Given that half of U.S. pregnancies are unintended and some prescription drugs are frequently used by reproductive-age women, there is an increasing interest in establishing pregnancy registries to monitor fetal exposures and pregnancy outcomes. Physicians report prenatal exposures and pregnancy outcomes, including birth defects, to these registries. We compared pooled data from four pregnancy registries with data from a population-based birth defects surveillance system, the Metropolitan Atlanta Congenital Defects Program (MACDP); specifically we compared the defect prevalence by organ system and severity, the number of defects per baby, and timeliness. We also compared the number of zidovudine exposures identified by a registry to the number identified by 29 states with HIV surveillance. The registries' overall defect prevalence (41/1471, 2.7%) was slightly lower than MACDP (6157/195642, 3.2%). The defect prevalence by organ system was similar, except for genitourinary defects which had a lower prevalence in the registries than in MACDP (RR = 0.22; 95% CI = 0.07,0.67). The prevalence of having an internal defect or severe defect reported was lower in the registries (RR = 0.75, 95% CI = 0.53,1.06, and RR = 0.82, 95% CI = 0. 57,1.19, respectively). The mean number of defects identified per affected infant was 2.82 in MACDP and 1.68 in the registries. Both systems received 69% of defect reports by 6 months after birth. In similar 6-month periods, U.S. HIV surveillance identified 300 prenatal zidovudine exposures, while the registry received 134 worldwide reports. If registries improve their ascertainment of defects and exposed pregnancies, they will increase their chance of detecting signs of possible teratogenicity. Teratology 60:356-364, 1999. Published 1999 Wiley-Liss, Inc.