BackgroundWhile efficacy data exist, there are limited data on the outcomes of patients on third-line antiretroviral therapy (ART) in sub-Saharan Africa in actual practice. Being able to identify predictors of switch to third-line ART will be essential for planning for future need. We identify predictors of switch to third-line ART among patients with significant viraemia on a protease inhibitor (PI)-based second-line ART regimen. Additionally, we describe characteristics of all patients on third-line at a large public sector HIV clinic and present their early outcomes.MethodsRetrospective analysis of adults (≥ 18 years) on a PI-based second-line ART regimen at Themba Lethu Clinic, Johannesburg, South Africa as of 01 August 2012, when third-line treatment became available in South Africa, with significant viraemia on second-line ART (defined as at least one viral load ≥ 1000 copies/mL on second-line ART after 01 August 2012) to identify predictors of switch to third-line (determined by genotype resistance testing). Third-line ART was defined as a regimen containing etravirine, raltegravir or ritonavir boosted darunavir, between August 2012 and January 2016. To assess predictors of switch to third-line ART we used Cox proportional hazards regression among those with significant viraemia on second-line ART after 01 August 2012. Then among all patients on third-line ART we describe viral load suppression, defined as a viral load < 400 copies/mL, after starting third-line ART.ResultsAmong 719 patients in care and on second-line ART as of August 2012 (with at least one viral load ≥ 1000 copies/mL after 01 August 2012), 36 (5.0% over a median time of 54 months) switched to third-line. Time on second-line therapy (≥ 96 vs. < 96 weeks) (adjusted Hazard Ratio (aHR): 2.53 95% CI 1.03–6.22) and never reaching virologic suppression while on second-line ART (aHR: 3.37 95% CI 1.47–7.73) were identified as predictors of switch. In a separate cohort of patients on third-line ART, 78.3% (47/60) and 83.3% (35/42) of those in care and with a viral load suppressed their viral load at 6 and 12 months, respectively.ConclusionsOur results show that the need for third-line is low (5%), but that patients’ who switch to third-line ART have good early treatment outcomes and are able to suppress their viral load. Adherence counselling and resistance testing should be prioritized for patients that are at risk of failure, in particular those who never suppress on second-line and those who have been on PI-based regimen for extended periods.Electronic supplementary materialThe online version of this article (10.1186/s12981-018-0196-9) contains supplementary material, which is available to authorized users.