OBJECTIVES: To assess persistence on protease inhibitors [PI; atazanavir/ritonavir (ATV±r), lopinavir/r (LPV/r), darunavir/r (DRV/r)]- or efavirenz (EFV)-based antiretroviral therapy (ART)] in the Women’s Interagency HIV Study (WIHS), a large, US, observational cohort. METHODS: This retrospective analysis of WIHS used bi-annually collected data from January 1998-April 2013. HIV+ women initiating ATV±r, LPV/r, DRV/r or EFV were followed until discontinuation of the  index PI or EFV, dropout from cohort, or end of study period. Persistence on ATV±r, LPV/r, DRV/r  or EFV over 50 months stratified by prior ART experience (ART naive, 0-1 prior PI, ≥2 prior PIs) was analyzed using Kaplan-Meier curves. In the analysis, we considered patient baseline variables (demographics, co-morbidities, viral load, CD4, and depression/ quality of life and  ART regimens used prior to initiating the index PI or EFV),  and follow-up variables (adherence, reasons for ART discontinuation).  RESULTS: In total, 661 women initiated ATV±r, 546 LPV/r, 203 DRV/r and 815 EFV-based regimens. Mean age was 41 years, >50% were Black, >15% were HCV+ and >50% had prior illicit drug use. For women with no prior ART, persistence at 50 months was 54% for EFV, 53% for ATV±r, and 14% for LPV/r (P = 0.0015); DRV/r had unreliable data in ART-naïve due to small sample size).  For ART exposed women with 0-1 PI, persistence was 64% for DRV/r, 50% for ATV±r, 49% for EFV, and 29% for LPV/r (P < 0.0001).  For women with ≥2 prior PI exposure, persistence was 70% for DRV/r, 41% for ATV±r, 28% for LPV/r, and 23% for EFV (P < 0.0001). CONCLUSIONS: Persistence differed among the ART regimens in analyses that stratified by prior ART/PI experience.   Proportional hazards regression will be used to examine if case-mix differences contribute to these patterns.