crossmark

Pharmacokinetics of Transfer of Azithromycin into the Breast Milk of

African Mothers
Sam Salman,a Timothy M. E. Davis,a Madhu Page-Sharp,b Bully Camara,c Claire Oluwalana,c Abdoulie Bojang,c
Umberto D’Alessandro,c,d,e Anna Rocac,d
School of Medicine and Pharmacology, University of Western Australia, Fremantle Hospital, Fremantle, Western Australia, Australiaa; School of Pharmacy and Curtin Health
Innovation Research Institute, Curtin University of Technology, Perth, Western Australia, Australiab; Medical Research Council Unit, Fajara, The Gambiac; London School of
Hygiene and Tropical Medicine, London, United Kingdomd; Institute of Tropical Medicine, Antwerp, Belgiume

Azithromycin (AZI) is used for its antibiotic and antimalarial properties in pregnancy. Reported estimates of AZI breast milk
transfer, based on concentrations in mostly single samples from small numbers of women, have suggested that infant intake is
safe. To better characterize infant intake and the associated potential benefits and risks, AZI was measured by liquid chromatog-
raphy-mass spectrometry in four breast milk samples taken over 28 days postpartum from each of 20 Gambian women given 2 g
AZI during labor. A population pharmacokinetic model utilizing published parameters for AZI disposition in pregnancy, the
present breast milk concentrations, and increasing/decreasing sigmoid maximum-effect (Emax) functions adequately described
temporal changes in the milk/plasma ratio. The median estimated absolute and relative cumulative infant doses were 4.5 mg/kg
of body weight (95% prediction interval, 0.6 to 7.0 mg/kg) and 15.7% (95% prediction interval, 2.0 to 27.8%) of the maternal
dose, respectively; the latter exceeded the recommended 10% safety limit. Although some infants with bacterial infections may
benefit from AZI in breast milk, there is a risk of hypertrophic pyloric stenosis with a worst-case number needed to harm of 60
based on the present and available epidemiologic data. (This study has been registered at ClinicalTrials.gov under registration
no. NCT01800942.)

A zithromycin (AZI) is a macrolide antibiotic that is used in a

variety of clinical situations, including young infants with
pertussis (1), and in pregnancy to treat sexually transmissible and
reproductive tract infections (2). It has antimalarial activity that,
as part of combination therapy (3), is beneficial in pregnant
women with Plasmodium infections and consequently at in-
creased risk of anemia and obstetric complications (4). When ad-
ministered at delivery, its elimination half-life of 3.3 days (5), to-
gether with high and sustained tissue concentrations relative to
those in serum (6), suggest that the breastfeeding infant could
ingest significant amounts of AZI. Markedly elevated AZI concen-
trations in maternal leukocytes, a component of breast milk (7)
that can increase in response to neonatal infection (8), represent a
potentially major contribution to postnatal AZI transfer.
Available data on concentrations of AZI in breast milk are
sparse. In a woman with cellulitis after tubal ligation treated with
1 g AZI orally, followed 48 h later by five 500-mg daily oral doses,
increasing breast milk concentrations of up to 2.8 mg/liter were
present in three samples, the last taken after a cumulative dose of
2.5 g (9). Using the concentration in the last sample, the authors
estimated that the infant consumed 1.3 mg/day or, after adjust-
ment for 37% bioavailability found in healthy adults (10), 0.2
mg/kg of body weight/day. This was not considered to be a clini-
cally significant infant dose (9). In a study of 8 women given a
500-mg intravenous (i.v.) AZI dose before cesarean section, single
breast milk samples were taken from each woman between 12 and
48 h postinfusion (11). Simulations from multicompartmental
pharmacokinetic (PK) modeling suggested that steady-state
breast milk concentrations after 500-mg i.v. doses every 12 hours
were achieved after 3 days. Under assumptions of 150 ml/kg/day
milk intake and bioavailability of 38%, the comparable daily i.v.
AZI dose was predicted to be 0.34 mg (0.097 mg/kg) in a 3.5-kg
infant (11). This corresponds to an oral dose of 0.25 mg/kg/day,
similar to the 0.2 mg/kg/day estimated in the single case report (9).
These predicted infant doses (9, 11) are based on simple PK
analyses that may not adequately capture temporal changes in
milk composition, volume, and AZI concentrations and, hence,
infant AZI exposure. The volume of milk produced during lacta-
tion increases to a maximum at day 7 postpartum (12), when the
estimated infant intake is 150 ml/kg/day (13). In addition, in the
report involving 8 women (11), there was no estimate of the pop-
ulation variability in infant AZI intake, an important consider-
ation in the assessment of the range of potential infant doses and
thus toxicity. A key aspect of quantifying exposure is the associa-
tion between macrolide therapy and infantile hypertrophic pylo-
ric stenosis (IHPS), the risk of which appears to be highest in the
first 2 weeks of life (14–16) but which is uncertain in the context of
AZI-treated breastfeeding women (17).
In view of these limitations, we measured AZI concentrations
in breast milk collected between 3 and 28 days postpartum from
20 Gambian women given a single 2-g oral dose of AZI during
labor and utilized a population PK model derived from these data
to predict absolute and relative total infant dose ranges for the
Received 2 November 2015 Returned for modification 5 December 2015
Accepted 13 December 2015
Accepted manuscript posted online 28 December 2015
Citation Salman S, Davis TME, Page-Sharp M, Camara B, Oluwalana C, Bojang A,
D’Alessandro U, Roca A. 2016. Pharmacokinetics of transfer of azithromycin into
the breast milk of African mothers. Antimicrob Agents Chemother 60:1592–1599.
doi:10.1128/AAC.02668-15.
Address correspondence to Timothy M. E. Davis, tim.davis@uwa.edu.au.
S.S. and T.M.E.D. contributed equally to this work.
Copyright © 2016, American Society for Microbiology. All Rights Reserved.

1592 aac.asm.org Antimicrobial Agents and Chemotherapy March 2016 Volume 60 Number 3

present maternal AZI regimen, as well as a higher dose (1 g daily
for 3 days) that has been used as part of antimalarial therapy in
pregnancy (18). In addition, we explored the association between
AZI exposure through breast milk ingestion and IHPS using avail-
able data.
MATERIALS AND METHODS
Study design and participants. The women in the present study were
recruited to a phase III, double-blind, placebo-controlled trial in which
women in labor were randomized 1:1 to receive a single oral dose of 2 g
AZI or placebo (ClinicalTrials.gov NCT01800942). The trial was con-
ducted at the Jammeh Foundation for Peace (JFFP), a government-run
health center in Western Gambia, approximately 20 km from the capital,
Banjul. Women aged 18 to 45 years who gave consent during antenatal
care visits and presented in labor between April 2013 and April 2014 were
eligible. Exclusion criteria included known HIV infection or other clini-
cally significant acute or chronic illness, planned cesarean section or
known required referral for hospital delivery, known multiple pregnancy,
known severe congenital malformation or intrauterine death, known
macrolide allergy, antibiotic treatment in the week before randomization,
and inability to attend all follow-up visits.
The study was approved by the joint Gambia Government-Medical
Research Council (MRC) Ethics Committee. A local safety monitor and a
Data Safety Monitor Board (DSMB) reviewed all the serious adverse
events during the trial, and the trial was monitored by an independent
clinical trials monitor.
Trial procedures. After confirmation of labor, a single 2-g oral AZI
dose or placebo was administered to each woman. The date and time of
dosing were recorded. Each woman was then observed closely by study
nurses, who documented episodes of vomiting. A detailed clinical exam-
ination of both mother and baby was carried out by one of the study
clinicians before discharge (between 6 and 24 h after delivery) and again 8
to 10 days after delivery. Additional active follow-up was conducted by
study nurses and field workers daily during the first week after delivery
and then weekly from week 2 to week 8.
Overall, 829 women and their 843 offspring participated in the main
study. The first 40 women who took the study drug without vomiting and
who were available to provide breast milk samples over 2 months partic-
ipated in the present substudy. Follow-up samples included breast milk
collected by manual expression on days 3 and 6 postpartum and at 2 and
4 weeks postdelivery. The first 0.5 ml was discarded, and the next 1 to 2 ml
was collected in a sterile plastic bottle, transported in a cold box within 8
h of collection, vortex mixed for 20 s, and stored at ⫺70°C before assay.
Azithromycin assay. Breast milk AZI concentrations were measured
using a validated ultra-high-performance liquid chromatography-tan-
dem mass spectrometry (UPLC–MS-MS) method with a triple-quadru-
pole MS (8030 plus; Shimadzu, Kyoto, Japan) and a deuterated internal
AZI standard, as described previously (5). The assays were performed
without knowledge of the allocated therapy. In brief, 20-␮l milk samples
were spiked with internal standard and extracted based on methodology
developed for plasma samples. After addition of ammonium hydroxide
and methanol, vortex mixing, and freezing, the thawed sample was cen-
trifuged, the organic supernatant was mixed with water, and an aliquot
was injected onto an Agilent Eclipse Plus C18 column (Agilent Technolo-
gies, Santa Clara, CA).
Mass spectrometric quantitation was performed in multiple-reaction-
monitoring mode using an electrospray ionization in positive ion mode
(ESI⫹) ion source. AZI USP was obtained from APAC Oargnaceytucak
LLC (Ellicott City, MD) and deuterated AZI from Toronto Research
Chemicals (North York, Canada). In brief, following the addition of an
internal standard, AZI was extracted from 1 ml breast milk by protein
precipitation. After centrifugation, supernatant (5 ␮l) was injected onto a
2795/Quattro Premier XE UPLC– electrospray ionization (ESI)–MS-MS
(Waters Corp., MA) using a Waters BEH C18 1.7-␮m column. Matrix
effect, process efficiency, and absolute recovery for AZI and deuterated
March 2016 Volume 60 Number 3 Antimicrobial Agents and Chemotherapy
Azithromycin Transfer into Breast Milk
AZI (AZI-d3) were within acceptable ranges (means, 92.5 to 107.9%).
Intra- and interday relative standard deviations were 3.1 to 5.2% and 6.6
to 9.7% at AZI concentrations between 100 and 5,000 ␮g/liter, respec-
tively. The mean accuracy was 94.5 to 103.8% over the same concentra-
tion range. The limits of quantitation and detection were 5 ␮g/liter and 2.5
␮g/liter, respectively.
Pharmacokinetic modeling. NONMEM (v 7.2.0; Icon Development
Solutions, Ellicott City, MD, USA) (19) with an Intel Visual FORTRAN
10.0 compiler was utilized for nonlinear mixed-effects modeling of the
natural logarithm (loge) breast milk concentration-time data set. The
first-order conditional estimates with interaction (FOCE INTER) estima-
tion method was used, with the minimum value of the objective function
value (OFV), goodness-of-fit plots, condition number (⬍1,000), and pre-
dictive checks used to determine suitable models. A significance level of a
P value of ⬍0.05 was set for comparison of nested models. Two structures
for residual variability (RV), equivalent to proportional and combined
RV structures on the normal scale, were tested using the log-transformed
data.
Given that only breast milk data were available for analysis, a previ-
ously developed validated population model of AZI disposition in preg-
nant women was utilized (5). It comprises a linear three-compartment
model with mixed zero- and first-order absorption. The structural-model
parameters were as follows: ka (rate of first-order absorption), DUR (du-
ration of zero-order absorption), CL/F (clearance relative to bioavailabil-
ity), VC/F (central volume of distribution relative to bioavailability), Q1/F
(intercompartmental clearance for VP1/F), VP1/F (first peripheral volume
of distribution relative to bioavailability), Q2/FAZI (intercompartmental
clearance for VP2/F), and VP2/F (second peripheral volume of distribution
relative to bioavailability). As it was not possible to estimate population
variability on the model parameters, the average population parameter
was utilized, with a single intraindividual variability (IIV) term estimated
for bioavailability (F). This IIV term is a scaling term encompassing vari-
ability in AZI absorption, as well as interindividual differences in the
milk/plasma ratio (MPratio). Given that this model was developed in a
different population (Melanesian women) earlier in pregnancy (second
trimester) (5), CL/F was allowed to be estimated, while other structural-
model parameters were fixed.
Breast milk concentrations were linked with the above-described
model using an MPratio. This method has been used successfully with
other antimalarial drugs (20) and pharmacotherapies for other diseases
(21). Given the temporal changes in this ratio (Fig. 1) and based on prior
experience (20), a sigmoid maximum effect (Emax) relationship was uti-
lized a priori to represent these changes. In this case, positive and negative
sigmoid Emax relationships were required to represent the data adequate-
ly: MP=M,t ⫽ MPM,t ⫻ [1 ⫹ SIGpos ⫻ (TbirthHill pos/TbirthHill pos ⫹
MAT50,posHill pos)] and MP=M,t ⫽ MPM,t ⫻ [1 ⫹ SIGneg ⫻ (TbirthHill neg/
TbirthHill neg ⫹ MAT50,negHill neg)], where MP=M,t is the MPratio after con-
sideration of the effect of time, SIGpos is the maximum effect on MPratio of
the positive curve, Tbirth is the time of birth, MAT50,pos is the time to 50%
of the positive curve, and Hillpos is the Hill coefficient for the positive
curve, while SIGneg is the maximum effect on MPratio of the negative
curve, MAT50,neg is the time to 50% of the negative curve, and Hilldec is the
Hill coefficient for the negative curve. The lack of concurrent plasma data
and the complex relationship over time did not allow informative testing
of other relationships. IIV terms were included where they could be esti-
mated. The effects of other covariates, including maternal age, gestational
age at birth, and infant birth weight, were identified by inspection of
individual-parameter-versus-covariate plots and use of the generalized
additive model within Xpose. Identified relationships were then tested
within NONMEM using a stepwise forward (P ⬍ 0.05) and backward
(P ⬍ 0.01) approach.
Model evaluation included goodness-of-fit plots with observed versus
individual and population predicted values and residual plots of time
from first dose. A bootstrap using Perlspeaks NONMEM with 1,000 sam-
ples was performed, and the derived parameters were summarized as me-
aac.asm.org 1593
Salman et al.

rates and macrolide exposure (without doses) for 1,074,000 infants, was
utilized, together with the present PK modeling.

RESULTS
Baseline characteristics. The 20 AZI-treated women had a mean
age (⫾standard deviation [SD]) of 27.7 ⫾ 6.1 years, and they
weighed 68 ⫾ 11 kg. There were 7 (35%) primigravidae (parity
range, 1 to 7), and their gestational age was 38 ⫾ 2 weeks at deliv-
ery. The mean birth weight was 3.2 ⫾ 0.4 kg (range, 2.5 to 4.0 kg).
Two of the 80 breast milk concentrations from the 20 AZI-
treated women (2.5%) were below the limit of quantitation and
were excluded from PK modeling. Observed breast milk con-
centrations plotted against the population average maternal
plasma concentration curve are shown in Fig. 1. Breast milk
AZI concentrations were undetectable in all 80 samples from
the 20 placebo-treated women.
Clinical course. None of the 40 women in the present study (20
randomized to AZI and 20 to placebo) required additional phar-
macotherapy during the 28-day follow-up period. All provided
four breast milk samples on days 3, 6, 14, and 28 postpartum. No
serious adverse events were reported. All the infants remained well
postpartum, with no cases of IHPS detected either in these 40
FIG 1 (Top) Population average maternal plasma azithromycin concentra- singleton infants or in the 419 infants whose mothers had received
tion, based on a previously published study (5) (line), together with breast milk
azithromycin concentrations from the 20 women given 2 g during labor in the
present study (Œ). (Bottom) The log10(milk/plasma ratio) for azithromycin
calculated using median parameters from a previously published study (5) (Œ) TABLE 1 Final population pharmacokinetic estimates and bootstrap
and combined model for changes in the milk/plasma ratio over time (black results for azithromycin in breast milk of lactating Gambian women
line), together with the effects of the positive and negative sigmoid Emax rela-
Bootstrap median
tionships for the milk/plasma ratio (gray lines).
Parametera Mean RSE%b (95% CI)
Objective-function value ⫺26.833 ⫺34.439 (⫺80.256–6.248)
Structural-model parameters
dian and 2.5th and 97.5th percentiles (95% empirical confidence interval CL/FAZI (liters/h/70 kg) 104 16 105 (76–181)
[CI]) to facilitate evaluation of final model parameter estimates. In addi- MPratio 2.49 40 2.84 (1.20–7.41)
tion, prediction-corrected visual predictive checks (pcVPCs) were per- SIGpos 30.9 80 31.2 (3.6–241.8)
formed for breast milk data, with 1,000 data sets simulated from the final MAT50,pos (h) 253 21 253 (137–438)
models. The observed 10th, 50th, and 90th percentiles were plotted with Hillpos 3.3 Fixed
their respective simulated 95% CIs to assess the predictive performance of SIGneg 13.6 40 13.2 (3.4–26.5)
the model and to evaluate bias. Shrinkage of population variability pa- MAT50,neg (h) 147 4 149 (136–167)
rameters and residual variability were incorporated to help determine Hillneg 3.3 Fixed
whether models were overparameterized and to determine the reliability ka (h⫺1) 0.525 Fixed
of diagnostic plots. DUR (h) 1.55 Fixed
The infant dose was estimated from the final model using previ- VC/FAZI (liters/70 kg) 714 Fixed
ously published data for piperaquine (20). Briefly, a sigmoid Emax Q1/FAZI (liters/h/70 kg) 325 Fixed
model was fitted to the mean milk transfer (in milliliters per kilogram) VP1/FAZI (liters/70 kg) 4,080 Fixed
for women with normal vaginal delivery, which was then incorporated Q2/FAZI (liters/h/70 kg) 67.2 Fixed
into the NONMEM control file to provide an estimate of the total infant VP2/FAZI (liters/70 kg) 5,070 Fixed
piperaquine dose as a continuous function. For the purposes of estima-
tion, discrete feeds were not modeled, given the longer half-life of AZI in Variable-model parameters
comparison to the short time between neonatal feeds. Absolute daily and (% shrinkage)
total infant AZI doses (in milligrams per kilogram), as well as relative daily IIV in F 63 (11) 30 63 (1–88)
and total infant doses (percentages of the daily and total maternal doses, IIV in SIGpos 60 (22) 20 58 (34–85)
respectively), were estimated for each participant to 28 days after birth. RV (%) 32 (19) 13 31 (21–38)
Simulations. Using the method described above, simulations were a
ka, rate for first-order absorption; DUR, duration of zero-order absorption; CL/FAZI,
performed to further define the range of expected neonatal AZI exposure. clearance relative to bioavailability; VC/FAZI, central volume of distribution relative to
One thousand woman-neonate pairs were simulated using a 2-g Stat dose bioavailability; Q1/FAZI, intercompartmental clearance for VP1/FAZI; VP1/FAZI, first
during labor and a 1-g daily dose over 3 days, as is used for intermittent peripheral volume of distribution relative to bioavailability; Q2/FAZI,
preventive treatment of malaria in pregnancy (18). Absolute daily and intercompartmental clearance for VP2/FAZI; VP2/FAZI, second peripheral volume of
distribution relative to bioavailability; SIGpos, maximum effect on MPratio of the
total infant doses (in milligrams per kilogram) and relative daily and total
positive curve; MAT50,pos, time to 50% of the positive curve; and Hillpos, Hill coefficient
infant doses (percentages of daily and total maternal doses, respectively), for the positive curve; SIGneg, maximum effect on MPratio of the negative curve;
were estimated for each simulated neonate to 28 days after birth. MAT50,neg, time to 50% of the negative curve; Hilldec, Hill coefficient for the negative
Risk of infantile hypertrophic pyloric stenosis. For the estimated risk effect; IIV, interindividual variability; RV, residual variability. The IIV is presented as
of IHPS in infants breastfed by an AZI-treated mother, the largest and 100% times the square root of the variability estimate.
most contemporary database available (14), which contains incidence b
RSE, residual standard error as a percentage of the mean.

1594 aac.asm.org Antimicrobial Agents and Chemotherapy March 2016 Volume 60 Number 3
 Azithromycin Transfer into Breast Milk

FIG 2 Goodness-of-fit plots for azithromycin in breast milk. (A and B) Observed breast milk concentration plotted against population (A) and individual (B)
predicted breast milk concentrations. (C and D) Weighted residuals against time from first dose (C) and time from birth (D).

AZI in the main trial and who were assessed weekly for 8 weeks were calculated with median values of 4.5 mg/kg and 15.7%, re-
after delivery. spectively (Table 2). From this, the estimated absolute and relative
Pharmacokinetic modeling. Initial modeling was performed daily infant doses were then calculated, with a median maximum
with all parameters fixed at their population means. Allowing daily intake of 0.7 mg/kg/day and 2.5%, respectively. The maxi-
clearance relative to bioavailability (CL/F) to be estimated resulted mum estimated cumulative infant dose was 27.8% (8.0 mg/kg),
in significant improvement in the model (P ⬍ 0.001). Using a while the maximum estimated relative daily infant dose was 4.5%
single sigmoid Emax model (either positive or negative) did not (1.3 mg/kg/day).
adequately describe temporal changes in MPratio and resulted in Simulations. The results of simulations are presented in Table
significant bias in the conditional weighted residual (CWRES) 2 and Fig. 4. The median estimated relative total dose was 13.7%
plots. Therefore, two linked relationships in opposite directions (95% prediction interval [PI], 3.5 to 64.7%) for 2 g maternal oral
were utilized, with resolution of the CWRES bias and a fall in the
OFV (P ⬍ 0.01). Attempts to model all parameters resulted in
unstable estimates and high condition numbers (⬎1,000). There-
fore, the Hill coefficient for the positive model was fixed at 3.3, as
obtained previously (20), and that of the negative model was as-
sessed at between 2 and 20, with a final value of 10 chosen, as it best
represented the data with no significant improvement in the fit of
the model or in model diagnostics. IIV was estimable for bioavail-
ability (F) and maximum effect on MPratio of the positive curve
(SIGpos) at 63% and 59%, respectively. No significant covariate
relationships were identified.
Final parameter estimates and bootstrap results are summa-
rized in Table 1. Bias was ⬍5.0% for fixed and random model
parameters, with the exception of MPratio, where it was 14%.
Goodness-of-fit plots are presented in Fig. 2 and the pcVPCs in
Fig. 3. The actual 10th, 50th, and 90th percentiles were within the
FIG 3 Prediction-corrected visual predictive check for azithromycin in breast
95% CI of simulated data and demonstrate reasonable predictive milk. Observed 50th (solid line) and 10th and 90th (dashed lines) percentiles
performance of the model. are shown within their simulated 95% CIs (gray shaded areas) with overlying
The estimated absolute and relative cumulative infant doses data points (Œ).

March 2016 Volume 60 Number 3 Antimicrobial Agents and Chemotherapy aac.asm.org 1595
Salman et al.

TABLE 2 Estimated infant doses of azithromycin to 28 days

Infant doseb
Total Highest daily
a
Source Absolute (mg/kg) Relative (%) Absolute (mg/kg/day) Relative (%)
Measured breast milk concn 4.5 (0.6–8.0) 15.7 (2.0–27.8) 0.7 (0.1–1.3) 2.5 (0.3–4.5)
Simulation for 2-g dose during labor 3.9 (1.0–18.5) 13.7 (3.5–64.7) 0.6 (0.2–3.1) 2.2 (0.6–10.7)
Simulation for three 1-g daily doses from labor 7.8 (2.0–35.9) 18.2 (4.6–83.9) 1.2 (0.3–4.0) 2.9 (0.8–13.8)
a
Derived from breast milk concentrations and from simulations and assuming 100% bioavailability.
b
The data are presented as median (range) for PK modeling based on measured concentrations and median (95% prediction interval) for simulations.

AZI, corresponding to a median absolute total dose of 3.9 mg/kg
(95% PI, 1.0 to 18.5 mg/kg). The highest median relative infant
daily dose was 2.2% (95% PI, 0.6 to 10.7%) or 0.6 mg/kg/day (95%
PI, 0.2 to 3.1 mg/kg/day) on day 6. This value was higher for the
daily 1-g dose over 3 days, with a median relative total dose of
18.2% (95% PI, 4.6 to 83.9%) or 7.8 mg/kg (95% PI, 2.0 to 35.9
mg/kg) and highest median relative daily dose of 2.9% (95% PI,
0.8 to 13.8%) or 1.2 mg/kg/day (95% PI, 0.3 to 4.0 mg/kg/day).
The maximum simulated relative total and daily doses (in only
1 in 1,000 simulations) were 104% and 18%, respectively, for the
2-g Stat dose and 169% and 29%, respectively, for 1 g daily for 3
days. These estimates are close to, or exceed, the suggested 10%
safety limit (13). Neonatal infections, such as pertussis, Urea-
plasma, and chlamydial conjunctivitis are treated with 10 and 20
mg/kg AZI daily for 5 or 3 days, respectively (total, 50 to 60 mg/kg)
(1, 22). Based on the present data, we estimate that the maximum
simulated total AZI doses in infants breastfed by mothers given 2
g AZI during labor or 1 g daily for 3 days are 32 mg/kg and 63
mg/kg, respectively, over 14 days, corresponding to doses that are
36% lower and 5% higher, respectively, than the total neonatal
treatment range (50 to 60 mg/kg).
Risk of infantile hypertrophic pyloric stenosis. The overall
IHPS rate in the large published database was 2.3/1,000, which
compares with 20.3/1,000 in infants given AZI between 0 and 14
days postpartum, with an adjusted odds ratio and 95% CI of 8.26
(2.62 to 26.0) (14). Based on these data, the number needed to
harm (NNH) is approximately 60. Assuming a dose-dependent
risk during the first 14 days of life, the additional cases of IHPS and
NNH by threshold cumulative dose (in milligrams per kilogram)
for both maternal AZI dose regimens (2 g Stat and 1 g daily for 3

FIG 4 Simulation results demonstrating the median (solid black lines), maximum (dashed black lines), and 95% prediction intervals (gray-shaded areas) for
absolute (in milligrams per kilogram) and percent relative total infant doses (A and C) and absolute (in milligrams per kilogram per day) and percent relative
daily infant doses (B and D) from time of dose for 2 g Stat (A and B) and 1 g daily over 3 days (C and D). The cross-hatched areas during the first 3 days represent
the lack of data during this period.

1596 aac.asm.org Antimicrobial Agents and Chemotherapy March 2016 Volume 60 Number 3

FIG 5 Additional cases of infantile hypertrophic pyloric stenosis (left y axis)
estimated based on simulations performed for a 2-g Stat maternal dose during
labor (solid black line) and 1 g daily for 3 days (dashed black line), assuming a
threshold effect for the cumulative dose over the first 14 days of life. The
corresponding numbers needed to harm (gray lines) are presented on the right
y axis.
days) are shown in Fig. 5. For a 10-mg/kg threshold dose in breast
milk, the NNH would be approximately 600 for the former dose
and 200 for the latter, with an additional risk of 1 to 5 cases/1,000
exposed children. The maximum NNH of 60 (14) applies if the
risk occurs with any AZI exposure. This highest possible risk is
unlikely in our infants, given that the rate of IHPS in the main trial
was 0 (95% CI, 0 to 11.3)/1,000 cases or, in the case of the upper
95% CI, an NNH of 111, assuming an index rate of 2.29/1,000
(14). This corresponds to a threshold dose of just under 5 mg/kg.
DISCUSSION
The present data and PK analyses extend the limited available
information relating to AZI transfer into the breast milk of lactat-
ing mothers and provide some indication of the potential benefits
and theoretical risks for the suckling infant. In contrast to previ-
ous studies involving ⱕ8 women who provided ⱕ3 samples taken
over ⱕ6 days postpartum (9, 11), the present analyses used 78
breast milk AZI concentrations from 20 lactating women studied
over 4 weeks after delivery. We utilized available data on AZI
disposition in pregnancy (5) and a modification of a prior popu-
lation PK model of piperaquine transfer into breast milk (20) to
predict infant exposure during 28 days postpartum. The data and
modeling show that some infants receive cumulative doses of AZI
that could help prevent bacterial infections, a major cause of neo-
natal mortality in Africa, but that could also be associated with
toxicity.
There was evidence of a time-dependent change in MPratio that
was adequately described by a combination of positive and nega-
tive sigmoid Emax curves. The positive curve may reflect the in-
creasing fat content of colostrum/milk (23) and the positive logP
value of AZI, while the negative curve could represent the decreas-
ing number of leukocytes in breast milk (7) that contain high AZI
concentrations (ⱖ100 times those in plasma) postdose (24). The
high MPratio is not compatible with passive processes of diffusion
and ion trapping alone (25). Although concentration-dependent
changes in the free fraction of AZI in blood may contribute to the
MPratio (10), an active process is likely, particularly as AZI is a
substrate of transporters, including P-glycoprotein and organic
March 2016 Volume 60 Number 3 Antimicrobial Agents and Chemotherapy
Azithromycin Transfer into Breast Milk
anion transporting polypeptides. A number of transporter genes
are expressed by mammary epithelial cells (26), with changes in
mRNA (both increasing and decreasing) as lactation is established
(27). The changes in the AZI MPratio from birth are, therefore,
likely to reflect a complex interplay between many factors.
The oral bioavailability of AZI given as a suspension or tablets
is not affected by coadministration with food in adults (10, 28),
with little evidence for acid degradation or extensive first-pass
metabolism. This suggests that the high concentration of fat in
breast milk and the changes in gastric pH during the first few
weeks of life may not influence AZI bioavailability in the suckling
infant. However, other factors that might alter neonatal gastroin-
testinal AZI absorption include intestinal transporters, such as
P-glycoprotein (29), the expression of which is reduced in the
neonate (30); increased intestinal permeability early in life (31);
and migration of maternal leukocytes with high AZI concentra-
tions from ingested breast milk into the neonatal circulation (7,
32). In the absence of a formal PK evaluation of the net effect of
these factors on the neonatal plasma AZI concentrations from
breast milk versus oral treatment, we assumed that bioavailabili-
ties were equivalent.
The median total infant dose was 3.9 mg/kg for 2 g AZI given to
pregnant women during labor and approximately double that for
1 g daily for 3 days. The highest median absolute daily infant doses
(on day 6) for the two maternal AZI regimens were 0.6 and 1.2
mg/kg/day, respectively, comparable to the 0.25 to 0.5 mg/kg/day
derived from simpler PK models after 0.5 g i.v. Stat and 2.5 g in
divided oral doses (9, 11). However, our PK modeling was also
able to investigate the total infant dose over longer periods, and
this suggested that some infants receive significantly more than
this, specifically, ⬎70% of an infant oral AZI treatment course (1)
over 28 days in up to 2.5% of children whose mothers were given
1 g daily over 3 days. This percentage may be greater with neonatal
infection and thus a stimulus to increased numbers of AZI-rich
maternal leukocytes in breast milk (8), a potential benefit.
There is, however, also concern regarding an association be-
tween macrolides and IHPS in neonates. Based on available data
(14) and the present PK modeling, the NNH for African infants
breastfed by a mother treated with 2 g AZI is, at worst (based on
the upper 95% CI), 111. Although this should be interpreted
against the possibility of higher maternal doses given for malaria
treatment/prevention (3 g [18] or even 4 g [33] total), the true risk
is likely to be much smaller than this and only estimable from
larger-scale intervention trials. This potential adverse effect needs
to be considered against AZI benefits for neonatal infections and
complications, such as bronchopulmonary dysplasia (34). In ad-
dition, there is some evidence that the incidence of IHPS may be
relatively low in developing countries, including those in sub-
Saharan Africa (35).
The present study had limitations. The MPratio, although pro-
viding an adequate characterization of the present data, was not
based on plasma AZI concentrations in the African women in the
present study. Although (and importantly) these data would not
affect the infant dose estimates, the true MPratio may be different
from the model-derived value. Alternative dynamic models of AZI
breast milk transfer may have been possible, but their complexity
would be limited by the numbers of available samples. Neverthe-
less, the present model fit the data well between 3 and 28 days.
From simulations, the extrapolated data between delivery and day
3 account for only 8 to 16% of the estimated total infant dose,
aac.asm.org 1597
Salman et al.
consistent with the fact that infant milk consumption during the
first 3 days is relatively low (36). Variations in milk crematocrit
could help explain some of the between-subject variability (37),
although the effect of coadministered fat on bioavailability ap-
pears limited (10, 28). Measurement of AZI concentrations in
maternal and neonatal blood, as well as breast milk, from birth
would provide the most robust data on which to base an inte-
grated PK model and thus recommendations regarding safety.
The present study represents the most comprehensive evalua-
tion of the transfer of AZI into breast milk performed to date. It
shows that infant versus maternal AZI dosing through lactation
may exceed the suggested 10% safety limit in a relatively large
proportion of neonates even if only a single dose is given at the
time of labor (13). There may be advantages for the infant in
reducing the risk of respiratory tract and other infections in the
small proportion of infants with high levels of exposure, but the
association with IHPS needs further evaluation in larger trials.
ACKNOWLEDGMENTS
We thank the mothers for their participation. We are also most grateful to
the staff of the JFFP and the Ministry of Health and Social Welfare of The
Gambia for their kind assistance and cooperation during the study. We
are especially grateful to the study field team (led by Edrissa Sabally and
Omar Jarra), the laboratory team (Isatou Jagne-Cox and colleagues), and
the data team (Bai Lamin Dondeh, Kodou Lette, and Maimuna Sowe
and the data entry team). Our thanks also extend to the daily support from
the Clinical Trials Office (led by Jenny Mueller) and to our internal mon-
itor (Vivat Thomas), the local safety monitor (Aderonke Odutola), and
the DSMB team chaired by Daniel Chandramohan (with Brian Green-
wood, Maria Quigley, Stephen Howie, and Hannah Blencowe).
T.M.E.D. was supported by a National Health and Medical Research
Council Practitioner Fellowship (number 1058260). The clinical trial was
jointly funded by the United Kingdom MRC and the United Kingdom
Department for International Development (DFID) under the MRC/
DFID Concordat agreement (reference number MR/J010391/1).
FUNDING INFORMATION
Medical Research Council (MRC) provided funding to Umberto
D’Alessandro and Anna Roca under grant number MR/J010391/1. De-
partment for International Development (DFID) provided funding to
Umberto D’Alessandro and Anna Roca under grant number MR/
J010391/1. Department of Health | National Health and Medical Research
Council (NHMRC) provided funding to Timothy M. E. Davis under grant
number 1058260.
REFERENCES
1. Tiwari T, Murphy TV, Moran J. National Immunization Program
CDC. 2005. Recommended antimicrobial agents for the treatment and
postexposure prophylaxis of pertussis: 2005 CDC Guidelines. MMWR
Recomm Rep 54(RR-14):1–16.
2. Chico RM, Hack BB, Newport MJ, Ngulube E, Chandramohan D. 2013.
On the pathway to better birth outcomes? A systematic review of azithro-
mycin and curable sexually transmitted infections. Expert Rev Anti Infect
Ther 11:1303–1332. http://dx.doi.org/10.1586/14787210.2013.851601.
3. Davis TM, Mueller I, Rogerson SJ. 2010. Prevention and treatment of
malaria in pregnancy. Future Microbiol 5:1599 –1613. http://dx.doi.org
/10.2217/fmb.10.113.
4. Desai M, ter Kuile FO, Nosten F, McGready R, Asamoa K, Brabin B,
Newman RD. 2007. Epidemiology and burden of malaria in pregnancy.
Lancet Infect Dis 7:93–104. http://dx.doi.org/10.1016/S1473-3099(07)
70021-X.
5. Salman S, Rogerson SJ, Kose K, Griffin S, Gomorai S, Baiwog F,
Winmai J, Kandai J, Karunajeewa HA, O’Halloran SJ, Siba P, Ilett KF,
Mueller I, Davis TM. 2010. Pharmacokinetic properties of azithromycin
in pregnancy. Antimicrob Agents Chemother 54:360 –366. http://dx.doi
.org/10.1128/AAC.00771-09.
1598 aac.asm.org Antimicrobial Agents and Chemotherapy
6. Wildfeuer A, Laufen H, Zimmermann T. 1996. Uptake of azithromycin
by various cells and its intracellular activity under in vivo conditions.
Antimicrob Agents Chemother 40:75–79.
7. Ho FC, Wong RL, Lawton JW. 1979. Human colostral and breast milk
cells. A light and electron microscopic study. Acta Paediatr Scand 68:
389 –396.
8. Riskin A, Almog M, Peri R, Halasz K, Srugo I, Kessel A. 2012. Changes
in immunomodulatory constituents of human milk in response to active
infection in the nursing infant. Pediatr Res 71:220 –225. http://dx.doi.org
/10.1038/pr.2011.34.
9. Kelsey JJ, Moser LR, Jennings JC, Munger MA. 1994. Presence of
azithromycin breast milk concentrations: a case report. Am J Obstet Gy-
necol 170:1375–1376.
10. Foulds G, Shepard RM, Johnson RB. 1990. The pharmacokinetics of
azithromycin in human serum and tissues. J Antimicrob Chemother
25(Suppl A):73– 82.
11. Sutton AL, Acosta EP, Larson KB, Kerstner-Wood CD, Tita AT, Biggio
JR. 2015. Perinatal pharmacokinetics of azithromycin for cesarean pro-
phylaxis. Am J Obstet Gynecol 212:812.e1– 6. http://dx.doi.org/10.1016/j
.ajog.2015.01.015.
12. Neville MC, Allen JC, Archer PC, Casey CE, Seacat J, Keller RP, Lutes
V, Rasbach J, Neifert M. 1991. Studies in human lactation: milk volume
and nutrient composition during weaning and lactogenesis. Am J Clin
Nutr 54:81–92.
13. Bennett P. 1996. Use of the monographs on drugs, p 67–74. In Bennett P,
Jensen AA (ed), Drugs and human lactation: a comprehensive guide to the
content and consequences of drugs, micronutrients, radiopharmaceuti-
cals, and environmental and occupational chemicals in human milk, 2nd
ed. Elsevier, New York, NY.
14. Eberly MD, Eide MB, Thompson JL, Nylund CM. 2015. Azithromycin
in early infancy and pyloric stenosis. Pediatrics 135:483– 488. http://dx.doi
.org/10.1542/peds.2014-2026.
15. Cooper WO, Griffin MR, Arbogast P, Hickson GB, Gautam S, Ray WA.
2002. Very early exposure to erythromycin and infantile hypertrophic
pyloric stenosis. Arch Pediatr Adolesc Med 156:647– 650. http://dx.doi
.org/10.1001/archpedi.156.7.647.
16. Lund M, Pasternak B, Davidsen RB, Feenstra B, Krogh C, Diaz LJ,
Wohlfahrt J, Melbye M. 2014. Use of macrolides in mother and child and
risk of infantile hypertrophic pyloric stenosis: nationwide cohort study.
BMJ 348:g1908. http://dx.doi.org/10.1136/bmj.g1908.
17. Goldstein LH, Berlin M, Tsur L, Bortnik O, Binyamini L, Berkovitch M.
2009. The safety of macrolides during lactation. Breastfeed Med 4:197–
200. http://dx.doi.org/10.1089/bfm.2008.0135.
18. Chico RM, Chandramohan D. 2011. Azithromycin plus chloroquine:
combination therapy for protection against malaria and sexually trans-
mitted infections in pregnancy. Expert Opin Drug Metab Toxicol 7:1153–
1167. http://dx.doi.org/10.1517/17425255.2011.598506.
19. Beal SL, Sheiner LB, Boeckmann A, Bauer RJ. 2009. NONMEM user’s
guides 1989-2009. Icon Development Solutions, Ellicott City, MD.
20. Moore BR, Salman S, Benjamin J, Page-Sharp M, Yadi G, Batty KT,
Siba PM, Mueller I, Davis TM. 2015. Pharmacokinetics of piperaquine
transfer into the breast milk of Melanesian mothers. Antimicrob Agents
Chemother 59:4272– 4278. http://dx.doi.org/10.1128/AAC.00327-15.
21. Tanoshima R, Bournissen FG, Tanigawara Y, Kristensen JH, Taddio A,
Ilett KF, Begg EJ, Wallach I, Ito S. 2014. Population PK modelling and
simulation based on fluoxetine and norfluoxetine concentrations in milk:
a milk concentration-based prediction model. Br J Clin Pharmacol 78:
918 –928. http://dx.doi.org/10.1111/bcp.12409.
22. Merchan LM, Hassan HE, Terrin ML, Waites KB, Kaufman DA, Am-
balavanan N, Donohue P, Dulkerian SJ, Schelonka R, Magder LS,
Shukla S, Eddington ND, Viscardi RM. 2015. Pharmacokinetics, micro-
bial response, and pulmonary outcomes of multidose intravenous azi-
thromycin in preterm infants at risk for ureaplasma respiratory coloniza-
tion. Antimicrob Agents Chemother 59:570 –578. http://dx.doi.org/10
.1128/AAC.03951-14.
23. Kociszewska-Najman B, Borek-Dzieciol B, Szpotanska-Sikorska M,
Wilkos E, Pietrzak B, Wielgos M. 2012. The creamatocrit, fat and energy
concentration in human milk produced by mothers of preterm and term
infants. J Matern Fetal Neonatal Med 25:1599 –1602. http://dx.doi.org/10
.3109/14767058.2011.648239.
24. Ballow CH, Amsden GW, Highet VS, Forrest A. 1998. Pharmacokinetics
of oral azithromycin in serum, urine, polymorphonuclear leucocytes and
inflammatory vs non-inflammatory skin blisters in healthy volunteers.
March 2016 Volume 60 Number 3

Clin Drug Investig 15:159 –167. http://dx.doi.org/10.2165/00044011
-199815020-00009.
25. Ito S, Lee A. 2003. Drug excretion into breast milk— overview. Adv
Drug Deliv Rev 55:617– 627. http://dx.doi.org/10.1016/S0169-409X
(03)00034-6.
26. Alcorn J, Lu X, Moscow JA, McNamara PJ. 2002. Transporter gene
expression in lactating and nonlactating human mammary epithelial cells
using real-time reverse transcription-polymerase chain reaction. J Phar-
macol Exp Ther 303:487– 496. http://dx.doi.org/10.1124/jpet.102.038315.
27. Lemay DG, Ballard OA, Hughes MA, Morrow AL, Horseman ND,
Nommsen-Rivers LA. 2013. RNA sequencing of the human milk fat layer
transcriptome reveals distinct gene expression profiles at three stages of
lactation. PLoS One 8:e67531. http://dx.doi.org/10.1371/journal.pone
.0067531.
28. Curatolo W, Foulds G, Labadie R. 2010. Mechanistic study of the azi-
thromycin dosage-form-dependent food effect. Pharm Res 27:1361–1366.
http://dx.doi.org/10.1007/s11095-010-0126-6.
29. Amsden GW, Nafziger AN, Foulds G, Cabelus LJ. 2000. A study of the
pharmacokinetics of azithromycin and nelfinavir when coadministered in
healthy volunteers. J Clin Pharmacol 40:1522–1527.
30. Tayman C, Rayyan M, Allegaert K. 2011. Neonatal pharmacology: ex-
tensive interindividual variability despite limited size. J Pediatr Pharmacol
Ther 16:170 –184. http://dx.doi.org/10.5863/1551-6776-16.3.170.
31. Kearns GL. 2000. Impact of developmental pharmacology on pediatric
study design: overcoming the challenges. J Allergy Clin Immunol 106:
S128 –S138. http://dx.doi.org/10.1067/mai.2000.109419.
32. Zhou L, Yoshimura Y, Huang Y, Suzuki R, Yokoyama M, Okabe M,
March 2016 Volume 60 Number 3 Antimicrobial Agents and Chemotherapy
Azithromycin Transfer into Breast Milk
Shimamura M. 2000. Two independent pathways of maternal cell trans-
mission to offspring: through placenta during pregnancy and by breast-
feeding after birth. Immunology 101:570 –580. http://dx.doi.org/10.1046
/j.1365-2567.2000.00144.x.
33. Unger HW, Ome-Kaius M, Wangnapi RA, Umbers AJ, Hanieh S, Suen
CS, Robinson LJ, Rosanas-Urgell A, Wapling J, Lufele E, Kongs C,
Samol P, Sui D, Singirok D, Bardaji A, Schofield L, Menendez C,
Betuela I, Siba P, Mueller I, Rogerson SJ. 2015. Sulphadoxine-
pyrimethamine plus azithromycin for the prevention of low birthweight
in Papua New Guinea: a randomised controlled trial. BMC Med 13:9.
http://dx.doi.org/10.1186/s12916-014-0258-3.
34. Nair V, Loganathan P, Soraisham AS. 2014. Azithromycin and other
macrolides for prevention of bronchopulmonary dysplasia: a systematic
review and meta-analysis. Neonatology 106:337–347. http://dx.doi.org/10
.1159/000363493.
35. Bickler SW, DeMaio A. 2008. Western diseases: current concepts and
implications for pediatric surgery research and practice. Pediatr Surg Int
24:251–255. http://dx.doi.org/10.1007/s00383-007-2095-3.
36. Evans KC, Evans RG, Royal R, Esterman AJ, James SL. 2003. Effect of
caesarean section on breast milk transfer to the normal term newborn over
the first week of life. Arch Dis Child Fetal Neonatal Ed 88:F380 –F382.
http://dx.doi.org/10.1136/fn.88.5.F380.
37. Salman S, Sy SK, Ilett KF, Page-Sharp M, Paech MJ. 2011. Population
pharmacokinetic modeling of tramadol and its O-desmethyl metabolite in
plasma and breast milk. Eur J Clin Pharmacol 67:899 –908. http://dx.doi
.org/10.1007/s00228-011-1023-6.
aac.asm.org 1599