Antibiotic use and resistance move together in several bacterial pathogens across countries, but a modeled reduction in consumption points to a relatively small effect on deaths from bacterial disease overall. A study published in PLOS Medicine estimates that 10% lower antibiotic use across drug classes would be associated with 38,000 fewer annual deaths attributable to antimicrobial resistance (AMR)—about 0.50% of all bacterial disease deaths.
Benjamin J. Singer and colleagues report the observational study, published October 26. They analyzed surveillance data on more than 30 million bacterial isolates from 114 countries between 2000 and 2018, alongside estimates of antibiotic consumption. The analysis covered 16 bacterial pathogens and seven antibiotic classes, comparing resistance prevalence with population-level use while adjusting for country-level health and development indicators.
The researchers found evidence of an association between antibiotic consumption and resistance for some, but not all, pathogen–drug combinations. The estimated relationship was strongest for quinolones among the drug classes studied. Yet after correcting for multiple statistical comparisons, only four pathogen–drug pairs showed significant associations: quinolone consumption with resistance in Acinetobacter species, Escherichia coli, Enterococcus faecalis and Pseudomonas aeruginosa.
The study measured “elasticity”: how a proportional change in antibiotic consumption corresponded to a proportional change in resistance prevalence. This offers a way to compare patterns across drug classes and bacteria, but it does not establish that changes in use caused changes in resistance. The data were aggregated by country and year, so the observed relationships cannot directly show what happened to individual patients or predict the effect of a particular policy in a particular place.
To estimate the possible consequences for mortality, the authors used those associations in a counterfactual scenario with 10% lower consumption of each antibiotic class. The resulting estimate—38,000 fewer AMR-attributable deaths a year, with a 95% uncertainty interval of 36,000 to 41,000—is a model projection, not an observed outcome of an intervention. Alternative assumptions produced different estimates: sensitivity analyses ranged from 37,000 to 54,000 fewer deaths, while scenarios concentrating reductions in drug classes with the highest or lowest estimated elasticity gave estimates of 110,000 and 7,400, respectively.
Even the central estimate is small relative to the wider toll of bacterial disease. The authors estimate that it represents 0.50% of global bacterial disease mortality. That comparison matters because antibiotics can treat bacterial infections, and the study’s authors say decisions about restricting or expanding their use must account for both resistance and the harms of inadequate access.
The evidence also has gaps. About 27.4 million isolates in the analysis came from high-income countries, compared with 2.7 million from low- and middle-income countries. The authors note that low-resource settings are underrepresented in surveillance of both antibiotic use and resistance, even though bacterial disease mortality is especially high in many such settings. Patterns found in the available country-level data may not capture differences within countries or over shorter periods.
The model is static and cannot predict long-term changes in disease transmission or bacterial evolution, or whether reductions in use would have effects that differ from increases. The findings therefore provide estimates for weighing possible benefits and harms, rather than a universal measure of what a reduction in antibiotic use would achieve. The authors argue that resistance burden is modifiable, while emphasizing that any assessment of antibiotic use must also consider the continuing burden of bacterial disease and access to effective treatment.