PK/PD and modelling
Critically ill patients
Pharmacokinetic alteration
A. Abdulla et al: University Medical Center Rotterdam; eposter 069; ECCMID 2017 Hosthoff et al, Swiss Med Wkly. 2016;146:w14368Roberts JA, Lipman J. Clin Pharmacokinetic 2006; 45 (8): 755-73
RRT: renal replacement therapyECMO: extra corporeal membrane oxygenation
Hyperdynamic statesIncreased cardiac out, and clearanceDecreased plasma concentrations
Altered fluid balance / Altered protein bindingIncreased volume of distributionDecreased plasma concentrations
Renal and hepatic impairmentDecreased clearanceIncreased plasma concentrations
Organ support (RRT/ECMO)Increased volume of distribution /clearanceIncreased/decreased plasma concentrations
Critically-ill patients
64Nov 2017 PK/PD and modelling
Consequences of PK alteration
Critically ill patients
Pharmacokinetic alteration
underdosing
Therapeutic failure/antibiotic resistance
overdosingTherapeutic
antibiotic concentration
toxic effects Therapeutic success
A. Abdulla et al: University Medical Center Rotterdam; eposter 069; ECCMID 2017 Hosthoff et al, Swiss Med Wkly. 2016;146:w14368Roberts JA, Lipman J. Clin Pharmacokinetic 2006; 45 (8): 755-73
Variability in antibiotic
concentration
65Nov 2017 PK/PD and modelling
The main objectives
• Current literature data are based mainly on TOTAL temocillin concentrations
• Only the free concentration is active !• Concentration in the infected tissue is important !
Part 1Pilot study
Population Pharmacokinetic Analysis and Protein Binding Characteristics of Free and total
Temocillin concentrations in Plasma of Healthy Volunteers and patients
66Nov 2017 PK/PD and modelling
Design of the in vitro study
Bound concentration vs free concentration of temocillin in plasma
Free Fraction at a given total concentration vs protein concentrations
67Nov 2017 PK/PD and modelling
Temocillin plasma protein binding In vitro study
0 25 50 75 100 125 150 175 200 22505
1015202530354045505560657075
Free fraction vs total concentration of temocillinin plasma for 4 healthy donors (D) compared with
5 patients donors (P) in vitro study
D1D2D3D4
P1P2P3P4P5
total concentration (mg/L)
free
frac
tion(
%)
For the patient donors High free fraction up to 65% Free fraction which increases with the total
concentration High variability between the patient donors.
For the healthy donors, except D4 Low free fraction between 5 to 8% Free fraction which is not influenced by the
total concentration Low variability between the healthy donors
D4: 57.03 g/LD1: 71.75 g/LD2: 84.91 g/LD3: 70.55 g/L
Plasma total protein level (mg/L)Reference range : 65-85g/L
P1: 52.89 g/LP2: 48.34 g/LP3: 61.17 g/LP4: 55.31 g/LP5: 55.53 g/L
68Nov 2017 PK/PD and modelling
Michaelis-Menten fitting of temocillin protein binding
Plasma protein binding of temocillin is saturable
Maximum binding is lower in patients
Bound concentration vs free concentration of temocillin in plasmaIn vitro study
0 20 40 60 80 100 1200
50
100
150
200
250
VolunPatie
Free Concentration (mg/L)
Bou
nd c
once
ntra
tion
(mg/
L)D; n=3 P; n=6
69Nov 2017 PK/PD and modelling
Design of the clinical study: « Phase1 »
8 healthy volunteers.
Single dose of 2g TMO in 40 min infusion; IV administration.
Blood sampling: 40min, 1, 2, 3, 4, 5, 6, 7, 8, 10, 12h.
Study of the relationship between free fraction of temocillin vs its total concentration.
Study of the relationship between bound concentration of temocillin vs free concentration.
Principal Investigator according to Austrian drug lawMarkus Zeitlinger,MDDepartment of clinical Pharmacology,Medical University of Vienna Graph Pad 4 software
70Nov 2017 PK/PD and modelling
Temocillin plasma protein binding
0 50 100 150 200 2500
5
10
15
20
25
V1 = 73.58g/LV2 = 77.45g/LV3 = 66.16g/LV4 = 74.08g/LV5 = 72.45g/LV6 = 64.50g/LV7 = 65.25g/LV8 = 69.45g/L
Plasma total protein levelReference range:
65-85g/L
Mean total concentration (mg/L)
Free
frac
tion
(%)
Low free fraction (3-8%) for total concentrationsbelow 150 mg/L, and increase in free fraction upto 20% for higher total concentrations
Free fraction vs total concentration of TMO in plasma for 8 healthy volunteers (V) in vivo study compared with healthy donors (D)
in vitro study
71Nov 2017 PK/PD and modelling
Michaelis-Menten fitting of temocillin protein binding
Bound concentration vs free concentration of temocillin in plasma
0 20 40 60 80 100 1200
100
200
300V (in vivo study; n=8)
Free concentration (mg/L)
Bou
nd c
once
ntra
tion
(mg/
L)
Protein binding saturation observed
72Nov 2017 PK/PD and modelling
Michaelis-Menten fitting of temocillin protein binding
Bound concentration vs free concentration of temocillin in plasma
0 20 40 60 80 100 1200
100
200
300V (in vivo study; n=8)D (in vitro study; n=3)P (in vitro study; n=6)
Free concentration (mg/L)
Bou
nd c
once
ntra
tion
(mg/
L)
Lower Bmax for patients !
Comparison of plasma protein binding in healthy volunteers (V), healthy donors (D) and patient donors (P)
Similar protein binding saturation observed
73Nov 2017 PK/PD and modelling
PK modeling approaches
“Data-rich” situation Simple to implementIndividual analyzes in descriptive statistics
Subject 1
Subject 2
Subject N
Adapted from I.Delattre. 2012
74Nov 2017 PK/PD and modelling
Plasma total and free concentration versus timePharmacokinetic profile of free and total concentration : individual data
Free concentration decreases with the total concentration Important variability in the pharmacokinetic profiles between
the volunteers
75Nov 2017 PK/PD and modelling
Comparison of pharmacokinetic parameters (free vs total)
Mean and IC95% of pharmacokinetic parameters of free and total TMO
p=0.07
Free Total3
4
5
6
Tem
ocill
in h
alf l
ife (h
)
The half-life of the free temocillin has an important numerical effect; But not significant
P<0.0001
Free Total0
10
20
30
40
50
Tem
ocill
in c
lear
ance
(L/h
)
The clearance of the free temocillin is very high compared to the total
P<0.0001
Free Total0
50
100
150
200
250
300
350
Tem
ocill
in v
olum
e of
dis
trib
utio
n (L
)
The volume of distribution of the free temocillin is very high compared to the total
t1/2 = 0.693 Vd / Cl
76Nov 2017 PK/PD and modelling
1. Structural pharmacokinetic model
This PK profile suggests that the kinetics of the TMO is Bi-compartmental
Intravenous administration of 2 g of TMOPlasma free concentration versus time
0 1 2 3 4 5 6 7 8 9 10 11 121
10
100
1000
Time (h)
log
free
con
cent
ratio
ns (m
g/L)
Distribution andelimination phase
Elimination phase
Visual evaluation of pharmacokinetic profile
77Nov 2017 PK/PD and modelling
1. Goodness-of-fit plotO
bser
ved
free
conc
entr
atio
ns (m
g/L)
Individual predicted free concentrations (mg/L)
Mono compartmental model tested without covariate
Obs
erve
dfr
ee co
ncen
trat
ions
(mg/
L)
Individual predicted free concentrations (mg/L)
Bi-compartmental model tested without covariate
The correlation is better in this case and with less variability
78Nov 2017 PK/PD and modelling
1. Goodness-of-fit plot
Bi-compartmental model + Proportional residual error model
WRE
= (C
pred
-Cob
s) /
Cpr
ed
Individual C predicted (mg/L)
Mono compartmental model tested without covariate
Time (h)
Freq
uenc
y
WRE
Shapiro-Wilk. P=0.002WRE vs. TimeWRE vs. Cpred
Bi-compartmental model tested without covariate
WRE
= (C
pred
-Cob
s) /
Cpr
ed
Individual C predicted (mg/L) Time (h)
Freq
uenc
y
WRE
Shapiro-Wilk. P=0.214WRE vs. TimeWRE vs. Cpred
Residues should be centered on 0 95% of the population residues should
be between approximately -2 and 2 Residue distribution should be normal
79Nov 2017 PK/PD and modelling
2. Covariate model Relevant physiological, biological and demographic parameters that could change
the pharmacokinetic parameters
Make it possible to explain the inter and / or intra-individual variability
Parameter Mean (sd) RangeAge (yr) 32.9 (12.1) 23.0-53.0Weight (kg) 81.9 (10.9) 70.2-105.6Height (m) 1.8 (0.1) 1.7-1.9BMI (kg/m2) 24.4 (2.9) 20.7-28.9GFR (mL/min)(Cockcroft-Gault)
135.7 (16.1) 108.2-153.4
ASAT (U/L) 23.8 (4.5) 14.0-30.0ALAT(U/L) 30.1 (9.2) 18.0-45.0LDH(U/L) 158.8 (28.4) 143.0-195.0Albumin (g/L) Not analyzedTotal protein (g/L) 70.4 (4.4) 64.5-77.5
Influence volume of distribution and clearance
Variable protein binding (-->93 %)
Renal excretion (80% found in the urine in 24h)
80Nov 2017 PK/PD and modelling
Validation population pharmacokinetic final model
Internal Validation: Monte Carlo simulations
• Simulated profiles (n=1000) compared to observed data.
• The observed concentrations should be distributed homogeneously around the median of the simulated concentrations
• Less than 5% of observed concentrations must be outside the 5th and 95th percentiles of the simulated concentrations
External Validation
81Nov 2017 PK/PD and modelling
Internal Validation: Visual Predictive Checks (VPC).
0 1 2 3 4 5 6 7 8 9 10 11 120.1
1
10
100
1000
Simulated free TMO profile
0.95
0.5
0.05
Observed free concentration
Free
con
cent
ratio
n (m
g/L)
Time (h)
82Nov 2017 PK/PD and modelling
Temocillin pharmacodynamic targets
As every β-lactam, temocillin is – bactericidal– time-dependent
(activity is driven by the time during which the drug plasma free concentration remains above the minimum inhibitory concentrations (MIC))
40% of time > MIC is enough for bacteriostatic activity acceptable for non-immunocompromised patients
70% of time > MIC is recommended for immunocompromised patients
100% of time > MIC is suggested For critically-ill patientsthis could not only maximize efficacy but also minimize emergence of resistance
Craig WA Diagn Microbiol Infect Dis. 1995;22:89-96. PMID: 7587056.
Delattre IK et al For submission to Expert Review on Antiinfective Therapy as Special Report
83Nov 2017 PK/PD and modelling
Probability of Target Attainment (PTA) of plasma free temocillin concentrations
For non-immunocompromised patientsTarget: fT > BSAC breakpoint = 8 mg/L of 40% of the time, based on a mean free fraction of 6.0 ± 1.4% (mean of values observed for total concentration < 150mg/L),
PTA
Target concentrations (mg/L)Standard dosing (2g/12h)PTA = 0
newly proposed dosing (2g/8h)
PTA = 0.5
84Nov 2017 PK/PD and modelling
For non-immunocompromised patientsTarget: fT > BSAC breakpoint = 8 mg/L of 40% of the time, based on a mean free fraction of 13.0 ± 4.0% (mean of values observed for total concentration > 150mg/L),
PTA
Target concentrations (mg/L) newly proposed dosing (2g/8h)PTA = 1
Standard dosing (2g/12h)
PTA = 0.99
Probability of Target Attainment (PTA) of plasma free temocillin concentrations
85Nov 2017 PK/PD and modelling
PTA
Target concentrations (mg/L)
For critically-ill patientsTarget: fT > BSAC breakpoint = 8 mg/L of 100% of the time, based on a mean free fraction of 35.0 ± 12.3% (mean of values observed for patient),
Standard dosing (2g/12h)
PTA = 0.7
newly proposed dosing (2g/8h)PTA = 1
Probability of Target Attainment (PTA) of plasma free temocillin concentrations
86Nov 2017 PK/PD and modelling
MIC distributions of E. coli : ESBL/AmpC (n=1155) vs non-ESBL/AmpC (n=1473)
Source: Eumedica (data on file)
Which are the actual (and recent) observations ?