J Antimicrob Chemother 2022; 77: 448–456

https://doi.org/10.1093/jac/dkab413 Advance Access publication 23 November 2021

Simultaneous pharmacokinetic/pharmacodynamic (PKPD) assessment

of ampicillin and gentamicin in the treatment of neonatal sepsis
Silke Gastine1*, Christina Obiero2, Zoe Kane 1,3
, Phoebe Williams2,4, John Readman1, Sheila Murunga3,
Johnstone Thitiri , Sally Ellis , Erika Correia , Borna Nyaoke6, Karin Kipper7, John van den Anker8,9,
3 5 5

Mike Sharland10, James A. Berkley3,4,11 and Joseph F. Standing 1,12

1
Infection, Immunity and Inflammation, Great Ormond Street Institute of Child Health, University College London, London, UK;
2
KEMRI-Wellcome Trust Research Programme, Kilifi, Kenya; 3Quotient Sciences, Mere Way, Ruddington, Nottingham, UK; 4Centre for
Tropical Medicine & Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, UK; 5Global Antibiotic Research &
Development Partnership (GARDP), Genève, Switzerland; 6Drugs for Neglected Diseases Initiative (DNDi), Nairobi, Kenya; 7Institute of
Chemistry, University of Tartu, Tartu, Estonia; 8Department of Paediatric Pharmacology and Pharmacometrics, University Children’s
Hospital Basel, University of Basel, Switzerland; 9Division of Clinical Pharmacology, Children’s National Hospital, Washington, DC, USA;
10
Paediatric Infectious Diseases Research Group, Institute for Infection and Immunity, St. George’s, University of London, London, UK;
11
The Childhood Acute Illness & Nutrition (CHAIN) Network, Nairobi, Kenya; 12Pharmacy Department, Great Ormond Street Hospital for
Children, NHS Foundation Trust, London, UK

*Corresponding author. E-mail: [email protected]

Received 17 May 2021; accepted 6 October 2021

Objectives: This study aimed to simultaneously investigate the pharmacokinetics of ampicillin and gentamicin,
currently the WHO standard of care for treating neonatal sepsis.
Methods: Pharmacokinetic data were collected in 59 neonates receiving ampicillin and gentamicin for
suspected or proven sepsis in the NeoFosfo trial (NCT03453177). A panel of 23 clinical Escherichia coli isolates
from neonates with sepsis, resistant to either ampicillin, gentamicin or both, were tested for susceptibility using
chequerboards. Pharmacokinetic/pharmacodynamic (PKPD) modelling and simulations were used to compare
single-agent (EUCAST MIC) and combination (chequerboard MIC) target attainment with standard dosing
regimens.
Results: A model was established that simultaneously estimated parameters of a one-compartment ampicillin
model and a two-compartment gentamicin model. A common clearance for both drugs was used
(6.89 L/h/70 kg) relating to glomerular filtration (CLGFR), with an additional clearance term added for ampicillin
(5.3 L/h/70 kg). Covariate modelling included a priori allometric weight and post-menstrual age scaling of
clearance. Further covariate relationships on renal clearance were postnatal age and serum creatinine.
Simulation-based PKPD assessments suggest good Gram-positive (MIC  0.25 mg/L) cover. However, less
than one-quarter of neonates were predicted to receive efficacious coverage against Enterobacterales
(MIC  2 mg/L). The benefit of the ampicillin/gentamicin combination was limited, with only 2/23 E. coli clinical
strains showing FIC index < 0.5 (synergy) and most in the range 0.5–1 (suggesting additivity). Simulations
showed that feasible dosing strategies would be insufficient to cover resistant strains.
Conclusions: PKPD simulations showed ampicillin and gentamicin combination therapy was insufficient to cover
Enterobacterales, suggesting the need for alternative empirical treatment options for neonatal sepsis.

Introduction carrying the largest burden.1 Microbial patterns vary globally.

In neonatal care, infectious diseases associated with antimicrobial In high-income countries, primary strains are Escherichia coli and
resistance (AMR) are of increasing concern and need to be group B Streptococcus (GBS, Streptococcus agalactiae),2–4 but
addressed on a global scale. Infection is one of the major causes of in low- and middle-income countries (LMICs) they are Klebsiella
neonatal deaths worldwide, with Asia and sub-Saharan Africa spp., E. coli and Staphylococcus aureus.

C The Author(s) 2021. Published by Oxford University Press on behalf of the British Society for Antimicrobial Chemotherapy.
V
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/
licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.
448
Neonatal ampicillin and gentamicin PopPK JAC
The WHO’s current recommended treatment for neonatal be significant at a level of P < 0.01 and >3.84 at a level of P < 0.05, for a
sepsis is a narrow-spectrum b-lactam agent in combination with change in 1 degree of freedom using the chi-squared distribution,
gentamicin.5,6 Combination of gentamicin with ampicillin has an respectively.
acceptable safety profile, but the limited access to therapeutic Allometric scaling of body weight was included using a fixed exponent
drug monitoring to monitor gentamicin exposure can be seen as a of 0.75 on clearance terms and linear scaling on volume terms.14 To com-
major safety concern in LMICs. Although resistance to ampicillin pare parameter estimates with other paediatric and adult studies, parame-
ters were scaled to a standard weight of 70 kg. A previously published
and gentamicin has been shown,7,8 in particular for Gram-
neonatal renal maturation function by Rhodin et al.15 was added to clear-
negative clinical isolates in Asia and Africa, this combination
ance a priori.10,16
therapy remains a vital part of standard-of-care (SOC) neonatal
Further renal clearance maturation after birth, regardless of gestational
antibiotic therapy.4,9 age, which occurs over the first few days/weeks of life10 was implemented.
Both drugs have previously been studied as monotherapy.10–12 This effect on clearance was detected by Kane et al.13 through analysis of
Pharmacokinetics/pharmacodynamics (PKPD) of the combination the fosfomycin data for this study (Equation 1) and was thus also tested
regimen and the degree to which the PK of both drugs are corre- in the ampicillin and gentamicin model.
lated within a patient and whether target attainment may differ
when accounting for both agents together, however, has not been PNAfunction ¼ hM þ ð1  hM Þ  ð1  ePNAi hN Þ (1)
described so far, to the best of our knowledge. We therefore
analysed ampicillin and gentamicin plasma concentrations from where PNAfunction is the introduced covariate relationship, hM is the popula-
the PK samples taken in the recent NeoFosfo study,13 where both tion value relating to PNA maturation, PNAi is the individual’s postnatal age
drugs were given as SOC treatment alongside fosfomycin. A PKPD and hN is the population value of the shape parameter. The ability of serum
model was established and evaluated with microbiological syn- creatinine concentration (SCR) to predict individual variability in renal clear-
ergy data from clinical strains. ance was tested according to Equation 2, where hSCr is the population value
of the detected creatinine effect and SCRi is the individual’s creatinine
measurement.
Methods
Study and drug details SCRi hSCr
SCRfunction ¼ (2)
An open-label randomized controlled trial was conducted to assess the TSCR
safety and PK parameters of IV followed by oral fosfomycin.13 In this trial,
The impact of deviation from age-related standard was tested on renal
fosfomycin was administered together with SOC antibiotics (ampicillin plus
gentamicin) and compared with SOC alone among hospitalized neonates clearance (Equations 1 and 2). The measured SCR was standardized to
with clinical sepsis. The NeoFosfo trial was conducted at Kilifi County typical serum creatinine concentrations (TSCR) for the respective age calcu-
Hospital (KCH) between March 2018 and March 2019. lated according to Ceriotti et al. (Equation 3).17
Ampicillin (50 mg/kg, administered 12 hourly for participants younger
than 7 days and 8 hourly if more than 7 days old), plus 24 hourly gentamicin TSCRðlmolÞ ¼ 2:37330  12:91367  lnðPNAyears Þ þ 23:93581

0:5
(3 mg/kg, for participants less than 2 kg, or 5 mg/kg if more than 2 kg) were  PNAyears (3)
prescribed as SOC antibiotics at admission.

PK sampling Model evaluation

Patients allocated to the SOC plus fosfomycin (SOC-F) arm were randomly Goodness-of-fit (GOF) plots and VPCs served as tools to evaluate improve-
allocated to one of three early (5, 30 or 60 min) and one of three late time- ment in model appropriateness and predictivity throughout model devel-
points (2, 4 or 8 h) for PK sample collection after the first parenteral and the opment steps and to assess the final model. A non-parametric bootstrap
first oral fosfomycin dose. A fifth sample was collected after the final dose (n = 1000) was performed on the final model to test parameter precision
of oral fosfomycin for participants still hospitalized on Day 7. Sample and robustness. Perl-speaks-NONMEM (PsN) was used in the bootstrap
processing and the analytical assay are described in the Supplementary
analysis and VPCs.
methods, available as Supplementary data at JAC Online.

Combination antibiotic testing

Data analysis and software
Clinical E. coli strains collected from neonates with sepsis18 from centres
Data pre-analysis and graphical output was created in R (version 3.6.1; R
with high prevalence of ESBL-producing Enterobacteriaceae were analysed
foundation for statistical computing, Vienna, Austria). Model building was
for antibiotic susceptibility when challenged with ampicillin, gentamicin
performed using NONMEM (Version 7.4; ICON Development Solutions,
and the combination of both drugs. The EUCAST disc diffusion test,19 with
Ellicott City, MD, USA) with the FOCE ! I estimation method. Visual predict-
reference to EUCAST clinical MIC breakpoints,20 was used to determine the
ive checks (VPCs) were performed with Perl-speaks-NONMEM (PsN, version
susceptibility of each strain to each drug.
4.8.1). Graphical output was created in R.
MICs for each strain were determined by broth microdilution meth-
ods in accordance with BSAC guidance.19 MICs were determined for the
PK model building single drugs ampicillin and gentamicin, as well as for the combination of
For each individual drug, one- and two-compartment structural models both drugs through synergy testing. Synergistic relationships were
were assessed. Inter-individual variability (IIV) was assumed to follow a tested against single E. coli isolates using a 96-well plate standard 2D
log-normal distribution and was tested on all parameters. An additive, chequerboard assay.20 The FIC index (FICI) was calculated and reported
a proportional and a combined error model were tested. For nested as synergistic when <0.5, additive when within the range of 0.5–2, and
models, a decrease in #2 times the log likelihood of >6.64 was needed to antagonistic when >2. The presence of commonly found resistance

449
Gastine et al.

genes was established by multiplex PCR with previously described measured below the lower limit of quantification (LLOQ) for ampicillin
primers and amplification conditions.21,22 (0.5 mg/L; n = 16) and gentamicin (0.1 mg/L; n = 18), respectively.

PK simulations PK model development

Simulations were carried out to explore ampicillin and gentamicin PTA using
Ampicillin was best described by a one-compartment model with linear
the registered dose and frequency as stated by the WHO Pocket Book of
elimination when modelled alone. Gentamicin was best described by a
Hospital Care for Children.5 Ampicillin was simulated at 50 mg/kg with a fre-
two-compartment model with linear elimination when modelled alone. IIV
quency of q12h for PNA  7 days and q8h for PNA > 7 days. Gentamicin was
was supported on clearance for ampicillin and clearance and central vol-
simulated with a single daily dose of 5 mg/kg/day in neonates with
ume of distribution (Vd) for gentamicin.
PNA  7 days weighing above 1.5 kg. Gentamicin dose was decreased to
Allometric scaling with fixed exponents of 1 and 0.75 for volume and
3 mg/kg/day in low-birthweight neonates and increased to 7.5 mg/kg/day
clearance terms, respectively, as well as including the Rhodin15 maturation
in neonates older than 7 days. Both drugs were simulated to be infused
function (Equation 3) on clearance improved each model fit.
over 5 min. As ampicillin dose recommendations vary across formularies23
and severity of infection, additional simulations were also carried out with The two individual models were subsequently linked in a single model
the highest proposed dose (100 mg/kg) using infusion rates of 5 min and, to with covariance estimated. To optimize the linked PK model, a common fil-
explore prolonged infusion, also with infusion rates of 1 h, 2 h and continu- tration clearance term was estimated for ampicillin and gentamicin, with
ous infusion. an additional ampicillin clearance accounting for non-renal pathways,
rather than having a separate clearance estimate for each drug (change in
The simulation population (n = 10 000) was created using observed
objective function value, DOFV = #1.5).
demographics from the present study combined with data from an inter-
national multi-centre neonatal observational study (NCT03721302).
Simulations were carried out in NONMEM with post-processing in R. Table 1. Study population demographics
Simulated PD targets were compared with EUCAST breakpoints.24
Overall target attainment was studied through graphical evaluation of the Parameter SOC arm (n = 59)
fraction of time above MIC at steady-state (%fT>MIC)25 for ampicillin and
Age, median (range)
the maximal concentration to MIC ratio (Cmax/MIC) for gentamicin at Day 2
PNA (days) 1 (0–23)
of treatment.26 In agreement with common practice, Cmax was defined as
Gestational age (weeks) 38 (34–44)
the gentamicin concentration 30 min after the end of injection. Simulation
results were also compared with results of the antibiotic synergy 2D Sex, n (%)
chequerboard study. PTA was calculated for ampicillin and gentamicin Female 24 (41)
MICs observed in combination with the respective other drug. Male 35 (59)
Covariates, median (range)
Weight (g) 2800 (1560–5670)
Results Creatinine at Day 0 (lmol/L) 96.5 (35–142)
Patients and demographics Study design
Total number of samples, n 273
The final modelling dataset included 373 (181 ampicillin and 192 gentami-
Samples per patient, median (range) 4 (3–5)
cin) samples from 59 patients (Figure 1). Demographics and covariate distri-
butions are displayed in Table 1. A total of 34 concentrations were

Figure 1. Observed plasma concentration versus time since last dose for ampicillin (left) and gentamicin (middle). Each line represents a single
patient’s profile. Correlation of observed ampicillin and gentamicin concentrations is shown on the right.

450
Neonatal ampicillin and gentamicin PopPK JAC
Table 2. Final model parameter estimates

Parameter Parameter estimates (RSE, %) IIV (%) (RSE, %) Bootstrap 95% CI

CLGFR (L/h) 6.89 (8) 38.5 (14) 5.48–8.01

Central VdGent (L) 26.4 (18) 40.6 (40) 9.81–31.2
QGent (L/h) 0.956 (28) 0.48–10.3
Peripheral VdGent (L) 6.69 (36) 3.10–16.3
CLnon-GFR (L/h) 5.3 (11) 45.4 (16) 4.25–6.54
Central VdAmp (L) 52.7 (11) 42.6–65.4
PNAfract (%) 44.9 FIX
PNAcoeff 0.117 FIX
Creatinine on CLGFR #0.423 (53) #1 to 0.05
Proportional error (%), gentamicin 29.3 (24) 20.6–34
Proportional error (%), ampicillin 53.6 (13) 46.6–60.4

Q, intercompartmental clearance; PNAfract, fraction of adult clearance at PNA; PNAcoeff, coefficient for PNA on clearance; FIX, parameter was fixed and
not estimated.

Figure 2. VPC for joint ampicillin and gentamicin model. Black dots, observed values; solid line, median; dashed lines, 5th and 95th percentiles for the
observed values; grey areas, 95% prediction interval of respective median and percentiles.

Subsequently, the effect of PNA on the combined clearance term, as Ampicillin dosing recommendations vary across guidelines and depend
proposed by Kane et al.,13 was included using a fixed fraction of clearance on the severity of the infection.23 Figure S2 shows the difference in target
at birth and fixed coefficient for clearance maturation (DOFV = #32.5). The attainment of the upper and lower end of dose recommendations (50 and
effect of creatinine in relation to the age-appropriate common creatinine 100 mg/kg dosing) when given by 5 min short infusion. Figure S3 shows the
levels according to Ceriotti et al.17 was estimated as an effect on the com- scenario of extending the infusion time from 5 min to 1 h and 2 h along with
bined clearance term and resulted in a significant model improvement a continuous infusion. Here, only the continuous infusion scenario was able
(DOFV = #7.0; P < 0.01). to bring the population median across the Enterobacterales’ MIC, with
Parameter estimates for the final model along with bootstrap results the green shaded area that represents the 90% prediction interval for the
simulated population still not entirely crossing the Enterobacterales MIC for
are shown in Table 2. VPCs for ampicillin and gentamicin from the final
50 mg/kg. For 100 mg/kg, a continuous infusion scenario was able to cover
combined model are shown in Figure 2.
the entire population.
A Cmax/MIC of 10 is regarded to best correlate with aminoglycoside anti-
Simulation results bacterial efficacy. Target attainment across the MIC range for 2 days of
gentamicin therapy is shown in Figure 4, along with the EUCAST breakpoint
%fT>MIC across MIC ranges, including EUCAST breakpoints (Gram-positive for Enterobacterales. Figure S4 translates this into a PTA for the simulated
strains, MIC  0.25 mg/L; Enterobacterales, MIC  2 mg/L) is shown for population. Both figures show that only a small fraction of low-birthweight
ampicillin in Figure 3. This was translated into PTA when aiming for neonates is covered against the Enterobacterales EUCAST breakpoint at
100%fT>MIC in Figure S1.

451
Gastine et al.

Figure 3. Target attainment as %fT>MIC against MICs for simulated ampicillin regimen. Solid line, population median; grey area, 90% prediction inter-
val; dashed lines, EUCAST breakpoints for Enterobacterales and CoNS.

Figure 4. Target attainment as Cmax/MIC ratio against MICs for simulated gentamicin regimen. Solid line, population median; grey area, 90% predic-
tion interval; dashed line, EUCAST breakpoint for Enterobacterales.

2 mg/L MICs. The PTA increased with the increased doses suggested for increasing renal clearance, most of the population, however, show
older and heavier neonates, yet even the 7.5 mg/kg dose in neonates sufficient trough concentrations by 2 months.
older than 7 days only covers the population median. In contrast, Figure
S5 shows the probability of reaching the toxicity target that limits the Combination effect assessment through in vitro 2D
aminoglycoside therapeutic window. Aminoglycoside toxicity is linked
to insufficient drug clearance thus trough concentrations should fall
chequerboard assay
below 1 mg/L before the next dose is administered. Fewer than 25% of Coverage for both organisms in relation to collected E. coli MICs from clinical
the simulated neonates achieve trough concentrations falling below specimens is shown in Figure 5. Ampicillin 100%fT>MIC and gentamicin
this target during the first days of life. With renal maturation and Cmax/MIC ratio of >10 were calculated for the isolates’ combined MICs.

452
Neonatal ampicillin and gentamicin PopPK JAC

Figure 5. Resistance testing, FICI results and PTA for coverage of MICs for neonatal clinical specimens with the respective combination treatment.
Light grey bars, gentamicin; dark grey bars, ampicillin.

Sensitive and intermediate-resistant strains were estimated to have We used a combined PK modelling approach to assess the
partial to full response when treated in this neonatal target population. common ampicillin/gentamicin SOC regimen’s PK performance in
The magnitude of additive (21/23 strains with FICI between 0.5 and 1) and neonates. We found a one-compartment model to best describe
synergistic effects (2/23 strains with FICI < 0.5) between the combination
ampicillin PK and a two-compartment model to best describe gen-
was insufficient to cover strains resistant to both agents.
tamicin PK. Both models were linked after introducing allometry.
Elimination of both drugs depends on passive filtration, which is
Discussion thought to be the main method of clearance for gentamicin, and
We here present a combined ampicillin and gentamicin population contributes to approximately half for ampicillin. In the initial com-
pharmacokinetic (popPK) analysis in neonatal sepsis. Our major bined model the estimated covariance between the sole
finding is that the current empirical guidance for neonatal sepsis clearances was around 60%. Therefore, ampicillin clearance was
appropriately manages the most common Gram-positive patho- set to the gentamicin clearance (assumed filtration component)
gens responsible for neonatal sepsis with MICs of 0.25 mg/L but plus an estimated non-filtration component. Upon making this as-
not common Gram-negative Enterobacterales strains with MICs of sumption, the covariance in the final model between the filtration
2 mg/L. This includes the most frequently detected Gram- and non-filtration pathways was estimated to be low at 30%, with
positive organism in the LMIC setting, S. aureus, which the WHO the non-filtration component of ampicillin clearance being around
Pocket Book is targeting (along with GBS and CoNS). These are 43% of the total, which is similar to previous reports.31–33
more likely to be found in infections from vertical transmission, The assumption that gentamicin is predominantly cleared by
premature neonates or neonates undergoing more invasive treat- glomerular filtration was confirmed by the estimated clearance of
ments. However, Enterobacterales strains, including Klebsiella spp. 6.89 L/h/70 kg (CLGFR) corresponding to a normal adult glomerular
and E. coli, are not well covered (Figures 3 and 4).27–30 When analy- filtration rate (GFR) ( 115 mL/min). For ampicillin, almost half of
sing clinical E. coli strains, we observe that despite finding additivity the clearance process can be attributed to other pathways.31–33
through in vitro 2D chequerboard assays for most tested strains Sjovall et al.32 determined that 80% of ampicillin can be found un-
(0.5  FICI  1) and two strains showing a trend towards synergy changed in the urine and that 49% of this excretion can be attrib-
(FICI < 0.5) for the combination, this did not bring MICs low enough uted to tubular secretion. The entire elimination pathway of
such that clinically achievable doses would provide adequate ampicillin, however, remains unknown in the neonatal population,
coverage (Figure 5). which is why we describe this with a glomerular filtration-

453
Gastine et al.

associated term and a second non-filtration term in our model, gentamicin dosing would need to be accompanied by extending
which accounts for these additional pathways. Maturation the dose interval beyond the simulated 24 h window, as gentami-
functions and the effect of serum creatinine normalized to the cin toxicity is linked to insufficient gentamicin clearance and a
age-dependent expected creatinine levels through the function trough concentration of <1 mg/L is used to guide adequate dose
from Ceriotti et al.17 was subsequently added as a covariate on frequency10 (Figure S5). During the first week of life the probability
the filtration-related clearance term only and this resulted in of reaching this trough increases, but still leaves more than half
ampicillin clearance estimates close to previously estimated adult the simulated population with a trough level above the threshold.
values.11 With the restricted access to therapeutic drug monitoring for neo-
Wang et al.33 have recently evaluated the application of matur- natal care in LMICs, this poses a safety concern in this vulnerable
ation functions to predict neonatal drug clearance, including pre- population. Bolus ampicillin dosing showed less than 30% PTA
dictions for ampicillin. They concluded that drugs that are only across the simulated population. Adequate ampicillin fT>MIC is only
partially cleared through glomerular filtration are less reliably pre- reached when applied as continuous infusion along with increas-
dicted through these scaling methods. For ampicillin, this meant ing the total daily dose (Figure S3), which is seldom feasible in the
Wang et al.33 predicted clearances being twice as high as the neonatal population and in a global setting.
observed clearance. In our model, when only applying maturation A neglected topic in antibiotic susceptibility testing is the possi-
to the filtration part, overall ampicillin clearance was estimated at bility of synergy. As b-lactamases are proteins and aminoglyco-
12 L/h/70 kg, which is very much in line with adult ampicillin clear- sides inhibit protein synthesis, a reasonable hypothesis is that the
ance stated in the summary of product characteristics (11.64 L/h)
combination therapy may be synergistic. However, despite FICIs
and observed in adult PK studies (10.7 L/h).11 This again highlights
being <1, suggesting at least additivity, only 2/23 tested were
the need to assess the feasibility of extrapolation when applying
<0.5 and so could be classed as synergistic. In Figure 5, where
covariate modelling to special population studies, especially when
E. coli strains extracted from clinical specimens were analysed
a priori effects are implemented. For other b-lactams, the use of
based on the results of 2D chequerboard synergy testing, the
generic maturation functions to describe exposure in neonates
interaction effect was insufficient to restore activity for strains
generally gives good predictions according to Wang et al.,33 who
successfully predicted meropenem clearance in neonates, and a where one or both drugs of the combination were resistant.
recent study on patients in neonatal, paediatric and adult ICUs Even the two strains that showed ampicillin susceptibility
showed a common maturation function was applicable to amoxi- and intermediate gentamicin susceptibility were not covered
cillin, benzylpenicillin, cefotaxime, meropenem and piperacillin/ throughout the entire neonatal population. Strains resistant to
tazobactam.34 It therefore seems that the non-filtration compo- both drugs (n = 8) did not show any coverage, with PTA being
nent of ampicillin clearance is close to maturity at birth and implies <1%. From these results, together with the explored modifica-
that higher doses or substitution for amoxicillin should be tions to the dosing regimen, as stated above, the ability to over-
considered. come these resistances does not seem feasible with this
The major limitation of our analysis is that the NeoFosfo study’s combination.
PK sampling was optimized to assess the plasma concentration Neonatal sepsis due to Gram-negative and MDR organisms
trajectories of the primary investigational drug (fosfomycin). with poor response to current empirical regimens is an increasingly
Ampicillin and gentamicin treatments were started at different significant global health threat.1,35 Current SOC regimens, includ-
timepoints compared with the investigated fosfomycin therapy, ing the studied WHO recommendation (ampicillin and gentamicin)
resulting in a broad range of sampling timepoints. This and the fact are now redundant in many areas and new regimens need to be
that more attention is usually paid to the documentation of the in- found, or older, less commonly used antibiotics re-evaluated.36,37
vestigational drug rather than the SOC drugs given may have led Combination therapy, which aims to cover multiple modes of anti-
to some uncertainty in this evaluation. We tried to overcome this, bacterial action, needs to be further explored not only for antibac-
however, through data cleaning prior to the analysis and address- terial efficacy, but also through analysis of combined PK behaviour.
ing uncertainties with the study site. In addition, whilst the in vitro For neonatal sepsis, the next step is exploring neonatal combined
chequerboard analysis was undertaken with organisms isolated PK, safety and efficacy of combinations of other drugs with
from neonates with sepsis and having a range of resistance genes improved coverage against MDR pathogens.
(Figure 5), these isolates came from European hospitals, and
hence may not be truly representative of the African setting.
Conclusions
Target attainment simulations with WHO empirical treatment
dose guidance for neonatal sepsis35 using the final model pre- A combined model describing ampicillin and gentamicin PK
dicted that Gram-positive coverage throughout the neonatal through joint renal filtration clearance and an additional non-
population would be provided by the regimen’s b-lactam compo- filtration term describes the observed neonatal sepsis popula-
nent ampicillin (Figure 3 and Figure S1) when considering a tion well. PKPD assessments suggest that good Gram-positive
common EUCAST24 clinical breakpoint of 0.25 mg/L to reflect coverage throughout the neonatal population is provided by the
susceptible GBS, CoNS and S. aureus infections. Current dosing regimen’s b-lactam component ampicillin. Less than one-
guidance, however, provides poor coverage against common quarter of neonates were, however, provided with efficacious
Gram-negative Enterobacterales pathogens. Our analysis reveals coverage against Enterobacterales. When analysing a range of
that less than one-quarter of neonates treated with this regimen exemplar clinical E. coli strains displaying resistance to one or
receive efficacious coverage against the gentamicin EUCAST24 both agents, we found only a very weak trend towards synergy,
clinical breakpoint of 2 mg/L (Figure 4 and Figure S4). Increasing with mostly additive effects according to FICI, indicating the

454
Neonatal ampicillin and gentamicin PopPK JAC
need to consider alternative antibiotics for empirical therapy in 7 Chaurasia S, Sivanandan S, Agarwal R et al. Neonatal sepsis in South Asia:
areas where the burden of antibiotic resistance is high. huge burden and spiralling antimicrobial resistance. BMJ 2019; 364: k5314.
8 Iroh Tam PY, Musicha P, Kawaza K et al. Emerging resistance to empiric
antimicrobial regimens for pediatric bloodstream infections in Malawi (1998-
2017). Clin Infect Dis 2019; 69: 61–8.
Acknowledgements 9 Sharland M, Pulcini C, Harbarth S et al. Classifying antibiotics in the WHO
We thank the study participants and parents who took part in the study, Essential Medicines List for optimal use - be AWaRe. Lancet Infect Dis 2018;
along with all the healthcare professionals involved in study delivery. 18: 18–20.
10 Germovsek E, Kent A, Metsvaht T et al. Development and evaluation of a
gentamicin pharmacokinetic model that facilitates opportunistic gentamicin
Funding therapeutic drug monitoring in neonates and infants. Antimicrob Agents
Chemother 2016; 60: 4869–77.
The NeoFosfo clinical trial was sponsored by DNDi/GARDP with funding
from the German Federal Ministry of Education and Research (BMBF), 11 Soto E, Shoji S, Muto C et al. Population pharmacokinetics of ampicillin
German Ministry of Health, South African Medical Research Council, and sulbactam in patients with community-acquired pneumonia: evaluation
Department for International Development (DFID) UK, Ministry of Health, of the impact of renal impairment. Br J Clin Pharmacol 2014; 77: 509–21.
Welfare and Sport of the Netherlands and Médecins Sans Frontières. 12 Nahata MC, Vashi VI, Swanson RN et al. Pharmacokinetics of ampicillin
Support at the institution level came from the National Institute for and sulbactam in pediatric patients. Antimicrob Agents Chemother 1999; 43:
Health Research Biomedical Research Centre (NIHR BRC) at Great 1225–9.
Ormond Street Hospital for Children NHS Foundation Trust and University 13 Kane Z, Gastine S, Obiero C et al. IV and oral fosfomycin pharmacokinetics
College London. J.S. was supported by a United Kingdom MRC Fellowship in neonates with suspected clinical sepsis. J Antimicrob Chemother 2021; 76:
(Grant MR/M008665/1). J.A.B. was supported by the MRC/DFID/Wellcome 1855–64.
Joint Global Health Trials scheme (Grant MR/M007367/1). Laboratory
14 Holford N, Heo YA, Anderson B. A pharmacokinetic standard for babies
work carried out by K.K. was supported by Analytical Services
International Ltd. and adults. J Pharm Sci 2013; 102: 2941–52.
15 Rhodin MM, Anderson BJ, Peters AM et al. Human renal function matur-
ation: a quantitative description using weight and postmenstrual age. Pediatr
Nephrol 2009; 24: 67–76.
Transparency declarations 16 Germovsek E, Lutsar I, Kipper K et al. Plasma and CSF pharmacokinetics
None to declare. For the purpose of open access, the authors have of meropenem in neonates and young infants: results from the NeoMero
applied a CC BY public copyright license to any Author Accepted studies. J Antimicrob Chemother 2018; 73: 1908–16.
Manuscript version arising from this submission.
17 Ceriotti F, Boyd JC, Klein G et al. Reference intervals for serum creatinine
concentrations: assessment of available data for global application. Clin
Chem 2008; 54: 559–66.
Supplementary data 18 Standing JF, Readman J, Hamawandi A et al. Identifying carbapenem
Supplementary methods and Figures S1 to S5 are available as and amikacin-sparing first-line combination therapy for neonatal sepsis/
Supplementary data at JAC Online. meningitis in centres with high prevalence of extended-spectrum beta-lacta-
mase-producing Enterobacteriaceae. ECCMID 2019, Amsterdam, The
Netherlands, 2019. Abstract O0827.
References 19 Andrews JM. Determination of minimum inhibitory concentrations. J
1 UNICEF, WHO, World Bank Group and United Nations. Levels and trends in Antimicrob Chemother 2001; 48 Suppl 1: 5–16.
child mortality. Report 2020. Estimates developed by the UN Inter-agency 20 Bonapace CR, Bosso JA, Friedrich LV et al. Comparison of methods of in-
group for child mortality estimation. https://www.unicef.org/reports/levels- terpretation of checkerboard synergy testing. Diagn Microbiol Infect Dis 2002;
and-trends-child-mortality-report-2020. 44: 363–6.
2 Bielicki JA, Lundin R, Sharland M et al. Antibiotic resistance prevalence in 21 Mir AR, Bashir Y, Dar FA et al. Identification of genes coding aminoglyco-
routine bloodstream isolates from children’s hospitals varies substantially side modifying enzymes in E. coli of UTI patients in India. Scientific
from adult surveillance data in Europe. Pediatr Infect Dis J 2015; 34: 734–41. WorldJournal 2016; 2016: 1875865.
3 Droz N, Hsia Y, Ellis S et al. Bacterial pathogens and resistance causing 22 Dallenne C, Da Costa A, Decre D et al. Development of a set of multiplex
community acquired paediatric bloodstream infections in low- and middle- PCR assays for the detection of genes encoding important beta-lactamases
income countries: a systematic review and meta-analysis. Antimicrob Resist in Enterobacteriaceae. J Antimicrob Chemother 2010; 65: 490–5.
Infect Control 2019; 8: 207. 23 Gastine S, Hsia Y, Clements M et al. Variation in target attainment of
4 Williams PCM, Isaacs D, Berkley JA. Antimicrobial resistance among chil- beta-lactam antibiotic dosing between international pediatric formularies.
dren in sub-Saharan Africa. Lancet Infect Dis 2018; 18: e33–44. Clin Pharmacol Ther 2021; 109: 958–70.
5 WHO. Pocket book of hospital care for children: guidelines for the manage- 24 EUCAST. Clinical breakpoint tables, Version 11. 2021. https://www.eucast.
ment of common childhood illnesses: second edition. 2013. https://www. org/fileadmin/src/media/PDFs/EUCAST_files/Breakpoint_tables/v_11.0_Break
who.int/publications/i/item/978-92-4-154837-3. point_Tables.pdf.
6 Hsia Y, Lee BR, Versporten A et al. Use of the WHO Access, Watch, and 25 Guilhaumou R, Benaboud S, Bennis Y et al. Optimization of the treatment
Reserve classification to define patterns of hospital antibiotic use (AWaRe):
with beta-lactam antibiotics in critically ill patients - guidelines from the
an analysis of paediatric survey data from 56 countries. Lancet Glob Health
French Society of Pharmacology and Therapeutics (Societe Francaise de
2019; 7: e861–71.
Pharmacologie et Therapeutique-SFPT) and the French Society of

455
Gastine et al.

Anaesthesia and Intensive Care Medicine (Societe Francaise d’Anesthesie et 32 Sjovall J, Westerlund D, Alvan G. Renal excretion of intravenously
Reanimation-SFAR). Crit Care 2019; 23: 104. infused amoxycillin and ampicillin. Br J Clin Pharmacol 1985; 19:
26 Moore RD, Lietman PS, Smith CR. Clinical response to aminoglycoside 191–201.
therapy: importance of the ratio of peak concentration to minimal inhibitory 33 Wang J, Kumar SS, Sherwin CM et al. Renal clearance in newborns and
concentration. J Infect Dis 1987; 155: 93–9. infants: predictive performance of population-based modeling for drug de-
27 Tumuhamye J, Sommerfelt H, Bwanga F et al. Neonatal sepsis at velopment. Clin Pharmacol Ther 2019; 105: 1462–70.
Mulago national referral hospital in Uganda: etiology, antimicrobial re- 34 Lonsdale DO, Baker EH, Kipper K et al. Scaling beta-lactam antimicrobial
sistance, associated factors and case fatality risk. PLoS One 2020; 15: pharmacokinetics from early life to old age. Br J Clin Pharmacol 2019; 85:
e0237085. 316–46.
28 Gezmu AM, Bulabula ANH, Dramowski A et al. Laboratory-confirmed 35 Wammanda RD, Adamu SA, Joshua HD et al. Implementation of the
bloodstream infections in two large neonatal units in sub-Saharan Africa. Int WHO guideline on treatment of young infants with signs of possible serious
J Infect Dis 2021; 103: 201–7. bacterial infection when hospital referral is not feasible in rural Zaria, Nigeria:
29 Sands K, Carvalho MJ, Portal E et al. Characterization of challenges and solutions. PLoS One 2020; 15: e0228718.
antimicrobial-resistant Gram-negative bacteria that cause neonatal 36 WHO. Guidelines for the prevention and control of carbapenem-
sepsis in seven low- and middle-income countries. Nat Microbiol 2021; resistant Enterobacteriaceae, Acinetobacter baumannii and
6: 512–23. Pseudomonas aeruginosa in health care facilities. 2017. https://apps.
30 Jatsho J, Nishizawa Y, Pelzom D et al. Clinical and bacteriological profile of who.int/iris/handle/10665/259462.
neonatal sepsis: a prospective hospital-based study. Int J Pediatr 2020; 2020: 37 Tacconelli E, Cataldo MA, Dancer SJ et al. ESCMID guidelines for the man-
1835945. agement of the infection control measures to reduce transmission of
31 Foulds G. Pharmacokinetics of sulbactam/ampicillin in humans: a review. multidrug-resistant Gram-negative bacteria in hospitalized patients. Clin
Rev Infect Dis 1986; 8 Suppl 5: S503–11. Microbiol Infect 2014; 20 Suppl 1: 1–55.

456