Mass distribution of azithromycin to children in rural Niger reduced all cause mortality. Can multiplex antibody testing shed light on the mechanism?

A substudy from a large clinical trial provides additional evidence for mechanism and highlights strengths and limitations of antibody responses as outcomes in clinical trials.

Like Comment
Read the paper

In 2018, members of our team published a surprising finding from a large simple trial that enrolled hundreds of thousands of children across rural Niger, Tanzania, and Malawi: biannual mass distribution of azithromycin to children 1-59 months old reduced all cause mortality by 13.5%. In Niger, where overall mortality was highest, mortality fell by 18% in the azithromycin group (The trial was called MORDOR: Macrolides Oraux pour Réduire les Décès avec un Oeil sur la Résistance.)1

Azithromycin presumably reduced mortality by treating infections and also, potentially, by reducing overall pathogen transmission in the community. A key question that arose from the trial was: what is the mechanism?  Azithromycin is a broad spectrum antibiotic that has antimicrobial activity against gram-positive and atypical bacteria. It is a highly effective treatment for bacterial enteric infections and even has modest antimalarial properties. This means there are many possible ways it could have reduced mortality. Understanding the mechanism is important in this context to help identify complementary interventions that could build upon its effects, or to identify substitute interventions that target the same pathogens but might be less likely to select for antimicrobial resistance. Alongside a mortality benefit, the MORDOR trial documented increased prevalence of genes associated with macrolide resistance in treated communities (azithromycin is part of a broader class of antibiotics called macrolides).2

Early efforts to study the mechanism of mass azithromycin distribution on child mortality have identified a broad set of potential pathways. In MORDOR Niger, an analysis of cause-specific mortality assessed from caregiver interviews using the verbal autopsy method showed reductions in several infectious causes of death, including: malaria, dysentery, meningitis, pneumonia, and diarrhea.3  Studies focused on specific pathogen infection demonstrated large reductions in Shigella 4 and more modest reductions in Campylobacter carriage 5 and malaria infection.6 All three pathogens are common in this setting and thought to be key contributors to mortality among young children. Together, these results suggest a diverse set of mechanisms and that the mortality benefit from azithromycin was unlikely to be attributable to a single pathogen.

Here, we took advantage of recent advances in multiplex antibody testing to study the intervention’s effect on population-level transmission of specific pathogens in a substudy of 30 villages from the Niger trial. We focused on immunoglobulin G (IgG) antibody responses because IgG can be measured in high-throughput assays and because our colleagues at the US Centers of Disease Control and Prevention (CDC) had developed a broad IgG panel with antigens for several species of malaria, bacteria, and protozoan pathogens of interest. The antibody assay works by exposing blood to pathogen-specific antigens — antigens are proteins that are recognized by our immune system — and measuring the quantity of antibodies that bind to each antigen. Antibodies can be useful outcomes in trials that have infrequent measurements because they provide a sensitive measure of whether a child has previously been infected, even if they have cleared the infection. In the intensive monitoring substudy of the MORDOR trial, the field team collected specimens from study children just once per year so we thought antibody measurement could complement molecular measures that detect current infections.

Since this was a masked, placebo controlled, randomized trial (fully masked field team, participants, lab team, data analysts) with consistently high levels of treatment coverage (71% to 92% of all children 1-59 months in the communities were treated), we could be confident that any difference in IgG responses between children in communities that received azithromycin compared with placebo were due to the azithromycin and not to some other factor.

After analyzing 5,642 blood specimens collected from 3,814 children on the multiplex assay, we used seroepidemiologic methods to summarize population-level patterns in antibody responses. There is an enormous amount of variation between individuals in immune response, but by averaging responses large numbers of children in each treatment group we could identify consistent patterns that reflect population-level transmission. One approach was to use age-structured seroprevalence to estimate each pathogen’s force of infection — the rate of new infections in a period of time among those in the population who are still susceptible.

The antibody measurements suggested that children in rural Niger live in an environment of intense pathogen exposure and experience a high burden of infectious disease. For bacterial pathogens like Campylobacter and enterotoxigenic E. coli (ETEC), which are major causes of diarrhea and dysentery among children, nearly 100% of children in the study were seropositive by age 18 months. By age 36 months, nearly 90% of children were seropositive to falciparum malaria — the most deadly form of malaria. Since IgG provides a durable measure of past infection, this means that nearly all children in the study had been infected at least once by these pathogens. 

Against this backdrop of very high transmission, we estimated that azithromycin reduced the force of infection of Campylobacter by 29% (a reduction from 1.8 to 1.3 incident seroconversions per year, hazard ratio = 0.71, 95% CI: 0.56 to 0.89; P=0.004) but we found no other significant differences between groups for other pathogens we studied after correcting for multiple testing. For example, although IgG responses suggested a transient reduction in falciparum malaria seropositivity between ages 12-36 months, the difference between groups was not significantly significant (Figure 1).


Figure 1. Seroprevalence by age and treatment group for Campylobacter jejuni and Plasmodium falciparum among children 1-59 months in the MORDOR Niger trial. Lines represent groups means, estimated using semiparametric splines. The large drop in seroprevalence for P. falciparum is from maternal IgG antibodies, which are acquired in utero and wanes after birth. The shaded region in each panel shows the age range in which the trial compared each pathogen’s serological force of infection, which differed by pathogen based on the rate of waning maternal IgG and the age range when seroprevalence increased to a plateau. Differences between groups were significant for C. jejuni (29% reduction in the force of infection), but not for P. falciparum (12% reduction in the force of infection).


One concern that we had after reflecting on the results was that IgG responses might be too sensitive in the context of extremely high pathogen transmission. Another concern we had was that the trial only collected blood once per year: the monitoring protocol was optimized to measure antimicrobial resistance, which meant that specimens were collected 6 months after the most recent round of treatment. Azithromycin treatment most likely has short-term effects on infection and child survival, on a timescale of days to weeks. Since antibody levels were tested 6 months after the most recent azithromycin distribution it was likely that children acquired new infections in the interim period, potentially obscuring part of the benefit when measured through IgG response. Furthermore, treatment of an infection might not reduce a child’s IgG levels enough for them to fall below seropositivity thresholds. Based on these results, we speculate that IgG-based endpoints might be most valuable for studying interventions in lower transmission settings and for studying interventions that primarily prevent new infections rather than treating existing infections. We initially thought IgG outcomes would be an advantage in a trial like this with infrequent measurements, but it is possible that in very high transmission settings that is not the case.

During peer review, a referee suggested that we further interrogate the data to see if we could provide more evidence about the utility of IgG in the context of different levels of transmission. Since the trial had collected independent, baseline measures of malaria parasitemia using thick smear microscopy, we could stratify communities by baseline malaria parasitemia to see if differences in IgG response were more pronounced in lower transmission communities. Indeed, we found modest, exploratory results to support this possibility (Figure 2).


Figure 2. Seroprevalence by age and treatment group for Plasmodium falciparum among children 1-59 months in the MORDOR Niger trial, grouped by baseline, community-level malaria parasitemia prevalence. There were 17 communities with baseline parasitemia ≤5% (n=2,174 children 12-59 months) and 13 communities with baseline parasitemia >5% (n=1,686 children 12-59 months). Lines represent groups means, estimated using semiparametric splines. The large drop in seroprevalence from 0 to 12 months is from waning maternal IgG, acquired in utero. The shaded region in each panel shows the age range where the trial compared each pathogen’s serological force of infection. Among communities with baseline parasitemia ≤5%, there was a 32% reduction in the force of infection (hazard ratio = 0.68, 95% CI 0.43 to 1.07), but no difference in communities with >5% baseline parasitemia.


Another value of serology in this context is that age-seroprevalence curves provide some insight into a pathogen’s overall, community-level transmission. Pathogens with very high force of infection will have age-seroprevalence curves that rise steeply and plateau at very high levels. Conversely, pathogens with low force of infection will result in age-seroprevalence curves that rise more slowly and plateau at lower levels (as demonstrated to some degree in Figure 2 among communities with low baseline parasitemia). Overall, the seroepidemiologic analyses in this trial suggest that overall transmission of enteric pathogens and malaria was intense and was not dramatically reduced by biannual distribution of azithromycin. However, it remains possible that the IgG measurements were too distant from the point of treatment, and that transient reductions in community level transmission — not detected through serology in the highest transmission settings — were still important for the mortality benefit.7 


Taken together, the evidence suggests that mass distribution of azithromycin likely reduces child mortality through broad effects on diverse pathogens, and that there is no single, dominant mechanism. The antibody data provide additional evidence that the intervention reduced Campylobacter infection, and our exploratory analyses of malaria IgG responses suggest that antibody-based measures of infection might be most useful for lower transmission settings, as evidenced by the larger downward shift in the antibody curve among children in communities with lower levels of malaria infection (Figure 2). Since accumulating evidence suggests several pathways through which azithromycin provides a mortality benefit in high mortality settings, further studies of mechanism may not be essential to inform public health programs: a population’s particular pathogen burden may not predict whether children will benefit from mass distribution of azithromycin. Instead, and consistent with current WHO guidelines,8 the decision of whether to administer mass distribution of azithromycin should likely focus on populations with highest child mortality rates, weighed against the potential for selection of antimicrobial resistance.

For more details, please see our paper with accompanying data and computational notebooks archived on Dryad (https://doi.org/10.7272/Q6VX0DSD) and the Open Science Framework (https://osf.io/954bt).

Photo creditDominique Catton

References

  1. Keenan, J. D., Bailey, R. L., West, S. K., Arzika, A. M., Hart, J., Weaver, J., Kalua, K., Mrango, Z., Ray, K. J., Cook, C., Lebas, E., O’Brien, K. S., Emerson, P. M., Porco, T. C., Lietman, T. M. & MORDOR Study Group. Azithromycin to Reduce Childhood Mortality in Sub-Saharan Africa. N. Engl. J. Med. 378, 1583–1592 (2018). PMID: 29694816
  2. Doan, T., Arzika, A. M., Hinterwirth, A., Maliki, R., Zhong, L., Cummings, S., Sarkar, S., Chen, C., Porco, T. C., Keenan, J. D., Lietman, T. M. & MORDOR Study Group. Macrolide Resistance in MORDOR I - A Cluster-Randomized Trial in Niger. N. Engl. J. Med. 380, 2271–2273 (2019). PMID: 31167060
  3. Keenan, J. D., Arzika, A. M., Maliki, R., Elh Adamou, S., Ibrahim, F., Kiemago, M., Galo, N. F., Lebas, E., Cook, C., Vanderschelden, B., Bailey, R. L., West, S. K., Porco, T. C., Lietman, T. M. & MORDOR-Niger Study Group. Cause-specific mortality of children younger than 5 years in communities receiving biannual mass azithromycin treatment in Niger: verbal autopsy results from a cluster-randomised controlled trial. Lancet Glob Health 8, e288–e295 (2020). PMID: 31981558
  4. Platts-Mills, J. A., Ayoub, E., Zhang, J., McQuade, E. T. R., Arzika, A. M., Maliki, R., Abdou, A., Keenan, J. D., Lietman, T. M., Liu, J. & Houpt, E. R. Impact of biannual mass azithromycin treatment on enteropathogen carriage in children younger than 5 years in Niger. Clin. Infect. Dis. (2022). doi:10.1093/cid/ciab1046 PMID: 35020888
  5. Doan, T., Hinterwirth, A., Worden, L., Arzika, A. M., Maliki, R., Abdou, A., Kane, S., Zhong, L., Cummings, S. L., Sakar, S., Chen, C., Cook, C., Lebas, E., Chow, E. D., Nachamkin, I., Porco, T. C., Keenan, J. D. & Lietman, T. M. Gut microbiome alteration in MORDOR I: a community-randomized trial of mass azithromycin distribution. Nat. Med. 25, 1370–1376 (2019). PMID: 31406349
  6. Arzika, A. M., Maliki, R., Boubacar, N., Kane, S., Cotter, S. Y., Lebas, E., Cook, C., Bailey, R. L., West, S. K., Rosenthal, P. J., Porco, T. C., Lietman, T. M., Keenan, J. D. & MORDOR Study Group. Biannual mass azithromycin distributions and malaria parasitemia in pre-school children in Niger: A cluster-randomized, placebo-controlled trial. PLoS Med. 16, e1002835 (2019). PMID: 31237871
  7. O’Brien, K. S. & Oldenburg, C. E. MDA and trial designs to evaluate the impact of azithromycin on child mortality. Lancet Glob Health 10, e183 (2022). PMID: 35063109
  8. WHO Guideline on Mass Drug Administration of Azithromycin to Children under Five Years of Age to Promote Child Survival. (World Health Organization, 2020). PMID: 32924384

Benjamin Arnold

Associate Professor, UCSF