Page 1
1
Population pharmacokinetics of gentamicin and dosing optimization for infants 1
Susanna E. Medellín-Garibaya#, Aída Rueda-Naharrob, Silvia Peña-Cabiab, Benito Garcíab, 2
Silvia Romano-Morenoa, Emilia Barciac 3
aDepartamento de Farmacia, Facultad de Ciencias Químicas, Universidad Autónoma de 4
San Luis Potosí, Av. Manuel Nava # 6, Zip Code: 78210, San Luis Potosí, S.L.P. México; 5
bServicio de Farmacia, Hospital Universitario Severo Ochoa, Avenida de Orellana, Zip Code: 6
28911 Leganés, España; cDepartamento de Farmacia y Tecnología Farmacéutica, 7
Universidad Complutense de Madrid, Plaza de Ramón y Cajal, Ciudad Universitaria, Zip 8
Code: 28040, Madrid, España. 9
#Address correspondence to Susanna E. Medellín-Garibay, 10
E-mail: [email protected] 11
Phone: +52 (444) 8262300 xt. 6572 12
Fax: +52 (444) 8262372 13
Running title: Gentamicin population PK in infants 14
Key words: gentamicin, population pharmacokinetics, infants, NONMEM. 15
16
AAC Accepts, published online ahead of print on 10 November 2014Antimicrob. Agents Chemother. doi:10.1128/AAC.03464-14Copyright © 2014, American Society for Microbiology. All Rights Reserved.
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 2
2
ABSTRACT 17
The aim of this study is to characterize and validate the population pharmacokinetics of 18
gentamicin in infants and to determine the influence of clinically relevant covariates to 19
explain the inter- and intra-individual variability associated to this drug. Infants receiving 20
IV gentamicin and with routine therapeutic drug monitoring were consecutively enrolled 21
in the study. Plasma concentration and time data were retrospectively collected from 208 22
infants (1-24 months old) of the Hospital Universitario Severo Ochoa (Spain) of whom 44% 23
were males (mean age 5.8 ± 4.8 months and mean body weight 6.4 ± 2.2 kg). Data analysis 24
was performed with NONMEM 7.2. One- and two-compartment open models were 25
analyzed to estimate gentamicin population parameters and the influence of several 26
covariates. External validation was carried out in another population of 55 infants. 27
The behavior of gentamicin in infants exhibits two-compartment pharmacokinetics with 28
total body weight being the covariate which mainly influences central volume (Vc) and 29
clearance (CL); this parameter was also related to creatinine clearance. Both parameters 30
are age-related and different to those reported for neonatal populations. 31
On the basis of clinical presentation and diagnosis, a once-daily dosage regimen of 7 32
mg/kg every 24 hours is proposed for IV gentamicin, followed by therapeutic drug 33
monitoring to avoid toxicity and ensure efficacy with minimal blood sampling. 34
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 3
3
Gentamicin pharmacokinetics and disposition was accurately characterized in this 35
pediatric population (infants), with the parameters obtained being different to those 36
reported for neonates and children. These differences should be considered for dosing 37
and therapeutic monitoring of this antibiotic. 38
39
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 4
4
INTRODUCTION 40
Aminoglycosides are among the most commonly used broad-spectrum antibiotics 41
in the anti-infective therapy (1, 2). Despite their potential for renal toxicity, ototoxicity, 42
and bacterial resistance, several aminoglycosides continue to play an important role when 43
treating infections (3, 4). Among aminoglycosides, gentamicin is one of the most 44
commonly used due to its low cost and broad-spectrum efficacy (5, 6). Gentamicin is 45
commonly used to treat infections due to both gram-negative and gram-positive bacilli 46
such as E. coli, Proteus, Pseudomonas, Serratia, and Staphylococcus. Other clinical 47
applications of gentamicin include infections in the CNS, respiratory, abdominal and 48
urinary systems, bone, skin and soft tissues, endocarditis, and septicemia; being used in 49
combination with ampicillin as empiric therapy for sepsis in newborns and infants (7). 50
Drug pharmacokinetics and disposition are different in pediatric patients when 51
compared to adults (8). In this context, the Food and Drug Administration (FDA) 52
considering the complex changes and the anatomic, biochemical and physiological 53
differences related to age, proposed to classify pediatric populations as neonate (from 54
birth to 1 month of life); infant (between 1 and 24 months); child and adolescent (9). 55
Growth, development and maturation of enzymatic processes are features with marked 56
physiological influence from birth to adult age (10). For these reasons, pharmacokinetic 57
differences found between newborns and adults justify the need for the performance of 58
pharmacokinetic studies especially during the first months of life. For instance, in very 59
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 5
5
premature neonates (gestational age <32 weeks), the pharmacokinetic characteristics of 60
antibiotics may vary from that in full-term neonates due to differences in drug absorption, 61
distribution, metabolism, and excretion; being all these processes rapidly developed 62
during the first year of life (11, 12). 63
In adults dosage regimens for gentamicin have evolved from multiple daily dosing 64
to extended-interval dosing. This change has also being applied to establishing gentamicin 65
dosing in pediatrics (1, 5, 13, 14). However, its pharmacokinetics in infants exhibits large 66
inter- and intra-individual variability mainly due to the developmental changes occurring 67
from the first month of life (15). In clinical practice individualization of dosage regimens 68
and monitoring of gentamicin concentrations are routinely performed in order to assure 69
peak blood concentrations sufficiently high to elicit the therapeutic response while 70
avoiding high trough concentrations which would be potentially toxic after prolonged drug 71
exposure (11, 16, 17). 72
Non-linear mixed-effects modelling methodology was introduced to population 73
pharmacokinetic analysis several years ago, and has become a standard procedure 74
employed to analyze sparse data (18). This situation applies to data collected from infant 75
populations where usually only one or two blood serum concentrations can be reasonably 76
obtained in clinical practice. A recent review conducted on population pharmacokinetic 77
analyses performed during the first two years of life showed antibiotics as the drug class 78
most cited (44%), with vancomycin and gentamicin representing the two agents most 79
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 6
6
studied. Age was one of the covariates more frequently evaluated during pharmacokinetic 80
analysis. The majority of the studies were conducted on populations composed of 25-100 81
neonates with a post-natal age (PNA) ranging from the first day of life to 1 month (19). 82
It was therefore the purpose of this study to calculate the population 83
pharmacokinetic parameters of gentamicin in a large population of infants (age range 1 – 84
24 months) and to validate the predictive performance of the population analysis in 85
another population with the same characteristics, in order to establish the most 86
appropriate dosage regimens to maximize safety and efficacy of gentamicin in infants. 87
PATIENTS AND METHODS 88
Infants from 1 to 24 months of age receiving gentamicin by intravenous infusion 89
were included in the current study with routine therapeutic drug monitoring (TDM) in the 90
Hospital Universitario Severo Ochoa (Leganes, Spain) between 1990 and 2011. Data 91
collection and handling was performed according to the ethical procedures of the hospital. 92
Total population was divided into two groups: one composed of 208 infants which 93
was used to build the population pharmacokinetic model (population study group), and 94
the other composed of 55 infants (validation study group), used for validating the 95
predictive performance of the former. Assignation of the infants to each group was carried 96
out randomly. 97
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 7
7
The following data were collected from each infant’s medical record: age, sex, 98
weight, height, serum creatinine (CRs), creatinine clearance (CLCR) and body mass index 99
(BMI). CRs was used to estimate the CLCR of gentamicin by the Schwartz equation: CLCR = K 100
· length/CRs, where K is set to 0.45 for term infants whose weight is appropriate for age; 101
to 0.33 for low weight infants; and to 0.55 for infants over one year (7, 20). This formula 102
applies to alkaline picrate ("Jaffe") methods, even those traceable to IDMS, as applied for 103
the current work for CRs measurement. Details of gentamicin administration was also 104
retrieved from each patient’s record (doses, start of infusion, infusion rate, dosing 105
interval) as well as data from TDM (serum sampling date and time, as well as assay 106
concentration). 107
Administration of gentamicin was performed by means of IVAC-syringe pump (IVAC 108
corp., USA) as intravenous infusions given over a period of 20 min. The empirical dosing 109
regimen for infants is 2.5 mg/kg/8h or 4.5 – 7.5 mg/kg/24 h (7). Nevertheless, infants who 110
received doses of 7 mg/kg or more and had a sample drawn one hour after the start of the 111
infusion were excluded (21). 112
Gentamicin analysis 113
Gentamicin serum concentrations were determined by fluorescence polarization 114
immunoassay (Abbott TDx, Abbott Park, IL, USA). Calibration curves for gentamicin in 115
human serum ranged from 0.5 mg/L to 10 mg/L, with a coefficient of variation of 1.4% for 116
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 8
8
the concentration of 4 mg/L. The intra- and inter-day coefficients of variation of the 117
method were 4.3% and 5.3%, respectively. All plasma concentrations below the limit of 118
quantification were not considered for the population analysis. 119
Pharmacokinetic analysis 120
Pharmacokinetic modeling was performed using nonlinear mixed effects analysis 121
with NONMEM 7.2 (Icon Development Solution, Ellucott City, MD, USA). Subroutines 122
ADVAN1 TRANS2 and ADVAN3 TRANS4 were used to evaluate the one- and two-123
compartment pharmacokinetic open models, respectively. First-order conditional 124
estimation (FOCE) was employed to calculate the mean and variance of the population 125
pharmacokinetic parameters. Fixed-effects parameters directly estimated from the one- 126
compartment open model were total clearance (CL) and volume of distribution (Vd). For 127
the two-compartment open model the pharmacokinetic parameters estimated were total 128
clearance (CL), central volume of distribution (Vc), intercompartmental clearance (Q), and 129
peripheral volume of distribution (Vp). 130
The statistical models used to account for inter-individual variability in the 131
pharmacokinetic parameters as well as for the residual error were modeled with 132
homocedastic (additive), heterocedastic (proportional) and exponential error. 133
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 9
9
The Akaike information criterion (AIC) was calculated based on the objective 134
function (OBJ) for comparing structural models; a drop of 2 was the threshold for 135
considering one model over another (AIC = OBJ · np; where np is the total number of 136
parameters in the model). Graphical analysis and a generalized additive model (GAM) 137
were employed to evaluate the influence of covariates on the population pharmacokinetic 138
parameters estimated from both base pharmacokinetic models. Covariates were 139
introduced into population model using linear, allometric or exponential functions. 140
Changes occurring in the OBJ and due to the addition of covariates in the regression 141
model are χ2 asymptotically distributed, with degrees of freedom equal to the difference 142
in the number of parameters between models. A difference (ΔOBJ) higher than 3.8 (P 143
<0.05, d.f. = 1) was considered significant for the addition of a covariate. An intermediate 144
model was obtained by first adding the continuous covariates that showed influence on 145
the base model, and keeping only those that significantly diminished OBJ as mentioned 146
before. Each covariate was added individually; firstly the most important ones with the 147
rest of them being added or dropped following the same criteria. The model was selected 148
when no further improvement occurred. Finally, the discrete covariate with influence over 149
the intermediate model was selected as mentioned before, followed by graphical 150
assessment. Evaluation of the final model was performed by means of backward 151
elimination in which each of the covariates was deleted in the same order as it was 152
introduced in the model, and considering ΔOBJ>10.836 (P<0.001, d.f.=1) as significant. 153
Therefore, clinically relevant covariates that explain inter-individual variability showing a 154
significant contribution to the estimation of the fixed parameters were kept (18, 22). 155
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 10
10
The precision of the parameter estimate is expressed as the standard error (SE) 156
reported in the covariance tables given by NONMEM. The confidence intervals were 157
estimated as parameter value ± 1.96 · SE. 158
Bootstrap was performed with 200 replicates to report the confidence intervals (CI) 159
for each parameter; the 95% CI must cover the true value of the parameter estimated to 160
prove the stability of the final model (18). 161
Validation 162
External validation was carried out in a separate group of infants (n=55) with 163
similar characteristics to the population study group. The predictive performance of the 164
population pharmacokinetic model was evaluated using a priori method. Gentamicin 165
serum concentrations of the validation group were compared to their predicted values in 166
order to estimate the precision of the population model built. The bias fit was evaluated 167
by means of the mean prediction error (MPE). Precision was estimated by means of the 168
absolute prediction error (APE), the mean-squared prediction error (MSPE), and the root-169
mean-squared of the prediction error (RMSPE) (23, 24). 170
171
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 11
11
RESULTS 172
Demographics 173
The population study group was composed of 208 infants and 55 infants in the 174
validation study group. Considering both populations 56% were females and 44% males. 175
Demographic characteristics of the groups assayed (population study group and validation 176
study group) are summarized in Table 1. 177
[Insert Table 1] 178
Both populations were classified by weight according to age (months) and sex 179
following the WHO child growth standards (25); 12.1% of the infants were 1 – 11 months 180
old with low BW and were classified as INF = 0; 72.9% of the infants were 1 – 11 months 181
old with normal BW (INF = 1); and 15 % were infants over one year old with normal BW 182
(INF = 2). 183
In the Hospital Universitario Severo Ochoa, different dosage regimens were used 184
along the period studied, with mean gentamicin doses of 6.0 ± 1.5 mg/kg/day and 5.5 ± 185
1.2 mg/kg/day for the population and validation study group, respectively. In the 186
population study group 4 infants received mean gentamicin doses of 2.3 mg/kg/6 h (total 187
daily d 188
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 12
12
189
ose 9.4 mg/kg), 98 received mean doses of 2.3 mg/kg/8 h (total daily dose 6.8 190
mg/kg), 3 received 3.05 mg/kg/12 h (total daily dose 6.1 mg/kg), and the rest (103 infants) 191
received single daily doses of 4.8 mg/kg. Regarding the validation study group (n=55), 26 192
infants received mean doses of 2.1 mg/kg/8 h (total daily dose 6.4 mg/kg/24h) whereas 29 193
infants received single daily doses of 4.6 mg/kg. As indicated before, mean doses of 6.0 ± 194
1.5 mg/kg/day and 5.5 ± 1.2 mg/kg/day were given to both population and validation 195
study groups, respectively. 196
For both populations a total of 421 serum gentamicin concentrations were 197
available, ranging from 1 to 6 gentamicin serum concentrations per infant. Of these, 335 198
serum concentrations corresponded to the population study group and 86 to the 199
validation study group. For most infants (96.2%), one or two gentamicin serum 200
concentrations were obtained; of these 45% had one and 51.2% two. 201
Figure 1 shows the number of gentamicin serum concentrations obtained related 202
to blood sampling times. A total of 335 gentamicin serum concentrations were available 203
from the population study group. Gentamicin serum concentrations ranged 0.5 – 15.9 204
mg/L with a mean value of 4.8 mg/L. Blood sampling times ranged from 1 to 321 hours 205
(mean 56.3 h); 8.6% of the samples were collected within the first 24 hours, 66.1% 206
between 24 and 72 hours, and 25.2% from 72 to 321 hours. 207
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 13
13
[Insert Figure 1] 208
Base pharmacokinetic models and covariate addition 209
The first pharmacokinetic model evaluated was the one-compartment open model 210
with exponential inter-individual variability associated to both CL (Ө1) and Vd (Ө2). 211
Residual variability was modeled as homocedastic (additive) error. For this model OBJ was 212
564.3. 213
When using a two-compartment open model OBJ decreased 43.5 units when 214
compared to the one-compartment open model. This difference in OBJ corresponds to a P 215
value lower than 0.005 (χ2 is assumed). Table 2 summarizes the pharmacokinetic 216
parameters estimated with both pharmacokinetic models (one- and two-compartment 217
open models). 218
[Insert Table 2] 219
Initially, the influence of covariates was evaluated with the one-compartment 220
model in which CL and Vd were assumed to be the same for all individuals in the 221
population study group; then the two-compartment open model was evaluated assuming 222
the same for CL, Vc, Vp and Q. When applying the two-compartment open model ΔOBJ 223
was statistically significant but the coefficients of variation (CVs%) obtained for residual 224
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 14
14
variability differ slightly (0.2%) from both pharmacokinetic models, with CVs (%) of inter-225
individual variability associated to both CL and Vd being higher for the two-compartment 226
open model. Table 3 summarizes the population parameters and inter-individual 227
variability values obtained for the final pharmacokinetic models. 228
[Insert Table 3] 229
For the one-compartment open model the influence of BW on Vd was evaluated, 230
taking into consideration in this analysis the classification of the population study group 231
according to differences in age and BW as stated previously. For this analysis inter-232
individual variability associated to Vd was significantly reduced. When introducing CLCR in 233
the base one-compartment model, a significant reduction of the inter-individual variability 234
associated to CL was also observed. However, sex as covariate did not have any influence 235
on the performance of the model. 236
When applying the same choice of covariates to the two-compartment model the 237
influence of both BW and CLCR on CL and Vc was demonstrated. In this case the inter-238
individual variability associated to CL and Vc is significantly reduced with independence of 239
the characteristics (age and BW) of the infants (population study group). The shrinkage for 240
inter-individual and residual variability obtained for the final two-compartment model was 241
29 – 17% and 48%, respectively. 242
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 15
15
Figure 2 shows the scatters plots of predicted (PRED) versus observed (DV) 243
gentamicin concentrations for the base and final two-compartment open models. As it can 244
be observed the introduction of covariates in the model clearly reduced the dispersion of 245
the points around the identity line. Moreover, the scatter plots of conditional weighed 246
residuals (CWRES) versus BW (including the identity line) are also shown demonstrating 247
the influence that BW has on the pharmacokinetic model. The incorporation of BW to CL 248
and Vd improves the goodness of fit for the group of infants with low BW for whom initial 249
overestimation occurred as for infants with normal BW for whom underestimation was 250
initially obtained. 251
[Insert Figure 2] 252
After the selection of the final model, a bootstrap analysis was performed with 200 253
replicates from the original dataset to adequately reflect the parameter distributions. The 254
results obtained are summarized in Table 3. In this table the same number of the fixed 255
parameters (Ѳ) corresponding to the final two-compartment open model with covariates 256
are indicated as median and 95% IC. The percentile bootstrap CI reported covers the mean 257
values shown in Table 3 for each parameter, which reflects the symmetry of the 258
distribution and allows assuming stability in the final model. 259
Validation 260
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 16
16
A recent literature review found that validations of PK models are only performed 261
in 17% of PK-PD published pediatric studies and in 28% of adult studies however, 262
adequate model validation is essential in model building (26). The characteristics of the 263
validation group are summarized in Table 1. The validation study group was composed of 264
55 infants whose mean age was 5.5 ± 4.4 months, similar to that of the population study 265
group. Mean BW was 6.7 ± 2.3 kg. 266
Figure 3 shows the scatter plot of predicted versus observed gentamicin 267
concentrations corresponding to the validation group and obtained for the final two-268
compartment open model. Table 4 summarizes the mean prediction errors obtained from 269
the validation of the final model when compared to those corresponding to the base 270
model. 271
[Insert Figure 3] 272
[Insert Table 4] 273
Mean prediction errors (MPE) for the two-compartment model include zero in the 274
95% CI, thereby indicating that from a statistical point of view the model is not 275
conditioned to over- or under-estimation. 276
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 17
17
It can be observed a decrease in the difference obtained between the upper and 277
lower limits of the 95% CI when comparing the final model to the base one, thereby 278
indicating higher precision of the model. This is also confirmed by the APE (absolute 279
precision error) value obtained. Similarly, MSPE (mean-squared prediction error) and 280
RMPSE (root-mean-squared prediction error) are clearly reduced, thereby confirming the 281
improvement of precision obtained in the prediction of the final pharmacokinetic model. 282
DISCUSSION 283
A population pharmacokinetic analysis of gentamicin in infants was performed in 284
order to define the disposition of aminoglycosides in newer populations that have not 285
been already recorded in the literature. This population of infants exhibit different 286
characteristics than newborns or adults (13, 27, 28). In the current study the population 287
pharmacokinetic parameters of gentamicin have been retrospectively estimated by 288
NONMEM 7.2 in a total population of 263 infants who were admitted to the Hospital 289
Universitario Severo Ochoa (Leganés, Spain) between the years 1990 and 2011 and that 290
received gentamicin as part of their treatment. 291
Both one- and two-compartment pharmacokinetic models can adequately describe 292
the pharmacokinetics of gentamicin in infants. However the two-compartment open 293
model showed improved goodness of fit to represent the disposition of gentamicin in the 294
population studied. Previous reports have also proposed the use of the two-compartment 295
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 18
18
model to describe the pharmacokinetic behavior of gentamicin in different populations of 296
infants (27, 28). 297
In order to explain the variability found, physiological variables (BW, height, age, 298
sex and CLCR) were analyzed to evaluate their influence on the pharmacokinetic 299
parameters of gentamicin. BW exhibited the highest correlation which was followed by 300
CLCR. Gentamicin CL and Vd are different in infants when compared to adults, showing a 301
wide inter-individual variability. 302
This difference can be attributed to several factors that contribute to modify the 303
distribution of gentamicin, such as protein binding, blood flow to tissues, membrane 304
permeability and drug affinity for different tissues (27-29). The fact that gentamicin is a 305
polar compound with low plasma protein binding (<30%) makes its Vd directly related to 306
extracellular body fluid (5). In this respect and during the first two years of life, 307
compartmentalization of body water changes continuously; total body water decreases 308
and adipose tissue increases (29), resulting in higher volume of distribution in newborns 309
and infants that in older children. BW also increases significantly during the first year of 310
life; therefore the introduction of such covariate in the population pharmacokinetic 311
analysis allows for the comparison of Vd with other populations. The mean value obtained 312
for this parameter is in accordance with other studies, being lower than that reported for 313
adult populations (0.28 L/kg) and higher than that given for newborns (0.48 L/kg) (27, 28). 314
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 19
19
Touw et al., have demonstrated that normalization of gentamicin Vd by BW 315
(Vd/kg) clearly decreases with age (11). After applying the two-compartment open model 316
to our population, differences in Vd are not observed, probably due to the age range of 317
the population assayed (1-2 years). Despite this fact, in the current study Vd was 318
determined for a population which has not been previously described in the literature 319
since most studies are carried out in populations ranging from 6 months to 4 years or up 320
to 18 years of age (30, 31). Differences in gentamicin Vd by age were not found when 321
applying the two-compartment pharmacokinetic model, which simplifies the population 322
model by considering only total BW. 323
The categorical variable sex did not exert any influence on the pharmacokinetic 324
parameters of gentamicin, which is in accordance with other studies that have 325
demonstrated that creatinine serum concentration and muscular body mass do not 326
exhibit significant differences between males and females before the adolescence age 327
(20). The variability found in Vd makes the need for monitoring plasma gentamicin 328
concentrations in order to improve its efficacy and avoid the toxicity effects associated to 329
aminoglycosides. 330
Gentamicin is excreted by the kidneys mainly by glomerular filtration, with 50-93% 331
of the drug recovered in urine 24 hours after administration in individuals with renal 332
impairment. Due to the immaturity of the newborns most of the processes involved in 333
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 20
20
renal excretion of drugs require several weeks to more than one year to reach maturity 334
thereby influencing the disposition of polar drugs, such as gentamicin (10, 11, 19). 335
For the current population the mean value of gentamicin CL obtained is 0.12 ± 0.01 336
L/h/kg, and comparable to that reported by Taketomo et al. which was 0.1 L/h/kg (7). 337
Drug clearance mainly depends on the functional capacity of the kidneys as well as renal 338
blood flow which increases from 12 mL/min at birth up to 140 mL/min at around one year 339
of age. In this population of infants, CL is influenced by both BW and CLCR. 340
Dosing and TDM recommendations 341
Through the estimation of the pharmacokinetic parameters of gentamicin and their 342
variability it is feasible to propose a dosing scheme based on simulations using the model 343
generated from our infant population. Once-daily dosing is preferred to avoid toxicity. In 344
order to reach a peak value greater than 10 mg/L a dose of 7 mg/kg of gentamicin should 345
be given followed by serum concentration monitoring. According to the simulated 346
population, 6–10 hours after starting the infusion gentamicin concentrations ranging 0.8-2 347
mg/L will be expected. This time range is optimal to ensure detectable serum 348
concentrations of gentamicin; 70% of the simulated infants were over the limit of 349
quantification (0.5 mg/L) of routinely used immunoassays. In fact, the Cmax should not be 350
measured until 2 hours after a 30 minutes infusion to ensure post-distributional samples 351
in 7 mg/kg once daily dosing (21). Despite the post-antibiotic effect, trough serum 352
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 21
21
concentrations < 0.5 mg/L should not last for more than 6 – 8 hours to ensure optimal 353
efficacy of this antibiotic (13, 14, 16, 32). 354
Figure 4 shows mean gentamicin plasma concentrations and percentiles 5 – 95% 355
corresponding to one thousand simulations obtained for the final two-compartment 356
model as generated by NONMEM and using data obtained from a standard infant 6 357
months old, 8 kg of BW, CRs of 0.4 mg/dL, and receiving 7 mg/kg/24 h of gentamicin IV. 358
Data from simulated low-weight 1-month old infant and low Crs 6-months old infant are 359
also shown in the same figure, were median values for plasmatic concentrations are no 360
significantly different from those obtained by a standard infant receiving the same dose. 361
[Insert Figure 4] 362
CONCLUSION 363
For infants, IV gentamicin is best represented by two-compartment 364
pharmacokinetics where total body weight and creatinine clearance exhibit influence on 365
volume of distribution and clearance of the drug. It was feasible to distinguish differences 366
between neonates and infants making it possible to individualize dosage regimens for this 367
population. The dosing scheme proposed is a once-daily dosage of 7 mg/kg followed by 368
therapeutic drug monitoring to avoid toxicity and ensure efficacy with minimal blood 369
sampling.370
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 22
22
ACKNOWLEDGMENTS 371
The authors acknowledge the assistance of the pharmacists, analytical technicians, 372
nurses and medical staff of the Hospital Universitario Severo Ochoa (Leganés, Spain) for 373
their contributions to the present study. 374
FUNDING 375
This study was conducted as part of routine work; no external funding was 376
received to support the current study. Financial support was given to Medellín-Garibay SE 377
by the National Research Council of Science and Technology (CONACYT) from Mexico 378
during her research stay performed in Spain (309952). 379
The authors declare no conflict of interest. 380
381
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 23
23
REFERENCES 382
1. Jackson J, Chen C, Buising K. 2013. Aminoglycosides: how should we use them 383
in the 21st century? Current opinion in infectious diseases 26:516-525. 384
2. Kumana CR, Yuen KY. 1994. Parenteral aminoglycoside therapy. Selection, 385
administration and monitoring. Drugs 47:902-913. 386
3. Warchol ME. 2010. Cellular mechanisms of aminoglycoside ototoxicity. Current 387
opinion in otolaryngology & head and neck surgery 18:454-458. 388
4. Huth ME, Ricci AJ, Cheng AG. 2011. Mechanisms of aminoglycoside ototoxicity 389
and targets of hair cell protection. International journal of otolaryngology 390
2011:937861. 391
5. Gilbert DN. 2005. Aminoglycosides, p. 307-336. In Mandell GL BJ, Dolin R (ed.), 392
Principles and Practice of Infectious Diseases, 6th ed, Churchill Livingstone, New 393
York. 394
6. Dobie RA, Black FO, Pezsnecker SC, Stallings VL. 2006. Hearing loss in patients 395
with vestibulotoxic reactions to gentamicin therapy. Archives of otolaryngology--396
head & neck surgery 132:253-257. 397
7. Taketomo CK, Hodding JH, Kraus DM. 2013. p. 787-790, The Pediatric & 398
Neonatal Dosage Handbook, 20th ed. Leximcomp, Hundson, Ohio. 399
8. Anderson BJ, Holford NH. 2009. Mechanistic basis of using body size and 400
maturation to predict clearance in humans. Drug metabolism and pharmacokinetics 401
24:25-36. 402
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 24
24
9. Administration FaD. 1998. General Considerations for Pediatric Pharmacokinetic 403
Studies for Drugs and Biological Products. In Services USDoHaH (ed.), Rockville, 404
MD. 405
10. Anderson GD, Lynn AM. 2009. Optimizing pediatric dosing: a developmental 406
pharmacologic approach. Pharmacotherapy 29:680-690. 407
11. Touw DJ, Westerman EM, Sprij AJ. 2009. Therapeutic drug monitoring of 408
aminoglycosides in neonates. Clinical pharmacokinetics 48:71-88. 409
12. Milsap RL, Jusko WJ. 1994. Pharmacokinetics in the infant. Environmental health 410
perspectives 102 Suppl 11:107-110. 411
13. Best EJ, Gazarian M, Cohn R, Wilkinson M, Palasanthiran P. 2011. Once-daily 412
gentamicin in infants and children: a prospective cohort study evaluating safety and 413
the role of therapeutic drug monitoring in minimizing toxicity. The Pediatric 414
infectious disease journal 30:827-832. 415
14. Contopoulos-Ioannidis DG, Giotis ND, Baliatsa DV, Ioannidis JP. 2004. 416
Extended-interval aminoglycoside administration for children: a meta-analysis. 417
Pediatrics 114:e111-118. 418
15. Rakhmanina NY, van den Anker JN. 2006. Pharmacological research in 419
pediatrics: From neonates to adolescents. Advanced drug delivery reviews 58:4-14. 420
16. Dersch-Mills D, Akierman A, Alshaikh B, Yusuf K. 2012. Validation of a dosage 421
individualization table for extended-interval gentamicin in neonates. The Annals of 422
pharmacotherapy 46:935-942. 423
17. Turnidge J. 2003. Pharmacodynamics and dosing of aminoglycosides. Infectious 424
disease clinics of North America 17:503-528, v. 425
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 25
25
18. Mould DR, Upton RN. 2013. Basic concepts in population modeling, simulation, 426
and model-based drug development-part 2: introduction to pharmacokinetic 427
modeling methods. CPT: pharmacometrics & systems pharmacology 2:e38. 428
19. Marsot A, Boulamery A, Bruguerolle B, Simon N. 2012. Population 429
pharmacokinetic analysis during the first 2 years of life: an overview. Clinical 430
pharmacokinetics 51:787-798. 431
20. Schwartz GJ, Haycock GB, Edelmann CM, Jr., Spitzer A. 1976. A simple 432
estimate of glomerular filtration rate in children derived from body length and 433
plasma creatinine. Pediatrics 58:259-263. 434
21. Demczar DJ, Nafziger AN, Bertino JS, Jr. 1997. Pharmacokinetics of gentamicin 435
at traditional versus high doses: implications for once-daily aminoglycoside dosing. 436
Antimicrobial agents and chemotherapy 41:1115-1119. 437
22. Mould DR, Upton RN. 2012. Basic concepts in population modeling, simulation, 438
and model-based drug development. CPT: pharmacometrics & systems 439
pharmacology 1:e6. 440
23. Sheiner LB, Beal SL. 1981. Some suggestions for measuring predictive 441
performance. Journal of pharmacokinetics and biopharmaceutics 9:503-512. 442
24. Williams PJ, Kim YH, Ette EI. 2007. The Epistemology of Pharmacometrics, p. 443
223-245. In Ette EI, Williams PJ (ed.), Pharmacometrics: the science of quantitative 444
pharmacology. John Wiley & Sons, New Jersey, U.S.A. 445
25. Organization WH. 2006. WHO Multicentre Growth Reference Study Group. 446
WHO Child Growth Standards: Length/height-for-age, weight-for-age, weight-for-447
length, weight-for-height and body mass index-for-age: Methods and development. 448
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 26
26
26. De Cock RF, Piana C, Krekels EH, Danhof M, Allegaert K, Knibbe CA. 2011. 449
The role of population PK-PD modelling in paediatric clinical research. European 450
journal of clinical pharmacology 67 Suppl 1:5-16. 451
27. Xuan D, Nicolau DP, Nightingale CH. 2004. Population pharmacokinetics of 452
gentamicin in hospitalized patients receiving once-daily dosing. International 453
journal of antimicrobial agents 23:291-295. 454
28. Garcia B, Barcia E, Perez F, Molina IT. 2006. Population pharmacokinetics of 455
gentamicin in premature newborns. The Journal of antimicrobial chemotherapy 456
58:372-379. 457
29. Koren G. 1997. Therapeutic drug monitoring principles in the neonate. National 458
Academy of CLinical Biochemistry. Clinical chemistry 43:222-227. 459
30. Miron D. 2001. Once daily dosing of gentamicin in infants and children. The 460
Pediatric infectious disease journal 20:1169-1173. 461
31. Bass KD, Larkin SE, Paap C, Haase GM. 1998. Pharmacokinetics of once-daily 462
gentamicin dosing in pediatric patients. Journal of pediatric surgery 33:1104-1107. 463
32. Kraus DM, Pai MP, Rodvold KA. 2002. Efficacy and tolerability of extended-464
interval aminoglycoside administration in pediatric patients. Paediatric drugs 4:469-465
484. 466
467
468
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 27
27
TABLES 469
Table 1. General characteristics of the populations assayed. 470
Variable
Population group
(n=208)
Validation group
(n=55)
Age (months) 5.8 ± 4.8 5.5 ± 4.4
BW (kg) 6.4 ± 2.2 6.7 ± 2.3
Height (cm) 62.6 ± 9.1 62.8 ± 8.8
BMI (kg/m2) 15.8 ± 2.1 16.5 ± 2.7
CRS (mg/dL) 0.42 ± 0.1 0.44 ± 0.2
CLCR (mL/min/1.73 m2) 76.7± 36.9 76.7 ± 44.3
Data are shown as mean ± standard deviation. BW (body weight), BMI (body mass 471
index), CRS (serum creatinine), CLCR (creatinine clearance). 472
473
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 28
28
Table 2. Pharmacokinetic parameters and inter-individual variability values obtained for 474
the base two-compartment open model without the inclusion of covariates. 475
PK MODEL PARAMETER Mean ± SE
Two-compartment open model (OBJ = 520.85) (AIC = 528.85)
CL (L/h) Ө1 0.83 ± 0.05
Vc (L) Ө2 1.9 ± 0.06
Q (L/h) Ө3 0.31 ± 0.06
Vp (L) Ө4 4.9 ± 1.0
Interindividual variability associated to CL (%)
ω2CL 59.3 ± 22.4
Interindividual variability associated to Vc (%)
ω2Vc 43.8 ± 14.1
Residual variability (%) σ 12.0 ± 4.0
476
477
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 29
29
Table 3. Pharmacokinetic parameters and interindividual variability values obtained for 478
the final two-compartment open model and Bootstrap results (n = 200). 479
PK MODEL PARAMETER MEAN ± SE BOOTSTRAP
MEDIAN PERCENTILES
Two-compartment open model (OBJ = 197.27) 2.5th 97.5th
CL = Ө1*BW+ Ө5 * (CLCR/75) (L/h•kg)
Ө1 0.12 ± 0.01 0.12 0.09 0.13
Ө5 0.06 ± 0.11 0.10 0.007 0.24
Vc = Ө2 * BW (L/kg) Ө2 0.35 ± 0.02 0.34 0.27 0.37
Q (L/h) Ө3 0.23 ± 0.05 0.24 0.15 0.45
Vp (L) Ө4 2.3 ± 1.0 2.9 1.3 6.7
Interindividual variability associated to CL
ω2CL 26.4 ± 14.1 28.3 22.4 34.6
Interindividual variability associated to Vc
ω2Vc 26.5 ± 14.0 24.5 20.0 33.2
Residual variability σ 11.2 ± 2.9 11.0 8.1 12.5
480
481
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 30
30
Table 4. Mean (± standard deviation) prediction errors and 95% CIs estimated for the 482
validation group with the base and final two-compartment open model. 483
484
Base model
Mean ± SD
(95% CI)
Final model
Mean ± SD
(95% CI)
MPE 0.3± 2.5
(-0.3, 0.8)
-0.2 ± 1.5
(-0.6, 0.1)
APE 1.6 ± 1.9
(1.2, 2.0)
0.9 ± 1.2
(0.7, 1.2)
MSPE 6.1 ± 11.6
(3.6, 8.1)
2.3 ± 5.7
(1.0, 3.5)
RMSE 2.5
(1.9, 2.8)
1.5
(1.0, 1.9)
MPE (mean prediction error), APE (absolute prediction error), MSPE 485
(mean-squared prediction error), RMSE (root-mean-squared prediction 486
error). 487
488
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 31
31
FIGURES 489
490
Figure 1. Gentamicin serum concentrations versus blood sampling times (n = 335) for the 491
population study group. 492
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 32
32
493
494
Figure 2. Scatter plots of predicted (PRED) versus observed (DV) gentamicin 495
concentrations (including the identity line) for the base (a) and final (b) two-compartment 496
open model. Scatter plots of conditional weighed residuals (CWRES) versus BW (including 497
the identity line) for the base (c) and final (d) two-compartment open model, and 498
corresponding to the population study group. 499
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 33
33
500
Figure 3. Scatter plot of predicted (PRED) versus observed (OBS) gentamicin 501
concentrations corresponding to the validation group and obtained for the final two-502
compartment open model. 503
504
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from
Page 34
34
505
506
Figure 4. Plasma concentrations shown as median and 5 – 95% percentiles for a 6-months 507
old infant receiving 7 mg/kg of gentamicin IV. Median values for a simulated 1-month old 508
infant with low BW; and data from simulated infant with low Crs are also depicted (n = 509
1000 infants simulated for each situation). 510
on January 29, 2018 by guesthttp://aac.asm
.org/D
ownloaded from