Azithromycin for child survival: benefit and resistance

In Niger, giving every child aged 1–59 months a dose of azithromycin twice a year cut under-5 mortality by 14% (95% CI 7–22) ¹. The AVENIR trial, run by Nigerien partners with the UCSF Proctor Foundation, followed 382,586 children in 3,000 communities and counted 5,503 deaths ¹. Evidence from MORDOR onward tests who benefits, how resistance changes with repeated treatment, and what a national programme would cost.

This article walks through that evidence trial by trial. It also sets out what did not work, what a programme costs, and what the World Health Organization (WHO) is now reviewing.

The burden

Most child deaths in the world now happen in one region. Sub-Saharan Africa accounts for 57% of under-5 deaths, even though it has only 30% of the world's live births ¹⁷. Its neonatal mortality rate is 27 per 1000 live births, against 21 per 1000 in central and southern Asia ¹⁷.

Progress has been real but has slowed. Newborn deaths fell from 5.0 million in 1990 to 2.3 million in 2022, but the rate of improvement has slowed a lot since 2010 ¹⁷. Those 2.3 million newborn deaths are 47% of all under-5 deaths ¹⁷. The other half are children aged one month to five years: the age group that azithromycin mass drug administration (MDA) targets.

In the places where azithromycin trials were run, mortality after the first month of life is still high. In the MORDOR trial, children aged 1–59 months in placebo communities in Niger died at 27.5 per 1000 person-years. In Malawi the rate was 9.6, and in Tanzania 5.5 ². Later, the AVENIR placebo arm in Niger still recorded 13.9 deaths per 1000 person-years ¹.

Most of these deaths are from infections. In the Burkina Faso CHAT trial, verbal autopsies found that the most common causes of death were malaria (34%), diarrhoea (24%) and pneumonia (9%) ⁵. These are exactly the conditions an antibiotic with some antimalarial effect might reach.

How it works

Azithromycin is a macrolide antibiotic. It is taken by mouth. In child-survival programmes, distributors go door to door twice a year. They give every eligible child a single dose of about 20 mg per kilogram of body weight, whether the child is sick or not [2,11].

The idea is simple. Many young children in high-mortality areas carry infections that have not yet become severe, or will catch one soon. A single dose may clear some of these before they kill. Researchers do not yet know exactly which infections matter most. The evidence so far points to several:

• Malaria. Azithromycin has some antimalarial effect. In Burkina Faso, malaria deaths were a third lower in treated communities ⁵. In Mali, a substudy of the LAKANA trial found malaria prevalence two weeks after a dose was 8.0% with azithromycin and 8.9% with placebo (risk ratio 0.96; 95% CI 0.63–1.48). The authors say this is consistent with, but does not prove, an antimalarial effect ²¹.
• Respiratory and gut infections. A trial in Niger found lower deaths from malaria, dysentery, meningitis and pneumonia with azithromycin, according to the CHAT authors' summary of that earlier work ⁵.

The cost of this approach is antibiotic resistance. Every dose given to a healthy child puts pressure on the bacteria that child carries. Bacteria in the gut and nose that survive are more likely to be resistant to macrolides. That is why every major trial now measures resistance alongside deaths [1,6].

For a health worker, the key facts are these. It is one oral dose. It goes to all children in the target age group in the community, not only sick ones. It is repeated every six months. And it only makes sense where child deaths are high.

What the evidence shows

MORDOR, 2018: the mortality result (Malawi, Niger, Tanzania)

MORDOR, published in the New England Journal of Medicine in 2018, was the trial that started this field ². It randomised 1,533 communities in Malawi, Niger and Tanzania to four twice-yearly doses of azithromycin or placebo for children aged 1–59 months ². It counted 190,238 children at baseline and followed 323,302 person-years. Coverage averaged 90.4% ².

Mortality was 14.6 deaths per 1000 person-years with azithromycin and 16.5 with placebo. That is 13.5% lower overall (95% CI 6.7–19.8) ². The effect was largest in Niger, where the rate fell from 27.5 to 22.5. In Malawi it went from 9.6 to 9.1, and in Tanzania from 5.5 to 5.4 ². The authors warned that any policy would need to weigh the effect on resistance ².

MORDOR is a central trial within this 15-year review. Later trials address the questions it left open.

In 2020, WHO responded with a cautious guideline. It advised against distribution to all children under five. It said countries may consider azithromycin MDA for infants aged 1–11 months in sub-Saharan African settings with very high infant or under-5 mortality ³. The aim was to limit resistance by treating fewer children, but that approach had not yet been tested ¹.

MORDOR, 2020: resistance after four years

In a parallel randomised study of 30 Nigerien villages, four years of twice-yearly azithromycin selected gut resistance to more than macrolides. At 48 months, macrolide-resistance determinants were 7.5 times as abundant as with placebo (95% CI 3.8–23.1), and resistance determinants for non-macrolide antibiotics, including beta-lactams, also increased ²³. These are resistance-gene measurements, not counts of treatment failures. They nevertheless show why the shorter AVENIR follow-up cannot settle the consequences of prolonged national use. Fold changes from different assays and trial populations should not be compared as though they were the same measurement.

AVENIR (Niger): the main mortality result

AVENIR was designed to test that WHO restriction head to head. Communities in Niger were randomly assigned to one of three arms: azithromycin for all children aged 1–59 months, azithromycin for infants aged 1–11 months only (with placebo for older children), or placebo for all ¹. Each arm received four twice-yearly distributions over two years. Census workers who did not know the group assignments recorded deaths ¹.

In total, 1,273 communities were assigned to the all-child arm, 773 to the infant arm and 954 to placebo ¹. The trial recorded 419,440 person-years and 5,503 deaths ¹.

Deaths among children aged 1–59 months fell from 13.9 per 1000 person-years with placebo (95% CI 13.0–14.8) to 11.9 with azithromycin for all under-5s (11.3–12.6) ¹. That is the 14% reduction. Treating all under-5s was also more effective than treating infants only ¹. The trial was funded by the Gates Foundation and registered as NCT04224987 ¹.

This is the largest confirmation so far of the MORDOR effect. It also answers the question the 2020 guideline raised: restricting treatment to infants did not do as well.

AVENIR: underweight infants

A secondary AVENIR analysis looked at 97,572 children who were weighed at least once, in 2,880 communities in the Dosso and Tahoua regions ⁴. Underweight children died at higher rates than other children ⁴.

Giving azithromycin to all under-5s lowered infant deaths by 6.2 per 1000 person-years (95% CI −9.3 to −2.6) overall. In severely underweight infants, the reduction was 11.2 per 1000 person-years (−26.0 to −2.1) ⁴. Underweight status did not significantly change the effect ⁴. The authors conclude that their results do not support restricting eligibility, because treating all children saves the most lives ⁴.

CHAT (Burkina Faso): causes of death where malaria prevention exists

The Community Health with Azithromycin Trial (CHAT) was run in Nouna District by the Centre de Recherche en Santé de Nouna (CRSN) and UCSF. It randomised 341 communities to azithromycin or placebo for children aged 1–59 months, with six rounds over 36 months ⁵. Unlike the earlier Niger trial of cause-specific deaths, these communities also received seasonal malaria chemoprevention (SMC) ⁵.

Verbal autopsies covered 992 of 1,086 deaths (91%) ⁵. Malaria deaths were a third lower with azithromycin (incidence rate ratio 0.67; 95% CI 0.50–0.90; P = 0.008) ⁵. Deaths from diarrhoea and pneumonia were also lower, but these differences were not statistically significant ⁵. The result matters because it shows azithromycin still adds benefit where SMC is already in place.

AVENIR: resistance in the gut and nose

AVENIR measured resistance as a co-primary outcome, not an afterthought. In 150 communities (50 per arm), researchers analysed 4,382 rectal and 4,402 nasopharyngeal samples ⁶. The results were published in Nature Medicine in 2026.

The trial met its resistance endpoint for the gut but not for the nose ⁶. In the gut, macrolide resistance was 1.16-fold higher with azithromycin for all under-5s than with placebo (95% CI 1.06–1.28). Compared with the infant-only arm it was 1.13-fold higher (1.02–1.23). Infant-only azithromycin against placebo gave 1.04-fold (0.94–1.15), which is not significant ⁶. The more children treated, the more gut resistance rose.

In the nose, no difference was statistically significant. For all under-5s against placebo the estimate was 2.14-fold (95% CI 0.93–4.99; P = 0.10). For infant-only against placebo it was 2.08-fold (0.93–4.69), and for all under-5s against infant-only it was 1.03-fold (0.46–2.30) ⁶. We return to these wide intervals below.

The authors report no detectable selection for resistance to any non-macrolide antibiotic class, in the gut or the nasopharynx ⁶. We checked this in the full text. They conclude that close monitoring of resistance should be an essential part of any programme ⁶.

CHAT: resistance after 36 months

In Burkina Faso, researchers took rectal swabs from 15 randomly selected children per community in 48 communities after 36 months and six rounds. They analysed 483 samples from 41 communities ⁷.

The gut macrolide resistance fold-change was 1.21 (95% CI 0.96–1.52). The beta-lactam fold-change was 1.05 (0.79–1.40). Neither was significant ⁷. The trial is registered as NCT03676764 ⁷.

Malawi: pneumococcal resistance 3.5 years after MORDOR

The Malawi-Liverpool-Wellcome Programme (MLW) went back to Mangochi, a MORDOR site, 3.5 years after distribution ended. In April and May 2021, it recruited 905 children aged 1–9 years and sequenced pneumococcal isolates from nose swabs ⁸.

In former azithromycin clusters, macrolide resistance was 21.7% at baseline, 31.9% six months after MDA and 32.1% at 3.5 years. In placebo clusters it went from 21.0% to 25.0% to 30.9% ⁸. The differences between arms were not significant at any time point ⁸. At a distant Malawian site with no MDA, resistance stayed at about 16.5–16.9% ⁸.

Among children born into azithromycin clusters after MDA ended, macrolide resistance was 36.0% ⁸. The authors conclude that MDA is associated with resistance that persists and may spread to untreated populations, and call for careful monitoring ⁸.

Niger: no spillover to untreated villages

A common worry is that trials underestimate resistance because it spreads to nearby untreated villages. In Niger, researchers sampled 300 children in 30 untreated villages surrounded by 594 treated ones, after two years of MDA ⁹. They found no sign that macrolide resistance spread. Resistance-gene levels in placebo villages stayed at baseline whatever the treatment intensity around them (Spearman ρ = −0.05, P = 0.83) ⁹. No spillover was found for other antibiotic classes either ⁹.

The Malawi and Niger results are not identical. Malawi saw resistance rise in both arms over time. Niger saw no spillover at 24 months. Both findings are useful, and both argue for keeping watch.

LAKANA (Mali) and MIRAMA (Burkina Faso): infant-only programmes

These two trials tested the infant-only strategy that the 2020 guideline allows. Both are described in detail in the next section, because neither found a benefit.

Results table

LimitsWhat didn't work (or the limits)

Treating only infants did not save lives. Two large trials tested the strategy the 2020 WHO guideline allows.

In Mali, LAKANA, run by CVD-Mali with Tampere University, enrolled 149,090 infants in 1,151 villages from December 2020 to December 2022 ¹¹. Villages were assigned in a 3:4:2 ratio to placebo, twice-yearly azithromycin or quarterly azithromycin. The trial recorded 82,600 person-years and 968 deaths ¹¹.

Mortality was 11.9 per 1000 person-years with placebo, 11.8 with twice-yearly azithromycin (IRR 1.00; 95% CI 0.83–1.19), and 11.3 with quarterly dosing (IRR 0.93; 0.75–1.15) ¹¹. Deaths among untreated children aged 12–59 months were also similar across groups ¹¹. Adverse events were rare and similar in all groups ¹¹. Commentators note that mortality in the placebo arm was well below what the trial planned for ¹².

In Burkina Faso, MIRAMA, led by Helen Keller Intl, delivered azithromycin to infants through the existing Child Health Days platform, alongside vitamin A. Primary health-centre catchment areas (CSPS) were randomised 2:1 ¹³. From September 2021 to January 2024, 201,709 children in 303 CSPS received azithromycin and 100,959 in 158 CSPS received placebo, with 85% coverage ¹³.

Mortality was 2.6 per 1000 person-years with azithromycin and 2.5 with placebo (IRR 1.04; 0.75–1.46) ¹³. There were 16 non-serious adverse events and no serious ones ¹³. The authors say the trial shows MDA can be built into existing child health programmes, and recommend that future policy consider treating all children up to five ¹³. The mortality rate in this trial was very low, which limits what it can show.

Nasopharyngeal resistance is not settled. In AVENIR, no difference in nasopharyngeal macrolide resistance was statistically significant. But the point estimates were high. For all under-5s against placebo the estimate was 2.14-fold (95% CI 0.93–4.99; P = 0.10). For infant-only against placebo it was 2.08-fold (0.93–4.69) ⁶. The intervals are wide and do not rule out a doubling.

Background resistance is rising anyway. In Malawi, pneumococcal macrolide resistance went from about 21% to about 31–32% in both arms over 3.5 years ⁸. The authors also found that resistance to at least three antibiotic classes was significantly higher over time in both arms, and they flag transmissible resistance genes as a public health concern ⁸. The CHAT gut result also cannot rule out a rise ⁷.

Other benefits did not appear. Some hoped azithromycin would also help children grow. AVENIR found no effect on weight-for-age in 51,687 infants (mean difference 0.00; 95% CI −0.05 to 0.04) ¹⁴. In Mali, a LAKANA growth substudy of 1,789 children found no difference in weight, length, underweight, stunting or wasting between groups ²⁰. Azithromycin is not a nutrition intervention. Children who are underweight still need food and nutrition care ¹⁴.

The effect depends on baseline mortality. Across MORDOR, the benefit was large in Niger and small in Malawi and Tanzania, where death rates were lower ². The two infant-only trials that failed also had low or lower-than-planned mortality [12,13]. The drug should not be expected to work the same way everywhere.

Cost and scale

The price is known. A micro-costing study with Niger's Ministry of Health used observations from 80 communities in Dosso region. It estimated that a national door-to-door programme for children aged 1–59 months in non-urban areas, at 90% coverage, would cost $12.5 million a year (95% UI $12.2M–13.0M) ¹⁰. That is $1.59 per dose delivered (UI $1.40–2.30) ¹⁰.

Cost per dose varied by region, from $1.17 (UI $1.03–1.69) to $3.61 (UI $3.20–5.16). Sparsely populated regions cost more ¹⁰. Training was 16.4% of total costs. The authors suggest combining it with training for other health programmes to save money ¹⁰.

Who delivers. In the trials, the drug was given door to door by community distributors ¹⁰, or through existing campaign platforms. MIRAMA showed that Burkina Faso's Child Health Days can carry azithromycin with vitamin A and reach 85% coverage ¹³.

Programmes already exist. Niger is expanding azithromycin MDA nationwide ¹⁶. Côte d'Ivoire runs twice-yearly campaigns in eight high-mortality regions ¹⁸. There, researchers monitored resistance in infants in 2022 and 2023. Erythromycin-resistant pneumococci were 32.8% in 2022 and 26.2% in 2023, not a significant change (p = 0.0537) ¹⁸.

Guideline status. The 2020 WHO guideline is still the official position. It supports infant-only MDA in very high-mortality settings and advises against treating all under-5s ³. In April 2026 WHO announced a Guideline Development Group (GDG) to update this guidance ¹⁵. It cited new evidence on effectiveness, safety, feasibility and antimicrobial resistance ¹⁵. The group will look at age-specific effects, baseline mortality, malaria endemicity, health-system capacity, and short- and long-term resistance ¹⁵. Public comment closed on 20 May 2026, and the group met in Geneva from 30 June to 2 July 2026 ¹⁵. As of this writing, no updated guideline has been published ¹⁵.

What to watch

AVENIR II in Niger (NCT06358872). Niger has built a randomised trial into its national rollout ¹⁶. All eligible health-centre catchment areas (Centres de Santé Intégrés, CSIs) will join twice-yearly MDA for children aged 1–59 months. A subset, expected to be 5–10%, will start two years later. Another subset will later pause for two years ¹⁶. Mortality and resistance are measured at baseline and every two years, and delivery and safety are tracked continuously. No area goes without MDA for more than two years ¹⁶.

The trial is sponsored by UCSF with Niger's Ministère de la Santé Publique, the Centre de Recherche Médicale et Sanitaire, the Centre de recherche et interventions en santé publique (CRISP), the national eye health programme and the Gates Foundation ²². It is recruiting, started on 29 April 2024, and has an estimated completion date of 29 April 2028 ²². It is designed as a platform, so other community health interventions can be added ¹⁶.

The WHO guideline. The key question for the GDG is whether to move from infant-only to all under-5s in the highest-mortality settings. The AVENIR, LAKANA and MIRAMA results all point that way [1,11,13]. The group must also decide how much resistance monitoring to require ¹⁵.

Nasopharyngeal resistance. The AVENIR estimate of about a two-fold rise in the nose was not significant, but it was not ruled out either ⁶. Longer follow-up and national surveillance, such as in Côte d'Ivoire and AVENIR II, should narrow this [16,18].

Humanitarian settings. A 2026 review in PLOS Global Public Health argues that azithromycin MDA could help reduce deaths in displaced populations, where mortality often exceeds WHO thresholds ¹⁹. It suggests a single round or a short course during an acute crisis may limit long-term resistance ¹⁹. This is a proposal, not a tested policy. The authors call for research on protocols and resistance monitoring ¹⁹.

What it means

For ministries of health. The evidence now points one way on who to treat. Twice-yearly azithromycin for all children aged 1–59 months saved lives in Niger. Infant-only programmes did not in Mali or Burkina Faso [1,11,13]. That is the reverse of the 2020 WHO position ³. Ministries in high-mortality Sahel countries should follow the WHO update closely and plan for an all-under-5 option.

Target matters. The benefit is largest where child deaths are highest ². Programmes should focus on areas where mortality is known to be high, not roll out everywhere.

Build resistance monitoring in from the start. Niger's AVENIR II and Côte d'Ivoire's surveillance show this can be done within a programme [16,18]. Gut resistance does rise with treatment; nose resistance may too ⁶. Four-year MORDOR follow-up found selection for resistance to other antibiotic classes ²³, even though AVENIR did not ⁶. The separate Niger spillover study found no spread to untreated villages at two years ⁹; that does not establish the long-term absence of spread.

For funders. The Niger estimates are about $1.59 a delivered dose and a 14% mortality reduction in a high-mortality setting ¹¹⁰. Resistance findings vary by trial and duration; prolonged treatment has selected resistance to other drug classes ²³. Costs are higher in sparse regions ¹⁰. Funding for surveillance should be part of every grant.

For health workers. Azithromycin MDA does not replace anything. It does not treat malnutrition [14,20]. It works alongside SMC, not instead of it ⁵. Sick children still need to be assessed and treated. The single twice-yearly dose is an extra layer of protection for children in the places where they are most likely to die.