Burden of bacterial antimicrobial resistance associated with modifiable human antibiotic consumption: An observational study

BACKGROUND: Human antibiotic use is widely believed to contribute to the evolution and maintenance of antimicrobial resistance (AMR) in bacterial pathogens. While antibiotic stewardship should thus discourage use in situations where AMR-related harms outweigh clinical benefit, evidence to inform this assessment of tradeoffs is lacking, most notably in low-resource settings where gaps in antibiotic access persist. Quantitative understanding of use-resistance relationships is thus necessary as countries navigate commitments from the 2024 UN General Assembly resolution to reduce AMR burden by 2030. METHODS AND FINDINGS: We used global surveillance data on AMR prevalence from 2000 to 2018 representing over 30 million bacterial isolates from 114 countries, along with antibiotic consumption estimates for each country and year, to estimate elasticity relationships between AMR and antibiotic exposure. Our analyses addressed 16 bacterial pathogens and antibiotic consumption across 7 drug classes, controlling for country-level health and development indicators via weighted regression models. Across pathogens, evidence that AMR is associated with population-level antibiotic exposure is strongest for Acinetobacter spp., Streptococcus pneumoniae, Staphylococcus aureus, Pseudomonas aeruginosa, and Escherichia coli. In a modeled counterfactual scenario, we estimate that 10% lower levels of antibiotic use, in comparison to current levels, across classes, would be associated with 38,000 (95% confidence interval: 36,000-41,000) fewer annual deaths attributable to AMR globally. Although substantial, this total represents only 0.50% of estimated bacterial disease mortality. In sub-Saharan Africa, where risk of death due to bacterial disease is greatest, the modeled avertible burden of AMR-attributable deaths comprises only 0.34% of bacterial disease mortality. Key limitations of the study include reliance on ecological relationships to infer modifiable burden, and under-representation of low-resource settings in surveillance data on antibiotic consumption and AMR. CONCLUSIONS: Our findings suggest near-term achievable reductions in AMR-attributable deaths may be modest relative to the persisting burden of bacterial disease mortality. These estimates should inform assessments of the potential benefits and harms of restrictive versus permissive antibiotic use paradigms in epidemiologically distinct settings.