Personalizing the empiric treatment of gonorrhea using machine learning models

Area of Science:

• Infectious Diseases
• Computational Biology
• Public Health

Background:

• Antimicrobial resistance (AMR) in Neisseria gonorrhoeae necessitates updated treatment strategies.
• Current gonorrhea treatment relies on standardized guidelines, which may not account for geographic and demographic variations in AMR prevalence.
• The Gonococcal Isolate Surveillance Project (GISP) collects national data on AMR gonorrhea strains.

Purpose of the Study:

• To investigate the feasibility of personalizing empiric gonorrhea treatment using machine learning models trained on national surveillance data.
• To compare the effectiveness and antibiotic usage of personalized treatment strategies against standardized guidelines.

Main Methods:

• Utilized GISP data from 2000-2010 to train and validate machine learning models for predicting resistance to ciprofloxacin (CIP).
• Developed personalized treatment algorithms incorporating sexual behavior and geographic location.
• Evaluated model performance and compared personalized treatments with standardized guidelines (CIP, ceftriaxone (CRO), cefixime (CFX)) from 2005-2010.

Main Results:

• Personalized treatments could have substituted 33% of CRO and CFX use with CIP between 2005-2010.
• The personalized approach ensured 98% treatment effectiveness, maintaining efficacy compared to standardized guidelines.
• Machine learning models demonstrated consistent performance across different time periods and geographic regions.

Conclusions:

• Predictive models trained on national AMR gonorrhea surveillance data can personalize empiric treatment at the point of care.
• Personalized treatment can optimize antibiotic selection, reducing unnecessary use of newer agents while preserving treatment effectiveness.
• This data-driven approach offers a promising strategy to combat gonorrhea in the era of increasing antimicrobial resistance.