OBJECTIVES: Rates of first-line treatment failure and switches to second-line therapy are key indicators for national HIV programmes. We assessed immunological treatment failure defined by WHO criteria in the Tanzanian national HIV programme. METHODS: We included adults initiating first-line therapy in 2004-2011 with a pre-treatment CD4 count, and ≥6-months of follow-up. We assessed subhazard ratios (SHR) for immunological treatment failure, and subsequent switch to second-line therapy, using competing risks methods to account for deaths. RESULTS: Of 121 308 adults, 7% experienced immunological treatment failure, and 2% died without observed immunological treatment failure, over a median 1.7 years. The 6-year cumulative probability of immunological treatment failure was 19.0% (95% CI 18.5, 19.7) and of death, 5.1% (4.8, 5.4). Immunological treatment failure predictors included earlier year of treatment initiation (P < 0.001), initiation in lower level facilities (SHR = 2.23 [2.03, 2.45] for dispensaries vs. hospitals), being male (1.27 [1.19, 1.33]) and initiation at low or high CD4 counts (for example, 1.78 [1.65, 1.92] and 5.33 [4.65, 6.10] for <50 and ≥500 vs. 200-349 cells/mm(3) , respectively). Of 7382 participants in the time-to-switch analysis, 6% switched and 5% died before switching. Four years after immunological treatment failure, the cumulative probability of switching was 7.3% (6.6, 8.0) and of death, 6.8% (6.0, 7.6). Those who immunologically failed in dispensaries, health centres and government facilities were least likely to switch. CONCLUSIONS: Immunological treatment failure rates and unmet need for second-line therapy are high in Tanzania; virological monitoring, at least for persons with immunological treatment failure, is required to minimise unnecessary switches to second-line therapy. Lower level government health facilities need more support to reduce treatment failure rates and improve second-line therapy uptake to sustain the benefits of increased coverage.