Early Viral Dynamics Predict Human Immunodeficiency Virus Posttreatment Control After Analytic Treatment Interruption

Area of Science:

• Virology
• Mathematical Biology
• Machine Learning

Background:

• Defining viral dynamics and biomarkers for sustained post-treatment control is crucial for human immunodeficiency virus (HIV) cure strategies.
• Predicting post-treatment control or non-control can optimize time off antiretroviral therapy (ART).

Purpose of the Study:

• To identify virologic predictors of long-term viral control after ART interruption.
• To develop a predictive model for classifying post-treatment control or non-control outcomes.

Main Methods:

• Utilized mathematical modeling and machine learning on viral kinetics data from ART interruption studies.
• Characterized predictors using data from treatment interruption, simulating real-time clinical data collection.
• Classified outcomes as post-treatment control (plasma viremia ≤400 copies/mL at 2/3 time points for ≥24 weeks) or non-control.

Main Results:

• Identified time to rebound, rate of initial rebound, and peak plasma viremia as potential predictors.
• Demonstrated that individuals likely to be non-controllers could be identified within 3 weeks of rebound with 80% accuracy.
• Achieved 82% sensitivity and 71% specificity in predicting non-controllers.

Conclusions:

• The identified predictors can enhance the safety of analytic treatment interruption in HIV cure trials.
• Early identification of individuals unlikely to achieve post-treatment control is feasible.
• These findings contribute to optimizing HIV cure strategies by managing ART interruption risks.