More so than any other class of drugs, the overuse of antibiotics directly contributes to the global decline of their own efficacy. The need for personalised and precision approaches to prescribing antibiotics is therefore important not only for individual patient outcomes, but also for preserving global antibiotic effectiveness and reducing the selection and spread of resistance. Despite this, treatment of bacterial infections lags well behind other areas of medicine when it comes to personalised medicine: the use of advanced diagnostics and data-driven approaches to predict which treatment strategies will be best for the individual patient.

Indeed, empirical treatment, without reference to pathogen susceptibility measurements, remains standard practice in many settings. Prescription guidelines often suggest switching to alternative antibiotics once resistance to first-line drugs reach a threshold rate at the population level. However, this overlooks a much more nuanced picture at the individual-patient level. Many patient-specific factors, such as age and past infection history, are strongly associated with the risk of resistance in an ongoing infection. Electronic patient records can be used to predict the personal risk that a patient’s infection is resistant to specific antibiotics, reducing the frequency of antibiotic susceptibility-mismatched empirical prescriptions¹.

When drug susceptibility measurements are available, antibiotic choice currently focuses on avoiding antibiotics to which the pathogen is resistant, with less attention paid to the risk that even susceptibility-matched treatments can fail as a result of resistance emerging in response to treatment. In a study focusing on urinary tract infections and wound infections, we found that treatment-induced emergence of resistance was remarkably common and driven not by de novo evolution but by rapid reinfection with a resistant strain pre-existing in the patient’s own microbiota². Despite these cases being treated “correctly”, with a susceptibility-matched antibiotic, the individual risk of an infection gaining resistance was not uniform: patients with past infections resistant to the currently prescribed antibiotic were at much higher risk of experiencing recurrence with gained resistance compared to patients whose previous infections were sensitive. Based on these associations, together with patient demographic information, machine learning–personalized antibiotic recommendations could minimize the risk of treatment-induced gain of resistance, offering a promising route to reduce the spread of resistant pathogens.

This study also highlights that the collateral effect of antibiotics on the patient’s microbiota can directly contribute spread of resistant strains. Future precision approaches to better target antibiotics and minimize collateral damage they cause, or decolonize resistant pathogens from the microbiota of high-risk patients, may be particularly useful not just for improving treatment outcomes, but also preserving antibiotic effectiveness.

1. Yelin, I. et al. Personal clinical history predicts antibiotic resistance of urinary tract infections. Nat. Med. 25, 1143–1152 (2019).

2. Stracy, M. et al. Minimizing treatment-induced emergence of antibiotic resistance in bacterial infections. Science 375, 889–894 (2022).

Dr Mathew Stracy is a Sir Henry Dale Fellow in the Dunn School of Pathology at the University of Oxford. His research focuses on antibiotic treatment failures and understanding how resistance spreads within a patient during treatment. His lab aims to develop new treatment strategies to improve outcomes and reduce the spread of resistance. Dr Stracy will be exploring this topic in more detail at the BSAC Winter Conference 2022. Register your place now.