Seminar: "Bayesian inference for unknown HIV infection times: biomarker-based models, public health applications and molecular-clock extensions"

AUEB STATISTICS SEMINAR SERIES 2025-2026

Presenter: Nikos Pantazis, Assistant Professor of Epidemiology and Medical Statistics, Department of Hygiene, Epidemiology and Medical Statistics, School of Medicine, National and Kapodistrian University of Athens

Bayesian inference for unknown HIV infection times: biomarker-based models, public health applications and molecular-clock extensions

Room 709

ABSTRACT

The exact time of HIV infection is rarely observed but plays a crucial role in understanding epidemic dynamics, evaluating prevention strategies, and informing public health policy. In this talk, we present a Bayesian framework for inferring the time since infection using routinely collected clinical data, primarily CD4 cell counts and viral load measurements. By learning the natural history of untreated HIV infection from well-characterized seroconverter cohorts, we reverse the usual modelling perspective and estimate infection timing probabilistically at the individual level. We show how this approach can be extended to incorporate additional sources of information, such as AIDS status and behavioural data, and how uncertainty can be propagated to population-level analyses. Applications include distinguishing between pre- and post-migration HIV acquisition and improving inference from sparse surveillance data. We also discuss integration into large-scale public health tools, such as the ECDC HIV Platform. Finally, we present recent methodological developments that combine biomarker-based models with molecular clock information derived from viral sequences. This extension is particularly relevant in modern settings where biomarker data are limited, highlighting the flexibility of the Bayesian framework to integrate complementary sources of information.