Assessment of Long-Term Macrolide Exposure on the Oropharyngeal Microbiome and Macrolide Resistance in Healthy Adults and Consequences for Onward Transmission of Resistance

ABSTRACT

While the use of long-term macrolide therapy to prevent exacerbations in chronic respiratory diseases is widespread, its impact on the oropharyngeal microbiota and macrolide resistance, and the potential for onward transmission of resistance to close contacts are poorly understood. We determined the effects of long-term exposure to azithromycin or erythromycin on phenotypic and genotypic macrolide resistance within the oropharyngeal microbiome of healthy adults and their close contacts in a randomized, single-blinded, parallel-group trial of 4 weeks of twice-daily oral 400 mg erythromycin ethylsuccinate or twice-daily oral 125 mg azithromycin. Using oropharyngeal swabs collected from 20 index healthy adults and 20 paired close contacts, the oropharyngeal microbial composition and macrolide resistance in streptococci were assessed by 16S rRNA sequencing and antibiotic susceptibility testing of oropharyngeal cultures, respectively, at baseline and weeks 4 and 8 (washout). Targeted quantitative PCR of antibiotic resistance genes was performed to evaluate paired changes in resistance gene levels in index patients and close contacts and to relate the potential transmission of antibiotic resistance. Neither azithromycin nor erythromycin altered oropharyngeal microbiota characteristics significantly. Proportional macrolide resistance in oropharyngeal streptococci increased with both erythromycin and azithromycin, remaining above baseline levels for the azithromycin group at washout. Levels of resistance genes increased significantly with azithromycin[erm(B) and mef] and erythromycin (mef), returning to baseline levels at washout only for the erythromycin group. We found no evidence of onward transmission of resistance to close contacts, as indicated by the lack of concomitant changes in resistance gene levels detected in close contacts. (This study has been registered with the Australian and New Zealand Clinical Trials Registry under identifier ACTRN12617000278336.)

INTRODUCTION

Long-term macrolide therapy, particularly that involving azithromycin or erythromycin, is a common strategy to reduce exacerbation frequency in those with chronic respiratory diseases, including asthma (1), bronchiectasis (2), cystic fibrosis (3), and chronic obstructive pulmonary disease (4). However, long-term exposure to azithromycin or erythromycin is known to result in increased macrolide resistance in oropharyngeal (OP) (5–7) and lower airway (8–10) microbiota, including in streptococci (6, 7). Macrolide exposure also has implications on the selection of bacteria with tetracycline resistance (11) due to the cocarriage of macrolide and tetracycline resistance determinants on mobile genetic elements (12).

The emergence of resistance does not appear to be associated with reduced efficacy of macrolides in preventing acute exacerbations of chronic lung disease (13). This is in keeping with a macrolide mode of action that is believed to primarily involve immune modulation (14, 15) or suppression of bacterial pathogenicity (16). However, with the increasing long-term use of macrolides to manage chronic respiratory diseases (17), there is concern about the potential for increased population-level carriage of resistance determinants. Such an increase would be problematic in regions where macrolide resistance in Streptococcus pneumoniae is relatively low (18), where macrolides are commonly used as first-line therapy for respiratory tract diseases such as nontuberculous mycobacteria (NTM) infections (19, 20), and in severe cases of community-acquired pneumoniae (21).

It has been suggested that the widespread use of azithromycin might contribute disproportionately to the emergence of resistance compared to that of erythromycin and other macrolides due to its long half-life and once-daily dosing regimen (22). However, direct comparisons between azithromycin and erythromycin on the induction of resistance in oropharyngeal streptococci, or the selection of transmissible resistance determinants in the wider oropharyngeal microbiota, have not been reported. Additionally, the potential for onward transmission of resistant bacterial populations that arise from macrolide exposure has also not been assessed directly.

Determining the relative impact of macrolides on the development of resistance would inform shared decision-making for safe and long-term prophylaxis. However, assessments of macrolide-associated effects in those with chronic lung conditions present considerable challenges that confound assessments of treatment impact. In particular, exposure to antibiotics (macrolide and nonmacrolide) is common in these individuals, which can disrupt commensal microbiology (23) and select for macrolide resistance genes (24). High levels of antibiotic exposure in those with chronic lung disease also reduce our ability to assess potential onward dispersion of antibiotic resistance to close contacts who have long-term exposure to those with increased resistance carriage (25).

It is evident that increases in macrolide resistance can arise from short periods of macrolide exposure in healthy adults (≤7 days) (6). Studies involving longer antibiotic exposure are critical in detecting microbiological changes and the potential for resistance transfer resulting from long-term macrolide therapy. We therefore tested the hypothesis that long-term exposure (4 weeks) with the long-acting macrolide, azithromycin, would result in greater levels of phenotypic and genotypic macrolide resistance in oropharyngeal microbiota than erythromycin (a short-acting macrolide) in healthy adults with no recent exposure to macrolides. In addition, we investigated whether macrolide therapy was associated with increased carriage of macrolide resistance in close contact of recipients, consistent with onward transmission.

RESULTS

The 20 index cases (healthy adults) were randomly assigned to one of the two study arms, azithromycin (AZM) or erythromycin ethylsuccinate (ERY). Treatment groups had similar baseline demographics, including age, sex, and body mass index (BMI) (Table 1). Average baseline corrected QT (QTc) intervals of the azithromycin and erythromycin groups were within the normal values expected for healthy individuals (≤440 ms) (Table 1), with no significant QTc changes observed in the index patients across the study period [ΔQTc(ABX − baseline), AZM = −4.5 ± 18.6; ERY = −6.2 ± 18.8; P = 0.84].

TABLE 1.

Baseline characteristics of participants^a

Characteristic	Data for:		P value
	Erythromycin (n = 10)	Azithromycin (n = 10)
Age (yrs)	36 ± 12	40 ± 15	0.113
Female (no. [%])	9 (90)	6 (60)	0.303
Body mass index (BMI) (kg/m2)	27.7 ± 5.9	28.5 ± 5.1	0.727
QTc (ms [median, IQR])	425 (421–431)	411 (392–430)	0.147
Heart rate (beats per minute)	76 ± 13	72 ± 12	0.575
Time spent with close contact (hours per day [median, IQR])	8.0 (4.8–8.1)	8.0 (6.9–9.1)	0.269

^aData are represented as mean ± standard deviation (SD) unless stated otherwise. P values were calculated using a t test (age, BMI, and heart rate), Mann-Whitney test (QTc and time spent with close contact), or Fisher’s exact test, according to data characteristics.

Impact of azithromycin and erythromycin on oropharyngeal microbiota.

The baseline oropharyngeal (OP) microbial community of healthy index subjects was dominated by commensal airway microbiota, including the genera Streptococcus, Veillonella, Prevotella, Rothia, Gemella, Leptotrichia, Actinomyces, Fusobacterium, Porphyromonas, and Neisseria (Fig. S1 in the supplemental material). Microbiota composition of the azithromycin and erythromycin groups was comparable at baseline.

Four weeks of oral azithromycin or erythromycin did not result in significant changes in either the number of bacterial taxa detected (species richness) or in OP microbiota diversity (Faith’s phylogenetic diversity) (Fig. S2). Macrolide exposure also did not alter OP microbiota composition (Fig. S3A). Despite not reaching statistical significance, the magnitude of changes in microbiota composition was greater in the azithromycin group than the erythromycin group. Changes in bacterial relative abundance associated with exposure to azithromycin (four bacterial genera) or erythromycin (two bacterial genera) were observed (Table S4), although they were not significant when corrected for false-discovery rate (FDR P > 0.05).

We further assessed levels of specific opportunistic pathogens that are commonly detectable within the oropharynx and are frequently associated with lower respiratory tract infection. OP levels of Streptococcus pneumoniae and Haemophilus parainfluenzae were not significantly altered by macrolide treatment or individually by erythromycin or azithromycin (Fig. S3B and C). No index individuals had detectable levels of Haemophilus influenzae or Staphylococcus aureus at any time point.

Phenotypic resistance in oropharyngeal streptococci following macrolide exposure.

Macrolide use has been associated with an expansion in macrolide-resistant streptococci carrying transmissible macrolide-resistant genes (6, 22). Given the high prevalence of Streptococcus within the oropharynx (detected in all index cases and at a mean relative abundance of 42.7%), the induction of phenotypic macrolide resistance in streptococci following 4 weeks of azithromycin and erythromycin treatment and the residual effects on macrolide resistance levels following antibiotic cessation (4 weeks postexposure) were assessed within and between groups.

Prior to exposure, subjects in both groups carried a similar proportion of macrolide-resistant streptococci (ERY median [interquartile range], 6.5% [1.8 to 34.4%]; AZM, 3.7% [2.9 to 10.6%]). Both erythromycin and azithromycin exposure resulted in a significantly higher proportion of macrolide-resistant streptococci (Fig. 1). The proportion of macrolide-resistant streptococci fell during the washout period to baseline levels in the erythromycin group but remained significantly higher than baseline levels for azithromycin (Fig. 1).

FIG 1.

Proportion of macrolide-resistant streptococci. Macrolide resistance in streptococci at baseline, at week 4 of macrolide treatment (ABX), and 4 weeks after termination of the antibiotics (washout) in the erythromycin (ERY) and azithromycin (AZM) groups were determined by culturing medium from oropharyngeal swabs on Streptococcus selective medium, with and without 2 μg/mL erythromycin. The median and interquartile range are indicated by the symbols and the error bars. Statistical comparisons were performed to the respective baseline values using the Wilcoxon test.

Changes in the proportion of macrolide-resistant streptococci between baseline and week 4, and baseline and washout, were also performed to assess for between-group differences. Consistent with the above findings, the magnitude of increases in the proportion of macrolide-resistant streptococci after 4 weeks of macrolide treatment was similar between the erythromycin and azithromycin groups (Table 2). The proportion of macrolide-resistant streptococci trended higher in the azithromycin group than the erythromycin group in the washout period but did not achieve statistical significance (P = 0.051).

TABLE 2.

Mean difference in the proportion of macrolide-resistant streptococci from baseline^a

Time of treatment	Difference from baseline (mean ± SD [%]) for:		P value
	Erythromycin (n = 10)	Azithromycin (n = 10)
ABX	24.1 ± 33.4	48.0 ± 30.7	0.113
Washout	−0.25 ± 30.5	23.1 ± 17.7	0.051

^aThe between-group comparison was performed using the unpaired t test.

Microbiota-wide carriage of transmissible resistance determinants.

To determine the genetic determinants associated with macrolide resistance, carriage of genes encoding macrolide resistance [erm(A), erm(B), erm(C), mef, and msrA] was assessed. Given that genes encoding tetracycline resistance are also carried by conjugative transposons associated with macrolide resistance and are prevalent in airway bacteria (26), carriage of tetracycline resistance [tet(M), tet(O), tet(L), and tet(K)] was also assessed. Of these, the erm(B), erm(F), mef, tet(M), and tet(O) genes were highly represented within the study cohort at all time points (Table 3). In contrast, erm(A), erm(C), msrA, tet(L), and tet(K) were not detected in any individuals. Erythromycin and azithromycin did not alter the OP bacterial load at any time point (Fig. 2A), and baseline levels of these resistance genes did not differ between groups (Fig. 2B to F). Levels of erm(B) and mef increased significantly following 4 weeks of azithromycin and remained significantly above baseline levels 4 weeks after antibiotic cessation (washout) (Fig. 2B and D, respectively). Only the mef gene significantly increased in response to erythromycin exposure and returned to baseline levels following washout.

TABLE 3.

Proportion of individuals carrying genes conferring resistance to macrolide or tetracycline

Resistance gene	% resistance to:
	ERY (%) (n = 10)			AZM (%) (n = 10)
	Baseline	ABX	Washout	Baseline	ABX	Washout
Index cases
erm(A)	0	0	0	0	0	0
erm(B)	90	80	90	100	100	100
erm(C)	0	0	0	0	0	0
erm(F)	70	80	70	80	80	80
mef	100	100	100	100	100	100
msrA	0	0	0	0	0	0
tet(M)	100	100	100	100	100	100
tet(L)	70	50	80	100	60	100
tet(O)	0	0	0	0	0	0
tet(K)	0	0	0	0	0	0
Close contacts
erm(A)	0	0		0	0
erm(B)	80	80		90	100
erm(C)	10	0		10	0
erm(F)	90	70		80	60
mef	100	100		90	100
msrA	10	0		0	0
tet(M)	100	90		90	100
tet(L)	80	90		90	100
tet(O)	0	0		0	0
tet(K)	10	10		0	0

FIG 2.

Levels of the bacterial load and resistance genes in index cases. (A) Total bacterial load was determined based on the levels of the 16S rRNA gene. Levels of the resistance genes encoding macrolide resistance, erm(B) (B), erm(F) (C), and mef (D), as well as tetracyline resistance, tet(M) (E) and tet(O) (F), were determined at baseline, at 4 weeks of macrolide treatment (ABX), and 4 weeks after antibiotic termination (post-ABX) using target-specific quantitative PCR. Resistance gene levels were normalized against the total bacterial load. Statistical comparison between paired samples to their baseline levels was performed using the Wilcoxon test.

Accordingly, at 4 weeks of treatment relative to baseline, the increases in erm(B) levels were significantly higher in the azithromycin group than erythromycin, while the changes in mef remained similar between the groups (Table S5). In the washout period, the azithromycin group also showed significantly greater increases in erm(B) levels from baseline compared to the erythromycin group and mef (Table S5). The levels of erm(F), tet(M), and tet(O) from baseline did not differ significantly between erythromycin and azithromycin groups after 4 weeks of antibiotic treatment or following the washout.

Potential for transmission of antibiotic resistance to close contacts.

The baseline carriage of macrolide and tetracycline resistance genes in close contacts was comparable to index cases (Table 3). Levels of macrolide and tetracycline resistance genes were assessed in OP swabs from close contacts to determine whether increases in resistance carriage in the index cases during macrolide exposure result in onward transmission (Fig. 3). Despite increased levels of erm(B) and mef in the index cases, no concomitant changes in the levels of these genes were detected in the close contacts. In addition, levels of erm(F), tet(M), and tet(O) detected in close contacts also remained similar to their baseline values during the period of macrolide treatment in index cases.

FIG 3.

Levels of the bacterial load and resistance genes in close contact of index subjects. (A) Total bacterial load was determined based on the levels of the 16S rRNA gene. Levels of the resistance genes encoding macrolide resistance, erm(B) (B), erm(F) (C), and mef (D), as well as tetracyline resistance, tet(M) (E) and tet(O) (F), were determined at baseline, and at week 4 (corresponding to 4 weeks of macrolide treatment in index cases) using target-specific quantitative PCR. Resistance gene levels were normalized against the total bacterial load. Statistical comparison between paired samples to their baseline levels was performed using the Wilcoxon test.

DISCUSSION

Culture- and sequencing-based studies have previously reported long-term macrolide therapy to alter airway microbiota characteristics in those with chronic lung diseases (5, 24, 27–29). However, distinguishing the direct impact of macrolide exposure from the effects of factors such as exposure to nonmacrolide antibiotics, nonantibiotic therapies, and underlying disease characteristics is extremely challenging. To our knowledge, this is the first study of long-term macrolide exposure to control for these potential exposures. We determined, in healthy adults, the impact of long-term exposure (4 weeks) to erythromycin or azithromycin on microbiota characteristics and resistance carriage.

Both azithromycin and erythromycin are known to have broad-spectrum activity against bacteria associated with respiratory tract infection, including Gram-positive pathogens (streptococci and staphylococci), Gram-negative respiratory pathogens (including Moraxella catarrhalis, Bordetella pertussis, and H. influenzae), as well as atypical pathogens such as Mycoplasma pneumoniae and Legionella species (14, 30). While the antimicrobial spectrum of macrolides against oral commensal organisms is less well established, there is in vitro evidence of activity against most oral commensal species within the genera Actinomyces, Peptostreptococcus, Prevotella, and Porphyromonas (31). Despite their spectra of activity suggesting that macrolides should have broad impacts upon the oral microbiome, 4-week exposure to azithromycin or erythromycin at doses comparable to those used in long-term treatment of chronic respiratory conditions resulted in no substantial change in OP microbiota characteristics in healthy individuals. Our results reflected the stability of the airway microbiome, which is consistent with the relatively modest changes to the OP microbiota observed in chronic respiratory disease populations following long-term erythromycin (5) or azithromycin (8, 32, 33), as well as during exacerbations (34). The lack of OP microbiota change in healthy individuals suggests that the clinical benefit exerted by macrolides may not occur through direct modulation of OP commensal microbiota composition, but, rather, macrolides may alter inflammation-related and functional signaling between the host and OP microbiota in preventing exacerbations (33).

In contrast to the limited impact of macrolides on OP microbiota composition, both azithromycin and erythromycin were associated with significant increases in macrolide-resistant streptococci and in OP carriage of macrolide resistance genes. However, changes were sustained beyond the antibiotic exposure period for azithromycin alone. The significant increase in proportional sensitivity in OP streptococci is consistent with that reported in both clinical populations receiving macrolides long term (7, 9) and in healthy populations following 1 week of macrolide exposure (6). Our findings extend this understanding in which differential effects in the macrolide resistance genes that were increased following azithromycin [erm(B) and mef] or erythromycin treatment (only mef) were identified. In keeping with the increase in streptococcal phenotypic resistance, both erm(B) and mef are prevalent in streptococci (35), as well as in several of the core OP bacteria, including Actinomyces (35) and Gemella species (36). The erythromycin ribosomal methylase (erm) gene and mef genes, which encode enzymes that methylate the 23S rRNA macrolide binding site and macrolide efflux pump protein, respectively, represent predominant mechanisms of macrolide resistance (14). The erm genes are associated with higher-level resistance to macrolides (MIC at which 90% of isolates are inhibited [MIC₉₀] to erythromycin or azithromycin of >64 μg/mL) than those conferred by mef genes (erythromycin or azithromycin MIC₉₀ between 0.5 to 16 μg/mL) (37, 38). These genes can be present on bacterial chromosomes or encoded on mobile genetic elements that facilitate resistance gene transmission between bacteria (35). Notably, while selected transposable elements within plasmids also cocarry the genes erm(B) or erm(A)/mef along with tetracycline resistance genes [tet(M) or tet(O), respectively] (35, 39), no significant change in the levels of tetracycline resistance genes was observed.

Whether azithromycin or erythromycin provides a relative advantage in relation to antibiotic selective pressure is the subject of ongoing debate (22). A higher incidence of macrolide resistance in the oral microbiota of children receiving azithromycin than erythromycin has been reported previously (40). Our findings support this observation, with only azithromycin resulting in significantly greater persistence of macrolide resistance genes erm(B) and mef and higher levels of macrolide-resistant streptococci following the washout period. In contrast, the selection, but not persistence, of only the mef gene was observed for erythromycin. Although the reason for this difference has not been fully established, azithromycin is known to have a higher bioavailability and with sustained concentrations detected in pulmonary tissues after 4 days following a single dose of 25 mg/kg azithromycin (41) due to its longer elimination half-life from serum and lung tissue than erythromycin (42). The higher tissue penetration of azithromycin may contribute to an increased risk for selection of resistance genes such as ermB, which confers resistance to high levels of macrolide (37, 38). Previous studies on S. pneumoniae isolates also reported higher mutation prevention concentration relative to the MIC for azithromycin than erythromycin (43). While longer periods of exposure within the mutant-selective window may promote bacterial resistance development with azithromycin, such phenomena may be primarily relevant in the context of acute treatments or where macrolides are used in on/off cycles. When macrolides are used in an ongoing manner to control chronic lung disease, such effects would be removed.

The potential clearly exists for transferable resistance elements to be transmitted to close contacts. For example, at least half of the increases in community carriage of methicillin-resistant Staphylococcus aureus (MRSA) can be attributed to transmission through household contacts (44). As demonstrated by baseline data from our study cohort, carriage of macrolide resistance determinants is prevalent in healthy individuals with no recent antibiotic exposure, an observation that is consistent with ongoing person-to-person spread. However, the scale at which any such transmission occurs, particularly in the absence of direct antibiotic selective pressure, is not clear. Our study explored this phenomenon, detecting no evidence of transmission of macrolide resistance genes from recipients to household contacts.

The choice between azithromycin and erythromycin as a strategy to reduce rates of exacerbation in those with chronic respiratory conditions is not informed appreciably by clinical efficacy or by the risk of adverse effects. Our investigation also showed that these macrolides were not separated by their impact on the composition of the oropharyngeal microbiota or their effect on onward transmission to a close contact. These findings provide further confidence on the safety of long-term macrolide therapies that have proven clinically effective for chronic lung disease, although long-term assessments in larger study populations are required to capture these potentially infrequent transmission events.

Our study had a number of limitations. Macrolide regimens for chronic lung diseases vary considerably in dose and frequency, as reflected by studies of azithromycin (1, 4, 6, 9). Assessments on the impact of macrolides on OP microbiota composition and resistance carriage may differ from those associated with other regimens in clinical use. The combined use of macrolides and nonmacrolide antibiotics was not assessed here but may influence overall impact. Additionally, the study sample size was based on adequate power to detect macrolide resistance induction, and therefore, potential alterations in the abundance of OP bacterial taxa abundance may have not been statistically detected. Although less common, other mechanisms that can confer macrolide resistance, including the inactivation of macrolides by esterases, phosphotransferases, and other methylases, were not assessed in our study. Macrolide resistance can also arise through point mutations in the 23S, L4, or L22 ribosomal proteins; however, they do not appear to be the main cause of macrolide resistance for S. pneumoniae (45). Point mutations also represent minimal risk to close contacts, given that both the transmission and replication of the bacteria would be required for increased incidence in these populations.

In conclusion, we directly compared the effects of long-term azithromycin and erythromycin on oropharyngeal resistance and transmission in healthy individuals where macrolide and nonmacrolide antibiotic exposure are known. We report that the administration of long-term (4 weeks) azithromycin or erythromycin in healthy individuals increases genotypic and phenotypic macrolide resistance, with resistance selection by azithromycin marginally greater than erythromycin. Notably, we found no evidence of onward transmission to close contacts.

MATERIALS AND METHODS

Study design.

A single-center, randomized, single-blinded, parallel-group trial was used to evaluate the impact of 4 weeks of twice-daily oral 400 mg erythromycin ethylsuccinate or twice-daily oral 125 mg azithromycin on oropharyngeal antibiotic resistance carriage in healthy adults (index cases) and their close contacts. Index cases were allocated to azithromycin or erythromycin at a 1:1 ratio, and one close contact was included in the study per index case. Between 1 February 2018 and 31 August 2018, 20 subjects and close contacts were recruited at the Mater Adult Hospital (Brisbane, Australia) and completed the study. Participant inclusion was based on previous antibiotic use (no antibiotics in the preceding 3 months, no macrolide antibiotic in the preceding 12 months), nonsmoker, no respiratory illness in the preceding month, and the ability to provide a close contact (at least 4 h of daily interaction over 5 days in a week) consenting to the study. Full inclusion and exclusion criteria are provided in the online supplement. The relationship between index cases and close contact was either spouse or partners, with a median interaction time of 8 h per day (Table 1). The trial was approved by institutional ethics committees (HREC/15/MHS/41) and registered with the Australian and New Zealand Clinical Trials Registry (ACTRN12617000278336) (Fig. 4).

FIG 4.

CONSORT flow diagram for index subjects and close contacts in this study.

Sample collection.

Subjects attended four clinic visits, baseline, week 1, week 4 (ABX), and week 8 (washout). Index cases were randomized at the baseline visit to receive either twice-daily azithromycin or erythromycin for a duration of 4 weeks. Safety assessments and adverse events were recorded at each visit (Table S1 in the supplemental material). Two oropharyngeal swabs were taken from the index patient and their close contact at each visit; one swab was placed in an aerobic medium for microbial culture, and the second flocked swab was stored in a dry tube for molecular-based assays.

DNA extraction, 16S rRNA sequencing, and bioinformatics analysis.

Bacterial DNA was extracted from oropharyngeal flocked swabs as described previously (8) and in the online supplement. Extracted DNA was used to generate amplicons of the 16S rRNA V4 hypervariable region according to previously validated methods (46) for microbiota profiling. The indexed library was sequenced using a MiSeq v3 kit on a MiSeq Illumina platform (Sahmri, Adelaide, Australia).

Bioinformatic analyses.

Paired-end 16S rRNA sequence reads were merged, and amplicon sequence variants (ASVs) were derived using Quantitative Insights Into Microbial Ecology 2 (QIIME2) (release 2019.4) and the DADA2 plugin as detailed in the online supplement. Taxonomic assignment was performed against the SILVA 16S ribosomal database (version 132) (clustered at 97% sequence identity) (47). Following the removal of spurious sequence variants, all samples were subsampled to the lowest sample read depth of 3,989 sequence reads for microbiome analysis.

Quantification of bacterial genes.

SYBR- and TaqMan-based quantitative PCR (qPCR) assays were applied to determine levels of resistance genes erm(A), erm(B), erm(C), erm(F), mef, msrA, tet(K), tet(L), tet(M), and tet(O), bacterial load (16S rRNA) (Table S2), and selected bacterial species (Table S3) (detailed in the online supplement).

Proportional sensitivity testing of streptococci.

Aerobic culture medium from oropharyngeal swabs was used to inoculate Streptococcus selective medium, with and without 2 μg/mL erythromycin (Sigma-Aldrich, St. Louis, USA) (online supplement). Plates were incubated overnight at 37°C under anaerobic conditions (10% CO₂, 10% H₂, 80% N₂). The proportion of macrolide-resistant streptococci was determined as previously (6).

Outcomes and statistical analysis.

Sample size calculation was performed for the primary outcome of identifying a change in macrolide resistance in the OP microbiota of index cases following 4 weeks of treatment with erythromycin or azithromycin. Based on a macrolide resistance induction rate of 27.7% (μ₀) and 56.6% (μ) (derived from the BLESS [15] and AZISTAST studies [5], respectively) and an estimated standard deviation of interquartile range (IQR)/1.35, the sample size required for a type 1 error rate of 5% and power of 0.9 was 10 participants in each arm. The secondary outcome (exploratory) was the difference between azithromycin and erythromycin to induce macrolide resistance in the oropharyngeal flora in the close (nonmacrolide-treated) contacts of those index case participants. Compositional differences were determined by permutational analysis of variance (PERMANOVA) using weighted UniFrac distances. Paired and unpaired nonparametric data were assessed using Wilcoxon and Mann-Whitney tests, respectively. Paired or unpaired t test was used for parametric data. Categorical data were assessed using Fisher’s exact test. The threshold of statistical significance was a P value of <0.05.

Data availability.

Sequence data are publicly accessible from the Sequence Read Archive (SRA) under BioProject accession no. PRJNA680665.