In the Kagera region of Tanzania, a population-based study was initiated in 1987 followed by the establishment of antenatal-clinic-based sentinel surveillance system in the town of Bukoba in 1990. Repeat studies in both populations in Bukoba in 1993 and 1996 made it possible to study the dynamics of HIV infection prevalence and incidence in the area. This study aims at comparing the findings from this sentinel surveillance system with those of cross-sectional studies in the general population to assess its validity in estimating HIV prevalence and their trends in the general population. A multistage cluster sampling technique was used in the population-based studies whereas the antenatal-clinic-based population was obtained by consecutively recruiting antenatal care attenders coming for the first time during a given pregnancy. Antibodies against HIV infection were tested using two independent enzyme-linked immunosorbent assay (ELISA) antibody detection tests. Unlinked anonymous testing strategy was adopted for the sentinel population. Age-adjusted prevalence among antenatal care attenders decreased from 22.4% (95% confidence interval [CI], 20.6-25.2) in 1990 to 16.1% (95% CI, 15.9-18.8) in 1993 and further to 13.7% (95% CI, 11.8-14.3) in 1996. These results closely resemble those of the general population of adult women in the clinic's catchment area (the town of Bukoba) where the age-adjusted prevalence of 29.1% (95% CI, 24.4-34.6) in 1987 showed a decrease in the studies in 1993 18.7% (95% CI, 15.1-23.0) and in 1996 14.9% (95% CI, 12.0-17.1). The study indicates that general population trend estimates can be generated using sentinel surveillance data based on pregnant women visiting an antenatal clinic for the first time during a given pregnancy. The benefits of using this group outweigh its limitations that are brought about by possible selection bias. Continued surveillance of the epidemic based on antenatal care patients as a sentinel population is therefore recommended.