Vaginal microbicide gel trials for HIV prevention may require withdrawal of study product when a woman becomes pregnant. We assessed the potential impact of withdrawals in four trials by comparing self-reported sexual behavior pre- and post-pregnancy detection: (1) behavior in the month prior to positive pregnancy test versus behavior reported at the subsequent monthly visit; (2) behavior changes according to pregnancy status at the subsequent visit (continuing pregnancy versus not); (3) average sexual behaviors reported for all months prior to pregnancy detection versus all months after pregnancy was no longer detected; and (4) behavior changes among participants never testing positive for pregnancy. Pregnancy detection was associated with immediate reductions in self-reported numbers of partners and sex acts. The proportion of acts in which study gel was used following a negative pregnancy test did not return to pre-pregnancy levels. Pregnancies complicate the conduct and interpretation of vaginal microbicide trials when product must be withdrawn.