Effect of Adding Azithromycin to Seasonal Malaria Chemoprevention
Daniel Chandramohan, Ph.D., Alassane Dicko, M.D., Issaka Zongo, Ph.D., Issaka Sagara, M.D., Matthew Cairns, Ph.D., Irene Kuepfer, Ph.D., Modibo Diarra, M.D., Amadou Barry, M.D., Amadou Tapily, M.D., Frederic Nikiema, M.D., Serge Yerbanga, Ph.D., Samba Coumare, M.D., et al.
Article
Figures/Media
Metrics
19 References
1 Citing Article
Comment
Comments open through June 12, 2019
Abstract
BACKGROUND
Mass administration of azithromycin for trachoma control led to a sustained reduction in all-cause mortality among Ethiopian children. Whether the addition of azithromycin to the monthly sulfadoxine–pyrimethamine plus amodiaquine used for seasonal malaria chemoprevention could reduce mortality and morbidity among African children was unclear.
METHODS
We randomly assigned children 3 to 59 months of age, according to household, to receive either azithromycin or placebo, together with sulfadoxine–pyrimethamine plus amodiaquine, during the annual malaria-transmission season in Burkina Faso and Mali. The drug combinations were administered in four 3-day cycles, at monthly intervals, for three successive seasons. The primary end point was death or hospital admission for at least 24 hours that was not due to trauma or elective surgery. Data were recorded by means of active and passive surveillance.
RESULTS
In July 2014, a total of 19,578 children were randomly assigned to receive seasonal malaria chemoprevention plus either azithromycin (9735 children) or placebo (9843 children); each year, children who reached 5 years of age exited the trial and new children were enrolled. In the intention-to-treat analysis, the overall number of deaths and hospital admissions during three malaria-transmission seasons was 250 in the azithromycin group and 238 in the placebo group (events per 1000 child-years at risk, 24.8 vs. 23.5; incidence rate ratio, 1.1; 95% confidence interval [CI], 0.88 to 1.3). Results were similar in the per-protocol analysis. The following events occurred less frequently with azithromycin than with placebo: gastrointestinal infections (1647 vs. 1985 episodes; incidence rate ratio, 0.85; 95% CI, 0.79 to 0.91), upper respiratory tract infections (4893 vs. 5763 episodes; incidence rate ratio, 0.85; 95% CI, 0.81 to 0.90), and nonmalarial febrile illnesses (1122 vs. 1424 episodes; incidence rate ratio, 0.79; 95% CI, 0.73 to 0.87). The prevalence of malaria parasitemia and incidence of adverse events were similar in the two groups.
CONCLUSIONS
Among children in Burkina Faso and Mali, the addition of azithromycin to the antimalarial agents used for seasonal malaria chemoprevention did not result in a lower incidence of death or hospital admission that was not due to trauma or surgery than antimalarial agents plus placebo, although a lower disease burden was noted with azithromycin than with placebo. (Funded by the Joint Global Health Trials scheme; ClinicalTrials.gov number, NCT02211729.)
Malaria transmission is concentrated during a few months of the year in much of the Sahel and sub-Sahel regions of Africa. In these areas, seasonal malaria chemoprevention — the administration of sulfadoxine–pyrimethamine plus amodiaquine to children at monthly intervals three or four times during the malaria-transmission season — has been a highly effective approach to malaria control.1 Seasonal malaria chemoprevention is now being implemented widely across these regions.2 The frequent contact between children and health care workers that is needed for seasonal malaria chemoprevention provides an opportunity for the delivery of other health interventions.
Mass administration of azithromycin has been a highly effective approach to trachoma control.3 Reductions in the incidences of skin, gastrointestinal, and respiratory infections have been recorded after mass administration of azithromycin.4-8 Nevertheless, the finding in Ethiopia of a 49% reduction in all-cause mortality among children 1 to 9 years of age during the year after mass administration of a single dose of azithromycin — a reduction that was sustained during a 26-month follow-up period — was surprising.9,10 Consequently, we conducted a randomized, double-blind, placebo-controlled trial to determine whether the addition of azithromycin to the sulfadoxine–pyrimethamine plus amodiaquine given for seasonal malaria chemoprevention could have a similar effect on overall child mortality and morbidity.
Methods
TRIAL OVERSIGHT
The trial was approved by the ethics committees of the London School of Hygiene and Tropical Medicine, London; the Malaria Research and Training Center, University of Bamako, Bamako, Mali; the Ministry of Health, Ouagadougou, Burkina Faso; and the national regulatory authorities of Burkina Faso and Mali. A data and safety monitoring board reviewed serious adverse events, monitored the overall progress of the trial, approved the statistical analysis plan, and archived the locked database before the data were unmasked. A steering committee reviewed the protocol (available with the full text of this article at NEJM.org) and provided overall advice. The authors vouch for the accuracy and completeness of the data and the fidelity of the trial to the protocol.
SITES AND POPULATION
The trial was conducted in the Houndé district of Burkina Faso and in the Bougouni district of Mali (Fig. S1 in the Supplementary Appendix, available at NEJM.org). Information about these communities and the children who live in them is provided in the protocol and the Supplementary Appendix.
ENROLLMENT AND RANDOMIZATION
A household census was conducted in June 2014, and children of either sex who were 3 to 59 months of age on August 1, 2014, were eligible for enrollment in the trial. After written informed consent was obtained from the child’s parent or guardian, the child received a long-lasting insecticide-treated bed net. Children were excluded if they had a chronic disease or a known allergy to sulfadoxine–pyrimethamine, amodiaquine, or azithromycin or if they were taking cotrimoxazole. The household census was repeated in May 2015 and in May 2016 to recruit additional eligible children and to detect any deaths that had been missed through the surveillance system. Each year, children who were still younger than 60 months of age on August 1 remained in follow-up for the subsequent trial year, and children who had reached 5 years of age on or before July 31 exited the trial on that date. Enrollment of children in the trial started on August 25, 2014, in Mali and on August 28, 2014, in Burkina Faso.
Randomization was performed according to household to avoid the potential effect of within-household transmission of infection; all eligible children who shared a kitchen were assigned to the same trial group. To mask the trial-group assignments for the trial team and caregivers, a placebo for azithromycin of identical appearance was used.
INTERVENTIONS
Children who were enrolled in the trial received the assigned preventive regimen during the annual peak malaria-transmission season (August to November). The drug combinations were administered in four 3-day cycles, at monthly intervals, for three successive seasons. Infants 3 to 11 months of age received a combined 250 mg of sulfadoxine and 12.5 mg of pyrimethamine plus 75 mg of amodiaquine on day 1 and received 75 mg of amodiaquine on days 2 and 3 (Guilin Pharmaceutical, Shanghai, China). In addition, they were randomly assigned to receive either 100 mg of azithromycin or matching placebo on days 1, 2, and 3 (Cipla, Mumbai, India). Children 1 to 4 years of age received double these doses. All doses were based on age and administered by trial staff. All trial drugs were purchased from the manufacturers with the use of a grant provided by the U.K. Medical Research Council and were provided to the children free of cost.
Each year, the drug combinations were prepacked in resealable plastic bags by pharmacists who were not part of the trial team. Each child was assigned one large bag that contained four smaller bags, each of which contained the sulfadoxine–pyrimethamine, amodiaquine, and either azithromycin or placebo for one of the four cycles. The child received a photo identification card that had a quick response code (known as a QR code) that encoded the child’s name, the mother’s name, the child’s date of birth, the census number, and a randomization number with a check digit. A label on the large bag also had a QR code that encoded the same variables. On the day of administration of the trial drugs, the QR codes on the identification card and on the large bag were scanned with a tablet computer to link the child to the correct bag. The trial drugs were kept at the trial office and were administered under direct observation by trial staff.
When the child was seen for administration of the trial drugs, if a diagnosis of malaria was confirmed with the use of a rapid diagnostic test, the child was not given the assigned regimen and instead received a dose of artemether–lumefantrine. Children with other illnesses were referred to a local health center for investigation and treatment.
END POINTS
The primary end point was death or hospital admission for at least 24 hours that was not due to trauma or elective surgery during the intervention period. The intervention period was defined as the period from the administration of the first dose of the first cycle of the trial drugs until 30 days after the administration of the first dose of the last cycle. For children who did not receive the first dose of the first or last cycle, the date that the dose was scheduled to be administered was used.
The prespecified secondary end points were the individual components of the primary end point; death or hospital admission for at least 24 hours during the entire trial period; parasitologically confirmed malaria, which was defined as a febrile illness (a history of fever within 24 hours or a measured temperature of ≥37.5°C) and either a positive rapid diagnostic test or a positive blood smear; radiographically confirmed pneumonia; clinically diagnosed pneumonia or lower respiratory tract infection; gastrointestinal infection; nonmalarial fever, which was defined as a febrile illness that was not due to malaria, lower or upper respiratory tract infection, or gastrointestinal infection; and anemia (hemoglobin level, <10 g per deciliter) or severe anemia (hemoglobin level, <7 g per deciliter) at the end of the malaria-transmission season. An exploratory analysis was performed to investigate the incidence of skin diseases.
SURVEILLANCE
Deaths and hospital admissions were recorded throughout the trial period, but only events that occurred during the intervention period contributed to the primary end point. Data regarding vital status were updated during an annual census and during an exit census that was conducted in March 2017. Deaths that occurred outside a health facility were assessed with the use of the World Health Organization verbal autopsy questionnaire.11 The trial-group assignments were masked for all assessments. Data regarding adverse events that occurred during the week after administration of the trial drugs were solicited from 800 children (a random selection of 200 children from each trial group in each country) on day 7 after each cycle in the first year of the trial. Details regarding surveillance are provided in the Supplementary Appendix.
In addition, 200 children (a random selection of 50 children from each trial group in each country) were visited each week during the malaria-transmission season for active detection of malaria infection. At the end of each malaria-transmission season (≥30 days after the administration of the first dose of the last cycle of the trial drugs), 4000 children (a random selection of approximately 1000 children from each trial group in each country) were included in a cross-sectional survey to assess the prevalence of malaria parasitemia. In addition, at the end of each malaria-transmission season, blood slides were obtained from 500 primary school children 5 to 12 years of age who lived in each trial area to provide data on the prevalence of malaria parasitemia among children who did not receive seasonal malaria chemoprevention.
STATISTICAL ANALYSIS
On the basis of data from previous seasonal malaria chemoprevention trials performed in Burkina Faso12 and Mali,13 we assumed that the incidence of death or hospital admission that was not due to trauma or elective surgery during the malaria-transmission season would be approximately 15 per 1000 children who received seasonal malaria chemoprevention plus placebo and the rate of loss to follow-up would be 10% per year. On the basis of these assumptions, we calculated that the enrollment of 19,200 children (9600 per country) for three malaria-transmission seasons would give the trial 90% power to detect a 25% lower incidence of the primary end point with azithromycin than with placebo.
The primary analysis was an intention-to-treat analysis of deaths and hospital admissions that occurred during the 4-month intervention period each year. The intention-to-treat population included all the children who had been screened and enrolled in the trial. A per-protocol analysis of the primary end point was also performed. Children who were seen on the first day of administration of the trial drugs for all four cycles of a particular year were included in the per-protocol population for that year. All analyses of secondary end points were performed on an intention-to-treat basis; these analyses were not adjusted for multiple comparisons, and thus P values are not reported for secondary end points.
For each child, person-time at risk was calculated as the time from the date of enrollment until 30 days after the date that the first dose of the last cycle was scheduled to be administered. If applicable, the following end dates were used instead: if the child was lost to follow-up, the date that the child was last seen; if the child emigrated, the date of permanent emigration; if the child died, the date of death; or if the child reached 5 years of age, the last day of the trial year in which the child reached 5 years of age.
The incidence rate ratio of the primary end point was estimated with the use of Poisson regression models, with a gamma-distributed random effect to account for clustering of episodes within households. Regression models were adjusted for trial site and stratified according to follow-up time with the use of Lexis expansion. As prespecified in the statistical analysis plan, effect modification according to trial site and year of age was assessed with the use of the likelihood ratio test, without adjustment for multiple comparisons, since only these two subgroup analyses were performed. Prevalence rate ratios were estimated with the use of Poisson regression models, with a robust standard error to account for randomization according to household.12
Results
CHILDREN AND COVERAGE WITH SEASONAL MALARIA CHEMOPREVENTION
Figure 1.
Screening, Randomization, and Follow-up.
In July 2014, a total of 19,578 children from 9618 households were randomly assigned to receive seasonal malaria chemoprevention plus either azithromycin (9735 children) or placebo (9843 children) (Figure 1). Each year, additional children in the specified age range were enrolled (6287 in the second year and 5748 in the third year), and children who were 5 years of age on August 1 exited the trial. Because of establishment of new households in the trial areas and migration of children into households that were originally included in the trial, the overall number of children in the trial increased each year. At the last follow-up visit, there were 10,885 children in the group that received seasonal malaria chemoprevention plus azithromycin and 10,852 in the group that received seasonal malaria chemoprevention plus placebo.
The two trial groups were well matched with regard to baseline characteristics (Table S1 in the Supplementary Appendix). Coverage with long-lasting insecticide-treated bed nets was high and similar in the two groups. The percentage of children who received at least three directly observed cycles of the assigned regimen was 92.8% in the first year, 86.8% in the second year, and 84.3% in the third year (Table S2 in the Supplementary Appendix).
EFFICACY
Table 1.
Incidence of Death or Hospital Admission That Was Not Due to Trauma or Elective Surgery during Three Successive Malaria-Transmission Seasons in the Intention-to-Treat Population.
In the intention-to-treat analysis, the overall number of deaths and hospital admissions that were not due to trauma or elective surgery was similar in the two trial groups: 250 in the azithromycin group and 238 in the placebo group (events per 1000 child-years at risk, 24.8 vs. 23.5; incidence rate ratio, 1.1; 95% confidence interval [CI], 0.88 to 1.3) (Table 1). In the per-protocol analysis, the overall number was 173 in the azithromycin group and 158 in the placebo group (events per 1000 child-years at risk, 19.8 vs. 18.2; incidence rate ratio, 1.1; 95% CI, 0.88 to 1.4) (Table S3 in the Supplementary Appendix).
Figure 2.
Incidence of Death or Hospital Admission That Was Not Due to Trauma or Elective Surgery during Three Successive Malaria-Transmission Seasons in Burkina Faso and Mali.
In the intention-to-treat population, there was evidence of an interaction between trial group and trial site (P=0.02 by the likelihood ratio test). The incidence of the primary end point was higher in the azithromycin group than in the placebo group in Burkina Faso (incidence rate ratio, 1.3; 95% CI, 1.0 to 1.7) but not in Mali (incidence rate ratio, 0.84; 95% CI, 0.64 to 1.1). The incidence of the primary end point according to year of age was similar in the two trial groups, both in the entire trial population (P=0.44 for interaction by the likelihood ratio test) and in each country individually (Figure 2, and Figs. S2 and S3 in the Supplementary Appendix). The causes of deaths and hospital admissions that occurred during the intervention period and during the entire trial period are shown in Tables S4 through S9 in the Supplementary Appendix. Malaria was the most prominent cause of death and hospital admission in each trial group.
Table 2.
Incidence of Clinical Events Treated at Health Centers or by Community Health Workers during Three Successive Malaria-Transmission Seasons in Burkina Faso and Mali in the Intention-to-Treat Population.
The incidence of clinic visits for the following events was lower with antimalarial agents plus azithromycin than with antimalarial agents plus placebo: gastrointestinal infections (incidence rate ratio, 0.85; 95% CI, 0.79 to 0.91), upper respiratory tract infections (incidence rate ratio, 0.85; 95% CI, 0.81 to 0.90), and nonmalarial fevers (incidence rate ratio, 0.79; 95% CI, 0.73 to 0.87) (Table 2). An exploratory analysis showed that, among children who had a clinic visit for nonmalarial fever, 269 children in the azithromycin group and 442 children in the placebo group had a skin condition (events per 1000 child-years at risk, 26.6 vs. 43.6; incidence rate ratio, 0.61; 95% CI, 0.53 to 0.73). Among children who had a clinic visit for nonfebrile illness, 428 children in the azithromycin group and 556 children in the placebo group had a skin condition (events per 1000 child-years at risk, 42.4 vs. 54.8; incidence rate ratio, 0.77; 95% CI, 0.67 to 0.89). The number of children with a skin condition that most likely had a bacterial cause was substantially lower in the azithromycin group than in the placebo group, especially among those with nonmalarial fever (66 vs. 145; events per 1000 child-years at risk, 6.54 vs. 14.3; incidence rate ratio, 0.46; 95% CI, 0.33 to 0.62).
Table 3.
Prevalence of Malaria Parasitemia and Anemia among Children in the Trial and Prevalence of Malaria Parasitemia among Primary School Children.
Among children who were randomly selected for weekly follow-up visits during the intervention period, the prevalence of malaria parasitemia ranged from 3 to 7% and was similar in the two groups. At the end of the malaria-transmission season, the prevalence of malaria parasitemia ranged from 4 to 10% and the prevalence of anemia ranged from 20 to 26%. For both variables, results were similar in the two trial groups (Table 3). Among primary school children who lived in the trial areas and did not receive seasonal malaria chemoprevention, the prevalence of malaria parasitemia at the end of the malaria-transmission season ranged from 50 to 65% (Table 3).
SAFETY
No severe adverse events that were judged by investigators to be related to the trial drugs were recorded. Diarrhea was the most frequent adverse event reported after administration of the trial drugs. The incidences of diarrhea and of other adverse events that occurred during the week after administration of the trial drugs were similar in the two trial groups (Table S10 in the Supplementary Appendix).
Discussion
In this trial, the incidence of death or hospital admission that was not due to trauma or elective surgery did not differ significantly between children who received seasonal malaria chemoprevention plus azithromycin and children who received seasonal malaria chemoprevention plus placebo when data from Burkina Faso and Mali were combined for the primary analysis. However, the incidence of death or hospital admission was higher with azithromycin than with placebo in Burkina Faso but not in Mali. No plausible mechanism to explain this difference was found, and given the borderline increased incidence in Burkina Faso, it may be a chance finding. The incidences of gastrointestinal infections, upper respiratory tract infections, and nonmalarial fevers were lower with azithromycin than with placebo (by 15%, 15%, and 21%, respectively), and in an exploratory analysis, the incidence of skin diseases, especially those that most likely had a bacterial cause, was also lower with azithromycin; these results are consistent with findings of previous studies in which azithromycin was used in trachoma control programs.4-8
The findings of our trial contrast with those of the MORDOR (Mortality Reduction after Oral Azithromycin) trial conducted in Malawi, Niger, and Tanzania, in which azithromycin was given to children younger than 5 years of age twice a year for 2 years and then was associated with a 13.5% (95% CI, 6.7 to 19.8) lower overall all-cause mortality than placebo, with the effect being most marked in Niger.14 Children who were assigned to the azithromycin group in our trial had greater exposure to azithromycin than those in the MORDOR trial (four courses each year for up to 3 years in our trial, as compared with two courses each year for 2 years in the MORDOR trial); such enhanced exposure might have been expected to have an effect on mortality that was at least equal to the effect seen in the MORDOR trial, but this was not the case.
There are several possible explanations for the different outcomes of these two trials. One possible explanation is that azithromycin, which has antimalarial activity,15 contributed to decreased mortality in the MORDOR trial partly through its effect on malaria, and this benefit was lost when an additional, effective antimalarial combination was given at the same time as azithromycin. However, the effect of azithromycin on malaria has been inconsistent when azithromycin has been given in mass drug administration programs.16-18 In addition, all the children in our trial received sulfadoxine–pyrimethamine, which has weak antimicrobial properties, and this may have reduced the potential benefit of adding another antimicrobial to the regimen. Finally, coverage with a pneumococcal conjugate vaccine was high among the children in our trial, and this may have reduced the potential benefit of azithromycin in lowering mortality from pneumonia.
In the MORDOR trial, the greatest effect of azithromycin on all-cause mortality was seen in the first year of life.14 This finding suggests that, despite the results of this trial, the addition of azithromycin to the sulfadoxine–pyrimethamine used for intermittent prevention of malaria in infants19 is an option worth investigating in countries with a high malaria burden in the first year of life.
The prevalence of malaria at the end of the malaria-transmission season was substantially lower among children who were enrolled in our trial than among primary school children who were living in the same areas and were not receiving seasonal malaria chemoprevention, a finding that indicates that seasonal malaria chemoprevention with sulfadoxine–pyrimethamine plus amodiaquine was having a major protective effect against malaria in these populations. Nevertheless, the proportion of deaths and hospital admissions that were attributable to malaria was still large in both trial groups, despite good access to treatment and high coverage with long-lasting insecticide-treated bed nets. Effective control of malaria in these and other, similar areas necessitates additional control measures.
A limitation of the trial is that randomization was performed according to household rather than village; randomization according to household reduced the potential for bias but precluded the potential for a herd effect that might have occurred had randomization according to village been performed. Only limited safety data were obtained because seasonal malaria chemoprevention with sulfadoxine–pyrimethamine plus amodiaquine and mass administration of azithromycin have now been given to millions of children with no major safety concerns.
In conclusion, among children in Burkina Faso and Mali, the addition of azithromycin to the antimalarial agents used for seasonal malaria chemoprevention did not result in a lower incidence of death or hospital admission that was not due to trauma or surgery than antimalarial agents plus placebo. We also noted that the incidences of clinic visits for gastrointestinal infections, upper respiratory tract infections, and nonmalarial febrile illnesses, without adjustment for multiple comparisons, were lower among children who received antimalarial agents plus azithromycin than among those who received antimalarial agents plus placebo.
Supported by a grant (MR/K007319/1) from the Joint Global Health Trials scheme, which includes the U.K. Medical Research Council, Department for International Development, National Institute for Health Research, and Wellcome Trust.
Disclosure forms provided by the authors are available with the full text of this article at NEJM.org.
No potential conflict of interest relevant to this article was reported.
This article was published on January 30, 2019, 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 the members of the trial steering committee (Feiko ter Kuile [chair], Kalifa Bojang, Kojo Koram, David Mabey, Morven Roberts, and Mahamadou Thera) and the members of the data and safety monitoring board (Blaise Genton [chair], Cheick Oumar Coulibaly, Umberto D’Alessandro, and Francesca Little) for their sustained assistance with the trial; Alice Greenwood for reviewing all the hospital records and verbal autopsies and validating causes of hospital admissions and deaths that were assigned by the trial team; the Ministry of Health staff in the Bougouni and Houndé districts for their assistance with running the trial; and all the caretakers and children for their participation.
Author Affiliations
From the London School of Hygiene and Tropical Medicine, London (D.C., M.C., I.K., P.M., B.G.); the Malaria Research and Training Center, University of Science, Techniques, and Technologies of Bamako, Bamako, Mali (A.D., I.S., M.D., A.B., A. Tapily, S.C., I.T., O.D.); and Institut de Recherche en Sciences de la Santé, Bobo-Dioulasso, Burkina Faso (I.Z., F.N., S.Y., A. Traore, H.T., J.-B.O.).
Address reprint requests to Dr. Chandramohan at the London School of Hygiene and Tropical Medicine, Keppel St., London WC1E 7HT, United Kingdom, or at daniel.chandramohan@lshtm.ac.uk.
Ogobara Doumbo, M.D., is deceased.
Longer-Term Assessment of Azithromycin for Reducing Childhood Mortality in Africa
Jeremy D. Keenan, M.D., Ahmed M. Arzika, M.S., Ramatou Maliki, M.S., Nameywa Boubacar, M.D., Sanoussi Elh Adamou, M.D., Maria Moussa Ali, M.P.H., Catherine Cook, M.P.H., Elodie Lebas, R.N., Ying Lin, M.P.H., Kathryn J. Ray, Ph.D., Kieran S. O’Brien, M.P.H., Thuy Doan, M.D., Ph.D., et al.
Article
Figures/Media
Metrics
28 References
2 Citing Articles
Letters
Abstract
BACKGROUND
The MORDOR I trial (Macrolides Oraux pour Réduire les Décès avec un Oeil sur la Résistance) showed that in Niger, mass administration of azithromycin twice a year for 2 years resulted in 18% lower postneonatal childhood mortality than administration of placebo. Whether this benefit could increase with each administration or wane owing to antibiotic resistance was unknown.
METHODS
In the Niger component of the MORDOR I trial, we randomly assigned 594 communities to four twice-yearly distributions of either azithromycin or placebo to children 1 to 59 months of age. In MORDOR II, all these communities received two additional open-label azithromycin distributions. All-cause mortality was assessed twice yearly by census workers who were unaware of participants’ original assignments.
RESULTS
In the MORDOR II trial, the mean (±SD) azithromycin coverage was 91.3±7.2% in the communities that received twice-yearly azithromycin for the first time (i.e., had received placebo for 2 years in MORDOR I) and 92.0±6.6% in communities that received azithromycin for the third year (i.e., had received azithromycin for 2 years in MORDOR I). In MORDOR II, mortality was 24.0 per 1000 person-years (95% confidence interval [CI], 22.1 to 26.3) in communities that had originally received placebo in the first year and 23.3 per 1000 person-years (95% CI, 21.4 to 25.5) in those that had originally received azithromycin in the first year, with no significant difference between groups (P=0.55). In communities that had originally received placebo, mortality decreased by 13.3% (95% CI, 5.8 to 20.2) when the communities received azithromycin (P=0.007). In communities that had originally received azithromycin and continued receiving it for an additional year, the difference in mortality between the third year and the first 2 years was not significant (−3.6%; 95% CI, −12.3 to 4.5; P=0.50).
CONCLUSIONS
We found no evidence that the effect of mass administration of azithromycin on childhood mortality in Niger waned in the third year of treatment. Childhood mortality decreased when communities that had originally received placebo received azithromycin. (Funded by the Bill and Melinda Gates Foundation; ClinicalTrials.gov number, NCT02047981.)
The MORDOR I trial (Macrolides Oraux pour Réduire les Décès avec un Oeil sur la Résistance) showed that twice-yearly azithromycin distributions reduced childhood mortality by 14% in communities in Niger, Malawi, and Tanzania1-8 (and unpublished data). The greatest observed benefit was seen in Niger, with 18% fewer deaths in communities that were randomly assigned to azithromycin than in those that were randomly assigned to placebo. This observed effect could decrease or increase over time for a number of reasons. For example, a beneficial effect of azithromycin might wane with an increase in bacteria that are resistant to it (i.e., through selection). This is possible because mass azithromycin distributions in trachoma programs have selected for macrolide-resistant strains of Streptococcus pneumoniae and Escherichia coli, and azithromycin clearly selected for resistant bacteria in Niger during MORDOR I. Alternatively, azithromycin might delay the death of a frail child but not ultimately prevent it. Such an effect could occur if antibiotic distributions diminished the development of protective immunity in a population by reducing its exposure to pathogens. On the other hand, the observed efficacy of azithromycin increased with each distribution during the 2 years of MORDOR I, which suggests the possibility of an enhanced effect with additional treatment.9 Such an effect could be explained by cumulative reduction in pathogens with each distribution or by antibiotic-resistant bacteria being less fit. Moreover, efficacy could improve over time if implementation improves with experience.
In the MORDOR II trial, we provided twice-yearly azithromycin for an additional year in the Nigerien communities that had participated in the MORDOR I trial; azithromycin was administered in both the communities that had originally received placebo and the communities that had originally received azithromycin. In this way, we were able to compare the first year of mass azithromycin treatment with the third year of treatment. Given the limited efficacy seen in Malawi and Tanzania, further study in those communities was not pursued. Because azithromycin affects transmissible diseases, treating an individual person may influence others in the same community. Thus, randomization and intervention were at the community level, and inference of efficacy was made at the community level. MORDOR II continued the cluster-randomized trial design of MORDOR I.10
Methods
ELIGIBILITY
This continuation study was planned only in the Niger districts that had participated in the MORDOR I trial — not in Malawi or Tanzania, where mortality was lower. The Niger component of the MORDOR I trial was conducted in the districts of Boboye and Loga. The randomization unit was the grappe (i.e., a cluster of households representing the smallest government health unit), and those with a population between 200 and 2000 inhabitants on the most recent pre-MORDOR I census were eligible for enrollment. Communities remained in the MORDOR II continuation study even if the population had drifted out of this range. All children 1 to 59 months of age (truncated to month) and weighing at least 3800 g were eligible for treatment.
RANDOMIZATION AND MASKING
The original MORDOR I randomization and interventions were performed at the community level. The randomization list was generated with the use of R software, version 3.5.1 (R Foundation for Statistical Computing). Although all communities received azithromycin in MORDOR II, participants and observers remained unaware of the original treatment assignments in MORDOR I.
CENSUS
In MORDOR II, a house-to-house census was performed during the two additional 6-month periods in the same manner as in MORDOR I.9 In each community, all households were recorded in a custom-built mobile application (Conexus), with the head of household and the global positioning system coordinates facilitating identification of the household at the subsequent census. All children in the household who were 1 to 59 months of age were identified. The vital status (alive, dead, or unknown) and residence (moved within community, moved outside community, or unknown) were recorded for each child. The vital status of children enrolled in the preceding census who were older than 59 months in the current census was also assessed, although these children were not included in the next study period. Children younger than 1 month and pregnant women were documented in the application in anticipation of their enrollment in the subsequent census. The census was conducted in the communities in the same general order each period. Data were uploaded to the Salesforce cloud database service (Salesforce), and data cleaning was performed with the use of the Salesforce platform, StataSE software, version 15.1 (Statacorp), and R software.
INTERVENTION
Each child who was 1 to 59 months of age at the time of the census was offered a single directly observed dose of oral azithromycin. Children were given a volume of suspension corresponding to 20 mg per kilogram of body weight; alternatively, in the case of children who could stand, the dose was approximated with the use of height-based data. Only azithromycin suspension was used (not tablets), and children known to be allergic to macrolides were not treated. Azithromycin was administered at the time of the census or during additional visits in an attempt to achieve at least 80% coverage. Administration of study medication was documented in the mobile application for each child, and community coverage was calculated relative to the census data. Serious adverse events other than death within 2 weeks after administration of azithromycin were reported. The parents or guardians of the children reported the events to the grappe chief, who reported to the Nigerien study coordinator, who in turn reported within 24 hours to the data coordinating center at the University of California, San Francisco (UCSF). Full study details and the statistical analysis plan are provided in the protocol, available with the full text of this article at NEJM.org.
PRIMARY OUTCOME
The prespecified primary outcome was the community-level, all-cause mortality rate determined by twice-yearly census. Each intercensal period was treated separately, with a death counted only when a child was recorded as being alive and living in the household at one census and recorded as having died while residing in the community at the subsequent census. By design, no attempt was made to track down the status of a child after the child had moved out of the community. Person-time at risk was calculated as days between consecutive censuses; children who moved, died, or had an unknown follow-up status contributed to one half of the intercensal period. All children documented as alive and present in the household at the initial census of each intercensal period were included in the analysis. No changes to trial methods or outcomes were made after the continuation trial had begun.
SECONDARY OUTCOMES
In addition to the primary comparison between community groups, mortality was compared longitudinally within each community group with the use of similar methods. Mortality during the first 2 years of assigned regimen (placebo or azithromycin) was contrasted with mortality during the third year of treatment (azithromycin in both groups), accounting for clustering at the community level.
TRIAL OVERSIGHT
Approval for the study was obtained from the ethics committees of the Niger Ministry of Health, the UCSF institutional review board, and Emory University. Oral informed consent was obtained from the local ministry of health, village leaders, and the children’s guardians. No incentives were offered for participation. The study was undertaken in accordance with the principles of the Declaration of Helsinki.
STATISTICAL ANALYSIS
This continuation study was prespecified before the start of the MORDOR I trial, contingent on that trial showing a significant beneficial effect. For the original sample size for MORDOR I–Niger, we estimated that 624 clusters would provide 80% power to detect 15% lower mortality in communities receiving azithromycin than in those receiving placebo, assuming a community size of 668, 17% of the population in the target range of 1 to 59 months of age, a death rate of 2% per year, a coefficient of variation of 0.51, and loss to follow-up of 10% per year (details are provided in the statistical analysis plan in the protocol). After we updated calculations on the basis of results from MORDOR I–Niger — using the observed coefficient of variation of 0.34 and a death rate of 2.5% per year — we estimated that the MORDOR II study would have 80% power to detect a 15.5% effect size. Because all communities were treated twice yearly, no interim efficacy analysis or futility stopping rule was prespecified, although a three-member data and safety monitoring committee (see the Supplementary Appendix, available at NEJM.org) reviewed the data at the end of the year. For the comparison of communities receiving their first year of azithromycin distributions with those receiving their third year of azithromycin, the prespecified primary analysis was negative binomial regression of the number of deaths per community, with the treatment group as a predictor and person-time at risk as an offset. Hypothesis testing was two-sided, with an alpha of 0.05. P values were determined with Monte Carlo permutation testing (10,000 replications). Intracluster correlations were accounted for through the use of community-level data, and community-level heterogeneity was taken into account by the dispersion parameter in the negative binomial regression. All statistical analyses were conducted with R software.
Results
PARTICIPATING COMMUNITIES
Figure 1.
Randomization and Follow-up in the MORDOR I and MORDOR II Trials in Niger.
Table 1.
Demographic Characteristics of Communities and Participants at the Start of MORDOR II.
As shown in Figure 1, all 594 communities from the Niger component of MORDOR I were followed in MORDOR II. Census periods were from February 2017 through August 2017, September 2017 through January 2018, and February 2018 through August 2018. Demographic characteristics of communities in both groups at 24 months are shown in Table 1.
In MORDOR II, the mean (±SD) azithromycin coverage was 92.0±6.6% of the targeted population in communities that had received azithromycin in MORDOR I and 91.3±7.2% in the communities that had received placebo in MORDOR I (Table S1 in the Supplementary Appendix). The census status was recorded as moved or unknown in 4079 of 64,225 cases (6.4%) in the communities receiving their first year of azithromycin, and as moved or unknown in 4685 of 72,108 cases (6.5%) in the communities receiving their third year of azithromycin, with no significant difference between the two groups (P=0.48) (Table S2 in the Supplementary Appendix).
PRIMARY OUTCOME
Figure 2.
Mortality Rate in Each Group over Time.
Table 2.
Mortality Rate over Time among Children 1 to 59 Months of Age.
Mortality rates in the two treated groups are shown according to intercensal period (Figure 2) and according to year (Table 2). A total of 24.0 deaths (95% confidence interval [CI], 22.1 to 26.3) per 1000 person-years were recorded in communities that were receiving the first year of azithromycin distribution, and 23.3 deaths (95% CI, 21.4 to 25.5) per 1000 person-years were recorded in communities receiving the third year of azithromycin distribution. We found no evidence that the first year of treatment had a greater effect on mortality than the third year of treatment, with 3.5% (95% CI, −8.3 to 14) more deaths in communities receiving the first year of treatment (P=0.55) (Table 2).
SECONDARY OUTCOMES
In the communities that originally received twice-yearly placebo distribution for 2 years, mortality decreased over the next year when children were treated with twice-yearly azithromycin (decrease in mortality, 13.3%; 95% CI, 5.8 to 20.2; P=0.007). In the communities that had originally received azithromycin, the difference in mortality between the third year and the first 2 years was not significant (−3.6%; 95% CI, −12.3 to 4.5; P=0.50).
SERIOUS ADVERSE EVENTS
Mortality rates are as reported. Medical review was unable to determine whether any additional serious adverse events were caused by azithromycin.
Discussion
In MORDOR I, twice-yearly oral azithromycin administered for 2 years to postneonatal preschool children resulted in significantly lower all-cause mortality (by 18%) in Niger than administration of placebo.9 In MORDOR II, both groups of Niger communities from MORDOR I (communities that had received azithromycin and those that had received placebo) received twice-yearly doses of azithromycin during the third year; this design allowed us to compare a third year of treatment with the first year of treatment. We found no evidence that the benefit of azithromycin waned in the third year. Some experts had hypothesized that there would be a decrease in efficacy of azithromycin with more distributions owing to the selection of antibiotic-resistant bacteria.11-13 Repeated mass azithromycin distributions for trachoma have indeed selected for macrolide resistance in nasopharyngeal S. pneumoniae and rectal E. coli.1-3,6,14 Resistance was noted in the nasopharynx and stool in children in Niger in MORDOR I (unpublished data). Resistance emerging during mass azithromycin distributions could theoretically have curbed or even reversed any potential survival benefit in this community.
We also found no evidence that the effect of azithromycin was enhanced with additional distributions. Enhancement was possible since the overall benefit in the three sites of MORDOR I increased from 7% to 22% with each of the first four twice-yearly distributions; however, that apparent increase was not statistically significant.9 Here, the comparison between the first and third years of treatments did not support either an increasing or decreasing effect on mortality with additional rounds of azithromycin. Longer follow-up will be necessary to determine whether the mortality effect is sustained past the third year of distributions and whether potential side effects, such as increased colonization with antibiotic-resistant bacteria, cause negative health outcomes.
The communities receiving their first year of treatment had 13% lower mortality than they had in the previous 2 years during which they had received placebo. Although this longitudinal analysis was not a randomized comparison and is therefore subject to confounding, the result does support the original MORDOR I finding of a reduction in mortality in the Niger analysis. Mortality decreased with the first of the two additional distributions, which suggests that cumulative treatments are not necessary to achieve efficacy. This is consistent with a secondary analysis of MORDOR I in which the number of deaths was relatively lower in the first 3 months after a twice-yearly distribution of azythromycin.15 The convergence of mortality rates in the two groups in MORDOR II — when both groups of communities received the same treatment — bolsters the argument that the difference in MORDOR I was caused by intervention and not by imbalanced randomization.
The study has a number of limitations. Because it is a large, simple trial, little information was collected on each child and community.10 Deaths were determined by consecutive censuses. Children who were born and died between censuses did not contribute to either the number of deaths or the person-time at risk for the primary outcome. Death rates may have differed among children who moved away or had an unknown census status. Although the comparison assessed whether a community’s prior treatment history affected the results, it was not designed to analyze a child’s prior treatment history. Communities in Malawi and Tanzania (part of MORDOR I) were not studied in MORDOR II, since the mortality rates were lower in those communities. Cluster-randomized trials run the risk of contamination between groups, which could dampen the observed effect. Although the intervention itself was not subject to contamination in MORDOR II, since all communities were given the same treatment, infections could spread between nearby communities and cause contamination. Although this could theoretically explain the MORDOR II findings, contamination did not prevent a highly significant result in MORDOR I, so invoking this explanation would require contamination in the third year only. No child in MORDOR I and II had ever received azithromycin as part of a trachoma program, but macrolide use outside the study was not recorded. Since distributions were offered only twice yearly, a child’s first treatment might not be until 7 months of age. Supplementary treatments given during a scheduled vaccination visit to a health clinic might prove to be a more reliable way of reaching younger infants. The assessments of years 1 and 2 as compared with year 3 were longitudinal within each group, and not randomized between groups. Conditions may have changed between these time periods. This study did not investigate whether morbidity increased or decreased with azithromycin. The study also did not evaluate the mechanism by which azithromycin reduced mortality, although its antimicrobial effect presumably plays a role, since a majority of child deaths in the geographic area of the trial are attributed to infectious disease.16 Smaller parallel trials with detailed microbiologic and anthropometric assessments were conducted, and these may provide insight into mechanism of action.17,18 Azithromycin has been linked to cardiac death in adults, although epidemiologic results are mixed and may not be relevant to children in this setting.19-22 Later development of atopic disease has been associated with infant antibiotic use in general, and macrolides in particular.23 Rare side effects or those apparent only later in life would be difficult to assess with this study design.
The International Trachoma Initiative has now distributed more than 800 million doses of oral azithromycin in the trachoma control programs sponsored by the World Health Organization.24 Azithromycin has proved quite effective in reducing the prevalence of, and in some cases completely eliminating, the strains of ocular chlamydia that cause trachoma.2,4,6,25-27 Annual trachoma case distribution numbers are now declining as countries continue to meet control criteria.24 Trachoma is no longer endemic — or never was — in many regions with high childhood mortality. Thus, the majority of children now being born in areas with the highest mortality among those younger than 5 years of age will not receive azithromycin as part of trachoma programs.28 The treatment regimen for trachoma and the regimen used in the MORDOR studies differed; the regimen for trachoma is annual mass azithromycin treatment of children 6 months of age or older, whereas the regimen in MORDOR I and II was twice-yearly distributions targeted to children 1 to 59 months of age. The MORDOR studies distributed approximately one third as many doses of azithromycin per community per year as would a trachoma program. If azithromycin for childhood mortality were targeted to areas with very high mortality, such as Niger, only a fraction of the total antibiotics used in trachoma programs would be required in those areas.
In summary, this study showed no evidence that the beneficial effect of mass twice-yearly azithromycin distribution on childhood mortality waned in the third year of distribution as compared with the first year. However, twice-yearly oral azithromycin distribution in year 3 resulted in significantly reduced mortality in communities that had previously received 2 years of twice-yearly placebo distributions in MORDOR I. This longitudinal observation supports the original MORDOR I community-randomized trial results. Selection of antibiotic-resistant strains of pathogenic bacteria may eventually reduce efficacy and should continue to be monitored with longer follow-up.
Supported by a grant (OPP1032340) from the Bill and Melinda Gates Foundation. Pfizer provided both the azithromycin and the placebo oral suspensions. The Salesforce Foundation provided user licenses to Salesforce.com and cloud storage.
Disclosure forms provided by the authors are available with the full text of this article at NEJM.org.
This is the New England Journal of Medicine version of record, which includes all Journal editing and enhancements. The Author Final Manuscript, which is the author’s version after external peer review and before publication in the Journal, is available under a CC BY license at PMC6512890.
A data sharing statement provided by the authors is available with the full text of this article at NEJM.org.
We thank That Man May See and Research to Prevent Blindness, as well as the staff of the Biblioteca Angelica Ministry of Cultural Heritage and Activities (MiBAC).
Author Affiliations
From the Francis I. Proctor Foundation (J.D.K., C.C., E.L., Y.L., K.J.R., K.S.O., T.D., C.E.O., T.C.P., T.M.L.), the Departments of Ophthalmology (J.D.K., T.D., C.E.O., T.C.P., T.M.L.) and Epidemiology and Biostatistics (K.J.R., C.E.O., T.C.P., T.M.L.), and the Institute for Global Health Sciences (C.E.O., T.M.L.), University of California, San Francisco, San Francisco, and the University of California, Berkeley, School of Public Health, Berkeley (K.S.O.) — both in California; and the Carter Center (A.M.A., R.M., N.B., S.E.A., M.M.A., E.K.C.) and Emory University (P.M.E.), Atlanta, and the International Trachoma Initiative, Decatur (P.M.E.) — all in Georgia.
Address reprint requests to Dr. Lietman at 513 Parnassus Ave., Medical Sciences Bldg., Rm. S309, University of California, San Francisco, San Francisco, CA 94143-0944, or at tom.lietman@ucsf.edu.
Daniel Chandramohan, Ph.D., Alassane Dicko, M.D., Issaka Zongo, Ph.D., Issaka Sagara, M.D., Matthew Cairns, Ph.D., Irene Kuepfer, Ph.D., Modibo Diarra, M.D., Amadou Barry, M.D., Amadou Tapily, M.D., Frederic Nikiema, M.D., Serge Yerbanga, Ph.D., Samba Coumare, M.D., et al.
Article
Figures/Media
Metrics
19 References
1 Citing Article
Comment
Comments open through June 12, 2019
Abstract
BACKGROUND
Mass administration of azithromycin for trachoma control led to a sustained reduction in all-cause mortality among Ethiopian children. Whether the addition of azithromycin to the monthly sulfadoxine–pyrimethamine plus amodiaquine used for seasonal malaria chemoprevention could reduce mortality and morbidity among African children was unclear.
METHODS
We randomly assigned children 3 to 59 months of age, according to household, to receive either azithromycin or placebo, together with sulfadoxine–pyrimethamine plus amodiaquine, during the annual malaria-transmission season in Burkina Faso and Mali. The drug combinations were administered in four 3-day cycles, at monthly intervals, for three successive seasons. The primary end point was death or hospital admission for at least 24 hours that was not due to trauma or elective surgery. Data were recorded by means of active and passive surveillance.
RESULTS
In July 2014, a total of 19,578 children were randomly assigned to receive seasonal malaria chemoprevention plus either azithromycin (9735 children) or placebo (9843 children); each year, children who reached 5 years of age exited the trial and new children were enrolled. In the intention-to-treat analysis, the overall number of deaths and hospital admissions during three malaria-transmission seasons was 250 in the azithromycin group and 238 in the placebo group (events per 1000 child-years at risk, 24.8 vs. 23.5; incidence rate ratio, 1.1; 95% confidence interval [CI], 0.88 to 1.3). Results were similar in the per-protocol analysis. The following events occurred less frequently with azithromycin than with placebo: gastrointestinal infections (1647 vs. 1985 episodes; incidence rate ratio, 0.85; 95% CI, 0.79 to 0.91), upper respiratory tract infections (4893 vs. 5763 episodes; incidence rate ratio, 0.85; 95% CI, 0.81 to 0.90), and nonmalarial febrile illnesses (1122 vs. 1424 episodes; incidence rate ratio, 0.79; 95% CI, 0.73 to 0.87). The prevalence of malaria parasitemia and incidence of adverse events were similar in the two groups.
CONCLUSIONS
Among children in Burkina Faso and Mali, the addition of azithromycin to the antimalarial agents used for seasonal malaria chemoprevention did not result in a lower incidence of death or hospital admission that was not due to trauma or surgery than antimalarial agents plus placebo, although a lower disease burden was noted with azithromycin than with placebo. (Funded by the Joint Global Health Trials scheme; ClinicalTrials.gov number, NCT02211729.)
Malaria transmission is concentrated during a few months of the year in much of the Sahel and sub-Sahel regions of Africa. In these areas, seasonal malaria chemoprevention — the administration of sulfadoxine–pyrimethamine plus amodiaquine to children at monthly intervals three or four times during the malaria-transmission season — has been a highly effective approach to malaria control.1 Seasonal malaria chemoprevention is now being implemented widely across these regions.2 The frequent contact between children and health care workers that is needed for seasonal malaria chemoprevention provides an opportunity for the delivery of other health interventions.
Mass administration of azithromycin has been a highly effective approach to trachoma control.3 Reductions in the incidences of skin, gastrointestinal, and respiratory infections have been recorded after mass administration of azithromycin.4-8 Nevertheless, the finding in Ethiopia of a 49% reduction in all-cause mortality among children 1 to 9 years of age during the year after mass administration of a single dose of azithromycin — a reduction that was sustained during a 26-month follow-up period — was surprising.9,10 Consequently, we conducted a randomized, double-blind, placebo-controlled trial to determine whether the addition of azithromycin to the sulfadoxine–pyrimethamine plus amodiaquine given for seasonal malaria chemoprevention could have a similar effect on overall child mortality and morbidity.
Methods
TRIAL OVERSIGHT
The trial was approved by the ethics committees of the London School of Hygiene and Tropical Medicine, London; the Malaria Research and Training Center, University of Bamako, Bamako, Mali; the Ministry of Health, Ouagadougou, Burkina Faso; and the national regulatory authorities of Burkina Faso and Mali. A data and safety monitoring board reviewed serious adverse events, monitored the overall progress of the trial, approved the statistical analysis plan, and archived the locked database before the data were unmasked. A steering committee reviewed the protocol (available with the full text of this article at NEJM.org) and provided overall advice. The authors vouch for the accuracy and completeness of the data and the fidelity of the trial to the protocol.
SITES AND POPULATION
The trial was conducted in the Houndé district of Burkina Faso and in the Bougouni district of Mali (Fig. S1 in the Supplementary Appendix, available at NEJM.org). Information about these communities and the children who live in them is provided in the protocol and the Supplementary Appendix.
ENROLLMENT AND RANDOMIZATION
A household census was conducted in June 2014, and children of either sex who were 3 to 59 months of age on August 1, 2014, were eligible for enrollment in the trial. After written informed consent was obtained from the child’s parent or guardian, the child received a long-lasting insecticide-treated bed net. Children were excluded if they had a chronic disease or a known allergy to sulfadoxine–pyrimethamine, amodiaquine, or azithromycin or if they were taking cotrimoxazole. The household census was repeated in May 2015 and in May 2016 to recruit additional eligible children and to detect any deaths that had been missed through the surveillance system. Each year, children who were still younger than 60 months of age on August 1 remained in follow-up for the subsequent trial year, and children who had reached 5 years of age on or before July 31 exited the trial on that date. Enrollment of children in the trial started on August 25, 2014, in Mali and on August 28, 2014, in Burkina Faso.
Randomization was performed according to household to avoid the potential effect of within-household transmission of infection; all eligible children who shared a kitchen were assigned to the same trial group. To mask the trial-group assignments for the trial team and caregivers, a placebo for azithromycin of identical appearance was used.
INTERVENTIONS
Children who were enrolled in the trial received the assigned preventive regimen during the annual peak malaria-transmission season (August to November). The drug combinations were administered in four 3-day cycles, at monthly intervals, for three successive seasons. Infants 3 to 11 months of age received a combined 250 mg of sulfadoxine and 12.5 mg of pyrimethamine plus 75 mg of amodiaquine on day 1 and received 75 mg of amodiaquine on days 2 and 3 (Guilin Pharmaceutical, Shanghai, China). In addition, they were randomly assigned to receive either 100 mg of azithromycin or matching placebo on days 1, 2, and 3 (Cipla, Mumbai, India). Children 1 to 4 years of age received double these doses. All doses were based on age and administered by trial staff. All trial drugs were purchased from the manufacturers with the use of a grant provided by the U.K. Medical Research Council and were provided to the children free of cost.
Each year, the drug combinations were prepacked in resealable plastic bags by pharmacists who were not part of the trial team. Each child was assigned one large bag that contained four smaller bags, each of which contained the sulfadoxine–pyrimethamine, amodiaquine, and either azithromycin or placebo for one of the four cycles. The child received a photo identification card that had a quick response code (known as a QR code) that encoded the child’s name, the mother’s name, the child’s date of birth, the census number, and a randomization number with a check digit. A label on the large bag also had a QR code that encoded the same variables. On the day of administration of the trial drugs, the QR codes on the identification card and on the large bag were scanned with a tablet computer to link the child to the correct bag. The trial drugs were kept at the trial office and were administered under direct observation by trial staff.
When the child was seen for administration of the trial drugs, if a diagnosis of malaria was confirmed with the use of a rapid diagnostic test, the child was not given the assigned regimen and instead received a dose of artemether–lumefantrine. Children with other illnesses were referred to a local health center for investigation and treatment.
END POINTS
The primary end point was death or hospital admission for at least 24 hours that was not due to trauma or elective surgery during the intervention period. The intervention period was defined as the period from the administration of the first dose of the first cycle of the trial drugs until 30 days after the administration of the first dose of the last cycle. For children who did not receive the first dose of the first or last cycle, the date that the dose was scheduled to be administered was used.
The prespecified secondary end points were the individual components of the primary end point; death or hospital admission for at least 24 hours during the entire trial period; parasitologically confirmed malaria, which was defined as a febrile illness (a history of fever within 24 hours or a measured temperature of ≥37.5°C) and either a positive rapid diagnostic test or a positive blood smear; radiographically confirmed pneumonia; clinically diagnosed pneumonia or lower respiratory tract infection; gastrointestinal infection; nonmalarial fever, which was defined as a febrile illness that was not due to malaria, lower or upper respiratory tract infection, or gastrointestinal infection; and anemia (hemoglobin level, <10 g per deciliter) or severe anemia (hemoglobin level, <7 g per deciliter) at the end of the malaria-transmission season. An exploratory analysis was performed to investigate the incidence of skin diseases.
SURVEILLANCE
Deaths and hospital admissions were recorded throughout the trial period, but only events that occurred during the intervention period contributed to the primary end point. Data regarding vital status were updated during an annual census and during an exit census that was conducted in March 2017. Deaths that occurred outside a health facility were assessed with the use of the World Health Organization verbal autopsy questionnaire.11 The trial-group assignments were masked for all assessments. Data regarding adverse events that occurred during the week after administration of the trial drugs were solicited from 800 children (a random selection of 200 children from each trial group in each country) on day 7 after each cycle in the first year of the trial. Details regarding surveillance are provided in the Supplementary Appendix.
In addition, 200 children (a random selection of 50 children from each trial group in each country) were visited each week during the malaria-transmission season for active detection of malaria infection. At the end of each malaria-transmission season (≥30 days after the administration of the first dose of the last cycle of the trial drugs), 4000 children (a random selection of approximately 1000 children from each trial group in each country) were included in a cross-sectional survey to assess the prevalence of malaria parasitemia. In addition, at the end of each malaria-transmission season, blood slides were obtained from 500 primary school children 5 to 12 years of age who lived in each trial area to provide data on the prevalence of malaria parasitemia among children who did not receive seasonal malaria chemoprevention.
STATISTICAL ANALYSIS
On the basis of data from previous seasonal malaria chemoprevention trials performed in Burkina Faso12 and Mali,13 we assumed that the incidence of death or hospital admission that was not due to trauma or elective surgery during the malaria-transmission season would be approximately 15 per 1000 children who received seasonal malaria chemoprevention plus placebo and the rate of loss to follow-up would be 10% per year. On the basis of these assumptions, we calculated that the enrollment of 19,200 children (9600 per country) for three malaria-transmission seasons would give the trial 90% power to detect a 25% lower incidence of the primary end point with azithromycin than with placebo.
The primary analysis was an intention-to-treat analysis of deaths and hospital admissions that occurred during the 4-month intervention period each year. The intention-to-treat population included all the children who had been screened and enrolled in the trial. A per-protocol analysis of the primary end point was also performed. Children who were seen on the first day of administration of the trial drugs for all four cycles of a particular year were included in the per-protocol population for that year. All analyses of secondary end points were performed on an intention-to-treat basis; these analyses were not adjusted for multiple comparisons, and thus P values are not reported for secondary end points.
For each child, person-time at risk was calculated as the time from the date of enrollment until 30 days after the date that the first dose of the last cycle was scheduled to be administered. If applicable, the following end dates were used instead: if the child was lost to follow-up, the date that the child was last seen; if the child emigrated, the date of permanent emigration; if the child died, the date of death; or if the child reached 5 years of age, the last day of the trial year in which the child reached 5 years of age.
The incidence rate ratio of the primary end point was estimated with the use of Poisson regression models, with a gamma-distributed random effect to account for clustering of episodes within households. Regression models were adjusted for trial site and stratified according to follow-up time with the use of Lexis expansion. As prespecified in the statistical analysis plan, effect modification according to trial site and year of age was assessed with the use of the likelihood ratio test, without adjustment for multiple comparisons, since only these two subgroup analyses were performed. Prevalence rate ratios were estimated with the use of Poisson regression models, with a robust standard error to account for randomization according to household.12
Results
CHILDREN AND COVERAGE WITH SEASONAL MALARIA CHEMOPREVENTION
Figure 1.
Screening, Randomization, and Follow-up.
In July 2014, a total of 19,578 children from 9618 households were randomly assigned to receive seasonal malaria chemoprevention plus either azithromycin (9735 children) or placebo (9843 children) (Figure 1). Each year, additional children in the specified age range were enrolled (6287 in the second year and 5748 in the third year), and children who were 5 years of age on August 1 exited the trial. Because of establishment of new households in the trial areas and migration of children into households that were originally included in the trial, the overall number of children in the trial increased each year. At the last follow-up visit, there were 10,885 children in the group that received seasonal malaria chemoprevention plus azithromycin and 10,852 in the group that received seasonal malaria chemoprevention plus placebo.
The two trial groups were well matched with regard to baseline characteristics (Table S1 in the Supplementary Appendix). Coverage with long-lasting insecticide-treated bed nets was high and similar in the two groups. The percentage of children who received at least three directly observed cycles of the assigned regimen was 92.8% in the first year, 86.8% in the second year, and 84.3% in the third year (Table S2 in the Supplementary Appendix).
EFFICACY
Table 1.
Incidence of Death or Hospital Admission That Was Not Due to Trauma or Elective Surgery during Three Successive Malaria-Transmission Seasons in the Intention-to-Treat Population.
In the intention-to-treat analysis, the overall number of deaths and hospital admissions that were not due to trauma or elective surgery was similar in the two trial groups: 250 in the azithromycin group and 238 in the placebo group (events per 1000 child-years at risk, 24.8 vs. 23.5; incidence rate ratio, 1.1; 95% confidence interval [CI], 0.88 to 1.3) (Table 1). In the per-protocol analysis, the overall number was 173 in the azithromycin group and 158 in the placebo group (events per 1000 child-years at risk, 19.8 vs. 18.2; incidence rate ratio, 1.1; 95% CI, 0.88 to 1.4) (Table S3 in the Supplementary Appendix).
Figure 2.
Incidence of Death or Hospital Admission That Was Not Due to Trauma or Elective Surgery during Three Successive Malaria-Transmission Seasons in Burkina Faso and Mali.
In the intention-to-treat population, there was evidence of an interaction between trial group and trial site (P=0.02 by the likelihood ratio test). The incidence of the primary end point was higher in the azithromycin group than in the placebo group in Burkina Faso (incidence rate ratio, 1.3; 95% CI, 1.0 to 1.7) but not in Mali (incidence rate ratio, 0.84; 95% CI, 0.64 to 1.1). The incidence of the primary end point according to year of age was similar in the two trial groups, both in the entire trial population (P=0.44 for interaction by the likelihood ratio test) and in each country individually (Figure 2, and Figs. S2 and S3 in the Supplementary Appendix). The causes of deaths and hospital admissions that occurred during the intervention period and during the entire trial period are shown in Tables S4 through S9 in the Supplementary Appendix. Malaria was the most prominent cause of death and hospital admission in each trial group.
Table 2.
Incidence of Clinical Events Treated at Health Centers or by Community Health Workers during Three Successive Malaria-Transmission Seasons in Burkina Faso and Mali in the Intention-to-Treat Population.
The incidence of clinic visits for the following events was lower with antimalarial agents plus azithromycin than with antimalarial agents plus placebo: gastrointestinal infections (incidence rate ratio, 0.85; 95% CI, 0.79 to 0.91), upper respiratory tract infections (incidence rate ratio, 0.85; 95% CI, 0.81 to 0.90), and nonmalarial fevers (incidence rate ratio, 0.79; 95% CI, 0.73 to 0.87) (Table 2). An exploratory analysis showed that, among children who had a clinic visit for nonmalarial fever, 269 children in the azithromycin group and 442 children in the placebo group had a skin condition (events per 1000 child-years at risk, 26.6 vs. 43.6; incidence rate ratio, 0.61; 95% CI, 0.53 to 0.73). Among children who had a clinic visit for nonfebrile illness, 428 children in the azithromycin group and 556 children in the placebo group had a skin condition (events per 1000 child-years at risk, 42.4 vs. 54.8; incidence rate ratio, 0.77; 95% CI, 0.67 to 0.89). The number of children with a skin condition that most likely had a bacterial cause was substantially lower in the azithromycin group than in the placebo group, especially among those with nonmalarial fever (66 vs. 145; events per 1000 child-years at risk, 6.54 vs. 14.3; incidence rate ratio, 0.46; 95% CI, 0.33 to 0.62).
Table 3.
Prevalence of Malaria Parasitemia and Anemia among Children in the Trial and Prevalence of Malaria Parasitemia among Primary School Children.
Among children who were randomly selected for weekly follow-up visits during the intervention period, the prevalence of malaria parasitemia ranged from 3 to 7% and was similar in the two groups. At the end of the malaria-transmission season, the prevalence of malaria parasitemia ranged from 4 to 10% and the prevalence of anemia ranged from 20 to 26%. For both variables, results were similar in the two trial groups (Table 3). Among primary school children who lived in the trial areas and did not receive seasonal malaria chemoprevention, the prevalence of malaria parasitemia at the end of the malaria-transmission season ranged from 50 to 65% (Table 3).
SAFETY
No severe adverse events that were judged by investigators to be related to the trial drugs were recorded. Diarrhea was the most frequent adverse event reported after administration of the trial drugs. The incidences of diarrhea and of other adverse events that occurred during the week after administration of the trial drugs were similar in the two trial groups (Table S10 in the Supplementary Appendix).
Discussion
In this trial, the incidence of death or hospital admission that was not due to trauma or elective surgery did not differ significantly between children who received seasonal malaria chemoprevention plus azithromycin and children who received seasonal malaria chemoprevention plus placebo when data from Burkina Faso and Mali were combined for the primary analysis. However, the incidence of death or hospital admission was higher with azithromycin than with placebo in Burkina Faso but not in Mali. No plausible mechanism to explain this difference was found, and given the borderline increased incidence in Burkina Faso, it may be a chance finding. The incidences of gastrointestinal infections, upper respiratory tract infections, and nonmalarial fevers were lower with azithromycin than with placebo (by 15%, 15%, and 21%, respectively), and in an exploratory analysis, the incidence of skin diseases, especially those that most likely had a bacterial cause, was also lower with azithromycin; these results are consistent with findings of previous studies in which azithromycin was used in trachoma control programs.4-8
The findings of our trial contrast with those of the MORDOR (Mortality Reduction after Oral Azithromycin) trial conducted in Malawi, Niger, and Tanzania, in which azithromycin was given to children younger than 5 years of age twice a year for 2 years and then was associated with a 13.5% (95% CI, 6.7 to 19.8) lower overall all-cause mortality than placebo, with the effect being most marked in Niger.14 Children who were assigned to the azithromycin group in our trial had greater exposure to azithromycin than those in the MORDOR trial (four courses each year for up to 3 years in our trial, as compared with two courses each year for 2 years in the MORDOR trial); such enhanced exposure might have been expected to have an effect on mortality that was at least equal to the effect seen in the MORDOR trial, but this was not the case.
There are several possible explanations for the different outcomes of these two trials. One possible explanation is that azithromycin, which has antimalarial activity,15 contributed to decreased mortality in the MORDOR trial partly through its effect on malaria, and this benefit was lost when an additional, effective antimalarial combination was given at the same time as azithromycin. However, the effect of azithromycin on malaria has been inconsistent when azithromycin has been given in mass drug administration programs.16-18 In addition, all the children in our trial received sulfadoxine–pyrimethamine, which has weak antimicrobial properties, and this may have reduced the potential benefit of adding another antimicrobial to the regimen. Finally, coverage with a pneumococcal conjugate vaccine was high among the children in our trial, and this may have reduced the potential benefit of azithromycin in lowering mortality from pneumonia.
In the MORDOR trial, the greatest effect of azithromycin on all-cause mortality was seen in the first year of life.14 This finding suggests that, despite the results of this trial, the addition of azithromycin to the sulfadoxine–pyrimethamine used for intermittent prevention of malaria in infants19 is an option worth investigating in countries with a high malaria burden in the first year of life.
The prevalence of malaria at the end of the malaria-transmission season was substantially lower among children who were enrolled in our trial than among primary school children who were living in the same areas and were not receiving seasonal malaria chemoprevention, a finding that indicates that seasonal malaria chemoprevention with sulfadoxine–pyrimethamine plus amodiaquine was having a major protective effect against malaria in these populations. Nevertheless, the proportion of deaths and hospital admissions that were attributable to malaria was still large in both trial groups, despite good access to treatment and high coverage with long-lasting insecticide-treated bed nets. Effective control of malaria in these and other, similar areas necessitates additional control measures.
A limitation of the trial is that randomization was performed according to household rather than village; randomization according to household reduced the potential for bias but precluded the potential for a herd effect that might have occurred had randomization according to village been performed. Only limited safety data were obtained because seasonal malaria chemoprevention with sulfadoxine–pyrimethamine plus amodiaquine and mass administration of azithromycin have now been given to millions of children with no major safety concerns.
In conclusion, among children in Burkina Faso and Mali, the addition of azithromycin to the antimalarial agents used for seasonal malaria chemoprevention did not result in a lower incidence of death or hospital admission that was not due to trauma or surgery than antimalarial agents plus placebo. We also noted that the incidences of clinic visits for gastrointestinal infections, upper respiratory tract infections, and nonmalarial febrile illnesses, without adjustment for multiple comparisons, were lower among children who received antimalarial agents plus azithromycin than among those who received antimalarial agents plus placebo.
Supported by a grant (MR/K007319/1) from the Joint Global Health Trials scheme, which includes the U.K. Medical Research Council, Department for International Development, National Institute for Health Research, and Wellcome Trust.
Disclosure forms provided by the authors are available with the full text of this article at NEJM.org.
No potential conflict of interest relevant to this article was reported.
This article was published on January 30, 2019, 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 the members of the trial steering committee (Feiko ter Kuile [chair], Kalifa Bojang, Kojo Koram, David Mabey, Morven Roberts, and Mahamadou Thera) and the members of the data and safety monitoring board (Blaise Genton [chair], Cheick Oumar Coulibaly, Umberto D’Alessandro, and Francesca Little) for their sustained assistance with the trial; Alice Greenwood for reviewing all the hospital records and verbal autopsies and validating causes of hospital admissions and deaths that were assigned by the trial team; the Ministry of Health staff in the Bougouni and Houndé districts for their assistance with running the trial; and all the caretakers and children for their participation.
Author Affiliations
From the London School of Hygiene and Tropical Medicine, London (D.C., M.C., I.K., P.M., B.G.); the Malaria Research and Training Center, University of Science, Techniques, and Technologies of Bamako, Bamako, Mali (A.D., I.S., M.D., A.B., A. Tapily, S.C., I.T., O.D.); and Institut de Recherche en Sciences de la Santé, Bobo-Dioulasso, Burkina Faso (I.Z., F.N., S.Y., A. Traore, H.T., J.-B.O.).
Address reprint requests to Dr. Chandramohan at the London School of Hygiene and Tropical Medicine, Keppel St., London WC1E 7HT, United Kingdom, or at daniel.chandramohan@lshtm.ac.uk.
Ogobara Doumbo, M.D., is deceased.
Longer-Term Assessment of Azithromycin for Reducing Childhood Mortality in Africa
Jeremy D. Keenan, M.D., Ahmed M. Arzika, M.S., Ramatou Maliki, M.S., Nameywa Boubacar, M.D., Sanoussi Elh Adamou, M.D., Maria Moussa Ali, M.P.H., Catherine Cook, M.P.H., Elodie Lebas, R.N., Ying Lin, M.P.H., Kathryn J. Ray, Ph.D., Kieran S. O’Brien, M.P.H., Thuy Doan, M.D., Ph.D., et al.
Article
Figures/Media
Metrics
28 References
2 Citing Articles
Letters
Abstract
BACKGROUND
The MORDOR I trial (Macrolides Oraux pour Réduire les Décès avec un Oeil sur la Résistance) showed that in Niger, mass administration of azithromycin twice a year for 2 years resulted in 18% lower postneonatal childhood mortality than administration of placebo. Whether this benefit could increase with each administration or wane owing to antibiotic resistance was unknown.
METHODS
In the Niger component of the MORDOR I trial, we randomly assigned 594 communities to four twice-yearly distributions of either azithromycin or placebo to children 1 to 59 months of age. In MORDOR II, all these communities received two additional open-label azithromycin distributions. All-cause mortality was assessed twice yearly by census workers who were unaware of participants’ original assignments.
RESULTS
In the MORDOR II trial, the mean (±SD) azithromycin coverage was 91.3±7.2% in the communities that received twice-yearly azithromycin for the first time (i.e., had received placebo for 2 years in MORDOR I) and 92.0±6.6% in communities that received azithromycin for the third year (i.e., had received azithromycin for 2 years in MORDOR I). In MORDOR II, mortality was 24.0 per 1000 person-years (95% confidence interval [CI], 22.1 to 26.3) in communities that had originally received placebo in the first year and 23.3 per 1000 person-years (95% CI, 21.4 to 25.5) in those that had originally received azithromycin in the first year, with no significant difference between groups (P=0.55). In communities that had originally received placebo, mortality decreased by 13.3% (95% CI, 5.8 to 20.2) when the communities received azithromycin (P=0.007). In communities that had originally received azithromycin and continued receiving it for an additional year, the difference in mortality between the third year and the first 2 years was not significant (−3.6%; 95% CI, −12.3 to 4.5; P=0.50).
CONCLUSIONS
We found no evidence that the effect of mass administration of azithromycin on childhood mortality in Niger waned in the third year of treatment. Childhood mortality decreased when communities that had originally received placebo received azithromycin. (Funded by the Bill and Melinda Gates Foundation; ClinicalTrials.gov number, NCT02047981.)
The MORDOR I trial (Macrolides Oraux pour Réduire les Décès avec un Oeil sur la Résistance) showed that twice-yearly azithromycin distributions reduced childhood mortality by 14% in communities in Niger, Malawi, and Tanzania1-8 (and unpublished data). The greatest observed benefit was seen in Niger, with 18% fewer deaths in communities that were randomly assigned to azithromycin than in those that were randomly assigned to placebo. This observed effect could decrease or increase over time for a number of reasons. For example, a beneficial effect of azithromycin might wane with an increase in bacteria that are resistant to it (i.e., through selection). This is possible because mass azithromycin distributions in trachoma programs have selected for macrolide-resistant strains of Streptococcus pneumoniae and Escherichia coli, and azithromycin clearly selected for resistant bacteria in Niger during MORDOR I. Alternatively, azithromycin might delay the death of a frail child but not ultimately prevent it. Such an effect could occur if antibiotic distributions diminished the development of protective immunity in a population by reducing its exposure to pathogens. On the other hand, the observed efficacy of azithromycin increased with each distribution during the 2 years of MORDOR I, which suggests the possibility of an enhanced effect with additional treatment.9 Such an effect could be explained by cumulative reduction in pathogens with each distribution or by antibiotic-resistant bacteria being less fit. Moreover, efficacy could improve over time if implementation improves with experience.
In the MORDOR II trial, we provided twice-yearly azithromycin for an additional year in the Nigerien communities that had participated in the MORDOR I trial; azithromycin was administered in both the communities that had originally received placebo and the communities that had originally received azithromycin. In this way, we were able to compare the first year of mass azithromycin treatment with the third year of treatment. Given the limited efficacy seen in Malawi and Tanzania, further study in those communities was not pursued. Because azithromycin affects transmissible diseases, treating an individual person may influence others in the same community. Thus, randomization and intervention were at the community level, and inference of efficacy was made at the community level. MORDOR II continued the cluster-randomized trial design of MORDOR I.10
Methods
ELIGIBILITY
This continuation study was planned only in the Niger districts that had participated in the MORDOR I trial — not in Malawi or Tanzania, where mortality was lower. The Niger component of the MORDOR I trial was conducted in the districts of Boboye and Loga. The randomization unit was the grappe (i.e., a cluster of households representing the smallest government health unit), and those with a population between 200 and 2000 inhabitants on the most recent pre-MORDOR I census were eligible for enrollment. Communities remained in the MORDOR II continuation study even if the population had drifted out of this range. All children 1 to 59 months of age (truncated to month) and weighing at least 3800 g were eligible for treatment.
RANDOMIZATION AND MASKING
The original MORDOR I randomization and interventions were performed at the community level. The randomization list was generated with the use of R software, version 3.5.1 (R Foundation for Statistical Computing). Although all communities received azithromycin in MORDOR II, participants and observers remained unaware of the original treatment assignments in MORDOR I.
CENSUS
In MORDOR II, a house-to-house census was performed during the two additional 6-month periods in the same manner as in MORDOR I.9 In each community, all households were recorded in a custom-built mobile application (Conexus), with the head of household and the global positioning system coordinates facilitating identification of the household at the subsequent census. All children in the household who were 1 to 59 months of age were identified. The vital status (alive, dead, or unknown) and residence (moved within community, moved outside community, or unknown) were recorded for each child. The vital status of children enrolled in the preceding census who were older than 59 months in the current census was also assessed, although these children were not included in the next study period. Children younger than 1 month and pregnant women were documented in the application in anticipation of their enrollment in the subsequent census. The census was conducted in the communities in the same general order each period. Data were uploaded to the Salesforce cloud database service (Salesforce), and data cleaning was performed with the use of the Salesforce platform, StataSE software, version 15.1 (Statacorp), and R software.
INTERVENTION
Each child who was 1 to 59 months of age at the time of the census was offered a single directly observed dose of oral azithromycin. Children were given a volume of suspension corresponding to 20 mg per kilogram of body weight; alternatively, in the case of children who could stand, the dose was approximated with the use of height-based data. Only azithromycin suspension was used (not tablets), and children known to be allergic to macrolides were not treated. Azithromycin was administered at the time of the census or during additional visits in an attempt to achieve at least 80% coverage. Administration of study medication was documented in the mobile application for each child, and community coverage was calculated relative to the census data. Serious adverse events other than death within 2 weeks after administration of azithromycin were reported. The parents or guardians of the children reported the events to the grappe chief, who reported to the Nigerien study coordinator, who in turn reported within 24 hours to the data coordinating center at the University of California, San Francisco (UCSF). Full study details and the statistical analysis plan are provided in the protocol, available with the full text of this article at NEJM.org.
PRIMARY OUTCOME
The prespecified primary outcome was the community-level, all-cause mortality rate determined by twice-yearly census. Each intercensal period was treated separately, with a death counted only when a child was recorded as being alive and living in the household at one census and recorded as having died while residing in the community at the subsequent census. By design, no attempt was made to track down the status of a child after the child had moved out of the community. Person-time at risk was calculated as days between consecutive censuses; children who moved, died, or had an unknown follow-up status contributed to one half of the intercensal period. All children documented as alive and present in the household at the initial census of each intercensal period were included in the analysis. No changes to trial methods or outcomes were made after the continuation trial had begun.
SECONDARY OUTCOMES
In addition to the primary comparison between community groups, mortality was compared longitudinally within each community group with the use of similar methods. Mortality during the first 2 years of assigned regimen (placebo or azithromycin) was contrasted with mortality during the third year of treatment (azithromycin in both groups), accounting for clustering at the community level.
TRIAL OVERSIGHT
Approval for the study was obtained from the ethics committees of the Niger Ministry of Health, the UCSF institutional review board, and Emory University. Oral informed consent was obtained from the local ministry of health, village leaders, and the children’s guardians. No incentives were offered for participation. The study was undertaken in accordance with the principles of the Declaration of Helsinki.
STATISTICAL ANALYSIS
This continuation study was prespecified before the start of the MORDOR I trial, contingent on that trial showing a significant beneficial effect. For the original sample size for MORDOR I–Niger, we estimated that 624 clusters would provide 80% power to detect 15% lower mortality in communities receiving azithromycin than in those receiving placebo, assuming a community size of 668, 17% of the population in the target range of 1 to 59 months of age, a death rate of 2% per year, a coefficient of variation of 0.51, and loss to follow-up of 10% per year (details are provided in the statistical analysis plan in the protocol). After we updated calculations on the basis of results from MORDOR I–Niger — using the observed coefficient of variation of 0.34 and a death rate of 2.5% per year — we estimated that the MORDOR II study would have 80% power to detect a 15.5% effect size. Because all communities were treated twice yearly, no interim efficacy analysis or futility stopping rule was prespecified, although a three-member data and safety monitoring committee (see the Supplementary Appendix, available at NEJM.org) reviewed the data at the end of the year. For the comparison of communities receiving their first year of azithromycin distributions with those receiving their third year of azithromycin, the prespecified primary analysis was negative binomial regression of the number of deaths per community, with the treatment group as a predictor and person-time at risk as an offset. Hypothesis testing was two-sided, with an alpha of 0.05. P values were determined with Monte Carlo permutation testing (10,000 replications). Intracluster correlations were accounted for through the use of community-level data, and community-level heterogeneity was taken into account by the dispersion parameter in the negative binomial regression. All statistical analyses were conducted with R software.
Results
PARTICIPATING COMMUNITIES
Figure 1.
Randomization and Follow-up in the MORDOR I and MORDOR II Trials in Niger.
Table 1.
Demographic Characteristics of Communities and Participants at the Start of MORDOR II.
As shown in Figure 1, all 594 communities from the Niger component of MORDOR I were followed in MORDOR II. Census periods were from February 2017 through August 2017, September 2017 through January 2018, and February 2018 through August 2018. Demographic characteristics of communities in both groups at 24 months are shown in Table 1.
In MORDOR II, the mean (±SD) azithromycin coverage was 92.0±6.6% of the targeted population in communities that had received azithromycin in MORDOR I and 91.3±7.2% in the communities that had received placebo in MORDOR I (Table S1 in the Supplementary Appendix). The census status was recorded as moved or unknown in 4079 of 64,225 cases (6.4%) in the communities receiving their first year of azithromycin, and as moved or unknown in 4685 of 72,108 cases (6.5%) in the communities receiving their third year of azithromycin, with no significant difference between the two groups (P=0.48) (Table S2 in the Supplementary Appendix).
PRIMARY OUTCOME
Figure 2.
Mortality Rate in Each Group over Time.
Table 2.
Mortality Rate over Time among Children 1 to 59 Months of Age.
Mortality rates in the two treated groups are shown according to intercensal period (Figure 2) and according to year (Table 2). A total of 24.0 deaths (95% confidence interval [CI], 22.1 to 26.3) per 1000 person-years were recorded in communities that were receiving the first year of azithromycin distribution, and 23.3 deaths (95% CI, 21.4 to 25.5) per 1000 person-years were recorded in communities receiving the third year of azithromycin distribution. We found no evidence that the first year of treatment had a greater effect on mortality than the third year of treatment, with 3.5% (95% CI, −8.3 to 14) more deaths in communities receiving the first year of treatment (P=0.55) (Table 2).
SECONDARY OUTCOMES
In the communities that originally received twice-yearly placebo distribution for 2 years, mortality decreased over the next year when children were treated with twice-yearly azithromycin (decrease in mortality, 13.3%; 95% CI, 5.8 to 20.2; P=0.007). In the communities that had originally received azithromycin, the difference in mortality between the third year and the first 2 years was not significant (−3.6%; 95% CI, −12.3 to 4.5; P=0.50).
SERIOUS ADVERSE EVENTS
Mortality rates are as reported. Medical review was unable to determine whether any additional serious adverse events were caused by azithromycin.
Discussion
In MORDOR I, twice-yearly oral azithromycin administered for 2 years to postneonatal preschool children resulted in significantly lower all-cause mortality (by 18%) in Niger than administration of placebo.9 In MORDOR II, both groups of Niger communities from MORDOR I (communities that had received azithromycin and those that had received placebo) received twice-yearly doses of azithromycin during the third year; this design allowed us to compare a third year of treatment with the first year of treatment. We found no evidence that the benefit of azithromycin waned in the third year. Some experts had hypothesized that there would be a decrease in efficacy of azithromycin with more distributions owing to the selection of antibiotic-resistant bacteria.11-13 Repeated mass azithromycin distributions for trachoma have indeed selected for macrolide resistance in nasopharyngeal S. pneumoniae and rectal E. coli.1-3,6,14 Resistance was noted in the nasopharynx and stool in children in Niger in MORDOR I (unpublished data). Resistance emerging during mass azithromycin distributions could theoretically have curbed or even reversed any potential survival benefit in this community.
We also found no evidence that the effect of azithromycin was enhanced with additional distributions. Enhancement was possible since the overall benefit in the three sites of MORDOR I increased from 7% to 22% with each of the first four twice-yearly distributions; however, that apparent increase was not statistically significant.9 Here, the comparison between the first and third years of treatments did not support either an increasing or decreasing effect on mortality with additional rounds of azithromycin. Longer follow-up will be necessary to determine whether the mortality effect is sustained past the third year of distributions and whether potential side effects, such as increased colonization with antibiotic-resistant bacteria, cause negative health outcomes.
The communities receiving their first year of treatment had 13% lower mortality than they had in the previous 2 years during which they had received placebo. Although this longitudinal analysis was not a randomized comparison and is therefore subject to confounding, the result does support the original MORDOR I finding of a reduction in mortality in the Niger analysis. Mortality decreased with the first of the two additional distributions, which suggests that cumulative treatments are not necessary to achieve efficacy. This is consistent with a secondary analysis of MORDOR I in which the number of deaths was relatively lower in the first 3 months after a twice-yearly distribution of azythromycin.15 The convergence of mortality rates in the two groups in MORDOR II — when both groups of communities received the same treatment — bolsters the argument that the difference in MORDOR I was caused by intervention and not by imbalanced randomization.
The study has a number of limitations. Because it is a large, simple trial, little information was collected on each child and community.10 Deaths were determined by consecutive censuses. Children who were born and died between censuses did not contribute to either the number of deaths or the person-time at risk for the primary outcome. Death rates may have differed among children who moved away or had an unknown census status. Although the comparison assessed whether a community’s prior treatment history affected the results, it was not designed to analyze a child’s prior treatment history. Communities in Malawi and Tanzania (part of MORDOR I) were not studied in MORDOR II, since the mortality rates were lower in those communities. Cluster-randomized trials run the risk of contamination between groups, which could dampen the observed effect. Although the intervention itself was not subject to contamination in MORDOR II, since all communities were given the same treatment, infections could spread between nearby communities and cause contamination. Although this could theoretically explain the MORDOR II findings, contamination did not prevent a highly significant result in MORDOR I, so invoking this explanation would require contamination in the third year only. No child in MORDOR I and II had ever received azithromycin as part of a trachoma program, but macrolide use outside the study was not recorded. Since distributions were offered only twice yearly, a child’s first treatment might not be until 7 months of age. Supplementary treatments given during a scheduled vaccination visit to a health clinic might prove to be a more reliable way of reaching younger infants. The assessments of years 1 and 2 as compared with year 3 were longitudinal within each group, and not randomized between groups. Conditions may have changed between these time periods. This study did not investigate whether morbidity increased or decreased with azithromycin. The study also did not evaluate the mechanism by which azithromycin reduced mortality, although its antimicrobial effect presumably plays a role, since a majority of child deaths in the geographic area of the trial are attributed to infectious disease.16 Smaller parallel trials with detailed microbiologic and anthropometric assessments were conducted, and these may provide insight into mechanism of action.17,18 Azithromycin has been linked to cardiac death in adults, although epidemiologic results are mixed and may not be relevant to children in this setting.19-22 Later development of atopic disease has been associated with infant antibiotic use in general, and macrolides in particular.23 Rare side effects or those apparent only later in life would be difficult to assess with this study design.
The International Trachoma Initiative has now distributed more than 800 million doses of oral azithromycin in the trachoma control programs sponsored by the World Health Organization.24 Azithromycin has proved quite effective in reducing the prevalence of, and in some cases completely eliminating, the strains of ocular chlamydia that cause trachoma.2,4,6,25-27 Annual trachoma case distribution numbers are now declining as countries continue to meet control criteria.24 Trachoma is no longer endemic — or never was — in many regions with high childhood mortality. Thus, the majority of children now being born in areas with the highest mortality among those younger than 5 years of age will not receive azithromycin as part of trachoma programs.28 The treatment regimen for trachoma and the regimen used in the MORDOR studies differed; the regimen for trachoma is annual mass azithromycin treatment of children 6 months of age or older, whereas the regimen in MORDOR I and II was twice-yearly distributions targeted to children 1 to 59 months of age. The MORDOR studies distributed approximately one third as many doses of azithromycin per community per year as would a trachoma program. If azithromycin for childhood mortality were targeted to areas with very high mortality, such as Niger, only a fraction of the total antibiotics used in trachoma programs would be required in those areas.
In summary, this study showed no evidence that the beneficial effect of mass twice-yearly azithromycin distribution on childhood mortality waned in the third year of distribution as compared with the first year. However, twice-yearly oral azithromycin distribution in year 3 resulted in significantly reduced mortality in communities that had previously received 2 years of twice-yearly placebo distributions in MORDOR I. This longitudinal observation supports the original MORDOR I community-randomized trial results. Selection of antibiotic-resistant strains of pathogenic bacteria may eventually reduce efficacy and should continue to be monitored with longer follow-up.
Supported by a grant (OPP1032340) from the Bill and Melinda Gates Foundation. Pfizer provided both the azithromycin and the placebo oral suspensions. The Salesforce Foundation provided user licenses to Salesforce.com and cloud storage.
Disclosure forms provided by the authors are available with the full text of this article at NEJM.org.
This is the New England Journal of Medicine version of record, which includes all Journal editing and enhancements. The Author Final Manuscript, which is the author’s version after external peer review and before publication in the Journal, is available under a CC BY license at PMC6512890.
A data sharing statement provided by the authors is available with the full text of this article at NEJM.org.
We thank That Man May See and Research to Prevent Blindness, as well as the staff of the Biblioteca Angelica Ministry of Cultural Heritage and Activities (MiBAC).
Author Affiliations
From the Francis I. Proctor Foundation (J.D.K., C.C., E.L., Y.L., K.J.R., K.S.O., T.D., C.E.O., T.C.P., T.M.L.), the Departments of Ophthalmology (J.D.K., T.D., C.E.O., T.C.P., T.M.L.) and Epidemiology and Biostatistics (K.J.R., C.E.O., T.C.P., T.M.L.), and the Institute for Global Health Sciences (C.E.O., T.M.L.), University of California, San Francisco, San Francisco, and the University of California, Berkeley, School of Public Health, Berkeley (K.S.O.) — both in California; and the Carter Center (A.M.A., R.M., N.B., S.E.A., M.M.A., E.K.C.) and Emory University (P.M.E.), Atlanta, and the International Trachoma Initiative, Decatur (P.M.E.) — all in Georgia.
Address reprint requests to Dr. Lietman at 513 Parnassus Ave., Medical Sciences Bldg., Rm. S309, University of California, San Francisco, San Francisco, CA 94143-0944, or at tom.lietman@ucsf.edu.
Δεν υπάρχουν σχόλια:
Δημοσίευση σχολίου