new england

The
journal of medicine
established in 1812 July 2, 2020 vol. 383  no. 1

Amoxicillin for 3 or 5 Days for Chest-Indrawing Pneumonia

in Malawian Children
Amy‑Sarah Ginsburg, M.D., M.P.H., Tisungane Mvalo, M.M.E.D., Evangelyn Nkwopara, M.S.,
Eric D. McCollum, M.D., Melda Phiri, M.B., B.S., Robert Schmicker, M.S., Jun Hwang, M.S.,
Chifundo B. Ndamala, Dip., Ajib Phiri, M.D., Norman Lufesi, M.Phil., and Susanne May, Ph.D.​​

a bs t r ac t

BACKGROUND
Evidence regarding the appropriate duration of treatment with antibiotic agents in From the Department of Biostatistics,
children with pneumonia in low-resource settings in Africa is lacking. University of Washington Clinical Trial
Center, Seattle (A.-S.G., R.S., J.H., S.M.);
METHODS the University of North Carolina Project,
Lilongwe Medical Relief Fund Trust,
We conducted a double-blind, randomized, controlled, noninferiority trial in Lilongwe, ­Tidziwe Centre (T.M., M.P., C.B.N.), and
Malawi, to determine whether treatment with amoxicillin for 3 days is less effec- Acute Respiratory Infection and Emer-
tive than treatment for 5 days in children with chest-indrawing pneumonia (cough gency Triage Assessment and Treatment,
Malawi Ministry of Health (N.L.), Lilongwe,
lasting <14 days or difficulty breathing, along with visible indrawing of the chest and the Department of Pediatrics and
wall with or without fast breathing for age). Children not infected with human Child Health, College of Medicine, Uni-
immunodeficiency virus (HIV) who were 2 to 59 months of age and had chest- versity of Malawi, Blantyre (A.P.) — all in
Malawi; Save the Children, Fairfield, CT
indrawing pneumonia were randomly assigned to receive amoxicillin twice daily (E.N.); and Eudowood Division of Pediat-
for either 3 days or 5 days. Children were followed for 14 days. The primary out- ric Respiratory Sciences, Department of
come was treatment failure by day 6; noninferiority of the 3-day regimen to the Pediatrics, Johns Hopkins School of Medi-
cine and Department of International
5-day regimen would be shown if the percentage of children with treatment failure Health, Johns Hopkins Bloomberg School
in the 3-day group was no more than 1.5 times that in the 5-day group. Prespecified of Public Health, Baltimore (E.D.M.). Ad-
secondary analyses included assessment of treatment failure or relapse by day 14. dress reprint requests to Dr. Ginsburg at
the Department of Biostatistics, Univer-
sity of Washington Clinical Trial Center,
RESULTS
Bldg. 29, Suite 250, 6200 NE 74th St., Seat-
From March 29, 2016, to April 1, 2019, a total of 3000 children underwent ran- tle, WA 98115, or at m ­ essageforamy@​
domization: 1497 children were assigned to the 3-day group, and 1503 to the 5-day ­gmail​.­com.
group. Among children with day 6 data available, treatment failure had occurred This is the New England Journal of Medi-
in 5.9% in the 3-day group (85 of 1442 children) and in 5.2% (75 of 1456) in the cine version of record, which includes all
Journal editing and enhancements. The
5-day group (adjusted difference, 0.7 percentage points; 95% confidence interval Author Final Manuscript, which is the au-
[CI], −0.9 to 2.4) — a result that satisfied the criterion for noninferiority of the thor’s version after external peer review
3-day regimen to the 5-day regimen. Among children with day 14 data available, and before publication in the Journal, is
available under a CC BY license at
176 of 1411 children (12.5%) in the 3-day group and 154 of 1429 (10.8%) in the PMC7233470.
5-day group had had treatment failure by day 6 or relapse by day 14 (between-
N Engl J Med 2020;383:13-23.
group difference, 1.7 percentage points; 95% CI, −0.7 to 4.1). The percentage of
DOI: 10.1056/NEJMoa1912400
children with serious adverse events was similar in the two groups (9.8% in the Copyright © 2020 Massachusetts Medical Society.
3-day group and 8.8% in the 5-day group).
CONCLUSIONS
In HIV-uninfected Malawian children, treatment with amoxicillin for chest-in-
drawing pneumonia for 3 days was noninferior to treatment for 5 days. (Funded
by the Bill and Melinda Gates Foundation; ClinicalTrials.gov number, NCT02678195.)

n engl j med 383;1  nejm.org  July 2, 2020 13

The New England Journal of Medicine
Downloaded from nejm.org on August 17, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l
E
ach year, approximately 920,000 chil-
dren die from pneumonia before the age of
5 years.1 Providing greater access to ap-
propriate and effective treatment is an important
goal. Treatment of bacterial pneumonia requires
an effective antibiotic agent administered in ade-
quate doses for an appropriate duration. Deter-
mining the appropriate duration of antibiotic
therapy is key to ensuring effective treatment
while maximizing adherence and minimizing
adverse drug effects, costs, and antimicrobial
resistance.
The World Health Organization (WHO) uses
clinical signs to classify the severity of pneumo-
nia (classifications of pneumonia and trial defi-
nitions are provided in Table S1 in the Supple-
mentary Appendix, available with the full text of
this article at NEJM.org). The WHO recommends
a 5-day course of amoxicillin, administered
orally twice daily at a minimum of 40 mg per
kilogram of body weight per dose (for a total of
80 mg per kilogram per day), as first-line treat-
ment for chest-indrawing pneumonia (cough
lasting <14 days or difficulty breathing, along
with visible indrawing of the chest wall with or
without fast breathing for age) in immunocom-
petent children younger than 5 years of age.2,3
However, whether a 5-day course is necessary or
a shorter duration of treatment would be as effec-
tive is unclear. On the basis of studies of 3-day
oral antibiotic treatment as compared with 5-day
treatment for fast-breathing pneumonia (cough
lasting <14 days or difficulty breathing, along
with fast breathing for age), the WHO recom-
mends a 3-day course of oral amoxicillin for
treatment of fast-breathing pneumonia in immu-
nocompetent children younger than 5 years of
age.2,4-7 A Cochrane review identified no qualify-
ing randomized, controlled trials that compared
a 2-day or 3-day regimen with a 5-day regimen
of intravenous antibiotics for chest-indrawing or
more severe pneumonia.8 Few data exist to in-
form duration of treatment for pneumonia, and
data are lacking from studies comparing 3-day
oral antibiotic treatment with 5-day treatment for
chest-indrawing pneumonia.9,10 International and
national pneumonia treatment guidelines rely on
expert opinion and limited and weak evidence.10,11
In light of the global threat of increasing anti-
microbial resistance, evidence-based recommen-
dations are needed to guide the proper duration
of antibiotic treatment for pneumonia. Given the
14 n engl j med 383;1  nejm.org  July 2, 2020
The New England Journal of Medicine
Downloaded from nejm.org on August 17, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
of m e dic i n e
paucity of data from Africa, African-specific re-
search in settings in which malaria is endemic
is needed to establish the appropriate treatment
in children with chest-indrawing pneumonia.
Me thods
Trial Design and Oversight
The primary objective of this prospective, double-
blind, randomized, controlled, noninferiority trial
was to determine whether 3-day treatment with
amoxicillin in children 2 to 59 months of age,
not infected with human immunodeficiency vi-
rus (HIV), who have chest-indrawing pneumonia
and who are residing in a region of Malawi in
which malaria is endemic is substantively less
effective than 5-day treatment (the null hypoth-
esis). A noninferiority design was chosen on the
basis of the hypothesis that a 3-day regimen of
amoxicillin would not be more beneficial than a
5-day regimen with respect to the primary out-
come of treatment failure by day 6, but it might
be only slightly less beneficial than a 5-day regi-
men (alternative hypothesis).12
Children 2 to 59 months of age who pre-
sented to the outpatient departments of Kamuzu
Central Hospital (KCH) or Bwaila District Hospi-
tal (BDH) in Lilongwe, Malawi, and who met the
WHO criteria for the case definition of chest-
indrawing pneumonia were screened by trial
staff to determine eligibility (Table S2). Screen-
ing included testing for malaria, HIV, and ane-
mia. No diagnostic testing for viruses, bacteria,
or tuberculosis was performed (Section S4).
The trial was conducted in accordance with
the International Conference on Harmonisation
Good Clinical Practice guidelines and the prin-
ciples of the Declaration of Helsinki. The trial
was approved by the Western Institutional Re-
view Board (United States); the College of Medi-
cine Research and Ethics Committee (Blantyre,
Malawi); and the Malawi Pharmacy, Medicines,
and Poisons Board. All the authors vouch for
the accuracy and completeness of the data and
analyses and for the adherence of the trial to the
protocol (available at NEJM.org).
Procedures
On day 1, eligible children were enrolled and
were randomly assigned in a 1:1 ratio to either a
3-day regimen of amoxicillin or a 5-day regi-
men. High-dose oral amoxicillin was provided in
 Amoxicillin for Chest-Indr awing Pneumonia in Children
250-mg dispersible tablets (tablets dissolved in
liquid) in two divided doses daily. Children in the
3-day group received amoxicillin tablets for 3 days
and placebo tablets for 2 days. Children in the
5-day group received amoxicillin tablets for 5 days.
The total dose was determined on the basis of
age group (children 2 to 11 months of age re-
ceived 500 mg per day, children 12 to 35 months
of age received 1000 mg per day, and children 36
to 59 months of age received 1500 mg per day),
according to the current WHO-recommended
therapy for HIV-uninfected children.2 The amox-
icillin and placebo were identical in appearance,
smell, taste, dispersion activity, and packaging.
Randomization was stratified according to age
group (2 to 11 months, 12 to 35 months, or 36 to
59 months) and trial phase (phase 1 [KCH site]
or phase 2 [BDH site]) with the use of block
sizes of 2, 4, and 6. Other than the biostatisti-
cians, pharmacists, trial monitors, and members
of the data and safety monitoring board, no mem-
bers of the trial team were aware of the treat-
ment assignments.
Enrollment was initially conducted solely at
KCH (phase 1) and was then transitioned (on
September 20, 2016) to BDH (phase 2), after
KCH introduced user fees that reduced patient
volumes. BDH enrollees were transferred to KCH
for additional evaluation and admission. For
safety reasons, most enrollees were hospitalized
for 2 days and discharged on day 3 if no criteria
for treatment failure had been met.
Children were evaluated on day 2 (while hos-
pitalized) and on days 4, 6, and 14 in the clinic
or at home. During follow-up, all children were
assessed at all scheduled and unscheduled visits
for treatment failure or relapse and for adher-
ence to the trial regimen. Most children who had
treatment failure or relapse were hospitalized
and were treated with intravenous antibiotics.
Once they received intravenous or other second-
line antibiotics, children were considered to have
had treatment failure.
Outcomes
The primary outcome was treatment failure by
day 6. Secondary outcomes included relapse (oc-
curring any time from day 7 through day 14
among children who did not have treatment
failure on or before day 6) and treatment failure
by day 6 or relapse by day 14. Treatment failure
was assessed in prespecified subgroups accord-
n engl j med 383;1  nejm.org  July 2, 2020
The New England Journal of Medicine
Downloaded from nejm.org on August 17, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
ing to age group (2 to 11 months, 12 to 35
months, or 36 to 59 months), malnutrition sta-
tus (as assessed by mid–upper-arm circumfer-
ence <11.5 cm, 11.5 to 13.5 cm, or >13.5 cm),
malaria status (positive or negative), and very
fast breathing for age (yes or no).
All adverse events were documented and were
assessed and managed according to KCH stan-
dard clinical practice; children were followed and
treated until resolution of the event or stabiliza-
tion of the child’s condition. All serious adverse
events were reported to the trial safety team for
review within 24 hours.
Statistical Analysis
The relative noninferiority margin for treatment
failure in the 3-day group — 1.5 times the inci-
dence of treatment failure in the 5-day group
— was chosen on the basis of anticipated treat-
ment failure in 8% of the children in the 5-day
group. This noninferiority margin, which repre-
sented a 50% higher incidence of treatment fail-
ure in the 3-day group than in the 5-day group,
was chosen after extensive discussions among the
investigators and with external experts regard-
ing the level of treatment failure in the 3-day
group as compared with the 5-day group that
might be an acceptable margin to clinicians,
considering the anticipated treatment failure in
the 5-day group and the potential enrollment
size in the trial. With initial adjustment for two
formal interim analyses (with an O’Brien–Fleming
boundary for early noninferiority13 and a Pocock
boundary for early inferiority14), we determined
that enrollment of 2000 children (1000 per group)
would provide the trial with 88.1% power to
show noninferiority of the 3-day regimen to the
5-day regimen if treatment failure occurred in
8% of the children in each group and 64.8%
power if treatment failure occurred in 4% of the
children in each group. A potential increase in
sample size was considered during planning of
the trial in the event that the overall incidence
of failure was much lower than the anticipated
8%. After the second formal interim analysis, it
was clear that the overall incidence of failure was
less than 6%. To maintain a power (with equal
incidence of failure in the two groups) of 80% or
higher, the maximum sample size was increased
to 3000 children (1500 per group), and a third
formal interim analysis was performed after
slightly more than 2000 children had been en-
15
 The n e w e ng l a n d j o u r na l
rolled. The decision to increase the sample size
was made by trial investigators who were un-
aware of the treatment assignments, and the
decision was made after consultation with the
funding agency. With the increase in the maxi-
mum sample size (and assuming an equal inci-
dence of failure in each group), the trial had
84.8% power to detect 5% treatment failure in
each group and 89.8% power to detect 6% treat-
ment failure in each group. Power calculations
took into account 5% dropout and assumed a
one-sided alpha level of 0.025 for a test of the
between-group difference in the percentage of
children who had treatment failure.
Primary analyses were performed according
to the intention-to-treat principle but included
only children with complete data; analyses used
linear regression (adjusted for age group, trial
phase, and sex) and robust standard errors
based on the Huber–White sandwich estima-
tor.15,16 Because the sample size was sufficiently
large, linear regression was used for this binary
outcome to model differences in the incidence of
treatment failure.17 Estimates of between-group
differences in the incidence of treatment failure
in prespecified subgroups are reported with 95%
confidence intervals, without adjustment for mul-
tiple comparisons. No per-protocol, as-treated,
or post hoc subgroup analyses were performed
(Section S5). The independent data and safety
monitoring board considered formal stopping
boundaries during their interim reviews but de-
cided not to follow them and rather to treat
them as a guide only (Section S6). Therefore, the
primary analysis was not adjusted for interim
monitoring. Sensitivity analyses were performed
with the use of multiple imputations and tipping-
point analyses.18 For multiple imputations, a hot-
deck approach (20 imputations) was used, with
matching for at least three of the following five
factors: age group (2 to 11 months, 12 to 35
months, or 36 to 50 months), sex, mother’s edu-
cation level (none, primary, or secondary or
higher), number of children in the home (1, 2,
3, or ≥4), and number of amoxicillin or placebo
doses taken (≤4, 5 to 7, 8 or 9, or 10). Analyses
of secondary outcomes used robust standard er-
rors unadjusted for interim analyses or other
factors. Six prespecified subgroup analyses were
performed, of which four are reported here. Pre-
specified subgroups of children with low oxygen
saturation (10 children) and of children with
16 n engl j med 383;1  nejm.org  July 2, 2020
The New England Journal of Medicine
Downloaded from nejm.org on August 17, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
of m e dic i n e
wheezing (49 children) are not reported because
of the small numbers.
R e sult s
Participants
Enrollment started on March 29, 2016. Formal
interim analyses were performed after one third,
two thirds, and slightly more than the original
maximum planned number of children had been
enrolled, and the last visit was completed on
April 14, 2019. In total, 3336 children were
screened (Fig. 1). A total of 265 children were
ineligible, and consent was declined for 82.
However, 11 children were enrolled in error (5 in
the 3-day group and 6 in the 5-day group). Thus,
3000 children were enrolled and underwent ran-
domization: 1497 were assigned to the 3-day
regimen of amoxicillin, and 1503 were assigned
to the 5-day regimen. Primary-outcome data were
available for 1442 children (96.3%) in the 3-day
group and 1456 (96.9%) in the 5-day group.
Primary-outcome data were missing primarily
because of loss to follow-up. Baseline character-
istics were similar in the two groups (Table 1
and Table S3).
Outcomes
By day 6, among children who were included in
the analysis and had complete outcome data,
5.9% of the children (85 of 1442) in the 3-day
group and 5.2% (75 of 1456) in the 5-day group
had treatment failure, which resulted in an ad-
justed absolute difference of 0.7 percentage points
(95% confidence interval [CI], −0.9 to 2.4 [with
a noninferiority margin for an upper limit of the
95% confidence interval of 2.6]) (Table 2).
Among children who did not have treatment
failure by day 6, a total of 91 of 1326 (6.9%) in
the 3-day group, as compared with 79 of 1354
(5.8%) in the 5-day group, had relapse by day 14
— an absolute difference of 1.0 percentage point
(95% CI, −0.8 to 2.9 [with a noninferiority mar-
gin for an upper limit of the 95% confidence
interval of 8.7]).
Before day 4, both groups were receiving
amoxicillin, and as such, we would expect the
incidence of treatment failure before day 4 to be
the same in the two groups. In a post hoc, de-
scriptive, unadjusted analysis, 2.3% (33 of 1442
children) in the 3-day group and 2.3% (33 of
1456 children) in the 5-day group had treatment
 Amoxicillin for Chest-Indr awing Pneumonia in Children

3336 Children were assessed for eligibility

265 Were ineligible

128 Had severe anemia
55 Had severe respiratory distress
43 Had HIV-1 seropositivity or exposure
25 Received antibiotic agent in previous 48 hr
25 Had previous participation in an ITIP trial
9 Had a cough lasting ≥14 days
41 Had other reason

3071 Were eligible

82 Did not have consent provided

11 Were ineligible and were incorrectly enrolled

3000 Were enrolled and underwent

randomization

1497 Were assigned to the 3-day 1503 Were assigned to the 5-day
amoxicillin group amoxicillin group

1 Was lost to follow-up 3 Were lost to follow-up

9 Had consent withdrawn 7 Had consent withdrawn
5 Had missing follow-up 4 Had missing follow-up
data data

1482 Were included in the day 2 follow-up 1489 Were included in the day 2 follow-up

4 Were lost to follow-up 2 Were lost to follow-up

8 Had consent withdrawn 10 Had consent withdrawn
9 Had missing follow-up 8 Had missing follow-up
data data

1482 Were included in the day 4 follow-up 1489 Were included in the day 4 follow-up

8 Were lost to follow-up 4 Were lost to follow-up

3 Had consent withdrawn 1 Had consent withdrawn
12 Had missing follow-up 11 Had missing follow-up
data data

1442 Had outcome data available at day 6 1456 Had outcome data available at day 6

Figure 1. Screening, Randomization, and Follow-up.

Children may have had more than one reason for ineligibility. Missing follow-up data may have been the result of
missed visits or visits that occurred outside the prespecified visit windows. Some children had missing follow-up
data for either day 2 or day 4 or both but had outcome data available for day 6. HIV-1 denotes human immunodefi-
ciency virus type 1, and ITIP Innovative Treatments in Pneumonia.

n engl j med 383;1  nejm.org  July 2, 2020 17

The New England Journal of Medicine
Downloaded from nejm.org on August 17, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

Table 1. Characteristics of the Children at Enrollment.*

3-Day Amoxicillin 5-Day Amoxicillin Overall

Characteristic (N = 1497) (N = 1503) (N = 3000)
Age — no. (%)
2–11 mo 867 (57.9) 869 (57.8) 1736 (57.9)
12–35 mo 509 (34.0) 514 (34.2) 1023 (34.1)
36–59 mo 121 (8.1) 120 (8.0) 241 (8.0)
Male sex — no. (%) 833 (55.6) 820 (54.6) 1653 (55.1)
Mid–upper-arm circumference — no. (%)
<11.5 cm 0 0 0
11.5–13.5 cm 321 (21.4) 304 (20.2) 625 (20.8)
>13.5 cm 1176 (78.6) 1199 (79.8) 2375 (79.2)
Respiratory rate — no./total no. (%)†
Age 2–11 mo
<50 breaths/min 300/867 (34.6) 303/869 (34.9) 603/1736 (34.7)
50–59 breaths/min 361/867 (41.6) 393/869 (45.2) 754/1736 (43.4)
≥60 breaths/min 206/867 (23.8) 173/869 (19.9) 379/1736 (21.8)
Age 12–59 mo
<40 breaths/min 160/630 (25.4) 164/634 (25.9) 324/1264 (25.6)
40–49 breaths/min 251/630 (39.8) 262/634 (41.3) 513/1264 (40.6)
≥50 breaths/min 219/630 (34.8) 208/634 (32.8) 427/1264 (33.8)
Oxygen saturation — no. (%)‡
<90% 0 0 0
90–92% 5 (0.3) 7 (0.5) 12 (0.4)
≥93% 1492 (99.7) 1496 (99.5) 2988 (99.6)
Axillary temperature — no. (%)†
<38°C 1020 (68.1) 1054 (70.1) 2074 (69.1)
≥38°C 477 (31.9) 449 (29.9) 926 (30.9)
Receipt of 13-valent pneumococcal
conjugate vaccine — no. (%)§
Received age-appropriate no. of doses 942 (62.9) 952 (63.3) 1894 (63.1)
Received less than the age-appropriate 555 (37.1) 551 (36.7) 1106 (36.9)
no. of doses or received unknown
no. of doses

* Percentages may not total 100 because of rounding.

† Shown is the value recorded at the time of the enrollment visit or the value recorded at the time of screening, whichever
was higher.
‡ Shown is the value recorded at the time of the enrollment visit or the value recorded at the time of screening, whichever
was lower.
§ The appropriate number of doses according to age is three doses in children 14 weeks of age or older, two doses in
children 10 weeks to 14 weeks of age, and one dose in children 6 weeks to 10 weeks of age.

failure before day 4. On days 4 and 5, the 3-day
group was receiving placebo, whereas the 5-day
group was continuing to receive amoxicillin. In
the analysis of days 4 through 6, the percentage
of children with treatment failure was 3.6% (52
of 1442 children) in the 3-day group and 2.9%
(42 of 1456 children) in the 5-day group.
A total of 176 of 1411 children (12.5%) in the
3-day group and 154 of 1429 (10.8%) in the
5-day group had either treatment failure before

18 n engl j med 383;1  nejm.org  July 2, 2020

The New England Journal of Medicine

Downloaded from nejm.org on August 17, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
 Amoxicillin for Chest-Indr awing Pneumonia in Children

Table 2. Outcomes According to Treatment Group.*

Between-Group
3-Day Amoxicillin 5-Day Amoxicillin Difference
Variable (N = 1497) (N = 1503) (95% CI)†
Primary outcome
Treatment failure on or before day 6 — no./ 85/1442 (5.9) 75/1456 (5.2) 0.7 (−0.9 to 2.4)
total no. with data (%)‡
Secondary outcomes
Relapse on or before day 14 — no./total no. 91/1326 (6.9) 79/1354 (5.8) 1.0 (−0.8 to 2.9)
(%)§
Treatment failure by day 6 or relapse on or 176/1411 (12.5) 154/1429 (10.8) 1.7 (−0.7 to 4.1)
before day 14 — no./total no. (%)
Sensitivity analysis of treatment failure — no. 55 47 0.8 (−0.9 to 2.4)
with missing primary-outcome data
imputed¶
Treatment failure according to prespecified
subgroups‖
Age — no./total no. (%)
2–11 mo 57/832 (6.9) 46/842 (5.5) 1.4 (−0.9 to 3.7)
12–35 mo 23/490 (4.7) 24/498 (4.8) −0.1 (−2.8 to 2.5)
36–59 mo 5/120 (4.2) 5/116 (4.3) −0.1 (−5.3 to 5.0)
Mid–upper-arm circumference — no./total
no. (%)
<11.5 cm 0 0
11.5–13.5 cm 25/309 (8.1) 17/297 (5.7) 2.4 (−1.7 to 6.4)
>13.5 cm 60/1133 (5.3) 58/1159 (5.0) 0.3 (−1.5 to 2.1)
Malaria status — no./total no. (%)
Positive 4/127 (3.1) 5/136 (3.7) −0.5 (−4.9 to 3.9)
Negative 81/1315 (6.2) 70/1320 (5.3) 0.9 (−0.9 to 2.6)
Very fast breathing for age — no./total no. (%)
Yes 5/68 (7.4) 5/59 (8.5) −1.1 (−10.6 to 8.3)
No 80/1374 (5.8) 70/1397 (5.0) 0.8 (−0.9 to 2.5)

* Secondary outcomes and subgroup analyses were prespecified.

† The between-group difference is the difference in percentage points. The values may differ from the expected value be-
cause of rounding.
‡ The between-group difference and the 95% confidence interval were adjusted for age, sex, and trial phase (i.e., enroll-
ment at either Kamuzu Central Hospital or Bwaila District Hospital).
§ This category includes children who had not had treatment failure on or before day 6.
¶ Covariates used in multiple imputation were treatment group, age group, sex, mother’s education level, number of chil-
dren in the home, and the number of doses taken.
‖ The total numbers are the same as those assessed for the primary outcome.

or on day 6 or relapse by day 14 (absolute differ-
ence, 1.7 percentage points; 95% CI, −0.7 to 4.1).
Additional results regarding secondary out-
comes are provided in Table 2. The incidence of
treatment failure was generally consistent across
prespecified subgroups defined according to age
group, malnutrition status, malaria status, and
very fast breathing for age. Most 95% confidence
intervals associated with outcomes in the sub-
groups included the noninferiority margin, and
any adjustment for multiple comparisons would
have resulted in the upper limit of all 95% con-
fidence intervals being larger than the noninfe-
riority margin. The amount of missing primary-
outcome data was low. Overall, data were
missing for 102 of 3000 children (3.4%): 55 of

n engl j med 383;1  nejm.org  July 2, 2020 19

The New England Journal of Medicine
Downloaded from nejm.org on August 17, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l
1497 children in the 3-day group and 47 of 1503
in the 5-day group. Estimates derived from mul-
tiple imputations for missing outcome data were
similar to those derived from the analysis of
complete cases. When we performed a tipping-
point analysis, we failed to show noninferiority
only when among the children with missing data,
at least 3 more children in the 3-day group than
in the 5-day group had treatment failure. If the
same incidence of treatment failure observed in
the analysis that included children with com-
plete data (5.9% in the 3-day group and 5.2% in
the 5-day group) were applied to an analysis
with imputation of missing data (5.9% × 55 in
the 3-day group and 5.2% × 47 in the 5-day group),
the expected average difference in the number of
children with treatment failure would be 0.8. As
such, we would have needed to observe a larger
difference among the children with missing data
(e.g., treatment failure in 3 of 55 children in the
3-day group and 0 of 47 children in the 5-day
group) to fail to conclude noninferiority. If the
results of the primary analysis had been adjusted
for sequential monitoring, the conclusion of
noninferiority would have remained the same.
Adverse Events
The percentage of children with at least one seri-
ous adverse event from the time of enrollment to
day 14 was 9.8% in the 3-day group and 8.8% in
the 5-day group (Table 3). There was one death
(from pneumonia) in the 3-day group (<0.1% of
the children), and two deaths (one from pneu-
monia and one from acute gastroenteritis) in the
5-day group (0.1% of children). In the 3-day
group, 91.6% of children received all doses, and
in the 5-day group, 91.8% of children received
all doses, with adherence reported by caregivers.
Discussion
We evaluated a 3-day regimen as compared with
a 5-day regimen of oral amoxicillin treatment in
3000 HIV-uninfected children 2 to 59 months of
age who presented with chest-indrawing pneu-
monia in a region of Malawi in which malaria
is endemic. Our results showed that 3 days of
amoxicillin treatment was noninferior to 5 days
of treatment with respect to treatment failure on
or before day 6. The criterion for noninferiority
continued to be met through day 14, but there
was no prespecified plan to adjust for multiple
20 n engl j med 383;1  nejm.org  July 2, 2020
The New England Journal of Medicine
Downloaded from nejm.org on August 17, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
of m e dic i n e
comparisons; therefore, definitive conclusions
should not be drawn regarding secondary out-
comes.
The results of this trial suggest that a 3-day
regimen of amoxicillin is not substantively worse
than a 5-day regimen for the treatment of chest-
indrawing pneumonia among HIV-uninfected
children. Keeping in mind the benefits to both
patients and the health care system of a shorter
course of antibiotic therapy and the fact that the
WHO already recommends 3 days of amoxicillin
for treatment of fast-breathing pneumonia,2,5-7 it
appears that a 3-day regimen of amoxicillin in
children with chest-indrawing pneumonia might
be sufficient. Currently, the WHO recommends
a 5-day course of twice-daily, high-dose oral
amoxicillin to treat chest-indrawing pneumonia
in children with cough or with difficulty breath-
ing.2,19 However, the findings of this trial may
allow for harmonization and simplification of
treatment courses to 3 days for both fast-breath-
ing and chest-indrawing pneumonia in HIV-
uninfected children. A study from Pakistan
showed that in cases of chest-indrawing pneu-
monia in patients without underlying complica-
tions, home treatment with a short course of
high-dose oral amoxicillin was preferable to par-
enteral treatment because of the associated reduc-
tion in referral, admission, and treatment costs.19
Home treatment of chest-indrawing pneumonia
with oral amoxicillin is effective across commu-
nities and geographic regions.20-22 In contrast to
criteria in low-resource settings, criteria for the
diagnosis of pneumonia in high-resource set-
tings often require confirmation on chest radi-
ography, especially in hospitalized children.23
Yet little evidence exists to dictate duration of
treatment.11 Of note, in a very small study from
Israel, a 3-day course of high-dose oral amoxicil-
lin was associated with a high incidence of treat-
ment failure (40% [4 of 10 children]) among
children with radiographically confirmed pneu-
monia.24
Poor adherence to antibiotic treatment has
been associated with treatment failure in WHO-
defined clinical pneumonia.25,26 Improving ad-
herence to a shorter course of treatment could
improve outcomes among children with chest-
indrawing pneumonia while also minimizing
adverse drug effects, costs, and the emergence
of antimicrobial resistance.7,25,26
Limitations of our trial included strict inclu-
 Amoxicillin for Chest-Indr awing Pneumonia in Children

Table 3. Serious Adverse Events and Common Nonserious Adverse Events According to Treatment Group.*

3-Day Amoxicillin 5-Day Amoxicillin Overall

Variable (N = 1497) (N = 1503) (N = 3000)

number (percent)
Children with ≥1 serious adverse event† 147 (9.8) 132 (8.8) 279 (9.3)
Children with ≥1 nonserious adverse event† 394 (26.3) 455 (30.3) 849 (28.3)
Serious adverse events
Pneumonia 135 (9.0) 118 (7.9) 253 (8.4)
Chest-indrawing pneumonia‡ 61 (4.1) 49 (3.3) 110 (3.7)
Pneumonia with any danger sign 49 (3.3) 51 (3.4) 100 (3.3)
Pneumonia with fast breathing for age§ 17 (1.1) 14 (0.9) 31 (1.0)
Pneumonia confirmed on chest 7 (0.5) 3 (0.2) 10 (0.3)
radiography¶
Pneumonia of unspecified type 1 (0.1) 1 (<0.1) 2 (0.1)
Nonpneumonia events 20 (1.3) 15 (1.0) 35 (1.2)
Gastroenteritis 8 (0.5) 6 (0.4) 14 (0.5)
Fever 3 (0.2) 5 (0.3) 8 (0.3)
Malaria 1 (0.1) 2 (0.1) 3 (0.1)
Meningitis 3 (0.2) 0 3 (0.1)
Otitis media 2 (0.1) 0 2 (0.1)
Conjunctivitis 1 (0.1) 0 1 (<0.1)
Edema 0 1 (<0.1) 1 (<0.1)
Febrile seizure 1 (0.1) 0 1 (<0.1)
Rectal prolapse 1 (0.1) 0 1 (<0.1)
Vomiting 0 1 (<0.1) 1 (<0.1)
Common nonserious adverse events
Gastroenteritis 176 (11.8) 223 (14.8) 399 (13.3)
Upper respiratory infection 113 (7.5) 114 (7.6) 227 (7.6)
Rash 32 (2.1) 50 (3.3) 82 (2.7)
Conjunctivitis 21 (1.4) 20 (1.3) 41 (1.4)
Rhinitis 22 (1.5) 15 (1.0) 37 (1.2)
Otitis media 13 (0.9) 21 (1.4) 34 (1.1)
Eczema 15 (1.0) 17 (1.1) 32 (1.1)
Oral candidiasis 13 (0.9) 15 (1.0) 28 (0.9)

* Children may have had more than 1 serious or nonserious adverse event and may have had multiple serious or nonse-
rious adverse events.
† Included are adverse events that occurred any time after the trial drug was administered, up to 14 days after enrollment.
‡ Chest-indrawing pneumonia is defined as cough lasting less than 14 days or difficulty breathing, along with visible in-
drawing of the chest wall with or without fast breathing for age.
§ A total of 37 events occurred on or before day 6 and were considered treatment failures. The remaining events occurred
after day 6 and thus were considered relapses.
¶ The events in this category did not include pneumonia manifested with fast breathing, chest indrawing, or any danger
signs; however, pneumonia was diagnosed by means of positive chest radiography.

sion and exclusion criteria, lack of laboratory or ity of our results to routine programmatic care
radiologic testing, and close monitoring and settings. It is important to note that children
follow-up — factors that limit the generalizabil- with severe disease were excluded; therefore, the

n engl j med 383;1  nejm.org  July 2, 2020 21

The New England Journal of Medicine
Downloaded from nejm.org on August 17, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

results of the trial have limited applicability in
this group. Pneumonia is frequently considered
a single entity, rather than a clinical syndrome
that encompasses several underlying factors, so
interpretation of the results of this trial is chal-
lenging. A multicountry case–control study that
evaluated the causes of chest-indrawing pneu-
monia in children younger than 5 years of age
across a range of geographic and epidemiologic
settings showed that viruses accounted for 61.4%
of cases, bacteria for 27.3%, and Mycobacterium
tuberculosis for 5.9% of cases.27 Without etiologic
information, including ascertainment of viral as
compared with bacterial causes, we could note
the effect of the intervention only on the clinical
syndrome of pneumonia. This approach is con-
sistent with nontrial conditions relevant to pedi-
atric care in low-resource settings, but it blunts
the ability to detect a difference between a 3-day
course and a 5-day course.
Follow-up care and monitoring of enrolled
children generally exceeded local standards of
care; therefore, the incidence of treatment fail-
ure may have been affected both by the high qual-
ity of care that was provided and by the high
awareness of and vigilance for detecting treat-
ment failure. It may be that the children in whom
treatment failure was identified would have re-
covered without a longer course of antibiotics if
we had used a watchful waiting approach and
not intervened with antibiotic treatment. How-
ever, opportunities for follow-up and access to
care are often challenging in low-resource set-
tings. In addition, treatment approaches vary
widely among countries and regions. The routine
pediatric HIV testing included in this trial pro-
tocol, although recommended, is not rigorously
implemented during routine care in low-resource
settings in which HIV is endemic.28 In areas
where pneumococcal immunization coverage is
lower, where HIV endemicity is high, or where
severe acute malnutrition or other conditions
that confer a predisposition to bacterial disease
are common, it may be reasonable to expect a
higher incidence of treatment failure among
patients who are not treated with a longer course
of antibiotics. As such, our results might not be
generalizable across different regions, settings,
or nontrial conditions. Specifically, the percent-
age of children who had treatment failure or re-
lapse that was observed in this trial might under-
estimate the true incidences of treatment failure
and relapse that occur in nontrial conditions.
Although pneumonia is a common and dead-
ly illness, the duration of antibiotic treatment for
community-acquired pediatric pneumonia has
not yet been clearly defined. In this population
in Malawi, a 3-day regimen of amoxicillin was
noninferior to a 5-day regimen in HIV-uninfect-
ed children with chest-indrawing pneumonia. In
considering policy changes regarding duration
of amoxicillin treatment for chest-indrawing pneu-
monia, further research may be needed to deter-
mine whether these results can be replicated in
other low-resource regions and pediatric popula-
tions.
Supported by a grant (OPP1105080) from the Bill and Melinda
Gates Foundation.
Disclosure forms provided by the authors are available with
the full text of this article at NEJM.org.
A data sharing statement provided by the authors is available
with the full text of this article at NEJM.org.
We thank Rasa Izadnegahdar for his technical support and
advice; Gwen Ambler for her help with the planning of trial
implementation; the dedicated trial staff at the University of
North Carolina Project, Lilongwe Medical Relief Fund Trust,
and at Kamuzu Central Hospital for providing patient care;
Triclinium Clinical Development for facilitating data manage-
ment and safety monitoring; the Malawi Ministry of Health for
their support; the members of the data and safety monitoring
board (Shamim A. Qazi [chair], Christopher T. Roberts, Grace J.
Malenga, and Harry Campbell); and the trial participants, their
caregivers, and the local community in Lilongwe, Malawi, for
their participation.

References
1. Liu L, Oza S, Hogan D, et al. Global,
regional, and national causes of child
mortality in 2000-13, with projections to
inform post-2015 priorities: an updated sys-
tematic analysis. Lancet 2015;​385:​430-40.
2. Revised WHO classification and treat-
ment of pneumonia in children at health
facilities:​evidence summaries. Geneva:​
World Health Organization, 2014.
3. Integrated management of childhood
illness:​chart booklet. Geneva:​World
Health Organization, 2014.
4. Lassi ZS, Das JK, Waqas Haider S,
Salam RA, Qazi SA, Bhutta ZA. System-
atic review on antibiotic therapy for pneu-
monia in children between 2 and 59
months of age. Arch Dis Child 2014;​99:​
687-93.
5. Haider BA, Saeed MA, Bhutta ZA.
Short-course versus long-course antibi-
otic therapy for non-severe community-­
acquired pneumonia in children aged
2 months to 59 months. Cochrane Data-
base Syst Rev 2008;​2:​CD005976.
6. Pakistan Multicentre Amoxycillin Short
Course Therapy (MASCOT) Pneumonia
Study Group. Clinical efficacy of 3 days

22 n engl j med 383;1  nejm.org  July 2, 2020

The New England Journal of Medicine

Downloaded from nejm.org on August 17, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.
 Amoxicillin for Chest-Indr awing Pneumonia in Children

versus 5 days of oral amoxicillin for treat-
ment of childhood pneumonia: a multi-
centre double-blind trial. Lancet 2002;​
360:​835-41.
7. Agarwal G, Awasthi S, Kabra SK, Kaul
A, Singhi S, Walter SD. Three day versus
five day treatment with amoxicillin for
non-severe pneumonia in young children:
a multicentre randomised controlled trial.
BMJ 2004;​328:​791.
8. Lassi ZS, Imdad A, Bhutta ZA. Short-
course versus long-course intravenous
therapy with the same antibiotic for severe
community-acquired pneumonia in chil-
dren aged two months to 59 months.
Cochrane Database Syst Rev 2017;​
­ 10:​
CD008032.
9. McMullan BJ, Andresen D, Blyth CC,
et al. Antibiotic duration and timing of
the switch from intravenous to oral route
for bacterial infections in children: sys-
tematic review and guidelines. Lancet In-
fect Dis 2016;​16(8):​e139-e152.
10. Das RR, Singh M. Treatment of severe
community-acquired pneumonia with oral
amoxicillin in under-five children in de-
veloping country: a systematic review.
PLoS One 2013;​8(6):​e66232.
11. Grimwood K, Fong SM, Ooi MH, Na-
than AM, Chang AB. Antibiotics in child-
hood pneumonia: how long is long
enough? Pneumonia (Nathan) 2016;​8:​6.
12. Ginsburg AS, May SJ, Nkwopara E,
et al. Methods for conducting a double-
blind randomized controlled clinical trial
of three days versus five days of amoxicil-
lin dispersible tablets for chest indrawing
childhood pneumonia among children two
to 59 months of age in Lilongwe, Malawi:
a study protocol. BMC Infect Dis 2018;​18:​
476.
13. O’Brien PC, Fleming TR. A multiple
testing procedure for clinical trials. Bio-
metrics 1979;​35:​549-56.
14. Pocock SJ. Group sequential methods
in the design and analysis of clinical trials.
Biometrika 1977;​64:​191-9.
15. Huber PJ. The behavior of maximum
likelihood estimates under nonstandard
conditions. In:​Proceedings of the Fifth
Berkeley Symposium on Mathematical
Statistics and Probability. Berkeley:​Uni-
versity of California Press, 1967:​221-33.
16. White HL Jr. Maximum likelihood es-
timation of misspecified models. Econo-
metrica 1982;​50:​1-25.
17. Lumley T, Diehr P, Emerson S, Chen
L. The importance of the normality as-
sumption in large public health data sets.
Annu Rev Public Health 2002;​23:​151-69.
18. Andridge RR, Little RJ. A review of hot
deck imputation for survey non-response.
Int Stat Rev 2010;​78:​40-64.
19. Hazir T, Fox LM, Nisar YB, et al. Am-
bulatory short-course high-dose oral amox-
icillin for treatment of severe pneumonia
in children: a randomised equivalency
trial. Lancet 2008;​371:​49-56.
20. Addo-Yobo E, Anh DD, El-Sayed HF,
et al. Outpatient treatment of children
with severe pneumonia with oral amoxi-
cillin in four countries: the MASS study.
Trop Med Int Health 2011;​16:​995-1006.
21. Bari A, Sadruddin S, Khan A, et al.
Community case management of severe
pneumonia with oral amoxicillin in chil-
dren aged 2-59 months in Haripur dis-
trict, Pakistan: a cluster randomised trial.
Lancet 2011;​378:​1796-803.
22. Soofi S, Ahmed S, Fox MP, et al. Ef-
fectiveness of community case manage-
ment of severe pneumonia with oral
amoxicillin in children aged 2-59 months
in Matiari district, rural Pakistan: a clus-
ter-randomised controlled trial. Lancet
2012;​379:​729-37.
23. Nascimento-Carvalho CM, Madhi SA,
O’Brien KL. Review of guidelines for evi-
dence-based management for childhood
community-acquired pneumonia in under-
5 years from developed and developing
countries. Pediatr Infect Dis J 2013;​32:​
1281-2.
24. Greenberg D, Givon-Lavi N, Sadaka Y,
Ben-Shimol S, Bar-Ziv J, Dagan R. Short-
course antibiotic treatment for community-
acquired alveolar pneumonia in ambula-
tory children: a double-blind, randomized,
placebo-controlled trial. Pediatr Infect
Dis J 2014;​33:​136-42.
25. Nightingale R, Colbourn T, Mukanga
D, et al. Non-adherence to community
oral-antibiotic treatment in children with
fast-breathing pneumonia in Malawi —
secondary analysis of a prospective co-
hort study. Pneumonia (Nathan) 2016;​8:​
21.
26. King C, Nightingale R, Phiri T, et al.
Non-adherence to oral antibiotics for
community paediatric pneumonia treat-
ment in Malawi — a qualitative investiga-
tion. PLoS One 2018;​13(10):​e0206404.
27. Pneumonia Etiology Research for
Child Health (PERCH) Study Group. Causes
of severe pneumonia requiring hospital
admission in children without HIV infec-
tion from Africa and Asia: the PERCH
multi-country case-control study. Lancet
2019;​394:​757-79.
28. Theodoratou E, McAllister DA, Reed
C, et al. Global, regional, and national
estimates of pneumonia burden in HIV-
infected children in 2010: a meta-analysis
and modelling study. Lancet Infect Dis
2014;​14:​1250-8.
Copyright © 2020 Massachusetts Medical Society.

n engl j med 383;1  nejm.org  July 2, 2020 23

The New England Journal of Medicine
Downloaded from nejm.org on August 17, 2020. For personal use only. No other uses without permission.
Copyright © 2020 Massachusetts Medical Society. All rights reserved.