Estimating Incidence of HIV Infection in Uganda

Abstract

To the Editor: Human immunodeficiency virus (HIV) surveillance is in need of a practical and valid method to measure recent HIV incidence rates and the BED test (a test developed for HIV subtypes B E and D) is increasingly being used for this purpose. However we are greatly concerned over claims of validity given the confusing results this test generates. In their study of risk factors for recent HIV infection in Uganda Dr Mermin and colleagues used this test to estimate incidence in Uganda as 1.8 per 100 personyears at risk (pyar) among 15- to 59-year-olds. The greatest incidence rates were for 35- to 39-year-old women (3.5/ 100 pyar) and 30- to 34-year-old men (2.8/100 pyar). The most recent direct empirical estimate of incidence in Uganda is 0.25 per 100 pyar in 2005 for a rural southwest region with a prevalence level similar to that observed nationally. 3 In these data the age groups with the highest incidence are 10 years younger. The study by Mermin et al points to the agreement between observed incidence in the Rakai cohort (1.7/100 pyar) and the corresponding BED estimate for the region. But although this is an atypically high prevalence area the BED incidence estimate is slightly lower than that for the whole country. To explore their results the authors used a lowerthreshold optical density (OD) to increase specificity at the expense of sensitivity identifying similar risk factors for recent infections. However for both analyses the comparison group included a vast majority of uninfected individuals. Perhaps a more telling analysis would be to compare the risk factors for HIV infections categorized as recent by the BED test with those infected with HIV but not identified as recent infections. Incidence in Uganda may have increased due to changes in risky sexual behavior3 but we would be surprised if it had almost tripled compared with rates that likely existed immediately prior to this study. While we accept the limitations inherent in tracking HIV epidemics through measures of prevalence we are concerned that implausible BEDderived estimates of incidence are being published and that that this could lead to mistaken assessments of the state of the HIV epidemic and trends in risk. (full-text)