Heterosexual transmission of HIV analysed by generalized estimating equations

Abstract

A longitudinal analysis of a partner study is compared with a cross-sectional analysis which identify behavioural and biological risk factors for heterosexual transmission of HIV. Using generalized estimating equations (GEEs) a random effects logistic model is used for the longitudinal analysis. These approaches are illustrated by the Edinburgh heterosexual partner study. The longitudinal analysis finds that 'high-risk' sexual practices, unprotected intercourse for HIV and a low CD4 count in the index case significantly increase the risk of HIV transmission. The cross-sectional analysis, however, only indicates 'high-risk' sexual practices as favourable for HIV transmission.