Background Although risk factors of recurrent and persistent bacterial vaginosis (BV) have been explored in the literature, the longitudinal incidence patterns of BV remain elusive. Methods We conducted a secondary analysis of longitudinal data from a randomized clinical trial of metronidazole treatment for asymptomatic BV. Repeated-measures latent class analysis was used to identify distinct longitudinal patterns of incident BV cases. Multinomial regression analysis was used to determine the predictors of class membership. The multivariable model included age, last BV treatment, douching frequency, birth control, sexual risk behavior, and assignment to treatment arm. Results A total of 858 African American women who were asymptomatic for BV were included in the analysis. Three emergent patterns of BV for 12 months were identified by repeated-measures latent class analysis: Persistent (55.9%), recurrent (30.5%), and clearance (13.5%). Participants who had douched at least once had significantly lower odds to be in the recurrent class versus the clearance class (adjusted odds ratio [adjOR], 0.55; 95% confidence interval [CI], 0.18-0.63). Women who had sex with women had significantly lower odds of belonging to the persistent class versus the clearance class (adjOR, 0.38; 95% CI, 0.22-0.68) and the recurrent class (adjOR, 0.43; 95% CI, 0.23-0.81). Those who were assigned to the treatment arm had significantly increased odds of being in the recurrent class versus the clearance class (adjOR, 1.92; 95% CI, 1.22-3.03). Women older than 21 years were significantly more likely to be in the recurrent class (adjOR, 1.88; 95% CI, 1.17-3.00) than in the clearance class. Conclusions Assessment of BV cases revealed distinct patterns of recurrence and persistence of BV, which were significantly associated with douching, being in the treatment arm, and being a woman who had sex with women.