Defining antibiotic treatment duration and its impact on antimicrobial resistance

Thesis

Reducing antibiotic treatment duration is a key stewardship intervention to mitigate antimicrobial resistance. In this thesis, I first address methodological issues introduced by non-adherence in non-inferiority trials, the commonest design for antibiotic duration randomised controlled trials. I conducted a simulation study and found that the probability of concluding non-inferiority when the treatment efficacy is actually inferior can be increased to as high as 0.1 from the acceptable 0.05 when adherence is relatively high at 90%. The simulations also highlighted the importance of anticipating patterns of non-adherence and accounting for this in power calculations. These findings were incorporated in the design of the "Reducing antibiotic treatment duration for ventilator-associated pneumonia (REGARD-VAP)" trial. The REGARD-VAP trial randomised patients with VAP to either a short (three to seven days) or a standard-of-care duration (eight days or more). The primary outcome was mortality or pneumonia recurrence by day 60. From May 2018 to July 2021, 340 patients were enrolled from seven hospitals in Nepal, Thailand, and Singapore. The second interim analysis, which included the first 231 patients, showed non-inferiority of the short duration arm based on a pre-defined Fleming-Harrington-O’Brien boundary. Adjusted analysis with inverse probability weighting suggested that a short duration was potentially superior to long duration. Lastly to examine shortened antibiotic duration as a strategy to reduce antimicrobial resistance in the hospital setting, I constructed agent-based models that incorporated within- and between-host dynamics of susceptible and resistant bacterial carriage in response to clinically-indicated antibiotic treatment. I found that shortening antibiotic duration is most effective at reducing resistance carriage in high transmission settings, and when resistant bacteria rapidly increase in abundance under antibiotic selection pressure and rapidly decrease when treatments are stopped amongst the treated individuals. A meta-analysis of antibiotic duration randomised trials showed an estimated odds ratio of 1.05 (80% credible interval 0.90 to 1.23%) with one additional day of antibiotic treatment. I conclude by discussing the future directions of the above work and the REGARD-VAP trial network.

Authors: Mo, Y

• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English