Mathematical Modeling Approaches for HIV Cure Drug Development
Speaker: Youfang Cao and Daniel Rosenbloom, Merck & Co., Inc.
Abstract: Several strategies have been proposed for achieving HIV cure, either by supporting the immune response to maintain virus at undetectable levels (“functional cure”) or by completely eliminating the low-level infection that persists despite long-term therapy (“eradication cure”). Development of HIV cure strategies requires reliable tools based on mechanistic understanding of proposed curative interventions to measure the impact of these interventions on already low levels of infection. We describe mathematical approaches in support of such efforts. First, we describe a HIV cure viral dynamics modeling platform, which was developed based on recently published models incorporating immune control of HIV infection. The model was fitted to large viral load dataset from clinical studies (N = 896 participants) of both antiretroviral treatment (ART) in treatment-naïve patients and analytical treatment interruption (ATI) in long-term suppressed patients. The model was also fitted to a SIV viral load dataset from a monkey immune perturbation study. Data heterogeneity and model identifiability present challenges; we found that 5 of the model parameters related to HIV cure can be estimated with high precision using the Monolix SAEM method. Simulations from the fitted model are used to identify strategies that may enable functional cure. Platform development will proceed by incorporating new mechanisms to identify targets and generate hypotheses about combination therapies, enabling design of HIV cure trials through simulations. Second, we describe how a pilot experiment involving diluted HIV samples – representing very low viral loads typically observed in long-term treated individuals – was used to characterize precision of a sensitive viral RNA assay. Although assay precision at these low levels is insufficient to quantify an individual subject’s change in infection size with high confidence, a Bayesian statistical method can characterize cohort-level changes over time, allowing incorporation of this data into mechanistic models. This method is being extended to characterize the dynamics of other sensitive infection biomarkers, including measurements of latent infection.
Room Reservation Information
Room Number: 106 McAllister
Time: 1:30pm - 2:30pm