 |
Center for Nonlinear Studies |
HIV surveillance typically entails estimation of the incidence and prevalence of cases based on diagnosis rates and the results of various antigen and antibody tests. Standard methods cannot estimate the timing of transmission events over the course of an infection; although this is believed to be a key determinant of new prevention modalities such as treatment-as-prevention. Multiple model-based analyses and longitudinal cohorts studies have come to highly variable and incongruous conclusions including both extremes: nearly all transmissions occurring either very early or very late during the course of an infection. In this talk I present a new method for estimating the timing of transmission events based on coalescent theory. This recently developed method provides a framework for describing the probability of a infection genealogy inferred from sequence relationships given a complex epidemiologic history. This framework can be used to estimate the parameters of an epidemiologic model in any of the standard representations used by epidemiologists (ODE, stochastic, agent-based) given a set of branching times inferred from a sample of HIV genetic sequences with known sample times. I also present our application of this framework to the timing of transmission events in Detroit, Michigan using both traditional surveillance data and genetic sequence data. We found that in the period 2003-2011, approximately 40% of all new infections were caused by individuals who were infected for less than one year. The sustained high rate of early-stage transmission may limit the efficacy of treatment-as-prevention programs.
Host: Thomas Leitner