Primary and secondary clarithromycin resistance in *Helicobacter pylori* and mathematical modeling of the role of macrolides

Helicobacter pylori infects over half of the global population and, while often asymptomatic, is implicated in chronic gastritis, peptic ulcer disease, gastric adenocarcinoma, and MALT lymphoma. Eradication therapy is required to clear infection, but rising antibiotic resistance—particularly to clarithromycin (Cla), a key macrolide in standard triple therapy—has reduced treatment efficacy. Cla resistance (Cla-res) is primarily driven by point mutations in the 23S rRNA peptidyl transferase region, with efflux mechanisms acting synergistically in mutant strains. Horizontal gene transfer can spread resistance within an individual’s intrahost population but is not known to transmit resistance between individuals. As Cla-res prevalence increases, empirical Cla-containing regimens become less effective, prompting guideline recommendations to avoid Cla without susceptibility testing when resistance rates are ≥15%. Prior work suggests cross-resistance among macrolides and links macrolide consumption to Cla-res, including bystander selection from non-eradication indications. The precise contribution of prior macrolide use (eradication-related versus other indications), heteroresistance, and transmission versus de novo mutation to primary resistance remains unclear. This study aims to accurately determine the true primary Cla-res rate by integrating individual macrolide use data, and to elucidate the population dynamics and sources of primary Cla-res using a mathematical model, including the impact of macrolide use for non-eradication purposes.

The paper situates its work within evidence that: (1) Cla-res in *H. pylori* is chiefly due to specific 23S rRNA point mutations; efflux pumps contribute but require 23S mutations; horizontal gene transfer is intrahost-limited. (2) Empirical triple therapy efficacy is undermined by increasing Cla-res; WHO prioritizes Cla-res *H. pylori* as a high-priority research target; guidelines discourage Cla use when resistance ≥15%. (3) Cross-resistance among macrolides is documented and macrolide consumption correlates with Cla-res prevalence (including ecological links to population-level consumption). (4) Prior epidemiologic studies often classified primary resistance without comprehensive accounting of non-eradication macrolide use, relying on sales data, recall, or incomplete records. (5) Modeling frameworks have addressed *H. pylori* transmission/dynamics, but prior models did not incorporate antibiotic resistance; broader AMR modeling reviews highlight gaps for pathogen-specific resistance dynamics. This study addresses these gaps by combining individual-level dispensing data with resistance phenotyping and a bespoke transmission model.

Study design and setting: Retrospective observational cohort (2005–2013) including 4,744 *H. pylori*-infected patients from Budapest and Central Hungary, identified via institutional databases. Ethical approval obtained (Semmelweis University EC #148/2007, #97/2012). Pediatric cases (0–19 years; n=51) were excluded from age-group analyses. Diagnostics: Gastric mapping biopsies obtained during upper endoscopy (16 institutions). Infection detection used modified Giemsa and/or *Helicobacter* immunohistochemistry (Dako B0471), confirmed by FISH (BACTFish *H. pylori* Combi Kit). *H. pylori* density was graded 1+/2+/3+ per updated Sydney system. Clarithromycin resistance testing: FISH targeted 23S rRNA mutations A2143G, A2144G, A2143C (ClaR1–3) with species probe (Hpy-1). Interpretation: green = susceptible; yellow (green + Cy3) = homoresistant; mixed green and yellow = heteroresistant. Both homo- and heteroresistant classified as Cla-res. Drug exposure ascertainment: Individual antibiotic dispensing and procedure/diagnosis codes retrieved from the Hungarian National Health Insurance Fund Administration (NEAK) prescription database (available since 2000). Clarithromycin dispensed for eradication was defined by proximity to gastroduodenoscopic biopsy or co-dispensation within 1 week with known eradication regimen components. Exposure groups: macrolide-naive; exposed to any macrolide; exposed for eradication; exposed for non-eradication; non-eradication exposure to clarithromycin versus other macrolides. Definitions: Primary resistance by customary clinical definition = no prior Cla-containing eradication regimen. Alternative ("true primary") definition = macrolide-naive (no prior macrolide exposure for any indication). Secondary resistance = macrolide-exposed (eradication-related or other indications). Mathematical model: Deterministic compartmental model of population in Central Hungary stratified by infection status (uninfected; infected with wild-type susceptible; homoresistant; heteroresistant) and medication history (macrolide-naive; macrolide for other indications; clarithromycin for eradication), yielding 12 compartments. Processes: acquisition of infection from wild-type or resistant sources (transmission rate β; relative transmission fitness of resistant β_r/β), superinfection leading to heteroresistance, spontaneous mutation from wild-type to heteroresistance (rate λ), macrolide use for non-eradication (rate m; fraction η of wild-type become heteroresistant), clarithromycin eradication attempts (rate δ) with outcomes: cure (p), no effect (p_u), induced homoresistance (p_r), induced heteroresistance (p_h). Demographic turnover μ and a slow declining *H. pylori* prevalence trend were included. Inputs estimated from cohort data, published sources, and public databases; model solved and scenarios projected using Wolfram Mathematica. Statistics: Categorical comparisons by Fisher’s exact test; generalized linear models for associations with age/sex; significance p<0.05. R 3.5.1 with epitools and ggplot2 used.

- Cohort and overall resistance: Among 4,744 infected patients, overall Cla-res prevalence was 17.2% (816/4,744). Nearly half of resistant cases were heteroresistant (47.2%). - Sex differences: Females had higher Cla-res than males (19.8% vs 13.7%, p<0.001). No sex difference in heteroresistance proportion (males 49.1% vs females 46.2%; p=0.46). Age and sex independently associated with resistance (both p<0.0001); heteroresistance proportion not associated with age/sex. - Bacterial density: Resistance was lower with low density (1+) vs 2+ or 3+ (12.9% vs 18.5% and 18.9%; p<0.001). Heteroresistance less frequent in 1+ vs 2+ (40.3% vs 50.0%; p=0.0452). - Age groups: Resistance <15% only in patients ≥70 years. In adult women, Cla-res >20% in all age groups <70; in men, generally <15% except ages 30–39. - Primary vs secondary by eradication history: Primary Cla-res (eradication-naive) was 13.3% (569/4,263). Secondary resistance in previously eradication-treated patients was 51.4% (247/481). - Primary vs secondary by any macrolide exposure: True primary resistance among macrolide-naive was 5.5% (140/2,532; p<0.001 vs conventional 13.3%). Secondary resistance among macrolide-exposed was 30.6% (676/2,212). - Sex and exposure: Among eradication-naive, females had higher resistance than males (15.1% vs 11.1%; p=0.0001). Among macrolide-naive, no significant sex difference (6.4% vs 4.6%; p=0.055). Among macrolide-exposed, women had higher resistance (32.4% vs 27.3%; p=0.011) and higher macrolide dispensing metrics. - Type of macrolide exposure: Secondary resistance was higher after eradication use vs non-eradication use (51.4% vs 24.8%; p<0.001). For non-eradication indications, clarithromycin caused more resistance than other macrolides (36.2% vs 13.4%; p<0.001). - Modeling outputs: Transmission of resistant strains accounts for ~5.7% (4.67–7.43%) of all new infections; resistant strains have relative transmission fitness β=0.72 (0.58–0.95) vs susceptible. Among macrolide-naive primary resistant infections (observed 5.5%), 98.7% are predicted to originate from transmission of resistant bacteria (including ~one-third via overcolonization leading to heteroresistance), and 1.3% from spontaneous mutation. - Projections: With current macrolide consumption, Cla-res is predicted to increase by ~0.1% (0.06–0.14%) per year. Discontinuation of non-eradication macrolide use would slow the long-term growth of resistance, yielding a very slowly increasing trajectory rather than faster growth.

By integrating individual-level macrolide dispensing data with FISH-based resistance profiling in a large regional cohort, the study refines the estimate of true primary Cla-res, showing it to be substantially lower (5.5%) than when defined solely by eradication history (13.3%). The disappearance of sex differences among macrolide-naive patients and higher resistance among macrolide-exposed women align with higher macrolide consumption in females, supporting bystander selection as a driver of secondary resistance. The high heteroresistance proportion underscores its epidemiological importance and potential therapeutic implications. The mathematical model indicates that most primary (macrolide-naive) Cla-res cases arise from transmission of resistant strains rather than spontaneous mutation, despite a modest transmission fitness cost (β≈0.72). This finding emphasizes that controlling transmission of resistant *H. pylori*—and reducing selective pressure from macrolide use—are key to curbing resistance growth. Scenario analyses suggest that limiting non-eradication macrolide use would slow the increase in Cla-res prevalence; however, due to the chronic nature of infection and ongoing transmission/superinfection, resistance would continue to rise slowly rather than decline, consistent with external observations (e.g., Taiwan policy changes). Clinically, given resistance ≥15% thresholds, susceptibility testing before Cla-containing eradication is advisable in subgroups with high prevalence (women <70 years, men 30–39 years) and in any patients with prior macrolide exposure.

This work provides precise estimates of primary and secondary Cla-res in *H. pylori* by leveraging individual macrolide exposure histories and introduces a dedicated transmission model for Cla-res dynamics. Key contributions include: (1) establishing that true primary Cla-res in Central Hungary is ~5.5% and largely transmission-derived, with spontaneous mutation contributing minimally; (2) quantifying the substantial impact of prior macrolide use—especially failed Cla-containing eradication—on secondary resistance; (3) highlighting a high burden of heteroresistance. The findings support stewardship to minimize non-eradication macrolide use and recommend susceptibility-guided therapy in groups with ≥15% resistance or any history of macrolide exposure. Future research should refine management strategies for heteroresistant infections, incorporate more granular exposure-response data (including dosing and repetition), and validate/extend the model across different populations to optimize intervention policies.

- Resistance detection targeted only the three most prevalent 23S rRNA mutations; rare mutations or uncommon resistance mechanisms were not captured, though literature suggests limited additional phenotypic impact. - Potential sampling error due to patchy mucosal distribution and low-density infections may underestimate heteroresistance or spontaneous mutation-derived cases. - Detailed homo-/heteroresistance ratios by all exposure subgroups were unavailable; some model parameters were inferred from prior work. - Modeling simplifications include homogeneous mixing, shared parameters across resistance statuses, ignoring unintended eradication, repeated macrolide courses, dosing differences, and differences in resistance induction between non-eradication clarithromycin vs other macrolides. - Assumptions about transmission from heteroresistant hosts (mutually exclusive transmission of wild-type or resistant per event) and that spontaneous/mu or macrolide-induced resistance in wild-type infections yields heteroresistance may not capture all biological nuances.