G
Time (h)2 4 10 12
npde
2
1
0
-1
-2
86
Normalized prediction distribution error (npde) diagnostic plots for unbound TMO:
Population pharmacokinetic modelling of total and unbound temocillin in the plasma of healthy volunteers after intravenous administration
Perrin Ngougni Pokem1, Peter Matzneller2, Beatrix Wulkersdorfer2, Arnaud Capron3, Laure Elens4, Pierre Wallemacq3, Paul M Tulkens1,Markus Zeitlinger2, Françoise Van Bambeke1
1Pharmacologie cellulaire et moléculaire, Louvain Drug Research Institute, Université catholique de Louvain, Brussels, Belgium; 2Department of Clinical Pharmacology, Medical University of Vienna, Austria; 3Clinical Chemistry Department, Cliniques Universitaires St. Luc, Université catholique de Louvain, Brussels, Belgium ; 4Integrated PharmacoMetrics, PharmacoGenomics and PhamacoKinetics, Université catholique de Louvain, Brussels, Belgium;
Temocillin (TMO) is a β-lactam antibiotic activeon Gram (-) bacteria (except most isolates ofPseudomonas aeruginosa [1]), including strainsproducing extended-spectrum β-lactamases(ESBL) and some carbapenemases [2-3]. It istherefore a useful alternative to carbapenemsand is indicated for the treatment of complicatedurinary tract infections (including pyelonephritis),low respiratory infections, bacteremia, and woundinfections [4].
Temocillin is eliminated unchanged by glomerularfiltration. It is highly protein bound (up to 85% [4])and only the unbound concentration isconsidered as potentially active.
The aim of this phase I study was to develop ajoined Population-based (Pop)PK model of totaland unbound TMO plasma concentrations inhealthy volunteers.
Subsequently, this model was used to testdifferent dosing regimens aiming at a 90%probability of target attainment (PTA), i.e.unbound concentrations at least 40% of thedosing interval above the minimal inhibitoryconcentration (40%ƒT>MIC) of susceptibleorganisms [5].
A copy of this poster will be made available after the meeting at
http://www.facm.ucl.ac.be/posters.htm
Background and Aims
Mailing address: Françoise Van Bambekeav. Mounier 73 B1.73.05; 1200 Brussels - [email protected]
Materials & Methods Single-center, open-label, non-randomized study.
8 male healthy volunteers received a single dose of 2 g of TMO as IVinfusion over 40 minutes. Plasma samples were collected from 40 min to12h after the end of infusion.
Plasma concentrations were determined by HPLC-MS/MS for total(sample preparation including protein precipitation with methanol) andunbound (sample preparation including ultrafiltration with Amicon filterUltra-15 device; NMWL 30K; Merck Millipore Ltd) TMO [6].
The PopPK model of total and unbound plasma concentrations werefitted using a non-parametric approach with Pmetrics softwareversion1.4.1 (LAPKB, Los Angeles, CA, USA.). One- and two-compartmental PK models were tested for total TMO plasmaconcentrations; the model that best described the data was selected toderive the TMO unbound concentrations. Final model selection wasbased on the Bayesian information criterion (BIC), goodness-of-fit plots,normalized prediction error (npde) distribution and visual predictive check(VPC).
1000 Monte Carlo simulations per subject receiving TMO 2g q12h1 and2g q8h2 were performed using the final PopPK model. The total andunbound concentrations of TMO were estimated and the 5th, 50th and 95th
percentiles compared to the observed concentrations.
PTA (fT > MIC of 40%) were then computed with 2 different dosingregimens (2g/12h1 or 2g/8h²) assuming either a mean free fraction of 6.0± 1.4% or of 13.0 ± 4.0% being the values observed for totalconcentrations < 150 mg/L or > 150 mg/L, respectively (see results).
1 conventional dose, ² dose registered for severe infections.
TMO shows a bi-compartmental pharmacokinetics. TMO protein binding is high but saturable, with unbound concentrations
ranging from 3 to 20% of total concentrations within the limits of the totalconcentrations observed in volunteers.
PTA is low for conventional dosage, arguing for the use of more frequentadministrations and/or continuous infusion.
Current licensed dosage regimen is suboptimal for MICs> 8 mg/L based on PK in healthy volunteers, due to high proteinbinding in this population. Further studies are needed to evaluate thePK of unbound TMO in target patients’ populations.
Discussion and Conclusions
References1. Chalhoub et al; Sci Rep. 2017;7:40208. PMID:
28091521 2. Livermore. J Antimicrob Chemother. 2006;57:1012-4.
PMID: 16531428 3. Livermore, Tulkens. J Antimicrob Chemother. 2009;
63:243-5. PMID: 190956794. Belgian SmPC, last revision 05/2017; Eumedica (data
on file)5. Glupczynski et al. Eur J Clin Microbiol Infect Dis. 2007;
26:777-83. PMID: 176682536. Ngougni Pokem et al. Clin Biochem. 2015;48:542-5.
PMID: 25712752
0 1 2 3 4 5 6 7 8 9 10 11 12 130
20406080
100120140160180200220240260280300320340360
Simulated total TMO profile
Simulated free TMO profile0.95
0.5
0.05
Observed total concentration
Observed free concentration
Time (h)
Tota
l and
free
TM
O c
once
ntra
tions
(mg/
L)
0.4
0.6
0.8
1.0
0.2
0.000.25 0.5 1 2 4 328 16 64
PTA
Target concentrations (mg/L)
------- 2g/8h2g/12h
Target concentrations (mg/L)
0.4
0.6
0.8
1.0
0.2
0.000.25 0.5 1 2 4 328 16 64
PTA
------- 2g/8h2g/12h
Results
0 50 100 150 200 250 300 3500
50
100
150
200
250
300
350R2 = 87.60% (p<0.0001)Slope = 0.98 (95% CI = 0.90 to 1.07)Inter = -10.20 (95% CI = -20.90 to 0.63)Bias = 4.11Imprecision = 23.9
Total TMO concentrations predicted (mg/L)
Tota
l TM
O c
once
ntra
tions
obs
erve
d (m
g/L)
0 50 100 150 200 250 300 3500
50
100
150
200
250
300
350R2 = 99.30% (p<0.0001)Slope = 1.01 (95% CI = 0.98 to 1.03)Inter = -0.17 (95% CI = -2.45 to 2.11)Bias = 0.02Imprecision = 1.18
Total TMO concentrations predicted (mg/L)
Tota
l TM
O c
once
ntra
tions
obs
erve
d (m
g/L)
A B
Linear regression of individual observed versus predicted total TMOconcentrations using (A) the median model PK parameter values and (B) themedian of the individual Bayesian posterior parameter distributions. The blackdashed lines and the open black circles represent the unity lines and the individualtotal concentrations, respectively.
0 5 10 15 20 25 30 350
5
10
15
20
25
30
35R2 = 86.50% (p<0.0001)Slope = 0.46 (95% CI = 0.42 to 0.50)Inter = 1.50 (95% CI = 0.43 to 2.57)Bias = 9.48Imprecision = 59.00
Unbound TMO concentrations predicted (mg/L)
Unb
ound
TM
O c
once
ntra
tions
obs
erve
d (m
g/L)
0 5 10 15 20 25 30 350
5
10
15
20
25
30
35R2 = 99.% (p<0.0001)Slope = 1.02 (95% CI = 1.00 to 1.04)Inter = -0.10 (95% CI = -0.36 to 0.20)Bias = -0.07Imprecision = 0.87
Unbound TMO concentrations predicted (mg/L)
Unb
ound
TM
O c
once
ntra
tions
obs
erve
d (m
g/L)
Linear regression of Individual observed versus predicted unbound TMOconcentrations using (C) the median model PK parameter values and (D) themedian of the individual Bayesian posterior parameter distributions. The blackdashed lines and the open red circles represent the unity lines and the individualunbound concentrations, respectively.
DC
Bi-compartmental model with linear elimination best fitted total and unbound TMO concentrations.
Equation output of final model: Y(1) = X(1)/V; Y(2) = Y(1)*Km/(Ymax-Y(1));Y(1) and Y(2) (total and unbound TMO concentrations respectively), X; TMO doseV (apparent volume of distribution)
VcC
Rate IV
ke=kei*(GFR/135)kes
KCP
KPC
VpP
o IV: Intravenous o C; P: Central; Peripheral compartmento Vc; Vp: Volume of the compartment o KCP; KPC : Transfer rate o Kd: unbound TMO concentration
corresponding to half of the maximal binding capacity
o Ymax: maximal binding capacityo ke: Elimination constanto kei: typical value of ke; kes: parameter
estimated by the model representing the effect of the GFR on the typical kei value.
o Glomerular filtration rate (GFR)o 135: mean value of the GFR of study pop.
Plasma unbound
concentrationKd
Ymax
Q-Q plot (E), Histogram with expected normal distributions indicated by thedashed lines (mean) and light blue boxes (CI95% ranges) (F), npde with respect topost-intake time (G), and to predicted total TMO concentrations (H) with individualobservations (blue dots), observed 5, 50, 95th percentiles (blue and red lines),predicted percentiles (dashed black lines) and CI 95% (blue and red ribbons).
Q-Q plot (I), Histogram with expected normal distributions indicated by the dashedlines (mean) and light blue boxes (CI95% ranges) (J), npde with respect to post-intake time (K), and predicted unbound TMO concentrations (L) with individualobservations (blue dot), observed 5, 50, 95th percentiles (blue and red lines),predicted percentiles (dashed black lines) and CI 95% (blue and red ribbons).
Theo
retic
al q
uant
iles
Sample quantiles (npde)
Q-Q plot versus N(0,1) for npde
-3
3
-2
-1
0
1
2
-3 3-2 -1 0 1 2
IM
N
PTA plots for standard dosing (2g/12h)versus higher dosing (2g/8h) considering astarget fT > MIC during 40% of the time, based(M) on a mean unbound fraction of 6.0 ±1.4% (mean of values observed for totalconcentration < 150mg/L),(N) on a mean unbound fraction of 13.0 ±4.0% (mean of values observed for totalconcentration > 150mg/L).
The probability of reaching a target of 8mg/L depends on TMO protein bindingfor both dosing options.
VPC of simulated (5th, 50th and95th percentiles) of total (solidblack lines) and unboundTMO concentrations (solidred lines) represented vstime. The individual values ofobserved concentrations areshown as black (total) or red(unbound) circles.
Less than 5 % of observedconcentrations are outsidethe 5th and 95th percentiles ofsimulated concentrations.
Pharmacokinetic parameters of the final model.Parameters Mean SD CV% Mediankei (h-1) 0.24 0.06 26.68 0.22kes 0.18 0.13 76.94 0.14KCP 0.35 0.21 60.93 0.25KPC 0.57 0.23 41.20 0.54V (L) 7.70 1.78 23.21 7.61Ymax (mg/L) 346.55 132.2 38.14 314.94Kd (mg/L) 14.63 9.60 65.59 12.11kei (typical value of ke) and kes (parameter estimated by the model representing the effect of the GFR on the typical kei value), KCP and KPC (transfer rate between the compartments), V (apparent volume of distribution), Ymax (maximal binding capacity), Kd (free TMO concentration corresponding to half maximal binding), SD (standard deviation), CV (coefficient of variation).
Schematic overview of the mechanism based final PK model for TMO in healthy volunteers
Probability of target attainment (PTA)
Plasma free fraction vs total TMO concentrations
0 50 100 150 200 2500
5
10
15
20
25HV1HV2HV3HV4HV5HV6HV7HV8
Total TMO concentrations (mg/L)
Unb
ound
frac
tion
(%)
Mean: 6%
Mean: 13%
Goodness-of-fit plot for total temocillin (TMO): Goodness-of-fit plot for unbound temocillin (TMO):
E
Theo
retic
al q
uant
iles
Sample quantiles (npde)
Q-Q plot versus N(0,1) for npde
-3
3
-2
-1
0
1
2
-3 3-2 -1 0 1 2
F
Sample quantiles (npde)
0
5
10
15
20
Freq
uenc
y
-2 -1 0 1 2
H
Predicted total concentrations (mg/L)50 100 150 200
npde
2
1
0
-1
-2
J
Sample quantiles (npde)
0
5
10
15
20
Freq
uenc
y
-2 -1 0 1 2
L
Predicted unbound concentrations (mg/L)10 20 30 40
npde
2
1
0
-1
-2
0
K
Time (h)2 4 12
npde
2
1
0
-1
-2
86 10
Normalized prediction distribution error (npde) diagnostic plots for total TMO:
Visual predictive check (VPC):
P 2220
Plasma bound
concentration
Covariate: GFR was retained as covariate in the final model (𝚫𝚫 -2LL = 138).
Error model: Polynomial error and proportional error (Gamma factor).
SDtotal = 0.19*Y(1) + 1.17*Y(1)2
SDunbound = 0.0018 + 0.24*Y(2) + 0.11*Y(2)2
Gamma error factor for the final model was 0.32.SD (standard deviation). Final model converged after 1171 cycles
This study was approved by the ethical committee of the Medizinische Universität Wien (Eudra CT 2015-003457-18)
AcknowledgmentsThe present work was performed without specific financial support. P.N.P., L.E., P.W. areemployees of the Université catholique de Louvain, F.V.B. is Research Director of the BelgianFonds de la Recherche Scientifique, P.M.T. is emeritus professor and unpaid collaborator.
