Background Limited viral load (VL) testing in HIV-infected individuals on treatment in low-income countries often results in late detection of treatment failure. The impact of remaining on failing second-line, protease inhibitor (PI) containing regimens is unclear. Methods We retrospectively tested VL from 2,164 stored plasma samples from 386 patients randomised to receive PI-monotherapy (ritonavir-boosted lopinavir, after initial PI+raltegravir induction) in the EARNEST trial. Protease genotypic resistance testing was performed in samples with VL>1000 copies/ml. We assessed evolution of drug resistance mutations from virological failure (confirmed VL>1000 copies/ml) until discontinuation of PI-monotherapy and examined associations using Poisson and linear mixed-effects models. Results 118 patients had a median 68(IQR 48-88) weeks on PI-monotherapy post-failure. At failure, 21/107(20%) had intermediate/high resistance to lopinavir. 40-48 weeks post-failure, 49/72(68%) and 36/71(51%) had intermediate/high-level resistance to lopinavir and atazanavir. Most remained susceptible to darunavir (12/72[17%] intermediate, no high resistance). Common PI mutations were