Antenatal clinic-based HIV prevalence in Zambia: Declining trends but sharp local contrasts in young women Ingvild F. Sandøy 1 , Gunnar Kva ˚le 1 , Charles Michelo 1,2 , Knut Fylkesnes 1 1 Centre for International Health, University of Bergen, Bergen, Norway 2 Department of Community Medicine, University of Zambia, Lusaka, Zambia Summary objectives To describe regional variation in human immunodefffeciency virus (HIV) prevalence trends in the period 1994–2002 and to assess the effects on prevalence trends of residence, educational level and age, and potential interaction between these variables. methods The data were from the national HIV sentinel surveillance system comprising information collected using interviews and unlinked anonymous testing of blood among pregnant women attending antenatal clinics in 22 sites in 1994, 1998 and 2002. results There was a decline in HIV prevalence in the age group 15–24 years in the period 1994–2002 both in rural (by 11%) and urban (by 26%) areas. The decline was strongest among highly educated women. However, this overall decline masked striking differences at community (site) levels with clearly declining epidemics in many sites contrasted by increasing epidemics in some and stability in others. Urban/rural residence, age, educational attainment, marital status and parity were factors closely associated with HIV infection. Having born many children was associated with lower risk of being infected by HIV, even in the age group 15–24. conclusions The HIV prevalence decline in young women is likely to reflect a drop in incidence during the period. However, there were sharp geographical contrasts in trends. Such local contrasts probably indicate differences in effectiveness of preventive interventions. Understanding factors and mechanisms explaining the differences will be of critical importance to better guide preventive inter- ventions. keywords human immunodeficiency virus, Zambia, prevalence, trends, sentinel surveillance, regional variation Introduction National estimates of human immunodeficiency virus prevalence from sub-Saharan Africa show sharp contrasts between the countries. The HIV prevalence among preg- nant women in 2003 was estimated to be 37.7% in Botswana, 22% in Namibia and 1.0% in Senegal. How- ever, the country estimates may disguise huge regional differences within the countries (Asamoah-Odei et al. 2004). Generally, urban areas have been more severely affected than rural areas (Asamoah-Odei et al. 2004). This is also the case in Zambia (Fylkesnes et al. 2001). Now, 20 years after the first HIV case was detected, however, the picture appears more complex. In many urban and rural areas a decline in prevalence has been observed, whereas in other areas the epidemic is still growing. A plausible explanation is that different areas of the country have reached different stages of the epidemic because of cultural and economic factors and differences in preventive efforts. An antenatal clinic (ANC)-based sentinel surveillance system is still the main instrument employed to estimate HIV transmission trends. In Zambia, the sentinel surveil- lance system was established in 1990 and has gradually expanded. The surveillance has been performed repeatedly at over 20 sites from 1994 with all provinces represented. As more than 90% of women in Zambia attend ANC- clinics at least once during pregnancy, the system provides a good basis for studies of transmission trends (Central Board of Health, Zambia 1997; Central Statistical Office, Zambia 2003; University of Zambia 1993). The main aim of this study is to describe the HIV- epidemic in the period 1994–2002 in Zambia based on sentinel surveillance data from pregnant women. We will focus on regional variation in prevalence trends in all women aged 15–49 and in the age group 15–24 years. The Tropical Medicine and International Health doi:10.1111/j.1365-3156.2006.01629.x volume 11 no 6 pp 917–928 june 2006 ª 2006 Blackwell Publishing Ltd 917