Estimating the proportion of bystander selection for antibiotic resistance among potentially pathogenic bacterial flora

Abstract

Bystander selection-the selective pressure for resistance exerted by antibiotics on microbes that are not the target pathogen of treatment-is critical to understanding the total impact of broad-spectrum antibiotic use on pathogenic bacterial species that are often carried asymptomatically. However, to our knowledge, this effect has never been quantified. We quantify bystander selection for resistance for a range of clinically relevant antibiotic-species pairs as the proportion of all antibiotic exposures received by a species for conditions in which that species was not the causative pathogen ("proportion of bystander exposures"). Data sources include the 2010-2011 National Ambulatory Medical Care Survey and National Hospital Ambulatory Medical Care Survey, the Human Microbiome Project, and additional carriage and etiological data from existing literature. For outpatient prescribing in the United States, we find that this proportion over all included antibiotic classes is over 80% for eight of nine organisms of interest. Low proportions of bystander exposure are often associated with infrequent bacterial carriage or concentrated prescribing of a particular antibiotic for conditions caused by the species of interest. Applying our results, we roughly estimate that pneumococcal conjugate vaccination programs result in nearly the same proportional reduction in total antibiotic exposures of Streptococcus pneumoniae, Staphylococcus aureus, and Escherichia coli, despite the latter two organisms not being targeted by the vaccine. These results underscore the importance of considering antibiotic exposures of bystanders, in addition to the target pathogen, in measuring the impact of antibiotic resistance interventions.

Keywords: antibiotic resistance; microbiome; vaccines.

Fig. 1.

Inputs and overall results of bystander analysis. (A) Heat map shading represents the proportion of visits (after weighting to be nationally representative) with a diagnosis of the specified condition, given that the visit resulted in a prescription of the specified antibiotic class. Results for TMP/SMX and nitrofurantoin are for the individual drug instead of an antibiotic class. Rows are not required to sum to 100%, as only a subset of conditions are shown, and each visit may be associated with more than one condition. Antibiotics included in each class are based on the Multum Lexicon classification system. Macr./Linc., Macrolides/lincosamides; TMP/SMX, trimethoprim–sulfamethoxazole. Diagonal lines indicate cells with value of 0. (B) Heat map shading represents the estimated etiology of each condition by species. If etiological data were available for multiple age groups, the weighted mean based on the relative frequency of visits (after weighting to be nationally representative) for that condition is shown. Diagonal lines indicate cells with value of 0. (C) Bars indicate mean carriage prevalence of each species across age groups, weighted by relative frequency of visits (after weighting to be nationally representative). (D) Bars indicate proportion of bystander exposures by antibiotic class and species (B_as) with 95% confidence intervals. “Overall” estimates reflect exposures to antibiotic in any of the classes shown in A.