HIV incidence estimation using biomarkers for recent infection

Abstract

Approximately one in six South African adults is infected with HIV, making it the country with the largest population of HIV positive individuals in the world. Strategies for monitoring this epidemic are an important area of research. In particular, estimation of incidence, the rate at which individuals are being infected, is a key indicator of the scale of the epidemic. Since it is cheaper, quicker, easier and potentially less biased than prospective follow-up, incidence estimation from the cross-sectional application of a biomarker that tests for recent infection has gained much attention. There is, however, controversy over how best to account for individuals that present anomalous biomarker responses. The central contribution of the thesis is to derive a consistent incidence estimation approach that accounts for anomalous responses. This approach is compared with other cross-sectional incidence estimators found in the literature and shown to be less biased. Implications of the new approach to survey design and the development of new biomarkers are explored. Application to survey data gathered by the Africa Center for Health and Population Studies showed consistent results when compared with incidence estimates derived from follow-up. Aside from other theoretical contributions, the thesis also provides a systematic review of the application of the BED assay in incidence estimation with recommendations on best current practice.