Antimicrobial Resistance · PertussisIs Azithromycin Still Enough? Mapping Macrolide Resistance in Bordetella pertussis

Pertussis is highly contagious. Its basic reproduction number sits at around 5.5, roughly four times that of influenza, and although mass vaccination from the mid-20th century drove incidence down dramatically, many countries have since documented a sustained resurgence — increasingly involving adolescents and adults as well as young infants.¹

Laboratory confirmation is slow. B. pertussis is a fastidious Gram-negative coccobacillus with demanding nutritional requirements, and culture can take five to seven days. Because antibiotics work best when started early, treatment is in practice empirical, and macrolides have been the drug class of choice for decades; erythromycin alone has more than 65 years of use in the United States.¹

That reliance has been under pressure since 1994, when the first erythromycin-resistant isolate was reported from Arizona. Resistant strains have since appeared on several continents, with a particularly steep rise in China over the past decade. Until now, no meta-analysis had quantified how common resistance has become or tracked how it has changed over time. Ma and colleagues set out to do both, and to collate minimum inhibitory concentration (MIC90) data that might guide the choice of alternatives.¹

The authors searched PubMed, Embase, the Cochrane Library and Web of Science from inception to December 2024 for studies reporting antimicrobial susceptibility of B. pertussis. There was no language restriction, although an English abstract was required. Two reviewers screened independently, with a third adjudicating disagreements. Studies were excluded if they tested fewer than five isolates, used non-human strains, reported resistance unclearly, or duplicated earlier data.¹

Because neither CLSI nor EUCAST has defined susceptibility breakpoints for B. pertussis, the authors followed earlier literature and classed an isolate as resistant to erythromycin, azithromycin, clarithromycin or clindamycin at an MIC of 32 mg/L or above. Other agents were interpreted against CLSI breakpoints for Haemophilus influenzae (35th edition). Clindamycin was included alongside the macrolides because of its shared mechanism of action and substantial cross-resistance.¹

Resistance rates were pooled with a random-effects model to allow for the expected between-study variation, and subgroup analyses examined country (China versus elsewhere), period (1994–2010 versus 2011–2024) and test method — the Epsilometer test (E-test, used in 40 studies) versus agar dilution (16 studies). Study quality was rated with an adapted Newcastle–Ottawa Scale: 36 studies (63.2%) were high quality and 21 (36.8%) moderate, with a mean score of 4.86.¹

Pooled across all settings, resistance was substantial for every macrolide studied. Erythromycin susceptibility was reported in 53 studies covering 5,337 isolates, of which 1,839 (34.5%) were resistant, giving a pooled resistance rate of 21% (95% CI 11–32%). Azithromycin, tested in 40 studies and 4,707 isolates, had a pooled rate of 25% (95% CI 12–40%), and clarithromycin, tested in 26 studies and 3,144 isolates, 15% (95% CI 4–30%). Clindamycin fared worst: 15 studies and 1,443 isolates yielded a pooled rate of 49% (95% CI 28–70%).¹

Those overall figures, however, conceal a sharply divided picture. Heterogeneity was extreme (I² above 97% for every agent), and the subgroup analyses explain why. The chart below lets you explore each split.

A concentrated, and recent, problem

The geographical divide is stark. For erythromycin, the pooled rate was 71% in China (95% CI 61–81%) against 0% (95% CI 0–1%) in all other countries combined, and the pattern repeated for azithromycin (72% vs 0%) and clarithromycin (64% vs 0%), each difference significant at p < 0.01. Almost every Chinese study reported azithromycin-resistant isolates.¹

The temporal shift is equally clear. Isolates collected and reported between 1994 and 2010 showed a pooled erythromycin resistance rate of 0%, rising to 42% (95% CI 26–58%) for 2011–2024. Azithromycin rose from 0% to 45% (95% CI 31–59%) and clarithromycin from 0% to 35% (95% CI 18–53%). The authors note that, although erythromycin resistance was first described in 1994, resistant strains were not identified in China until 2013.¹

Resistance has not stayed confined to one country. The review records macrolide-resistant isolates from the United States, Russia, Iran, France, Japan, Vietnam and Finland. Among the most recent studies, a 2024 Finnish series found one resistant isolate among 462, and a French series one among 67 — small numbers, but a signal of spread into settings that previously reported none.¹

The test-method effect

Resistance rates also differed by laboratory method: erythromycin resistance pooled at 33% (95% CI 19–49%) in E-test studies but only 1% (95% CI 0–3%) with agar dilution. The authors attribute much of this to who uses which method rather than to the tests themselves — Chinese laboratories, which report most resistant strains, overwhelmingly use the E-test, while agar dilution is better suited to large batches of isolates. Method and geography are therefore heavily entangled.¹

The molecular basis of macrolide resistance in B. pertussis is largely understood. In 2003, Bartkus and colleagues identified an A2047G point mutation in the 23S rRNA gene as conferring erythromycin resistance, and the great majority of resistant strains since then have carried either A2047G or A2058G. A small number of resistant isolates lack both mutations, which suggests that other, as yet unidentified, mechanisms exist. For that reason the authors argue that phenotypic susceptibility testing remains necessary even where molecular resistance assays are available.¹

Resistance outside the macrolide class is, so far, rare. Only one study reported quinolone resistance: six Japanese isolates collected between 2004 and 2006 showed high-level resistance to nalidixic acid (MIC >256 mg/L), attributed to a mutation in the quinolone resistance-determining region of the gyrA gene. No other quinolone-resistant strains had been reported at the time of the review, and no study reported resistance to trimethoprim–sulfamethoxazole.¹

To support decisions when a macrolide is unsuitable, the authors collated MIC90 values — the concentration inhibiting 90% of isolates — from all 77 studies. Lower values indicate greater in vitro potency. The chart compares the reported MIC90 ranges for agents the authors discuss as possible alternatives, alongside the oral cephalosporins they caution against.

Children

Trimethoprim–sulfamethoxazole (co-trimoxazole), already a recognised second-line agent, stands out. Twenty-eight studies covering 2,790 isolates reported MIC90 values of 0.01–0.75 mg/L, and no resistant isolate has been described. Some studies recommend it for children who have failed macrolide therapy. It is, however, not recommended for neonates or infants younger than two months.¹

Among the β-lactams, MIC90 values for piperacillin (0.006–0.156 mg/L), piperacillin–tazobactam (<0.016 mg/L), ampicillin–sulbactam (0.19 mg/L) and amoxicillin–clavulanate (0.5–2 mg/L) suggest likely susceptibility when read against H. influenzae breakpoints, as do those for intravenous third-generation cephalosporins including ceftriaxone (0.04–2 mg/L), cefotaxime (0.064–0.625 mg/L), ceftazidime (0.125–0.78 mg/L) and cefoperazone–sulbactam (0.047–0.094 mg/L).¹

Oral cephalosporins are a different story. B. pertussis is intrinsically less susceptible to first- and second-generation agents — cephalexin is in fact used as the selective agent in charcoal horse-blood culture medium — and the newer oral agents also showed high MIC90 values: cefixime 1.25–16 mg/L, cefpodoxime 16–100 mg/L and cefdinir 32–128 mg/L. The authors conclude that oral cephalosporins may have limited efficacy for pertussis.¹

Adults

With pertussis increasingly recognised in adults, the authors also examined the quinolones. Several showed low MIC90 values — moxifloxacin 0.012–0.03 mg/L, levofloxacin 0.01–1 mg/L and ciprofloxacin 0.01–2 mg/L — and a cited pharmacokinetic study found sputum levofloxacin concentrations of 1.27–4.36 mg/L, above the reported MIC90 range. Given that adult pertussis is generally mild, the authors suggest that quinolones, co-trimoxazole and certain β-lactams are adequate options, and that quinolones currently remain effective.¹

The authors argue that the near-zero resistance seen outside China may be partly an artefact of how pertussis is diagnosed. In China, culture remains a primary diagnostic method, and culture yields the live isolates needed for susceptibility testing. Most other countries rely on PCR or serology, which are faster and more sensitive but provide no susceptibility profile. Resistant strains could therefore circulate undetected, and the authors suggest the true global prevalence of macrolide resistance is likely underestimated.¹

Other limitations temper the pooled estimates. Susceptibility testing for B. pertussis is performed in relatively few tertiary laboratories, largely for research, so the evidence base is small and many studies reported a resistance rate of 0%. Heterogeneity was very high throughout. The breakpoints are borrowed rather than validated, and resistance thresholds varied between primary studies. Egger’s test showed no publication bias for the three macrolides but did suggest bias for clindamycin (p = 0.04). All included studies were cross-sectional, and the MIC90 compilation describes in vitro activity only — it cannot establish clinical cure, eradication or transmission benefit.¹

The authors also attribute China’s high resistance burden to extensive macrolide use there, which is a plausible explanation rather than one the review was designed to test.¹

1. Ma K, Sun W, Pan G, Xu Y, Shi Y, Chen Y. Antimicrobial resistance in Bordetella pertussis: a systematic review and meta-analysis. Epidemiology and Infection. 2026;154:e25, 1–11. doi: 10.1017/S0950268826101010
2. Bishop K, Moosa F, Makhasi M, et al. Comparing respiratory illness surveillance case definitions to detect Bordetella pertussis in children aged <5 years with respiratory illness in South Africa, 2017–2023. The Journal of Infectious Diseases. 2025;233(1):e34–e43. doi: 10.1093/infdis/jiaf501
3. Moosa F, du Plessis M, Wolter N, et al. Challenges and clinical relevance of molecular detection of Bordetella pertussis in South Africa. BMC Infectious Diseases. 2019;19:276. doi: 10.1186/s12879-019-3869-7