Previous studies suggested that some groups of HIV-infected women were underrepresented in studies of antiretrovirals (ARVs). We assessed rates of and reasons for nonenrollment in a U.S. prospective cohort study (protocol P1025), and differences in the characteristics of HIV-infected pregnant women who were and were not enrolled. Forty-one percent of women invited to participate were not enrolled. Clinic-related reasons for nonenrollment included staffing or site resources (26.7% of women) and clinician refusal because of the woman's nonadherence to prenatal care and/or poor research candidacy (10.8%). Patient-related reasons for nonenrollment included unavailability of women for enrollment (e.g., difficulty enrolling during labor/delivery, loss of clinic contact) (20.3%), refusal because of mistrust (10.1%), refusal because of time requirements (8.3%), refusal because of distance to the clinic (4.7%), and spontaneous abortion (4.7%). P1025 participants (N = 530) were significantly more likely to be Hispanic (32.1% vs. 19.8%), and less likely to be non-Hispanic black), to present in the first or second trimester for prenatal care (91.5% vs. 77.6%), and to be ARV-naive (32.8% vs. 23.0%) than nonparticipants (N = 2222). This high rate of nonenrollment can bias study results and generate findings that are applicable only to particular groups of women. Efforts should be taken to design protocols that facilitate enrollment of HIV-infected pregnant women.