Pediatric Pharmacology

Population Pharmacokinetics of Milrinone The Journal of Clinical Pharmacology

2019, 0(0) 1–14
in Infants, Children, and Adolescents 
C 2019, The American College of

Clinical Pharmacology
DOI: 10.1002/jcph.1499

Christoph P. Hornik, MD, MPH1 , Ram Yogev, MD2 , Peter M. Mourani, MD3 ,
Kevin M. Watt, MD, PhD1 , Janice E. Sullivan, MD4 , Andrew M. Atz, MD5 ,
David Speicher, MD6 , Amira Al-Uzri, MD7 , Michelle Adu-Darko, MD8 ,
Elizabeth H. Payne, PhD9 , Casey E. Gelber, MS9 , Susan Lin, MPH9 , Barrie Harper, MT1 ,
Chiara Melloni, MD1 , Michael Cohen-Wolkowiez, MD, PhD1 , Daniel Gonzalez, PharmD,
PhD10 , and Best Pharmaceuticals for Children Act–Pediatric Trials Network Steering
Committee∗

Abstract
Milrinone is a type 3 phosphodiesterase inhibitor used to improve cardiac output in critically ill infants and children. Milrinone is primarily excreted
unchanged in the urine, raising concerns for toxic accumulation in the setting of renal dysfunction of critical illness. We developed a population
pharmacokinetic model of milrinone using nonlinear mixed-effects modeling in NONMEM to perform dose-exposure simulations in children with
variable renal function. We included children aged <21 years who received intravenous milrinone per clinical care. Plasma milrinone concentrations
were measured using a validated liquid chromatography–tandem mass spectrometry assay (range 1-5000 ng/mL). We performed dose-exposure
simulations targeting steady-state therapeutic concentrations of 100-300 ng/mL previously established in adults and children with cardiac dysfunction.
We simulated concentrations over 48 hours in typical subjects with decreasing creatinine clearance (CrCl), estimated using the updated bedside
Schwartz equation. Seventy-four patients contributed 111 plasma samples (concentration range, 4-634 ng/mL). The median (range) postmenstrual
age (PMA) was 3.7 years (0-18), and median weight (WT) was 13.1 kg (2.6-157.7). The median serum creatinine and CrCl were 0.5 mg/dL (0.1-
3.1) and 117.2 mL/min/1.73 m2 (13.1-261.3), respectively. A 1-compartment model characterized the pharmacokinetic data well. The final model
parameterization was: Clearance (L/h) = 15.9*(WT [kg] / 70)0.75 * (PMA1.12 / (67.71.12 +PMA1.12 )*(CrCl / 117)0.522 ; and Volume of Distribution (L) =
32.2*(WT [kg] / 70). A loading dose of 50 µg/kg followed by a continuous infusion of 0.5 µg/kg/min resulted in therapeutic concentrations, except
when CrCl was severely impaired at 30 mL/min/1.73 m2 . In this setting, a 25 µg/kg loading dose and 0.25 µg/kg/min continuous infusion resulted in
therapeutic exposures.

Keywords
children, creatinine clearance, infants, milrinone, pharmacokinetics

Milrinone is a type 3 phosphodiesterase inhibitor Milrinone is eliminated via the kidneys, with up to
approved by the US Food and Drug Administra- 90% of the drug recovered in the urine after 8 hours,
tion (FDA) for up to 48 hours of intravenous (IV)
treatment in adults with acute decompensated heart 1 Duke Clinical Research Institute, Duke University School of Medicine,
failure.1 Milrinone is not approved for use in patients Durham, NC, USA
2 Ann and Robert H. Lurie Children’s Hospital of Chicago, Chicago,
under 18 years of age, but is frequently adminis-
IL, USA
tered off-label to infants and children for the treat- 3 The Children’s Hospital Colorado, Aurora, CO, USA
ment of pulmonary hypertension, low cardiac output, 4 University of Louisville Norton Children’s Hospital, Louisville, KY, USA
and shock with vasoconstriction.2–5 Phosphodiesterase 5 Medical University of South Carolina Children’s Hospital, Charleston,

inhibition by milrinone raises intracellular cyclic SC, USA

6 Rainbow Babies and Children’s Hospital, Cleveland, OH, USA
adenosine monophosphate levels, which increases in- 7 Oregon Health and Science University, Portland, OR, USA
tracellular calcium concentrations and promotes con- 8 University of Virginia Children’s Hospital, Charlottesville, VA, USA
tractile protein phosphorylation, resulting in improved 9 The EMMES Corporation, Rockville, MD, USA
systolic cardiac contractility, diastolic cardiac relax- 10 Division of Pharmacotherapy and Experimental Therapeutics, UNC

ation, and vascular relaxation.6,7 Due to a lack of direct Eshelman School of Pharmacy, The University of North Carolina at
adrenoreceptor stimulation, milrinone does not raise Chapel Hill, Chapel Hill, NC, USA
myocardial oxygen consumption, a particularly favor- Submitted for publication 4 April 2019; accepted 3 July 2019.
able property in the setting of impaired myocardial Corresponding Author:
oxygen delivery. Christoph P. Hornik, MD, MPH, Duke Clinical Research Institute, 2400
In adults, milrinone displays 70% plasma protein Pratt Street, Durham, NC 27710
binding, predominantly to albumin, and has a volume Email: christoph.hornik@duke.edu
of distribution (V) of approximately 30 L/70 kg.8,9 ∗ See Acknowledgments for listing of committee members.
2 The Journal of Clinical Pharmacology / Vol 0 No 0 2019

predominantly in the parent form (80%) or as an O- milrinone CL. We then leveraged the model to conduct
glucuronidated metabolite (10%), which is not known simulations across the pediatric age continuum, high-
to be pharmacologically active. Milrinone elimination lighting the need for dose reductions in the setting of
clearance (CL) is 300 mL/min in adults, indicative renal impairment.
of active tubular secretion, though the specific trans-
porters involved are not known. Elimination CL is Methods
significantly reduced in adults with moderate to severe The study protocol was reviewed and approved by
renal impairment and an estimated creatinine clear- the institutional review boards of Duke University
ance (CrCl) between 30 and 80 mL/min, including (coordinating center) and all participating sites. All
those with congestive heart failure.10–12 Plasma con- participants and participants’ parents/legal guardians
centrations associated with therapeutic effects in adults provided written informed consent or assent, as
range from 100 to 300 ng/mL.13,14 Higher concen- applicable.
trations (>500 ng/mL) are associated with toxicity,
primarily systemic hypotension resulting from excessive Pediatric Patient Population
vasodilation.13 To limit the risk of toxic accumulation PK samples used to develop the PopPK model de-
in the setting of impaired renal function and decreased scribed in this report were collected through the Pe-
milrinone elimination, the product label recommends diatric Trials Network’s (PTN) Pharmacokinetics of
reducing the continuous infusion rate by up to 66% in Understudied Drugs Administered to Children per
adults based on estimated CrCl.1 Standard of Care trial (PTN POPS; ClinicalTrials.gov
Milrinone pharmacokinetics (PK) have been previ- no. NCT01431326; protocol: NICHD-2011-POP01).
ously studied in pediatric patients, including premature The PTN POPS trial is a multicenter, prospective,
infants and those with congenital heart disease.5,13,15–17 PK, and safety study of understudied drugs adminis-
Population PK (PopPK) models have been developed tered to children (<21 years of age) per standard of
in different pediatric populations with both weight and care. Subjects who received IV milrinone per standard
age as covariates relevant for characterizing CL. The of care as administered by their treating caregiver were
effects of variable renal function on milrinone CL in eligible for enrollment. Exclusion criteria included fail-
children are complex, involving an interplay between ure to obtain consent or assent, or a known pregnancy
maturation of organ function, underlying disease pro- as determined by interview or testing. Standard-of-
cesses, therapeutic benefits of milrinone, and milrinone care laboratory assessments (eg, basic metabolic panel
toxicity, all affecting renal CL.17–19 In a retrospective including serum creatinine) were recorded if collected
analysis of therapeutic drug monitoring data from a within 72 hours of a study dose of the drug. Gestational
single center, a significant correlation between esti- age was documented in infants with a postnatal age
mated CrCl and milrinone CL was observed, a PopPK (PNA) of <120 days, and height and weight were
model was developed, and the need for dose individual- measured around the time of the first PK sample
ization was highlighted.18 While dose individualization collection. Subjects were enrolled in the study for up to
has the potential to improve therapeutic outcomes of 90 days.
milrinone, turnaround times of quantification assays
Drug Dosing and Sample Collection
may prohibit its widespread implementation in several
All milrinone doses were administered via the IV route,
settings of care. More recently, a PopPK model in-
but the dose amount was per standard of care. Dosing
corporating renal dysfunction as measured by Kidney
information was collected for up to 8 doses before the
Disease Improving Global Outcomes (KDIGO) criteria
sampling dose (last dose before first biological sample
was developed in infants <12 months of age receiv-
collection). PK samples were collected optimally with
ing milrinone after cardiac surgery.20 In this patient
standard-of-care lab collections or at different times
population, the authors recommended lower milrinone
from standard-of-care collections if allowed per con-
infusion rates in infants with KDIGO stage 3 renal in-
sent (Table S1). Because this was a standard-of-care
sufficiency based on their model-derived dosing simula-
study, dosing and PK sample collection times varied
tions. A PopPK model that incorporates renal function
among subjects. Actual sampling time points were used
parameters across a broader pediatric age range could
in the PK analysis.
be leveraged to perform similar dosing simulations that
may a priori inform milrinone dosing in children with Analytical Methods
variable age and renal function. Blood was collected (0.5 mL in subjects <1 year
In this PK analysis, we used opportunistically col- of age and 3 mL in subjects 1 year of age) in
lected samples from infants and children covering a an ethylenediaminetetraacetic acid–K2 microtainer or
wide range of ages and renal function to develop a vacutainer and was processed into plasma immedi-
PopPK model and characterize the effect of CrCl on ately before freezing (–70°C) at the study sites. PK
Hornik et al 3

samples remained frozen (–70°C) and were sent to where PARij denotes the estimate of parameter j in the
a PTN-contracted commercial laboratory (Frontage ith individual, θ Pop,j is the population value for param-
Laboratories, Exton, Pennsylvania) for storage and eter j, and ηij denotes the deviation from the average
analysis. Stability data confirmed milrinone stability at population value for parameter j in the ith individual
storage conditions and for the entire storage duration with mean zero and variance ω2 . The apparent IIV was
(up to 24 months). Milrinone concentrations were converted to percent coefficients of variation by taking
quantified using validated liquid chromatography– the square root of ω2 and multiplying it by 100.
tandem spectrometry assays.13 Reversed-phase high- Proportional, additive, and combined (additive plus
performance liquid chromatography separation was proportional) residual error models were explored.
achieved with a Phenomenex Synergi, Polar-RP 80A
column (50 × 2.0 mm, 4 micron) (Phenomex, Torrance, Covariate Analysis
California). The total validation range for the assay was Actual body weight (WT) was assumed to be a
1 to 3800 ng/mL. Accuracy and precision assessed were significant covariate for CL and V, and was included in
within the FDA’s bioanalytical assay validation criteria the base model. The relationship between WT and PK
(eg, ±15%). Additional details of the bioanalytical parameters was characterized using an allometric rela-
method are provided in Table S2. tionship for CL and V parameters (scaled to a 70-kg
standardized WT), according to equations 2 and 3:
Statistical Analysis  
W Ti θW T,C L
Using the value at the time of first recorded dose C L = C L std ∗ (2)
(bolus of continuous infusion), the median and range 70 kg
were calculated and presented for demographic and
dosing variables. Counts and percentages are calculated  θW T,V
W Ti
and presented for categorical variables. Distributions V = Vstd ∗ (3)
70 kg
of study variables are compared using nonparametric
Kruskal-Wallis, chi-square, or Fisher’s exact tests where where CLstd and Vstd represent population estimates
appropriate. The distributions of individual empirical of CL and V in a 70-kg adult and WTi denotes
Bayesian estimates calculated from the final PopPK WT for the ith subject, and θW T,C L and θW T,V denote
model were compared across covariates using Wilcoxon the exponents for the effect of weight on CL and V
rank-sum tests. With the exception of the PK modeling, parameters, respectively. Estimation of these exponents
all statistical analyses were performed using Stata (ver- was compared to using fixed values of 0.75 and 1 for
sion 14.2, College Station, Texas). θW T,C L and θW T,V , respectively. Scaling to 70-kg WT
was chosen to facilitate comparisons with adult PK
PopPK Model Development parameters and to comply with efforts to standardize
Milrinone plasma PK data collected after IV admin- the reporting of pediatric PK parameters.24
istration were analyzed with a nonlinear mixed effects After accounting for differences in body size
modeling approach using the software NONMEM using WT, additional covariates were tested for model
(version 7.4.3, Icon Development Solutions, Ellicott inclusion. Determination of which covariates to test for
City, Maryland). The first-order conditional estimation model inclusion was based on physiological relevance
method with eta-epsilon interaction was used for all and by visual inspection of scatter and box plots
model runs. Run management was performed using (continuous and categorical variables, respectively;
Pirana (version 2.8.2).21 Bootstrap analyses were per- Figure S1) of the individual deviations from the
formed with Perl-speaks-NONMEM (version 3.7.6).22 population-typical value PK parameters against
Data manipulation and visualization were performed covariates. The following covariates were explored:
using the packages Xpose in the software R (version sex, race, ethnicity, obesity (defined as body mass
3.0.3, R Foundation for Statistical Computing, Vienna, index 95th percentile for age and sex in children 2
Austria), RStudio (version 0.97.551, RStudio, Boston, years of age), postmenstrual age (PMA) and PNA,
Massachusetts), and Stata (version 14.2, College Sta- extracorporeal membrane oxygenation (ECMO) at
tion, Texas).23 any time during PK sampling, indication, surgery
One-, 2-, and 3-compartment PK models with as- history, serum creatinine, and CrCl. To estimate CrCl
sumed linear PK were explored.19 Interindividual vari- we used the updated bedside Schwartz equation.25–29
ability (IIV) was assessed for PK model parameters With the exception of WT, which was standardized to
using an exponential relationship (equation 1). a 70-kg body weight, all other continuous covariates
were normalized to the population median value of
 
P A Ri j = θ Pop, j ∗ exp ηi j (1) the trial population. When covariates were missing at a
specific time point, the last available value was carried
4 The Journal of Clinical Pharmacology / Vol 0 No 0 2019

forward. When covariates were entirely missing, the were normalized using the median prediction for each
median value from the study population was imputed, bin across time after dose.32 The dosing and covariate
except for gestational age, where a value of 40 weeks values used to generate the simulations in the VPC were
(term infant) was imputed when missing. the same as those in the study population. We chose a
The relationship between age and CL was evaluated pcVPC, rather than standard VPC, because of its ability
using linear, power, and sigmoidal Emax maturation to remove variability resulting from binning across
functions (equations 4–6). Maturation function param- independent variables, a concern in an opportunis-
eters were both estimated and fixed to values as pub- tic PK data set with standard-of-care drug dosing.32
lished by Mizuno et al,20 Mahmood and Staschen,30 The number of observed concentrations outside of
and Rhodin et al.31 As a measure of age, PNA and the 90% prediction interval for each time point was
PMA were explored. quantified.

Fage = Coefficient ∗ age + intercept (4)

Dosing Simulations
Fage = age Coefficient
(5) Plasma concentrations of milrinone were simulated in
virtual subjects with variable renal function as defined
ageHILL by the CrCl. A total of 4 virtual subjects were created
Fage = (6) (subject 1: weight 3 kg, PMA 40 weeks; subject 2: weight
T M50 HILL + ageHILL
13 kg, PMA 144 weeks [PNA 2 years]; subject 3: weight
The relationship between CrCl and CL was eval- 40 kg, PMA 664 weeks [PNA 12 years]; subject 4:
uated using linear and power relationships. For di- weight 70 kg, PMA 1600 weeks [PNA 30 years]). Within
chotomous categorical covariates (eg, ECMO), a power each subject, estimated CrCl was imputed at 4 differ-
relationship was used (equation 7). ent levels—30, 60, 120, and 160 mL/min/1.73 m2 —
thereby creating a total of 16 virtual subjects. These sce-
Fcov = coefficient covaraite (7) narios were chosen to represent and compare weights
and renal function of infants, children, and adolescents
For categorical variables with n distinct values (race, who may be exposed to milrinone in clinical practice,
ethnicity), additive coding with n-1 indicator variables and were within the range of observed values in our
(1 = yes, 0 = no) was used (equation 8). study cohort. The PNA and subsequently calculated
PMA were chosen to represent approximately the 50th
Fcov = Coefficient1 ∗ Covariate level1 + · · · percentile on the Centers for Disease Control and
+Coefficientn−1 ∗ Covariate leveln−1 (8) Prevention weight-for-age percentiles. For each virtual
subject, 2 dosing scenarios frequently used in clini-
A forward inclusion (P < .05 and change in objective cal practice were evaluated: an IV loading dose of
function value [OFV] >3.8) and backward elimination 25 µg/kg/min administered over 30 minutes followed
(P < .01 and change in OFV >6.6) approach was by a continuous IV infusion of 0.25 µg/kg/min for
used to evaluate statistical significance in the covariate 48 hours, and an IV loading dose of 50 µg/kg/min
analysis (Table S3). administered over 30 minutes followed by a continuous
IV infusion of 0.5 µg/kg/min for 48 hours. Starting
PopPK Model Evaluation with the completion of the loading dose and at regular
Standard model diagnostic methods were used and in- intervals up to 48 hours, we simulated individual pre-
cluded successful minimization, diagnostic plots, plau- dicted plasma concentrations at 10 time points. A total
sibility, and precision of parameter estimates, as well of 1000 simulations were performed for each virtual
as OFV and shrinkage values. Generated diagnostic subject for each dosing scenario, and the median values
plots included individual predictions and population are presented in concentration time curves plotted over
predictions vs observations, and conditional weighted the plasma milrinone concentration range associated
residuals vs population predictions and time after first with efficacy per the FDA label and prior publications:
dose. 100 to 300 ng/mL.1,13,14 To illustrate the effect of IIV
Parameter precision for the final PopPK model was on simulated milrinone exposures, plasma concentra-
evaluated using nonparametric bootstrapping (1000 tions at steady state (Css ) during continuous infusion,
replicates) to generate the 95% confidence intervals calculated as Css = infusion rate / CL, were determined
for parameter estimates. A prediction-corrected visual and plotted as box plots stratified by virtual subject
predictive check (pcVPC) was performed using the final WT, estimated CrCl, and dosing regimen. Simulations
model (N = 1000 simulations). To generate the pcVPC were also repeated for a virtual adult-size (70-kg body
plots, both the simulated and observed concentrations weight) subject (Figure S2).
Hornik et al 5

Table 1. Clinical Data From 74 Subjects in 1 subject at >1800 hours after the first milrinone
Median (Range) or dose.
Characteristic N (%)
PopPK Model Development and Evaluation
Age, years 2.9 (0.01-18) A 1-compartment model characterized the milrinone
28 days 13 (18)
>28 days to 2 y 19 (26)
concentration-versus-time data well (Figures 1 and 2).
>2 years to 12 y 29 (39) For the base model, we included allometrically scaled
>12 y 13 (18) WT, standardized to a 70-kg WT, with fixed allometric
Gestational age, wk 39 (37-41) exponents for both CL at 0.75 and V at 1. Estimation
Postmenstrual age, wk 192 (37-977) of the allometric exponents was also performed but did
Weight, kg 13.1 (2.6-157.7)
Male 39 (53)
not improve model fit or significantly reduce OFV com-
Race pared to inclusion of other covariates as outlined below
White 52 (70) (OFV change with estimated allometric exponents: –
Black 17 (23) 0.02). The IIV for V could not be accurately estimated
Unknown/not reported 4 (5) due to excessively high shrinkage values (>50%); there-
Asian 1 (1)
Ethnicity
fore, this random-effect parameter was fixed to zero in
Hispanic or Latino 12 (16) both the base and final models.
Not Hispanic or Latino 62 (84) After accounting for body size, use of Emax mat-
Serum creatinine, mg/dL 0.5 (0.1-3.1) uration functions for PMA on CL most significantly
Creatinine clearance, mL/min/1.73 m2a 117.2 (13.1-261.3) reduced the OFV (–24.02), followed by inclusion of
30 mL/min/1.73 m2 4 (5)
>30 to 60 mL/min/1.73 m2 9 (12)
estimated CrCl as a covariate on CL using a power
>60 to 120 mL/min/1.73 m2 32 (43) relationship (–22.97). In the multivariable step, addition
>120 mL/min/1.73 m2 29 (39) of CrCl as a covariate to a model containing the
On ECMO 17 (23) Emax maturation function further significantly reduced
Obese 11 (15) the OFV (–10.77). In the backward elimination step,
Infusion rate (µg/kg/min) 0.5 (0.1-41)
Number of pharmacokinetic samples per subject 1 (1-11)
removal of either covariate increased the OFV by >6.6.
As a result, the final model included WT, a maturational
ECMO, extracoporeal membrane oxygenation. function that characterizes the relationship between CL
a
Calculated using the modified bedside Schwartz equation. and PMA, and a power relationship that characterized
the relationship between CL and CrCl (Table 2 and
Table S3):
Safety Analysis
In the PTN POPS study, only adverse events associated r CL(L/h) = (15.9)*(WT[kg]/70)0.75 *(PMA
with study procedures (phlebotomy) and suspected [weeks]1.12 /(67.71.12 +PMA[weeks]1.12 ))*(CrCl
unexpected serious adverse reactions were recorded. [mL/min/1.73 m2 ] / 117)0.522
r V (L) = (32.2)*(WT [kg] / 70)
Results
Patient Characteristics A proportional error model with a value of 32%
A total of 74 pediatric patients contributed 111 evalu- characterized the residual error well. The shrinkage
able plasma samples for PK analysis (Table 1 and Fig- estimate for the parameter estimating the IIV in CL
ure S3). All subjects received IV milrinone at a median was 6%. Shrinkage for the proportional residual error
infusion rate of 0.5 µg/kg/min (range, 0.1-41 µg/kg/min) parameter was 27%. The percentage difference between
for a median duration of 12 hours (range, 0.02-2647 model and bootstrapped median parameter estimates
hours). The median number of dose adjustments per was 5% for all parameters.
pediatric patient, including bolus doses, defined as Diagnostic plots for the final PopPK model showed
infusions of 1 hour duration, or changes in infusion underprediction at higherconcentrations (400 ng/mL)
rate, was 2.5 (range, 1-17). Renal function as assessed by (Figure 1). A total of 6 subjects had conditional
serum creatinine and estimated CrCl varied across sub- weighted residuals >2. These subjects ranged in WT
jects (Table 1). From the 125 plasma PK samples drawn, from 3 to 16 kg and PMA from 0.7 to 5.4 years, 1 was
a total of 14 (11%) were excluded for the following rea- supported with ECMO, and none were obese. Eight
sons: 9 due to dilutions greater than the validated range percent of observed concentrations were outside of
(50× dilution); 3 were below the limit of quantification, the pcVPC 90% prediction interval, indicating slight
were the only sample obtained in 3 subjects, and were overestimation of the IIV by the final PopPK model
not imputed given their low number; and 2 were sig- (Figure 2 and Figure S4). Results were similar when the
nificant outliers and were the only 2 samples obtained final model parameters were reestimated using all of the
6 The Journal of Clinical Pharmacology / Vol 0 No 0 2019

Figure 1. Final population pharmacokinetic model goodness-of-fit plots: observed concentrations vs population predictions (A) and individual
predictions (B), and conditional-weighted residuals against time after first dose (C) and population predictions (D). The solid black line represents the
line of unity, the dashed line is a linear regression line (A and B) or a Loess smoother (C and D).

available PK samples (with the exception of those below

the quantification limit and those measured outside of
the validated dilution range). Population estimates of
CL (17.1 L/h/70 kg) and V (29.5L/70 kg) were within
10% of values obtained from the final model.
Individual empirical Bayesian estimates of milri-
none half-life were compared between subjects with
variable estimated CrCl and age (Figure 3 and Ta-
ble S4). As expected, the distribution differed signif-
icantly between groups of estimated CrCl (P < .001
from a nonparametric Kruskal-Wallis test).

Simulated Dose-Exposure Relationship

A loading dose of 50 µg/kg resulted in milrinone plasma
concentrations at the end of the load within the thera-
peutic range of 100 to 300 ng/mL, except for the 40-kg
adolescent (12 years of age) with an above-normal
Figure 2. Prediction corrected visual predictive check for the final
CrCl of 160 mL/min/1.73 m2 (Figure 4). Following ad-
population pharmacokinetic model. Seven percent of the observations
are outside of the 90% prediction interval. The shaded area represents ministration of a 50-µg/kg loading dose, a continuous
the 90% prediction interval, the dashed line the median predicted infusion of 50 µg/kg/min resulted in therapeutic plasma
concentrations, and the solid line the median observed concentrations. concentrations in the virtual 3-kg newborn, 13-kg child
Hornik et al 7

Table 2. Parameter Estimates and Bootstrap Results for the Final Safety
Population Pharmacokinetic Model No study procedure-related adverse events or serious
Bootstrap Median adverse events were reported in the 74 subjects enrolled
Parameters Estimate (RSE) (5th, 95th percentile) in the PTN POPS study exposed to milrinone.
CL70kg (L/h) 15.9 (23%) 15.9 (12.2, 33.5)
V70kg (L) 32.2 (30%) 31.6 (12.4, 153.9)
Exponent for power function 0.522 (27%) 0.54 (0.30, 0.82)
Discussion
characterizing the effect of We developed a pediatric PopPK model using milrinone
CrCl on CL concentrations collected opportunistically following IV
TM50 (weeks PMA) 67.7 (64%) 68.4 (39.5, 1158.6)
Hill coefficient 1.12 (70%) 1.1 (0.44, 24.7)
administration in a group of children with a wide range
IIV CL (CV%) 70 (27%) 67 (48, 83) of age and renal function. Similar to some previously
Residual proportional error (%) 32 (29%) 26 (20, 37) published PopPK models of milrinone in adults and
children, a 1-compartment model best characterized
CL70kg , elimination clearance standardized to a 70-kg body weight; CrCl,
creatinine clearance; CV, coefficient of variation; IIV, interindividual variability;
the data.9,15,16 Our model incorporates the effect of
PMA, postmenstrual age in weeks; RSE, relative standard error; TM50 , PMA at PMA and estimated CrCl on milrinone CL. This fea-
which 50% of adult CL is reached; V70kg , volume of distribution standardized ture makes it particularly applicable to pediatric clinical
to a 70-kg body weight. practice by simulating optimal milrinone dosing for
Over 95% of the 1000 bootstrap runs minimized successfully. children with variable renal function.
Our population estimate for milrinone V scaled to a
70-kg WT (32.2 L) is comparable to values previously
reported in adults, in the FDA drug label (26.6 L/
70 kg), and in a pediatric population 0 to 6 years of
age after cardiac surgery using a 1-compartment PK
model (V = 33.7 L/70 kg).1,15 Interestingly, neither our
study nor any prior published PopPK study of milri-
none has, to our knowledge, identified any significant
covariates affecting V estimates beyond WT or related
anthropomorphic measurements such as body surface
area. Plasma concentrations of albumin in particular
could affect estimates of V given the moderate degree of
binding of milrinone to this protein.8,9 Unfortunately,
Figure 3. Empirical Bayesian estimates of milrinone half-life by esti-
we did not have access to plasma albumin concentra-
mated creatinine clearance (CrCl).Outer limits of the boxes are 25th and tions in these opportunistically collected data to assess
75th percentiles, vertical lines inside the box the 50th percentile, while its effect on milrinone V. Our estimate of CL scaled to
whiskers are 1.5 times the interquartile range. Outlier values are shown. a 70-kg WT (15.9 L/h) falls within the wide range of
adult estimates (18 L/h in healthy volunteers, 9.1 L/h in
(2 years of age), and 40-kg adolescent (12 years of patients with congestive heart failure, 1.4 L/h in patients
age) when the CrCl was 60 mL/min/1.73 m2 or greater, with congestive heart failure, and CrCl <30 mL/min)
except for the 40-kg adolescent (12 years of age) with and is similar to prior pediatric estimates.10–12,33 In a
an above-normal CrCl of 160 mL/min/1.73 m2 , where study of 19 children 0 to 13 years of age (3.5-40 kg
concentrations remained below the 100 ng/mL thresh- WT) recovering from cardiac surgery, milrinone CL
old. Most importantly, the 50 mcg/kg/min continuous was estimated at 24 L/h/70 kg using a 2-compartment
infusion exceeded the upper limit of the therapeutic PK model.34 In a more recent 2-compartment PopPK
range for a 3 kg newborn, when the estimated CrCl was model developed in infants <12 months of age who
severely impaired at 30 mL/min/1.73 m2 . Adjusting received milrinone after cardiac surgery, milrinone CL
the infusion rate to 0.25 µg/kg/min after 12 hours of was estimated at 7.91 L/h/70 kg.20 Importantly, milri-
treatment results in a return to therapeutic plasma none CL has been previously shown to increase with
concentrations within 4 hours of dosing adjustment age in both infant cohorts and in cohorts including
(Figure S5). When CrCl was 30 mL/min/1.73 m2 , a children up to 6 years of age.15,18,20 In our covariate
continuous infusion of 0.25 µg/kg/min maintained analysis, the effect of age on CL was best represented by
plasma concentrations in the therapeutic ranges in all an Emax maturation model using PMA, and estimation
3 virtual subjects. Box plots of simulated Css during of maturation function parameters rather than fixing to
the continuous infusion support these conclusions but previously published values resulted in superior model
also highlight the wide range of simulated exposures fit. Inclusion of a maturation function is consistent with
resulting from the significant IIV (Figure 5). a recent PopPK model developed in infants <12 month
8 The Journal of Clinical Pharmacology / Vol 0 No 0 2019

CrCl=30 mL/min/1.73 m2
CrCl=60 mL/min/1.73 m2
CrCl=120 mL/min/1.73 m2
CrCl=160 mL/min/1.73 m2

CrCl=30 mL/min/1.73 m2
CrCl=60 mL/min/1.73 m2
CrCl=120 mL/min/1.73 m2
CrCl=160 mL/min/1.73 m2

CrCl=30 mL/min/1.73 m2
CrCl=60 mL/min/1.73 m2
CrCl=120 mL/min/1.73 m2
CrCl=160 mL/min/1.73 m2

Figure 4. Median simulated milrinone plasma concentration time curve in individual subjects with variable creatinine clearance following a 25-µg/kg
loading dose followed by a 0.25-µg/kg/min continuous infusion in a 3-kg neonate (A);following a 50-µg/kg loading followed by a 0.5-µg/kg/min continuous
infusion in a 3-kg neonate (B); following a 25-µg/kg loading dose followed by a 0.25-µg/kg/min continuous infusion in a 13-kg child (2 years of age) (C);
following a 50-µg/kg loading dose followed by a 0.5-µg/kg/min continuous infusion in a 13-kg child (2 years of age) (D); following a 25-µg/kg loading dose
followed by a 0.25-µg/kg/min continuous infusion in a 40-kg adolescent (12 years of age) (E); following a 50-µg/kg loading followed by a 0.5-µg/kg/min
continuous infusion in a 40-kg adolescent (12 years of age) (F).

of age who received milrinone after cardiac surgery.20 lower number of subjects 1 year of age (N = 23)
In this study, the reported PMA at which 50% of adult in our cohort. Our Hill coefficient estimate similarly
CL is reached (TM50 ) of the maturation function was lacked precision, as illustrated by the wide range of
47 weeks PMA. We observed a higher TM50 value of values estimated during bootstrapping. Despite this
68 weeks PMA, though bootstrapping results spanned lack of precision, we can speculate that maturational
a wide range that included previously published values. effects primarily responsible for milrinone CL likely
This is likely due to the wide age range and the include changes in renal function, in particular the
Hornik et al 9

CrCl=30 mL/min/1.73 m2
CrCl=60 mL/min/1.73 m2
CrCl=120 mL/min/1.73 m2
CrCl=160 mL/min/1.73 m2

CrCl=30 mL/min/1.73 m2
CrCl=60 mL/min/1.73 m2
CrCl=120 mL/min/1.73 m2
CrCl=160 mL/min/1.73 m2

CrCl=30 mL/min/1.73 m2
CrCl=60 mL/min/1.73 m2
CrCl=120 mL/min/1.73 m2
CrCl=160 mL/min/1.73 m2

Figure 4. Continued.

ontogeny of renal transporters that may be responsible to our observations, with a reported TM50 of 58 weeks
for active secretion, given that >80% of the parent PMA.24,35 This is significantly greater than the TM50
drug is recovered unchanged in the urine and that previously reported to characterize the maturation of
mean renal CL in healthy adults (300 mL/min) is in glomerular filtration rate.30,31 A greater sample size
excess of glomerular filtration.18 Because the primary of subjects, particularly neonates, would be needed to
transporters that may be responsible for active secretion more precisely estimate the TM50 value and clarify
of milrinone are unknown, characteristics of the matu- whether maturation of milrinone clearance is in fact
rational function of milrinone CL remain speculative. reflective of transporter maturation and to what extent
Interestingly, acyclovir, a drug cleared through active this maturation differs from that of glomerular filtra-
renal secretion, likely organic anion transporters 1 and tion processes. Alternatively or perhaps concomitantly,
3, exhibits maturational properties that are more similar additional covariates beyond age and renal function
10 The Journal of Clinical Pharmacology / Vol 0 No 0 2019

Figure 5. Simulated steady-state milrinone plasma concentration (Css ) by weight, age, and creatinine clearance following a 0.25-µg/kg/min continuous
infusion (A) and during a 0.5-µg/kg/min continuous infusion (B). Dashed lines represent the upper and lower limit of the therapeutic range (100 ng/mL
and 300 ng/mL, respectively). Outer limits of the boxes are 25th and 75th percentiles, vertical lines inside the box are the 50th percentile, while
whiskers are 1.5 times the interquartile range. Outlier values are not shown.

may affect milrinone elimination and could be further estimated CrCl of 117 mL/min/1.73 m2 (5.9 L/h/70 kg
elucidated in a larger, more granular data set. vs 15.9 L/h/70 kg), the median estimated value in our
Based on our final PopPK model, children with study cohort.36 This decrease in CL is comparable
an estimated CrCl of 30 mL/min/1.73 m2 , indicative to findings from a prior study of 11 children aged
of severe renal impairment as defined by the FDA, 3 weeks to 20 years of age diagnosed with acute kidney
had a 65% lower milrinone CL than children with an injury with or without continuous renal replacement
Hornik et al 11

therapy requirement (estimated CrCl 2.5-103 mL/min), suggesting that this dosing may be adequate. In
where the mean milrinone CL was 4.7 L/h/70 kg.18 all simulation scenarios where estimated CrCl was
In this prior study, milrinone CL was estimated using 60 mL/min/1.73 m2 or greater, a bolus dose of 50 µg/kg
PopPK modeling based on scavenged sampling, with followed by a continuous infusion of 0.5 µg/kg/min
only allometrically scaled weight tested as a covariate, achieved plasma concentrations below the upper limit
and CL being the only population parameter estimated. of the therapeutic concentration range. Overall, our
The authors reported a strong correlation (R2 = 0.7) proposed doses are within the range commonly used in
between milrinone CL and CrCl estimated by the critically ill infants and children.2,37,38
bedside Schwartz equation but did not leverage their Side effects associated with higher milrinone plasma
model to perform dosing simulations at the population concentrations in children are mostly unknown. In
level. More recently, the same authors developed a adults, plasma concentrations >500 ng/mL have been
PopPK model in 92 infants <12 months of age who associated with adverse events including hypotension,
received milrinone after cardiac surgery and reported a arrhythmias, and thrombocytopenia.33,39,40 The PTN
population estimate for CL of 7.91 L/h/70 kg. In this POPS study that generated the data used in this analysis
analysis, renal insufficiency as measured by KDIGO is not designed to rigorously capture adverse events.
criteria was included as a covariate on CL. Consistent However, our simulations suggested that at the pro-
with our findings, milrinone CL was significantly lower posed dosing regimens, most infants would maintain
in infants with KDIGO stage 3 renal insufficiency. milrinone Css below the 500 ng/mL cutoff.
The overall higher population estimate of CL in our Despite its strengths, our study has several important
analysis is likely related to the difference in our study limitations. Our model underpredicts higher milrinone
population, which included older children. Indeed, the plasma concentrations. The limited number of PK
median empirical Bayesian estimates for CL in the samples with high concentrations likely contributed
subset of infants <1 year of age included in our analysis to this bias. Despite a comprehensive covariate anal-
(11 L/h/70 kg) is more comparable to the value reported ysis, IIV variability on CL remains high in our fi-
in this recent publication. nal model (70%). While this value is comparable to
We are able to leverage our model to perform dosing PopPK models of other drugs developed using op-
simulations that take into account various levels of portunistically collected PK data, overprediction of
renal impairment as quantified by estimated CrCl. IIV is evidenced by the VPC. We draw attention to
Since we estimated CrCl with the bedside Schwartz this limitation by plotting simulated Css that incorpo-
equation, our model has applicability to daily clinical rate IIV to demonstrate the range of concentrations
practice in children. Our simulations support the notion that may be anticipated at a specific infusion rate for
of dose reduction in children, with renal impairment a child of given WT and age, and estimated CrCl
and an estimated CrCl of 30 mL/min/1.73 m2 , based on our model. This wide range is comparable
a CrCl that is typically observed in the immediate to the dosing simulation results recently reported in
neonatal period. This finding is comparable to a infants after cardiac surgery. Of note, the model used
prior study of 16 neonates with complex single- for dosing simulations in that study was developed
ventricle congenital heart disease undergoing cardiac from a rich data set of >1000 plasma samples in
surgery on cardiopulmonary bypass who received 92 infants. Recently published PopPK models have
milrinone in the immediate perioperative settings.17 characterized milrinone disposition in children using
While the 2-compartment PopPK model developed a 2-compartment approach, though a 1-compartment
in this study did not incorporate a measure of renal model has also been previously described. A richer PK
function as a covariate, the authors observed a data set with more frequent sampling per subject may
gradual increase of milrinone CL over time, which have allowed us to identify a multicompartment model
correlated with improvements in renal function. structure, which may in turn have improved precision
Based on their model, the authors recommended of individual estimates. Future efforts pooling existing
an infusion rate of 0.2 µg/kg/min for infants with PK data sets may prove beneficial in developing a model
postoperative renal dysfunction. In the previously that most comprehensively characterizes milrinone dis-
mentioned recent study of infants after cardiac surgery, position in children by incorporating effects of size,
the recommended infusion range for those with maturation, renal function, and other covariates such
KDIGO stage 3 criteria was 0.2 µg/kg/min.20 Our as use of ECMO across a wide pediatric age range. We
dosing simulations arrived at a similar conclusion for calculated estimated CrCl using the modified bedside
infants (3-kg WT, 40 weeks PMA). In a toddler Schwartz equation. The simplicity of this calculation
(13-kg WT, 144 weeks PMA), an infusion of lends greater applicability to our model in clinical
0.5 µg/kg/min reached a median Css at the upper limit practice, but does carry with it the limitations of CrCl
of the therapeutic concentration range (300 ng/mL), estimation based on this method. These limitations
12 The Journal of Clinical Pharmacology / Vol 0 No 0 2019

may be more pronounced in pediatric subpopulations, Institute, Little Rock, AR; Ian M. Paul, Penn State College
including infants <2 years of age, obese children, of Medicine, Hershey, PA; Christoph Hornik, Duke Clinical
and those supported with ECMO, all of whom were Research Institute, Durham, NC; Kelly Wade, Children’s
included in our study cohort. Newer renal biomarkers Hospital of Philadelphia, Philadelphia, PA.
may be more accurate and timely in the diagnosis of The Eunice Kennedy Shriver National Institute of Child
renal insufficiency, and better take into consideration Health and Human Development: David Siegel and Perdita
effects of renal dysfunction on active secretion of drugs, Taylor-Zapata.
rather than glomerular filtration alone. We were forced The EMMES Corporation (Data Coordinating Center):
to impute estimated CrCl in 5 subjects included in Ravinder Anand and Gina Simone.
our analysis, as a result of missing serum creatinine Pediatric Trials Network’s Pharmacokinetics of Understud-
concentrations. While the PopPK parameters estimated ied Drugs Administered to Children per Standard of Care Study
from the final model did not differ significantly (<20%) Team and Study Coordinators
when these subjects were excluded, lack of complete Duke Clinical Research Institute: Chiara Melloni (PI),
covariate data is an important potential limitation of Barrie Harper (PL), Adam Samson (L-CRA), Tammy Day
opportunistic clinical trials. We chose target plasma (CRA)
concentrations based on previously published cutoffs Ann and Robert H Lurie Children’s Hospital of Chicago:
but acknowledge that these are based on limited phar- Ram Yogev (PI), Laura Fearn (SC)
macodynamic and efficacy data in this population.41 We Children’s Hospital Colorado: Peter Mourani (PI), Megan
are also unable to definitively comment on exposures Dix (SC), Matthew Steinbess (SC)
associated with adverse events, as safety data were not Duke University Medical Center: Kevin Watt (PI), Saman-
comprehensively collected in our study. While some tha Wrenn (SC), Christie Milleson (SC), Melissa Harward
of these limitations may be overcome in a prospective (SC)
PK, pharmacodynamics/efficacy, and safety trial of University of Louisville and Norton Children’s Hospital:
milrinone, consideration should also be given to the Janice E. Sullivan (PI), Tressa Bratton (SC), Carla Rich (SC),
use of therapeutic drug monitoring for the titration Carrie Schanie (SC), Mitzi Smith (SC), Melissa Thomas (SC)
of milrinone in children with or at risk for renal Medical University of South Carolina Children’s Hospi-
impairments, as previously suggested by others.18 tal: Andrew Atz (PI), Hibah Al Nasiri (SC), Patricia Infinger
(SC), Layla Al Sarraf (SC)
Rainbow Babies and Children Hospital: David Speicher
Conclusions (PI), Sue Bergant (SC)
We used opportunistic PK data from a cohort of Oregon Health and Science University: Amira Al-UZri
children with a wide range of age and renal func- (PI), Kira Clark (SC)
tion to develop a pediatric PopPK model for IV- University of Virginia Children’s Hospital: Michelle Adu-
administered milrinone. Our model is the first to in- Darko (PI), Robin Kelly (SC)
corporate estimated CrCl as a covariate on CL. Using Alfred I. DuPont Hospital for Children: Marisa Meyer
our model for dosing simulations, we recommend (PI), Glen Stryjewski (PI), Kimberly Klipner (SC), Ramany
that infants and small children with an estimated John (SC)
CrCl of 30 mL/min/1.73 m2 receive infusions of University of Maryland: Susan Mendley (PI), Donna
0.25 µg/kg/min. Future studies of milrinone in criti- Cannonier (SC)
cally ill children should incorporate markers of renal Cincinnati Children Hospital Medical Center: Stuart
function to further elucidate the complex relationship Goldstein (PI), Teresa Mottes (SC), Tara Terrell (SC)
between dynamically changing renal function, matura- KK Women’s and Children’s Hospital, Singapore: Jan Hau
tion, and growth and milrinone exposure. Lee (PI), Kathy Liaw (SC), Cecilia Anggraini Chandra (SC)
University of California San Diego Medical Center: Adri
Tremoulet (PI), Wade Rich (SC), Jason Sauberan (SC),
Acknowledgments Pamela Emery (SC)
The assay measuring milrinone plasma concentrations was University of Florida Center for HIV/AIDS Research
developed and performed at Frontage Laboratories Inc. (Ex- Education and Service (UF CARES: Mobeen H. Rathore,
ton, Pennsylvania). MD (PI), Nizar Maraqa, MD (PI), Kathleen Thoma (SC)
The Best Pharmaceuticals for Children Act–Pediatric Trials University of North Carolina at Chapel Hill: Matthew
Network Publication Committee: Gary Furda, Duke Clinical Laughon (PI), Janice Bernhardt (SC)
Research Institute, Durham, NC; Danny Benjamin, Duke University of Utah Hospitals and Clinics: Catherine Sher-
Clinical Research Institute, Durham, NC; Edmund Cappar- win (PI), Sharada Dixit (SC)
elli, University of California San Diego, San Diego, CA; Wesley Medical Center: Paula Delmore (PI), Barry Bloom
Gregory L. Kearns, Arkansas Children’s Hospital Research (Sub-I)
Hornik et al 13

Declaration of Conflict of Interest 9. Lehtonen LA, Antila S, Pentikainen PJ. Pharmacokinetics and
pharmacodynamics of intravenous inotropic agents. Clin Phar-
Disclosures are available at www.dcri.duke.edu/research/ macokinet. 2004;43(3):187-203.
coi.jsp. C.P.H receives support for research from the 10. Das PA, Skoyles JR, Sherry KM, Peacock JE, Fox PA, Woolfrey
Eunice Kennedy Shriver National Institute for Child SG. Disposition of milrinone in patients after cardiac surgery.
Health and Human Development (NICHD) (5K23HD Br J Anaesth. 1994;72(4):426-429.
11. Stroshane RM, Koss RF, Biddlecome CE, Luczkowec C,
090239). D.G. receives support for research from the
Edelson J. Oral and intravenous pharmacokinetics of milri-
NICHD (5K23HD083465 and 1R01HD096435). Portions none in human volunteers. J Pharm Sci. 1984;73(10):1438-
of this work were funded under NICHD contract 1441.
HHSN275201000003I for the Pediatric Trials Network 12. Larsson R, Liedholm H, Andersson KE, Keane MA, Henry G.
(PI Danny Benjamin). M.C.W. receives support for research Pharmacokinetics and effects on blood pressure of a single oral
dose of milrinone in healthy subjects and in patients with renal
from the National Institutes of Health (1R01-HD076676-
impairment. Eur J Clin Pharmacol. 1986;29(5):549-553.
01A1), National Institute of Allergy and Infectious Diseases 13. Bailey JM, Levy JH, Kikura M, Szlam F, Hug CC Jr. Phar-
(HHSN272201500006I and HHSN272201300017I), NICHD macokinetics of intravenous milrinone in patients undergoing
(HHSN275201000003I), the Biomedical Advanced Research cardiac surgery. Anesthesiology. 1994;81(3):616-622.
and Development Authority (HHSO100201300009C), and 14. Baruch L, Patacsil P, Hameed A, Pina I, Loh E. Pharmaco-
dynamic effects of milrinone with and without a bolus loading
industry for drug development in adults and children.
infusion. Am Heart J. 2001;141(2):266-273.
15. Bailey JM, Hoffman TM, Wessel DL, et al. A population
Source of Funding pharmacokinetic analysis of milrinone in pediatric patients after
cardiac surgery. J Pharmacokinet Pharmacodyn. 2004;31(1):43-
This work was funded under National Institute of Child 59.
Health and Human Development contract (HHSN2752 16. Paradisis M, Jiang X, McLachlan AJ, Evans N, Kluckow M,
01000003I or HHSN275201800003I) for the Pediatric Trials Osborn D. Population pharmacokinetics and dosing regimen
Network (PI Danny Benjamin). The content is solely the re- design of milrinone in preterm infants. Arch Dis Child Fetal
Neonatal Ed. 2007;92(3):F204-209.
sponsibility of the authors and does not necessarily represent
17. Zuppa AF, Nicolson SC, Adamson PC, et al. Population phar-
the official views of the National Institutes of Health. macokinetics of milrinone in neonates with hypoplastic left
heart syndrome undergoing stage I reconstruction. Anesth Analg.
Data Sharing 2006;102(4):1062-1069.
18. Gist KM, Mizuno T, Goldstein SL, Vinks A. Retrospective eval-
All data relevant to this manuscript will be made publicly uation of milrinone pharmacokinetics in children with kidney
available. injury. Ther Drug Monit. 2015;37(6):792-796.
19. Gist KM, Goldstein SL, Joy MS, Vinks AA. Milrinone dosing
issues in critically ill children with kidney injury: a review.
J Cardiovasc Pharmacol. 2016;67(2):175-181.
References 20. Mizuno T, Gist KM, Gao Z, et al. Developmental pharma-
cokinetics and age-appropriate dosing design of milrinone in
1. Milrinone [package insert]. Irvine, CA: B. Braun Medical Inc; neonates and infants with acute kidney injury following cardiac
2006. surgery. Clin Pharmacokinet. 2019;58(6):793-803.
2. Bishara T, Seto WT, Trope A, Parshuram CS. Use of milrinone 21. Keizer RJ, van Benten M, Beijnen JH, Schellens JH, Huitema
in critically ill children. Can J Hosp Pharm. 2010;63(6):420-428. AD. Pirana and PCluster: a modeling environment and clus-
3. Burkhardt BE, Rucker G, Stiller B. Prophylactic milrinone for ter infrastructure for NONMEM. Comput Methods Programs
the prevention of low cardiac output syndrome and mortality Biomed. 2011;101(1):72-79.
in children undergoing surgery for congenital heart disease. 22. Lindbom L, Pihlgren P, Jonsson EN. PsN-Toolkit–a collection
Cochrane Database Syst Rev. 2015(3):CD009515. of computer intensive statistical methods for non-linear mixed
4. Lindsay CA, Barton P, Lawless S, et al. Pharmacokinetics and effect modeling using NONMEM. Comput Methods Programs
pharmacodynamics of milrinone lactate in pediatric patients Biomed. 2005;79(3):241-257.
with septic shock. J Pediatr. 1998;132(2):329-334. 23. Jonsson EN, Karlsson MO. Xpose–an S-PLUS based popula-
5. McNamara PJ, Shivananda SP, Sahni M, Freeman D, Taddio tion pharmacokinetic/pharmacodynamic model building aid for
A. Pharmacology of milrinone in neonates with persistent pul- NONMEM. Comput Methods Programs Biomed. 1999;58(1):51-
monary hypertension of the newborn and suboptimal response 64.
to inhaled nitric oxide. Pediatr Crit Care Med. 2013;14(1): 24. Holford N, Heo YA, Anderson B. A pharmacokinetic standard
74-84. for babies and adults. J Pharm Sci. 2013;102(9):2941-2952.
6. Honerjager P, Nawrath H. Pharmacology of bipyridine phos- 25. Schwartz GJ, Work DF. Measurement and estimation of
phodiesterase III inhibitors. Eur J Anaesthesiol Suppl. 1992;5:7- GFR in children and adolescents. Clin J Am Soc Nephrol.
14. 2009;4(11):1832-1843.
7. Yano M, Kohno M, Ohkusa T, et al. Effect of milrinone 26. Gretz N, Schock D, Sadick M, Pill J. Bias and precision of
on left ventricular relaxation and Ca(2+) uptake function of estimated glomerular filtration rate in children. Pediatr Nephrol.
cardiac sarcoplasmic reticulum. Am J Physiol Heart Circ Physiol. 2007;22(2):167-169.
2000;279(4):H1898-1905. 27. Hellerstein S, Berenbom M, DiMaggio S, Erwin P, Simon
8. Shipley JB, Tolman D, Hastillo A, Hess ML. Milrinone: basic SD, Wilson N. Comparison of two formulae for estima-
and clinical pharmacology and acute and chronic management. tion of glomerular filtration rate in children. Pediatr Nephrol.
Am J Med Sci. 1996;311(6):286-291. 2004;19(7):780-784.
14 The Journal of Clinical Pharmacology / Vol 0 No 0 2019

28. Mattman A, Eintracht S, Mock T, et al. Estimating pediatric 36. U.S. Department of Health and Human Services, Food and
glomerular filtration rates in the era of chronic kidney disease Drug Administration, Center for Drug Evaluation and Re-
staging. J Am Soc Nephrol. 2006;17(2):487-496. search (CDER). Guidance for Industry: Pharmacokinetics in
29. Zappitelli M, Joseph L, Gupta IR, Bell L, Paradis G. Val- patients with impaired renal function—study design, data anal-
idation of child serum creatinine-based prediction equations ysis, and impact on dosing and labeling. Silver Spring, MD;
for glomerular filtration rate. Pediatr Nephrol. 2007;22(2): 1998.
272-281. 37. Czaja AS, Reiter PD, Schultz ML, Valuck RJ. Patterns of off-
30. Mahmood I, Staschen CM. Prediction of human glomerular label prescribing in the pediatric intensive care unit and priori-
filtration rate from preterm neonates to adults: evaluation of tizing future research. J Pediatr Pharmacol Ther. 2015;20(3):186-
predictive performance of several empirical models. AAPS J. 196.
2016;18(2):445-454. 38. Rizza A, Bignami E, Belletti A, et al. Vasoactive drugs and
31. Rhodin MM, Anderson BJ, Peters AM, et al. Human re- hemodynamic monitoring in pediatric cardiac intensive care: an
nal function maturation: a quantitative description using Italian survey. World J Pediatr Congenit Heart Surg. 2016;7(1):
weight and postmenstrual age. Pediatr Nephrol. 2009;24(1): 25-31.
67-76. 39. Taniguchi T, Shibata K, Saito S, Matsumoto H, Okeie K.
32. Bergstrand M, Hooker AC, Wallin JE, Karlsson MO. Prediction- Pharmacokinetics of milrinone in patients with congestive heart
corrected visual predictive checks for diagnosing nonlinear failure during continuous venovenous hemofiltration. Intensive
mixed-effects models. AAPS J. 2011;13(2):143-151. Care Med. 2000;26(8):1089-1093.
33. Cox ZL, Calcutt MW, Morrison TB, Akers WS, Davis MB, 40. Young RA, Ward A. Milrinone. A preliminary review of
Lenihan DJ. Elevation of plasma milrinone concentrations in its pharmacological properties and therapeutic use. Drugs.
stage D heart failure associated with renal dysfunction. J Car- 1988;36(2):158-192.
diovasc Pharmacol Ther. 2013;18(5):433-438. 41. Bailey JM, Miller BE, Lu W, Tosone SR, Kanter KR, Tam VK.
34. Ramamoorthy C, Anderson GD, Williams GD, Lynn AM. The pharmacokinetics of milrinone in pediatric patients after
Pharmacokinetics and side effects of milrinone in infants and cardiac surgery. Anesthesiology. 1999;90(4):1012-1018.
children after open heart surgery. Anesth Analg. 1998;86(2):283-
289.
35. Tod M, Lokiec F, Bidault R, De Bony F, Petitjean O, Au-
Supporting Information
jard Y. Pharmacokinetics of oral acyclovir in neonates and in Additional supporting information may be found on-
infants: a population analysis. Antimicrob Agents Chemother. line in the Supporting Information section at the end
2001;45(1):150-157.
of the article.