The appropriateness of sentinel sero-surveillance based upon ante-natal clinic (ANC) attendees to estimate HIV-1 prevalence in the general population has been questioned. In Ethiopia, where the population is heterogeneous and where economic and practical barriers to ANC attendance exist, problems of extrapolation may be exacerbated. We planned an unlinked anonymous sero-survey which included data on basic population characteristics to investigate whether sero-surveillance data from ANCs in Afar Region might be taken to represent the situation among the general population of the Region. 371 pregnant women attending Dubti Hospital and Assayta Health Centre were tested for HIV-1 (using a single ELISA test) and active syphilis (RPR test). Socio-demographic characteristics were collected for each woman. Of the women tested, 278 (75%) were 28 years of age or younger. Two hundred eighty (76%) were urban residents and 237 (64%) described themselves as being of Amhara ethnicity. Overall, 73 (19.7%) were HIV-1 positive, but prevalence was three times higher among the 237 women of Amhara ethnicity compared to the 112 of Afar ethnicity (24.9% vs 8.0%, p < 0.001), and almost three times higher for urban compared to rural residents (23.2% vs 8.8%, p < 0.001). Positive RPR results were strongly associated with HIV-1 infection (OR 3.37, 95% CI 1.47-7.71). According to the Demographic and Health Survey (DHS) 2000, only 4.5% of the population of Afar Region is of Amhara ethnicity, and 7.8% urban residents. We have demonstrated that basing ANC sero-surveillance in urban areas of Afar Region over-samples urban residents of Amhara ethnicity and yields a major over-estimation of overall HIV-1 prevalence for the Region. Reliable estimation of HIV-1 prevalence in Afar Region will require more flexible strategies that permit sampling of rural Afar residents.