BACKGROUND: Most data on HIV prevalence in Malawi come from antenatal clinic (ANC) surveillance and are, therefore, subject to bias. OBJECTIVES: HIV prevalence and risk factors were measured using population-based data to assess the accuracy of ANC surveillance and changes in prevalence and risk factors for HIV over time. METHODS: HIV prevalence was measured in 1988-1993 and 1998-2001 in community controls from case-control studies of mycobacterial disease in Karonga District, Malawi. ANC surveillance studies in the district began in 1999. RESULTS: Age and area-standardized HIV prevalence in women aged 15-49 years in the community was 3.9% in 1988-1990, 12.5% in 1991-1993 and 13.9% in 1998-2001. For men, HIV prevalence was 3.7%, 9.2% and 11.4% in the same periods. In 1988-1993, HIV positivity was associated with occupations other than farming, with increased schooling and being born outside Karonga District. In 1998-2001, non-farmers were still at higher risk but the other associations were not seen. The age- and area-adjusted HIV prevalence in the ANC in 1999-2001 was 9.2%. The underestimate can be explained largely by marriage and mobility. Reduced fertility in HIV-positive individuals was demonstrated in both ANC and community populations. A previously recommended parity-based adjustment gave an estimated female HIV prevalence of 15.0%. CONCLUSIONS: HIV prevalence has increased and continues to be higher in non-farmers. The increase is particularly marked in those with no education. ANC surveillance underestimated HIV prevalence in the female population in all but the youngest age group. Although there were differences in sociodemographic factors, a parity-based adjustment gave a reasonable estimate of female HIV prevalence.