HIV Post-Treatment: Stochastic Modeling Insights on Rebound and Control

Antiretroviral drug therapy (ART) can effectively control HIV infection and reduce plasma virus load from very high, to undetectably low, levels. Cessation of ART is typically followed by rapid viral rebound (VR) to the high, pre-ART viral loads. However, recent observations give nuance to this pattern. In 2013, reports of the 'Mississippi baby' emerged, a child born with HIV and treated with ART shortly after birth. Taken off ART at 18 months of age, she appeared to be cured, only for HIV to reappear and rebound after 27 months. The following year, a study revealed 14 HIV+ individuals - 'the VISCONTI cohort' - who continued to control their HIV infection in spite of having stopped ART 4-10 years before study publication. These and other examples show that in some HIV-infected individuals, there can be significant delays between the suspension of ART and VR. Delayed VR and the potential for post-treatment control (PTC) are poorly understood but suggest exciting possibilities for treatment of HIV without ART, thereby avoiding side effects and high drug costs. Further, people who control HIV to undetectable levels are unlikely to infect others, thus reducing the possibility of HIV transmission in the population as a whole. The goal of this project is to better understand delayed VR and PTC by analyzing the following three questions: Under what circumstances will ART suspension lead to delayed VR or PTC? If ART suspension leads to VR, when will it occur? Can one predict when a patient can control infection off ART, permanently?

Central to this project is the development of mathematical viral dynamics models to investigate VR in HIV+ individuals following cessation of ART. The models will make use of stochastic multi-type branching processes, since viral populations in treated patients are small. While stochastic methods have been used to model viral dynamics, the standard approach is extensive simulation; in this project, the Principal Investigator will emphasize the derivation of analytic results instead, thereby introducing novel tools from applied probability to viral dynamics problems. The objectives will be addressed by stochastic models that will increase in complexity and realism through projects focusing on different biological aspects of HIV control such as immune responses. These models will be validated using data on HIV dynamics and VR from experimental collaborators and published data. The aim of this modeling is to better characterize VR dynamics, make testable predictions on times to VR, and gain insights into post-treatment control of HIV. In medical settings, the models developed may inform clinical guidelines with regards to ART suspension, lead to optimizing testing frequency during the ART interruption to detect VR and predicting PTC or 'HIV remission' where VR will be unlikely.