Background: Accurate and efficient identification of seroconverters in antiretroviral-based HIV prevention trials is critical for minimizing emergence of resistance. The use of second-generation (IgG) HIV-1 rapid tests followed by confirmation by Western blot and/or HIV-1 RNA is the most common strategy to identify seroconverters in HIV prevention studies. We evaluated 150,125 HIV rapid tests performed in MTN 003 (VOICE) to assess the performance of this detection algorithm. Methodology: VOICE evaluated the safety and effectiveness of oral tenofovir disoproxil fumarate (TDF), oral TDF-emtricitabine and vaginal tenofovir 1% gel for the prevention of HIV infection in 5029 women from 15 clinical sites in South Africa, Uganda and Zimbabwe. Participants underwent monthly HIV testing with one or two concurrent rapid tests including Alere DetermineTM HIV-1/2, Orasure Oraquick AdvanceR Rapid HIV-1/2 and/or Trinity BiotechTM Unigold RecombigenR HIV. Confirmation of infection following a positive rapid test was done using Western blot (WB) (Bio-Rad). HIV RNA PCR (Abbott M2000 or Roche TaqMan) was used to resolve negative or indeterminate WB following one or two positive rapids. Results: Rapid testing correctly identified HIV infection status (positive or negative) at 77,795/77,915 (99.85%) follow-up visits in VOICE. Of the 161 rapid tests from 104 participants at 120 visits where rapid test results were discrepant with WB and/or RNA PCR test results, 51/161(32%) were due to a false positive rapid and 110/161 (68%) were due to a false negative rapid. Of note, 48 false-positive cases had rapid test results for subsequent visits; of these, 14 (29%) cases (from 7 participants) resolved by switching the falsely positive rapid kit to a different rapid kit. The resolution of HIV status after discrepant results took a median of 32 days (IQR 27-48 days). The overall sensitivity for Determine, Unigold and OraQuick was 88.78%, 85.86% and 61.54% respectively, and specificity was 99.96%, 99.97% and 99.99% respectively. The positive predictive values of the tests were 91.75%, 94.89% and 94.12% respectively. Conclusions: In a diverse clinical trial setting, the combination of Determine, Unigold and Oraquick accurately (99.85%) characterized HIV infection status in VOICE, but had limited sensitivity (61-88%) for identifying acute seroconverters. Resolution of discrepant results prolongs the confirmation of seroconversions. The use of newer antigen/antibody rapid tests could improve current HIV testing algorithms by reducing the time from infection to detection.