OBJECTIVE: To assess neurocognitive function at the first-line antiretroviral therapy failure and change on the second-line therapy. DESIGN: Randomized controlled trial was conducted in 5 sub-Saharan African countries. METHODS: Patients failing the first-line therapy according to WHO criteria after >12 months on non-nucleoside reverse transcriptase inhibitors-based regimens were randomized to the second-line therapy (open-label) with lopinavir/ritonavir (400 mg/100 mg twice daily) plus either 2-3 clinician-selected nucleoside reverse transcriptase inhibitors, raltegravir, or as monotherapy after 12-week induction with raltegravir. Neurocognitive function was tested at baseline, weeks 48 and 96 using color trails tests 1 and 2, and the Grooved Pegboard test. Test results were converted to an average of the 3 individual test z-scores. RESULTS: A total of 1036 patients (90% of those >18 years enrolled at 13 evaluable sites) had valid baseline tests (58% women, median: 38 years, viral load: 65,000 copies per milliliter, CD4 count: 73 cells per cubic millimeter). Mean (SD) baseline z-score was -2.96 (1.74); lower baseline z-scores were independently associated with older age, lower body weight, higher viral load, lower hemoglobin, less education, fewer weekly working hours, previous central nervous system disease, and taking fluconazole (P < 0.05 in multivariable model). Z-score was increased by mean (SE) of +1.23 (0.04) after 96 weeks on the second-line therapy (P < 0.001; n = 915 evaluable), with no evidence of difference between the treatment arms (P = 0.35). CONCLUSIONS: Patients in sub-Saharan Africa failing the first-line therapy had low neurocognitive function test scores, but performance improved on the second-line therapy. Regimens with more central nervous system-penetrating drugs did not enhance neurocognitive recovery indicating this need not be a primary consideration in choosing a second-line regimen.