The pill that replaced the needle

In a trial across 13 hospitals in five sub-Saharan African countries, children with severe pneumonia who switched early from drips to oral amoxicillin did at least as well as children kept on intravenous antibiotics: 5.6% were readmitted or died by day 28, against 6.3% on IV-only treatment ⁴. That result builds on earlier community-treatment and short-course trials. The distinction is which children can start with oral treatment and which need hospital care before stepping down.

The burden

Pneumonia remains one of the biggest killers of young children. The latest estimates of the causes of child death, published in the BMJ in June 2026 by Perin and colleagues and built on the all-cause death counts of the UN Inter-agency Group for Child Mortality Estimation (IGME), put total deaths in children under 5 at 4.9 million in 2024 ². Lower respiratory infections, the category that includes pneumonia, caused 0.66 million of those deaths (90% uncertainty interval 0.60 to 0.71 million) ². Only complications of preterm birth killed more ².

The Global Burden of Disease (GBD) study at the Institute for Health Metrics and Evaluation (IHME) uses different methods and arrives at a similar order of size. It estimated 502,000 deaths from lower respiratory infections in children under 5 in 2021 (95% uncertainty interval 406,000 to 611,000) ¹. About half of them, 254,000, were in countries with a low Socio-demographic Index, that is, the poorest countries ¹.

WHO's own fact sheet, based on 2019 data, gives 740,180 pneumonia deaths in children under 5, or 14% of all under-5 deaths and 22% of deaths in children aged 1 to 5 years ³. Deaths are highest in southern Asia and sub-Saharan Africa ³.

The different figures come from different years and models. All of them agree on the size: hundreds of thousands of young children die each year from an illness that is largely preventable and treatable.

The trend is down, but slowing. The GBD team found that the all-age death rate from lower respiratory infections fell by 41.7% between 1990 and 2019 ¹. The BMJ analysis found that the decline in most causes of under-5 death has slowed since 2016 ².

The most common bacterial cause is Streptococcus pneumoniae. In 2021 it accounted for an estimated 505,000 lower respiratory infection deaths across all ages ¹. This is the bacterium that amoxicillin is chosen to treat.

The gap in treatment is large. According to WHO, only one third of children with pneumonia receive the antibiotics they need ³.

How it works

Pneumonia is an infection of the lungs. The small air sacs fill with fluid and pus, so the child cannot take in enough oxygen. The child breathes faster to compensate. In more severe cases, the lower chest wall pulls in with each breath. Health workers call this "chest indrawing".

WHO's Integrated Management of Childhood Illness (IMCI) approach uses these two signs to sort children aged 2 to 59 months into groups:

• Fast breathing pneumonia. The child breathes faster than a set rate for their age, but has no chest indrawing and no danger signs.
• Chest-indrawing pneumonia. The lower chest pulls in, but the child has no danger signs.
• Severe pneumonia. The child has a general danger sign. These children are referred for hospital care.

Amoxicillin is a penicillin-type antibiotic, given to young children as a dispersible tablet.

WHO changed its guidance in 2012 to allow outpatient oral amoxicillin for chest-indrawing pneumonia [7,10]. WHO's 2024 guideline on pneumonia and diarrhoea in children up to 10 years confirmed oral amoxicillin as first-line treatment for children aged 2 to 59 months with fast breathing or chest indrawing, and said these children can be managed well in community settings ¹¹.

In hospital, WHO recommends 5 days of intravenous antibiotics for children admitted with severe community-acquired pneumonia ⁴.

Across this 15-year window, trials ask which children can safely receive oral treatment, when hospital treatment can step down, and how long a course needs to be.

What the evidence shows

2012–2020: community treatment and shorter courses

In Matiari, Pakistan, a 2012 cluster trial tested oral amoxicillin delivered by Lady Health Workers against a first antibiotic dose followed by referral. Among 4,410 children aged 2–59 months, treatment failure was 8% versus 13%; the cluster-adjusted difference was −5.2 percentage points (95% CI −13.7 to 3.3) ¹⁵. Both groups received repeated home follow-up. This was evidence about a community-care strategy where referral was difficult, not a direct comparison with completed inpatient treatment.

Two 2020 trials sharpened the prescribing question. In Karachi, placebo failed the non-inferiority test against three days of amoxicillin for fast-breathing pneumonia: treatment failure was 4.9% versus 2.6%, with 44 children needing treatment to prevent one failure ¹⁶. In Malawi, three days were non-inferior to five days for chest-indrawing pneumonia in 3,000 HIV-uninfected children: day-6 failure was 5.9% versus 5.2% ¹⁷. Neither result establishes that antibiotics can be omitted, or that oral treatment alone is adequate for every child with danger signs. PediCAP extends this evidence to selected hospitalised children who improve enough to switch treatment.

PediCAP: step down early, stop sooner

PediCAP was led by Julia Bielicki and colleagues at City St George's, University of London, with partner hospitals in five sub-Saharan African countries ⁴. It was funded by the Second European and Developing Countries Clinical Trials Partnership (EDCTP2) and registered as ISRCTN63115131 ⁴.

The trial enrolled children aged 2 months to 6 years admitted with severe community-acquired pneumonia without complicating factors ⁴. They had to weigh 3 to 30 kg and have a point-of-care C-reactive protein (a blood marker of inflammation) above 10 mg/L ⁴. Between December 2020 and August 2023, 2,248 children were screened and 1,101 were randomised ⁴.

The design was factorial, which means it answered several questions at once ⁴. Children were randomised to step down from IV antibiotics to either oral amoxicillin or oral co-amoxiclav (amoxicillin with clavulanate), each given as dispersible tablets twice a day ⁴. They were also randomised to a total treatment length of 4, 5, 6, 7 or 8 days ⁴. A small group stayed on a fixed 5-day IV-only course, the current WHO approach ⁴. Children stepped down when they improved and could take tablets ⁴.

The primary outcome was readmission or death by day 28 ⁴. The primary outcome was available for 1,055 children (95.8%) ⁴.

• Amoxicillin step-down: 5.6% (27 of 484) ⁴.
• Co-amoxiclav step-down: 6.9% (33 of 475) ⁴.
• IV-only: 6.3% (6 of 96) ⁴.

Both oral strategies were non-inferior to IV-only care ⁴. Co-amoxiclav was not better than amoxicillin (adjusted risk difference 1.3%, 95% CI -1.8 to 4.4; p=0.40) ⁴.

For duration, the primary outcome occurred in 4.1% of the 4-day group, 5.3% of the 5-day group, 8.5% of the 6-day group, 8.3% of the 7-day group and 5.2% of the 8-day group ⁴. All shorter durations were non-inferior to 8 days ⁴.

Adverse events told a mixed story. Antibiotic-related or serious adverse events occurred in 1.0% of the IV-only group and 2.5% of the amoxicillin group (adjusted risk difference -2.3%, 95% CI -4.2 to -0.3) ⁴. They occurred in 3.4% of the co-amoxiclav group ⁴. The rate of these events rose as the randomised duration got longer ⁴. That is one more reason to prefer the shorter course.

The authors concluded that stepping down to oral amoxicillin, with a total of 4 to 5 days of treatment, is non-inferior to the WHO-recommended 5 days of IV treatment ⁴.

Caveats. The trial was open-label, so staff and families knew which treatment each child received ⁴. The IV-only group was small (96 children), by design ⁴. Children did not always stop at their assigned duration: in the 4-day group, only 68.6% stepped down within their randomised duration, and the mean total antibiotic exposure was 6.0 days ⁴. The trial excluded children with complicating factors, so the results apply to uncomplicated severe pneumonia ⁴.

The WHO-commissioned review: tablet against injection, and against nothing

For the 2024 WHO guideline, the Society for Applied Studies in New Delhi pooled the randomised trials ⁵. The review searched trials published from January 2000 to August 2024 ⁵.

Fast breathing pneumonia. Four trials with 7,699 children compared oral amoxicillin with no antibiotic ⁵. Amoxicillin cut treatment failure at day 14 by 16% (RR 0.84, 95% CI 0.75-0.94; moderate certainty) ⁵. So for these milder cases, the antibiotic does help.

Chest-indrawing pneumonia. Three trials with 4,864 children compared outpatient oral amoxicillin with inpatient injectable antibiotics ⁵. Treatment failure was similar at day 6 (RR 0.96, 95% CI 0.83-1.11) and at day 14 (RR 0.94, 95% CI 0.79-1.13) ⁵. Mortality was lower with oral amoxicillin (RR 0.28, 95% CI 0.09-0.86), though the absolute differences were small ⁵. Serious adverse events did not differ between groups ⁵.

Caveats. The mortality finding is based on few deaths, so the authors themselves stress the small absolute difference ⁵. The trials were run under research conditions with trained staff and close follow-up.

Community health workers: treatment at the doorstep

A second review by the same New Delhi team asked whether community-level health workers could manage chest-indrawing pneumonia, rather than sending the child to a facility ⁶. It found three randomised trials in Africa and Asia with 11,779 children ⁶.

Treatment failure by day 6 was 34% lower with community-worker care than with standard care (RR 0.66, 95% CI 0.44-0.98) ⁶. Mortality did not differ significantly (RR 0.85, 95% CI 0.31-2.30) ⁶. Other serious adverse events also did not differ (RR 1.20, 95% CI 0.34-4.26) ⁶. All results were rated low certainty ⁶.

WHO's 2024 guideline reflects this: children with chest indrawing and no danger signs can be managed in the community ¹¹.

Ethiopia: University of Gondar

WHO coordinated programme studies in several countries to check whether the trial results hold in routine government clinics. These studies were registered together as ISRCTN12687253 [7,8].

In Ethiopia, Tigabu and colleagues at the University of Gondar worked in five health centres in northern Ethiopia from November 2022 to November 2023 ⁷. Health workers screened 3,492 children with cough or difficult breathing and enrolled 345 with chest-indrawing pneumonia ⁷.

All 345 received oral amoxicillin, and 98.6% received a 5-day prescription ⁷. Of 333 children assessed on day 15, two died, a case fatality of 0.6% ⁷. Most children (94.6%) completed the 5-day course ⁷. Thirteen children (3.9%) were taken to hospital; all 13 were alive and well on day 15 ⁷. Twelve children (3.5%) were lost to follow-up ⁷.

Pakistan: Trust for Vaccines & Immunization, Karachi

In Pakistan, the government revised its national IMCI chart booklet in 2019 to follow the WHO outpatient recommendation ⁸. Suhag and colleagues at the Trust for Vaccines & Immunization studied three primary care facilities in Thatta district, Sindh province, from December 2022 to March 2024 ⁸.

They enrolled 456 children. Two died (0.4%) ⁸. Of 452 followed up on day 15, 96.3% were cured, 2.7% did not improve, and two worsened and were admitted ⁸. All children were treated with oral antibiotics ⁸. But only 62.4% received amoxicillin; of those, 83.7% took it for five days or more ⁸. A quarter (25.2%) received cefixime ⁸.

Zambia: University of Zambia

Zambia has about 6,000 under-5 pneumonia deaths each year, according to the study authors ⁹. Jacobs and colleagues at the University of Zambia worked in eight primary health facilities in Lusaka, Chibombo and Chongwe districts from October 2022 to April 2024 ⁹.

They enrolled 335 children; 23.6% were severely stunted ⁹. All 314 children with follow-up data survived, and none failed therapy ⁹. Most (84.1%) received amoxicillin, with a cure rate of 82.2% ⁹. The rest were "clinically better". Cure rates were lower with cotrimoxazole (60%) and erythromycin (26.7%) ⁹. Fully vaccinated children had a higher cure rate (79.6%) than partially or unvaccinated children (48.0%) ⁹.

Results at a glance

Case fatality below 1% in routine care is the signal that the trial results hold up outside research conditions [7-9].

LimitsWhat didn't work (or the limits)

Diagnosis is the bottleneck. The most important negative result comes from Lagos. A University of Ibadan team worked within the INSPIRING Lagos project in 16 primary health centres in Ikorodu Local Government Area, from September 2021 to September 2023 ¹⁰. Research staff identified 247 children with chest-indrawing pneumonia, while the clinic health workers, all trained in IMCI, identified only 24 ¹⁰. Only 1 of those 24 children (4%) received care aligned with IMCI ¹⁰. Of the children prescribed antibiotics, only half (4 of 8) completed the course ¹⁰. Only 1 of 6 referred children was admitted to hospital ¹⁰.

The case fatality estimates there are fragile: one death among 16 followed-up clinic-identified cases (6.3%, 95% CI 0.2-30.2), and the same single death among 197 research-identified cases (0.5%) ¹⁰. One-third of patients were lost to follow-up ¹⁰. A pill cannot help a child whose pneumonia is never recognised.

The wrong drug still gets prescribed. In the Pakistan cohort, a quarter of children received cefixime, a broader-spectrum antibiotic, instead of amoxicillin ⁸. In Zambia, children given cotrimoxazole or erythromycin had much lower cure rates than those given amoxicillin ⁹. Drug choice matters for the child and for antibiotic resistance.

Broader is not better. PediCAP found no benefit from co-amoxiclav over amoxicillin in hospital, and a slightly higher rate of adverse events ⁴.

Not every result is certain. The community health worker review rated all its evidence as low certainty and found no significant difference in mortality ⁶. The lower mortality with oral treatment in the clinic meta-analysis came with small absolute differences ⁵. PediCAP was open-label, and many children in the shortest arms took antibiotics for longer than assigned ⁴.

Cohorts are not trials. The Ethiopia, Pakistan and Zambia studies had no comparison group [7-9]. They show that outpatient care was safe in these clinics, not that it was better than referral. Their outcomes at day 15 relied partly on caregiver reports ⁸. They were also run with research follow-up, which may have improved care compared with an ordinary clinic.

Older children remain a gap. WHO's 2024 update notes that big knowledge gaps remain on the best antibiotic regimens and diagnosis for pneumonia in children aged 5 to 9 years ¹¹.

The treatment gap is still large. WHO estimates that only a third of children with pneumonia get the antibiotics they need ³. Better evidence on the drug does not close that gap on its own.

Cost and scale

Amoxicillin dispersible tablets are cheap and widely made. A 2022 market report by Results for Development (R4D) found that the range in manufacturers' average retail price fell by 35% between 2016 and 2019 ¹². In 2014, partners estimated there was enough manufacturing capacity to meet five times global demand ¹².

UNICEF is a major buyer. Its procurement volumes of amoxicillin dispersible tablets increased threefold after 2015 ¹². Across four high-burden countries, the number of registered amoxicillin DT products doubled between 2015 and 2021, and local manufacturing capacity improved in some sub-Saharan African countries ¹². In three high-burden countries, public-sector tenders from 2016 to 2020 went to seven different manufacturers ¹².

Quality assurance is a weak point. R4D found only four manufacturers of amoxicillin DT 250 mg held a global quality designation for the finished product, and none had WHO prequalification at the time ¹². About one-third of government-funded tenders in three high-burden countries since 2016 went to international manufacturers without a global quality designation ¹².

Funding is the biggest barrier. R4D reported that pneumonia received only 7% ($7.1 billion) of global donor funding given to malaria and HIV/AIDS combined over 2007-2018 ¹². Of the donor money for pneumonia, most went to vaccines and only 1% ($80.1 million) went to treatment such as amoxicillin DT ¹².

Governments are increasingly paying. R4D reports that government funds are the main source for maternal, newborn and child health medicines, including amoxicillin DT ¹². In one high-burden country, the government allocated $1.1 million since 2017, about 5.6 million treatment courses; before 2017 the drug had been fully donor-funded there ¹². Uganda and some other countries have used the Global Financing Facility to fund commodity procurement, and since 2021 countries can use their Global Fund co-financing requirement for non-malaria commodities including amoxicillin DT ¹². R4D also found 13 to 16 countries that did not include amoxicillin DT in their clinical treatment guidelines and essential medicines lists ¹².

The newer studies point to further savings. PediCAP's authors say that earlier step-down and shorter courses free beds, reduce IV exposure and lower cost ⁴. The Pakistan and Zambia teams say outpatient treatment reduces costs for health systems and families [8,9]. None of these studies published a full cost analysis, so we do not give a cost per life saved.

Delivery happens through ministries of health and their IMCI programmes. Ethiopia adopted the revised WHO guidelines after 2012 ⁷. Pakistan revised its national IMCI chart booklet in 2019 ⁸. Community case management programmes, where they function, can extend this to the doorstep [6,11].

What to watch

Even shorter courses. WHO's 2024 guideline already allows 3 days of amoxicillin for fast-breathing pneumonia in areas with low HIV prevalence ¹¹. At Khyber Teaching Hospital in Pakistan, a randomised trial (NCT07647835) plans to compare 3 and 5 days of amoxicillin in 250 children aged 3 months to 5 years with uncomplicated community-acquired pneumonia; it was listed as not yet recruiting in June 2026 ¹³. Its primary outcomes are clinical cure at day 14 and treatment failure ¹³.

Short courses for hospitalised children. At the Medical College of Wisconsin in the United States, a non-inferiority trial (NCT06494072) is comparing a 5-day course with a standard 7 to 14 day course in 236 hospitalised children aged 3 months to 18 years with uncomplicated pneumonia ¹⁴. It was recruiting in 2026, with primary completion expected in May 2028 ¹⁴. It is a high-income setting, but it tests the same question PediCAP answered for Africa.

Uptake of PediCAP. PediCAP was published in July 2026 ⁴. The current WHO recommendation for hospitalised children is still 5 days of IV antibiotics ⁴. Whether WHO and national programmes adopt early oral step-down is the next thing to watch.

Diagnosis tools and training. The Lagos findings show that clinic staff miss most cases ¹⁰. Training and supervision in IMCI, and simple tools that help health workers count breathing and spot chest indrawing, are the obvious next target.

Quality-assured supply. Watch whether more manufacturers gain WHO prequalification and whether national tenders start to require it ¹².

What it means

For ministries of health. The drug question is largely settled. Dispersible amoxicillin, given early and for short courses, is as safe as injections for most children with pneumonia, from district hospitals to community health workers [4-6]. Ethiopia, Pakistan and Zambia have shown it works in government clinics, with case fatality below 1% [7-9]. Ministries that have not yet put amoxicillin DT in their treatment guidelines and essential medicines lists should do so ¹².

The remaining gains are in the people and systems around the pill. Clinic staff need the skills, time and supervision to count breaths and look for chest indrawing, as the Lagos findings make plain ¹⁰. Supply chains and prescribing habits should favour amoxicillin over cefixime, cotrimoxazole or erythromycin [8,9].

For hospitals and clinicians. PediCAP offers a practical change: once a child with uncomplicated severe pneumonia improves and can swallow, switch to amoxicillin tablets and stop at 4 to 5 days in total ⁴. Co-amoxiclav adds nothing for these children ⁴. Follow national guidance while it catches up with the trial.

For community programmes. Where community health worker programmes work well, children with chest indrawing and no danger signs can be treated close to home [6,11]. Referral should be kept for children with danger signs.

For funders. Pneumonia treatment has received a tiny share of donor money for decades ¹². The case is simple: a cheap tablet, already on essential medicines lists, now has trial and programme evidence behind it at every level of care. The best investments are in quality-assured supply, health-worker training and supervision, and co-financing that helps governments take over procurement ¹².

Every child who is diagnosed and given a five-day box of tablets, rather than referred to a hospital they cannot reach, is a child with a better chance of surviving.