Population HIV prevalence estimates rely heavily on sentinel surveillance in antenatal clinics (ANCs), but because HIV reduces fertility, these estimates are biased. To aid interpretation of such data, we estimated HIV-associated fertility reduction among pregnant women in ANCs in Yaoundé (Cameroon), Kisumu (Kenya), and Ndola (Zambia). Data collection followed existing HIV sentinel surveillance procedures as far as possible. HIV prevalence among the women was 5.5% in Yaoundé, 30.6% in Kisumu, and 27.3% in Ndola. The birth interval was prolonged in HIV-positive multiparous women compared with HIV-negative multiparous women in all three sites: adjusted hazard ratios of pregnancy were 0.84 (95% confidence interval [CI]: 0.62-1.1) in Yaoundé, 0.82 (95% CI: 0.70-0.96) in Kisumu, and 0.74 (95% CI: 0.61-0.90) in Ndola, implying estimated reductions in the risk of pregnancy in HIV-positive women of between 16% and 26%. For primiparous women, the interval between sexual debut and birth was longer in HIV-positive women than in HIV-negative women in all sites, although the association was lost in Ndola after adjusting for age and other factors. Consistent results in different study sites help in the development of standard methods for improving ANC-based surveillance estimates of HIV prevalence. These may be easier to devise for multiparous women than for primiparous women.