BACKGROUND: New designs of female condom have been developed to reduce costs and improve acceptability. To secure regulatory approvals, clinical studies are needed to verify performance. We aimed to assess the functional performance and safety of three new condom types-the Woman's Condom, the VA worn-of-women (wow) Condom Feminine, and the Cupid female condom-against the existing second-generation female condom (FC2). METHODS: We did a randomised controlled, non-inferiority, four-period crossover trial at three sites in Shanghai, China, and one site in Durban, South Africa, between May 1, 2011, and Jan 31, 2012. Participants aged 18-45 years who were sexually active, monogamous, not pregnant, and not sex workers, were eligible for inclusion if they were literate, had no known allergies to the study products; used a reliable, non-barrier method of contraception, and had no visible or reported sexually transmitted infections. We used a computer-generated randomisation sequence with a Williams square design of size four to assign patients (1:1:1:1) to the FC2 control device, or the Woman's, VA wow, or Cupid condoms, with 12 potential allocations. Randomisation was stratified by site. Participants were not masked to condom type, but allocation was concealed from study investigators. The primary non-inferiority endpoints were total clinical failure and total female condom failure, with a non-inferiority margin of 3%. Women were asked to use five of each condom type and were interviewed after use of each type. We also assessed safety data for each type. We did both per-protocol and intention-to-treat analyses. We calculated frequencies and percentages for each failure event and estimated differences in performance with a generalised estimating equation model. This study is registered, number DOH-27-0113-4271. FINDINGS: 616 women were assessed for eligibility, of whom 600 were randomly assigned to condom-type order (30, 120, and 150 women in the three sites in China, and 300 women in the site in South Africa). 572 women completed follow-up, with at least one condom of each type. Total female condom failure was 3·43% for FC2, 3·85% for the Woman's Condom (difference 0·42%, 90% CI -1·42 to 2·26), 3·02% for VA wow (-0·42%, -1·86 to 1·32), and 4·52% for Cupid (1·09%, -0·60 to 2·78); total clinical failure was 2·88%, 3·05% (0·17%, -1·48 to 1·81), 2·49% (-0·25%, -1·75 to 1·26), and 3·87% (0·99%, -0·55 to 2·52), respectively. Only two (<1%) participants, in South Africa, reported serious adverse events, unrelated to use of the study products. INTERPRETATION: Non-inferiority was shown for all condom failure events for the three new devices versus the FC2, within the predefined margin. FUNDING: Universal Access to Female Condoms (UAFC).