application of pk/pd modeling for optimization of linezolid therapy
DESCRIPTION
Application of PK/PD modeling for optimization of linezolid therapy. Julia Zayezdnaya Zack. Background: MRSA & linezolid. Methicillin Resistant S.aureus (MRSA) is a major nosocomial pathogen that has caused severe morbidity and mortality Linezolid - PowerPoint PPT PresentationTRANSCRIPT
Application of PK/PD modeling for optimization of linezolid
therapy
Julia Zayezdnaya Zack
Background: MRSA & linezolid
Methicillin Resistant S.aureus (MRSA) is a major nosocomial pathogen that has caused severe morbidity and mortality
Linezolid newer antibiotic: first drug of a new class-
oxazalidinone activity against Gram-positive bacteria: used mainly
for MRSA and VRE infections and in patients with hypersensitivity
MOA: binds to the bacterial 50S ribosome subunit and inhibits the initiation of protein synthesis
Goal
To use a PD model based on kill-curves and PK in humans to predict the impact of differing dosage regimens on timecourse of MRSA CFU
To design and validate these predictions using an in vitro PK/PD model
Methods: kill-curve experiments
PD kill-curve experiments: fixed initial inoculum (~107) constant drug concentrations: 0-10XMIC sampling over 24 hours were fit by a PD mixture model
PD mixture model: capacity limited replication 1st order elimination, effect of LZD as a Hill-type model inhibiting
replication
Methods: PD model-Dynamics of Bacterial Growth and Death
Time course of total bacteria growth is a result of a mixture of homogenous sub-populations (mixture model)
Model incorporates bacterial replication modelled as a capacity limited function
1st order rate constant for death Drug effect enhancing bacterial death or inhibiting replication
BacteriaBacteriaCFU/mLCFU/mL
Pop 1Pop 1 Pop2Pop2 Pop3Pop3
KDReplication
IC50IC50
Drug(+)
(-)
Methods: PD model-Dynamics of Bacterial Growth and Death
The differential equation, for each bacterial subpopulation, is as follows:
d CFUi/dt = VGmax·CFUi/[CFUM + CFUTOT] – kd·CFUi
CFUi , CFU/mL of the i th subpopulation Vgmax, maximum velocity of growth (CFU/mL/hr) CFUM, CFU/mL associated with half-maximal growth
CFUTOT, sum total of all subpopulations kd, drug-free 1st-order death rate constant of the bacteria (hr-1) all subpopulations were assumed to share a common VGmax,
CFUM, and kd
Methods: PD model-Dynamics of Bacterial Growth and Death
Drug effect (E) was modelled as a Hill-type function that either decreased bacterial replication or enhanced the 1st order death rate constants, as follows:
E(t) = 1± [Emax·(C/MIC)H]/[SITMiH + (C/MIC)H]
E(t) is multiplied by the replication term or the rate constant for death Emax is the maximum drug effect C/MIC is ~ the inverse serum inhibitory titre (SIT-1) SITMi is the SIT at which E is 50% of the Emax, for the ith subpopulation H is the Hill’s constant (reflects slope) SITMi and initial conditions were allowed to differ between subpopulations
Results: kill-curve experiments
LZD vs. M R S A 0-10xM IC
Hours
0 5 10 15 20 25 30
CF
U1
0Lo
g
0
2
4
6
8
10
12
0.5 x MIC
GC
1 x MIC2 x MIC5 x MIC10 x MIC
Methods: in silico simulations
Two clinical MRSA isolates each with two sub-populationsMIC 2 mg/L: “sensitive” subpopulation SITM
of 0.4 X MIC and “resistant” subpopulation SIT of 3X MIC
MIC 4 mg/L: “sensitive” subpopulation SITM of 0.6 X MIC and “resistant” subpopulation SIT of 6 X MIC
Methods: in silico simulations
Use human PK model to predict concentration profiles and the PD mixture model to predict responses to different dosing regimens: 600 mg PO q12h (BID) 900 mg PO at time 0, followed by 600mg PO q12h
(BIDDL) 600 mg PO q8h (TID) 1200 mg PO at time 0, followed by 600 mg PO q8h
(TIDDL)
Results: in silico predictions
0 10 20 30 40 50 60 70 80 90 100TIME (hr)
3
4
5
6
7
8
9
Lo
g1
0(C
FU
)
MIC_2_TIDDLMIC_2_TIDMIC_2_BIDDLMIC_2_BIDMIC_2_GCMIC_4_TIDDLMIC_4_TIDMIC_4_BIDDLMIC_4_BIDMIC_4_GC
Results: in silico predictions
0 10 20 30 40 50 60 70 80 90 100TIME (hr)
-4
-3
-2
-1
0
Lo
g1
0(D
i ffe
ren
ce f r
om
GC
)
MIC 4 mg/L BID
MIC 2 mg/L BID
MIC 4 mg/L TID
MIC 2 mg/L TID
Results: in silico predictions
0 10 20 30 40 50 60 70 80 90 100TIME (hr)
3
4
5
6
7
8
9
Lo
g1
0(C
FU
)
MIC_2_TIDDLMIC_2_TIDMIC_2_BIDDLMIC_2_BIDMIC_2_GC
Results: in silico predictions
0 10 20 30 40 50 60 70 80 90 100TIME (hr)
4
5
6
7
8
9
Lo
g1
0(C
FU
)
MIC_4_TIDDLMIC_4_TIDMIC_4_BIDDLMIC_4_BIDMIC_4_GC
Methods: in vitro PK/PD model
Bacterial strains: MRSA, MIC 2 and 4 mg/L Drug: linezolid In vitro PK/PD model: series of flasks with multiple
ports for delivery of the drug and media and for removal of waste
Methods: in vitro PK/PD model
What we are simulating:normal volunteer PK parameters—
clearances, volumes, etc.dosing regimens:600 mg PO q12h (BID)
and 600 mg PO q8h (TID)
2
3
4
5
6
7
8
9
0 10 20 30 40 50 60
Time (hr)
Results: in vitro activity
GCs
BID MIC4
TID MIC2
TID MIC 4BID MIC2
Results: in vitro activity
MIC 4 mg/L
MIC 2 mg/L
Inoculum changes over 48 hrs for BID regimen
-4.5
-4
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
12h 24h 40h 48 h
Time (hr)
Ch
ang
e (L
og
10(C
FU
))
Results: in vitro activity
BID MIC 4 mg/L
TID MIC 4 mg/L
TID MIC 2 mg/L
BID MIC 2 mg/L
Inoculum changes over 48 hrs
-4.5
-4
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
12h 24h 40h 48 h
Time (hr)
Ch
ang
e (L
og
10(C
FU
))
Conclusions
In silico and in vitro simulations: traditional regimen is predicted to be ineffective against MRSA with MIC 4 mg/L
Mutant selection phenomenon Predictive value of in silico simulations:
despite deriving from very sparse kill-curve experiments and extrapolating to 96 hrs
Challenges translating these results into biological systems
Future work