The Impacts of Sexual Role, HIV Treatment, and Pre-Exposure Prophylaxis Use on HIV Set Point Viral Load: A Network Modeling Study

Abstract

The analyses in this dissertation investigate how sexual role, treatment increases, and pre-exposure prophylaxis (PrEP) increases impact HIV virulence. All analyses used set point viral load (SPVL), the viral load (VL) found shortly after the period of acute HIV infection, as a proxy for virulence. This is a good proxy because higher SPVLs are associated with faster progression to AIDS and higher probability of transmission. The aims of this dissertation were: 1) to understand the effects of different sexual roles in MSM (exclusively insertive, exclusively receptive, or versatile) on the virulence of the viruses men acquired; 2) to understand the effects of increasing treatment on HIV virulence evolution; and 3) to understand the effects of increasing PrEP on HIV virulence evolution. All chapters used a stochastic, dynamic network model (EvoNetHIV) to partially or fully complete their analyses. EvoNetHIV is based in temporal exponential random graph models (ERGMs) and uses the statnet suite of R packages and the EpiModel package API. The findings in these studies help to overturn the previously modeled trade-offs in which higher treatment or PrEP coverage resulted in higher HIV virulence. Instead, increasing treatment or PrEP coverage can lead to lower virulence levels becoming evolutionarily advantageous, allowing their influences to positively affect the entire population. They also show that, because different sexual roles are associated with different HIV acquisition probabilities, there are significant differences in the virulence of the viruses men with different sexual roles acquire. These studies also show how virulence evolves differently when contrasting pressures are present: infected individuals going on treatment creates a different evolutionary environment than do susceptible individuals going on PrEP. Finally, these studies emphasize the impacts of model assumptions on all outputs. Careful consideration of and transparency about model assumptions are important for realistic and replicable models.