Impact of the addition of azithromycin to antimalarials used for seasonal malaria chemoprevention on antimicrobial resistance of Streptococcus pneumoniae

Abstract Objective A trial was conducted in Burkina Faso and Mali to investigate whether addition of azithromycin to the antimalarials used for seasonal malaria chemoprevention reduces mortality and hospital admissions of children. We tested the sensitivity of nasal isolates of Streptococcus pneumoniae obtained during this trial to azithromycin and other antibiotics. Methods Azithromycin or placebo was administered monthly, in combination with the antimalarials used for seasonal malaria chemoprevention, for four months, over the annual malaria transmission seasons of 2014, 2015, and 2016. Nasopharyngeal swabs were collected from 2773 Burkinabe and 2709 Malian children on seven occasions: in July and December each year prior to and after drug administration, and at a final survey in early 2018. Pneumococci were isolated from nasopharyngeal swabs and tested for sensitivity to azithromycin and other antibiotics. Results A total of 5482 samples were collected. In Burkina Faso, the percentage of pneumococcal isolates resistant to azithromycin among children who had received it increased from 4.9% (95% CI: 2.4%, 9.9%) before the intervention to 25.6% (95% CI: 17.6%, 35.7%) afterward. In Mali, the increase was from 7.6% (95% CI: 3.8%, 14.4%) to 68.5% (95% CI: 55.1%, 79.4%). The percentage of resistant isolates remained elevated (17.7% (95% CI: 11.1%, 27.1%) in Burkina Faso and 19.1% (95% CI: 13.5%, 26.3%) in Mali) among children who had received azithromycin 1 year after stopping the intervention. An increase in resistance to azithromycin was also observed in children who had received a placebo but it was less marked. Conclusion Addition of azithromycin to the antimalarial combination used for seasonal malaria chemoprevention was associated with an increase in resistance of pneumococci to azithromycin and erythromycin, which persisted 1 year after the last administration of azithromycin.


Introduction
In Burkina Faso and Mali, malaria continues to be a burden with a large number of cases and high mortality rates despite control efforts. In 2016, WHO estimated 7.9 million malaria cases with 21 300 fatalities in Burkina Faso and 7.2 million cases with 12 400 fatalities in Mali [1]. Malaria is highly seasonal; 60% to 80% of cases occur during the raining season in both countries.
Mass drug administration (MDA) with azithromycin (AZ) is being widely deployed as a highly effective method for the control of trachoma [2]. The incidence of respiratory, gastrointestinal and skin infections, and malaria [3][4][5][6][7][8] is lower in children who participated in mass AZ campaigns. An additional, surprising finding was detection of a more than 50% reduction in mortality in children who participated in an AZ MDA program in Ethiopia, a reduction that was sustained over a period of 26 months [9,10]. This unexpected finding led to the MORDOR (Macrolides Oraux pour R eduire les D ec es avec un Oeil sur la R esistance) trial, which investigated the impact of two rounds of AZ MDA on mortality in children under the age of 5 years in Malawi, Niger, and Tanzania [11]. An overall reduction in mortality of 13.5 % was observed, with the reduction being most marked in Niger and in children under the age of 1 year.
On the basis of the findings of the study in Ethiopia, we hypothesized that adding AZ to the antimalarial combination sulphadoxine-pyrimethamine and amodiaquine (SP + AQ) used for seasonal malaria chemoprevention (SMC) would further reduce child mortality and severe morbidity, and we conducted a trial in 19 200 children aged 3-59 months in Burkina Faso and Mali to investigate this hypothesis. Modest reductions in the incidence of gastrointestinal, respiratory tract and skin infections, and non-malaria fevers were seen, but the addition of AZ to the antimalarial used for SMC had no impact on child mortality or hospital admissions [12].
A concern over the use of AZ for MDA programs is that this will enhance resistance to macrolide antibiotics. An increase in the resistance of Streptococcus pneumoniae to AZ has been observed in several MDA studies, although this has often been only short term [5,[13][14][15][16][17][18][19]]. An increase in resistance of Staphylococcus aureus to AZ after MDA for trachoma control has also been reported [20]. For these reasons, the antibiotic sensitivity of nasopharyngeal isolates of S. pneumoniae was studied during the SMC + AZ trial and the results from this study are presented in this paper.

Design and conduct of the SMC + AZ trial
Details of the trial in Burkina Faso and Mali to investigate the impact of adding AZ to the antimalarial drugs used for SMC have been published previously [12]. In brief, 19 200 children aged 3-59 months were randomised to receive SP and AQ with either AZ or placebo. Randomization was by household. Infants aged 3-11 months received SP 250 mg/12.5 mg and AQ 75 mg on day 1 and AQ 75 mg on days 2 and 3. In addition, they received AZ 100 mg or matching AZ placebo on days 1, 2, and 3. Children aged 1-4 years received double these doses. SP + AQ was supplied by Guilin Pharmaceutical (Shanghai, China), and AZ and matching placebo were supplied by CIPLA (Mumbai, India). All doses of treatments were given by trial staff. Coverage with the monthly treatments was high, with more than 80% of children receiving three or four rounds of treatment each year. Deaths, hospital admissions, and attendances at clinics were recorded throughout the study period as described earlier [12]. Cross-sectional surveys were undertaken at the end of each malaria transmission season. The overall outline of the study is shown in Figure S1.

Nasopharyngeal carriage surveys
Nasopharyngeal swabs were collected from 2773 Burkinabe and 2709 Malian children at cross-sectional surveys at seven time points. Children were randomly selected by an independent statistician, with a new sample drawn for each time point. Swabs were collected in July 2014, 2015, and 2016 just before the first round of AZ or placebo was given and in December of each of these years 4-6 weeks after the last round of AZ or placebo had been given (hereafter referred to as pre-and post-2014 etc.). Swab samples were also taken early in 2018, 1 year after the last administration of AZ or placebo.
Randomized children were gathered in a health facility and trained staff took samples on-site while completing the required documents. A sample was taken from the posterior wall of the child's nasopharynx using a calcium alginate swab (FLOQSwabs, Copan Diagnostics Inc., Murrieta, CA, USA) and immediately transferred to a cryotube containing skimmed milk-tryptone-glucose-glycerol medium (STGG). The cryotubes were labeled and placed in a cold box prior to transfer to the laboratory within 8 h of collection and stored at À80°C until analyzed [21,22].

Laboratory methods
Standard protocols for the analysis of nasopharyngeal samples were used [21]; these are described in the supplement.

Statistical analysis
The primary study endpoint was the prevalence of nasal carriage isolates of S. pneumoniae resistant to AZ at the seven time points described above. A secondary endpoint was the overall prevalence of pneumococcal carriage in the two intervention groups at the same time points. Exploratory endpoints included the analysis of the sensitivity of pneumococci to other antibiotics.

Sample size
A sample size of 400 children per survey per country was chosen for the nasopharyngeal substudy on the basis that the pneumococcal carriage prevalence would be 50%, and thus, 200 samples would be positive and available for resistance assays at each survey in each country. Assuming that the prevalence of resistance was 50% among the 200 samples available for resistance assays, the precision of the estimate of resistance for each country, at each time point, would be within 15% of the true value. Written consent was obtained from parents or guardians for inclusion of a child in the overall trial, and further consent was obtained from parents or guardians of children selected for the pneumococcal carriage substudy. A data safety monitoring board reviewed serious adverse events and monitored the trial's overall progress. An international steering committee reviewed the protocol and provided advice throughout the course of the study.

Study population and samples
A total of 5482 nasopharyngeal specimens were collected, 2773 from children in Hound e, Burkina Faso (1379 from the AZ and 1394 from the placebo group) and 2709 from children in Bougouni, Mali (1346 from the AZ and 1363 from the placebo group). Sex and age distribution of study children were well balanced between the AZ and placebo groups, and the prevalence of pneumococcal carriage was comparable in each group at baseline in both countries (Tables 1 and 2). The numbers of swabs obtained at each survey and tested for antibiotic resistance are shown in Tables 3 and 4.
A history of consumption of antibiotics other than AZ in the 30 days prior to the collection of a swab was reported by 4.1-14.6% of children in Burkina Faso and by 6.9-11.9% of children in Mali in different surveys (Table 1). Antibiotic use among the subset of children with a recent morbidity episode was higher, often exceeding 50%. Amoxicillin was the most commonly prescribed antibiotic, accounting for 120 of the 140 (85.7%) reported antibiotic treatments of participants in Burkina Faso and for 112 of 180 (62.2%) treatments in Mali. Erythromycin (10.0% of antibiotic treatments), co-trimoxazole (10.6%), and metronidazole (21.7%) were also commonly used in Mali.
Pneumococcal conjugate (PCV13) vaccination was introduced into the Expanded Programme of Immunization in 2013 in Burkina Faso and in 2011 in Mali. Vaccination cards were often not available at the time of the annual census of all children in the trial. Vaccine coverage with PCV13 in Burkina Faso among children whose vaccine status was known was 58.3% when first measured before the 2015 malaria transmission season, but improved substantially in 2016 (Table 1). In Mali, PCV vaccine coverage was >90% at all survey contacts.

Prevalence of nasopharyngeal carriage of Streptococcus pneumoniae
The overall prevalence of nasopharyngeal carriage of S. pneumoniae at baseline was 67.4% in Burkina Faso and 63.5% in Mali (Tables 1 and 2; Figure 1a,b). Overall carriage remained >50% at the five subsequent surveys during the study period, with the exception of the post-2016 survey, when overall carriage was 46.7% in Burkina Faso and 40.4% in Mali. Carriage was generally lower in the AZ group in Burkina Faso but differences between the two study groups were not marked, with overlapping confidence intervals, apart from two occasions when prevalence was significantly lower in the AZ group: post-2014 in Burkina Faso, with a prevalence ratio (PR) of 0.79 (0.68, 0.92; P = 0.003), and post-2016 in Mali (PR 0.62 (0.47, 0.80), P < 0.001).
One year after the last administration of SMC with AZ, the prevalence of pneumococcal carriage was 60.0% in Burkina Faso (Table 1 and Figure 1a) and slightly lower in the group that had previously received AZ than in the placebo group (PR 0.86 (0.73, 1.03); P = 0.095). In Mali, the prevalence of pneumococcal carriage was 88.3% one year after the last administration of SMC with AZ (Table 2 and Figure 1b), again slightly but not markedly lower in the group that received AZ (PR 0.94 (0.88, 1.01); P = 0.096). Because of the unexpectedly high isolation rate in the last survey conducted in Mali, these samples were retested and a high isolation rate was confirmed (Table S1).
3.1% using the E-test (Figure 2a,b). In Mali, the E-test was used only to confirm resistance in positive samples identified by the disk diffusion assay and, at baseline, only two samples were tested, both of which were positive with each assay. In Burkina Faso, results obtained using the disk diffusion assay showed that the prevalence of resistance to AZ increased over time in both study groups, reaching approximately 10% or more from the post-2015 survey onwards. Prevalence was higher in the AZ group at each of the first six surveys, apart from the pre-2016 survey, although differences were only marked at two time points to be positive by E-test at the 2016 surveys. Because of the very high level of resistance to AZ found in the post-2016 survey, these samples were retested and a high level of concordance between the two sets of testing was found (Table S2). Analysis of changes in the pattern of resistance to AZ over time by age group did not show any age effect in either Burkina Faso or Mali ( Figure S2).

Resistance of pharyngeal isolates of Streptococcus pneumoniae to erythromycin
Resistance to erythromycin, assessed with a disk diffusion assay, showed a very similar pattern to that seen for AZ ( Figures S3 and S4). In Burkina Faso, the prevalence of resistance increased from a prevalence at baseline of 3.5%, exceeding approximately 10% at all time points after the post-2015 survey. The prevalence of resistance was higher in the AZ than in the placebo group at all but one survey, but there was only weak statistical evidence of a true difference on two occasions: PR 1.73 (1.00, 2.99; P = 0.049) at the post-2015 survey, and PR 1.86 (1.00, 3.47; P = 0.051) at the post-2016 survey. There was no evidence of a difference persisting between the groups at the 2018 survey.
In Mali, the results for erythromycin also closely mirrored those obtained for AZ: There was a relatively slow increase in the prevalence of resistance during the first . At the 2018 survey, the prevalence of erythromycin resistance had dropped below 17% in both study groups and there was no evidence of a difference between groups.

Resistance of pharyngeal isolates of Streptococcus pneumoniae to penicillin
In Burkina Faso, the E-test was used to determine sensitivity to penicillin. The overall prevalence of resistance at the first survey was 13.5%; during the remainder of the study period, this ranged between a low of 5.1% at the pre-2015 survey and a high of 22.5% at the post-2015 survey, with no evidence of any difference between the study groups (

Resistance of pharyngeal isolates of Streptococcus pneumoniae to other antibiotics
Resistance to ceftriaxone, norfloxacin, and vancomycin was measured only in isolates obtained in Burkina Faso. Overall, a very low prevalence of resistance to ceftriaxone was found using either a disk diffusion method or an E-test over the whole study period (0.61% and 0.31% respectively). Resistance to norfloxacin, tested with a disk diffusion test, was slightly more frequent and found in 85 samples (5.3% of those tested) over the study period, ranging from 11 positives (3.8% of those tested) at baseline to a low prevalence at the end of the study period, with only one positive at the post-2016 survey and three positives at the 2018 survey. No resistance to vancomycin was found in any of the 1633 isolates tested.

Discussion
This study evaluated the impact of AZ, combined with SMC, given once a month for 4 months (August to November) over a 3-year period (2014-2016) on the resistance of S. pneumoniae to AZ. This was a more intense treatment schedule than that used in previous MDA studies employing AZ for control of trachoma [7,8,[13][14][15][16][17][18][19][20]23]. The two regions of Hound e in Burkina Faso and Bougouni in Mali had received the last distribution of AZ (Zithromax) for control of trachoma in 2007 and 2011, respectively, and this is, therefore, unlikely to have affected the results of this study [24]. The overall prevalence of carriage of S. pneumoniae declined modestly over time in children who had received either AZ or placebo with the exception of an unexpected increase in the 2018 survey in Mali; this increase was confirmed on retesting so is likely to be a true, but unexplained, finding. Carriage tended to be lower in the AZ group than in the placebo group, but apart from two time points (post-2014 in Burkina Faso and post-2016 in Mali), differences were not large and may have been due to chance.
In both Burkina Faso and Mali, resistance of pharyngeal isolates of S. pneumoniae to AZ and erythromycin increased substantially during the course of the study and this persisted for a year after the last drug administration. This contrasts with the findings in most other studies in which the prevalence of resistance has usually returned close to baseline at surveys some months after the last drug administration [16,18]. There was strong statistical evidence of a difference between the AZ and placebo groups only at the post-2015 survey in Burkina Faso and at the post-2016 survey in Mali. Although study children were randomized by household, rather than individually, there may have been sufficient mixing between young children in neighboring households to dilute differences between the intervention groups; a cluster-randomized village trial might have found more marked differences between study groups. As expected, patterns of resistance to erythromycin matched those seen for AZ. A modest level of resistance to penicillin, as assessed by the E-test, was found but resistance to other antibiotics tested was rare.
Incorporation of AZ into the SMC treatment regimen did not have any significant impact on deaths or hospital admission due to non-traumatic causes but addition of AZ did reduced the incidence of visits to a health facility or community health worker due to an acute respiratory tract, gastrointestinal or skin infection, and of nonmalaria fever by about 20% [11]. These gains will need to be balanced against the costs of adding AZ to the SMC regimen, currently being assessed, and against the risk of inducing widespread resistance of S. pneumoniae, and perhaps other bacterial pathogens, including gastrointestinal pathogens, to macrolide antibiotics. Erythromycin and (particularly) AZ are currently used infrequently for treatment of young children at government-supported clinics in the study areas, but they may be prescribed more frequently in pharmacies and private clinics and their loss to the list of effective and affordable antibiotics would be a significant one.
This study has some weaknesses. Although efforts were made to standardize laboratory procedures in Burkina  Faso and Mali, some differences in methods emerged, in part because of difficulties in obtaining reagents at the appropriate time. Hence, it was decided to undertake separate analyses in each country rather than merging the data. Nevertheless, very similar results were found in each country. In addition, data on coverage with pneumococcal conjugate vaccine were not recorded for each study participant, but the information that was available indicates that a high proportion had received at least one dose of this vaccine. Policymakers are currently considering the potential of widespread deployment of AZ mass drug administration as an infant survival strategy in countries where infant mortality remains high. The results of this study suggest that the potential for inducing resistance to macrolide antibiotics in important pathogens will need to be taken into consideration when policy decisions are being made on the costs and benefits of this intervention.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Figure S1. Schematic showing timing of surveys for pneumococcal sampling in relation to study interventions and the malaria transmission seasons. Figure S2. Results of resistance to azithromycin by age obtained during three annual pre-and post-intervention surveys and 1 year after the last post-intervention survey was done in Burkina Faso (a) and Mali (b). Figure S3. Results of disc diffusion assays for testing for resistance to erythromycin and its comparison to resistance to azithromycin in isolates obtained during three annual pre-and post-intervention surveys and 1 year after the last post-intervention survey was done in Burkina Faso. Figure S4. Results of disc diffusion assays for testing for resistance to erythromycin and its comparison to resistance to azithromycin in isolates obtained during three annual pre-and post-intervention surveys and 1 year after the last post-intervention survey was done in Mali.