Transcript Slide 1

Evaluation of Current Vancomycin Dosing Practices and Pharmacokinetic Neonatal
Infant Models with Therapeutic Drug Monitoring Data from a Pediatric Population
Craig M. Comisar, Bhuvana Jayaraman, Jeffrey S. Barrett
Laboratory for Applied PK/PD, Division of Clinical Pharmacology and Therapeutics, The Children's Hospital of Philadelphia; Philadelphia, PA
OBJECTIVES
BACKGROUND
• Vancomycin General Information:
• A glycopeptide that is the first-line treatment for coagulase-negative Staphylococci and
Staphyloccus aureus infections involved in children.
•To utilize published models to evaluate the clinical performance of the current dosing
recommendations at Children’s Hospital of Philadelphia (CHOP) across a strata of
dosing regimens using simulations of ideally dosed patients.
• Therapeutic drug monitoring (TDM) is used in the clinical setting because high levels
are associated with nephrotoxicity while underdosing can lead to bacterial resistance
and ineffective treatment.
•To determine the suitability of published neonatal vancomycin models to evaluate CHOP
therapeutic drug monitoring (TDM) data.
• The drug is almost entirely renally cleared as parent compound.
SIMULATION RESULTS
• Current clinician driven dosing strategies result in inconsistent achievement of even
the lowest acceptable (5 mg/mL) vancomycin trough levels days into the treatment
regimen (Fig. 2). This is most likely due to a combination of physician underdosing
(47% of doses were under the recommended amount), systematic underdosing
guidance, and inaccessibility of vancomycin pharmacokinetic data.
CHOP Dosing (less than 1000 gm Birth Weight)
150%
100%
Percent of Children Achieving Vancomycin
trough levels of >5 ug/mL
CHOP Dosing (1000-2000 gm Birth Weight)
130%
CHOP Dosing (greater than 2000 gm Birth
Weight)
Median Glomular Filtration Rate for children
>34 weeks Gestational Age
110%
90%
70%
50%
30%
10%
90%
80%
70%
60%
50%
0
100
200
Postnatal Age (days)
300
3
Figure 1: CHOP vancomycin dosing as a function of glomular filtration rate.
8
13
Dose Number
18
Figure 2: Percentage of patients achieving the lowest acceptable
vancomycin trough levels (>5 ug/mL).
• Proposed Alternative Modeling-based Dosing Strategies for Vancomycin Dosing:
• Several empirical models have been proposed to aid dosing of individual neonatal
vancomycin patients using various covariates (weight, age, creatinine concentration,
etc.).
Authors
de Hoog et al.
(1)
Grimsley et
al. (2)
Capparelli et
al. (3)
Anderson et
al. (4)
n # obs Postnatal
108 0-29 days
59
347
0-76 days
374 1103 0-730 days
214 604
1-27 days
Dosing
Covariates in the final model
Weight
Weight, Creatinine Concentration
Weight, Creatinine Concentration
Postnatal Age, Gestational Age
Weight, Creatinine Concentration,
Postmenstrual Age, Presence of a
Ventilator, Inotrope Levels
Current Study 444 1611 1-417 days
Additional demographic information- Weight: 0.45-12 kg, Gender Distribution: 73% male
•Figure
4
shows
the
mean
concentration vs. time profile for a
simulated patient in which the patient
fails to achieve the vancomycin
trough target concentration (8-15
ug/mL)
using
the
Lexicomp
guidelines (5). The models show
very good agreement in trough
predictions though the Anderson et
al. and Capparelli et al. data show
slightly higher trough concentrations.
This is most likely due to those
models ability to account for early
gestational age patients.
•Figure
5
shows
the
mean
concentration vs. time profile for a
simulated patient in which the patient
has
greatly
increasing
serum
creatinine levels representing a loss
of kidney function and lowered drug
clearance. The models provide large
differences in this scenario. DeHoog
et al. don’t include creatinine in their
model and the functional form of
creatinine is different in the other
models resulting in a different
response to the creatinine spike.
Patient #1
Dosing: b.i.d. and t.i.d.
Weight: 3.27 kg
Gestational Age: 38 wks
Postnatal Age:
4-11 days
Creatinine Conc:
0.5-0.6 mg/dL
Grimsley Model
Anderson Model
deHoog Model
25
Capparelli Model
20
Concentration (mg/L)
• Clinicians alter dosing to achieve vancomycin trough concentrations between 8 and 15
mg/mL by the third dose, with adjustments made for low/high trough levels or evidence
of renal impairment (sudden spikes of serum creatinine).
•Figure
3
shows
the
mean
concentration vs. time profile for a
simulated patient in which the patient
is appropriately dosed (trough
concentrations between 8-15 ug/mL)
using the Lexicomp guidelines (5).
The models show very good
agreement in trough predictions.
This is most likely due to a near full
term child and stable creatinine
concentration.
30
15
10
Authors
deHoog et al. (1)
Grimsley et al. (2)
Capparelli et al. (3)
Anderson et al. (4)
Number of
parameters
5
5
12
11
AIC from parameters fixed to AIC from the parameters
published results
optimized for CHOP data
1077
519
503
479
637
478
1661
484
CHOP therapeutic drug monitoring data was fit using the published models. Model fit
was evaluated using the akaike information criterion (AIC), a least squares analysis
adjusted to the number of fitted parameters. When the parameters were fixed, the
Grimsley and Capparelli models fit the CHOP data most precisely. The large
differences in the fits are most likely due to two main factors. TDM data includes
mostly trough vancomycin levels versus a rich data set which could encompass a
more valid pharmacokinetic profile. Additionally the data from the TDM data did not
include any information on inotrope levels or ventilator presence. Postmenstural
age/gestational age was also estimated using other covariates. When parameters
were allowed to optimize, all but deHoog et al. model, which did not include
creatinine concentration, fitted the data equally well.
METHODS
5
0
0
25
50
100
time (hr)
150
Grimsley Model
Patient #2
Dosing: q.d.
Weight: 0.795 kg
Gestational Age: 25 wks
Postnatal Age:
7-14 days
Creatinine Conc:
0.7-1.0 mg/dL
Anderson Model
deHoog Model
20
Concentration (mg/L)
• Dictated by an institution specific version of the Lexi-Comp’s Pediatric Dosing Guide
(5) based primarily on patient postnatal age and weight. Although glomular filtration
rate roughly follows median glomular filtration rate, the range of glomular filtration rates
among children is highly variable (Fig. 1).
Capparelli Model
15
10
•Eight CHOP patients representing every one of the different CHOP dosing regimens
shown in figure 2, were simulated using NONMEM. The patients’ real demographic
information (age, weight, visit history) and serum creatinine levels were used in the
simulations. 1000 simulations were run for each individual using the structural
models and parameters proposed by the authors listed above. Dosing was set to be
the exact dosing schedule recommended by the Lexi-Comp’s Pediatric Dosing
Guide for CHOP (5). Simulations were set to collect data in a one week period of
intravenous vancomycin treatment.
•Simulation results were evaluated to observe deviations from the target trough
serum vancomycin concentrations and to look at simulation performance between
the various published models.
•The structural models were then fit to actual CHOP TDM data using NONMEM using
two scenarios. The first fit fixed parameters to the published results and the second
fit allowed parameters in the various model to optimize to the CHOP data.
5
0
0
60
50
100
time (hr)
150
Grimsley Model
Anderson Model
deHoog Model
Capparelli Model
50
Concentration (mg/L)
• Vancomycin Dosing at Children’s Hospital of Philadelphia (CHOP):
Percent of Adult Dose or Glomular Filtration Rate
FITTED MODEL RESULTS
CONCLUSIONS
Patient #3
Dosing: b.i.d.
Weight: 1.9 kg
Gestational Age: 31 wks
Postnatal Age:
18-25 Days
Creatinine Conc:
1.0-3.0 mg/dL
40
• Current non-modeling based vancomycin dosing practices at CHOP result in poor
achievement of target vancomycin trough concentrations.
• Simulation of vancomycin dosing identified potential underdosing situations in
pediatric patients following Lexi-Comp guidelines.
• When optimized in NONMEM, all models except the deHoog et al. model fit the
vancomycin therapeutic drug monitoring data with equal precision. Published
models parameters should only be used when all covariate information is available.
30
20
10
REFERENCES
0
0
50
100
time (hr)
150
(1) M. de Hoog, R. C. Schoemaker, J. W. Mouton, and J. N. van den Anker, “Vancomycin population pharmacokinetics in neonates”, Clinical Pharmacology &
Therapeutics, 2000, 67(4), 360-367.
(2) C. Grimsley and A. H. Thomson, “Pharmacokinetics and dose requirements of vancomycin in neonates”, Arch Dis Child Fetal Neonatal Ed, 1999,81, F221–F227.
(3) E. V. Capparelli, J. R. Lane, G. L. Romanowski, E. J. McFeely, W. Murray, P. Sousa, C. Kildoo and J. D. Connor, “The influences of renal function and maturation on
vancomycin elimination in newborns and infants”, J. Clin. Pharmacol, 2001, 41, 927-934.
(4) B. J. Anderson, K. Allegaert, J. N. Van den Anker, V. Cossey and N. H. G. Holford, “Vancomycin pharmacokinetics in preterm neonates and the prediction of adult
clearance”, Br J Clin Pharmacol, 2006, 63, 75–84.
(5) Lexi-Comp Online, Pediatric Lexi-Drugs Online, Hudson, Ohio: Lexi-Comp, Inc.; 2004; May 30, 2008. http://www.crlonline.com/crlsql/servlet/crlonline
Dosing
Simulated assuming
complete adherence to
Lexi-Comp guidelines (5)
Simulation Objectives
Therapeutic Drug Monitoring (TDM)
Data from 8 CHOP patients
representing the major different
Lexi-Comp dosing categories (Fig. 2)
• Demographic Information
(Sex, Weight, Postnatal Age, Height, etc.)
• Creatinine Serum Concentrations
• Vancomycin Serum Concentrations
• Test efficacy of Lexi-Comp (5) to
produce required trough
concentrations in a patients
• Evaluate model performance in
different types of patients
• Tested 4 models using NONMEM
• 1000 simulations were run for each
individual
• Simulations were set to collect data
in a one week period
Findings
Simulated assuming
complete adherence to
Lexi-Comp guidelines (5)
Model Fit Objectives
• Compare the “off-the-shelf” utility of
each model (parameters unchanged)
• Compare the CHOP-optimized fit of
each model
• Tested 4 models using NONMEM
• Fixed parameter and CHOP-optimized
fits were run separately
Findings
Actual from clinical
records
Dosing
Actual from clinical
records
•Simulation results were evaluated to observe deviations from the target trough se
concentrations and to look at simulation performance between the various publish
•Fixed parameter model results allowed for comparison of the “off-the-shelf” utility o
•CHOP optimized parameter results allowed for the maximized fit comparison of ea
•Eight CHOP patients representing every one of th
shown in figure 2, were simulated using NONMEM
information (age, weight, visit history) and serum
simulations. 1000 simulations were run for ea
models and parameters proposed by the authors l
the exact dosing schedule recommended by t
Guide for CHOP (5). Simulations were set to co
intravenous vancomycin treatment.
Therapeutic Drug Monitoring (TDM) Data from 8 neonatal CHOP patients
representing the major different Lexi-Comp dosing categories (Fig. 2)
• Demographic Information
(Sex, Weight, Postnatal Age, Height, etc.)
• Creatinine Serum Concentrations
• Vancomycin Serum Concentrations
Fitted Model Objectives
• Compare the “off-the-shelf” utility of each model (parameters
unchanged)
• Compare the CHOP-optimized fit of each model
Dosing
Actual from clinical
records
• Tested 4 models using NONMEM
• Fixed parameter and CHOPoptimized fits were run separately
Findings
• Compared how well each model fit CHOP data by analyzing
the akaike information criterion (lower value indicates better fit)
Simulation Objectives
• Test efficacy of Lexi-Comp (5) to produce target trough
concentrations in patients
• Evaluate model performance in different types of patients
• Tested 4 models using NONMEM
• 1000 simulations were run for
each individual
• Simulations were set to collect
data in a one week period
Dosing
Simulated assuming
complete adherence
to Lexi-Comp
guidelines (5)
Findings
• Simulated vancomycin troughs from each model and
compared results to target trough levels
• Compared the predicted vancomycin levels between the
different models
•Simulation results were evaluated to observe d
serum vancomycin concentrations and to look at
the various published models.
•The structural models were then fit to actual CHOP
two scenarios. The first fit fixed parameters to the
fit allowed parameters in the various model to optim