Predicting human-immunodeficiency virus rebound after therapy initiation/switch using genetic, laboratory, and clinical data

Optimization of combination antiretroviral therapy (CART) for treatment of human immunodeficiency virus (HIV) infection usually targets achievement of suppression of plasma viral load at specific time points after therapy initiation (e.g. 12 or 24 weeks). In the majority of antiretroviral-naïve patients, modern CARTs are usually powerful enough to push the HIV-RNA load down to undetectable levels using standard detection assays (i.e. 50 copies/mL). However, emergence of drug resistance and viral load rebound is still a concern, especially in patients who failed multiple lines of therapy. Nowadays, understanding and predicting CART durability, i.e. how long an antiretroviral therapy can be sustained without changes, in absence of viral load rebound or other adverse events, has become an important challenge to address. In this work, we develop and implement a model to predict CART durability using multiple input domains (demographics, clinical, laboratory, and virus genetics), with data extracted from one of the largest HIV cohorts worldwide, the EuResist integrated database.